JAMES MILROY

BORE VS ;
eh Ebude infidle quinque
*kOCEANY S
—

Mirela meridsansee
as:redipe adeseno
mercer rrtsane 54 Bk
& paraleloram
ated
Atrebaty

VARTAIIG®
 Digitized by the Internet Archive
in 2023 with funding from
Kahle/Austin Foundation

https://archive.org/details/linguisticvariatO000milr
 LANGUAGE IN SOCIETY 19

Linguistic Variation and Change

ES
Language in Society
GENERAL EDITOR
Peter Trudgill, Professor in the Department of Language and Linguistics,
University of Essex

ADVISORY EDITORS
Ralph Fasold, Professor of Linguistics, Georgetown University
William Labov, Professor of Linguistics, University of Pennsylvania

Language and Social Psychology 10 Dialects in Contact

Edited by Howard Giles and Peter Trudgill
Robert N. St Clair
11 Pidgin and Creole Linguistics
Peter Miihlhdusler
Language and Social Networks
(Second Edition) 12 Observing and Analysing Natural
Lesley Milroy Language
A Critical Account of Sociolinguistic
The Ethnography of Communication Method
(Second Edition) Lesley Milroy
Muriel Saville-Troike
Bilingualism
Discourse Analysis Suzanne Romaine
Michael Stubbs
14 Sociolinguistics and Second
The Sociolinguistics of Society Language Acquisition
Introduction to Sociolinguistics, Dennis R. Preston
Volume I
Ralph Fasold 15 Pronouns and People
The Linguistic Construction of Social
The Sociolinguistics of Language and Personal Identity
Introduction to Sociolinguistics, Peter Miihlhdusler and Rom Harré
Volume II
Ralph Fasold Politically Speaking
John Wilson
The Language of Children and
Adolescents 17 The Language of News Media
The Acquisition of Communicative Allan Bell
Competence
Suzanne Romaine 18 Language, Society and the Elderly
Nikolas Coupland, Justine Coupland
Language, the Sexes and Society and Howard Giles
Philip M. Smith
Linguistic Variation and Change
The Language of Advertising James Milroy
Torben Vestergaard and Kim Schroder
 Linguistic
Variation and
Change
On the Historical
Sociolinguistics of
English

James MShaka
EC ERC MelineBetas SWNT Arena cecal RTH Uae
Ruts
4
RF oe Bees 1
} aa PH SOT ete, germ are ere ae on
5 Ree
ia iia anes: em ee y Go ee
fm A
1 PA ss , Be i vs i 5
ce 4 u fe=4 F A ae
Rot
Sloat Pxy
, Uh
©
lie
AE
Was
ice
,
oe
a
SR
x iN
oy ne ee \ be ‘at ‘
2 — i
ex wr:
vu
J
i y

IB
BLACKWELL
Oxford UK & Cambridge USA 4 @ & ¢ 1 4
 Copyright © James Milroy 1992

James Milroy is hereby identified as author of this work in accordance with section 77 of the
Copyright, Designs and Patents Act 1988.

First published 1992

Basil Blackwell Ltd

108 Cowley Road, Oxford, OX4 1JF, UK

Basil Blackwell, Inc.

3 Cambridge Center
Cambridge, Massachusetts 02142, USA

All rights reserved. Except for the quotation of short passages for the purposes of criticism
and review, no part may be reproduced, stored in a retrieval system, or transmitted, in any
form or by any means, electronic, mechanical, photocopying, recording or otherwise, without
the prior permission of the publisher.

Except in the United States of America, this book is sold subject to the condition that it
shall not, by way of trade or otherwise, be lent, re-sold, hired out, or otherwise circulated
without the publisher’s prior consent in any form of binding or cover other than that in
which it is published and without a similar condition including this condition being imposed
on the subsequent purchaser.

Library of Congress Cataloging in Publication Data

Milroy, James.
Linguistic variation and change: on the historical
sociolinguistics of English/James Milroy.
p. cm. — (Language in society; 17)
Includes bibliographical references and index.
ISBN 0-631-14366-1. — ISBN 0-631-14367-X (pbk.)
1. English language—Grammar, Historical. 2. English language-
Social aspects. 3. English language—Variation.
4. Sociolinguistics. I. Title. II. Series: Language in society
(Oxford, England); 17
PE 1101.M55 1922
306.4'4—dce20 91-18703
CIP

Typeset in Ehrhardt 11 on 13 pt
by TecSet Ltd
Wallington, Surrey
Printed in Great Britain by Hartnolls Ltd, Bodmin

This book is printed on acid-free paper.

 Contents

Editor’s Preface
Preface
1 Introduction: Language Change and Variation
2 Social and Historical Linguistics
3 Analysing Language in the Community: General
Principles
4 Interpreting Variation in the Speech Community
5 On the Time-depth of Variability in English 123
6 Speaker-innovation and Linguistic Change 164
ip Towards an Integrated Social Model for the Interpretation
of Language Change 206
Notes 223
References dh)
» Index 240
 Editor’s Preface

The impressive and exciting breakthroughs developed in this book

are due not only to the fact that the research presented here is very
much in historical sociolinguistics, as opposed to historical linguistics
pure and simple, but also to the fact that it involves work where, more
than in any other previous historical linguistic research, a consistent
distinction is drawn between linguistic systems, on the one hand, and
speakers on the other. Obviously, languages without speakers do not
change. Linguists, however, have not always drawn the correct
conclusion from this truism, namely that it is speakers who change
languages. A language changes as a result of what its speakers do to it
as they use it to speak to one another in everyday face-to-face
interactions. One of the emphases of this book is therefore on the use
of detailed analyses of data, derived from empirical observations of
vernacular language varieties in actual everyday use, to shed light on
historical linguistic problems, some of them associated with the
interpretation of historical data from English and other languages,
most of which is derived, of course, from non-vernacular written
sources.
Further vital insights are due to an insistence on making a
distinction between linguistic innovations and linguistic changes, and
a related distinction between speakers as innovators and speakers as
‘early adopters’. Crucial, too, is the emphasis on the importance of
weak social ties between individuals in the spread of linguistic
innovations.
Most impressive in this book, however, is the way in which it
combines erudition in historical-linguistic, sociolinguistic and social
 Editor’s Preface vii

theory with a detailed knowledge of linguistic varieties that can come

only from the experience of working closely with speakers themselves
— as sociolinguists do — rather than with languages as purely abstract
systems that one may or may not have intuitions about. Linguistic
change constitutes one of the great unsolved mysteries of linguistic
science, and the puzzle of the causation and function of language
change is a challenge that generations of linguists and philologists
have wrestled with. Perhaps the biggest mystery of all, especially as
far as sound change is concerned, is what we have come to call,
following the seminal 1968 paper by Weinreich, Labov and Herzog,
the actuation problem. We have become quite expert at explaining why
particular linguistic changes occur, and why certain changes are more
to be expected than others. We are still very inexpert, however, at
accounting for their actuation: why do particular changes occur in a
particular language variety at a particular time, and not in other
varieties or at other times?
Professor Milroy would not claim to have answered this question
here. Indeed, he makes it clear that the final answer to this question,
if we ever achieve one, will be extraordinarily complex. His book,
however, does penetrate further than any other research that I am
familiar with into the heart of this problem. For what Jim Milroy has
done in this book is to ensure that, henceforth, in seeking the answer
to this question, we shall be looking in the right place. We will not in
future be wasting our energies in seeking purely linguistic explana-
tions for what, as he has shown with exemplary clarity, is a supremely
sociolinguistic phenomenon.
As an acknowledged expert on the history of the English language
and as one of the best-known of all the world’s practising empirical
sociolinguists, Jim Milroy is uniquely equipped to tackle the massive
problems associated with the study of change in human language. It
should not surprise us, therefore that he has produced a work that
future generations of scholars will surely come to regard as one of the
‘ most important works on linguistic change ever written.

Peter Trudgill
 Preface

This book is undertaken as a result of a long-term interest in the task

of explaining how linguistic change comes about. In particular, it is an
attempt to explore the extent to which language change can be said to
be a social phenomenon, and it is focused mainly on variation in the
English language.
In this task I rely quite heavily on a generation of work in social
dialectology extending back to the 1960s, and especially on the
sociolinguistic research projects that I undertook at Queen’s Univer-
sity, Belfast, between 1975 and 1982. The specific purpose of these
was to study social aspects of language change by investigating spoken
vernacular English systematically in social and situational contexts of
use. These projects were based on a ‘coming-together’ of a number
of my own research interests, and they were designed as part of a
longer-term research programme. It is appropriate here to sketch in
briefly the relationship between the Belfast work — including the ideas
behind it and the continuing work that has arisen out of it — and the
general topic of this book. This is further justified by misunderstand-
ings that have arisen about the responsibility for, and scope of, this
research.
I was chief investigator in the first project (1975-7) and jointly
chief investigator with Lesley Milroy in the second project (1979-82).
The work was undertaken on the basis of prior interests in language
change, language standardization and historical English dialectology,
extending back over a number of years, and some ideas about
linguistic norms and language maintenance that I had derived from
these interests. I had been particularly interested in language standar-
 Preface ix

dization and prescription, and the way in which vernacular norms

seem to oppose these processes, presumably in an_ identity-
maintaining function. These observations suggest that norm-
enforcement in language might often come about in a non-
institutional way, through informal pressures within social groups in
order to maintain the vernacular, and I thought it might be worth
investigating these things empirically. Orthodox historical linguistics,
however, appeared at that time to have no place at all for such matters
as prescription and norm-enforcement, even though these must be in
some way implicated in language change. But as non-standard norms
are not codified in books, the only way in which vernacular norm-
enforcement can be investigated is by exploring language in the
community — hence the particular form that the Belfast projects took.
My main interest in undertaking this work was in explaining how
non-standard norms are maintained by social pressures, and the
relevance of this to linguistic change and the history of English; this is
broadly the topic of this book. For a number of reasons — which may
have had to do with the fact that these ideas were unfashionable at the
time — I found it difficult to get support for the proposed projects. In
1975, I was finally successful in obtaining a modest research grant
from the Social Science Research Council. Lesley Milroy joined me
at the preparatory stage, and together we prepared the detailed design
of the inner-city project.
It is the interest in language maintenance (as a basis for studying
language change) that chiefly differentiates this research from other
quantitative sociolinguistic projects of that time, in that this particular
interest controls the research design and the methodology adopted. If
language maintenance is emphasized, questions of the following kind
become accessible to investigation: why do different dialects remain
divergent from ‘mainstream’ norms of language despite the low status
usually accorded to them (why do they not all become ‘standard-
ized’?), and why do many of these divergent forms and varieties
' persist for generations and even for centuries? These questions have
a broader applicability, of course, to bilingual and multilingual
situations, to the question of how minority languages survive and to
many associated social, legal and educational issues. We always had
these possible applications in mind, but in this book I shall be
focusing mainly on historical English language studies. Within this
narrower scope, we can pose further questions: why, for example, do
changes in the ‘prestige’ norm over the centuries seem to originate in
x Preface

‘lower-status’ varieties of English rather than in elite ones? The rise

of urban varieties (such as London English in the late Middle Ages
and the ‘new’ dialects of nineteenth-century industrial cities) are
relevant to historical developments in English: therefore, using
Belfast as a general case-study, I was additionally concerned with
exploring the rise of urban vernaculars to see what kinds of thing
happen in ‘live’ urban vernacular development. Traditional historical
descriptive linguistics has no satisfactory answers to the questions |
have raised in this paragraph: to attempt to answer them, we focused
in our socio-historical work on language in day-to-day interaction
rather than on past records of language.
Because of the concern with language maintenance we studied in
the first phase of the research the forms of speech and social
interaction that differ most from ‘mainstream’ norms, that is, highly
divergent and non-standard forms of the language. We studied
inner-city locations before extending the analysis in later stages
to the outer city (we were of course concerned with non-
institutional norms of language, which I shall discuss further in
chapters 1 and 4). However, it is important to notice that, unlike
Labov’s (1966) work, our research did not depend in the first place
on projecting a stratificational social class model on to the linguistic
data collected. Thus, our interpretation of the sociolinguistic patterns
does not depend on this either. During the course of the inner-city
research, we looked around for a social variable that we might adapt
to quantitative use to measure language maintenance, and decided
jointly to use social network in connection with other social variables,
rather than place prior emphasis on social class. I shall look at the
relationship of class and network in chapter 7.
The findings of the projects depended on collaboration between a
number of people. The main day-to-day task throughout most of
these years was the quantitative linguistic analysis of the data, which
was carried out at Queen’s University. I am grateful to Rose
Maclaran, Domini O’Kane, Linda Policansky, John Harris, Brendan
Gunn, Anne Pitts, and Zena Molyneux, all of whom worked - either
full-time or part-time —on the data as research assistants in my
department at Queen’s, and to Maire Burke, who carried out the
Andersonstown fieldwork; the late Sue Margrain gave statistical help
and advice, and Irene Dempster helped for many years as typist and
secretary to the projects. Most of all, I am grateful to Lesley Milroy.
 Preface xi

She and I shared the main tasks equally and are jointly responsible for
all aspects of the work. Decisions on research design, sampling,
fieldwork strategy, linguistic analysis, use of speaker variables, quan-
titative analysis, and on all other matters, were in practice taken
collaboratively. More importantly, we developed the ideas in the
research collaboratively also. But as some division of labour was
necessary in practice, my main day-to-day task was to direct and carry
out the linguistic side of the work, and Lesley’s was mainly on the
ethnographic and social side. These two sides of the work are of
course interdependent, and our findings depend on both.
These comments are fuller than I would have liked, but they are
necessary because the Belfast research has now been quite commonly
misrepresented as wholly or mainly the work of my colleague, Lesley
Milroy. In fact, her book Language and Social Networks develops the
ethnographic side of the argumentation within the general framework
of the joint research, and is dependent for its linguistic database on
my analysis of the language. It also depends on my interests in many
other ways, particularly in the emphasis on language norms and
maintenance. In Chapters 3 and 4, I shall review the central concerns
of the Belfast research, focusing on the linguistic interests rather than
ethnography, and emphasizing the methodological independence of
the linguistic analysis, on the one hand, from the socially-based
interpretation, on the other.
A few comments are also required here on the treatment of
linguistic change in this book. I am concerned with change in the
basic structural parts of language, and — following Neogrammarian
concerns — mainly with sound-change. Sound-change has always
been the testing-ground of historical linguistic theory, chiefly
because — in a language-internal account — sound-change, more so
than other kinds of change, appears at first sight to be quite
mysterious: there is no obvious reason why it should happen at all. In
a sound-change from [a] to [o], for example, there is apparently no
- {mprovement’ in the language — we cannot demonstrate by empirical
methods that the new state of language brought about by this change
is ‘better’ or ‘more efficient’ than the state that went before. It is more
difficult to explain why sound-change should happen than to explain
some other aspects of language change, and I am keeping the mystery
of sound-change very much in mind at every point throughout this
book. The book is also intended to contribute to the development of
xii Preface

historical sociolinguistics, a field of interest that Romaine (1982a)

labels socio-historical linguistics. It will be evident that my approach is
rather different from hers.
I am extremely grateful to the Social Science Research Council
(now renamed ESRC) for their generous financial support. The first
project (HR 3771: ‘Speech Community and Language Variety in
Belfast’ — the inner-city work) was directed by me and carried out
between 1975 and 1977; the second project (HR 5777: ‘Sociolin-
guistic Variation and Linguistic Change in Belfast’) was directed by
both of us and carried out between 1979 and 1982. I also
acknowledge with gratitude the support of Queen’s University and
the Ulster Polytechnic (at which Lesley Milroy was then a Senior
Lecturer). Since 1982 we have of course continued to work on
matters arising from these projects, and many chapters of this book
draw on my continuing collaboration with Lesley. Chapters 6 and 7,
in particular, are largely based on jointly-authored papers written
since we completed the analytic work in the Belfast projects, and this
is acknowledged in the text. In the projects themselves, however, so
many people gave advice and help that I cannot acknowledge all of
them here: I hope they will know that I am grateful. But I would
particularly like to mention Roger Lass for the much-needed encou-
ragement that he has always given to me, and Peter Trudgill for his
constant support and, more recently, his immense patience as editor
in dealing with an increasingly wayward author. My thanks also to
Philip Carpenter for his patience. But there is one other person
without whose help the Belfast research could hardly have got off the
ground. I owe a special debt of gratitude to William Labov for his
example and inspiration, and for his unfailing personal support and
encouragement.

James Milroy
March 1991
 ]
Introduction: Language Change
and Variation

1.0 Language Change

/One of the most important facts about human language is that it is

continuously changing. Everyone knows that languages have changed
in the course of history: it is easy to see from a distance in time that
there are differences between Shakespeare’s English and present-day
English, but it can also be shown from close at hand that language is
continuing to change in the present just as it did in the past. At this
very moment changes are being implemented and diffused: old
varieties are dying out and new varieties are springing up; pronuncia-
tions are changing, new words and constructions are being adopted
and old ones adapted to new uses. Sometimes change is rapid, and
sometimes it is slow, and at any given time some linguistic structures
are changing while others remain stable. Indeed, change seems to be
inherent in the nature of language:/there is no such thing as a
perfectly stable human language.
It is also true that at any given time a language is variable.
’ Languages are never uniform entities; they can be observed to vary
geographically and socially, and according to the situational contexts
in which they are used. In the study of linguistic change, this
heterogeneity of language is of crucial importance, as change in
progress can be detected in the study of variation. It is also important
to remember this when we look at past states of language: we have to
accept that, just as language is variable when observed at the present
day, so it must also have been variable in the past. The history of any
2 Introduction: Language Change and Variation

language is therefore not the history of one ‘variety’, but is a

multidimensional history.
It follows, therefore, that at any time we care to look at a
language — or a dialect— it is variable and in a state of change. We
may, of course, choose to ignore this and treat language for descript-
ive purposes as ifit were a uniform and unchanging phenomenon, and
there are often good practical reasons for adopting this convenient
idealization. For instance, we may want to write a grammar of English
for the use of foreign learners, and it will be more helpful to our
readers if we focus on what is constant rather than what is changing.
However, the idea that language is static or uniform has also
penetrated into the roots of theory, and much of the linguistic theory
of this century has been based on the perceived need to treat language
as ifit were static and uniform. It should, however, be borne in mind
that technological advances since the earlier part of this century have
made it much more feasible for us to focus on variable states of
language than it was for de Saussure, Sapir and Bloomfield.

1.1 Uniformity and Variation

Early in the century de Saussure emphasized the priority of synchro-

nic descriptions and from that time onwards the dominant trends in
language description were synchronic: they focused on states of
language at given times as finite entities. Historical (or diachronic)
linguistics was relegated to a subsidiary role and was often conceived
of as an exercise in comparing these finite states of language at
different times. From this point of view we can liken a synchronic
description to a still photograph and a historical description to a
comparison of a series of still photographs taken at different times. In
reality, however, the history of language is a continuous process: it is
not a series of stills, but a moving picture. If we are to come closer to
understanding why and how languages change, we need to bear this
in mind.
It has often been pointed out (for example, by Bynon, 1977: 104-7)
that Saussurean structuralism can lead to a paradox for historical
linguistics. According to the Saussurean view, a language is at any
given time a system ow tout se tient—in which everything holds
together in a coherent self-contained structure of interdependent
 Introduction: Language Change and Variation 3

parts. Historical linguistic inquiry then proceeds by comparing

different states of language that have been attested from different
periods. However, a difficulty arises in the (unattested) intervening
stage between one state (state A) and another (state B), as it appears
that in this transitional stage the cohesiveness of the state A structure
has been to some extent violated. From this point of view, therefore,
we may be inclined to think of language as being perfectly structured
at some times but flawed at other times. Now, if linguistic change
were an abnormal state of affairs, this would not be an unreasonable
way to look at language: change could then be seen as something that
strikes a language from time to time like a disease. We could talk of
healthy languages (where everything holds together) and sick lan-
guages (where it does not). But this is not how things are: no real
language state is a perfectly balanced and stable structure, linguistic
change is always in progress, and all dialects are transitional dialects.
Synchronic states, as we observe them at a given time, are therefore
changing states, and stable states of language of the kind postulated in
Saussurean theory are idealizations.
The same, of course, applies to uniform states of language. As Sapir
observed long ago (1921: 147): ‘everyone knows that language is
variable’. No language is ever uniform. Linguistic theorizing,
however, has often proceeded as if languages were uniform entities.
Indeed, according to Weinreich, Labov and Herzog (1968), it has
gone further than this and has assumed that structuredness is found
only in uniformity: thus, variability in language has often been
discounted as unstructured. In fact, the equation of uniformity with
structuredness or regularity is most evident in popular (non-
professional) attitudes to language: one variety— usually a standard
language—is considered to be correct and regular, and
others — usually ‘non-standard’ dialects — are thought to be incorrect,
irregular, ungrammatical and deviant. Furthermore, linguistic
changes in progress are commonly perceived as ‘errors’. ‘Thus,
' although everyone knows that language is variable, many people
believe that invariance is nonetheless to be desired, and professional
scholars of language have not been immune to the consequences of
these same beliefs. We have discussed these points elsewhere (J.
Milroy & L. Milroy, 1985a), and I shall return to them in later
chapters. Here my concern is merely to point out that uniform states
of language are idealizations and that variable states are normal;
furthermore, variation in language may itself be structured and
 4 Introduction: Language Change and Variation

regular. Languages are not in reality completely stable or uniform,

and there is absolutely no reason why they should be.
The discussion above has raised some basic questions about lan-
guage. Is a language actually a system ou tout se tient, in which all the
structural parts are interdependent? Is variation in a language also
totally structured and systematic? Finally, if languages are actually like
this, how can change occur at all? In this book, I shall take the view
that these structuralist assumptions are not self-evidently true and
that to establish the degree to which language is structured within
itself — without reference to outside (for example, social) factors — is
an empirical task. Indeed, if language states are not to some extent
open-ended — if there are no ‘leaks’ in the system — it is difficult to
see how linguistic change can take place.

1.2 Foundations for the Social Modelling of Language

Change

One way of investigating these matters is to focus on the social nature

of language change, and this is the purpose of this book. My aim is to
examine the extent to which the origins of linguistic change can be
shown to be social; to put it in a slightly stronger form, I want to
~ examine the thesis that(linguistic change is a product of speaker-
activity in social contexts, which cannot be wholly explained from
within the properties of language systems themselves. This approach
, 1s justified on the grounds that language is a social phenomenon: it is
~ used by speakers to communicate with one another in social and
cultural contexts in which the language system (narrowly defined as a
“‘grammar’) is not the sole means of communication and personal
interaction., Furthermore, it is commonly observed that languages
which have no speakers do not change;' therefore, it seems reasonable to
inquire into the role of speakers in language change. As socially-
based arguments of this kind have not been widely favoured by
historical linguists over the last century or so, I shall attempt in
Chapter 2 to relate this social theme to the context of intra-linguistic
historical argument. First, however, I want to explain more directly
why I have felt justified in developing this socially-based approach to
historical linguistics. The key point to bear in mind here is that
 Introduction: Language Change and Variation 5

language throughout history has been primarily a spoken, and not a

written, phenomenon — my thesis is largely a matter of following out
the implications of that fact.
‘The drama of linguistic change’, according to Wyld (1927: 21), ‘is
enacted not in manuscripts nor inscriptions, but in the mouths and
minds of men’, and historical linguists have generally insisted that the
history of language is primarily the history of spoken language.
Traditionally, however, it was not possible to follow this out very
thoroughly because investigators did not have the technology to study
spoken discourse in extenso, and could hardly have imagined how
complex the patterning of spoken interaction in situational contexts
would actually turn out to be when it did become possible to analyse
it. As a result of these limitations, much of the generally accepted
body of knowledge on which theories of change are based depends on
quite narrow interpretations of written data and decontextualized
citation forms (whether written or spoken), rather than on observation
of spoken language in context (‘situated speech’). However, it is in
spoken, rather than in written, language that we are able to detect
structural and phonetic changes in their early stages; for this reason
and others, our understanding of the nature of linguistic change can
be greatly enhanced by observing in a systematic way recurrent
patterns of spoken language as it is used around us in day-to-day
contexts by live speakers. I would therefore like to suggest here three
general principles, or foundations, for the social modelling of change
that arise directly from this emphasis on spoken interaction, and that I
have used as guidelines in research into language change.
Speech is a.social activity in a sense that writing is not, and the
primary locus of speech is conversation. Conversations take place
between two or more participants in social and situational contexts,
and linguistic change is one type of phenomenon that is passed from
person to person in these situations. The first principle for a
socially-based model of language change therefore concerns the
' observation of language in use: it is the principle that speech-
exchanges can be observed only within social and situational con-
texts — they can never be devoid of such a context. To express this
more fully:

Principle 1 As language use (outside of literary modes and

laboratory experiments) cannot take place except in social and
situational contexts and, when observed, is a/ways observed in these
6 Introduction: Language Change and Variation

contexts, our analysis — if it is to be adequate — must take account of

society, situation and the speaker/listener.

This first principle carries with it a number of implications, the

most important of which is that generalizations about language
structure depend on a process of abstracting ‘language’ from the
situational contexts in which it naturally occurs. We do not actually
observe ‘the language’ or ‘language’ in the abstract: we observe
people talking. In a social account of language change, therefore, we
have to explain how changes get into this abstract structure that we
call language (which we cannot observe directly) as a result of the
activities of people talking (which we can observe more directly).
Furthermore, unstructured observations of very selective phenomena
will not be enough here: our descriptions of sociolinguistic patterns
will depend on observing recurrent patterns and will have to be
systematic and accountable to the data. In chapter 2 I will suggest a
distinction in principle between system and speaker, which arises from
this discussion; it is a distinction that I think we need to bear in mind
when we are analysing language in use. It also follows from this first
principle that close attention to methods of data collection and
analysis (and the relation of one to the other) is crucial; we regarded
this as very important in our work in Belfast, which I describe more
fully in chapters 3 and 4.
Whereas Principle 1 concerns the impossibility of observing langu-
age independently of society, Principle 2 concerns the impossibility of
describing language structures independently of society. This is not as
controversial as it may seem.

Principle 2 A full description of the structure of a variety (whether

it is ‘standard’ English, or a dialect, or a style or register) can only
be successfully made if quite substantial decisions, or judgements,
of a social kind are taken into account in the description.

The word ‘social’ here does not mean social class or prestige — the
decisions (or judgements) we are talking about are decisions (or
judgements) about the ‘norms’ of the variety concerned, and these
norms are social in the sense that they are agreed on socially — they
depend on consensus among speakers within the community or
communities concerned and will differ from one community to
another. The accuracy of the linguist’s description must therefore be
 Introduction: Language Change and Variation ?

judged on how closely it coincides with the socially agreed norm for
the relevant community.
Most language description encounters this problem of ‘norms’, and
although it is not always acknowledged, it can be detected in many
descriptive accounts of English. Even a statement that Received
Pronunciation (RP) of English has a long diphthong with an open first
element in such words as tve and tight depends on observing a sample
of people who are considered to be speaking this variety and on the
linguist’s judgement that this vowel is the majority usage among these
persons. But as a more general example of judging the norm, let us
consider Palmer’s (1965: 72-7) characterization of the English
perfect tense/aspect. Palmer cites sentences in which the adverbs just
and already occur with the perfect, but he gives no examples of their
occurrence with the simple past tense. Thus, forms like ‘I just did it’
and ‘I did it already’ are not given as possible sequences. A normative
judgement is implicit here, and this is probably a correct judgement
for many varieties, chiefly southern English ones. However, it is
certainly not correct for all varieties: the [past tense + just/already]
collocation is frequently observed in American, Irish and Scottish
English. Therefore, the accuracy of Palmer’s characterization has to
be assessed, not in terms of some absolute standard of ‘grammatical-
ity of the construction, but in terms of the speech community to
which it is relevant. It is not a matter of grammaticality or ungramma-
ticality of the usage for all speakers of English; it is a matter of
accurately describing what is agreed on by speakers in the community
concerned as the consensus norm of that community.
The interpenetration of social and linguistic judgements is easily
demonstrated in the work of linguists who are ostensibly non-social in
their approach. Smith (1989: 111-12), for example, comments that
‘for most speakers of (British) English’ He ate the pie already is ‘barely
acceptable’, whereas He has eaten the pie already is ‘fine’. This involves
the same kind of normative judgement that I have discussed above,
' and it is more or less correct for English in England and Wales. But
much more dubiously, Smith further comments that ‘for a// speakers’
(my italics) He has eaten the pie yesterday is ‘ungrammatical’. However,
this construction does occur in SBE (Southern British English). As
Trudgill (1984a: 42) points out, “The rules governing the use of the
present perfect in Standard English seem to be altering somewhat,
and there appears in particular to be an increase in the usage of such
forms as: I’ve seen him last year; He’s done it two days ago.‘ Noting that
8 Introduction: Language Change and Variation

[perfect tense + yesterday, last year etc.] is spreading in SBE and is

‘srammatical’ in other languages, such as French, it is advisable to
side with sociolinguists and not with Chomskyan linguists. We are not
dealing with ungrammaticality, but with a change in the norms of
usage for some part of the community. What this demonstrates is how
easy it is for the non-social linguist to appear to propose prescriptive
judgements. These typically appeal to some idealized superordinate
norm which is part of the ‘standard’ or literary language, rather than a
consensus community norm, but, although they are not enunciated as
social, they are also social judgements.
But what is true of ‘standard’ English norms (as described by
Palmer, Trudgill and Smith) is also true of non-standard norms, no
matter how violently deviant they appear to be to the prescriptively-
inclined observer. For example, if everybody in a social group says we
was there, then we was is the consensus norm. To take one of our own
examples, it is clear that for many Belfast speakers (and indeed for
many speakers of Irish English generally), the pronoun yous (plural) is
categorical, contrasting with you (singular): ‘So I said to our Trish and
our Sandra: “Yous wash the dishes.” Sure, I might as well have said
“You wash the dishes”, for our Trish just got up, put her coat on and
went out.’ The categorical distinction here between you and yous can
be said to be a norm for some community of speakers. The difficulty
that arises for the descriptive linguist is not so much to determine the
extent of what is ‘grammatical’ (on which see especially Labov, 1973),
as to determine the extent of the community of speakers within which
this particular structure is the consensus norm. It is clear that many
people in Belfast have categorical yous, but that many others vary in
the plural between you and yous; indeed, there are some who have
categorical you (as in standard English). Therefore, a description
which states that yous (plural) is categorical in Belfast English will be
valid for some part of the community, but not for all speakers or all
styles, and the variability in you/yous usage will certainly exhibit a
socially structured pattern.
Thus, although linguists have generally described differences
between varieties of language as /inguistic facts, these differences are
also social facts. The preference for ‘I did it already’ in dialect A as
against ‘I’ve done it already’ in dialect B, for example, arises from
differences in speaker-agreement within communities and is to that
extent a social fact. It follows from this that all language descriptions,
no matter how objective they are, must be normative. But although
 Introduction: Language Change and Variation o

linguists have often equated normative with prescriptive, no such

equation is intended here. Language descriptions are normative
because to be accurate they have to coincide as closely as possible
with the consensus norms of the community concerned. To be
normative, the linguist’s account of a variety does not have to be
prescriptive; that is, it does not have to prescribe how people in a
community should speak. The distinction I am making here can be
described as the distinction between observing a norm for descriptive
purposes and enforcing a norm prescriptively; but as this is not a
familiar distinction, I should perhaps discuss it a little more fully.
Linguistic scholars commonly contrast ‘descriptive’ with ‘prescrip-
tive’. Daniel Jones, for example, has this to say in the preface to his
English Pronouncing Dictionary (1955): ‘No attempt is made to decide
how people ought to pronounce; all that the dictionary aims at doing is
to give a faithful record of the manner in which certain people do
pronounce.’ This is as good a statement of the descriptivist position as
I can think of, and it seems to me to be irrelevant to point out (as
Haas, 1982, does) that people will nevertheless treat the dictionary as
prescriptive: they will use it to find out how they ought to pronounce.
If they do, this will not be because the presentation is prescriptive, but
because the pronunciation that happens to be described here is
viewed as one that it is desirable to acquire. If some other dialect had
been described, people would not use the description prescriptively. I
doubt, for example, if many people will use my own description of
Belfast pronunciation (1981) in order to acquire fluency in inner-city
Belfast English, although in principle they could. Jones’s English
Pronouncing Dictionary is therefore descriptive, and not prescriptive, in
exactly the same way as the description of any other dialect is. But it is
also normative (as all such descriptions must be) in the sense that it
attempts to reflect the socially agreed norms of some particular
community of speakers—in this case the community (and social
network) of RP speakers.
These first two principles can be described as requirements to
acknowledge the importance in language descriptions of: (1) the
social and situational context of speech-exchanges, and (2) consensus
on linguistic norms of usage within speech communities. There is a
third principle, which can be seen as an extension of the idea of
consensus norms into the diachronic dimension, and it is based on
the notion of language maintenance. If we assume that the natural
tendency of language is to diverge, relatively convergent states can be
10 Introduction: Language Change and Variation

described as arising from language maintenance through agreement

on, or acceptance of, particular norms of usage in the community. To
the extent that linguistic changes take place in speech communities,
however, they take place against a background of language mainte-
nance, and the extent to which they are successful depends on the
interplay of these two sets of social influences — those that encourage
maintenance (or stability), on the one hand, and those that encourage
change (or divergence), on the other. Principle 3 is fundamental to
the design of the Belfast research projects which were initiated in
order to follow out some arguments about linguistic change. Our
various discussions since then, about ‘language loyalty’, ‘focusing’,
‘social identity’, social network and related matters, all grow directly
out of the idea of language maintenance in the research design itself.

1.3 Language Maintenance and Language Change

The third principle can be stated thus:

Principle 3 In order to account for differential patterns of change

at particular times and places, we need first to take account of those
factors that tend to maintain language states and resist change.

This is closely related to the actuation problem, which we discuss more

fully in chapter 2, and the emphasis on language maintenance is the
most salient difference between the way I have approached historical
linguistic change and the approach of most other historical linguists.
It also differentiates our sociolinguistic research from other work in
that subject, including the large urban projects (such as Labov, 1966)
that influenced the Belfast research in the first place. It gives rise to a
number of consequential differences in approach. Historical linguists
do not generally describe patterns of maintenance: they tend to focus
on those things that are known to have changed and ignore those
things that have not, and they can often explicate historical changes
very elegantly without any reference at all to the social embedding of
the changes concerned. What strikes me as important here, however,
is the fact that if we focus exclusively on change and ignore
maintenance, these non-social procedures can be quite easily just-
ified: we can indeed propose sophisticated descriptions and highly
 Introduction: Language Change and Variation 1]

constrained theories of linguistic change, without taking any account

of social factors, and this is frequently done. However, if we pose the
more basic question why some forms and varieties remain stable
while others change, we cannot avoid reference to society. This is one
of the justifications for Principle 3 in a socially-based model of
change. Let me clarify it briefly.
If we are interested in how language states can remain stable and
how speech communities resist change, we have almost no alternative
but to take account of social factors. Suppose we notice that the
structure of language X has remained stable for a century: it is not
very interesting to point this out and then to leave it at that. We
naturally want to know why it has remained stable when other states
of language have changed, but in order to do this we have to study the
social and speaker-based reasons that may account for the fact that it
has not changed. In reality, languages change at given times in some
ways and not in others, sometimes they change rapidly and sometimes
slowly, some varieties are divergent and some convergent, and so on.
Thus, the third principle that I have proposed above is clearly
relevant in a range of very diverse language situations, and at widely
differing levels of generality, and I shall have much more to say about
this in later chapters.
I noted above that our Belfast research projects depended
especially on this third principle. It is the idea of language mainte-
nance that is most immediately relevant to the /istorical interests of
this research, and that is what this book is also about. If you look at
historical states of English, it is clear that some characteristics of the
language have persisted through time while others have changed, and
it is also clear, even in written documents, that early states of English
were variable just as present-day English is. Therefore, we want to
know how divergent forms and varieties of the language can be
maintained across considerable periods of time, and how structured
variability can persist through time. These interests influenced the
. original design of the Belfast research, and there were some other
related interests (including dialectological interests, such as cross-
dialectal comprehension and the question of access of non-standard
speakers to so-called ‘standard’ languages). What these situations all
have in common is the maintenance of distinctive norms of language
and (very often) persistence of divergent varieties through time, but
historical linguistic theorists haven’t shown much interest in matters
like this. Therefore, as a historical linguist, I thought that we might
12 Introduction: Language Change and Variation

get a better understanding of what linguistic change actually is, and

how and why it happens, if we could also come closer to specifying the
conditions under which it does not happen — the conditions under
which ‘states’ and forms of language are maintained and changes
resisted.
This means, amongst other things, that we can see some of the
traditional problems of historical explanation in a different light: for
example, we might want to ask why apparently ‘low prestige’ varieties
of language can persist over centuries, and why dialects of the same
language can be maintained for long periods in forms that differ so
much from one another that they are mutually incomprehensible. But
the most general consequence of an interest in maintenance is the
one I have mentioned above: it forces us to ask questions about society
and to investigate the structure of the societies in which norms of
language are maintained and changes implemented. If we focus on
change alone, we can propose explanations that are language-internal
without systematic reference to social processes, but we cannot do this if
we focus on maintenance: our answers have to be in some way socially based.
Of course, none of this implies that historical linguists never appeal
to social explanations: sometimes they do, but the appeals tend to be
ad hoc appeals to ‘prestige’, ‘the standard language’ and the like,
which assume the existence of speaker-links and power-structures in
society, but which do not investigate these systematically. This has
various consequences, but the one I need to call attention to here is
that, as a result of the superficiality of the social analysis, accounts of
the histories of particular languages have often been very heavily
coloured by the social attitudes of the investigators themselves.
Frequently, the researchers have not been able to observe social
structures and processes in an impartial way: their subjective social
attitudes have often been based on ideological positions which they
have simply assumed to be ‘common sense’, and so not easily open to
rational examination. They have then imposed these ideological
positions on to the analysis ‘from above’, as in the following: ‘Just as
fashions in dress are binding upon all members of a given class and
are imitated by all who look up to that class, so fashions in language
are binding upon all people of culture and are followed by other
members of the community to the best of their ability’ (Sturtevant,
1917: 26). This is a unidimensional imposition of standards from the
top; it is basically elitist, but quite mildly expressed. In some cases, as
 Introduction: Language Change and Variation 13

we shall see, the standard language ideology is much more strongly

expressed, and in some accounts ‘non-standard’ forms and varieties
are rejected as if they were not really ‘language’ at all. In chapter 5 we
shall discuss the effects of standard-based attitudes on historical
descriptions of English.
What we seem to need, therefore, is a theoretical orientation to the
study of language maintenance that takes full account of social
processes and therefore of social theories. Such an orientation, in
contrast with theories that have focused on change alone, is in the
fullest sense sociolinguistic. In this book I shall explore the social side
of our subject more fully than has been usual, and in chapters 6 and 7
I shall attempt to develop an integrated social model for the
interpretation of language change. In section 1.4, we first
acknowledge the importance of the ‘empirical foundations’ of Wein-
reich, Labov and Herzog (1968).

1.4 The ‘Empirical Foundations’

Whereas the three principles proposed above can be seen as basic in a

context of language stability and maintenance, the ‘empirical founda-
tions’ of Weinreich, Labov and Herzog (1968) are quite specifically
directed towards a theory of language change. Furthermore, whereas
our three principles are very definitely about the social nature of
language variation, their principles are in practice more directly
focused on locating the /inguistic patterns of change. They are devoted
to supporting the claim that linguistic innovations move in an orderly
manner through space (social, geographical, historical) affecting
‘linguistic structure also in an orderly manner. Thus, although there is
no necessary contradiction between their principles and ours, we have
‘ tended over time to reinterpret their principles in the light of our own
experiences in social and sociolinguistic analysis. The essential
difference here can perhaps best be understood by recalling that
whereas Labov’s New York City study was specifically directed
towards locating linguistic changes, our inner-city study of Belfast
was primarily a study of maintenance, and therefore necessarily social
for the reasons given above. Although it is assumed that change is in
14 Introduction: Language Change and Variation

progress at any time in any speech community, we would not have

been greatly concerned if, in this limited study, we had not found any
evidence for change at all: we would have expected it to appear at
some later stage.
According to Weinreich, Labov and Herzog, the task of explaining
linguistic change is best divided into five parts—the problems of
constraints, embedding, evaluation, transition and actuation. The first
three of these are interrelated and we have tended to treat them as
aspects of the same thing. The problem of the universal constraints
on linguistic change would, if it were solved, specify which changes
are possible and which are impossible, and predict which changes
would happen in particular circumstances: much historical linguistic
theorizing has been directed towards a solution to this problem. Here,
however, we need to notice that although the term ‘constraints’ has
often been understood as intra-linguistic, it is obviously possible to
speak also of social constraints on change. Let us consider this briefly.
‘Avoidance of homophony’ (following Martinet’s (1955) argu-
ments, and see further chapter 2) may be considered to be an
intra-linguistic constraint on change. In this spirit, I have pointed out
(J. Milroy, 1976a) that the development of [e:] (as in bait, sane) in
Belfast vernacular (henceforth BV) to [ei] (as in RP) may be blocked
by the fact that the vernacular pronunciation of the word-class of bite,
sign is already pronounced with [ei] in BV. Thus, if the RP-like
pronunciation were adopted, there would be merger of two distinct
lexical sets. However, following Principle 2 (which is concerned with
consensus norms), the decisive constraint here and elsewhere is just
as likely to be social as linguistic, because despite homonymic clash
(of the meat/meet type), mergers do commonly occur in languages.
Furthermore, in this case there would have to be some social
motivation for moving in the RP direction in the first place. Our
evidence suggests, however, that there is little or no such motivation:
throughout most of the Belfast community there is no discernible
movement in phonetic realization towards RP. Therefore, we can
suggest very plausibly that the so-called ‘prestige’ motivation to adopt
RP forms is overridden here by the solidarity constraint, which
requires the speaker to conform to local community norms rather
than to norms that are viewed as ‘external’. It is very striking, after all,
that in our inner-city work there were no examples at all of RP-like
[ei] in closed syllables of the type bait, sane, and very similar points can
 Introduction: Language Change and Variation IVa,

be made about a number of other vowels (for further discussion see

chapter 3).
In the Weinreich, Labov and Herzog programme, all aspects are
said to be both social and linguistic, but it is the evaluation problem
that is most clearly designated as social. This pertains principally to
social responses to language change ‘at all levels of awareness, from
overt discussion to reactions that are quite inaccessible to introspec-
tion’ (Labov, 1982: 28). It embraces notions of prestige, attitudes to
language (both overt and covert), as well as linguistic stereotyping and
notions of correctness. In practice, we have taken a rather wider view
of what goes on in speech communities than the Weinreich, Labov
and Herzog principles imply, and have tried to look at evaluation
within the context of broader structural principles such as power and
solidarity (Brown and Gilman, 1960), and interactional factors such
as ‘politeness’ (Brown and Levinson, 1987) and ‘accommodation’
(Giles and Smith, 1979, Trudgill, 1986a). We have also been
interested in how speech communities can reach consensus on the
evaluation of linguistic forms and how this consensus can shift in the
course of time.
It seems, however, that the problems of evaluation and constraints
can be viewed as constituent, or contributory, parts of the more
general problem of explaining the embedding of linguistic changes in
pre-existing states of language and society. Labov’s original contribu-
tion here (principally in the New York City study) has been to provide
a general model of the social location of a linguistic innovation and of
the manner in which it spreads from a central point upwards and
downwards through a speech community. But clearly, this overlaps
conceptually with the transition problem, in so far as the transition
from one state to another must be described here also. Thus, the
graphs and diagrams of the quantitative paradigm, when they show
stable patterns, can be interpreted as displaying aspects of the
linguistic embedding of a variable; when there is a crossover pattern,
~ however, they also display transition.
Transition concerns ‘the intervening stages which can be observed,
or which must be posited, between any two forms of a language
defined for a language community at different times’ (Weinreich,
Laboy and Herzog, 1968: 101). Transition is what most historical
description has been about, mapping (as it has usually done) the
transition between state A at one period and state B at a later one.
16 Introduction: Language Change and Variation

Quantitative analysis here allows a gain in sophistication: the process

by which one form gradually gives way to another can be demons-
trated in fine-grained detail in what is generally an orderly progres-
sion through different social groups and speech-styles (for a study of
this kind, see Eckert, 1980).
The fifth problem — actuation — is a very different kind of problem
from these others, and I shall not consider it here. We shall return to
it in Chapter 2 and later chapters, and I shall suggest that a solution to
it must be based on the behaviour of speakers rather than primarily on
the properties of languages. Here, it is appropriate to comment on
some general differences that seem to exist between the Weinreich,
Labov and Herzog methods of accounting for change in language
systems and our own approach to it. The chief difference, as I have
suggested above, is that whereas their programme and Labov’s work
within it are based on first being able to locate linguistic changes in
progress, our work has been more generally based on describing
variation in the speech community and accounting for differing
patterns (whether or not they exhibit change in progress) in social
terms. In other words, our notion of the embedding of change in the
speech community is broader, and as a result our idea of what a
linguistic change actually is is broader and less traditional than
Labov’s view appears to be. It also raises the question of what a
sound-change actually is: how do we know when we have located one,
and how does the pattern of a change differ from other patterns that
we might locate?
I shall discuss these matters more fully in later chapters, but I can
lay some of the groundwork here. Traditional codifications of
sound-change have generally focused on sound-segments as they
‘change’ across time. Thus — to simplify — a linguistic change can be
described as a change from A to B in some lexical set, such as that of
Old English [a:] in stan, ham, which in the course of time ‘becomes’
an [o:]-like vowel in PresE (Present English) stone, home. The
transitional stages can be postulated in the Middle English period or
studied directly, as in Kristensson’s (1967) study of onomastic
sources from abround 1300, from which figure 1.1 is compiled,
showing the northward progress of the ‘new’ rounded vowel in this
set at that time (the rounded vowel has penetrated further northward
in the west, that is, in Lancashire and the West Riding, than in the
east, that is, in Lincolnshire). Labov’s treatment of sound-change
seems to be quite similar to this traditional treatment in that a change
 Introduction: Language Change and Variation 1

OE ac, brad, Brading, *ca, da, draf, han, hlaford, *hlafording, (ge)lad, ra, rap, stan
ON bar, gas, grar, pa, skali, vré in uncompounded names and ME pacok/pécok

o o-form
e@ a-form

ORKSHIRE ~
gwe RIDING) |
) 6

Figure 1.1 Northward progression of 6 for OE 4a (in words of the type ac:
‘oak’) c.1290-1350. (Adapted from Kristensson, 1967)

is generally located by comparative methods (comparing different

social groups in real and apparent time) within a single segment or a
very limited class of sounds.
We have also used these methods, but our conceptualization of
linguistic change is broader and is largely based on the normative
principle that I proposed above: linguistic change is to be understood
. more broadly as changes in consensus on norms of usage in a speech
community. During the process there will be some disagreement or
conflict on norms at some levelsin the community, but if a change is
ever ‘completed’, then it will be possible to say that some community
of speakers agrees that what was formerly A is now B. But this can
apply at different levels of generality — from a single sound-segment
up to a language state as a whole. Thus—to take a much more
generalized case than, say, post-vocalic /r/ in New York City —if a
18 Introduction: Language Change and Variation

language state is observed to become more (or less) homogeneous

within itself in the course of time, then the trend to greater or lesser
homogeneity is itself a pattern of linguistic change that has to be
accounted for in terms of consensus or conflict amongst speakers
within the speech community.
We might wish to look, for example, at Australian English in this
way. It is much more homogeneous than British English, even though
the early settlers came from many different places. If it has moved
from an early heterogeneous state to a more homogeneous one, this is
itself a linguistic change relevant to the history of Australian English
and aspects of colonial language development in general. Thus, many
of the detailed patterns that we are likely to find in sociolinguistic
inquiries will not be unidimensional (as more traditional work often
suggests), but bidimensional or multidimensional, showing trends
towards greater agreement or greater disagreement on norms within
the communities (see furtherJ.Milroy, 1982b, and chapters 3 and 4).

1.5 Synopsis

The purpose of chapter 2 is to place the general theme of this book in

perspective by considering the relationship between historical linguis-
tic explanations of a non-social kind and sociolinguistic explanations
of language change. I do not attempt to review all current intra-
linguistic work on change, as my main aim is to propose a different
(social) way of looking at change, but I discuss some of the main
trends. Following out my social arguments and Principle 1, I shall
focus on conversational settings as the locus of change, and I shall
suggest that language change is made possible to the extent that it is
passed from person to person in speaker encounters, in which the
apparently dysfunctional nature of language change is counteracted
by features of the communicative context (this idea is developed from
our early work in Belfast on speech and context: L. Milroy and J.
Milroy, 1977).
Chapters 3 and 4 are chiefly about analysing language in the
community and interpreting the patterns revealed, bearing in mind
Principles 2 and 3 (on linguistic norms and language maintenance).
The main database is the Belfast research, from which I have selected
examples. My purpose is to build the general foundations in these
 Introduction: Language Change and Variation 19

chapters, starting with the observation and analysis of language in the

community. In chapter 5, I extend the perspective backwards in time,
and consider some case-studies of language maintenance over peri-
ods of time that are relevant to historical interpretation. In this
chapter we have to consider what we mean by a sound-change in
history, and we conclude with a brief discussion of Neogrammarian
views on how sound-change is implemented.
Chapters 6 and 7 are the main theoretical chapters and are devoted
to the social modelling of linguistic change. In chapter 6, I outline a
social model which is derived from the social network model that we
have used to study language maintenance, and I propose a model of
‘weak ties’ to account for the possibility of language change. As this
model does not account for broader social structures and processes, I
explore in chapter 7 the links between network and social class in an
attempt to build up a more comprehensive social model for the
interpretation of language change.
 2
Social and Historical Linguistics

2.0 Explaining Language Change

The ultimate aim of historical linguistics is to explain the causation of

linguistic change. The question of causation is beset with difficulties,
but we can focus on it here by stating Weinreich, Labov and Herzog’s
formulation of the actuation problem (1968: 102): ‘Why do changes in a
structural feature take place in a particular language at a given time,
but not in other languages with the same feature, or in the same
language at other times? This actuation problem may be regarded as
the very heart of the matter.’ In attempting to solve the actuation
problem we are concerned with no less than the origin of change: we
want to locate its beginnings and by any means possible attempt to
explain why that particular change was initiated and diffused at some
particular time and place. It seems clear that to tackle it, we must take
account of how speakers initiate changes, and I shall treat it in later
chapters in these terms.
The actuation problem, however, is so challenging that historical
linguists do not usually address it directly; this is hardly surprising as,
when it is formulated in this way, it is actually insoluble: a solution to
it implies the capacity to predict, not only what particular change will
happen, but also when and where it will happen. However, the
probability of any event in life actually taking place at some particular
and specified place and time is close to zero. Weather prediction is a
convenient analogy here: we can predict from meteorological obser-
vations that it will rain on a particular day with a high probability of
 Social and Historical Linguistics 21

being correct, but if we predict that in a particular place it will start

raining at one minute past eleven and stop at six minutes past twelve,
the probability of the prediction being correct is vanishingly low.
Nevertheless, we would be bad meteorologists if we did not try to
improve the accuracy of our predictions, and of course this greater
accuracy includes the ability to specify the conditions under which
something will not happen as well as the conditions under which it
will happen. In view of all this, we have no excuse as linguists for not
addressing the actuation problem.
The Weinreich, Labov and Herzog formulation has several impli-
cations that are important for a theory of language change, and some
of these can be understood fairly readily if we cite as an example a
kind of sound-change that is frequently observed in languages and is
sometimes called ‘natural’. So let us consider here the palatalization
of /k/ before front vowels. Suppose it happens (as it often does) that
one particular language (or dialect) undergoes this palatalization,
whereas a closely related language (or dialect) of very similar
structure does not. Following the Weinreich, Labov and Herzog
principles, we have to ask why it happened in one variety but not in
the other. We also have to ask why it should have happened at some
particular time and not at some other time, when the structure of the
relevant language presumably exhibited suitable conditions for the
change at times when it did not happen as well as at the time when it
did. There are, of course, well-known examples of varying develop-
ments of this kind: amongst the continental Scandinavian languages,
Swedish and Norwegian have palatalization of Old Norse /k/,
whereas Danish now usually has a velar; Old English underwent
palatalization before front vowels whereas Old High German and Old
Norse did not: hence PresE cheese for German kdse and English/
Norse doublets in PresE such as shirt/skirt; many Hiberno-English
dialects (J. Milroy, 1981, and elsewhere) have [k]-palatalization in
words of the type car, cart, whereas most other English dialects do
‘not.! What we observe here are conflicting patterns of change and
stability in languages and dialects of similar structure. In these
examples it seems that the proximity of the velar consonant to a front
vowel may be a necessary condition for palatalization, but as it does not
happen in every case, it is not a sufficient condition. We need to find
out what the other conditions favouring or preventing the change
might have been, and it seems that in cases where the change was
adopted the social conditions must have been favourable; conversely,
22 Social and Historical Linguistics

when it was not adopted, it may again have been social conditions that
prevented the change. This suggests that to make progress in
understanding actuation we must take into account the activities of
speakers in social contexts in addition to the internal structural
properties of language.
Indeed, although linguistic changes are observed to take place in
linguistic systems, they must necessarily come about as a result of the
activities of speakers. As we have noted in chapter 1, languages which
have no speakers (or—sometimes — writers) do not change, and so
these remarks may well seem uncontroversial to any non-specialist
who has given thought to the matter. After all, there is no point in
having a language if it is not used by human beings. It seems to be
specialists, rather than non-specialists, who think that language
change can be explained without reference to society. Within ortho-
dox historical linguistics, the emphasis has generally been on the
properties of linguistic systems, and speaker-roles have been referred
to indirectly and sometimes very vaguely. As Lass (1980: 120-2)
points out, historical linguists have tended to regard language as an
‘autonomous formal system’ or natural object and have preferred to
believe that it is ‘languages that change and not speakers that change
languages’. Thus, historical descriptions and theories of historical
change have generally focused on the structural properties of langu-
age and not on its speakers. This tendency is very deeply embedded
in historical theorizing, and it is appropriate to look at it a little more
closely here.

2.1 System-based Accounts of Change

The orthodox position as stated by Lass (1980) is not entirely a

twentieth-century phenomenon: the separation of language from
speakers has an ancient and honourable pedigree, and the
nineteenth-century emphasis on the independent ‘life’ of language is
by present-day standards very striking. According to Trench (1888:
223-4), language has a life ‘as surely as a man or a tree’, and creativity
in language in developing new forms is attributed by Max Miiller (1881:
33) not to the creativity of speakers, but to the ‘marvellous power of
language’ itself: according to him (1861: 36) ‘it is not in the power of
man either to produce or prevent’ linguistic change. Miiller’s adop-
 Social and Historical Linguistics 23

tion of the biological metaphor is so strongly stated that for him it

does not seem to have been a metaphor at all: linguistics, according to
Miller, is literally a physical science on a par with geology, botany and
biology, and not a historical science, such as art, morals or religion.
‘Physical science’, including linguistics, ‘deals with the works of God’
whereas ‘historical science deals with the works of man’ (1861: 22).
Language therefore does not have history, it has growth. The meta-
phor has weakened since Miiller wrote, but there have been many
publications on language history since then that have been based on
the idea of the independent ‘life’ of language. Indeed, the metaphor is
by no means dead: this is amply demonstrated by continued refe-
rences in recent work to ‘language birth’, ‘language death’ and the
‘roots’ of language.
Of course, it is not true that language is a living thing (any more
than swimming, or birdsong, is a living thing): it is a vehicle for
communication betmeen living things, namely human beings. Hence,
the metaphor has been largely replaced since the nineteenth century
by a new metaphor based on the machine age: language is now more
often seen as a self-contained system, like a working machine. The
acceptance of this metaphor is widespread enough for it to appear in
the title of a book on linguistics — The Twitter Machine (Smith,
1989) —and it is clearly greatly encouraged by developments in
computer modelling of language. However, whereas the nineteenth-
century metaphor could readily incorporate the idea of change (as
language was said to have ‘growth’ within it, like a plant), this
system-based approach cannot so easily do so. Internal combustion
engines, for example, do not initiate structural changes within
themselves. From this point of view, therefore, the system-based
model may seem to be an unsuitable one to use as the basis for
studying language change. What is certainly clear is that within this
perspective our attempts at explanation continue to be essentially
language-internal. When speakers are referred to, they are decontex-
. tualized and asocial abstractions.’
Possibly as a result of the emphasis on internal language systems,
descriptive accounts (such as histories of English) commonly separate
the internal history of a language from its external history (that is, the
political, social and attitudinal contexts of language). Thus, some
historical accounts of English, such as Wyld (1927), have been mainly
internal (typically focusing on sound-change and morphological
change), whereas others (such as McKnight, 1928) have been about
24 Social and Historical Linguistics

the external history of the language, discussing, for example, speaker-

attitudes to variation as they were expressed by seventeenth- and
eighteenth-century commentators. Both of these approaches can of
course yield insights; however, it is commonly believed that the ‘real’
history of a language is its internal system-based history and that the
external history is relatively unimportant. The traditional position on
internal and external histories has again been clearly stated by Lass
(1987: 34-5), who claims quite explicitly that in most respects
‘external’ accounts do not help to explain changes in linguistic
structure. According to him, ‘there was nothing in the 17th century
English political or social climate’ that could account for, for example,
the merger of the formerly distinct vowels in words of the type bird,
fern, hurt; he further states that ‘at the structural-level there is no
connection between language and society’ and that ‘the internal life of
language is close to autonomous.’ My position, which | shall further
develop below, is — on the contrary — that we cannot hope to explain
change without inquiring into social factors.°
One reason for this is that intra-linguistic arguments with only
vague references to speakers, or accounts that explicitly reject
speakers, are not in themselves capable of dealing with actuation, as it
is speakers who actuate changes. Nevertheless, the causes of change
(like the causes of illness) are multiple; therefore, we need to take
both speakers and systems into account and, if possible, specify the
link between speaker-activity and change in language systems. As for
intra-linguistic theorizing, its main contribution has been to specify in
a more and more refined way the linguistic constraints on change, not
its causes. In order to exemplify the mode of argument used in
system-based accounts, we can consider here Kiparsky’s recent
review (1988) of progress in the study of phonological change.
This account is system-based and set in the traditional controversy
about whether sound-change operates blindly and without exceptions
(the Neogrammarian exceptionlessness hypothesis), or whether other
approaches over the last century (such as Jevical diffusion) have
invalidated the hypothesis. The lexical diffusion model (Wang, 1969)
holds that sound-changes may be lexically gradual: thus, in a change
from /e:/ to /i:/ (such as the EModE (Early Modern English) change
in words of the type meat, peace, leave), items are transferred to the
new class at differential rates, often leaving a residue of items that do
not get transferred (in this case such words as great, break, steak).
Neogrammarian theory, however, has generally been interpreted to
 Social and Historical Linguistics ji)

mean that the relevant class of items all undergo the change at the
same time, that is, that sound-change is phonetically gradual and
lexically sudden.
Kiparsky reconciles these approaches by arguing in terms of lexical
phonology: those changes that appear not to fit into the Neogramma-
rian hypothesis (including instances of lexical diffusion, in which
items in a class are not affected simultaneously by a change) are part
of the /exical rule component, whereas Neogrammarian exceptionless
change is accounted for by post-lexical rules. There is more to his
argument than this, but I am not concerned here, of course, with the
precise content of the argument — although I have tried not to do any
gross injustice to it—but with its intellectual background, and
specifically its intra-linguistic nature. In this argument, certain points
are evident. First, the argument is system-oriented and not speaker-
oriented (specifically, it is about phonological rules), and its goal is a
‘grammar’ of linguistic change; the activities of speakers are not given
prominence in the argument. Second, it is set in the traditional
controversy about the regularity of sound-change in language systems
(and certain other binary distinctions that arise from it, which attempt
to specify constraints on change: for example, whether sound-change
is lexically gradual or sudden). Third, it is proposed that the two
patterns of change can be accounted for by fitting them into a new
binary taxonomy based on lexical phonology. Thus, the discussion
casts new light on an old controversy by redefining the controversy
within a non-social system-based linguistics; it specifies the problem
in a more refined and elegant way, and its proposals are then subject
to critical discussion, testing and further refinements.
But this account, like many others, does not primarily address the
actuation problem, and the question why (and how) speakers initiate
changes is not central to the intellectual context in which it is
conceived. It is true that there are two pages on ‘causes of sound
change’ that acknowledge the work of sociolinguists in this
‘area — particularly Labov, Yaeger and Steiner (1972) and Kroch
(1978) — but the discussion of ‘causes’ does not form the backbone of
the account. The main aim of historical phonology, as represented
here, is the construction of sophisticated system-based grammars of
change, and the problem of how speakers actuate changes is not
directly addressed. It is interesting (and much to be welcomed),
therefore, that Kiparsky includes an extensive and skilful discussion
‘of the Labov—Sankoff variable rule paradigm. The significance of this
26 Social and Historical Linguistics

here is that variable rules are, of course, themselves explicitly

system-based (Cedergren and Sankoff, 1974, and see chapter 6,
below, for a fuller discussion of this point): they are grammars of
language and not accounts of actuation. Thus, whereas speaker-
based studies cannot easily be incorporated into orthodox system-
based accounts, variable rules can fit in perfectly and add sophistica-
tion to grammars of change.
A second quite recent example of intra-linguistic argumentation is
Lass’s On Explaining Language Change (1980), to which I have
referred above. Lass states that linguists have proceeded language-
internally and have taken the view that it is languages that change, not
speakers that change languages. But he is also critical of sociolingu-
istic accounts of change and seems to dismiss them. He says that
when attempts have been made to introduce the speaker into
explanations of language change, these attempts have been ‘super-
ficial and otiose’, and he considers (p. 121) even Labov’s notion of
‘speech community’ to be ‘a very tenuous abstraction’ (more on this
below). Taking his cue from the tradition, he points out that the most
fruitful results for theories of language change have come about
precisely because historical linguistics has studied ‘formal objects and
their mutations over time, not... their inventors or users’.
The point that great advances in the past have come about through
intra-linguistic argumentation cannot be disputed. One _ thinks
especially of the great innovators of the nineteenth century, such as
Bopp, Rask, Grimm and Verner, but also of modern advances, such
as work on language universals, lexical phonology and many other
areas. However, it plainly does not follow from any of this that we
should therefore neglect the role of speakers in linguistic change. If
we do not know what role speakers play, it seems appropriate that we
should investigate it empirically.
Before we look at this more fully, however, we need to notice that it
would not be correct to infer from what Lass says that older
generations of scholars always neglected the role of speakers. On the
contrary, there are several great names who assumed that linguistic
change must have social origins (amongst others) and who did not
think it beneath them to write extensively on this, while at the same
time contributing fully to system-based accounts of language. Among
these are Henry Sweet, H. C. Wyld, E. H. Sturtevant and Otto
Jespersen. In Sturtevant’s Linguistic Change (1917), we find an
emphasis on the idea of social norm-enforcement, childish errors and
 Social and Historical Linguistics |

slips of the tongue, and (incidentally) a plea for the study of universals
of language change. All this is presented in a framework that
distinguishes primary change (compare our idea of speaker-innovation)
from secondary change (effectively, linguistic change as admitted into
grammars of language). Thus, a speaker/system distinction similar to
that which we have proposed (J. Milroy and L. Milroy, 1985b) is
considered by Sturtevant, and he attempts to integrate the two sides
of the question. Jespersen’s Language (1922), which is a better-known
book, reads in places like a research proposal for modern sociolingu-
istics and language acquisition studies: the possible causes of change
include features of children’s language, sex-differences (there is a
chapter on “The Woman’), taboo and euphemism (unfashionable at
the moment, but unquestionably very important), language contact,
and Pidgin and Creole development.
It seems that if speaker-based arguments have since that time been
found unsatisfactory, there may have been contingent reasons for this.
The correct generalization seems to be, not that speakers are
irrelevant to change in language systems, but simply that it has in the
past been extremely difficult to study speaker-behaviour in a systema-
tic and accountable way. As we noticed in chapter 1, early invest-
igators, such as Wyld, did not have the necessary technology — in
particular they did not have tape-recorders. So, until about 1960 it
was very difficult to explore their ideas further by experiment or
systematic observation and impossible to study conversational inter-
action in a reliable way. Dialect investigations were usually limited to
single-word citation forms, and the empirical study of discourse in
situational contexts could hardly have been contemplated. This,
clearly, is no longer true. Although we are still interested in ‘formal
objects and their mutations over time’, advances in data collection
from live speakers, and the analysis of such data, have put us in a
position to inquire into the role of speakers in change. If we do not
know what role speakers play in implementing changes, we can
inquire further into that role — we do not need any longer to dismiss it
as impossible to explain.
Before we go on in the next section to amplify this discussion by
looking at the functions of change, we can draw together the strands of
the argument so far. Historical linguistics in this century has largely
depended on the idea that language is a self-contained system, and
investigators have generally not systematically addressed the question
‘of how speakers can introduce changes into the structural parts of
28 Social and Historical Linguistics

language. Thus, complaints about the superficiality of ‘external’

explanations of change are in a sense self-fulfilling prophecies.
Clearly, if you concentrate exclusively on abstract systems of language
and do not develop a coherent and accountable theory of the social
embedding of language change, your comments on the possible social
reasons for changes will inevitably be ad hoc and superficial.
But although the above remarks are concerned with theory rather
than description, they are relevant to descriptive accounts also: for
example, to historical descriptions of English. It is impossible to write
a reasonably full history of English without making at least some
reference to social categories, such as class, and to institutional
aspects of language, such as standardization. Indeed, system-based
accounts, such as that of Lass (1987) himself, do routinely make
reference to ‘class’, ‘prestige’ and other socially-based ‘external’
categories. What most of these histories of language have done is to
refer to these social matters in an ad hoc way, without contemplating
an accountable theory of the social embedding and motivation of
language change. Indeed, as we noticed in chapter 1, some invest-
igators, being themselves embedded in a social matrix, have filtered
their social explanations through their own ‘common sense’ views of
social class and social prestige, without clearly acknowledging that
these are themselves complex theoretical concepts, and certainly not
‘common sense’.
There seems to be no reason why we should not inquire further
into social structures and processes as part of an inquiry into
linguistic change, and no reason why we should. not also think of
linguistic change as being an aspect of social change in general, that
is, to think of it in an entirely new perspective. But to do these things
is not to exclude the possibility of also developing sophisticated
internal accounts of language change. Both kinds of approach are
needed — and one should contribute to the other — because although
linguistic change must be initiated by speakers (and is therefore a
social phenomenon) it is manifested as internal to language. Bearing
these questions in mind, therefore, I shall move on in section 2.2 to
consider in what senses linguistic variation and change can be shown
to be functional. It is important to consider this question, because from
an intra-linguistic point of view variation and change can actually
appear to be dysfunctional.
 Social and Historical Linguistics 29

2.2 The Functions and Malfunctions of Change and

Variation

In the introductory sections of chapter 1, we were mainly concerned

with structural aspects of language, exploring the apparent conflict
between the structuralist axiom —that a language is a coherent
self-contained system of interdependent parts —and the fact that
language is continuously changing. In what follows I am concerned
with function: | want to consider whether linguistic change is func-
tional, that is, whether it serves a purpose of some kind, and if so in
what sense it is functional. Here we need to recall that it is very widely
asserted by linguistic scholars that a language system is at any given
time equally well adapted to the functions for which it is used:
sometimes it is said to be perfectly adapted. Whether or not this is true,
it is reasonable to assume that linguistic structure is very sensitive to
the changing social and communicative needs of speakers. Further-
more, Weinreich, Labov and Herzog (1968) have claimed that
linguistic variation (which can be a symptom of language change) is
not only structured, but also functional, and that it 1s the absence of
variation that would be dysfunctional.* In order to investigate the
functions of language, however, we have to look at language in use,
but as the functions of language in use are social and pragmatic, they _
are not readily accounted for within an exclusively system-based
theory of language structure. We therefore have to look at speakers in
addition to systems. Traditional speculation as to whether linguistic
change is functional or not has, however, been based on the
configuration of systems rather than on speaker-use of language.
Speculation about the functions of change has been common in the
past: it has, for example, been suggested that there is a teleology, or
overall purpose, in language change (for a discussion of this, see Lass,
1980), and even that change in language structures may follow a
predestined path. This is plainly related to the idea of language as an
independent ‘growth’ with a ‘life’ of its own. In this view of language,
however, speakers would play little part in its development and could
do very little about linguistic change, as the blind force of language in
its purposive quest would overrule them (recall again the views of
Miiller, 1861). Related to these ideas in a general way is Sapir’s
(1921) notion of ‘drift’. It is striking, as Prokosch (1939) points out,
30 Social and Historical Linguistics

that related languages can undergo the same changes apparently

independently: all the early Germanic languages, for example (except
Gothic, records of which are too early), independently underwent the
important change known as front-mutation or i-umlaut. It is as if the
‘parent’ language was programmed in such a way that the conditions
were already present in it for the change to take place in all the
‘daughter’ languages.
It has also been suggested that language can make progress and
improve in the course of time: for Henry Sweet (Henderson, 1971),
the loss of grammatical gender in English constituted progress, as
grammatical gender, according to him, is ‘illogical’; Jespersen (1922)
argued that modern languages of ‘analytic’ (weakly inflected) struc-
ture, such as English and Danish, have evolved to a higher stage than
‘synthetic’ (heavily inflected) languages, such as classical Latin, and
are more efficient instruments of communication, mainly because
they are thought to have a more transparent one-to-one relationship
between meaning and form.’ For example, an auxiliary verb phrase,
such as English J have said... , is more transparent than the Latin
equivalent dixi because person, tense and aspect in English are
expressed in separate units, whereas in Latin they are all carried in
the second syllable of dixi and cannot be unravelled from its surface
form. Jespersen’s judgement here depends, of course, on the assump-
tion that transparency in language structure is a desirable thing.
Clearly, however, if some other criterion were used (and such criteria
as euphony, economy and elegance have also been used), completely
different conclusions might be drawn, and Latin might then be held
to be superior to English, as it often has been; therefore, it is safer to
assume that differences in overall grammatical structure are neutral.
They can all be used equally efficiently or inefficiently by the speakers
of the languages concerned.
I have described these kinds of argument as speculative because it
is impossible to test them empirically: they are based on value
judgements, which are often ethnocentric or class-based, and in the
wrong hands they can sometimes lead to quite damaging opinions.
For example, if it is believed that the structure of one language is
superior to that of another, it can then be suggested that this is due to
the cognitive superiority of its speakers. In fact, there seems to be
nothing inherent in the structural properties of language to suggest
that change has a positive function within language structure.® Within
these structural parts — phonology, grammar, lexical and semantic
 Social and Historical Linguistics 31

structure —it is quite impossible to demonstrate empirically that

language systems have in-built tendencies towards progress or decay,
that one language is more or less ‘efficient’ than another, or that there
is a teleology in linguistic change. If change is functional, it must be
speaker-functions that are involved.
In fact, if we focus exclusively on the internal properties of
language, it is much easier to make a prima facie case for the argument
that variation and change in language are dysfunctional, rather than
functional. This is because dialect divergence and language change
lead to difficulties in communication between speakers — a fact that is
obvious to any fieldworker who studies a ‘divergent’ dialect, and to
any analyst who transcribes the tape-recordings (they can be very
difficult to understand); it is also familiar to anyone who tries to read
an early English text. Therefore, if linguistic structure exists for the
purpose of successful communication, why does language change?
Why do languages diverge from one another in the course of time,
and why are some dialects of a language partly or wholly incompre-
hensible to speakers of other dialects? The apparent paradox has of
course been noticed by many, and it has been quite recently
commented on by Francis (1983: 15-16):

If the purpose of language is communication, it would seem that the

more homogeneous the language, the more efficient the communica-
tion. Why, then, does the increasing variation resulting from differen-
tial change not make communication difficult, unreliable, and even-
tually impossible? If so, the propensity for language to change would
ultimately lead to a breakdown in the principal function for which
language exists.

It is interesting that Francis — as a dialectologist — seems to presup-

pose here that increasing divergence does not lead to breakdown of
communication. One only has to look at the histories of related
‘languages that were at one time dialects of the same language to
realize, of course, that it does lead to breakdown. There is no need to
disagree with Wyld (1936: 7): ‘the process of differentiation is almost
infinite, and the tendency of language is not, as it has sometimes been
wrongly said, in the direction of uniformity, but of variety.’ Language
change can, and does, lead to breakdown of communication.
The belief that language change is dysfunctional is most clearly
expressed in popular attitudes to language. These commonly con-
32 Social and Historical Linguistics

ceive of languages as ideal and perfect structures, and of speakers as

awkward creatures who violate these perfect structures by misusing
and corrupting ‘language’: this is essentially a belief in the rape of
languages by speakers. These attitudes are strongly expressed and
highly resistant to rational examination. So strong is this intolerance
of speaker-variation and change that in many countries academies
have been set up to enforce a uniform ‘correct’ usage and to prevent
uncontrolled divergence; indeed, in the biblical story of the Tower of
Babel, language diversity is attributed to original sin. There are, of
course, socio-political and economic reasons for these attitudes,
which we have discussed elsewhere (J. Milroy and L. Milroy, 1985a),
but they are powerful and deep-seated and they cannot be ignored.
In so far as these prescriptive agencies have a rational purpose, this
purpose seems to be the maintenance of communicative efficiency in
carrying information-bearing messages over long distances and peri-
ods of time. For conveying information in these ways, uniformity and
standardization of language are highly valued, and it is usually in the
written channel that the highest level of this kind of efficiency is
achieved. Spoken language, as far as we can tell, continues to vary and
change, and it is in spoken language, and not in writing, that
structural (for example, phonological) change is implemented. Writ-
ing systems promote uniformity and suppress variation and change: it
is typically in the day-to-day situational contexts of speaker-
interaction that structural changes take place, and it is in these
contexts that they have to be investigated. That is why Principle 1,
which | put forward in chapter 1, should be borne in mind.
Popularly expressed attitudes to correctness and uniformity, there-
fore, do not square with what human beings actually do in conversa-
tional contexts. On the one hand they believe in uniformity, while on
the other they promote diversity. Clearly, if speakers consistently
carried their expressed beliefs into practice, the result would be a
uniform and stable state of language —the world of the idealized
native speaker in a perfectly homogeneous speech community — and
not the diversity that exists in the real world. It seems, therefore, that
in expressing adverse judgements on variation, speakers are subscrib-
ing to the notion that the main function of language should be the
successful communication of information-bearing messages (as in the
writing system), and there is plenty of evidence that from this point of
view variation and change can indeed be dysfunctional. But there is
also plenty of evidence that the successful communication of
 Social and Historical Linguistics 33

information-bearing messages is not the only function (or necessarily

the main function) of language in use. To clarify the argument, I shall
now consider some examples of communication difficulties that arise
from language diversity.

2.3. Malfunctions of Language Diversity

Cross-dialectal miscomprehension was one of the interests that led to

the setting-up of the Belfast research programme.’ We were inter-
ested in the first place in how commonly differential linguistic
structures could lead to miscomprehension, and more specifically
whether ‘non-standard’ speakers actually have the easy access to
standard English that is so often assumed. This has further theore-
tical implications — especially for the idea of the ‘polylectal’ or
‘pan-lectal’ grammar of English, which was a current interest in the
1970s. Within the projects, we collected many examples of cross-
dialectal divergence and miscomprehension, of which the following is
an example. Speaker A is a speaker of a Hiberno-English dialect, and
B and C are ‘standard’ speakers:

(1) A How long are yous here?

B Oh, we’re staying till next week.
(silence of about 2 seconds)
C We’ve been here since Tuesday.
A Ah well, yous are here a while then.

The miscomprehension here is indicated by the period of silence

after B’s reply, and it arises from the clash between different
linguistic systems. In certain constructions Hiberno-English dialects
consistently use the present tense where standard English uses the
present perfect. Thus, ‘How long are yous here?’ (in HibE) means How
long have you been here?, whereas in SE (Standard English) the present
tense form means (or at least implies) How long are you going to stay
here? or What is the total length ofyour stay? Speaker B construes the
HibE utterance according to the SE rules. An ‘appropriate’ response,
however (one which would be immediately perceived as relevant to the
question in this conversation), would in HibE refer to the past (for
example, ‘since Tuesday’) and not to the future. While it is true that
34 Social and Historical Linguistics

breakdowns occur for many reasons other than this, there is no doubt
that breakdowns arising from the different structures of divergent
dialects are quite common (see also Berdan, 1977; Trudgill, 1981),
and that they are naturally perceived as inconvenient when they are
noticed.
Sometimes, however, there is no period of silence or hesitation, as
in (1), and the conversation proceeds apparently normally, even
though there actually has been a misunderstanding. ‘The hearer may
continue to believe that the first speaker has said something quite
different from what was intended, or the hearer may wait for the
miscommunication to be ‘repaired’, or it may simply not seem to
matter very much at the time. But we do have attested cases where the
miscomprehension could be quite serious. In some Hiberno-English
(and Scottish) dialects, the conjunction mhenever is used almost
equivalently to when. Thus, if someone says: “Whenever my husband
came in he beat me’, we know from observing Ulster usage that the
reference is to one occasion only: the speaker is not suggesting that
her husband beat her every time that he came in. An outsider,
however, is likely to interpret this to mean that the husband beat her
many times. Consider also the following exchange:

(2) A Do you think he’s going to die?

BI doubt so.

B, who is a speaker of Hiberno-English, does not mean that he

believes that the person will not die; his meaning is the opposite. He is
saying that he is afraid that the person mill die (the utterance means
approximately ‘I’m afraid so’). In this case, | was Speaker A, and the
usage was known to me as a widely distributed one in Scots and Irish
dialects. If the interlocutors do not have access to the two different
systems, however, there will be a miscomprehension. Thus, if a
newscaster says ‘Mr Major doubts that the economy will deteriorate’,
speakers of these dialects may interpret this to mean that he is afraid
the economy mill deteriorate, when the speaker actually means that he
thinks that it won’t deteriorate.
These cases of breakdown arise from deep-lying differences in the
syntactic structures of the different dialects, which make the pan-
dialectal grammar (Bailey, 1973) of English a dubious proposition.®
The semantic distinctions carried in the Hiberno-English tense/
 Social and Historical Linguistics 35

aspect system are structured totally differently from standard English

(Harris, 1984;J.Milroy and L. Milroy, 1985a); thus, the miscommu-
nication in example (1) is due to major differences in the abstract
rule-systems of these dialects. Furthermore, all those involved in the
miscommunications reported above are native speakers of English,
but it appears that their ‘native speaker intuitions’ (or competences)
do not extend to the rules of all relevant varieties. These examples
also show, therefore, that our native speaker competence does not
necessarily guarantee comprehension of varieties that are removed
from us in time, space or social space. To put it simply — we do not
have total comprehension of dialects divergent from our own.
The assumption that speakers of a language are (or should be)
mutually comprehensible is rather basic in popular attitudes to
language, and it also seems to underlie some professional linguistic
approaches (for example, Smith and Wilson, 1981). Clearly, if there
can be miscomprehensions arising from structural differences in the
dimensions of spatial and social variation, it is likely that in the
chronological dimension of change there will also be miscomprehen-
sions. A change entering the language of a younger generation, for
example, may well be miscomprehended by older generation
speakers. However, the idea that mutual comprehensibility between
generations is a constraint on possible changes is quite deep-rooted
amongst historical linguists (see Lightfoot, 1979: 376, for a justifica-
tion), and is often taken for granted. But it is dangerous to assume
this too lightly: we cannot demonstrate that mutual comprehensibility
between generations (or between groups of other kinds) is always a
necessary factor in determining which linguistic changes are possible
and which are impossible, and sociolinguistic investigations (from
which most of the data reported in this book are derived) strongly
suggest that it is not always primary. It may be that some innovative
groups do not particularly wish to be comprehensible to others, or
that rapid social change (for example, in the genesis of Pidgin/Creole
languages) overrides the principle of mutual comprehensibility. The
constraint proposed by Lightfoot is, it seems, a variant of the idea that
heterogeneity is necessarily dysfunctional, and that mutual intelligi-
bility is functional. It seems to me, however, that the question we
should be asking is why linguistic changes that lead to miscompre-
hension (and divergence into mutually incomprehensible varieties)
can happen at all. What is the motivation for such changes, and what
is their function?
36 Social and Historical Linguistics

The examples discussed above concern variations in language

structure that lead to comprehension difficulties, but they have also
raised, in a much more general way, the question of what actually
happens in conversational exchanges between speakers. In section
2.4, I shall be concerned with this latter point: I shall attempt to show
that it is the multiple speaker-functions of language in use that make
linguistic change possible and suggest that we must look at these
speaker-functions if we are to make progress in understanding the
nature of language change. This implies a modelling of the locus of
linguistic change that differs from system-based models of the kind
that we have discussed The language-internal presuppositions of
these models have been projected on to the mental capacities of
human beings, and linguistic change has thus been seen as consisting
primarily of changes in the mental representations of the speaker in
the form of rule-addition, rule-loss, and other rule-changes. But as
we cannot directly observe mental representations (whether they are
described as rules or in some other way, for example, as parameter-
settings), suggestions of this kind are somewhat more speculative than
matters that can be verified by observing speaker-interaction. In
order to propose a more socially realistic account of change, there-
fore, Labov has argued that the locus of change is not in the
individual speaker, but in the group, or at least that we have to look
for it in group behaviour. What is implied here is more specific than
that: it is that linguistic change is located in speaker-interaction and is
negotiated between speakers in the course of interaction, much as
other aspects of discourse are negotiated between them. Bearing in
mind this speaker/system distinction, therefore, I shall introduce the
discussion of speaker-function in section 2.4 by first considering the
way in which functional change has been handled in system-based
historical linguistics.

2.4 The System-oriented Approach to Function

The system-oriented (as opposed to speaker-oriented) approach to

functional change originates with Martinet (1955), whose arguments
depend on the information-bearing function of language and the
presumed need to preserve mutual intelligibility. | have commented
briefly on this in chapter 1. Martinet proposes that in phonetic/
 Social and Historical Linguistics 7)

phonological change a phonetic opposition that is useful in maintain-

ing meaning-bearing distinctions will, other things being equal, resist
neutralization and loss of distinctiveness. Thus, for example, if the
speakers of English had found it useful in communicating informa-
tion to maintain the EModE phonetic distinction between words of
the meet, beet class and words of the meat, beat class, they would,
according to the theory, have been inclined to maintain it. The fact
that they did not (in some dialects) requires a functional explanation,
which in effect proposes that the distinction between the two
phoneme classes was no longer useful in maintaining meaning-
bearing distinctions, and this is generally argued in terms of functional
load (see especially Samuels, 1972). It is suggested that one of the two
categories had low functional load in distinguishing between words in
ordinary usage; for example, the words in that class might not have
been used very often, and when they were used, they were perhaps
used in contexts where confusion with the other class was unlikely to
occur (thus, as meet is a verb, whereas meat is a noun, it is unlikely that
you will misunderstand [ meet him as *I meat him).
This type of functional argument can also be used to explain
exceptions to ‘regular’ changes. For example, the idea of homonymic
clash may be used to explain why certain items in early English did not
undergo regular development: if they had, they would have become
phonetically identical with other forms. Lass (1980) cites the item
shut, which if it had developed regularly would have become identical
with shit: according to the theory this was prevented by the functional
need to keep the items distinct. However, it is easy to find counter-
examples, where the need to prevent homonymy did not operate: for
example, homonymy of rush (‘hasten’) with rush (a plant) has not been
prevented. In some circumstances quite dramatic loss of distinctions
can take place, and there are examples in the literature of wholesale
merger of previously distinct classes of items: Labov, Yaeger and
Steiner (1972) cite the reduction of five distinct vowel phonemes of
ancient Greek to one (/i/) in modern Greek. Because of apparent
counter-examples like these, Lass (1980: 75-80) points out that these
functional arguments as applied to given cases are unsatisfactory.
Social dialectologists are in little doubt that speakers will happily
tolerate a great deal of phonemic merger, allophonic overlap and
approximation, and homonymic clash.
None of this, however, detracts from Martinet’s original insight,
which recognizes the importance of speaker-function — in this case
38 Social and Historical Linguistics

the information-bearing function: the apparent circularity in arguing

about given instances is probably very largely due to the limitations of
the historical database (which is incomplete and deprived of situa-
tional context), and to some reluctance to investigate what happens in
the language of live speakers, where hypotheses such as functional
load can be more fully tested. Historical linguists know that language
is used to convey information, but they cannot specify very easily what
additional social and pragmatic functions might have been involved in
particular changes (which after all took place long ago in cir-
cumstances that we cannot fully investigate). Yet, although the
information-bearing function is the one that comes most easily to
mind, it is only one of the functions of language in use: other functions
interact with it. Thus, in historical change, the need for mutual
intelligibility over distances and the maintenance of meaning-bearing
distinctions can be overridden by these other functions — by the
identity-function, for example. This is suggested by the cross-
dialectal miscomprehensions discussed above, by the pattern of
multiple merger in Greek cited by Labov, Yaeger and Steiner (1972),
and by many other instances.
In this connection, it is useful to recall the metaphor that I appealed
to in chapter 1 — the idea that in viewing language as a system ow tout
se tient we are implicitly comparing it to a machine. The function of a
machine is known beforehand: the function of an internal combustion —
engine is to propel a vehicle, and if it does not succeed in propelling
the vehicle, we know that it has malfunctioned. If we believe that we
know beforehand that the principal function of language in use is to
communicate decontextualized information-bearing messages expli-
citly and unambiguously, then — pursuing the system-based meta-
phor — we shall have to view communicative breakdowns as malfunc-
tions. Of course, in given situations, they may be perceived in this
way, and if we put the argument on a more abstract and general plane,
we shall then have to conclude that language change and variation are
in general dysfunctional. But if this is so, why then do human
languages vary and change?
We can draw only one conclusion, which is that variation and
change must also be functional for speakers of languages: if this were
not so, languages would be uniform and they would not change.
Indeed, there is a case for claiming that if linguistic change were
impossible, speakers could not function adequately in speech com-
munities. But the functions of language in speech communities are
 Social and Historical Linguistics 39

multiple and are not limited to the information-bearing function. In

social dialectology we attempt to explore the question of function by
observing and analysing the language of ordinary speakers in conver-
sational contexts. ‘I’o return to the machine metaphor: we can take for
granted the function of a machine, but we cannot take for granted the
functions of language. Thus, as we do not know beforehand what all
the functions of language in the speech community might be, we have
to find out what they are and how they interact with one another by
exploring the speech community. This is by far the single strongest
justification for research in social dialectology of the kind that we and
others have carried out.

2.5 Speaker Functions in Discourse and Conversation

What I intend to do here is to tackle the question of discourse

functions in an introductory way, without venturing too far into the
enormous literature on discourse and conversational analysis. We can
start by noticing some of the binary distinctions between types of
discourse that have been suggested. Ochs (1979) proposes that
discourse can be divided into two broad types: planned and unplanned.
Brown’s (1982) distinction between message-oriented and _listener-
oriented speech is based explicitly on the functions of discourse in social
settings. Message-oriented speech is characterized by explicitness
and independence of situational context or shared knowledge be-
tween participants. According to Brown (p. 77), message-oriented
speech is ‘goal-directed. It matters...that the listener under-
stands... and that he understands... correctly. The point of the
utterance is ...communication of a propositional or cognitive
(information-bearing) message to the listener.’ Listener-oriented
speech, on the other hand, is characterized by inexplicitness and
vagueness, with primary attention to the feelings and attitudes of
conversational partners. As Brown comments: ‘it is often the case that
speakers in primarily listener-oriented dialogue don’t seem to be
talking about anything very much . . . we may judge it to be successful
if the participants succeed in maintaining friendly relationships.’
Historical linguistic views of language function tend to assume
without comment that the function of language is something like the
message-oriented function — which is often described as communica-
40 Social and Historical Linguistics

tive- and not like the listener-oriented function (in which the
meanings conveyed are social or context-dependent); they assume that
communication between speakers is, or ought to be, explicit and
context-independent in the interests of conveying cognitive proposi-
tions efficiently from speaker A to speaker B, in a context where the
aim is to convey new information to someone who does not already
know it. Let us consider, however, what this would actually mean in
terms of the use of language in social contexts.
Clearly, it would allow the written medium, as the function of
writing is to communicate messages outside the immediate inter-
personal context, and, to be effective, it must be explicit. It would also
allow certain kinds of speech event, such as reports or lectures, for
much the same reasons. In conversational exchange,’ however, such a
view of language use would account for only those parts of conversa-
tions in which ‘new’ and explicit information is given and received; for
example, the information that is passed in question—answer adjacency
pairs of the kind:

(3) A Where were you on the night of 15 August?

BI was looking after Mother at the motel.

Such exchanges are, however, especially characteristic of, and fre-

quent in, formal settings, such as classrooms, courtrooms and
interviews of various kinds (including some sociolinguistic ones) and
may indeed be required in such settings: they are not especially
characteristic of conversation. The conversations that we collected
and analysed in the inner-city Belfast projects (L. Milroy and J.
Milroy, 1977), for example, could not have been adequately
described in these terms. Speakers in most conversational contexts
are not solely (or even mainly) concerned with passing decontextual-
ized new information to one another, and the casual conversations
that sociolinguists record are not mainly made up of question—answer
adjacency pairs or elicitation—response sequences. On the contrary,
much of our discourse is unplanned and listener-oriented, in which
speakers ‘don’t seem to be talking about anything very much’ and in
which the primary goal seems to be the maintenance of ‘friendly
relationships’ (Brown, 1982: 77). It can be assumed, therefore, that
speakers in casual social contexts are not usually concerned with
avoiding homonymic clash or with being especially clear and explicit:
 Social and Historical Linguistics 4]

they are satisfied if the conversation progresses successfully, and the

success of the conversation is judged in social terms. If misun-
derstandings occur because of homonymic clash or for any other
reason, they can be repaired if necessary: speakers appear to accept
the results of vagueness and ambiguity on the assumption that
‘intended’ meanings will be clarified if necessary as the conversation
proceeds.
Underlying this distinction between discourse functions there is
another more general distinction. This is the stark contrast between
what is desirable in the written medium (or context-independent
speech-styles such as lectures) and what is desirable in, and charac-
teristic of, speech-exchanges in social settings. It is very clear that
much of the historical linguistic tradition has been based on assump-
tions derived from the functions of writing, rather than speech.
However, such features as redundancy, vagueness and ambiguity,
which are disfavoured in writing, are wholly characteristic of everyday
speech. Furthermore, many of the features that are positively dys-
functional in context-independent language are actually functional
and necessary in the conduct of successful conversation: lack of
explicitness, hesitation, ambiguity, incompleteness and repetition are
themselves very important aspects of how conversation is organized.
This has been clearly demonstrated by conversational analysts such
as Schegloff (1979). Far from being random and disorganized,
ambiguity and the other characteristics we have mentioned are
systematic strategies of conversational interaction. They are used for
monitoring the reactions of conversational partners and for clarifying
and repairing the mistakes or misunderstandings that might have
occurred in the interaction. In fact, one of the most important aspects
of conversation is the very high value that is placed on indirectness: this
is in obvious contrast to message-oriented styles, in which directness
is valued and indirectness disvalued. Direct imperatives (demanding
actions), for example, and direct interrogatives (demanding relevant
responses, as in (3)) are quite rare in conversation. This is because
exchanges in speech are social and personal: the high value placed on
indirectness is a result of the speakers’ concern for their own ‘face’
and that of their partners (Brown and Levinson, 1987): they are often
much more concerned with being polite, that is, avoiding threats to
‘face’, than with passing information efficiently and economically. In
spoken contexts, the directness that is so highly valued in
information-bearing styles is perceived as threatening.
42 Social and Historical Linguistics

These remarks may be sufficient here to draw our attention to

some of the functions of language in use apart from the message-
oriented function, and they have some consequences for our ideas
about how linguistic changes are implemented. Theories based
exclusively on the message-oriented function of language must
plainly be insufficient: it seems very unlikely that linguistic changes
over time could have been implemented mainly in this function, that
is, in formal styles such as lectures, or in formal settings such as
courtrooms or classrooms, especially since this function values
stability and resists change; for the most part, changes have been
initiated in countless millions of casual (mainly unplanned, listener-
oriented and context-dependent) encounters between speakers. It is
in these casual exchanges (and not primarily in formal settings) that
the sociolinguist looks for evidence of change in progress. It is the use
of speech in these context-tied situations that actually allows lin-
guistic change to be negotiated between speakers, and we need to
emphasize this here, because linguistic innovations plainly belong to
the class of phenomena that may be miscomprehended in context.
If we can accept that it is characteristic of conversation that some
utterances will be miscomprehended, we can also presumably accept
that it is the principles of conversational organization that permit
repair of such miscomprehensions. Clearly, it is possible that in the
course of a conversation innovations may be introduced and, as they
may be unfamiliar to some participants, they may be miscompre-
hended; however, as we have seen, the principles of conversation
allow for clarification of these miscomprehensions. For this reason,
and for other contextual reasons, the principles of conversation
permit linguistic changes to be negotiated between speakers and thus
admitted into the language system: it is the conversational context that
provides the conditions for change to be accommodated. Decontex-
tualized discourse, on the other hand, does not in principle cater for
these misunderstandings: it therefore resists ambiguity and vague-
ness, and in the present perspective it is hardly surprising that in its
written forms especially it has also been observed to resist structural
change.
 Social and Historical Linguistics 43

2.6 Speaker-Functions: Marking Social Roles

The discussion so far has focused on the functions of discourse in

conversational settings, and I shall continue to refer to these pragma-
tic aspects of language from time to time. However, there is another
relevant aspect which is largely independent of the idea of discourse
types, and which has been prominent in social dialectology since
Labov’s (1966) New York City study. This proposes that variation
within the structural parts of language (such as phonological and
morphological variation) is used by speakers to mark varying social
roles. Social meanings—in the words of Blom and Gumperz
- (1972) - are carried in linguistic structures. Style-shifting and code-
switching (switching from one language or dialect to another) are
socially functional: they are related to changes in the situational
contexts of speech events, to the social characteristics of the partici-
pants, and to the varying purposes of exchanges in speech. Further-
more, speakers normally attach great importance to this kind of
variation and assign strong social values to what are essentially
arbitrary differences. To exemplify this kind of function, I shall now
briefly consider the question of style-shifting (which functions in
essentially the same way as code-switching).
The fact that style-shifting is functional is rather neatly demons-
trated in cases where the expected variation is, apparently, absent. We
have discussed such a case in some detail elsewhere (J. Milroy and L.
Milroy, 1985a: 123-5), and this example is a useful demonstration of
Weinreich, Labov and Herzog’s (1968) point— that the absence of
variation is dysfunctional. According to Lavandera (1978), members
of the Argentinian Italian community, who are bilingual speakers of
Italian and the cocoliche dialect of Spanish, are perceived by mono-
linguals as deficient in their Spanish. However, there is actually
nothing deficient about the structure of their Spanish (as Lavandera
shows): what is lacking is the sty/istic variation, sensitive to occasion of
use (and other factors), that monolinguals observe. There is a lack of
stylistic variability, not a deficiency in command of the ‘core’ structure
of the language. The perceived absence of variation in this case
demonstrates, of course, that stylistic variation is —for the mono-
linguals — functional; that is why they notice it. For cocoliche speakers,
on the other hand, it is their bilingualism that is functional: their
aA Social and Historical Linguistics

communicative competence is exhibited in their command of two

languages, rather than in observing the stylistic variation inherent in
one of them.
But it is also clear that speakers may have very strong feelings about
particular regional or social dialects, even though in linguistic terms
the differences between dialects are arbitrary. One of the first things I
noticed in my early descriptions of Belfast vernacular was the strong
‘stigma’ associated with certain non-standard pronunciations. In a
dynamic account of the phonology J. Milroy, 1976a), I found it most
convenient to think in terms of the avoidance of stigma rather than
convergence towards a higher-class or standard form, but it was not
clear why some non-standard forms were avoided and others fav-
oured. Amongst younger inner-city speakers, it was clear that stigma
was attached to certain forms of rural origin (such as palatalized /k/
and dental /t/), but not to others (the raising of /a/ before velars
showed no recessive tendencies). Thus, it seems that communities
can disfavour pronunciations that were formerly favoured; and
despite the fact that sound-segments do not carry meaning in the
ususal sense, this phenomenon seems to have something in common
with the operation of taboo, which is so well attested in the
vocabulary. It is also associated with ‘face’ and politeness.
These brief discussions of social dialect, style-shifting and conver-
sational functions are sufficient, I think, to indicate that there is more
to language in use than the communication of decontextualized
information of a purely cognitive kind. It now seems appropriate to
summarize the main points, as I have not been trying to argue merely
that language in use has multiple functions — this is a view that will be
readily accepted by any sociolinguist. I have been concerned with the
limitations of what I have called system-based historical linguistics.
By this I mean virtually any approach to historical linguistics (tradi-
tional or current) that is centred entirely, or almost entirely, on the
properties of language as an abstract object, and that excludes the
systematically observed behaviour of speakers of languages. These
language-internal approaches have made immense progress in pro-
ducing sophisticated ‘grammars’ and models of linguistic change, but
they have not come very close to the actuation problem and the causes
of change. Furthermore, as we have noticed above, when speaker-
roles are referred to in system-based arguments, they tend to be
referred to in a rather ad hoc and unsystematic way, usually on the
assumption that the message-oriented function of discourse is the
 Social and Historical Linguistics 45

one that matters. We therefore need a theory of the embedding of

language change in society, but we do not so far have a social theory
of this kind that can rival the sophistication of system-based linguistic
theory. Yet, it seems that we cannot develop such a theory if we
remain wholly within the constraints of orthodox historical linguistics.
In section 2.7, therefore, I shall conclude this chapter by commenting
on a state of affairs that underlies some of the matters raised in this
introduction, that is, the limited nature of the database of historical
linguistics. By discussing this we can focus on the methodological
interface between historical and social linguistics and go on in later
chapters to suggest what would be involved in a socially-based theory
of language change.

2.7 Limitations of Historical Inquiry

It is obvious that data preserved from the past are likely to be more
limited in certain ways than data collected at the present day. Here we
notice two major limitations. The first is that past states of language
are attested in writing, rather than in speech. This has many
consequences, of which the most general ones depend on the fact that
written language tends to be message-oriented and is deprived of the
social and situational contexts in which speech events occur. ‘This is
relevant, of course, to our discussion of function, above, and to other
matters that will arise in later chapters: for example, interpreting
written texts as evidence for pronunciation.
The second limitation is that historical data have been accidentally
preserved and are therefore not equally representative of all aspects of
the language of past states. Thus, whereas research into present-day
states proceeds in a controlled way by collecting and analysing data
for the specific purpose of drawing generalizations about language
and about specified aspects of language, the researcher into past
states must use materials which were not in the first place collected
for this purpose. Some styles and varieties may therefore be over-
represented in the data, while others are under-represented. For
some periods of time there may be a great deal of surviving
information: for other periods there may be very little or none at all. It
is reasonable to say that the database of historical linguistics, as
compared with that of sociolinguistics, is impoverished.
46 Social and Historical Linguistics

To the extent that historical linguistics is subject to these limita-

tions, it is what Diaconis (1985) has called an ‘uncomfortable’
science. In this respect it is similar to some aspects of other sciences
such as geophysics, macro-economics or astronomy, in which the
scientist has relatively little control over the database. The astron-
omer, for example, does not have experimental control over the visits
of Halley’s Comet: thus, just as the astronomer does not have control
over space, so the historical linguist does not have control over time;
to be more specific, historical linguistics does not have experimental
control of its database, and so it cannot always isolate the variables
that may be involved in an explanation. It is quite appropriate here to
mention the analogy of the blind men and the elephant, which has so
often been mentioned before, or to use metaphors of the ‘tip-of-the-
iceberg’ type, because it is very much a matter of proceeding from a
base of very limited knowledge. Thus, whereas social dialectology can
plausibly claim to be to some extent an experimental science (because
it is possible to control some variables in the frame of the investiga-
tion), historical linguistics cannot. Sometimes the data may be so
impoverished that decisions cannot be made as to what is the best
description amongst a set of possible descriptions, or what is the best
explanation amongst a set of possible explanations. The result is that
interpretations of the surviving evidence are often strongly dependent
on current theoretical assumptions (which may of course be dubious) ~
and, more widely, on current ideological positions (which are even
more dubious). Many examples of this difficulty can be cited:!° here I
shall refer only briefly to a rather general one.
One example of a difficult area for the historical investigator is the
chronology of sound-changes in the history of a language: this can
often be uncertain and controversial. As a result of the limitations we
have noticed, and the imposition of certain theoretical orthodoxies on
what is or is not possible in sound-change, there has been a strong
tendency in historical descriptions to assign a date to a sound-change
at what seems to be the time of its completion, and (until recently) for
relatively little interest to be shown in the earlier stages of change.
Indeed, sometimes it seems as if change in a whole phoneme class is
believed to have taken place all at once with simultaneous actuation
and completion — perhaps overnight at some date in the early seven-
teenth century. However, if we are to understand the nature of
change, we want to know as much as possible about its actuation,
implementation and diffusion; therefore, we want to explore the early
 Social and Historical Linguistics 4]

stages if possible (I shall discuss this further in later chapters). But the
limitations of historical databases often make this difficult. Because of
these limitations, therefore, it seems that our understanding of the
nature of linguistic change will ultimately depend, not mainly on
historical data, but (recalling Principle 1) on our ability to observe it
systematically at the present day in social contexts of use, because that
is where we can most readily locate change in progress in a specifiable
social context.
This in turn makes it possible to project backwards. By using the
insights we are able to derive from the much richer data of
present-day researches we should be able to understand more fully
what happened in the history of the language, and I shall return to this
in later chapters. However, the main point here is that in order to
observe in a detailed way the contexts in which linguistic change takes
place we need to focus on present-day data. Accordingly, in the
following chapters, my aim is to build up a theoretical approach to the
social origins of linguistic change by focusing on present-day data. In
chapters 3 and 4, we focus on analysing and interpreting patterns of
variation in the speech community.
 3
Analysing Language in the
Community: General Principles

3.0 Introduction

The purpose of the next two chapters is to consider how far the }r
practical analysis of language use in live speech communities can }
contribute to our understanding of language change. I am concerned |):
largely with laying a basis for this by first determining the general }
patterns of language variation and social variation in which changes in
progress may then be discovered —in other words I am concerned ~
with the embedding of language variation in society. My concern in }
the present chapter is mainly with /inguistic patterns, which amounts
to a description of the main methods and principles of social
dialectology as we have tried to develop them, especially as they
concern the analysis of what are usually called ‘non-standard’ |)
varieties. In social dialectology we need to abstract from speaker-
interaction and the social contexts of speech events as we observe |
them, to project the data on to the configurations of language systems
(recall Principle 1). Analysis of purely linguistic variation is, however, 7
an essential part of a study of language change, because language
change is manifested as change in systems; therefore, in this chapter,
I am mainly concerned with the principles of observing and analysing
language variation in the community. In chapter 4 we shall consider
the sociolinguistic interpretation of these patterns.
As the language system is the focus of our study here, we start with
the idea of the linguistic variable (for example the vowel /a/ in such
items as cat, bad, have), and our first task is to discover what the
 Analysing Language in the Community 49

possible variants of a variable may be: subsequently, the quantitative

distribution of linguistic variants may be demonstrated by reference
to the familiar speaker-variables of social dialectology, such as age and
sex of speaker. But the important point to remember is that we do not
observe these sociolinguistic patterns directly: it is the speech of
individuals in conversational settings that we observe and describe,
and it is by analysing a large quantity of spoken language from many
speakers that we can then demonstrate the patterns that emerge from
our data. I shall be discussing the general principles of this kind of
analysis, and I shall start with some remarks about the general context
within which the Belfast research (1975-82) was designed — the
context of historical linguistic description.
Here, we need to recall the very broad distinction that has
been mentioned in chapter 2, between historical linguistic theory
(and the theory of change) on the one hand, and historical linguistic
description, on the other. Of course this is not an absolute distinction,
as historical descriptions depend on some kind of theoretical orienta-
tion, and theory depends on systematic observation and description of
linguistic forms (otherwise there is nothing to have a theory about).
For practical purposes, however, we can accept this broad distinction
here. But it is clear that the data collected in close investigations of
live speech communities are much richer than the data preserved
from early language states, and they are observable in a larger number
of dimensions and at a much finer level of detail; thus, the patterns
revealed in systematic investigations of live communities appear to the
observer as much more variable and multidimensional than historical
patterns (as these are usually reported). That is to say, the variation is
not necessarily patterned in one single linguistic dimension (for
example, it does not necessarily move in a single phonetic direction: it
may diverge in two or more directions), nor does it necessarily display
a unilinear or unidirectional pattern in terms of any independent
‘social’ variable: on the contrary, the patterns shown in relation to
different social variables may conflict and interact in a variety of ways.
In the Belfast research programme one of our aims was to use the
complex patterns discovered in a live speech community to throw
light on the kind of movements that might have taken place in
linguistic change in the past.
In conventional historical description, three very prominent inter-
related tendencies had been noticed. These are, first (as I have
mentioned in chapter 1), a tendency to focus on patterns of change
50 Analysing Language in the Community

alone with little or no attention to stable patterns of language through |:

time; second, a tendency to unidimensionality, that is, an inclination
to think of the history of a language as the history of a single
homogeneous variety and of sound-changes as proceeding in straight |
lines; and third, as noted in chapter 2, a tendency to impose
theoretical and ideological orthodoxies on (sometimes rather sparse)
data that might often be open to alternative kinds of interpretation. Of —
these orthodoxies, the ideological ones are perhaps the easiest to
explain here, and they are relevant to what I have to say later, so let us
consider here a very general point about the effects of ideology — the
apparent contrast between typical models of ancient language states, |?
on the one hand, and recent language states on the other.
It is noticeable that for the distant past of language orthodox
models of change, such as the Indo-European family-tree model, are }
mainly models of divergence (in these models languages are envisaged, |)
like galaxies, as moving away from one another at considerable }
speed), whereas for recent centuries models of language history are
predominantly convergent. Specifically, the history of English since
about 1550 is often presented as what Lass (1976: xi) has called a
‘single-minded march’ towards RP and modern standard English,
with divergent developments either excluded or admitted only in so
far as they throw light on ‘standard’ English. This is in clear contrast
to the divergence model in Indo-European studies, and as a result of
this contrast, the shape that emerges from historical language |
description (from ancient times to the present day) is not so mucha }|
pyramidal shape (with gradual convergence at the top) as a funnel
shape (the kind that is used for pouring liquids), as in figure 3.1.
At about the year 1550 the pyramidal base of the funnel suddenly |)
narrows and from that time proceeds in a straight and narrow path to
the present day. That is to say, scholars have assumed that around >
1550 the English language became much more convergent than it had
been before even though there can be no direct evidence that spoken
English did become more convergent. Following this type of model, |
conventional histories of English have customarily given considerable ¥
attention to phonological and morphological change and diversity in
Indo-European, Germanic and Early English. At the Middle English
stage, the description of divergence is still very salient (partly because |
the states attested in writing are unquestionably divergent states), but |
we also begin to notice attempts to launder the data retrospectively in |
such a way as to focus on those features that lead to modern |
 Analysing Language in the Community 51

Present
English

European

Figure 3.1 The divergence/convergence model, from Indo-European to

the present day.

‘standard’ English and to ignore, reject or explain away those features

that deviate from it. Examples are noted below and in chapter 5.
Then, at around 1550, the story becomes exclusively about standard
English. A model like this cannot be a sufficient basis for a convincing
explanation of a recurrent phenomenon like linguistic change because
both divergence and convergence can be present in language states at
any time, or, perhaps, at all times. But what we have to notice here is
that it cannot be right for mere historical description either: as there is
no such thing as a uniform language or dialect (and standardization
implies uniformity), and as sound-changes do not proceed in straight
lines, this cannot possibly be an adequate conceptualization of
_ English phonological history.
There are traditional influences at work here, and they are
ideologically loaded. H. C. Wyld, for example, was a very great expert
on the grossly divergent regional dialects of Middle English, but
when it came to post-sixteenth-century English, he was quite insis-
tent that the only object worthy of our study was Received Standard
English. Such views have an ideology behind them (though not
necessarily a conscious one), involving as they do a decision that the
language of ‘the Oxford Common Room and the Officers’ mess’ is an
appropriate object of study, whereas that of ‘illiterate peasants’ is not
(Wyld, 1927). From the point of view of standardization theory, this
D2 Analysing Language in the Community

backward projection of Wyld’s ‘Received Standard’ on to earlier

states can be seen as an attempt to historicize the standard (literary)
language — to create a past for it and determine a canon, in which
canonical (‘genuine’) forms are established and from which unortho-
dox (‘non-genuine’ or ‘corrupt’) forms are rejected. This retrospect-
ive model has been influential and pervasive in historical description:
since Wyld wrote there has been further important work on Early
Modern Standard English and its history, notably that of Dobson
(1955, 1968). Naturally, the ideological assumptions behind this
unilinear model have led to dismissive attitudes to ‘non-standard’
dialects. In his account of EModE pronunciation, Dobson (1968:
551) dismisses ‘vulgar’ and ‘dialectal’ sources, and I shall comment
further on this tendency in chapter 5. Even so, it’s surprising to find
the following comment in the work of a distinguished historical
linguist: ‘Nonstandard forms may be found among antisocial groups,
such as criminals, or a rebellious younger generation, or among
rustics’ (Lehmann, 1962: 142). This has the effect of marginalizing
non-standard vernaculars — appearing to present them as abnormal
or pathological language states - when the majority of human beings
throughout history must have used varieties that were, to a greater or
lesser extent, non-standard. Thus, the imposition of these precon-
ceptions on the object of study necessarily leads to a distorted view of
the history of a language. As a language is variable at all times, the
many different varieties can each be seen as having continuous
histories, with influences passing to and fro between them, as
represented in figure 3.2. If our ideological preconceptions incline us
to an exclusive interest in standard English, we will produce what is in
effect a history of literary English; this will exclude and neglect other
historical patterns that are capable of enriching our description of the
history of spoken English and, ultimately, of adding to our un-
derstanding of the general phenomenon of linguistic change.
Clearly, the difference here between traditional approaches (as
exemplified by Wyld, Dobson and others) and _ sociolinguistic
approaches is that, whereas the former project a uniform state
(standard English) on to the past, the latter project linguistic variabil-
ity on to the past. This backward projection is one of the things that
has motivated my own work on present-day speech communities, and
my interest in how the authority of the legitimized variety is promoted
in linguistic scholarship (J. Milroy and L. Milroy, 1985a). I shall have
more to say about it in later chapters.
 Analysing Language in the Community 53

Present English

Early English
Variety Variety Variety
A B (c

Figure 3.2 Simplified model of multidimensional change in the history of

English. (The dotted lines represent influences passing from one variety to
another.)

3.1 Shapes and Patterns

My purpose in the remainder of this chapter is to set out the main

principles of exploring multidimensional language states in present-
day communities, on the assumption that historical states of language
must also have been multidimensional. A major contribution of
Weinreich, Labov and Herzog (1968) and the quantitative paradigm
has been to demonstrate that heterogeneity in the speech community
can be shown to be have patterns in it. But we noticed in chapter | that
linguists have often lived by metaphors — the biological metaphor in
the nineteenth century and the machine metaphor today. If we speak
of patterns and shapes in speech communities, we are again talking in
a
metaphors, but this spatial metaphor is one that I have found
congenial in speaking of linguistic history and the complexities of
usage in live speech communities. The phonological system, for
example, can be described as having a ‘shape’, but there is also a
sociolinguistic shape of the speech community, involving society as well
as language. If we can describe this (and the Brown and Gilman
(1960) ‘power and solidarity’ model is suggestive as a beginning), we
can use it to help to model the embedding of language in the
community.
oa Analysing Language in the Community

The perception of shape and pattern in apparently disorderly (but

dynamic and mobile) things is usually mentioned with reference to
visual perception, and it is commented on in the sciences and the arts
alike: it is prominent, for example, in the notebooks of the poet
Gerard Manley Hopkins (J. Milroy, 1977), in his careful descriptions
of cloud formations, waterfalls and other dynamic phenomena, and
much of the poet’s imagery depends on a kind of ‘observer’s paradox’
(rather different from the familiar Labov version), through which a
dynamic phenomenon can nonetheless appear to have stable shapes
and patterns within it and, conversely, a static phenomenon may
appear to contain mobility. But it is also apparent in the perceptions
of some scientists, especially in subjects that are, like historical
linguistics, ‘uncomfortable’ sciences — meteorology and some aspects
of astronomy, for example (see Diaconis, 1985; Gleick, 1988). At the
most general level, therefore, it is convenient to think of the speech
community as having a ‘shape’ and of language in the community as
being capable of displaying patterns, much as we might think of these
other dynamic phenomena as displaying shapes and patterns. To the
extent that language is perpetually shifting and changing, it plainly
resembles dynamic states such as weather and cloud formations more
closely than it resembles static objects. In the case of language,
however, the patterns are not immediately observable in the way that,
say, cloud formations are: as we have noted above, they are observed
in the configuration of the data after the analysis has been carried out.
But as there must always be some kind of observer-effect in patterns
revealed in this analytic way, the position of the observer, and the
preconceptions that he or she brings to the act of observing, must be
accounted for at every phase of the research.
Observation of historical states is affected by the limitations of the
database, and we can liken this to a situation in which the object is
viewed from a distance. When this is the case, some patterns will be
obscured or invisible, and many details in the patterning will be
unobservable. The present-day sociolinguist, however, who is ob-
serving and exploring a ‘new’ language situation is in a close-up
position; he or she can observe the phenomena at successively finer
levels of detail and thus reveal fine-grained patterns that are not
accessible in historically attested states. But first let us stand back a
little and, starting at the surface of the phenomenon, let us consider
the most general ‘shapes’ and patterns that speech communities may
exhibit.
 Analysing Language in the Community JD)

At this superficial level, different kinds of speech community shape

can be readily distinguished. For example, some states that can be
easily identified are bilingual or multilingual, others can be viewed as
bidialectal, and yet others as monodialectal (but still variable).
Although the matter is much more complicated than this, with many
subtle gradations within and between these different shapes, this
rough classification is sufficient for our present purpose. The Belfast
community is a broadly monodialectal one, but within this it can be
described as a divergent-dialect community (Johnston, 1983). It is this
typological point that I would like to consider now, bearing in mind
the question of historical divergence and convergence that I referred
to above.
From a static point of view, the term ‘divergent-dialect situation’
can be understood in two senses: first, Belfast dialect (for example) is
observed to be divergent from other dialects and, particularly, from
‘mainstream’ norms of language, such as RP and standard English;
second, the dialect exhibits a great deal of internal variation — much
greater than is generally reported for these better-known varieties
(this is approximately the phenomenon that Le Page (1975) calls
diffuseness). Thus, it is not a uniform state (as standard English
appears to be in handbook descriptions), but a highly variable state,
with many complex patterns observable within it. As compared with
handbook accounts of supposedly homogeneous varieties, therefore,
it can be characterized as a ‘normal’ situation, as it is the normal state
of language to be heterogeneous. But what is important here is that,
just as present-day language states are normally heterogeneous, so
historical language states must also have been heterogeneous in
similar ways; hence, unilinear historical descriptions of single varie-
ties (such as ‘standard’ English) cannot be adequate descriptions of
the history of a language. In section 3.2, we shall briefly consider
these two aspects of divergence: external and internal.

3.2 Describing Divergent Language States

As we do not have intuitive knowledge in detail about states of

language that are divergent from us in time, space or social space, we
cannot reliably project our own linguistic intuitions on to them. In
order to describe them, therefore, we may be inclined to use as a
56 Analysing Language in the Community

reference point the linguistic norms of some well-described dialect

such as ‘standard English’. But this is not acceptable either, because
there will then be a danger of projecting the norms of standard or
‘mainstream’ varieties on to the dialect (imposing them from above, as
it were), and this will result in a distorted account of patterns
observed within the community, which may bear no direct relation to
these other varieties. We cannot assume that a divergent phonological
system, for example, is structurally similar to or derivative from RP,
or that lexical items belong to the same phonemic sets, or that the
tense/aspect system is structured in the same way as that of standard
English. In Belfast, for example, all these things are quite different.
Nor are we likely to know beforehand the sociolinguistic functions of all
the linguistic variants identified - whether they mark age, sex or
contextual style differences, for example. Thus, the exploration of a
present-day divergent state of language resembles the exploration of a
historically attested state to the extent that we cannot successfully
describe either of these in terms of an external (and usually superor-
dinate) variety. That is to say, you cannot describe sixteenth-century
English as if it were a variant of RP any more than you can do this for
divergent present-day varieties.
Let us consider briefly what difference all this makes to our
analysis and interpretation. Specifically, what interpretations of data
would be available to us if we relied on ‘mainstream’ norms, and how
would these differ from internally-based interpretations? Here, I shall
look at some examples of internal linguistic variation in Belfast in
(roughly) ascending order of generality in order to show how little we
can rely on outside norms in preparing a sociolinguistic analysis. First,
a particular instance: one of the many cases of ‘hypercorrection’ that
were noticed.
In exploratory recordings (1975), the word queue occurred several
times pronounced as [ku:] (without the [j] glide), by a middle-aged
female in a careful style. There is no externally-based explanation for
this, as there is no sign of [j]-deletion as a known process in the
history of Ulster dialect, and it is not favoured by RP, which on the
contrary replaces the [j| of [j]-deleting dialects (Trudgill, 1974; Wells,
1982). Nor is it found in the more ‘standardized’ forms of Ulster
English. To understand what is happening here, we have to know that
Ulster vernaculars have palatalization of initial [k] in certain prevoca-
lic positions (as in [kjat, kja:r] ‘cat, car’), and furthermore that this is
recessive in inner-city Belfast. Figure 3.3 shows the recessive pattern:
 Analysing Language in the Community 57
100%

Clonard Hammer Ballymacarrett

Figure 3.3 — Recession of palatization in middle-aged males in car, cat, etc.

(Clonard: 62%; Hammer: 14%; Ballymacarrett: 0% palatization.

palatalization is hardly present at all except in males over 40, and the
area which has the most recent immigration from the country is the
most conservative.
From this internal knowledge of Belfast English we can conclude
that the speaker who says [ku:] for queue is using a careful style
strategy to avoid the palatalized segment (see my comments on
avoidance of stigma in chapter 2.6, above), which is a strongly
regional marker and also a marker of older male speech. Therefore,
we can say that the speaker’s ‘knowledge’ includes the knowledge that
/k/ is palatalized in the dialect. But we could not draw these
conclusions if we relied on external norms, on general historical
information as codified in historical textbooks, or on our own
‘intuition’. We must, therefore, use such instances empirically to help
to establish what the internal linguistic norms of the community
actually are, and to do this is to be accountable to the data in quite a
strong sense.' In fact, we made considerable use of hypercorrection
phenomena of this kind in preparing our phonological analysis.
A second example concerns the pronunciation of /a/ before velar
-consonants in words of the type rag, pack, bang. In Belfast vernacular
(henceforth BV) the vowel is front-raised as far as [e] in this
environment. However, there is also a substantial incidence of a low
vowel with a prominent closing glide: approximately [a1]. This has the
effect that such items as back can sound very similar to RP dike, and
the outsider may easily miscomprehend ‘I was on my back’ as ‘I was on
my dike’. It is clear that the ‘underlying’ norm: ([e€]), is not the RP
norm, even though the movement towards [ai] may be modelled on
58 Analysing Language in the Community

some external norm. However, as it results in near-homophony with

RP bike, it seems that the constraints of the local system are much
more powerful than any superordinate norm: whereas back can be
[batk], bike is not normally [bark], but [berk]: there is a partial
avoidance of homophony. Clearly, the explanation for such pheno-
mena must be based on an analysis of the local vowel system, not that
of RP.
My next examples are more generalized, and they concern the
imposition of external RP stereotypes on the vowel phoneme /a/. One
common local belief about Belfast English is that upper-middle-class
people tend to front-raise /a/ (as in bat) towards the conservative RP
value: [z] (but we found little sign of this in any part of our research).
However, inner-city speakers normally, and quite consistently, have
{z] and [e] in items where the vowel precedes a velar consonant (as
noted above), and I have sometimes been told that this is affected by
RP. Our quantified investigations (for example, J. Milroy, 1984),
however, have repeatedly demonstrated that the only environments
that show a consistent front-raising pattern are these velar environ-
ments; thus, while it affects the whole phoneme /z/ in conservative
RP, it is confined to velar environments for most Belfast speakers.
Clearly, if it were based on RP, there would be no reason why it
should affect this environment and not other environments, and these
internal facts about BV are quite sufficient to establish independence
from RP. And this independence is further supported by the
existence of the ‘velar-raising’ rules in other Ulster varieties: they are
carefully described by Gregg (1964) for Larne, and by Patterson
(1860) for nineteenth-century Belfast, and so in this case they are
plainly of some antiquity.”
Also within the /a/ system, however, there is a high incidence of
back varieties of /a/: in BV, this vowel is realized in certain following
consonantal environments as a long, back vowel, which is in many
contexts virtually identical to RP (as in RP dance, bath); on a
superficial view we may therefore be tempted to believe that speakers
are adopting this external norm. The consonant environments that
encourage backing of /a/ are: (1) following voiceless fricatives (as in
grass, path); (2) following voiced obstruents generally (as in has, bad),
(3) nasals (as in Sam, man); and (4) /r/ (as in car); but backing does
not occur before voiceless stops (as in bat), or nasals followed by a
voiceless obstruent (as in dance). Of these environments, however,
only the voiceless fricative and /r/ environments coincide with the
 Analysing Language in the Community 59

backing environments of RP, so we are clearly dealing with a different

/a/-backing rule. The difference from RP can be appreciated quite
easily by reference to the pair can, can’t: can has the long back vowel
predicted by the single nasal environment, whereas can’t has the short
front vowel predicted by the nasal + voiceless obstruent environment
(compare also dance, ranch, etc., which are short, front). This is the
opposite of RP (where can is front and can’t is back), so we plainly
cannot explain this back /a/ pattern by relating it directly to RP.
There are many other examples that show the RP (and general
southern British) rules as ‘flip-flopped’ in certain environments in
Belfast: they are of historical interest, and we shall return to them in
chapter 5. What is important here, however, is the fact that a
description of these divergent configurations in the phonological
system is an essential basis for subsequent quantitative analysis: the
further away the variety is from mainstream norms, the more
important this prior description becomes.
The above examples were noted prior to quantitative analysis, and
were used, with other observations, as the basis of an internal
phonological description on which all our subsequent work has been
based. However, the general independence of the /a/ system from
mainstream influence is supported even more conclusively by the
broader community norms as they emerge from later quantitative
analysis. We have demonstrated in various publications (for example,
J. Milroy and L. Milroy, 1978) that in the inner city there is a change
in progress moving from front values of /a/ towards back values,
beginning to affect even those environments that were formerly
always front (for example pre-voiceless stop environments, as in map,
that). However, as we move out of the inner city the trend to backing
is reversed, and the system swings back again to front realizations (J.
Milroy, 1982b): it converges on front-vowel realizations to the extent
that all items (except those with following /r/) including items such as
grass, path (which are back-vowel items in both RP and inner-city
vernacular) have the front vowel. Thus, if we project this information
on to the socio-economic class dimension, the movements of /a/
show a zig-zag (or a split-level) pattern from front to back and then
from back to front, as in figure 3.4. RP thus has no direct effect on the
convergence pattern displayed by middle-class speakers: therefore, if
we start with the RP norm and derive these speech community
patterns from it, we can falsify the situation totally. In this case the
‘main falsification would be that hundreds of thousands of inner-city
60 Analysing Language in the Community

High
social
groups

Lower
social
groups
Front [a] Back [a]

Figure 3.4 The zig-zag pattern in the community.

speakers who are not remotely interested in RP have ‘borrowed’ back

{a] from RP. It might also appear from such an analysis that
working-class people are converging towards RP back [a] in grass,
path, and middle-class people diverging from it, when in fact there is
no evidence that the usage of these speakers is influenced by RP in
any way at all.
To this extent, of course, a synchronically divergent language state
requires the same kind of treatment as a historically divergent state: in
neither case can it be assumed that the norms of some ‘standard’
variety can be successfully projected on to it. In section 3.3 we turn to
the principles of exploring linguistic variation in present-day vernacu-
lars.
(

3.3. Exploring Language in the Speech Community: Some

General Principles

It is clear from the above examples (and from many others) that we
have to describe the speech community in depth and in its own terms.
That is to say, our conclusions about phonological structure must be
accountable to, and ‘warranted’ by, the data. If we do not do this, we
shall not find out very much about the social functions of variation in
the community or about the sources and motivations of change, and
our results will therefore be superficial. They may also be misleading.
For example, if we observe only public speakers in Belfast (news-
casters, reporters and the like), we will notice that some of them adopt
 Analysing Language in the Community 61

quasi-RP forms in their public styles. In the inner- and outer- city
work in Belfast, however, these forms (for example, fronting, un-
rounding and diphthongization of /o/ to /au/, as in home, stone) did
not occur at all in hundreds of hours of recording. Therefore, if we
had taken our cue from these public styles, we would have produced a
top-heavy account that would have been of only marginal importance
to the social ‘life’ of the speech community. From this perspective,
the success of Labov’s work in New York City, which is based on
more widely-known variables, suggests that New York City speech is
much less divergent from mainstream American English than Belfast
is from mainstream British English. To that extent, therefore, we are
dealing here with a different kind of analytic task. It is also relevant to
notice that Horvath and Sankoff (1987) in their Australian work
found the social variables the most difficult to deal with and the
linguistic variables ‘well defined’. In these dialect-divergent studies
(J. Milroy, 1981; Johnston, 1983; Newbrook, 1986), however, it is the
linguistic analysis that is the greatest immediate challenge facing the
investigator: if we do not investigate the internal linguistic structure of
the speech community itself, the sociolinguistic interpretation of the
data will be at best superficial, and, at worst, wrong.
In addition to this external divergence, however, communities like
Belfast also exhibit massive internal linguistic divergence, and we
must attempt to find out what are the internal social functions of this
divergence. In New York City Labov was able to show important
patterns of variation on the basis of only five linguistic variables; in
contrast to this, Pellowe et al. (1972) identified in Tyneside (a
divergent dialect community) no less than 303 linguistic variables that
they considered worthy of study. In Belfast we assumed that the
number of socially-patterned variables that we might uncover could
well run into the hundreds. If we adopt here Le Page’s distinction
between focusing and diffuseness in language states, a community like
Belfast is relatively diffuse (divergent within itself) in comparison with
‘more homogeneous communities, but relatively focused when com-
pared with communities in which there is extensive code-switching
and code-mixing (such as the Caribbean communities studied by Le
Page and Tabouret-Keller, 1985). But to suggest that a speech
community is diffuse (or internally divergent) is not to imply that it is
necessarily unpatterned or unstructured: on the contrary, our task is
to find out how the variation is structured by demonstrating what the
patterns in the community are like. However, our descriptive task will
62 Analysing Language in the Community

be more difficult, as it involves dealing with greater complexity of

variation than would be the case in more focused communities, and it
has to be carried out at a very fine-grained level. There are many
different ways in which speakers may exploit, in varying social
functions, the resources of variation that are available to them, and we
do not necessarily know beforehand what these are, how they interact,
or what the limitations on possible variants may be. However, as it is
the ‘normal’ state of language to exhibit structured variation, I have
suggested above that speech communities like Belfast can be re-
garded as ‘normal’ speech communities.
I mention this here because it has sometimes been thought that this
type of community is abnormal or atypical. Romaine (1982b), for
example, suggests that these may be atypical communities and,
further, that such communities may exhibit two ‘norms’ within them
(we shall return below to the question of internal norms). Similarly,
Johnston (1983) describes the patterns discovered in his Edinburgh
work as ‘irregular’. But the Belfast and Edinburgh situations are
abnormal and irregular only if it is assumed that the patterns revealed
in the New York City study are the normal and regular ones, and that
Labov’s findings on the structure of that speech community are
universally applicable. There is of course no reason why the patterns
revealed in one pioneering study should be taken as definitive and
exhaustive for all communities: it is an empirical matter to determine
what kinds of pattern can be revealed in speech communities and
hence to determine what the norms of particular speech communities
may be. And this, of course, is in accordance with Principle 2 b
suggested in chapter 1.
The guidelines that we rely on here are parallel to those of
‘exploratory’ science, as described by Tukey (1977), Diaconis (1985)
and others, and also to those of ethnographers who investigate the
internal structures of non-mainstream societies. The papers in
Cohen (1982), for example, describe research into geographically
peripheral communities in the British Isles, such as Shetland, Tory
Island and the Isle of Lewis. Their aim is to explore and describe the
internal social systems that govern behaviour in these communities,
and of course they gain access to these systems through studying
overt behaviour in the communities. Being aware of the (external)
divergence between these cultures and ‘mainstream’ culture, they are
anxious to avoid imposing what they call ‘crude mainstream stereo-
types’ on to the interpretation of these internal systems: they want to
 Analysing Language in the Community 63

get access to what the patterns of behaviour mean internally and how
these are perceived by in-group members. As the investigators are
themselves out-group members, they cannot have reliable prior
intuitions as to the social meanings conveyed and must investigate
these by observation and analysis. This is very much the position of
the observer of “divergent” speech communities: the difference is that
the linguist’s ultimate goal is to draw generalizations, not about
behaviour in general, but about language behaviour.
I have implied above that a major difference between more diffuse
and more focused communities is relative complexity of variable
linguistic structure: Cohen (1982) is very clear on the question of
internal complexity. The ‘voice’ to the in-group is, according to him,
more complex in structure than the ‘voice’ to the outside world, and
the internal meanings and functions of variation in behaviour are not
easily accessible to outsiders. This last point is a familiar one, which
fits in well with the remarks on observing patterns that I have
discussed above. In the Belfast research, we wished, like Cohen, to
approach an explanation for ‘the complex differentiation within’
communities, by exploring and analysing in-group patterns and
functions of differentiation; in the event, we showed that a great deal
of variation has in-group functions that are not readily accessible to
outsiders, and we shall review this in chapter 4.
In the analysis of linguistic variation, therefore, we were looking for
fine details of variation — variation that shows internal patterning
within the speech community, but which may have no social meaning
for outsiders — and in the first application for funding it was specified
that the research was intended to extend the quantitative methodo-
logy to a type of community that had not been studied in this way
before. It was of course necessary to introduce modifications in
method and procedure, which have sometimes been (wrongly)
thought to be in opposition to Labov’s work. These modifications are
required because of the similarity of our task to the ethnographic
‘tasks described in Cohen (1982). As these anthropologists chose to
investigate geographically peripheral communities, so we chose in the
first instance to investigate communities that are marginal in a
different sense. As hundreds of thousands of speakers use Belfast
vernacular, it is hardly marginal as a linguistic variety; but in the
tradition of linguistic inquiry it seems to be marginal. Linguists have
taken little notice of such communities in the past — their language is
not part of the central ‘canon’ —and so such vernaculars are still
64 Analysing Language in the Community

largely peripheral to the general body of linguistic knowledge. We

chose to investigate in the first place the speech of those who are not
very likely to be directly affected by mainstream norms, and whose
speech is of the type that had been least explored by descriptive
linguists — the urban working class. The intention was to relate the
linguistic variation revealed to social patterns within the communities
themselves, avoiding prior assumptions about social norms that might
seem to be imposed from outside. In this work we give full weight to
the speaker/system distinction that has been introduced above, and
hence to the general principle that speakers are ultimately responsible,
not only for introducing and adopting linguistic changes, but also for
maintaining diversity in language states. In section 3.4, I move on to
an outline of the methods of the Belfast research.

3.4 The Belfast Research Programme: Principles and

Methods

The empirical research in Belfast was carried out in two stages. The
first stage was the inner-city project, “Speech community and
language variety in Belfast’ (October 1975 — June 1977). The second
stage (1979-82) consisted of two major projects: the outer-city
speech community studies and a random-sample ‘doorstep’ survey.
There were also a number of minor studies in connection with the
main research programme.
As we have noticed above, the success of any social dialectological
project in a dialect-divergent community depends crucially on ade-
quate methods of analysing /inguistic variation, and it is this analysis
that takes most of the time. For methodological reasons, it is
important to distinguish between this and the socially based interpre-
tation of the results, because if the social factors involved are seen as
the motivating principle for the selection and analysis of linguistic
variables, the investigators may be accused of biased selectivity and
circularity. Within the inner-city study, we believed that the language
might show patterns of differentiation between older and younger
speakers and between the sexes, but we did not know that it would, or
if it did, what the precise differences would be. No one had
investigated inner-city language in this way before, so for all we knew
there might have been no significant differences at all, and we were
 Analysing Language in the Community 65

told by at least one colleague — not very encouragingly — that ‘Broad

Belfast’ dialect was all very much the same. But it is important to
affirm the independence in principle of the linguistic and social
analysis at this stage of the work, especially as this does not always
seem to be recognized. Wardhaugh’s (1986: 175-7) reference to the
Belfast research, for example, appears to suggest that the whole thing
was undertaken as an investigation of social network and the linguistic
variables selected specifically for the purpose of demonstrating
network patterns quantitatively.’ If this had been so, however, it
would have been a weakness: the strength of our arguments on social
network (and other social aspects of the research) depends precisely
on the fact that these socially-based criteria are not the motivating
factors for our selection of linguistic variables.
The research consists of different methodological phases, which
we can describe as collection, classification, analysis and interpretation. In
practice they overlap: for example, while the inner-city data was still
being collected, we were analysing the tape-recordings that had
already been obtained. It is, however, useful to think of these phases
as separate in principle. The analytic and interpretative phases are the
most relevant to the subject-matter of this book. In this chapter, I am
mainly concerned with analytic methods, and in succeeding chapters
with interpretation; I shall confine my remarks about collection and
classification to those points that affect the interpretation of the
findings.

3.5 Collection and Classification of Data

The type of sampling used in the five community studies is quota-

sampling, and our main concern is not to claim absolute representa-
tiveness for the whole city, but to guard ourselves against the
accusation that our informants might be hand-picked from amongst
friends and neighbours—or, worse, from our students. In this
respect, therefore, our work conforms to the general principles of
Labov’s work and is not comparable with studies (experimental or
otherwise) that are based on the language of, for example, RP-
speaking or near RP-speaking persons from amongst the analyst’s
own university students.
66 Analysing Language in the Community

For these reasons, access to the communities was obtained through

suitable intermediaries, who were not used as informants. Thereafter,
the fieldworkers were passed on from person to person within the
communities; thus, the informant groups were self-recruited in that
the speakers were not known to the investigators beforehand. In J.
Milroy and L. Milroy (1977), we justified this method, somewhat
retrospectively, in terms of the social network model (the fieldworker
can be described as a second-order network contact), and it is useful
methodologically to think of it in this way (indeed, many other
investigators have successfully used the idea of social network as an
explicit part of their fieldwork strategy). But the essential point here,
whether or not we describe it in terms of the network model, is that
we are concerned not with a superficial survey of the communities,
but with obtaining depth of coverage. In the perspective of this book,
the contrast with traditional dialectological method, and with data-
gathering in orthodox historical linguistics, is clear and requires no
further comment.
As for the fieldworker’s position vis-d-vis informants, our concern
is to account for this (this can be seen as an extension of the notion of
accountability to the data into the fieldwork phase) and to use it as part
of our method of data classification. We do not use Labov’s concept
of the vernacular— the language used by speakers when they are not
being observed (which is described by Labov as the most ‘regular’
style) -—and this is in accordance with Principle 1, suggested in
chapter 1: language in use is always observed within a social context
of some kind. The immediate reason for not using Labov’s concept is
that the vernacular must be an idealization-on a par with other
idealizations such as dialect or speech community —and so as an
idealization it must be inaccessible in practice (recall that you do not
observe the language: you observe people talking). But if by any
chance it were ever possible to locate this ‘vernacular’ in the usage of
an informant, we would have no criteria for demonstrating that we
had located it, any more than we could locate ‘the dialect’ in the
speech of a single informant. We have used the term ‘vernacular’ in a
different way: for us it is a ‘primitive’ term roughly synonymous with
‘real language in use’, and it is interpreted on a continuum of relative
closeness to, or distance from, the idealized norm, or (in some cases)
the idealized standard language. | shall have much more to say below
about vernaculars, norms and standards, and that is one reason why I
have to comment here.
 Analysing Language in the Community 67

Otherwise, the fieldwork strategy in large-scale quantitative studies

must be broadly the same as that of Labov: we need to go as far as
possible in obtaining casual styles from informants and to develop
ways of distinguishing styles on a continuum from ‘careful’ to ‘casual’
style. But we do not claim that we have reached the ‘vernacular’, or
the most casual of possible styles, for any informant (although some
reports on our work have stated that we have): we merely claim that
our data is rich and variable enough to enable us to classify styles on
the stylistic continuum in an extremely well motivated way.
I have noted above that in both social and historical linguistics the
position of the observer is crucial and is relevant to the interpretative
phase in that the observer may affect the data in some way. It is
consistent with this that we should use the role of the fieldworker in
speech-exchanges as a means of classifying different speech-styles.
We explained this during the course of the research (L. Milroy andJ.
Milroy, 1977). Briefly, when the fieldworker is actually interviewing
the informant and thus controlling the exchange, the style is labelled
‘interview style’. It is possible to distinguish this style from others in
that the discourse turns in interview style are alternating and relatively
brief; they follow a question—answer discourse structure, and they are
often of roughly equal length. Here, the degree of control exerted by
the fieldworker over the speech-exchange is the guiding principle,
and the fieldworker’s aim is to get away from interview style (that is, to
lose control) as soon as possible. In our 1977 article, we use
transcripts from our data in which: (1) the question—answer discourse
structure is overridden; and (2) in which the fieldworker is not in
control, but is a participant/observer. The differences in discourse
structure are quite obvious (the reader is referred to L. Milroy’s
(1987) report of our field methods, where the same transcripts are
used).
Much of what I have said here is relevant to the second methodolo-
gical phase (classification), but it should also be noted that we do not
make prior classification of social levels of language in the way that
many other urban dialectologists do. Bertz (1975: 77), for example, in
his study of Dublin, uses a prior classification into ‘drei Sozio-
stilistischen Typen’: ‘educated’, ‘general’ and ‘popular’ Dublin English,
and his informants are pre-classified in these groups. We are of
course aware that educational levels may be important at some stage,
but we do not pre-classify in this way because we do not assume
beforehand that we know the sociolinguistic structure of the speech
68 Analysing Language in the Community

community: this is what we are trying to find out! Our classification is

based solely on the speaker-variables of age, sex and area, and on
contextual styles, and this is regarded as a procedure whereby we can
establish the internal linguistic norms of the community. The
classifications we use normally depend on fine-grained analysis of
language, and | move on in section 3.6 to the analytic aspects of the
work.

3.6 Analysis

Whereas for Horvath and Sankoff (as noted above) the linguistic
variables are ‘well defined’, this is not so in a dialect-divergent
community: in such a community few of the linguistic variables can be
said to be defined at all. This difference seems to be partly a
consequence of the relative homogeneity of Australian English on the
one hand, and the heterogeneity of Belfast English on the other. But
there is another reason why communities like Belfast are difficult to
deal with, and this is that the phonology of such communities is not
usually adequately described and codified beforehand, whereas more
information of this kind is available for varieties closer to ‘standard’
English. The result of all this is that because of the divergence of
Belfast English from other varieties and the internal divergence
within it, we do not know beforehand what is the correct lexical input
to any phonological variable, we do not necessarily know what the
variants of the variable are, and we may not be at all certain about
what precisely might count as a variable. Therefore, if we are to find
out what linguistic variation means to in-group members and how it
functions in the community, much more prior observation and
analysis is needed than would be required for better described
varieties.*
In order to locate the variables that might be worthy of study, and
to define the variants of the variables, a broad description of the
phonological system has to be prepared beforehand. A ‘practical
phonological’ description of Belfast was prepared in March 1975 and
updated while the research was in progress (J. Milroy, 1976a, b). The
main points arising from this are that: (1) the vowel system is totally
different from mainstream British English in terms of vowel-length,
vowel-height, diphthongization and other properties (for example,
 Analysing Language in the Community 69

vowel-length is not usually contrastive, as it is alleged to be in RP, and

so most vowel-phonemes, such as /e/, as in gate, save, are realized as
considerably longer or shorter allophones according to consonantal
environment); (2) allophones of phonemes can overlap phonetically
with allophones of other phonemes in a manner that is not permitted
by classical phoneme theory (Bloomfield, 1933); (3) lexical items do
not necessarily belong to the same vowel phoneme classes as they do
in RP and SBE (for example, whereas good and food have different
vowels in most SBE, they have the same vowel in Ulster English); and
(4) many sets of lexical items exhibit vowel alternations, in that the
vowels in these items are realizations of two different phonemes.
More detailed accounts of Ulster phonology are now available in J.
Milroy (1981) and Harris (1985), and we shall return to aspects of
phonological variation at various points in this book. Here, I want to
discuss the principles on which variables are selected for quantitative
analysis.
As there is probably a large number of variables that might be
socially interesting, it is necessary to select a representative sub-
sample of different types of variable in a well-motivated way, and to
specify the range of potential linguistic variation that lexical items in
these sets may undergo. The second point is crucial for the reason
that if we include in the set items that cannot undergo the variation in
question, or items that undergo different patterns of linguistic
variation, the quantitative results will be false. A final point that has to
be borne in mind is that in order to make generalizations based on the
type of quantitative analysis pioneered by Labov, a large number of
tokens must be analysed (usually thousands); however, it happens that
some variables that are quite salient in the community occur relatively
rarely, and so we cannot make reliable quantitative statements about
these covering the range of speaker variables, even though they may
be involved in linguistic change and may be important for historical
projections on to earlier English. It is for this reason that for the
central part of the projects we selected representative variables that
occur frequently and can therefore be quantified in terms of the full
range of speaker-variables.
The main variables for the inner-city research (most of which were
reported in J. Milroy and L. Milroy, 1978) were therefore selected on
the general principle that a large number of tokens must be identifi-
able in the data, and it is clear that this stark necessity must override
any other consideration. Within this general constraint, however, we
70 Analysing Language in the Community

attempted to find representative variables that belonged to two

broadly-defined types (the second of which can be subdivided) that
we describe in terms of their linguistic (and not their social)
characteristics. These are: (1) variable sets in which the lexical items
are very numerous and are distributed throughout all or most of the
range of a phonemic class (we can call these ‘large-set variables’), and
(2) variables that are restricted to a small class of lexical items
(small-set variables). Within this second class, there are two sub-
types. The first type (2a) is phonologically defined and includes
variables such as (th), which concerns deletion of (6) intervocalically
in a small defined set of lexical items of the type mother, gather,
together. If enough tokens occur, this type is the easiest one to deal
with, and usually exhibits very sharp patterns of social divergence
according to age and sex of speaker in studies such as ours. The
second type (2b) consists of lexical sets, the membership of which
cannot be reliably predicted on phonological grounds alone, and this
type is additional to the types of variable that were studied by Labov
(1966) in New York City (these were types 1 and 2a). We do not know
for certain what the total membership of any such set is, how many
items belong to it, or the extent to which speakers vary in assigning
items to the set. These are open-ended sets, and they can be called
sets of phono-lexical alternants. Realizations of the vowels in the
relevant items belong to two different phonemes; these alternants
(like type 2a variants) are usually extremely socially salient in the
community, but for obvious reasons it can be difficult to deal with
them quantitatively. Of these, the set that I selected for careful
examination within the main projects is the (pu//) set, which deals with
alternation between [u] and [a] in a set of lexical items such as pull,
put, foot, shook. 1 shall return to the (pull) set as an example of
phono-lexical alternation, largely because such alternation appears to
be typical of dialect-divergent communities, and is extremely impor-
tant in the arguments that I shall develop in later chapters on
phonological change and on the effects of strong and weak social ties
in communities. But the (pull) set is only one of many: the others,
although often salient, are usually unsuitable for quantification
throughout the range of speaker variables because of the low numbers
of tokens attested (many examples are listed in J. Milroy, 1981).
This is not a watertight classification, and it is not exhaustive.’ It is
procedural only: we used it in Belfast as a guideline to help us to
select variables that would represent broader and narrower patterns
 Analysing Language in the Community a

of lexical coverage. However, the quantitative methods of the 1960s

and 1970s had successfully dealt with large-set vowel-variables (type
1), quantifying them through the whole range (or most of it) in a
single unilinear dimension of phonetic variation (such as raising or
backing). In the dialect-divergent community we found this proble-
matic, and we located only two vowel variables that could be reliably
quantified in this way. These are the two categorically short vowel
phonemes //, as in cut, dull (quantified as a binary counting rounded
v non-rounded realizations) and /1/, as in hit, fill (counting front v
centralized realizations). The difficulty in finding such variables is in
itself a measure of how internally divergent BV actually is, and it is
striking that the variables that show the sharpest patterns of social
variation are subsets within the larger sets. For consonant phonemes,
quantification throughout the phonemic range is of course im-
possible, since the potential for variation in consonants differs
according to the position of the segment within the word or
morpheme, but it is the difficulty with vowel variables that is most
significant here. In other studies, investigators have often been able to
assume that they know the lexical input to commonly occurring vowel
variables such as (e) and (a) and also that all the lexical items counted
have the much the same potential for variation (that is, the direction
of variation will be consistent throughout). In Belfast (and in
divergent dialect situations generally) the matter is much more
complicated.
In purely linguistic terms, there are two reasons for this difference
between Belfast and certain other studies. The first arises from the
fact that most BV vowels are long or short according to following
consonantal environment. The length difference is much greater than
allophonic length differences in RP and is easily audible. But the long
and short allophones are also qualitatively quite different. The low
and mid vowels, such as /e/ (as in set, bed), are short before voiceless
stops and before any sonorant + voiceless obstruent (for example, in
set, went), but long elsewhere (for example, in mess, bed, men). They
are also short in certain conditions in disyllables and polysyllables.
But for /e/, the short variant can be a low vowel ([a, 2]) and the long
variant is a mid vowel that is easily heard to be qualitatively distinct.
The pattern, which is more fully described in J. Milroy (1981) and
Harris (1985), is shown in table 3.1.
This speaker shows a very regular pattern with certain environ-
ments categorically low and others categorically mid. The result of all
V2 Analysing Language in the Community

Table 3.1 MC Senior (53) ‘formal’ style

Monosyllables All disyllables/polysyllables

Low Mid Mid Low

went (4) left (5) — electric

get men (2) heavy centre
set then (4) textile Wednesday
met well (3) very (8) fellows
next them terrible (3) definition
less (5) Devlin (2) — fun-
reckon damentals
twenty yourself
genera- fundamental
tions myself
general (2) Grenville
intellectual

this is that there are two subsets of /e/ with very different potentials
for variation. Our figures for short (€) in figures 3.5 and 3.6 show a
pattern of variation between low and mid realizations with the mid
realizations beginning to occur in environments that were formerly
categorically low, and with females leading in this change. The long
allophone of /e/, however, cannot be low, but varies between mid and
higher-than-mid realizations and can develop a centring glide. In the
simple type of quantification that we used, it is clear that we would
need to quantify the long-vowel environments separately from the
short-vowel ones, as the potential for variation is different in each
case. We can regard (€) as representative in two ways: it is represen-
tative of our method of analysis, but it also represents a pattern of
linguistic change that is valid for all the mid and low vowels: /a/,/e/,
and /9/. Where there is change in progress at the moment, we have
found that it is manifested in all these vowels as a pattern of
lengthening (or ‘tensing’) of previously short vowels.
The second reason for this relative complexity is that, as we have
already noticed, given lexical items do not necessarily belong to the
same sets as in ‘standard’ English. Thus, because of the divergent
history of the dialect, there can be subsets within larger classes (such
as (€)) which do not vary in the same way as other items in the set. In
 Analysing Language in the Community 7)
(e):
100 %

75 \.,
N+ Clonard
Hammer

Ballymacarrett
50

25 Sedpytee SES ! !
Men Women Men Women
40-55 40-55 18-25 18-25

Figure 3.5 Percentage low vowel in variable monosyllables and prefixed and
inflected disyllables.

100 %[- (e):

Celia

Clonard

50; Ballymacarrett

25 1 ! =f !
Men Women Men Women
40-55 40-55 18-25 18-25

Figure 3.6 Percentage low vowel in all other disyllables. The Hammer
figures (which have been excluded) correspond closely to those of the
Clonard.

the (€) set, for example, some short-vowel items (including get, never,
ever, yes) that should exhibit variation between [e] and [a] also have an
alternant in [1]. Thus, occurrences in [1] have to be excluded from the
main quantification, and the (get) subset, provided that we can
accurately specify its membership, can in principle be quantified
separately as a set of /exical alternants of the (pull) type (2b). It is in fact
74 Analysing Language in the Community

very common to find subsets like this within larger classes, and when
we quantify them they usually show sharp patterns of social differen-
tiation. But, as we have noticed above, we often cannot quantify these
subsets meaningfully throughout the range of speaker-groups and
styles because occurrences of the relevant variants are relatively rare.
Yet, it often happens that these variables are socially salient and
important for historical projection on to earlier states of language.
Therefore, as it takes a considerable amount of time to scan the
tapes looking for rarely occurring variants, we have selected one
particular variable of this kind, which we call meat/mate, and have
studied this quantitatively outside the main projects in a different way.
The variation here concerns the incidence of [e]-type pronunciations
in words of the type meat, leave, peace, easy, seize, and more particularly
the question whether the [e] variants of this class are merged with the
class of mate, save, lace, daze, or not. We are not primarily concerned
here with the distribution of this variable through the range of
speaker-groups and styles, mainly because this is quite evident from
our tape-recordings before we start, and without formal quantifica-
tion. The [e] variant is exclusively a casual style variant, which does
not occur at all in more formal styles and is unobtainable in word-list
elicitations. It is also much less common in female speech than in
male; therefore, our quantification is based on male speech alone.
And this selectivity is of course justified by my immediate aim, which
is phonological rather than sociolinguistic — to discover whether we
can reasonably speak of a meat/mate merger in Belfast English.
Clearly a number of factors that have to do with the history of English
(chiefly the reported merger of meat/mate in the sixteenth century),
with patterns of language maintenance, and with phonological theory
and description, have also motivated the choice of this variable, and
we shall return to these in chapter 5. However, the main reason for
mentioning it in this methodological discussion is to show how subtle
and fine-grained our analysis must be if we are to give an adequate
account of the function of language variation in close-tie communi-
ties. The paradox here is that this variable is very difficult to access
and would not be accessed at all if our field methods and analytic
methods were inadequate; yet, the variation has strong social meaning
within the community. Indeed, our quantified results suggest that the
[e]-type vowel of the meat and mate classes may have been maintained
as separate since Middle English, functioning as a social marker,
 Analysing Language in the Community 75

without merger, but with close approximation and overlap (J. Milroy
and Harris, 1980; but see further chapter 5).
The discussion so far has focused on the sociolinguistics of
dialect-divergence and the methods that can be used in order to
establish in-group norms of language use that are not previously part
of the knowledge of the investigator. In the discussion I have taken
the necessity of quantification for granted (although it sometimes
happens in such communities that some patterns of variation are
obvious without formal quantification). Before we proceed in chapter
4 to discuss the interpretative phase of social dialectological research,
however, it seems to be appropriate to add a few further remarks on
the use of quantitative methods.

3.7. Excursus: On the Uses of Quantification

We noted in chapter 1 that when we observe language in use, what we

actually observe is the speech of individuals in conversational sett-
ings—we do not directly observe overall patterns of use in the
sociolinguistic dimension. Thus, if—as an outside observer — you
listen to ‘core’ vernacular Belfast speech, it will actually sound all
much the same to you: you will not easily observe without systematic
analysis many of the differences that exist and you will not usually be
able to tell how far these co-vary with social factors, such as age, sex
and area. This is because the differences are in general not categor-
ical: the ‘dialect’ is a property of the community, and every native
speaker has roughly the same kind of access to it and roughly the
same knowledge about it. But the speakers that you are observing
‘know’ how to use the resources of variation available to them, and
they use them for many purposes, including the marking of varying
social roles and functions. Thus, the incidence of particular variants
differs according to these roles and functions. This is why a very
fine-grained analysis is needed, and why we will not get very far in
synthesizing our findings in relation to large databases like this if we
do not use quantification: as these differences are not categorical, we
usually will not be able to demonstrate them convincingly unless we
quantify.
76 Analysing Language in the Community

Quantification is usually taken for granted in social dialectology,

but it is not used in some other branches of sociolinguistics (for
example, those researches that follow Gumperz’s model), and there
can be disputes about whether or not it should be used in given
instances. We shall briefly consider some criticisms from within
sociolinguistics below. But questions have also been raised about it in
non-social areas of linguistics, and these are often attached to
reservations about the theoretical nature of sociolinguistics as a
discipline. Chomsky (1975) expressed the view that sociolinguistics is
a harmless activity like butterfly-collecting, but that it is not ‘theoreti-
cal’. Although this was some time ago, similar dismissals have
continued to surface in the work of those who adopt Chomsky’s
theoretical position. A recent example is Smith (1989), who thinks
that Labov’s work is not theoretical because it does not address ‘any
linguistic issue’. It is surprising that remarks of this kind from
influential linguists are still appearing, and that is why some com-
ments are needed here, not only on quantification, but also on the
relation between ‘method’ and ‘theory’.
Smith says that quantitative methods are inappropriate to linguistic
inquiries and cites as a warning Labov’s work in New York City,
which he represents thus: ‘The classic example of a quantificational
linguistic study is provided by Labov’s investigation of the determi-
nants (age, class and casualness) of New Yorkers’ pronunciation of
/r/, pre-consonantally and finally, in words like “fourth” and “floor”’
(1989: 180-1). Apparently, this refers not to the Lower East Side
survey, but to Labov’s preparatory department store survey (Laboy,
1966; 1972b). This of course was not the main part of the work, but a
pilot study used to test and refine some hypotheses about the wider
sociolinguistic situation, which was then investigated more fully over a
much longer period of time. But the whole study was set up within a
much more general theoretical orientation: the theory of linguistic
change; it was not merely an attempt to demonstrate age, class and
style differences, as Smith implies. So this account seems to be based
on a misunderstanding of what Labov’s work has been about. But
other commentators have also had something to say about the use of
quantification in linguistics. Lass (1980), for example, has pointed out
that stochastic (probabilistic) laws are not predictive, and has seemed
to suggest that the quantitative paradigm is theoretically uninteresting
for this reason. There have also been criticisms using similar lines of
 Analysing Language in the Community ig)

argument from within sociolinguistics (Romaine, 1981, 1984b;

Cameron, 1990).
But there is nothing to argue about here, as the act of quantifying is
not intended to be predictive, explanatory or theoretical: it is a
methodological tool that is used by those who wish to make account-
able statements about the distribution of linguistic forms in real
speech communities in cases where this is not evident without
quantification. In many situations, you will not be able to demonstrate
beyond reasonable doubt what the linguistic patterns in the commun-
ity are — what speaker-knowledge of those patterns is like and which
of the patterns show that linguistic changes are in progress — unless
you quantify. And you certainly will not be able to demonstrate these
things by relying on your own introspections! What must be under-
stood, therefore, is that there is a difference between the act of
quantifying data and the substantive claims made by the investigator
after the patterns have been established by quantification. These
claims (for example, that such-and-such a change is in progress) are
not in themselves quantitative claims: they are substantive and are
independent of the quantitative method. If the claims were made
simply on the basis of introspection without systematic observation in
the community, they would be of precisely the same order (except that
they would be unaccountable and probably much less reliable). So I
have treated quantification here as methodological: it is part of the
analytic phase of social dialectology, and it is parallel in this way to the
use of quantification and statistics in certain other sciences.
Both within sociolinguistics and outside of it there have been
discussions that appear to attack quantitative method for making
claims about causation and correlation. Romaine (1981, 1984b) and
Cameron (1990), for example, specifically cite our Belfast research as
an example of the ‘correlational fallacy’. Clearly, this kind of work
-would not be undertaken unless it was thought possible that causal
relationships between variables might exist, just as a medical res-
earcher might hypothesize a causal relationship between cigarette
smoking and lung cancer, or as a historical linguist might hypothesize
that a contributory cause of palatalization of [k] is an adjacent front
vowel. It is natural to look for causative factors, and we have treated
causality much as it is treated in experimental research generally (see
Plutchik, 1974: 174-87). But we do not generally fall into the trap of
believing that a statistical correlation between two variables demon-
78 Analysing Language in the Community

strates that one is the cause of the other; it is assumed that the
actuation of language change is multi-causal,° and we have frequently
demonstrated that the speaker-variables interact with one another
(that is, that no speaker-variable all by itself can ‘explain’ a given
configuration of language).
However, the misunderstanding about quantitative method can be
quite serious and can also get tied up with problems about statistical
significance testing. Smith, for example, further points out that Labov
did not use ‘tests of statistical significance, so his interpretation of the
figures is anyway suspect’. It is surprising that an anti-quantitative
linguist should advocate confirmatory statistical testing, but it is very
important to understand that the proposition put forward here is
simply wrong. If Labov’s interpretations were suspect (and of course
they are not), this would not arise from the fact that he failed to test
for significance. There was no reason for him to do so because the
claims that he wished to make were quite simple (they did not involve
complex interactions of different speaker variables), and because in
his analysis the same patterns were repeated for every variable
studied, tending in the same direction in every case in terms of both
class and style (several classes and several styles). Thus, the results are
very impressive: there is obviously a regular pattern in them, and as
they are accountable to observed data, they are much more persuasive
than categorical statements about language that are made on the basis
of introspection alone. The same applies to the examples that I
discuss in chapter 4: some of these show gross differences between
social groups on the basis of hundreds or even thousands of tokens,
which would not be apparent without quantification. Where there are
dramatic differences of incidence for different groups, where the
social pattern investigated is simplex (for example, class difference
only), and where the differences virtually always tend in the same
direction, it is often unnecessary to test for significance, because the
patterns revealed are so clear that no one could believe they are the
results of pure chance. In fact, since social dialectology is an
exploratory science, many important patterns of variation are ade-
quately demonstrated by exploratory statistical techniques, which are
designed to reveal patterns in the data rather than test for significance
(Tukey, 1977; Diaconis, 1985). The graphs and diagrams of the
quantitative paradigm are, in effect, examples of this.
Where there are numerous relevant variables, however, confirma-
tory statistical testing will help to give us confidence in our results,
 Analysing Language in the Community 79

and it is up to the analyst to decide in a given case whether these

methods are needed. In sociolinguistics they are normally used in
cases where several different independent variables (such as sex, age
and network) may be interacting, and therefore possibly cancelling
out the effects of one another. I explained this (J. Milroy, 1982b: 43)
in relation to the simplification pattern in short /a/ (see chapter 4.5
for the findings of this study), which we measured by using a
three-way analysis of variance. To simplify: it might happen that there
is no significant difference in terms of one social variable (such as
class), but that this statistic conceals other differences or similarities
within the results in terms of other social variables (such as age or
sex). Thus, it could turn out that a difference is best accounted for
not by the first social variable that you quantify (for example, social
class), but by a second or third one (for example, sex of speaker).
Since we carried out these studies, much more sophisticated
computational programs have been developed for the social sciences.
There have also been considerable advances in ‘variable rule’
methodology, with the development of GOLDVARB (Sankoff,
1986), and the journal Language Variation and Change has been
founded, with an editorial policy that concentrates on quantitative
methods of linguistic analysis. Properly used, it seems to me that
quantitative methods can only lead to advances in our subject. As for
my own research, there are some areas of it in which quantification
has not been used because in these areas it was not necessary or
appropriate for the purpose I had in mind (see for example the
discussions in chapter 5). In the social dialectology carried out in
Belfast, however, quantification was necessary: if it had been possible
to do the work without quantifying, we would have done it that way!

3.8 Conclusions

The purpose of this chapter has been to outline the main principles of
analysing language in the community, and I have paid particular
attention to exploring dialect-divergence. In such situations we
typically do not know beforehand what all the linguistic variants
involved actually are, so we must use very careful and accountable
methods of selection and analysis. But in previously unexplored
speech communities, we do not have reliable social intuitions either.
80 Analysing Language in the Community

That is to say, we do not know how these patterns of variation

function within the community: we do not know what social meanings
the variation carries for in-group members. To explore the situation,
therefore, and establish the relation of linguistic variables to speaker
variables, we normally require quantitative methods, which in some
cases may involve sophisticated statistical techniques. But as quanti-
tative method is not a primary interest of this book, I shall move on in
chapter 4 to the interpretative phase of this kind of research.
 4
Interpreting Variation in the Speech
Community

4.0 Introduction: Investigating Community Norms

Up to now, we have been mainly concerned with analysis of data. In

this chapter the emphasis is on the interpretative phase of the work.
What does this linguistic variation mean, and what contribution can
our analysis of variable states of language make to understanding
variation and change in language generally?
We have approached this task by focusing on the question of
linguistic norms, and it is relevant to recall here Principle 2 (chapter
1). In the lay person’s view, and in the view of some professional
linguists, the norms of language are associated with notions of
standardization and ‘correctness’ or with hierarchical dimensions of
social structure (or all of these), and they are usually felt to be
institutional: that is, they are thought of as being prescribed by
authority through the writing system, the educational system and
other agencies (for a relevant discussion, see J. Milroy and L. Milroy,
1985a). There is no doubt that factors like these are important in
what we may call the social ‘life’ of language, and they are implicated
in many aspects of linguistic variation and change. To understand
them properly, however, we need to relate them to wider social
structures, and we shall consider this aspect of sociolinguistics in
chapter 7. However, these institutional norms do not tell anything like
the whole story, and this is particularly true if we focus on spoken
language in casual conversation and on phonetic and phonological
variation: as we noticed in chapter 3, the norms of a superordinate
82 Interpreting Variation in the Speech Community

variety cannot be projected on to the norms of a speech community

without distorting our description. But more generally, if we use
superordinate norms as our main reference point, we cannot explain
why, despite superordinate pressures towards uniformity, varieties of
English and other languages can still remain so astonishingly diver-
gent from one another and so variable within themselves. It must be
the case that the norms of these variable states are agreed on by
internal consensus in the communities concerned.
These divergent states are often subjectively perceived as having
distinctive characteristics that mark them out as discrete varieties:
people can recognize regional varieties such as ‘Birmingham’ English,
‘Yorkshire’ English and so on, and they often have a fairly clear idea
of how such varieties are distinguished from one another. If this is so,
these varieties must incorporate within themselves sets of recurrent
and distinctive norms, through which they can be characterized, but
which do not usually coincide with the norms of the standard
language. This chapter is about determining the norms of usage that
characterize real speech communities and about the relevance of this
to determining the direction of change within the communities. The
kind of norms we are concerned with here are sometimes called
community norms in order to distinguish them from the superordinate
norms that I have mentioned, and I shall suggest below that a major
difference between superordinate and community norms is that,
whereas ‘standard’ norms are uniform, community norms are some-
times more aptly described as variable norms.
Underlying all this there is a more general question that impinges
very directly on the explanation for linguistic changes. This is the
second principle for the social modelling of change that I discussed in
chapter 1. As I pointed out there, it seems to be impossible to give a
full description of a linguistic variety (whether it is ‘standard’ English,
or a dialect, or a style or register) without making decisions about
what the ‘norms’ of the variety are, that is, without making decisions
about what the speakers agree on as structures that are appropriate for
that variety.
To take a simple example: suppose we are comparing two dialects,
one of which has the vowel /ei/ (in words of the type gate, place) where
the other has /ai/. In an intra-linguistic description, we will simply
state this difference as a structural linguistic fact (and do the same for
the other differences we notice) and that will be the end of it: dialect
A has one structure, and dialect B has another. But this difference is
 Interpreting Variation in the Speech Community 83

also a social fact: people are normally aware of differences of this kind,
and they often attach the utmost importance to them. We can easily
understand this if we assume that dialect A is ‘standard SBE’, and
dialect B is ‘Cockney’.’ However, in an objective language-based
description, one variant is just as ‘good’ or ‘efficient’ as the other. But,
as we noticed in chapter 1, it is a social fact in a much more
deep-lying sense, because these dialects are the possessions of their
speakers. The occurrence of /ai/ in one dialect and /e1/ in the other is
a result of the agreement within each community on a consensus norm of
usage within that community, and the difference cannot be explained
without some reference to this fact. There would be little point in
having these different norms (which are arbitrary in linguistic terms)
if they did not carry social meaning, distinguishing between one
community and another and carrying a sense of community identity
for speakers.
The same general point applies also to patterns of linguistic
change, and is most evidently relevant in the case of sound-change.
Suppose that in the course of time the vowel /a:/ in a particular
language becomes open /9:/, as it did in Southern Middle English in
words of the type home, stone (OE ham, stan). No purely internal
linguistic explanation can account for the fact that the change
happened in this way in some dialects and in different ways in others.
Nor can it explain why indeed the change should have happened at
all: why did the speakers not simply retain /a:/? Again, this
change, like synchronic variation between dialects, is a social, as well
as a linguistic, fact. It came about through changes of agreement
amongst speakers over the course of time involving greater and
greater consensus on open /9:/ as the appropriate realization, and it
is not fully explainable in purely linguistic terms. But just as speakers
attach great importance to variation in the structural parts of lan-
guage, so they must also attach great importance to change: if they did
not, then there would be no reason why changes like this one should
be accepted in speech communities.
But there is still more to be said about linguistic norms. Although
some of these community norms that we have mentioned above may
be recognized by outsiders as regional or social markers, many others
are better described as internal: they mark social differences inside the
community. These internal norms are the special property of the
community itself, and they are not evident to the outside observer;
therefore, to find out what they are, we must investigate the
84 Interpreting Variation in the Speech Community

communities in depth and at a fine-grained level of detail. The three

small inner-city communities that we studied in this way would have
been assumed by traditional scholars to be homogeneous in language,
and the fine-grained but systematic variation I am referring to here
would not have been thought possible. We were principally interested
in how linguistic norms that are highly divergent from ‘mainstream’
English are maintained in communities of this kind. As social class is
irrelevant in this case (there is little, if any, difference in social class
amongst our inner-city informants), the social groupings we used to
approach the sociolinguistic structure of the community were the
uncontroversial ones of age, sex and area, and social network was
adopted as an additional speaker variable during the course of the
research. We used this in conjunction with the other variables to
assist in exploring internal community norms of language.

4.1 Social Network

We have used social network in two ways within the research: first, as a
quantitative speaker variable and, second, as an interpretative cate-
gory, and its use in the projects arises from our interest in exploring
these community norms. As an interpretative category, it has impor-
tant advantages over other social models that we might wish to use.
The fundamental one in an investigation of this kind is that we do not
need to accept any prior assumption about how society at large is
organized or structured, and so in our interpretation we do not need
to import any presuppositions from theories of social class and social
structure or taxonomies of class or status, which may of course be
controversial. We need only accept the fact that individuals have
social contacts with other individuals, because social network is about
individuals and the relationships that can be contracted between
them, and not primarily based on pre-defined group structures. To
that extent it does not matter in principle whether the individuals are
described in a particular society as ‘upper class’, ‘middle class’ or
‘lower class’, or whether the society is rural or urban: it is a universal
that all individuals in all societies have contacts with other individuals
(even the exceptional case — say, a hermit — has occasional societal
contacts or has had them in the past, and ‘isolates’ are special cases).
The advantage over stratificational social class (as in Labov, 1966) as
 Interpreting Variation in the Speech Community 85

a principal variable is the universality of the network concept: whereas

stratificational social class theory cannot be universally applicable to
all historical states, to bilingual situations, or to language- (that is,
speaker-) contact situations (and all of these are relevant in historical
description), all speakers at all times have had ties of some
kind — weak or strong-—with other speakers. Thus, unlike social
class, social network does not require us to project at the initial stage a
fully-fledged theory of social structure on to the linguistic data. We
can therefore view social network, or some similar model of relation-
ships between individual speakers, as the basis of a strictly sociolin-
guistic theory.
Furthermore, the idea of network impinges fairly closely on the
notion of sociolinguistic functions, as we have presented this in chapter
2, because it is a fundamental postulate of network analysis that
individuals create personal communities that provide them with a
meaningful framework for solving the problems of their day-to-day
existence (Mitchell 1986: 74). So the network relationships con-
tracted between individuals are functional, just as linguistic variation
is also functional (L. Milroy, 1987, discusses examples in which social
network relationships are important enough to be viewed as survival
mechanisms). There are other conceptual similarities between ne-
twork analysis and linguistic analysis.
One of these is the fact that the social networks of individuals are,
like language itself, open-ended and changing, and they cannot be
precisely delimited for this reason. Furthermore, social network is
conceptually at a more generalized level than are definable groupings,
such as church or political organizations or street gangs (and the
model may in principle embrace these within it). Thus, a person may
be conscious of, for example, family and friendship relationships, or
membership of institutional groups, but he/she is not fully aware at
any point of the multiple web of (mainly informal) relationships that
constitute a ‘social network’. As it is conceptually so different from
the idea of the ‘peer-group’ used by Labov (1972a) and Cheshire
(1982), it is relatively difficult to operationalize social network as a
quantitative speaker variable.
In order to adapt the idea for use in an monolingual urban
situation, we have adopted for practical purposes certain procedures
from ethnographic work, for example, the idea of measuring the
intensity of network contacts; on the basis of this we can identify
clusters of individuals within which the network links are relatively
86 Interpreting Variation in the Speech Community

strong. Thus, for us, the idea of varying strength of network links is
the nearest we can get to understanding the notion of a speech
community. I shall return to, and elaborate on, these points. Here, the
main issue arising is that, following Bott (1971), strong network ties
can also be seen as norm-enforcement mechanisms. In this way, social
network gives us a model for demonstrating how states of language
are maintained through normative consensus within the communities,
leaving external influences out of the argument at this stage.

4.2 Variable Norms in the Speech Community

An important general finding of the inner-city research is that in most

of the linguistic variables quantified, there is a contextual style
difference for all twelve groups of speakers, and there is also a
sex-based difference. We have reported on the stylistic pattern quite
fully in a number of places (for example, J. Milroy, 1981; L. Milroy,
1987), and I will not go into detail on it here. Although in some
instances the sex/style differences are very slight, they are quite
consistent in that they usually tend in the same direction, much as
Labov (1966: 7) found for social class and style: ‘Native New Yorkers
differ in their usage in terms of absolute values of the variables, but
the shifts between contrasting styles follow the same pattern in almost
every case.’ The conclusion we are entitled to draw from these
findings is that, in general, female usage tends towards the more
‘careful’ end of the stylistic continuum and male usage towards the
more ‘casual’, and it seemed at this stage of our research that we had
some justification for the claim that in linguistic variation, sex-
differentiation is prior to class differentiation and need not be
interpreted as subsidiary to class (as it normally has been). Figure 4.1
is an example of clear sex differentiation in all three inner-city areas.
Notice, however, that although we may have views on the social
class distribution of [6] deletion (it can be readily observed to be less
common in middle-class Belfast English), social class is at this stage
an external category. We are interested here in finding out what the
norms are mithin a single social stratum and how they function for the
speakers: at this point in our work we have no justification for relating
this pattern to social class differences. Figure 4.1 shows only a
difference according to sex of speaker, but as this is quite consistent
 Interpreting Variation in the Speech Community 87

100 % (th):

75

Ne
\ *Clonard
50 <
Hammer

oe DOE , Ballymacarrett
Men Women Men Women
40-55 40-55 18-25 18-25

Figure 4.1 Deletion of medial [6] in words of the type mother, together.

between generations, it has another quality: it is also a stable variable.

It is this stability, and not any externally imposed category, that gives
us the important clue to how the sociolinguistic patterns are main-
tained in speech communities. The norms of the community include
stable differences between the sexes, with females tending towards
careful style variants and males towards casual style (thus, when
males are speaking in relatively formal settings, they will use a higher
proportion of ‘casual’ variants than females will). But the most
general conclusion that we draw from this kind of pattern is that the
norms of a speech community are not necessarily uniform within that
community, with every group agreeing on a single appropriate
realization. We can therefore interpret our findings as evidence that
what the community agrees on in this case is a pattern of stable
differentiation over two generations between male and female usage.”
It seems, therefore, that focused patterns in real speech communi-
ties are not patterns in which all groups speak in the same way, but
patterns of relatively stable differentiation within the community. In
Belfast, however, we found that the three inner-city communities
differed in the degree of focusing they exhibited. The community that
showed the most stable and consistent patterns of gender-
differentiation was the East Belfast Ballymacarrett community. It was
immediately hypothesized that the relative degrees of focusing and
diffuseness might be related to different degrees of social stability in
the communities. In Ballymacarrett, there was less unemployment,
traditional gender roles were more clearly differentiated, the popula-
88 Interpreting Variation in the Speech Community

tion was resident in traditional streets of terraced houses, and for the
most part it had been established there for generations. In the other
communities, unemployment was rife, traditional gender roles were
less clearly differentiated (with many women working while the men |
were unemployed), and there had been population movement. Addi-
tionally, the Hammer community was in the process of housing |
redevelopment, and many in-group members had moved out of the
immediate area. Figure 4.2 shows the flattening-out effect — an effect _
that was also noticed (but usually to a lesser extent than this) in other
variables: there is sharp differentiation in one community, but much
less in the others. In fact almost all the variables quantified display the
sharply differentiated pattern most clearly in Ballymacarrett.
What this suggests is that if there ever were a completely stable
community which was totally insulated from the outside world, there
would be stable differentiation within it— and gender-difference
might often be one aspect of this. Stable differentiation is the normal
state of affairs, and the existence of this variation is one of the things |
that makes linguistic change possible, in that the different variants
can be latched on to by different groups and for different social ©
functions: thus, the patterns of consensus can change in the course of
time. Conversely, a community in which large-scale quantification
reveals blurred and uncertain patterns of differentiation is likely to be
relatively unstable and, in Le Page’s terms, diffuse.

(a) Index score

300

oe Clonard
Brose? aa ~Hammer

ee = ade ; Ballymacarrett

100 = 2 aes 1
Men Women Men Women
40-55 40-55 18-25 18-25

Figure 4.2 ‘Flattening-out’ effect in gender variation in the Hammer as

against Ballymacarrett.
 Interpreting Variation in the Speech Community 89

The manner in which we used the social network variable has been
fully described elsewhere (L. Milroy, 1987), and I shall limit my
comments here to the question of focusing, stability and norm-
enforcement. Ballymacarrett was the only one of the three areas in
which there was a clear and consistent correspondence between the
strength of a person’s network structure and his or her use of
vernacular (or casual) variants. It was this kind of evidence that led us
to use the social network model in a systematic way: as Ballymacarrett
is the most stable and well-established of the communities, we can
conclude that the social conditions there are favourable to the
emergence of a close-knit network structure of the kind often found
in low-status communities (Young and Wilmott, 1962), and there is
ample ethnographic evidence that a close-knit structure of this kind is
capable of imposing normative consensus on its members. In other
words, relatively close socializing patterns have the effect of maintain-
ing traditional norms, including stable patterns of internal differentia-
tion, and resisting change from outside. What we have tried to show
here is that in this respect linguistic behaviour conforms to the same
principles as the other patterns of social behaviour that have been
studied by anthropologists.
But we have also noted that gender-differentiation in language is
sharper in Ballymacarrett than in the other communities. Here, male
network patterns prove to be more close-knit (as measured by a
network strength score: NSS) than female networks, whereas this
contrast is not so marked in the Hammer and is cancelled and partly
reversed in the Clonard (one of the highest-scoring groups for NSS
in West Belfast is the young female group in the Clonard, and this
largely accounts for the Clonard score in Table 4.1). In fact,
Ballymacarrett is the only area in which the gender difference in NSS
is statistically significant. This is clear in Table 4.1.

Table 4.1 Mean network scores (maximum 5)

Male Female

Ballymacarrett 3.9583 1.3333

Hammer 2.1250 1.8750
Clonard 2.7500 2.8750
90 Interpreting Variation in the Speech Community

Thus, as we have noted above, agreement on norms (as conse-

quence of close-knit and stable social patterns) results not in
uniformity of usage within a community, but on agreement on a
pattern of stable differentiation. In a relatively focused community
such as Ballymacarrett, therefore, we observe much greater regularity
in the patterning of the variable elements than we do in less focused
communities, and we may further suggest that the variants used in
such a focused community develop a clear and consistent pattern in
their social functions. That is to say, it is more generally agreed here
than elsewhere that preference for back realizations of /a/, for
example, indicate close personal ties, casualness of conversational
type, and/or male identity, and that other realizations indicate greater
social distance and/or female identity. The less focused the commun-
ity is, the less clear will be the relationship of specific linguistic
variants to their various social functions. These points are of the
greatest importance in developing a social model for the study of
linguistic change, and there is much more to say about relative
degrees of network strength as they affect patterns of change: I shall,
therefore, refer back frequently to these findings.
As our work has sometimes been interpreted as conflicting with
that of Labov, it should be noted here that the above characterization
of the speech community is fundamentally of the same type as his and
can be seen as derivative from his insights. Older characterizations of
‘speech community’, such as that of Wyld (1927: 47), assume that
everybody speaking a ‘dialect’ speaks in the same way: these scholars
would therefore have believed that inner-city Belfast is homogeneous
and would simply not have expected to find the enormous diversity
that actually does exist, so they probably would not have bothered to
investigate it. Labov’s (1966) definition, like ours, is based on
agreement on norms of variation within the community rather than on
the absolute values of the linguistic forms themselves. We can regard
his clear statement as a hypothesis that can be empirically tested (and
not a ‘tenuous abstraction’, as Lass, 1980: 121, views it), and our own
investigation (as reported so far) as empirical testing of his claim. The
difference that emerges is that whereas Labov’s statement is based on
speaker-agreement within the dimension of social class (and hence
derivative from the ‘shared value’ hypothesis of stratificational class
theory, on which see chapter 7), the characterization that I have built
up thus far is based on the variables of age, sex, area and network,
within a single social class, and without reliance on a prior model or
 Interpreting Variation in the Speech Community 91

theory of social class. It is this, and not any dispute about the
existence of variable norms, that leads in practice to a number of
differences in our treatment of ‘speech community’ and to some
differences in our model of linguistic change.
It follows from what has been said so far that, as stable norms can
be observed through analysis of linguistic patterns, change in progress
will show up as violations of the expected ‘normal’ patterns. More
generally, as linguistic stability consists in agreement on patterns of
variation and on the social functions of the variants, linguistic change
will appear as changes in agreement: as older patterns of agreement
recede, so new patterns of agreement will emerge, and in the interim
stage there may be some apparent randomness (or lack of agreement,
or conflicting patterns) in the distribution of variants. Linguistic
change, therefore, is change in agreement on norms of usage, and what we
observe in our quantified data is difference in the quantitative incidence
of certain variants in particular social groupings. This typically
appears as a gradual pattern: therefore, from a sociolinguistic point of
view, sound-change is most definitely a gradual process.
This view of what is meant by sound-change is very different from
the traditional view, because it is speaker-based and not system-
based. Let us explain this briefly. A language-centred description
typically suggests that speech-sounds are objects that can actually
change ([a] can ‘become’ [o], and so forth), when this is not what
happens at all. Old English [a:], for example, did not actually
‘become’ open [9:] in Southern Middle English (as in OE ham >
ME hom(e)). What happened was that speakers gradually and variably
began to use open [9:] in environments where [a:] had formerly been
used: the process was one of substitution rather than change sensu
stricto.
For all these reasons, we have treated sound-change as something
rather wider in scope than a simple unilinear process in a purely
phonetic/phonological dimension. It is accepted that change in this
unilinear aspect will indeed appear in our quantified findings — as a
crossover pattern which violates the pattern of stable norms (as in
Labov’s classic example of class/style overlap for post-vocalic /r/ in
New York City), and I shall discuss examples of this in later chapters.
However, it is characteristic of quantified patterns that they can
conceal what lies behind them and so they can be interpreted much
more deeply than in this unilinear way. In the remainder of this
‘chapter, therefore, I am concerned with the wider patterns of
o2 Interpreting Variation in the Speech Community

consensus and conflict in speaker-evaluation of the linguistic variants

that we have studied, and I shall relate these to the concept of strong
and weak ties in communities. | shall start by considering in section
4.3 some variables which are characteristic of Belfast English, but
which seem to function at a somewhat higher level of generality than
those that we have mentioned so far —as identity markers for the
community as a whole rather than for internal differentiation within
it.

4.3 Variables as Markers: the Identity Function

As we have seen, linguistic variables may be markers of certain

fine-grained social functions within the community,’ which can be
broadly considered to be identity functions, and which may appear as
age, sex or areal differences, for example. However, it may happen
that particular variables selected for analysis turn out to be especially
markers of one of these functions rather than another: one variable
may be chiefly a close-tie network marker and another a marker of
gender-differentiation. It is for this reason that in our quantitative
analysis we have treated speaker variables as independent of one
another. In fact, a particular variable may have more than one
function, but it is only by a close analysis that treats speaker variables
as independent that we will be able to determine whether this is so. I
shall discuss an extended example of this in sections 4.6 and 4.7.
Here, I have another purpose: I want to draw attention to the fact that
different variables may be used by the community to mark functions
of higher and lower levels of generality. For example, some variables
(or variants of them) may be markers of gender-difference in the
close communities, whereas others may not show fine-grained in-
ternal differentiation, but may be best interpreted as variables that
mark Belfast vernacular as a whole as different from other varieties.
These are not so much part of the ‘inner voice’: in Cohen’s (1982)
terms, they are perhaps more readily interpreted as the ‘voice to the
outside world’.
It is clear that some variables in English do function in a very
general way: for example, variation between the alveolar and velar
nasal in the present participle ending (ing) is universal and can be said
to mark the whole English-speaking world as a single speech
 Interpreting Variation in the Speech Community 93

community. More narrowly, but still at a high level of generality, some

variables may function as identity markers for whole status groups or
for larger regions than merely an inner-city community. In British
English, examples of this are [h]-dropping and glottalling, which are
regionally widespread and usually marked for social status or style. As
these broader functions are of importance in a theory of language
change, I shall discuss them briefly here.
Some of Labov’s New York City variables are of this broader type,
and the larger number used by Trudgill (1974) are mixed in this way:
some of them (such as [h]-dropping and (ing)) are common to many
varieties, whereas others have social meaning only, or mainly, for
Norwich and Norfolk speakers. As our inner-city work was mainly
focused on very fine-grained differentiation, we have so far paid less
attention to these more generalized markers in Belfast English, and in
some cases we have not published the quantitative findings that arose
from studying these. One such variable that has a wide distribution is
the lateral consonant (L). In a number of dialects in northern England
and peripherally in southern Scotland (J. Milroy, 1982c) there is a
high incidence of noticeably ‘clear’ (palatal) variants of /1/, chiefly in
pre-vocalic syllable-initial position, but also in other positions. As for
Irish English, commentators routinely state that clear [l] is ‘ubiqui-
tous’.* For Belfast, however, this is certainly not so. It is characteristic
of the inner-city vernacular that a noticeably clear (palatal) [I] is
pre-vocalic but that a noticeably dark (velar) [Il] is post-vocalic. But
what is of interest here is that there is no difference between the three
inner-city areas in the use of initial clear [I]. It is firmly established,
and when there is a tendency to ‘darkening’ of initial [I], the
conditions that apply are linguistic and not social: relatively ‘dark’ [I]
occurs mainly before relatively retracted vowels. Thus, the clear [I] in
initial pre-vocalic position is a broad regional marker, and we
conclude from this relative invariance that it is not undergoing
change. The Belfast contrast between clear and dark [I], however, is
not characteristic of all Irish English or of all Ulster English. It is a
feature of Belfast and surrounding areas and marks the speaker as a
native of this region. However, at a more fine-grained level, there are
differences in the incidence of the dark [I] variant (but not initial clear
[l]) in the inner city according to age, sex and area. The findings for
initial (L) are given in figure 4.3: as the total possible score is 200 (for
maximum incidence of dark [I]), these scores show a very marked
preference for clear [I] in initial position for all groups with no
94 Interpreting Variation in the Speech Community

80-

dark ih

5 all CT cr. beeri

aS ee

30, DreBALIe see3

i —__‘*-,Clonard
Hammer
sh Ballymacarrett

10
clear [
[beset N fy Sa =e :
Men Women Men Women
40-55 40-55 18-25 18-25

Figure 4.3 Average index scores by area for (L) in initial positions
(maximum score: 200).

significant differences, and only a tendency for the young females to

increase in their clear [l] usage (Owens, 1977).
At a somewhat less general level, however, there are also variables
that function as markers of BV (as against more general Ulster or
Irish varieties) and that do not vary significantly within the inner city
in social terms. An example of this is variation between low and
mid-front realizations of /a/ before velar consonants in words of the
type rag, pack, bang, which show only slight differences in terms of age
and sex within inner-city Belfast. However, the front-raised realiza-
tion is easily perceived to be a very salient marker of vernacular
speech in Belfast: as an identity marker, therefore, it marks nativeness
to Belfast and adjacent areas, and it seems to be very firmly
established. It contrasts in this way both with middle-class speech and
with many of the dialect regions of south, mid and west Ulster.
I have discussed these matters briefly here because the broad
distinction I have drawn is relevant to describing the ‘shape’ of the
speech community and locating patterns of change within it. The
markers that are characteristic of a variety perform social functions at
lower and higher levels of generality. In the main parts of the Belfast
projects we were more concerned with these lower levels of gener-
 Interpreting Variation in the Speech Community 95

ality, because the findings have consequences for the notion of

‘speech community’. As we have noted, if there were ever such a
thing as a variety that was completely insulated and unaffected by
external influence, these findings suggest very strongly that it would
not be uniform and that there would still be variable community
norms within it. But in reality there is no such thing as a wholly
insulated variety: however strong the links may be that bind a
population together, there will always be some consciousness of
external norms, and this will have two kinds of effect on in-group
behaviour. First, there will be resistance to external influence, and
this will manifest itself as consensus norms within the dialect that
affirm solidary behaviour. The agreement throughout the inner city
on the (bag) variable can, for example, be interpreted in this way.
Second, the external influence will in some cases actually be success-
ful, and we may therefore be able to demonstrate how the new forms
penetrate the solidary community. This implies that we need to take
conflict into account here as well as changes in consensus.
I shall return in chapters 6 and 7 to consider these points more
fully in the context of a model of weak and strong ties. In the
remainder of this chapter, | want to use the solidarity model
suggested by ‘social network’ to help to interpret patterns of variation
in Belfast, referring to the inner-city and outer-city community
studies and to our random sample ‘doorstep’ survey.

4.4 Investigating Patterns in the Inner City

In this section and the following ones I am concerned with the ‘shape’
of the Belfast speech community as a whole, and in particular with a
kind of patterning that is not directly accessible by the classic methods
of Labov (1966). These methods typically use a hierarchical model of
society and measure linguistic variation in a single phonetic dimen-
sion (such as, vowel-height) against a unilinear social dimension (in
most cases socio-economic class). The result of this is that linguistic
variation and change can appear to be unidimensional. or example,
within a particular phoneme class, social groups can appear to be
differentiated by greater or lesser preference for one particular
phonetic realization of a single sound-segment. In this respect, these
methods fit in well with traditional views of sound-change also, for
96 Interpreting Variation in the Speech Community

these too are frequently unilinear (vowel A ‘becomes’ vowel B in the

course of time). What these methods do not directly tell us is how
wide or narrow the range of internal linguistic variation is within each
social group, and how much the groups differ in the range of variants
used. It may, for example, happen that social group A uses only two
variants of a variable, whereas social group B uses three or four, and
this type of difference may well be of just as much interest as an
absolute difference in phonetic realization.
In fact, these differences in extent of variability within groups are
sociolinguistically very important. It was clear from a very early stage
of our Belfast work that to describe the inner-city phonology was a
more complicated task than to describe ‘middle-class’ phonology,
because there appeared to be much more variability within inner-city
language than in higher status language. In this dimension, therefore,
the speech community can be envisaged as being shaped like a
pyramid, with greater variability at the lower end and greater
convergence (or relative uniformity) at the upper end. It seems to be
worth inquiring why this should be so, and what generalizations about
sociolinguistic variation and change we can draw from this fact.
To approach this question, we bear in mind the principle that
variability in language is socially functional. We also bear in mind the
principles enunciated by Cohen (1982: 8) of ethnographic work on
small solidary communities, which I have referred to in section 4.2
and which distinguish between the ‘voice to the outside world’ and
the ‘much more complicated’ voice of the community ‘to its own
members’. According to this view in-group variation (in social affairs
generally) is not only functional within communities, but also subtle
and complicated, and difficult for outsiders to access. We are
interested here in understanding the functional pressures that main-
tain these complex patterns of variation.
For our present purposes, there are two main senses in which the
inner-city phonology can be said to be complicated: first, there is a
much higher degree of ‘low-level’ allophonic variation in the inner-
city than in outer areas, resulting in a wide range of variation and
frequent overlap between phonemes; yet, this variation can be shown
to be rule-governed; second, there is a high incidence of what I have
called phono-lexical alternation (as measured in variables of type 2b) in
the inner-city, which is much reduced in outer-city communities. In
chapter 3, we noticed that lexical alternants (of the pull type) are
highly salient in the community, and it has been shown in various
 Interpreting Variation in the Speech Community 97

studies (such as J. Milroy, 1981) that these alternants encode values

of greater and lesser solidarity. In what follows my main purposes are:
(1) to demonstrate the patterns of simplification that can be traced by
comparing our inner-city data with that of the city-wide random
sample ‘doorstep’ survey and the outer-city community studies; and
(2) to consider how far a theory of strong and weak ties can account
for the maintenance of complex patterns and the development of
simpler ones. My first example concerns allophonic complexity. I
shall use here the data on short /a/, but similar conclusions hold true
of other vowels (for greater detail, seeJ.Milroy, 1976a, 1981; Harris,
1985).
The Belfast /a/ system varies considerably in terms of length,
height, backness, rounding and diphthongization, and shows variation
to a degree not attested in standard accounts of English phonology: if
we include pre-velar items (as in table 4.2), the range (in place of
articulation) is from mid-front to low-mid back. Table 4.2 is a
simplified representation of the range of /a/, in terms of following
consonantal environments. It is simplified in many ways: it ignores
patterns of length and diphthongization and excludes tokens of
post-velar environments in which front-raising occurs (chiefly among
older speakers). Yet, it is still sufficiently detailed to make our main
points.

Table 4.2 Range of realizations for /a/ in inner-city Belfast

fs Be a a 9

bag back back grass bad

bang flash bag bad man
flash man can
chap pal hand
hat hand
can’t can
ant
aunt
dance

The pattern displayed here incorporates variation, but notice that the
distribution of variants is rule-governed. The pre-velar vowel can
vary between mid-front and low-front, but it cannot undergo backing.
98 Interpreting Variation in the Speech Community

Short vowels before voiceless stops (or sonorant + voiceless obs-

truent) are also front, but non-velar items cannot undergo front-
raising. Before fricatives (except palatal fricatives) and voiced conso-
nants generally, a long back vowel is favoured, with labial and nasal
environments most inclined to display the rounded back vowel. In
short, the heterogeneity in the system is orderly: it can, in principle,
be described in terms of variable rules. Similarly, it can be described
phonologically in terms of a set of sub-scales, in which the potential
for lowering and backing will vary in degree for different subsets.
We had much evidence during fieldwork that these complicated
rules are maintained by community pressures: for example, when a
young East Belfast informant applied a strategy of front-raising
(similar to RP) to the item chap (on a word-list), he was loudly mocked
by his companions, who perceived his pronunciation as an attempt at
RP (which for them is an effeminate stereotype). The item can be
realized with a fully low front vowel, and a strategy of backing (but not
of front-raising) can also be used in this social group.
To return to the main point about allophonic complexity: it is
obvious, even from the simplified displays in the table, that these
complex patterns are not easily learnable for outsiders, even though
they are regular. How, for example, could an outsider know that
normative consensus in the speech community does not permit
backing in velar environments or fronting in most other environ-
ments, or that sonorants followed by voiceless obstruents predict a
front vowel? It seems that these patterns are maintained by insider
knowledge depending on the extent to which speakers belong to
relatively close-knit groups. To the extent that these close-knit social
relations are weakened, we may expect that these linguistic complex-
ities will be reduced. In section 4.5 I shall adduce evidence from the
outer city that suggests that this view is correct.
A second sense in which the inner-city phonology may be said to be
complex is in the incidence of lexical items that have two alternative
vowel pronunciations quite distinct phonetically and phonemically
from each other. There are many such sets, and most of them (such
as the meet/meat alternation) are highly salient. They are discussed in
chapter 5 and by J. Milroy (1980, 1981);J.Milroy and Harris (1980);
Harris (1985).
As exemplification of this, we have usually used the (pull) set.
These items can be realized with a low-mid back short vowel
(unrounded or slightly rounded), alternating with a high central to
 Interpreting Variation in the Speech Community vas)

front rounded short vowel. In one realization they rhyme with cut,
dull, and in the other with good, pool (both of which have short [u] in
Scots and Ulster English). The set has deep historical roots and
consists mainly of ME /u/ items with initial labials (such as pull, bush,
full), and certain ME /o:/ items (such as shook, foot). Since 1860, as
far as we can tell, three items that were formerly alternants have been
reclassified as categorical [u] items (Patterson, 1860); thus, the speed
of transfer has been very slow, and the set as a whole has been quite
resistant to change. As we noticed in chapter 3, the full membership
of this limited set is not predictable by environmental phonological
rules: for example, whereas shook, look belong to it, cook, book do not.
The items wood, wool, hood and probably soot used to belong to it, but
do not now (Patterson, 1860). Yet, although mood no longer alter-
nates, the homophonous item mould does alternate. Furthermore, if
hood could belong to the set, there seems to be no reason why good
should not also have belonged to it; yet, as far as we know, it never
did. Nevertheless, inner-city speakers know that some /u/ items can
alternate and that others cannot. Thus, although some environmental
phonological factors (such as initial labials) are involved in defining
the set, speaker-knowledge of which items can alternate must de-
pend largely on memorizing the items singly. It seems that the best
explanation for the survival of this kind of insider knowledge is that
the knowledge is reinforced by frequent close contact with other
speakers, and this is best accessed by a theory of network ties, which
when they are strong function as norm-enforcement mechanisms.
Furthermore, as we might expect, these alternations are functional.
The existence of these sets provides an important sociolinguistic
resource for inner-city speakers, and their survival must surely
depend on this. First, in a broad sense, the ‘in-group’ variant: [A], can
be held to affirm group identity. More importantly, I think, the
alternative choices can encode messages of social nearness or social
distance — or, if you like, degrees of social distance. The in-group
form encodes messages of intimacy and closeness, and, like the T
pronoun (Brown and Gilman, 1960), can actually be required amongst
close friends in casual circumstances: its strong affective meanings,
however, are not easily accessible to outsiders. The other alternant is
the out-group form, and this of course is used in interactions with
those who have relatively weak ties with the speaker, or in situational
contexts in which social distance is present. Again, we have evidence
from fieldwork about the functions of these alternants: one man of 27
100 Interpreting Variation in the Speech Community

commented that during his years away from Northern Ireland, he had
stopped using what I have called the in-group alternants of the (pu/l)
set. The most obvious explanation for this is that during his years
away from home, he would develop a large number of relatively weak
ties: in such circumstances, the in-group alternant would cease to
have any function for him and so could be abandoned. This
functional explanation seems to be more satisfactory than, and
logically prior to, one based on prestige: as it happens, this individual’s
activities away from home had not been upwardly mobile.
If weak-tie situations can exist even in the inner-city communities
themselves (L. Milroy (1987: 131) discusses the case of Hannah McK
from the Clonard area who has a density/multiplexity score of 0), it is
to be expected that within the wider social structure of a city, social
distance between groups will develop even more. In such cir-
cumstances, the norm-enforcement mechanisms that maintain the
complex structure of inner-city variation (and other patterns of
language use) will be weakened, and simplification is likely. In section
4.5 I shall use some of the findings of the doorstep survey and the
outer-city community studies to inquire further into this.

4.5 Linguistic Patterns in the Wider Speech Community

The traditional view of urban dialect was that it is not ‘dialect’ at all.
This term was reserved for the rural dialects, which had been
legitimized by nineteenth-century investigations, and which were
believed to be ‘genuine’ in a way that urban dialects are not. The
latter were widely regarded as corrupt and impure (see for example
the views on Cockney cited in note 1, above). This attitude to urban
dialect seems to have been given further impetus by the view of Wyld
(1927) that urban English can be described as the ‘Modified
Standard’ of ‘city vulgarians’. This implies that it can be characterized
as an unsuccessful, or partly successful, effort by urban dwellers to
achieve competence in ‘standard’ English (which for Wyld is the
English of the upper and upper-middle classes). Our work in Belfast,
however, and that of colleagues in other cities, makes it clear that this
cannot be convincingly demonstrated. The language of urban people,
to the extent that it is ‘modified’, is modified vernacular: it cannot be
adequately explained in terms of modifications to the ‘standard’ (in
 Interpreting Variation in the Speech Community 101

which many speakers show no interest at all). It is easier to

understand it as an attempt to move away from something already
characteristic of the community than as an attempt to move towards
some outside idealization (compare my comments on avoidance of
stigma in Chapter 2.6, above). In this section, I am concerned with
the modifications to the inner-city vernacular that can be observed in
outer-city language.
In our second project, we selected our two outer-city areas rather
carefully. We wanted them to represent a section of the community
that was quite close in the social hierarchy to the inner-city popula-
tion, and in the event many speakers in the samples had inner-city
social and family contacts. The intention here was to test our theory
by making things difficult for ourselves: it would have been much
easier to select upper-middle class communities and to demonstrate
gross differences between them and the inner city. This, however,
would have been to trivialize: the results would be of little interest.
But the speakers in Andersonstown and Braniel are strong Belfast
speakers: an outsider would be unlikely to notice much difference
between their speech and that of the Clonard and Ballymacarrett. In
order to show that there are differences in language, therefore, we
have to listen to many hours of tape-recorded speech, select variables
that we believe may pattern in some way, and quantify. The most
salient difference between these outer-city informants and the inner-
city ones is not linguistic, but social, and of course we knew this. As
for the doorstep survey, this was carried out in order to generalize
about the speech of the city as a whole, and in this way to be able to
describe modifications to the vernacular that might be of general
import. Here, I shall first consider some results of the doorstep
survey.
I have elsewhere shown (J. Milroy, 1982b) that the range of
variation in (a) is greatly reduced in speakers outside the inner city.
Table 4.3, which is from the doorstep survey of speakers throughout
the city, shows total convergence on a low front vowel for /a/, and
table 4.4 demonstrates another interesting point. In it the two
front-raised items (castle, dabble) appear to be randomly front-raised
rather than governed by any systematic rule. The vernacular rules
predict that dabble will be subject to backing — certainly not front-
raising— and it is unlikely that this speaker is following the highly
recessive rule for front-raising after velars in the item castle (this has
been largely abandoned by younger low-status speakers). Table 4.5,
102 Interpreting Variation in the Speech Community

Table 4.3 /a/ range for a middle-class Belfast speaker: word-list style
(random sample survey).

2) ree aeagia:a

bag
back
cap
map
passage
cab
grass
bad
man
castle
dabble
passing +++++4¢+4+4+4+4+4

Index score of convergence on [a]: 0 (max. convergence)

Range score: 0 (min. range)

Table 4.4 /a/ range for a middle-class Belfast speaker: word-list style
(random sample survey).

(a ee PEL aT Es

bag
back
cap
map
passage
cab
grass
bad
man ++t¢4¢4+4+4+4
castle at
dabble ai
passing -

Range score: |
 Interpreting Variation in the Speech Community 103

Table 4.5 Range of variation in /a/ showing simplification pattern by

status group and gender

Average range scores (maximum 5):

Lower group (LWC-MWC) 2.83
Upper group (UWC-MC) 1-97

No. of speakers with range of 1 or less:

Lower group 1 ( 4 per cent)
Upper group 1] (31 per cent)

No. of males and females with range of 1 or less:

Males 3 (10 per cent)
Females 9 (30 per cent)

No. of speakers with range of 3 or more:

Lower group 16 (66.7 per cent)
Upper group 10 (27.7 per cent)

based on the output of 60 speakers, shows that the reduction in range

for this vowel throughout the city is clear and consistent (statistically,
the pattern is very highly significant). Furthermore, it is sex-graded:
females are significantly more likely to simplify than males. Although it
is impossible in a doorstep survey to estimate network strengths for
individuals, it is extremely likely that varying network strengths are
also implicated in the simplification pattern. The more dense and
multiplex the ties, the more likely it is that complexity will be
maintained.
In the doorstep survey, two further variables were fully quantified,
and others less closely examined. The pattern for (€) (as in went, bed)
proved to be virtually identical to that for (a). Superficially, the results
for the lexical alternant class of (pull) were much the same also. The
lower-status group figures were again very much higher than those
for the upper group (2.17 as against 0.91, out of a possible 4.00): this
means that lower group speakers, on average, use the vernacular
alternant for two or more of the four lexical items, even in word-list
style — a surprising finding in terms of Labov’s predictions on contex-
tual style. Word-list style is a careful style in his terms (but as we have
shown in a number of publications, Labov’s predictions about the
stylistic continuum did not work in this divergent-dialect community
104 Interpreting Variation in the Speech Community

in the way that they had worked in New York City). The figures for
males and females are also consistent with our findings for (a) and (e):
females lead in the move away from traditional vernacular forms,
and this in the long term tends to reduce allophony and simplify
alternations. Thus, it may well be worth considering the possibility
that sex-differentiation in language contributes to simplification
patterns and hence to the establishment of supra-local norms. We
now turn to the findings of the two outer-city community studies,
which also show a pattern of reduction of allophony.
The social characteristics of the two outer-city communities differ
quite considerably from those of the inner city. They are upper-
working to lower-middle class. The people were not necessarily born
and brought up in the same neighbourhood; many are upwardly
mobile (unlike the inner-city people). Above all, very few of them
meet more than one of the criteria used in the inner-city for
measuring network strength, and the essential reason for this is that
their network ties are not mainly territorial. The inner-city indicators
were based largely on territorial assumptions, for example that people
might have close relatives living in the same street or the next street,
and that people would work in the same places as some of their close
neighbours (L. Milroy 1987: 141-2). This was seldom true in the
outer city. Those people whose families had moved out from the
inner areas still retained some ties with relatives in the inner city, but
clearly such ties are by definition weaker in quality than ties with
immediate neighbours, and they were dismissed as relatively weak in
our inner-city network analysis. In other cases, people had no
inner-city ties, and ties with neighbours tended to be uniplex and
open-ended, rather than multiplex and dense. Sometimes people did
not even know their neighbours. Thus, although these outer-city
neighbourhoods do not to a casual observer differ much from the
inner city in language, they do exhibit clear social differences. In our
project this meant that measures of network strength that we used in
the inner city could not be readily operationalized for these speakers.
In this respect, therefore, the outer-city studies lay the basis for the
arguments about weak ties put forward in J. Milroy and L. Milroy
(1985b) and further discussed in later chapters.
In such conditions, one important prediction is that the weakening
of norm-enforcement mechanisms will bring about a reduction in the
everyday functional value of the use of in-group linguistic variants.
Thus, we are prepared to find that the simplification pattern apparent
 Interpreting Variation in the Speech Community 105

in the findings of the doorstep survey may also apply to the outer-city
neighbourhoods. In addition, the breakdown of inner-city consensus
may lead to patterns of usage that are less predictable by linguistic
rule or by sociolinguistic pattern. Thus, some apparently random
variation, or some lack of agreement on norms of usage (as in table
4.5) may be encountered.
It had been noticed in the inner city that many vowel variables
could be described in terms of sub-scales according to following
consonantal environments (as in table 4.5); we attempted, therefore,
to operationalize this perception by distinguishing following environ-
ments roughly based on the sonority scales discussed by Taylor
(1973) and others adapted to an ‘allophonic length’ dialect (J. Milroy,
1976a). They were as follows:

T = voiceless stop (incl. affricate) or sonorant + voiceless obstruent

TS = environment described in T + a following syllable in same
morpheme
D = voiceless fricative or any voiced consonant not immediately
followed by a voiceless segment
DS = environment described in D + a following syllable in same
morpheme

These categories were used in the quantification of (a), (9) and (e) (as
in cap, stop, bet) in the outer city (a modified version is used in sections
4.6 and 4.7), and a particular aim of the method was to distinguish
between variation in quality and variation in length. Hence, in the
range diagrams (tables 4.6 and 4.7), tokens are listed in columns
showing length as well as place of articulation. It is immediately

Table 4.6 /2/ in Andersonstown

a a a a: a) 2H

T got (2) T shop

DS Polytech
T shop DS probably D job (3)
T pot DS concentrated D of
DS vodka D God
TS bottom
106 Interpreting Variation in the Speech Community

Table 4.7 /e/ in Andersonstown

a € E: € €: €:

T set-up (2) DS specials D

T lent T went red
T went (2) D tell
DS specials (3) D ten
DS remember
TS twenty

obvious here that, for (9), the extreme left-hand column is empty, and
that, for (¢), both the extreme left- and extreme right-hand columns
are empty. In the inner city, however (see the Clonard figures in table
4.8), not only is the full range exploited, but the empty left-hand
columns (for environment T in the inner city) are in the order of 100
per cent (that is, categorical) for virtually all the males (young and old)
in the sample. This compares with quantities close to zero in the
outer city, for most speakers. Similarly, the range from low to
high-mid (in the case of (€)) is reduced to a much narrower range in
table 4.7. In this connection, Gunn (1982) reports that for (a) also the
range in Andersonstown is reduced to variation between [z] and [a]
(the latter sometimes slightly retracted), but he adds that this is
particularly the case for young speakers; older people frequently
retain the vernacular rules of the inner city. Thus, these findings
support the findings of the random sample, in which, as we have seen,
the range of realizations is reduced in ‘upper group’ speakers.
There is also some indication that the appearance of particular
items in particular columns is less easily shown as rule-governed than
it is in the inner city. It is rather surprising to find, in table 4.6, that
the item shop appears with a low back slightly rounded Jong vowel, and
the rounded vowel in the disyllable/polysyllables probably, bottom is
also rather surprising. These trends might of course represent the
beginning of a change in progress, but if this is so, it is not yet
established as a pattern that we can show by our methods as regular,
and so we cannot demonstrate that it is a change. But from this, and
from table 4.8, it is reasonable to conclude that the situation is in this
 Interpreting Variation in the Speech Community 107

respect less focused than in the inner city. The agreement on norms
that we noticed in the inner city seems here to have been weakened.
Table 4.8 measures in broad terms low as against non-low
realizations of (9) in the two outer-city areas as against the inner-city
Clonard area. Although this table ignores backness, roundness and
other phonetic categories, it also shows us something of interest. The
rule constraining low v non-low realizations is much closer to being
categorical in the low-status inner-city area than in the outer city;
indeed, for the T and TS environments, it is categorical. As we have
noted above, it is qualitatively a near-front unrounded vowel categor-
ically for most male inner-city speakers, but table 4.8 shows that,
regardless of fronting, it is still always a low vowel in Catholic West
Belfast. This finding may be interpreted as supporting the view
expressed above that the inner-city vernacular, although it shows
more complex patterns, is more amenable to description in terms of
linguistic rules than outer-city speech. We can add that it may be the
effects of network strength that maintain this relative consistency.
The general picture that emerges from these patterns of variation is
one in which simplification patterns and loss of regularity and
consistency are related to the weakening of network strength and the
development of patterns of weak ties. There may well be other
patterns of simplification besides these, and one of these seems to be
the merger of classes that were previously distinct. The gradual loss
of alternating classes, such as (pull) is also related to merger patterns,
as the end result of this long historical process is loss of a distinction
which at the moment is still socially functional. In the present
perspective, I would propose that resistance to merger, lexical

Table 4.8

Andersonstown Braniel Clonard

(+ low) 81 (+ low) 64 | ,_
Tot 19 erbs OEP how 36 Jie
(+ low) 74 /-TS (+ low) 74 /-TS {+ low] (100%) /-TS
(— low) 26 (— low) 26 He
(— low) 86 (— low) 93 { ,_ (— low) V8
OMe elon To |ce lows 13 es
(+ low) 80 /-DS (+ low) 51 /-DS (+ low) 91
(— low) 20 (— low) 49 (— low) 9
108 Interpreting Variation in the Speech Community

transfer and restructuring is promoted by the existence of strong

network ties, and that types of change that result in simplification are
encouraged by weakening of ties. Complex patterns, which (as we
have argued) are functional, are maintained by the norm-
enforcement function of dense and multiplex networks. For those
whose ties are uniplex and (relatively) open-ended, these patterns are
no longer functional and it is for this reason, and not primarily
because of speakers’ desire for ‘prestige’, that they disappear.
The purpose of our discussion so far has been to demonstrate that
there are important patterns of change that are not readily captured in
a paradigm that measures a unilinear scale of phonetic variation
against a unilinear scale of socio-economic class. I have also avoided
using the concept of prestige in the discussion because it appears that
the patterns can be interpreted without reference to prestige. If we
argue, nevertheless, that prestige models can still account for the
trend to simplification and uniformity that we have revealed, we have
to explain in this case why simplification should carry prestige. The
trend to uniformity (not always necessarily simplification) that is often
found in so-called prestige accents (such as RP) would seem itself to
be related to functional factors. As Jakobson perceived long ago,
those varieties that have supra-local functions and that tend to
develop in the direction of koines display simpler phonemic systems
than varieties that have purely local functions (for an excellent
discussion of simplification and complexity in a range of language
situations, see Andersen, 1986). It may well be that, in varieties that
have supra-local functions, a high degree of complexity (at any level)
is indeed dysfunctional.
In the social status dimension, of course, the network model is also
explanatory, as different status levels in society are normally charac-
terized by different degrees of network density and multiplexity (we
shall discuss this more fully in chapter 7). At the lowest and highest
levels, network density and multiplexity is normally high: it is chiefly
in the middle ranges of society that social and geographical mobility
lead to the development of large numbers of relatively weak ties:
individuals at these levels have more uniplex and diffuse ties. At these
levels, the complex variation which encodes rich social meanings to
insiders simply becomes progressively more and more redundant: it is
no longer functional for these mobile people.
Finally, in the dimension of sex-differentiation, it would appear
from our studies of (a), (€) and the (pu//) alternant (in addition to
 Interpreting Variation in the Speech Community 109

many similar cases in the inner city) that females lead in the
development of supra-local norms (including those that involve
systemic simplification). This perception seems to come closer to
providing an explanation for sex-differentiation in language than
explanations based on prestige. In the following sections, we look
more generally at patterns of change in two Hiberno-English vowels.

4.6 Reviewing Evidence for Change in Progress in Two

Hiberno-English Vowels

In this section and the next I am concerned with tracking linguistic

change in the vowels /a/ and /e/ in Belfast English, relating these
changes both to historical/ geographical background and to network
structure. First, it is important to recall that in BV (as compared with
textbook descriptions of standard English) many vowels have a
startlingly wide range of realizations (see the range for /a/ in table
4.2). This results in overlapping: thus, some realizations of /¢/ are
like /a/, and vice versa, and this applies particularly to the short vowel
environments (following voiceless stop, following sonorant + voice-
less obstruent). For some groups of inner-city male speakers vowels
in words of the type met, went, are very consistently realized as [a, z],
and this often seems to be merged with /a/, as in sat, want. First, we
review the regional and social range of realizations of /e/.
Figure 4.4 shows the result of a quantitative analysis of /e/
realizations in two Belfast outer-city communities (Andersonstown
and Braniel) and a smaller town (Lurgan) situated 17 miles south-
west of Belfast. The symbol T indicates a following voiceless stop or
sonorant + voiceless stop cluster; CS indicates that the vowel is in
‘the stressed syllable of a polysyllabic word (this environment tends to
favour short realizations); D indicates following fricative or voiced
consonant (excluding /r/). Notice that the lowest short realization,
[a], is not favoured, but that in Lurgan short and low realizations in
short environments (T, CS) are more favoured than elsewhere (see
also below), and that long realizations [z:, ©:] in these short
environments are rarer in Lurgan. The inner-city figures (Ballyma-
carrett, Clonard, Hammer) in table 4.9 clearly show some contrasts
with the outer-city figures. Before voiceless stops, a low short
realization ([a], [z]) is categorical for many male speakers, while the
110 Interpreting Variation in the Speech Community

% 10- Andersonstown (n = 1104)

60+

50}-

EY
TC$D TC$SD. TCED -TC$O,. TesD. -TCSD
a ee Ze: € i

Braniel (n = 800)

VY
a7 zA
10

gual A pot
TC$D C$D TC$D C$D TC$D TC$D
a ee e: € € e

% 70- Lurgan (n = 1484)

40+

30+

20

10- g

BM os TC$D TC$D TC$D TC$D

:
TC$D TC$D
y
a e C=2} € €: e

Figure 4.4 Percentage distribution of /e/ (bed, bet) variants by following

environment in outer-city Belfast (Andersonstown, the Braniel) and Lurgan.
(Adapted from Harris, 1985)
 Interpreting Variation in the Speech Community Lt

Table 4.9 Percentage low realizations of /e/ in typically ‘short’ phonetic

contexts in three inner-city Belfast communities,
Ballymacarrett (B), the Clonard (C) and the Hammer (H)

Men 40-55 Women 40-55 Men 18-25 Women 15-25

‘odes 100 68 100 56

er 97 81 84 igs
H 97 75 98 67
CS B 73 56 78 50
C 81 67 75 60
H 76 68 76 ae

women more often prefer higher and often lengthened realizations.

Thus, for typically low vowel environments, as in wet, went, females
often have [we:t, we:nt] for ‘vernacular’ [wat, want]. In this respect
the inner-city female pattern is similar to that found generally in
outer-city communities.
These variable data give us a basis for examining processes of
change, since they suggest initially that either the higher or lower
variants are innovatory, or — more properly —that the direction of
change is either raising or lowering of /e/.
In fact, an examinations of historical documentation (real-time
evidence) suggests that the direction of change is towards raising.
Moreover, it appears that mid realizations are gradually appearing in
environments (such as pre-voiceless stop) where low realizations were
once the norm. It also appears that as the low variants are replaced by
higher ones, the relevant vowels are lengthened and sometimes
diphthongized: thus, as the rules are applied, conservative variants
‘such as [rant, rent]: ‘rent’, are replaced by [rent] (raising and
lengthening) and [re-ont] (diphthongization). The options open to
speakers for the realization of /e/ before voiceless stop or before
consonant + voiceless stop may be described as follows:

Choose either mid or low;

If low, realize as short;
If mid, realize as long;
St
ie If mid-long, realize as monophthong or diphthong.
jz Interpreting Variation in the Speech Community

This is of course an idealized and simplified account, and the aim of

listing such options is descriptive only. Nor is there any implied claim
that all individual speakers have the same rules or rule-order— far
from it. Accepting this as a broad description of the current state, we
now examine some real-time data in order to confirm the direction of
change.
Patterson gives a list of five words of the /e/ class, which were then
(1860) pronounced in Belfast with low realizations: wren, wrestle,
wretch, grenadier, desk. These few examples are enough to show that
the low realization was then more widespread than today: mren and
desk do not satisfy the voiceless stop or sonorant + voiceless stop
condition in monosyllables, and are now categorical [e:] or [€-9]
environments. Even the disyllable wrestle is unlikely to appear with [a],
as the rule for raising and lengthening before [-s] now almost always
overrides the tendency to lower and shorten in disyllables and
polysyllables. Items like wretch and grenadier are now variable. Staples
(1898) and Williams (1903) additionally give quite detailed descrip-
tions of the vowel in the city, which allow us to infer that low variants
had a much wider distribution then than they do today. The complete
list taken from those early writers allows us to see that the low vowel
appeared in environments where it would not appear now — for
example, before voiceless fricatives and voiced stops. The distribution
in present-day Belfast is quite different, as is shown by table 4.9 and
figure 3.3. In conservative working-class speech, low variants are
maintained in ‘short’ environments, very much as in the nineteenth
century: but low realizations have been almost entirely replaced in
long environments by mid realizations of /e/. More ‘prestigious’ and
less conservative speakers are less likely to use ‘low’ realizations, even
in short environments.
It is evident that over the last hundred years or so mid realizations
have been spreading at the expense of low realizations. Mid /e/ has
now almost totally replaced low /e/ in ‘long’ contexts (pre-voiced
stop, pre-sonorant + voiced stop, and pre-fricative). Low-status
inner-city speakers (males) sometimes still have categorically low
realizations in short environments, but in the more progressive
outer-city housing estates the vowel is now categorically mid for some
speakers. Interestingly, the distribution of variants in Lurgan is more
similar to that of the inner-city areas than that of the outer areas (a
pattern that applies also to other vowel and consonant variables). This
relatively rapid linguistic change in Belfast has accompanied its rise in
 Interpreting Variation in the Speech Community 113

population from about 120,000 in 1860 to nearly half a million in the

early years of this century, and Belfast may be taken as an exemplar of
linguistic change in fast-growing communities (while rural towns and
villages adhere to older patterns). The characteristic network struc-
tures of these different types of community are also relevant to the
manner in which change may come about, in so far as urban growth
tends at first to weaken strong pre-existing rural networks.
We may supplement our observations on /e/ by considering
evidence from present-day Ulster dialects. These are divided into two
distinct types. Ulster Scots dialects are found in east Ulster in a belt
extending from around Coleraine in the north, through most of
County Antrim and much of County Down (which is south of
Belfast — see figure 4.5). Most of Ulster to the west of this belt is
English-based or mixed Scots-English. Present-day Belfast dialect is
often described as an intrusion of this Mid-Ulster type into the
Scottish eastern belt. Now, the long mid variants of /e/ are
overwhelmingly associated with present-day Ulster Scots dialects
(Gregg, 1972) and are characteristic of modern central Scots dialects
generally. Traditional Mid-Ulster English, on the other hand, is
characterized by lower realizations in all environments. The pattern
of distribution in these dialects is remarkably similar to that of
nineteenth-century Belfast vernacular as described in Patterson,
Staples and Williams. We may infer that this pattern is a residue of
some earlier English vowel pattern that has not been well identified or
described by historical linguists. There is sixteenth-century orthogra-
phic evidence, which we shall further discuss in chapter 5, that
suggests some distribution of low vowel realizations for /e/ in
London English of the period: it seems possible that this pattern of
lowering of historic short vowels has been overtaken in recent
standard English and Central Scots by a pattern of raising and (in the
latter case) lengthening. The Mid-Ulster dialects may therefore have
preserved to a great extent an older general English vowel pattern,
and they may help us to project knowledge of the present on the past.
The historical and geographical evidence then both suggest that
the low realizations of /e/ (conservative English in background) are
giving way in a linguistically ordered way to the long mid realizations
characteristic of present-day Scots. It is clear that this change is
highly evaluated in Belfast in terms of social class hierarchy and
status, as it is the more prestigious groups that tend to adopt it and the
more ‘advanced’ (generally female and younger) group who introduce
114 Interpreting Variation in the Speech Community

[ Rathlin

North
Channel

Figure 4.5 The ‘core’ Ulster Scots areas of north-east Ulster (shaded
areas). (Adapted from Gregg, 1972)

it to the conservative inner-city communities (which are characterized

by dense and multiplex network ties that tend to resist innovation and
maintain conservative forms). The tension between innovative and
conservative social mechanisms gives rise to a identifiable pattern of
gradual diffusion, which may be represented as a historical shift from
an older English-type pattern towards a pattern characteristic of
modern Scots. As we have implied, the manner in which the change
proceeds is conditioned by both social and phonological factors. We
now turn to a description of change in /a/, with which the /e/ system
can be compared.
 Interpreting Variation in the Speech Community 115

As indicated in table 4.2, the range of realizations of /a/ in

present-day Belfast vernacular is considerable — from [e] through [a]
and [a] to back raised and rounded [9]. Again, as for /e/, patterns of
lengthening and diphthongization are present, with long vowels being
associated mainly with back realizations and with the higher front
realizations before voiced velars (see table 4.2). In what follows, we
are concerned only with backing and retraction, and we therefore
largely exclude the pre-velar environments (in which backing is not
found).
Table 4.2 also shows that back realizations are favoured by
following fricatives, non-velar voiced stops and non-velar nasals.
Nasals favour backing particularly strongly. Middle-class urban
speakers, as we have seen, tend to narrow the extreme range
described above. The widest range is found mainly in the speech of
inner-city male speakers. Furthermore, it is the males of Ballymacar-
rett (East Belfast) who use the backed variants most and who show
evidence of spreading the backed realizations into voiceless stop
environments (as in that, wrap), where short, front variants are
expected. If there is evidence of change in progress towards backed
varaints of /a/, it will therefore be male speakers who are leading it,
rather than the females who lead the change towards raised /e/.
Historical documentation suggests that /a/ backing is a recent
trend. Patterson does not comment on /a/ backing at all. On the
contrary, his remarks suggest that the Ulster tendency was towards
fronting and raising and that the most salient Belfast feature was
fronting and raising in velar environments (1860: 15).

In some places [presumably in the north of Ireland: JM] the short sound
of e¢ is improperly substituted for a, in almost every word in which it
occurs; in Belfast, however, this error is almost exclusively confined to
those words in which a is preceded by c or g, or followed by the sound of k,
hard g or ng.

A very few of Patterson’s spellings may indicate that /a/ backing and
rounding had been observed sporadically in -r and -/ environments:
he has form for ‘farm’ and canaul for ‘canal’. However, examples of
this kind are so few that they indicate only a slight tendency (possibly
confined to some pre-sonorant environments), which is not enough
for /a/ backing to be discussed as a stereotype. The item car appears
in Patterson as ‘care’, in which the now highly recessive rule for
116 Interpreting Variation in the Speech Community

fronting and raising after velars is clear. Items like hand, band, in
which [9] is now stereotypically expected, are given simply as han,
ban, etc. Frequently, however, items that now have low and/or back
vowels, are given with [e]: these include rether for ‘rather’ (a rural
Scots residue), e for a in single nasal environments in polysyllables
such as exemine, Jenuary and in nasal cluster environments such as
demsel, exemple, Entrim (‘Antrim’), slent, bendy (‘bandy’), brench.
Whereas Patterson’s account indicates a system generally inclined
towards front-vowel realizations, Staples (1898), writing nearly 40
years later, reports a ‘low back wide’ vowel before non-velar nasals,
for example in man, hand, land. Since Patterson’s time — /r/ environ-
ments have become categorically back realizations. Otherwise, the
figures on present-day variations strongly suggest that since then it is
nasal environments that have subsequently led the change, closely
followed by fricative and voiced stop environments.
Thus, although raising and lengthening of /e/ and backing of /a/
are both changes associated with modern Central Scots, the former is
at present led in Belfast by females and the latter by males. It is clear
from patterns of stylistic variation that (as we might already have
inferred) the two changes have different social evaluations. As table
4.10 indicates, the backing of /a/ tends to be resisted by speakers in
careful ‘interview’ style (whereas raising of /e/ is more likely in careful

Table 4.10 Incidence of retraction and backing of /a/ by age, sex and
conversational style in two Belfast communities, calculated by
an index score ranging from 0 (minimum) to 4 (maximum).
IS, interview style; SS, spontaneous style.

East Belfast (Ballymacarrett)

Men (40-55) — Women (40-55) | Men (18-25) | Women (18-25)

IS 3.03 1.75 2.89 1.89

oS) 3.58 2.58 ae) 2.10

West Belfast (Clonard)

Men (40-55) — Women (40-55) — Men (18-25) | Women (18-25)

IS 2.79 ih 2.36 236

SS 2.79 1.85 233 2.61
 Interpreting Variation in the Speech Community 117

styles). Thus, men seem to be principally associated with a change

that speakers do not consciously evaluate highly, while women are
associated with one adopted by speakers in their more carefully
monitored styles. We shall return to the table 4.10 pattern in chapter
6.
Our real time evidence confirms that the movement in /a/ is
phonetically from front to back. This means that sporadic front-
raising (found mainly in West Belfast) in such words as flat, trap ([flet,
trep]) must be seen as residues and not as innovations. The fronting
and raising rule in Belfast vernacular is virtually confined to velar
environments and cannot apply to such words as bad, hand, stab
(which are front in RP). The diachronic evidence shows that, for a
century or more, the trend has been towards retraction and backing.
The evidence also indicates that the rule for backing diffuses
geographically from East to West Belfast (see table 4.10). Scores for
/a/ backing are higher for East Belfast males than for any other
groups studied, and the range of environments in which backing
operates is extended to voiceless stops amongst younger East Belfast
males. It appears to be inner East Belfast (Ballymacarrett) that
provides the model for working-class speech in the city (L. Milroy,
(1980), 1987); this is represented by the (relatively) fully employed
Protestant population of East Belfast.
Both /a/ backing and /¢/ raising are relatively recent phenomena in
Belfast (but see below), and both are associated with a background in
Scots. Patterson’s account of Belfast shows characteristics of conser-
vative rural Scots lexical distribution, much of which appears to have
been residual and is now obliterated by restructuring. However, leng-
thening and raising of /e/ and /a/-backing are modern Scots. Gregg’s
(1972) account of Ulster Scots gives overwhelmingly back realizations
of /a/ and describes /¢/ as often long in realization (contrast the very
short low realizations in conservative Belfast vernacular, such as [stap,
dzat] for step, jet). Similarly, /a/ backing seems to be a very general
modern Scots feature (Lass, 1976). East Belfast adjoins the Ulster-
' Scots region of North Down (where backing is strong), whereas West
Belfast points south-west down the Lagan Valley, the speech of which
is Mid-Ulster with less Scots influence; furthermore, immigration to
West Belfast is recent and is largely from a Mid- and West-Ulster
non-Scots hinterland. Present day quantitative studies in Lurgan, a
small country town south-west of Belfast in the Lagan Valley, confirm
the existence of an /a/ system with little backing (front vowels have
118 Interpreting Variation in the Speech Community

been noted in that area even before [r] and finally), which is quite
similar to Patterson’s (1860) account of Belfast in this respect (Pitts,
1982).

4.7 Social Network Structure and Speaker Innovation: an

Analysis of /a/ and /e/

In addition to the variables of age, sex and status, a further social

variable associated with a speaker’s degree of integration into his
close-knit community appears to affect the probability of his being
linguistically innovative with respect to choice of vowel variants. This
variable is social network.
A major point emerging from our analysis of language/network
relationships is that the variable network needs to be considered in
relation to the variable sex of speaker. Indeed, as Gumperz has
remarked (1982: 71), the network variable is in general closely
associated with many others, including generation cohort, geogra-
phical location, and social status. Thus, our next task here it to pick
out briefly the relevant parts of our analysis of the social distribution
of innovatory realizations of /a/ and /e/, as identified in section 4.6.
First of all, realizations of /a/ and /e/ are strongly affected by the
variable sex of speaker. Thus, although incoming variants of both
vowels appear to have originated in the same hinterland Scots dialect,
each has assumed a diametrically opposed social value in its new
urban setting.
Raised variants of /e/ are, in the low-status inner city, associated
particularly with women and with careful speech styles. They are also
associated generally with outer-city speech, and data collected by
survey methods confirm that the higher the status of the speaker, the
more likely he is to use raised variants (see J. Milroy et al., 1983).
Different levels of use according to sex of speaker are particularly
evident in Ballymacarrett, where it appears to be younger female
speakers who are most strongly associated with the incoming raised
variants.
The incoming variants of /a/ show an almost perfectly converse
pattern of social distribution. High levels of backing are associated
with males (particularly Ballymacarrett males, although levels in other
inner-city areas are still quite high) and with casual styles appropriate
 Interpreting Variation in the Speech Community 119

to interaction between areas. The most extremely backed variants do

not appear at all in outer-city speech. Interestingly, the sex-
differentiation pattern across the three inner-city areas is not as
consistent for /a/ as it is for /¢/; there is some indication that the
young Clonard women are increasing their use of back realizations
when compared with other female groups (see table 4.10). They also
use these variants more than their male counterparts, although they
follow the expected sex-differentiation patterns with respect to other
phonological variables (see chapter 6 for a discussion of the Clonard
pattern).
In summary, then, it appears that incoming variants of /a/ are
associated with core Belfast vernacular, while incoming variants of
/e/ are associated with careful higher-status speech.
If we look at the relationship between speaker-choice of variant and
individual network structure, the picture becomes even more compli-
cated. With respect to both vowels, choice of variant shows a
correlation with personal network structure in some subsections of
the inner-city communities; but the details of this correlation are
quite different for each vowel.
The vowel /a/ is particularly sensitive to variation according to the
network structure of the speaker; but women appear to correlate their
choice of variant more closely with their personal network structure
than do men. This means that among women a relatively large
amount of /a/ backing is more likely to be associated with a high level
of integration into the network than is the case among men-—a
relationship analysed by Spearman’s Rank Order Correlation (L.
Milroy, 1987: 155). Although, as we have noted, women are much
less likely than men to select back variants of /a/, this generally lower
level of use does not prevent individual women from varying their
realization of /a/, within the female norms, according to their social
network structure. Thus, the degree offit between phonological choice
and network structure may be seen as an issue quite separate from the
absolute level of use of a particular range of variants. We may thus argue
‘that /a/ functions for women as a network marker to a greater extent
than it does for men; by this we mean that there is for them a higher
correlation between choice of variant and network structure, a
tendency to select relatively backed variants being associated with
higher levels of integration into the community.
When we look at the relationship between choice of /e/ realization
and individual social network structure, we find a pattern emerging
120 Interpreting Variation in the Speech Community

converse to the one described for /a/; recall also that the incoming
variants of the two vowels showed an almost converse distribution
with regard to status, sex of speaker and speech styles.
Most importantly, there appears to be no tendency at all for women
to use /e/ as a network marker in the sense described above; but
there is a significant correlation between network scores of male
speakers (particularly young male speakers) and choice of /e/
realization. A tendency to select relatively low (conservative) variants
is associated with a relatively high level of integration into the
community.
This complex relationship between network structure, sex of
speaker and language use is summarized in table 4.11. However, our
interest here is in a generalization which we are now able to make
concerning on the one hand the relationship between language and
network structure, and on the other the social identity of the
innovating group. In the case of both /e/ and /a/ it is the persons for
whom the vowel has less significance as a network marker who seem
to be leading the linguistic change. It is as if absence of this
language/network relationship (a relationship that fulfils a cohesive
social function) enables a particular social group to adopt the role of
linguistic innovators. This appears to be the case regardless of
whether the innovation is evaluated by the wider urban community as
being of high or of low status. For although it is clear that /e/ raising
is diffusing on a much broader social front than /a/ backing, the
generalization still seems to hold true that it is those persons in the
inner city for whom the vowel functions less clearly as a network
marker who are the principal innovators in their own communities.

Table 4.11 Contrasting patterns of distribution of two vowels involved in

change, according to sex of speaker, relative frequency of
innovatory variants and level of correlation with network
strength.

High correlation with

Change led by network strength

/a/ Males Females

VAS Females Males
 Interpreting Variation in the Speech Community 121

It is important to note that even though back variants of /a/ are

strongly emblematic of vernacular speech, they are nevertheless
spreading to higher-status groups in the wider community. But this
diffusion is being implemented in a manner very different from that
affecting /e/. We have noted that [e] raising is characteristic both of
low-status female speech and more generally of higher-status speech.
The diffusion of [e] raising on this wide social front is confirmed both
by linguistic survey data and by more detailed outer-city community
studies.
We noticed in section 4.4 that, by way of contrast, higher-status
Belfast speakers avoid both extreme front and extreme back realiza-
tions, as they converge around cardinal vowel 4 in the middle of the
phonetic range. However, a very interesting group of young, male,
middle-class speakers can be identified in the sample of speakers
studied in the survey. They also show the characteristic middle-class
tendency to converge around a limited phonetic area, with relatively
little conditioned variation. However, phonetically, the point at which
they converge is further back than that characteristic of older
middle-class speakers.
It appears, therefore, that the mechanism of diffusion associated
with each of the vowels is different. Raised variants of /e/ are
apparently spreading in a linguistically ordered way, with ‘long’
environments affected first. For many outer-city and middle-class
speakers, a raised vowel is already categorical in all environments.
Although back variants of /a/ appear to be diffusing historically and
laterally (through the low-status inner-city communities) in a lin-
guistically order manner parallel to the processes affecting /e/, the
mechanism of diffusion upwards (socially) through the community is
quite different. What seems to be involved here is a ‘drift’ phon-
etically to the back of the characteristic middle-class realization.
The data presented here suggest that social network structure is
implicated in processes of linguistic change in at least two ways. First,
a strong close-knit network may be seen to function as a conservative
force, resisting pressures to change from outside the network. ‘Those
speakers whose ties are weakest are those who approximate least
closely to vernacular norms, and are most exposed to pressures for
change originating from outside the network.
Second, a detailed sociolinguistic analysis of [€] raising and [a]
backing— processes which have a common dialectal point of origin
122 Interpreting Variation in the Speech Community

but have taken on very different social values in their new urban
context —suggests that the vernacular speakers associated most
strongly with the innovation are in each case those for whom the
vowel functions least prominently as a network marker. It is as if a
strong relationship between the network structure of a given group
and choice of phonetic realization of a particular vowel disqualifies
that group from fulfilling the role of innovators with respect to that
vowel. Conversely, it may be the case that dissolution of the
language/network relationship with respect to a group of speakers is a
necessary condition for that group to fulfil the role of linguistic
innovators.

4.8 Concluding Comments

We shall return to the question of linguistic innovation in chapter 6.

In this chapter we have been concerned with interpreting patterns of
variation in speech communities with reference to the norms that can
be shown to exist at varying levels of abstraction and generality. We
have also been concerned with locating patterns of change in relation
to the social variables of sex and network, in particular.
In this discussion, however, we have found it necessary to turn to
‘real-time’ evidence (Patterson, 1860; Staples, 1898) in order to help
us to locate the direction of change. It is this time-depth that will
interest us in chapter 5, in which we turn to evidence for vernacular
variation in earlier English.
 5
On the Time-depth of Variability
in English

5.0 Introduction

My purpose in this chapter is to extend the scope of the argumenta-

tion developed in chapter 4 by projecting a variationist view on to past
states of language change and variation. The variationist account
developed so far is built on the axiom that language is variable at all
times. The potential for change is therefore always present in
variation, and may appear as a progressively greater or lesser
favouring by the speech community of particular linguistic variants
from among the variants that are available in the community at some
particular time: to that extent change can be said to consist of change
in community norms. In this chapter I shall be concerned with the
description and interpretation of past states of language, bearing in
mind that these were variable states, and with ‘the use of the present
to explain the past’.
However, although I am proposing a variationist view of language
change, I have to deal with this in a context where the dominant
tendency has been to focus on uniform states (Weinreich, Labov and
Herzog, 1968). In practice, the kind of uniform state that historical
scholars have had in mind has frequently been the uniform standard
language; for this reason it is appropriate to start with some
comments on the tendency to envisage the history of English as a
unilinear history of the standard language.
124 On the Time-depth of Variability in English

5.1 The Consequences of Standardization: 1

Projecting Backwards

We noted in chapter 2 that traditional accounts of language history

from around 1550 have tended to be unidimensional and convergent.
We also noted, in chapter 3, that it can be misleading to apply the
norms of mainstream and centralized varieties to the description of
divergent language states. Thus, just as it is dangerous to superim-
pose standard-based analyses on a present-day vernacular, so it must
be at least equally dangerous to do this in dealing with past states,
which are of course also divergent. Indeed, the limitations of
historical databases (noted in chapter 2) would seem to make it
considerably more dangerous.
Despite this, standard-based interpretations have been used so
frequently in diachronic description that they have greatly affected
the conceptualization of language history that we inherit from older
generations of scholars. These scholars tended — seemingly without
giving much thought to the matter — to accept the standard ideology
and the doctrine of uniformity that is associated with it; indeed, as we
have noticed in chapter 3, they were also influenced by ideological
positions which they believed to be ‘common sense’. To H. C. Wyld,
for example —surely the best historian of English —it probably
seemed to be common sense to believe that RP was the most
important accent of Modern English, and that a-historical account of
English pronunciation should therefore be, in effect, a history of RP.
But only in certain areas of historical linguistic research is it
appropriate to use standardization and ‘the standard language’ as our
main reference point and focus of interest, and sound-change is not
one of these areas. It is justified to use the standard language in this
way if our primary interest is in those developments that themselves
largely depend on the standard ideology (J. Milroy and L. Milroy,
1985a), for example, elaboration of vocabulary and certain develop-
ments in clausal syntax, because literacy—and hence standardi-
zation — are plainly involved in these, and their role can be investi-
gated. Indeed it is possible that we do not understand these things
very well unless standardization is taken into account. But it is a
different story when we focus on phonological change. As Dr Johnson
observed: ‘sounds are too volatile and subtile for legal restraints’
 On the Time-depth of Variability in English rH
(Bolton, 1966: 152), and the history of English phonology is not a
history of the standard at all, but a history of vernaculars. Thus, the
most general consequence of concentrating on standard English here
is that a multidimensional history of phonology is made to appear as
unidimensional — it becomes ‘a single-minded march’ towards RP
and standard English (Lass, 1976, xi).
This unidimensionality is imposed on history by a backward
projection of present-day standard phonology on to the past, and
according to the theory of language standardization that we have
tentatively advanced elsewhere (J. Milroy and L. Milroy, 1985a), it
can be seen as an attempt to historicize the standard language — to
create a past for it and determine a canon, in which canonical forms
are argued for and unorthodox forms rejected. This in itself is part of
the process of J/egitimization of the standard language, in which the
accepted norms are determined, not primarily by consensus amongst
speakers, but by legislation. But the tendency to project backwards in
a single (standard-based) dimension can be seen in the work of many
scholars, and it has two consequences that are of interest here.
The first of these consequences relates to our knowledge of the
structural forms of language in older stages of English. The linguistic
forms that are recorded in handbook accounts of change are, for the
most part, those that lead to present RP, and not usually those that
lead to other modern accents of English; as we shall notice below, the
latter forms are often explicitly rejected from the historical account.
Furthermore, when dates are suggested for changes in these forms,
they are usually given, without comment, as the dates at which the
changes took place in this standard variety or its unilinear precursor,
and not the dates at which they might have taken place in some other
variety. Despite this, however, they are then usually cited as the dates
at which the changes took place ‘in English’. Let us consider a simple
example.
In many historical accounts, it is stated that late ME [a] developed
_ into [a] in the early seventeenth century (it does not matter here what
date is cited; what is important is simply that a relatively exact date is
cited). However, it appears that this change must relate to certain
dialects of Southern British English at that date, and not necessarily
to other dialects. After the given date, the same change may well have
diffused to other dialects that were not affected at first, and Labov’s
well-known work (for example, 1980) on short /a/ in the northern
United States suggests that this EModE change is still in progress
126 On the Time-depth of Variability in English

there today. So —strictly speaking —it is not correct to say that it

‘happened’ in the seventeenth century and to leave it at that. This
caution is all the more necessary when we realize that there are
varieties in which this change has not yet taken place, more than three
centuries later, and that there are yet other varieties in which different
changes have taken place. Lowland Scots and Ulster Scots dialects,
for example, have had backing of [a], and we have to work out from
residues that in some ancestral forms of these, there may also have
been front-raising (Lass, 1976; J. Milroy, 1981; and see chapter 4.6
and 4.7, above, and section 5.7, below).
Thus, in general, it seems that backward projection on to historical
states of language has tended to be based on present-day standard
English and SBE, rather than other dialects, and as in the examples
cited from Dobson (below) there has been a tendency to think of a
phonemic set (such as short /a/) as being invariant or nearly invariant
within itself. But what is important here is that where allophonic
variation in a phoneme class is discussed in the main handbooks and
histories, this is usually variation that leads to a present-day charac-
teristic of the standard variety. Typically, the changes discussed
include such examples as the lengthening and backing process that
led to ‘broad’ [a] in the RP class of dance, path and the rounding after
[w] that led in mainstream accents to present-day wasp, swan (many
British English dialects do not have either the ‘broad’ [a] or rounding
after [w]). The moral of all this, of course, is that, as language is
variable at all times, there is something unsatisfactory about applying
a unilinear and uniform-state style of interpretation to an account of
language history. This seems to suggest that EModE was consciously
directing itself towards modern RP — in a state of ‘becoming’ RP, so
to speak.
The second consequence of this unilinearity is that the idea of the
standard is projected backwards on to states of language and society
in which that idea may not have existed, or — if it did exist — may have
been different in important ways from the idea of the standard as it
exists today. We can demonstrate this backward projection of the idea
of the standard by referring to the influential work of E. J.Dobson on
EModE pronunciation.
Dobson (1955, 1968) continues the tradition of his predecessor,
Wyld, in focusing very closely on the history of ‘standard’ English. It
is pronunciation that Dobson is interested in, and the EModE
phonological standard is referred to very frequently in terms that
 On the Time-depth of Variability in English 127
suggest that such an entity had some real existence at that time.
Sometimes there are several references to the standard on one page,
for example: ‘ME a had become [z] in...less careful Standard
English in the late sixteenth century . . . In the sixteenth and seven-
teenth centuries there were two pronunciations of ME a in use in
Standard English’ (1968 II: 548). These comments do not seem to be
based primarily on the notion of standardization as a process (J. Milroy
and L. Milroy, 1985a) that might have been beginning to have an
effect about this time: they present the standard language as a
coherent entity —a variety, like any other variety. But they do make
sociolinguistic assumptions: in context it is clear that they also assume
the early development of a socially elite variety, and we can see from
the first sentence of the quotation that ‘carefulness’ is probably also
involved. But, although these concepts (standardness, eliteness,
carefulness) are of different orders (and we shall return to these
distinctions), Dobson does not apparently see any reason to keep
them separate.
What is more interesting here, however, is the method of reasoning
by which Dobson reaches his conclusions on how EModE pronun-
ciation is to be codified and legitimized as part of the canon. It is wise
to bear in mind that similar reasoning has been used by many
others — so much so that much of our supposed ‘knowledge’ of the
history of English phonology is filtered through this kind of reason-
ing. Essentially, any attested forms that can be characterized as
‘vulgar’ or ‘dialectal’ are rejected, much as the apocryphal books of
the Bible are rejected. For example, Dobson notes that one source
(Thomas Pery) ‘shows the vulgar raising of ME a to [e]’. This,
according to Dobson, is not surprising because Pery’s speech ‘was
clearly Cockney . . . The evidence of such a writer does not relate to
educated StE’ (1968 II: 551). On the same page, evidence from the
Paston Letters is rejected as ‘dialectal’, and the Diary ofHenry Machyn
is rejected on the grounds that Machyn was a Yorkshireman. Dobson
‘seems to be quite sure of this, but Machyn was almost certainly a
Londoner, and his Diary is one of the most valuable primary sources
for the study of EModE London pronunciation. But as he was not an
upper-class Londoner, we can guess that his testimony would have
been rejected in any case.
Perhaps the most startling piece of reasoning occurs in Dobson’s
long argument against accepting early ‘occasional spelling’ evidence
for raising of [a, x] to [e]. It goes as follows (1968 II: 549):
128 On the Time-depth of Variability in English

The most important objection, and it alone is a decisive one, must be that
no Englishman could conceivably use ¢ as a means of representing [z]. It
may seem natural to a foreign scholar [presumably Zachrisson: JM] to
suppose that the sound [x]... might be spelt ¢; but it is little short of
incredible that native English-speaking scholars [presumably Wyld: JM]
should have accepted this view. No English-speaking child learning to
spell... would write ket for [ket]; the distinction between [z] and [e] is
an absolute one for him (since otherwise he could not distinguish, for
example, man from men)...

But clearly this is not decisive, as no evidence is given to prove the

negative — that English speakers do not ‘confuse’ [e] with [z, a]. In
fact, native speakers do overlap here; they do have differential
phonological rules controlling the distribution of /a/ and /e/, and
they do make spelling mistakes involving a and e. Furthermore, it is
likely that the allophonic distribution of these sounds in EModE was
quite different from the RP distribution today, and that the doctrine
of (retrospective) phonemic purity (as displayed in the remark about
man and men) is therefore inappropriate (supporting examples from
the Belfast projects are cited in chapter 4.6 and below). Moreover, it
is also certain that English-speaking scholars do not have reliable
retrospective intuitions about EModE pronunciation any more than
they have reliable intuitions about divergent states at the present day.
There is no reason to think that their intuitions about /a/ are
necessarily better than those of foreign scholars. So this elaborate
argument has no reliable foundation: in short, it is wrong.
But it is of great interest, because it demonstrates a familiar mode
of argumentation that has been more widely used in historical
descriptions and that can be found in many places, and I shall
comment on a similar case in ME below. The reasoning is not aimed
at supporting a positive argument: it is negative and adversarial, and
aimed at rejecting the arguments of other scholars and at excluding
evidence that cannot easily be accommodated into a unilinear
historical canon. The underlying idea is simply that some forms are
‘genuine’ and some are not. Ultimately, it does not matter very much
what arguments are used to reject evidence: the point is that all
possible post hoc arguments must be marshalled to defend one’s
position against other scholars: the more arguments that can be used,
the better. In this case the immediate position defended (contra the
work of Wyld and Zachrisson) is that orthoepic evidence is better
 On the Time-depth of Variability in English 129
than other evidence, such as ‘occasional spellings’ (Wyld, 1936), but
the underlying assumption (which is more relevant here) is that it is
possible to write a continuous unilinear history of ‘standard’ English
pronunciation.
It is certainly true that some of the processes associated with
standardization (on which see Haugen, 1972; Leith, 1980; J. Milroy
and L. Milroy, 1985a) were well under way around 1600 (elaboration
of function, use of a supra-regional writing system), but it looks as if
some of them were still at the stage of being localized developments
associated with the establishment of consensus on local norms, and
this applies particularly to pronunciation, which is the level of
language that is least uniform. In the phonetic/phonological dimen-
sion, I do not think we should speak too readily of Early Modern
‘Standard’ English. Perhaps I can make this clear by distinguishing
tentatively between the notion of prestige and the notion of standard
language.

5.2 Prestige Norms and Standard Norms

A standard variety or form is not conceptually the same thing as a

prestige variety or form (on prestige, see J. Milroy, 1989): the main
linguistic symptom of standardization is invariance. Standardization
comes about for functional reasons, and its effect is to make a
language serviceable for communicating decontextualized
information-bearing messages over long distances and periods of
time. In this respect its functions differ greatly from those of
vernaculars (recall our discussion of functions in chapter 2). It is
imposed through its use in administrative functions by those who
have political power. Once it spreads into other functions, it acquires
what we usually call ‘prestige’, in the sense that those who wish to
-advance in life consider it to be in their interests to use standard-like
forms. Prestige, however, is a different concept altogether, as it can be
subjectively attached by speakers to forms and varieties which are very
distant from, and in conflict with, the codified norms of the standard.
In recent years, Arab linguists (Abdel Jawad, 1987; Alahdal, 1989)
have found it difficult to reconcile the concepts of ‘standard’ and
‘prestige’, because rural norms that are identical to standard Arabic
are shown to be dispreferred in favour of urban ‘non-standard’ forms.
130 On the Time-depth of Variability in English

Our own work and that of Trudgill (1986a) has tended towards very
similar conclusions. In this context, the well-known late sixteenth-
century comments about the best English being spoken in the
London area should be understood for what they are, and not
necessarily as a sign of standardized pronunciation. They arise from a
consciousness of the development of a stable (but variable) vernacular
norm in London, knowledge that its norms differ from those of other
varieties, and a belief in the superiority of London norms over those
of other regions. What we are dealing with here is sociolinguistic: the
development of supra-local regional and social attitudes to language.
What we are not dealing with is a fully-fledged ‘standard’ pronuncia-
tion that can then be most usefully described in terms of a unilinear
history since that time. If we do think in this unilinear way, we will be
inclined to dismiss or devalue evidence that is important for our
understanding of language history, just as some of our predecessors
have done.
In order to understand the history of so-called standard English
pronunciation, therefore, we need to assume — paradoxically — that
there is no such thing. Standardization is not primarily about varieties
of language, but about processes. Therefore, it must be treated as a
process with an underlying socio-political motivation, which attempts
to promote uniformity and suppress variability for reasons that are
considered functional. Similarly, the standard can be regarded as the
ultimate in the development of a supra-local norm of language. If we
take this point of view, it seems that we will gain a better understand-
ing of the history of what is usually called standard English, because
we will be able to separate out those issues that primarily involve the
development of standard (prescriptive or codified) norms from those
that do not. The history of pronunciation is one of those issues that
do not primarily involve standardization, because it is about the
history of speech in face-to-face interaction and because standardiza-
tion has always had less effect on pronunciation than on other
linguistic levels. The history of standard pronunciations is not
therefore something that we can usefully chart in a unilinear temporal
continuum, as though co-existing varieties had no role except to
‘feed’ the standard ‘variety’ from time to time. This is an over-
simplification — and a distortion — of language history.
Bearing these points in mind, the remainder of this chapter
consists chiefly of a discussion of backward projection of variability,
including case-studies from my own research. I shall start by
 On the Time-depth of Variability in English 131

considering the effects of the standard ideology on analyses of Middle

English and go on to review aspects of variation in Middle English.

5.3 The Consequences of Standardization: 2

Variation in Middle English

One of the advantages of studying Middle English is that its written

forms are highly variable. Before that time — in late Old English — the
West Saxon literary standard was well established, and after-
wards — from around 1550—- most printed documents appear in a
(relatively) standardized form. The diversity of Early ME, however, is
so great that—apart from the so-called AB Language—no two |
substantial literary documents are in exactly the same ‘dialect’. If we
compare two twelfth-century texts, such as the Peterborough contin-
uation of the Anglo-Saxon Chronicle (East Midland) and Layamon’s
Brut (South-west Midland), we could not be blamed for believing that
they are in different languages. The differences between them are of
the same order as the differences between modern Dutch and
standard German: the Brut retains the OE case-inflexions and
grammatical gender, whereas the Peterborough Chronicle, even
though it is an earlier document, has lost most of these. But not only
is there considerable divergence between different texts, there is also
normally great variability (particularly in spelling and inflexional
forms) within the texts. Thus, ME language states, being so variable,
should in principle be suited to the same kind of analysis that we use
in present-day social dialectology, and by using variationist methods
we should be able to explore at least some of the constraints on
variation that might have existed in ME. In a present-day community,
these constraints can be observed in spoken language: in ME we must
locate these constraints initially through the writing system.
We can briefly demonstrate this point by comparing a modern case
with a medieval case. Suppose we show that in a present-day
vernacular there is structured variation in verb-forms of the type he
does/he do (see Cheshire, 1982, for a relevant study), with one form
perhaps being preferred in formal styles and the other in casual styles:
we may also — by comparing the speech of different social groups and
age-groups — show that one form is progressing at the expense of the
other. As it happens, ME texts also frequently exhibit variation in
132 On the Time-depth ofVarsadeltty in Engitsh

verbal inflexions. Suppose, for example, a text (such as the Resa

thirteenth century, East Anglia) exhibits 3rd singular verd-form
variation in -es, -ep (such as stemdes, stumdep) and syncopated forms |
(such as stamé): this may indicate that the text is composite and has _
been copied by scribes from different dialect areas (for discussions of
these scribal questions, see especially McIntosh, Samuels and Bens-
kin, 1986; Laing, 1989; McIntosh and Wakelin, 1989). But & could
also be the case that all three forms (or perhaps twe ofthem) were
current in the underlying dialect of the sembe (er of the auther),
or — more properly — of the speech community to which he beleaged.
Indeed, as the writing system was not standardized, it 8 bkely that
variation of this kind would enter more readily inte the texts than R
would today, and that it may therefore be possidle by comparing texts
to trace the pattern by which one variant recedes and another spreads.
However, we have to bear in mind that, because of the complexity ef
scribal histories and the fact that the provenance (date and place af
writing) of the text is net usually divulged, the medievalist fees 2
number of difficulties that the present-day dialectolegist dees net
face (for discussion see McIntosh, Samuels and Benskin, 1986 k J
Milroy, forthcoming), but if these can be solved,
the general similarity
that I have suggested here is still valid.
Variability in ME, however, has sometimes been perceived as an
obstacle rather than a resource, and as a result information that might
be of some value to the historian has been rejected tram the canen. Ik
is interesting that although no ene suggests that there actually was 2
‘standard English’ in Early ME times, the doctrine ofuniformity has
nevertheless been applied to ME. In editorial and descripaive cam-
mentary, there are many comments about chaotic or ‘wless” spel-
lings (for example, Sisam, 1915: .xxxvii) and even editorial judgements
to the effect that a given scribe could not have been a native Exgiish
speaker, so variable is his spelling (Hall, 1920 I: 037). This st
judgement (although it is commonly made) is speculative, af course,
as the scribe is normally anonymous, but it 8 analogous t rejecting
live speakers trom a random sample on the grounds that they de nee
speak as we think they ought to speak, or rejecting attested spoken
forms on the grounds that they are not what we would expect im some
particular location. However, judgements of this kind can efftetively
block further investigation of variable ceastraints in the maw ie
question: they can be dismissed as ‘corrupt or ‘unreliable’ specimens
of language. One way in which variation of this hind is discounted &
 On the Time-depth of Variability in English 133

to claim that the scribe was Anglo-Norman, or that the spellings are
Anglo-Norman and therefore not valid evidence for the history of
English.
The Anglo-Norman argument goes back to Skeat (1897), who
specified particular features of spelling as Anglo-Norman. These are
discussed by J. Milroy (1983), and it is noticeable that many of these
features, such as » for wh (in words of the type what, which), have
reflexes in later English pronunciation. The comments above by
Sisam (1915) and Hall (1920) arise directly from Skeats’s views, and,
although these comments were made a long time ago, it would be a
mistake to think that the Anglo-Norman argument has now been
abandoned. Clark (1990) has recently referred to the ‘myth’ of the
Anglo-Norman scribe and has collected a large number of comments
from early in the century up until very recently from the work of
distinguished scholars, in which attested forms are typically said to be
‘Anglo-Norman’ and therefore rejected. As a result of this ‘myth’,
however, the very fact of variable spelling in an Early Middle English
document becomes in itsel/fa reason for concluding that the scribe was
Anglo-Norman and therefore that his spelling can be corrected by
editors and ignored by historical commentators and dialectologists.
Even the work of scribes writing centuries after the Conquest has
been dismissed in this way, seemingly mainly because it is variable,
and not because we can (usually) know whether the scribe was a
first-language speaker of Anglo-Norman, or whether it would have
been relevant if he had been.
Leaving aside this argument, however, we must also recognize that
scholars have sometimes been more generally influenced by the
notion that written language should be uniform, even in a period in
which it plainly was not uniform, and they sometimes appear to chide
the scribes for spelling variably. Scragg (1974: 26), for example,
comments that ‘the existence of regional orthographies, and their
confusion in the copying of texts resulted in a very lax attitude to
. spelling in most scribes.’ In the context, this ‘very lax attitude’ seems
to be measured against circumstances (such as late Old English or the
present day) in which there is a uniform standard of spelling: thus,
what this really means is that in Early ME there was no uniform
standard, and indeed Scragg adds that these scribes had ‘no concep-
tion of a spelling standard’. But he further comments — much more
dubiously — that they used ‘variant forms at will’. However, if the
scribes really had used variants ‘at will’, we would actually be unable
134 On the Time-depth of Variability in English

to read the texts, as there would be no system in the spelling; but

there must always be some order in any spelling system that we can
read, even if it is a variable system. Therefore, the scribes did not
spell ‘at will’, but according to variable (and historically mixed)
conventions. It is our task to attempt to specify the constraints on
spelling under which they were working, always admitting that even
after we have done this, there may well be residues of apparent
randomness that we cannot explain.

5.4 Orderly Variation in Spelling

The existence of variable orthographies is an advantage to the ME

dialectologist in exactly the same way that the existence of spoken
variation is an advantage in present-day research. Although the
scribes no doubt made ‘errors’, it should be possible to investigate
variable texts in extenso to determine the extent to which variation in
spelling (or indeed in other linguistic dimensions) is in fact orderly,
and whether this variation can help us to work out what might have
been happening in spoken English at the time. As an example, let us
briefly consider some aspects of spelling in Havelok the Dane.
The Havelok text (Bodleian MS Laud Misc 108) is one of those
sources that has been traditionally thought to be the work of an
Anglo-Norman scribe (Sisam, 1915) on the grounds that the spelling
is highly variable in the respects specified by Skeat and indeed in
some other respects also. However, although it doubtless contains
some forms that are simply ‘errors’, it also exhibits the kind of orderly
variation that could be captured within a variable rule framework, but
in spelling variation rather than phonology. The scribe does not have
a free hand with spelling variation: there are constraints on the
variants he uses. OE post-vocalic /ht, xt/, for example — in words of
the type rht, niht, ‘PresE ‘right, night’ — can be represented in the
spelling of Havelok by st, ht, th, cht, cth, but not by, for example gt, ght,
or by random and unpredictable forms such as tc or m. The variation
is constrained in much the same way as present-day phonological
variation in speech communities is observed to be constrained. For
example, we can specify the constraints on /a/ before velars in
inner-city Belfast by stating that it can be realized as a mid-front, low
mid-front, or low-front vowel, but not as a low-back vowel. Thus, just
 On the Time-depth of Variability in English 135
as present-day phonological variation can be used as a clue to change
in progress, so it may be possible here to use orthographic variation in
the same way.
The spelling variants for OE (ht) overlap with spelling variants for
other forms (from different sources in OE), just as phonological
variants in present-day studies are found to overlap. Thus, if we take
the realization th, we find that this can be used word-finally, not only
for (ht), but also for (t) and (th). The result of this is that a spelling
like with can realize three separate classes: OE wiht (‘wight, person’),
OE mip (‘with’), and OE hwit (‘white’), and this of course applies to
other items of these types. To formalize this: the following (OE)
classes can appear with final th:

1 Final (post-vocalic) dental fricatives:/0, 0/, for example with (OE

wip, PresE with);
2 Final (post-vocalic) dental stop: /t/, for example mith (OE hwit,
PresE white);
3 Final /ht, xt/: for example, with (OE wiht, PresE might).

The potential realizations of these three classes are, however, differ-

ent: (ht) items can also appear with st, cht, etc. (such as micht): the
other two classes cannot; (th) items can also appear with final p,0
(such as mip): the other two classes cannot; (t) items can appear with
final single t (such as wit, whit): the other two classes cannot. Thus,
‘with, wight’ cannot appear as wit, whereas ‘white’ can. To this extent,
therefore, the variation is constrained, and not random. Applying the
principle that change in progress is manifested in variation, let us
consider its possible implications for spoken variation in ME.
The study of (ht) in Havelok is of course relevant to the date at
which the velar fricative [x] before [t] (in right, might, etc.) was lost in
English. The prima facie conclusion to be drawn is that in the variable
phonology of the ‘underlying’ (East Midland/East Anglian) speech
community, loss of the fricative and merger of wight, white, or close
approximation and overlap, had already taken place. It is also possible
that in this variable phonology there was some tendency to merge
final/6/with/t/. If developments of this kind were not in some sense
in progress, then there would have been less likelihood of the scribe
observing precisely this pattern of orderly spelling variation, because,
given the variable state of the orthographic conventions known to
him, he could have chosen to vary in other ways. Of course, it is quite
136 On the Time-depth of Variability in English

another matter to go on to argue from this very limited piece of

evidence that loss of the fricative in /xt/ was embedded in the English
language as a whole as a completed sound-change at this early date.
Yet, if we take this together with the fact that many other forms
characteristic of modern English spread in these centuries from the
East Midlands and the North, we can advance the hypothesis that this
change was in progress in the East Midlands around 1300, and look
for further evidence to support or refute this. If, however, we insist
that the scribes were simply careless or poorly acquainted with
English, we shall be inclined (as many scholars have been) to reject
the evidence and date this sound-change much later — at a time when
it was actually completed in ‘standard’ English. This, of course, will not
bring us anywhere near the origin of the change. Furthermore,
bearing in mind the reservations about changes in different varieties
that I expressed in section 5.2 above, the early date suggested here for
the loss of the velar fricative does not affect the fact that there are
dialects of English (in Lowland Scotland) which have not yet lost the
fricative.
Loss of the velar fricative is a change that was finally adopted in
near-standard vernaculars and formal styles. ME sources, however,
also contain variation that may be relevant to non-standard varieties
and casual styles of speech; hence, there may be considerable
time-depth to these variables also. In Section 5.5, therefore, we
consider some non-standard examples.

5.5 The Time-depth of Present-day Non-standard Variants

A number of present-day non-standard and casual speech forms

appear to be indicated by some features of variable ME spelling.
Some of these are recognized as regional and have been studied as
such (for example in Wakelin and Barry, 1968, a study of the voicing
of initial fricatives in South-west England): others are more wide-
spread in English. One of these is ‘final stop deletion’ (loss of /t, d/,
and sometimes other stops, in final clusters in words such as mist,
mend). This is today very common in many varieties of English (Guy,
1980; Romaine, 1984a), but not common in careful styles of RP
(hence its exclusion from many accounts which claim to be accounts
of ‘English’). The maps of the Linguistic Atlas of Late Medieval English
 On the Time-depth of Variability in English 137

(McIntosh, Samuels and Benskin, 1986) show a distribution of final

consonant loss also in medieval written English, and I have noted a
number of examples in Havelok and other texts. Thus, the phenom-
enon may have been part of variability in English for many centu-
ries — more common perhaps in some dialects than in others, reced-
ing at some periods and progressing at others. Yet, it plays little part
in standard accounts of the history of English before about 1600, and
in ME stop-deleted forms (such as bes, lan: ‘best, land’) are amongst
the forms that are typically corrected by textual editors as errors.
There are other features that may have much earlier origins than is
generally believed. These include: 1) the (casual style) -in’ ending on
present participles (Houston, 1987; Labov, 1989); 2) certain wides-
pread socially or regionally marked alternations in modern English,
such as ‘stopping’ of dental fricatives in, for example thick, that, and
[h]-dropping. I shall now consider in some detail this last phenome-
non — variable loss of [h] in stressed syllables initially before vowels.
I selected this particular variable as a case-study for a number of
reasons. The chief one is that it is a very salient stereotype popularly
characterized as non-standard or ‘vulgar’, and is extremely wide-
spread in England and Wales. The map (figure 5.1) shows that [h] is
preserved in rural dialects only in the extreme north-east, and in parts
of East Anglia and the West Country; but, if anything, the map
probably understates the extent of [h]-loss. Most large urban areas
south of the River Tees have considerable [h]-loss in vernacular
speech, and this includes even Norwich in East Anglia -an area
shown by the map as, at least partly, [h]-ful. As Trudgill (1974) has
shown (see figure 5.2), [h]-loss in Norwich is socially stratified, with
lower-class speakers tending strongly to loss of [h], despite the fact
that surrounding rural dialects tend to preserve [h]. In general,
[h]-loss is well established in urban vernaculars in England and
Wales; thus, a majority of the population of the country have
(categorical or variable) loss of [h].
However, it is also well known that social stereotyping of [h]-loss
does not apply in Scotland, Ireland, North America and colonial
Englishes generally, for the reason that [h] is stable in stressed
syllables in these varieties (although [h]-loss is found in English-
based Creoles). In this sense, the English-speaking world can be
divided into two broad speech-communities: one in which [h]-
dropping is widespread and has social significance, and one in which
it is so rare (if it happens at all) that it is socially irrelevant.
138 On the Time-depth of Variability in English

e(h) in hammer

Figure 5.1 h-pronouncing areas of England. (Adapted from Orton et al.

1963-9)

Although scholars have noticed instability in initial / spellings in

early English, the traditional view is that there is little reliable
evidence for ‘[h|-dropping’ in English much before the end of the
eighteenth century, and earlier instability in spelling is usually
dismissed as unreliable in handbook accounts. One reason given for
the alleged lateness of the phenomenon is its apparent absence from
colonial English (Wyld, 1927: 220). From a variationist point of view,
however, this is not necessarily conclusive, as language is variable at
all times; thus, it could be the case that modern [h]-ful and [h]-less
varieties are each equally derived from varieties in which [h]|-loss was
 On the Time-depth of Variability in English 139
Index

80

LWC
te _-—— MWC
40 a _——UWC
a“ -

20 2 ee, = Hee
a See rae LMC
0 eS SS ee ee MMC
WLS RPS FS CS
Style

Figure 5.2 Variable (h) by class and style in Norwich (LWC/MWC/

UWC = lower/middle/upper working class; LMC/MMC = lower/middle
middle class; WLS = word-list style; RPS = reading-passage _ style;
FS = formal (conversational) style; CS = casual (conversational) style).
(Adapted from Trudgill, 1974)

variable
— not categorically absent or categorically present. In any
case, colonial forms of English may have undergone change in (h); for
example, there is evidence that, although Australasian English is
[h]-ful now, it used to have [h]-dropping (Trudgill, 1986a: 138-9).
The evidence of variable spelling in ME seems to point to an early
origin, and if the arguments for this can be sustained, they have a
clear relevance to understanding historical patterns of variation.
In modern times [h]-dropping— like -im’ for (ing) —is extremely
widespread and well established: as we have noted above, it is not
confined to a particular region (as voicing of initial fricatives is, for
example). In fact, most people in England and Wales drop their [h]s
to a greater or lesser extent. Therefore, if the origin of the phenom-
enon is as recent as the late eighteenth century, it is difficult to
explain how it could have become so geographically widespread in so
short a time: it was already highly salient and overtly stigmatized by
. the latter half of the nineteenth century (for some citations see
Phillipps, 1984, 136-9). Consider the following comment by the
linguistic scholar Oliphant (1873: 226), made in the context of
objecting to ‘Americanisms’:

I ought in all fairness to acknowledge that no American fault comes up

to the revolting habit . . . of dropping or wrongly inserting the letter /.
Those whom we call ‘self-made men’ are much given to this hideous
140 On the Time-depth of Variability in English

barbarism . . . Few things will the English youth find in after-life more
profitable than the right use of the aforesaid letter.

Whatever we may think of Oliphant’s views, we have to assume there

would be little point in attacks on [h|-dropping by the educated elite
unless it was highly salient and widespread, and it is reasonable to
assume for these reasons that it probably has quite a long history in
the language. The late eighteenth-century evidence adduced by Wyld
and others is therefore likely to indicate the date at which it had
become stigmatized as a ‘vulgarism’, rather than its date of origin. By
that time — and not necessarily before — [h]-loss was recognized as an
identifying characteristic of certain salient social groupings.
The most important reason for questioning the traditional view,
however, is that variation in initial / usage is a very common pattern in
ME texts. Whereas we have discussed orderly variation in spelling
(above) by looking at distribution within a single text, the evidence for
early [h]-loss depends on spelling variation across a number of texts.
Many Early ME sources exhibit variable use of the letter / in
syllable-initial positions before vowels (that is, in such words as hate,
hopper). Sometimes it is omitted where it is historically expected to be
present, and sometimes it is added where it is not expected.
This pattern of variation is widespread in Early ME, and the maps
of the Linguistic Atlas ofLate Medieval English also show a distribution
at later periods. It has been very widely noted by careful editors such
as Hall (1920), and (although the atlas map shows some West
Midland distribution) it seems in the early part of the period to be
most common in texts originating in the East Midlands, East Anglia
and the South. It is quite common in southern texts of c.1200, such as
Poema Morale and The Owl and the Nightingale, and in early East
Midland/East Anglian texts such as Genesis and Exodus, King Horn,
Havelok. It is found in the Otho text of Layamon’s Brut, but not in the
Caligula text, which is certainly south-west Midland. It is not
characteristic of early texts known to be West Midland, such as those
of the Katherine Group. The geographical distribution of relevant
texts from c. 1190-1320 is from Lincolnshire or Norfolk (in the
north) to the southern counties, but the instability seems to be
greatest in the East Midlands. Certain later texts, mostly of a
non-literary kind, display the same phenomenon. It is found in
Kristensson’s (1967) northern onomastic sources in the period
1290-1350 (he notes that it may be ‘Anglo-Norman’), and Wyld
 On the Time-depth of Variability in English 141

(1927, 1936) documents a number of later examples, from sources

that include the Norfolk Gilds (late fourteenth century), The Paston
Letters (fifteenth century), and the mid-sixteenth-century Diary of
Henry Machyn (for a fuller discussion, see J. Milroy, 1983: 48-9). In
my own investigations of many of these texts, I have noted additional
examples. The following selective lists are from the thirteenth-
century Genesis and Exodus (Morris, 1873), which is believed to
originate in East Anglia. They include examples additional to those
given by Wyld. List A documents omission of h, and list B addition of
‘unhistorical’ h:
List A a, adde, adden, as, aue, auede, aued, auen, aue (parts of the
verb ‘have’: lines 239, 240, 1251, 1505, 1760, 2388, 2425, 2720, and
very commonly — considerably more so than forms with h; algen,
aligen, (‘hallow’): 258, 918: ail (‘hail’): 3066, 3183; ate (‘hate’): 373,
3638; alt (< infin ‘hold’): 924; atted(‘is called’ < OE hatan): 813; e
(‘he, they’): 2341, 2708, 4094; egest (‘highest’): 143, 1224; eld (‘held’):
2999; elles (‘of hell’): 4157; ere (of them’ < OE heora): 2855, 3773;
eOen (‘hence’): 2188; eui (‘heavy’): 2559; is (‘his’): 482, etc.; opperes
(‘hoppers’, that is, ‘locusts’): 3096; oste/ (‘hostel’, that is, ‘lodging’):
1056; om (‘home’): 2270; oten (‘called’): 1131.
List B hagte (‘wealth’): 431; hagt (grief): 486, 2044, 2082; halle
(‘all’): 2340; ham (‘am’): 926; helde (that is, elde: ‘age’): 457, 1527; her
(‘before’): 801; herf (that is, erf: ‘cattle’): 2991; herde (that is, erde:
‘land’): 806; hic (‘I’): 34, 2783; his (‘is’): 2935; hore (that is, or:
‘before’): 958; hunframe (unframe): 554; hunkinde (unkinde): 534;
hunne: (‘grant’): 2249; hunwreste (‘wicked’): 537; hure (‘our’): 322,
2206.
The most immediate ‘explanation’ for such substantial instability in
the use of / is that syllable-initial [h] was not present, or only variably
present, in the speech of the relevant regions. The letter was,
“however, present in the orthographic tradition (regardless of the
mixed origins of the tradition in Old English, Anglo-Norman and
. Latin orthography): thus, in the absence of strong orthographic
standardization, the scribes would omit it on some occasions and
insert it ‘hypercorrectly’ on others.
As instability of 4 is extremely common, it is remarkable that
careful scholars could have been so much aware of this type of
evidence, but could nevertheless have rejected it. Wyld (1936), for
example, cites a large number of spellings from around 1200 onwards
in which the letter / is ‘wrongly’ omitted or inserted, but concludes
142 On the Time-depth of Variability in English

that there is no reliable evidence for ‘the present day vulgarism’

before the eighteenth century (p. 296). Writers of standard hand-
books are also inclined to dismiss the early evidence. Ekwall (1975),
citing Walker, notes that [h]-loss might have occurred as early as
1791; Brunner (1963: 5) dismisses early ME spelling evidence with
the remark that Anglo-Norman scribes were prone to use initial /
‘incorrectly’. Dobson is also inclined, for similar reasons, to give the
impression that ME evidence for [h]-loss is sporadic and unreliable
(see, for example, Dobson, 1968 II: 991). Many more authorities who
accept the same view could be cited. In fact, the evidence for [h]-loss
in the texts I have mentioned above is anything but sporadic.
However, as the citation from Brunner suggests, /-instability is one of
the putative ‘Anglo-Norman’ features distinguished by Skeat (1897),
and this is another of the reasons why it has been dismissed.
Frequently, this orthographic evidence for variation in Middle
English is rejected not on the grounds that the scribe was literally an
Anglo-Norman (which is what Skeat argued), but that uses such as
variable / are originally scribal importations from French or Latin
usage. However, the origin of scribal habits is not in itself valid proof
that variable use of the conventions in written English do not also
relate to variable usages in spoken English. This is because variable
scribal usage is likely to be functional in some way, just as spoken
variation is functional (as suggested in chapter 2), and the most
immediately obvious function of an alphabetic writing system is to
relate writing to speech-forms, however complicated this relationship
may be. Thus, especially in a time of unsettled orthography, it is extremely
likely that current sound-changes will be admitted into writing,
whatever the historical origins of the writing conventions may be.
Moreover, the prima facie evidence for [h]-dropping continues well
into EModE —long after there can be any suspicion of direct
Anglo-Norman scribal interference. In addition to the non-literary
sources cited above, there is rather strong evidence in Shakespeare
and Marlowe — especially in puns of the type air, heir, hair —that
[h]-dropping was salient in the speech community: clearly, the puns
could not have worked if the ‘groundlings’ had not recognized them
(see, for example, Comedy of Errors, WL.ii.122-3; Dido and Aeneas,
1.i.10). But as it was accepted into this relatively formal literary genre,
[h]-loss seems to have been much less overtly stigmatized than it is
today. All this evidence strongly suggests that (h) has been a variable
in English for many centuries: [h]-loss may have gone to completion
 On the Time-depth of Variability in English 143

in some varieties at particular times and places, but in general speech

communities have used the variation over these centuries for stylistic
and social marking. In other words, whatever the linguistic origin of
the phenomenon may be — in phonotactic constraints (Lutz, 1985), in
rapid speech processes, in language contact, or in a combination of
these — variation in (h) has probably had social and stylistic functions in
the language for centuries.
We should bear in mind, however, that underlying the ‘colonial’
chronological argument and the Anglo-Norman argument, there is a
more fundamental reason why scholars have been so willing to reject
the evidence for [h]-loss. Many historians of English seem to have
shared the attitude of Oliphant (above) and have therefore been
inclined to think that apparently non-standard linguistic forms are
somehow not to be taken seriously as evidence for ‘genuine’ linguistic
change. Even such an excellent scholar as Wyld, who was extremely
interested in the social motivations of change, did not give sufficient
weight to the evidence for [h]-loss. What these scholars did was to
place their own (negative) evaluations of [h]-loss on to the speech
communities of earlier centuries in which evaluations of this pheno-
menon were not necessarily the same. It was difficult for them to
appreciate that [h]-loss could ever have been anything else but a
stigmatized form: in so far as they knew of evidence for it in earlier
centuries, they tended to dismiss it, seemingly in the belief that
‘vulgar’ and ‘careless’ usage is not implicated in linguistic change.
This attitude to [h]-dropping is, of course, symptomatic of a more
general attitude to non-standard English, and it can be seen as an
effect of the ‘standard ideology’.
Plainly, there is no compelling /inguistic reason why, in a particular
language at a particular time, loss of a segment should be considered
less beautiful or less ‘correct’ than its insertion. For example, loss of
‘pre-consonantal [r] (as in car, card) is widespread in many English
vernaculars, but it is not stigmatized in southern Britain. Although it
is a consonantal loss, it is not said to be ugly or careless in England
generally. Unlike /h/, however, [r] seems to have been lost quite early
in a forerunner of the institutional British accent (RP) at a time when
consciousness of the standard ideology was beginning to develop, and
it is this difference in the social evaluation of (r) and (h) which seems
to be the explanation for dominant attitudes to it. In dismissing
[h]-loss in the manner described, however, scholars may also have
dismissed important evidence for the study of how linguistic changes
144 On the Time-depth of Variability in English

are implemented and diffused. It is also relevant to observe, in a wider

dimension, that English is the only Germanic language that is widely
subject to [h]-loss.
It has also been clearly established that in the course of time
evaluation of particular variants can change or even be reversed. It is
therefore most unlikely that present-day stigmatized forms have
always been stigmatized, or that present-day elite forms have always
been elite. Even the RP ‘broad’ [a] (as in path, dance) seems to have
acquired its high evaluation only recently: Mugglestone (1989) cites
evidence from the nineteenth century to the effect that it was
stigmatized as a vulgarism by some commentators: it looks as though
it may have been ‘borrowed’ from a low-status dialect (such as
‘Cockney’). As for [h]-dropping-I have suggested elsewhere (J.
Milroy, 1983) that in the Middle Ages it may have been a marker of
more cultured speech.
What is clear is that, when compared with literary OE, some
varieties of Early ME (including those that show variation in (h)) were
contact varieties. They were spoken in those parts of the country that
had been massively subject to Scandinavian settlement in the late OE
period, and they were subsequently used in areas where the elite
language became Norman French. It is not impossible that [h]-loss
became fashionable because of its loss in French, but this cannot be
demonstrated. However, one outcome of contact is simplification and
loss of distinctions, and [h]-loss results in loss of a distinction — as in
the pair hall and all. Thus, the loss of [h] could be associated with the
contact situation in general, regardless of the specific properties of
the contact languages. But there is another circumstance that strongly
suggests that [h]-dropping was not a low-status feature in ME, and
that is its association at that time with East Anglian texts.
Figure 5.1 shows that East Anglia is one of those areas in which the
rural dialects have remained generally [h]-ful up to the present day.
The evidence for early /oss of [h], however, is quite strongly associated
with this part of the country. The texts of Genesis and Exodus, Havelok
the Dane and certain other Early ME works that I have cited, are
almost certainly East Anglian. As for late ME, among the best sources
for variable spelling (including variation in h) are the Norfolk Gilds
(late fourteenth century) and the Paston Letters (fifteenth century,
Norfolk). The Gilds contain records from Norwich, (King’s) Lynn
and other towns, and these include some quite magnificent spellings
such as alpenie (‘halfpenny’), hoke lewes (‘oak leaves’), and many more.
 On the Time-depth of Variability in English 145

We might therefore advance the hypothesis that the medieval evi-

dence (unlike the modern dialect evidence for [h|-fulness in Norfolk)
is from a relatively high social stratum — this is certainly true of the
Paston Letters at the very least — and that [h]-loss was current at that
level but not amongst the rural population. This is the kind of
hypothesis that would be suggested by a specifically sociolinguistic
interpretation of the evidence, but it is offered here as a suggestion,
not as a definite conclusion. In the following sections, I shall revert to
questions that are less socially-oriented than this, but still variationist.
I want to consider how far we can use descriptions of the Belfast and
Hiberno-English vowel system to reflect on the past — specifically on
aspects of the Early ModE vowel system.

5.6 Projecting Backwards: Vowel Systems

So far we have focused on the time-depth of vernacular variants in

English, using (h) as an example and treating it as a binary variable
(we have assumed that in such words as hall, hit it is either
pronounced or dropped). In this section I turn to the configuration of
vowel systems, and this of course is phonologically more complicated
than (h). As we noticed in chapter 3, vernacular vowel systems such as
the Belfast one may display patterns that are not comfortably
accounted for by standard or traditional methods. In particular,
traditional descriptive and analytic methods find it difficult to con-
template such phenomena as: (1) close approximation of phones
without merger; (2) overlapping allophones; (3) flip-flops, and (4)
reversal of merger. Indeed, in some cases evidence that seems to
point to one of these things may be rejected on the grounds that the
phenomenon is impossible. In what follows, I would like to bear in
mind two general points. The first is the difficulty of stipulating when
a sound-change is completed (this is amply demonstrated by the
history of (h), above), and hence the difficulty of saying precisely what
is meant by ‘sound-change’. The second is the traditional binary
distinction between ‘sound-change proper’ and ‘borrowing’: this
distinction, or something similar to it, still seems to be be assumed in
many orthodox accounts. That is to say that some changes are
believed to arise internally within the system and are motivated by
system-internal factors, whereas others are said to arise from the
146 On the Time-depth of Variability in English

influence of neighbouring languages or dialects. Leaving these (and

other) theoretical generalizations aside for the moment, we turn first
to the history of what is usually known as short /a/ in English, and to
possible historical patterns of overlap and merger with /e/ and /9/. In
later sections I present a fuller discussion of merger.
It is generally known, of course, that standard English enshrines
sporadic residues of approximations between these vowels, for
example in doublets of the type strop/strap, catch/ketch, but there has
been a good deal of dispute about EModE evidence for overlap and
merger in this part of the vowel-system. Orthodox accounts, as we
have seen, tend to ignore or explain away evidence for variability
when it does not have reflexes in present-day RP. The systematic
study of a non-standard system, however, supports the argument that
there could have been orderly variation in EModE involving merger
and reversal patterns of /a/ and /e/ ( or /9/) in certain consonantal
environments, and suggests that the usual account of the history of
/a/ (fronting to [z] and subsequent split into two RP phonemes) is
oversimplified. Amongst other things, this type of study can contri-
bute to problems in English etymology, for example in dealing with
pairs of the type: pack/peck. The following account will suggest that
many such pairs are etymological doublets, and we shall consider this
further in an appendix to the chapter.
The history of ModE /a/ is a traditional bone of contention, and
views have been expressed on it by a gallery of famous names from
Henry Sweet onward (for a review, see Lass, 1976). Some scholars
have argued that despite the front quality of OE /x/, ME a was a
back vowel (or perhaps merely a fully low vowel: it is not always clear
what is meant by ‘back vowel’), and on this basis have postulated a
change around 1600 from a back (or low) value to front-raised /x/,
which is of course the modern conservative RP value. Others (such as
Lass, 1976) have preferred to argue that a generally front vowel had
already existed in ME. Much of this argumentation is, of course,
concerned with very broad states of language, and, as I have pointed
out, can apply only at a rather high level of generality, because it is
likely that in EModE, as in PresE, there were varying conditioned
allophones of /a/ (perhaps some front and some back, or some low
and some low-mid). The regular patterns of variability observable in
(a) at the present day suggest this strongly. In what follows, we focus
on one particular modern vowel system, that of Belfast vernacular.
 On the Time-depth of Variability in English 147

5.7 Orderly Heterogeneity in Vowels: the Low Vowels in

Belfast

One of the purposes of the Belfast study of variation in (a) was to

project our observations on to past language states in which patterns
of heterogeneity must also have existed. In the event, the inner-city
study demonstrated that (a) in this vernacular is phonetically and
phonologically much more complicated than standard accounts of
EModE /a/ (such as Dobson, 1968). It is more complicated also than
(a) as studied in New York (Labov, 1966), and more complicated than
descriptions of /z/ in present-day RP (Gimson, 1970, etc.). We
noticed the wide range of realizations in chapter 4 (table 4.2, p. 97)
and commented on its regularity. The variation exhibited is con-
strained within definable limits: following velars predict a front vowel,
following voiceless stops predict a low front vowel; fricatives and
voiced obstruents allow variation between front and back, but strongly
favour backing. What table 4.2 does not show, however, is that there
is also variation in length in the /a/ system, which is equally
rule-governed. As many of the apparent overlaps between the low
vowels occur in environments where the vowel is short, a brief review
of the vowel-length system is necessary.
Like Scots, the Belfast vowel system does not use length phonem-
ically: most vowels have long and short allophones, and the variation
in length can be very marked. Thus, /a/ in monosyllables such as
back, bat, can’t (before voiceless stops and before sonorants +
voiceless obstruent) is very short, whereas before fricatives and voiced
obstruents (as in grass, mad, bag) it is markedly longer. We remarked
in chapter 3, however, that within the vowel system as a whole
(including /a/), length tends to co-vary with quality, that is, long-
vowel realizations are usually noticeably qualitatively different from
short-vowel allophones of the same phoneme. Furthermore, whereas
- the short-vowel allophones commonly exhibit patterns of overlap,
merger and near-merger with one another, the long-vowel realiza-
tions of different phonemes tend to be much more distinct from one
another: short-vowel realizations of /a/ overlap with /e/ and /9/ in
the same environments (for example, in pet, pat, pot), whereas the long
vowel realizations are fully distinct. Thus, whereas pairs such as
sat/set, can’t/Kent are commonly realized with the same, or very
148 On the Time-depth of Variability in English

similar, low-front short vowel, the pairs sand/send, bad/bed have

markedly different vowels, /a/ being long low-back and /e/ long
mid-front in such pairs. This commonly results in vowel alternations
in morphologically related pairs with /e/ and /9/, as in table 5.1.
Within the /a/ system, and in terms of its overlap with the other
low vowels, the areas most relevant to interpreting EModE data are:
(1) the short-vowel environments; and (2) the velar-raising rules. In
section 5.8, we examine these patterns in Belfast vernacular in a
degree of time-depth by appealing again to the evidence of Patterson
(1860).

Table 5.1

mid, long | low, short

send sent
twelve twelfth
Ed Eddy
Tom Tommy
Prod* Protestant

(* ‘Prod’ is a slang term for Protestant)

5.8 Short /a/ in Belfast Vernacular

One of the most salient characteristics of present-day Belfast ver-

nacular is the rule for raising /a/ to [z, €] before the three velar
consonants /k, g, n/ (these consonants are themselves normally
fronted to some degree). As we noticed in chapter 4.2, this is quite
well preserved in BV as a stable marker of membership of inner-city
communities. However, there is also a rule for raising afier /k, g/,
which predicts that such items as cabbage, castle, cab will appear with
front vowels. Unlike pre-velar raising, post-velar raising is highly
recessive. A common inner-city pattern is for the fronting rule to be
superseded by a lowering rule (before voiceless stops and in polysyll-
ables), or a backing rule (in other environments) in the speech of
younger speakers. Thus cab for example will appear as [ke-b] for some
speakers on some occasions and [ka-b] for others or on other
occasions. In the community as a whole, therefore, we can speak of
 On the Time-depth of Variability in English 149

alternation between front-raised values and lowered (mainly backed)

values, according to which of the rules is applied. The ‘change’ here
is phonetically sudden, not gradual, and the alternation is age-graded.
However, real-time evidence shows that in nineteenth-century Belfast
raising of /a/ was well established both before and after velars: the
rules are clearly stated by Patterson (1860) and cited in chapter 4.6.
Patterson attests that /a/ was raised after both velars (/k/ and /g/),
and Gregg’s (1964) lists for Larne (20 miles north of Belfast) indicate
that the same pattern has applied there. In present-day Belfast,
however, we have attested no cases of raising after /g/ or before /r/.
Thus, although Patterson has gellon for ‘gallon’ and care for ‘car’, and
although similar pronunciations are heard in Ulster rural dialects, our
Belfast data show that the rule has receded in an orderly way through
a series of environments. It now remains as a variable rule after /k/
before all consonants except /r/. The rule for raising before velars,
however, is strong and active and does not seem to have begun to
recede lexically: in word-list style (which is usually considered to be
formal) it persists much more strongly than post-velar raising;
neologisms undergo the rule, and it affects spelling (see the discus-
sion of ‘occasional spellings’ below). In 1976, for example, I noticed a
car-park sign in West Belfast which displayed the legend: NO
EXCESS (‘no access’), and some of our inner-city informants
refused to believe that there is an orthographic item fag (‘cigarette’) in
the language: for them, it was feg.
The fact that velar-raising is, or has been, so well established, bears
directly on the question of time-depth. What we are observing is a
late stage in the recession of formerly well-established rules. And
there is one well-known fact of present-day RP and SBE that points
to the importance of the velars in the ancestral forms of those dialects
also. Just as the velar environments prevent backing in BV, so they
have also prevented rounding after [w] in RP: hence, max, quack, but
want, swat. This suggests that, as in present BV, so also at some point
_ in the development of SBE, velar environments were fronted and
raised ahead of other environments and were not accessible to the
back-rounding rule. However, if raising of /a/ is so well established
in BV velar environments, the tendency to front-raising may formerly
have been present in. both BV and EModE in other environments
also, receding from these environments before receding from the
velars. There are indéed indications in Patterson (1860) that raising
of /a/ was, or had been, more widespread.
150 On the Time-depth of Variability in English

Patterson comments that the sound of ¢ is substituted for a ‘in some

places’ in almost every relevant word. It is not clear what this means,
but it is possible today to observe mid-front values of /a/ before all or
most consonants in many rural west of Ireland dialects. We shall see
that there are other indications that the Belfast /a/ system was
formerly more inclined to the front than it is now, and as we have
noted above, the EModE front-raising rule may have affected some
dialects later than others. What we observe in Belfast may therefore
be the reflex of a raising rule that was originally ‘metropolitan’ or
‘mainstream’, together with some admixture of Scots raising rules
which account for Scots [gles, kert] ‘glass, cart’, and which are now
partly superseded in Scots by lowering and backing rules. Patterson’s
evidence plainly indicates that there has been a gradual swing away
from front values in the past century or so. What is relevant to
EModE, however, is that Patterson’s word-lists suggest that a raising
rule still applied in pre-nasal environments (which often have a back
vowel today).
Patterson spells many a words with e, and some of these (such as
rether ‘rather’) are plainly sporadic residues of rural Scots (which has
[e, e] in, for example, father, rather, gather and in other pre-dental/
alveolar positions); before nasals, however, e for a is more evident in
Patterson. It appears in polysyllables before single nasals in exemine,
Jenuary, and before nasal + voiceless obstruent in demsel, exemple,
Entrim (‘Antrim’); it also appears in monosyllables before nasal +
voiceless obstruent in slent, brench: all of these now have a fully low
vowel in Belfast. We shall see repeatedly in what follows that
polysyllables are particularly resistant to change and that a following
sonorant + voiceless obstruent is also relatively resistant. For this
reason, Patterson’s data (being chiefly polysyllables) most probably
attest to a late stage in the loss of raising before nasals, a change that
had formerly affected a wider range of environments. We shall not be
surprised, therefore, to find evidence of pre-nasal raising in related
dialects, at earlier times and in other environments. For the moment,
however, it is reasonable to conclude that raising of /a/ formerly
affected other environments besides velars — the pre-nasal environ-
ment in particular. In the next section, we proceed to consider
variation in /e/, as our discussion up to this point has obviously raised
the possibility that tokens of /a/ and /e/ can merge in certain
environments.
 On the Time-depth of Variability in English 151

5.9 Short /e/ in Belfast: Overlap and Merger

Although ‘classical’ phoneme theory does not allow allophonic

overlapping and approximation without merger, the Belfast research
suggests that overlap and merger of phones in particular environ-
ments are common, and that they are amongst the patterns noticed
when linguistic change is observed in progress. An example of this is
the pattern of allophonic interchange between /a/ and /e/.
Just as /a/ can be raised to [e€] in velar environments, so /e/ can be
lowered to [a, x] before voiceless stops and before sonorant +
voiceless obstruent. These latter environments are precisely those in
which /a/ is also most likely to have a low front short value. Thus,
such pairs as sat/set, bat/bet commonly occur with a vowel around [z,
a], and there is little, if any, difference between the vowels in ant,
can’t, want, dance, on the one hand, and those in Kent, went, rent, pence
on the other. The pattern of /e/ lowering in these ‘short’ environ-
ments is extremely well established: as table 4.9, p. 111, shows, it is
categorical in some inner-city male groups. However, the pattern of
merger/approximation that this suggests is complicated by the
existence of the raising rule by which /a/ is raised before velars. The
lowering and raising rules can be stated as follows:

Rule 1 /e/ > <a>/ {voiceless stop}

Rule 2. /a/ > <e>/ {velar}

If these rules are interpreted categorically, they predict a flip-flop in

the case of voiceless velar stops: whereas the vowel in neck will lower
to [a] (Rule 1), the vowel in back will raise to [e] (Rule 2). This does,
in fact, happen very commonly in casual speech, as in the following
examples from the Belfast data:

‘1 The bfe]ck of my n[a]ck;

2 Will you pay by [ek]cess (‘Access’) card or by ch[a]que?;
3 T[a]xtile fle]ctory.

These rules, however, are not categorical, but variable. Thus,

speakers can also realise /a/ as [a] and /e/ as [e], although inner-city
lov On the Time-depth of Variability in English

speakers very rarely do this in casual speech. More importantly,

speakers can also use a further strategy in this case: merger. On
word-lists inner-city speakers all merged pack and peck ; when pressed
to differentiate, they often seemed unable, in the interview situation,
to do so. However, as items that were merged on word-lists were
commonly differentiated in conversational styles (and speakers some-
times actually commented on that fact), it seems possible that some
speakers may favour the flip-flop rule in casual style and the merger
in word-lists. Thus, the voiceless velar environment exhibits merger
in some instances and flip-flops in others; no doubt, it would be
possible to observe other varying patterns of approximation (as for the
meat/mate ‘merger’, below). Before we discuss further implications of
these patterns, however, we need to look briefly -at raising before
voiced velars, and at the rules for short /9 /.
Unlike the /e/ lowering rule, the velar-raising rule is not confined
to short-vowel environments (that is, before the voiceless velar stop),
but also applies when a voiced velar follows (that is, in long-vowel
environments): thus bag, bang are realized with [¢:]. When the vowel is
long, it is usually front-raised much farther than when it is short and
may be as high as [e:]. The routine application of the rule, however
results in flip-flops with /e/, so that jet-/ag, for example, is realized
with [a] in jet and [e:] in Jag. The short /95/ rules, however, are
formally identical to those for /e/: /9/ is realized as a short vowel near
[a] when it occurs before a voiceless stop or sonorant + voiceless
obstruent, for example in pot, font; both /e/ and /3/, therefore, show
patterns of overlap with /a/ in these environments and, under certain
conditions, in polysyllables, as we shall see below.

5.10 Syllabic Alternations

For the sake of clarity in the argument so far, I have largely ignored
the disyllabic/polysyllabic rules that affect /a/, /e/ and /9/, and have
confined most of the discussion to monosyllables. However, the
effects of polysyllabic environments are of the greatest interest, both
for the study of sound-change in progress and for historical explora-
tion. In the early stages of the inner-city study it was already evident
that these rules were of historical and theoretical importance, as they
lead to a regular series of alternations in derivationally related items.
 On the Time-depth of Variability in English 153

(They are discussed in J. Milroy, 1976a.) It seems that when a change

is in progress, polysyllabic environments will be amongst the last to be
affected: thus, the vowel observed in the polysyllable today may be the
vowel that was formerly found in other environments. It is possible to
suggest an implicational series, using the symbols suggested in
chapter 4.5: as the rule diffuses through the lexicon, monosyllabic T
environments are the first to be affected by it, followed by monosyl-
labic D environments, with the polysyllables being the last to be
affected. In preserving the conservative form in polysyllables, the
language enshrines time-depth differences in the phonology — a kind
of ‘apparent time’.
In the inner-city data, the (synchronic) rule for shortening of /e/ and
/9/ applies not only in the monosyllabic environments described, but
also in the stressed syllable of disyllables, when that syllable is
penultimate or pre-penultimate. The rule applies in all pre-
consonantal environments including those that would have a long
vowel in monosyllables (fricative and voiced obstruent environments
and liquids), except apparently before [s] clusters (as in hospital).
Thus, shortening applies to the stressed syllables of fellow, intention,
intentional, but not to fell, intend (as in the last case the stressed
syllable is not penultimate). These short disyllabic realizations of /a,
€, 0/ converge on a vowel near [a] (as in such monosyllables as went,
pot). One result is homophony or near-homophony in such pairs as
phonetic/fanatic, erratic/erotic.
The rule is plainly phonological, as it affects phonetically similar
pairs that are not morphologically or semantically related, as in table
5.2. However, as it applies to all phonologically similar pairs, it
inevitably has consequences for derivationally related items, such as
in table 5.3. The rule does not apply, however, in disyllabic/
polysyllabic inflected forms. Thus, while mess/message, intend/intention
have the alternation, mess/messing, intend/intended, have [e:] in both
items in each pair. Furthermore, if /a/ is followed by a velar in these
disyllabic environments, we again have a flip-flop: words like factory,
access have [e], whereas intention, message have [a]. The remarkable
predictability and regularity of these inner-city rules suggests that
they have deep historical embedding: their existence in well-
established form may help to interpret evidence for the configuration
of the low vowels of EModE.
The evidence that we have briefly reviewed above for various
patterns of flip-flop and phonologically conditioned interchanges
154 On the Time-depth of Variability in English

Table 5.2

[e] [a]
mess message
fell fellow

Table 5.3

[e, >] [a]

Ed Eddy
Tom Tommy
John Johnny
offend offensive
intend intention
intentional
intensive
superintendent

between adjacent vowels is much richer than anything we can hope to

recover from history. Nevertheless, we shall see in the next section
that EModE evidence strongly suggests a vowel system that exhibits
patterns of a similar type to those we have reviewed. These are
patterns which resemble vernacular states of language in which
change and variation are inherent, and which do not resemble
standard language states, which are uniform.

5.11 The Belfast Pattern and the Vowels of Early Modern

English

Amongst traditional scholars, H. C. Wyld was particularly inclined to

place a high value (for reconstructing EModE pronunciation) on
‘occasional spellings’ and rhymes/puns in written English, and to
consider the evidence of contemporary commentaries (the work of
the ‘orthoepists’) to be of less immediate value, as these can often be
difficult to evaluate. Dobson’s study (1968) of orthoepic evidence, on
 On the Time-depth of Variability in English 155

the other hand, relegates occasional spellings to a secondary position,

and relies more closely on the descriptions of the ‘best’ orthoepists.
Although such testimony is important, it is a little unfortunate that
Dobson’s influence led for a time to an undervaluing of occasional
spellings, rhymes and puns. The kind of variation revealed by
occasional spellings as primary evidence does not always fit comfor-
tably into the standard historical linguistic mould, and so it has often
seemed convenient to ignore it or explain it away, sometimes on the
grounds that variability of the kind apparently attested is ‘impossible’.
The best-known difficulty of this kind is the apparent EModE merger
of ME ‘open’ /e:/ (the meat class) with ME /a:/ (the mate class),
followed by apparent reversal of merger and subsequent merger of
the MEAT class with ME close /e:/ (the meet class). In the
concluding sections of this chapter we briefly consider in sociolin-
guistic terms: (1) phonemic overlap of /a/ and /e/ in EModE; (2) the
question of reversibility of merger, with reference to the meat/mate
‘merger’; (3) lexical diffusion patterns and the Neogrammarian
problem of gradual versus sudden phonetic change.
The value of EModE occasional spellings is given considerable
support by the Belfast evidence, put together as it is on the basis of
detailed observation of the spoken variety and real-time evidence
(Patterson, 1860). In view of this, many of the spellings cited by Wyld
from sixteenth-century sources (1936: 198-9) seem oddly familiar. It
is not necessary to be unduly selective to find indications of pre-velar
and pre-nasal raising of /a/, and polysyllabic lowering of /e/, as in
present-day Ulster English. Spellings that fall into these categories
are the majority of those cited by Wyld. Velar items include mex
(‘wax’), seck (‘sack’), thenking (‘thanking’), wexe (verb), renk (‘rank’),
beck (‘back’), ectes (‘acts’). Pre-palatal raising may also be indicated by
wesshe (‘wash’): this would not be surprising, as it is also characteristic
of stereotypical Lowland Scots and is found in the Belfast data. Wyld
also cites a number of such spellings in nasal environments that
- recapitulate Patterson’s fenuary, exemine: they are fenewary, axemyne
(‘examine’), exemynyde, Crenmer, Frencis. Apart from the proper
names, the items cited are the same as those in Patterson. All of them
are in polysyllables before the single nasal and before nasal clusters. It
is possible that the range of the nasal-raising rule was wider than in
nineteenth-century Belfast, as some other spellings indicate it in
monosyllables before nasal + voiced obstruent: these are hendes
(‘hands’), bend (‘band’). We cannot at the moment reconstruct with
156 On the Time-depth of Variability in English

certainty what the ordering of relevant environments in EModE

might have been, or whether it was always nasals that were decisive:
as Lowland Scots preserves evidence of a raising rule before
dentals/alveolars (for example in glad, gather, Saturday), it is not
impossible that hendes, bend attest to pre-dental raisings. However,
there is no doubt that the EModE items cited coincide closely with
Patterson’s environments (many similar examples are cited by other
commentators, such as Kokeritz and Dobson, and they are quite
common in texts with non-standard spellings from the Paston Letters
to the Diary ofHenry Machyn). As the EModE evidence for /a/ raising
is systematically related to the Belfast evidence, it is worth consid-
ering whether the flip-flop pattern affecting /a/ and /e/ in Belfast can
also be paralleled in EModE.
Wyld (1936: 198) provides evidence for this. He also cites a series
of what he calls ‘inverted spellings’ that strongly suggest a ‘flip-flop’
lowering of /e/ to /a/: these are also in the environments most
affected in Belfast, namely polysyllables. They include a for e in, for
example, Wanysday (‘Wednesday’), massynger, massage (‘message’),
zastyrday (‘yesterday’), mantion (‘mention’). The similarities to mo-
dern Ulster vernacular (see p. 151, above) are very striking to anyone
who knows Ulster vernacular: lowering of /e/ seems to have applied
in polysyllables in EModE (as in BV), and there seems to have been a
flip-flop, as the lowering of /e/ applies in environments where raising
of /a/ also applies — again as in Belfast. Thus, we have apparent
flip-flops before nasals as between Jenewary, Frencis, in which /a/ is
raised to [e¢], and Wanysday, mantion, in which /¢/ is lowered to [a]: if
raising before dentals also applied, then massage and others suggest a
further possibility of a flip-flop rule, and a wider range of environ-
ments in which the mid and low vowels are exchanged. Thus, there is
strong evidence that EModE, like modern Ulster, had rules for
raising of /a/ before velars, palatals and nasals, together with a rule
for lowering of /e/ in certain polysyllabic environments, and possibly
elsewhere. It also appears that just as Belfast speakers can merge
pairs like pack and peck, so it is possible that patterns of merger, or
near-merger, of such pairs might have been observable in EModE.
The most celebrated case of dispute about an alleged EModE
merger, however, is not pack/peck, but what we have elsewhere called
the meat/mate problem (J. Milroy and Harris, 1980). This is the
problem mentioned above — the question whether the ME open /e:/
class merged in EModE with ME /a:/, only to separate again and
 On the Time-depth of Variability in English iy)

undergo subsequent merger with ME close /e:/. Evidence from

rhymes and occasional spellings gives some support to this, and the
view taken by Wyld (1936) and followed by Kokeritz (1953) and
Dobson (1968) is that there was a merger followed by a reversal and
re-merger. Other prominent historical scholars (Jespersen, Luick and
others) however, have rejected the evidence for the meat/mate merger
by invoking the theoretical principle of irreversibility of merger. This
principle, which became strongly characteristic of structuralist think-
ing, continues into post-generative ‘mentalist’ historical linguistics.
Clearly, if the knowledge that classes A and B were once distinct is no
longer present in speakers’ minds, they cannot pick out the class B
items and so cannot separate them from the merged class and then
re-merge them with a completely different class. Thus, if the
historical evidence seems to suggest that meat merged with mate, this
cannot be a valid interpretation according to this principle, because if
it had happened then both meat and mate would subsequently have
been eligible for merger with meet: speakers would not have known
the difference.
The Belfast and general Hiberno-English system seems to display
a situation similar to EModE, and they cast considerable doubt on the
principle of irreversibility of merger. Hib-E speakers appear to have
access to two systems here, one in which meat merges with mate and
one in which meat merges with meet; that is, they know both mergers.
For most speakers the meat class belongs to both systems in that it is
an alternating class; thus, on different occasions the same speaker
may pronounce words of the type seat, peace, leave with [e:] or with [i].
Hib-E scholars have apparently always been quite happy to accept
that the meat/mate merger is a true merger (see citations in J. Milroy
and Harris, 1980: 200), and Bliss (1979: 208-10) cites evidence to
show that these two classes had merged in Hib-E by around 1700.
Seemingly, they are also happy to accept that the meat/ meet merger is
also a true merger, and that some people alternate between these two
' merged classes. If we take a social view of language, and accept that it
is the possession of the community as well as internal to speakers’
minds, we must be prepared to accept that reversal of merger can take
place in this way, so long as speakers know that meat is an alternating
class. The EModE scenario as envisaged by Wyld, Kokeritz and
others is therefore a possible scenario, and there is a sense in which
merger is reversible. Normal language states are not uniform, and in
vernacular (non-standardized) states we expect alternations and other
158 On the Time-depth of Variability in English

kinds of variation to exist. Thus, the potential for this kind of reversal
is always likely to be present in vernaculars. Whether a second merger
(in this case meat to meet) occurs in the community, and the direction
it moves in, will be determined not linguistically but socially.
Generally relevant here also is the (pull) set, which I discussed in
chapters 3 and 4. In BV the RP long and short /u/ classes (as in pool,
pull) are the same and are both realized as a short vowel; however, the
restricted lexical set of pull, bull, foot etc., is an alternating set,
rhyming alternately with cut, dull and with soon, pool, good. As we also
noticed in chapter 3, we cannot predict the full membership of the set
on phonological grounds. One group of alternants clearly originates
with the extension of the general unrounding of ME u (as in cut, dull)
to initial labial items, such as pull, but it is not clear how the words
originating from shortening in the ME 6 set can be predicted
phonologically (for example, if took belongs to the set, why not cook?). I
have listed the number of occurrences of each individual word
occurring in the data in table 5.4. Although it looks from this as
though each word has its own history, certain generalizations can be
based on these figures. First, disyllables are in general more resistant
to transfer into the incoming /u/ class than are monosyllables, and
this is in line with our findings for other vowels. The fact that in this
case the transfer is trans-phonemic whereas the other cases are
usually tensing rules within the same phoneme does not make any
difference. It seems that we reach a better generalization about
direction of vowel change in BV if we override contrastive phoneme
theory here. Second, items with initial labials are more resistant to
transfer than the other items; we can suggest, therefore, that the
initial labial is a constraint on change — it encourages maintenance.
Finally, labial items with initial [b] are more resistant to transfer than
other initial labial items. We have noted elsewhere (J. Milroy, 1980)
that the vernacular unrounded vowel is persistent in word-lists, that it
also persists into middle-class speech, and that polysyllables are the
most resistant environments. A rough implicational ordering of
resistance to change would be bulletin—bullet—bull-pull—football—foot,
with the leftmost item the most resistant and the rightmost the most
susceptible to transfer. The disentangling of ancient mergers that we
observe here has taken a very long time, and the best explanation for
the persistence of this alternating class is again a social explanation:
the ‘vernacular’ alternant carries an identity function and strong
connotations of closeness and intimacy.
 On the Time-depth of Variability in English 159

Table 5.4 The (pull) alternating set in Belfast.

Occurrences of [a]: reading styles

Items from ME short u Items from ME 6 and other sources

[ul] -N-— [a] % fu) -N-— [a] %

butcher 23 27 54 shook 40 ere

bush 1A 11 34 took 64 i we
bull Wg| 14 34 foot 35 070
pudding 36 4 10 could 56 0 0
put 49 4 8 look fi 0 0
pull 34 1 3 would 4] 0 0
Totals should 10 0 0
b-items 71 52 42 Total 253 1 0.4
p-items 119 Ds #i.
disyllables 59 S189
monosyllables 131 3421

Occurences of [a]: conversational styles

Items from ME short u Items from ME 6 and other sources

fu] —-N- [a] % fu] —-N- [a] %

bullet Z 8 80 football 5 4 44
pull 18 51 74 stood 10 7 41
full il 15 47 foot 22 Weis
put 189 120 39 took 99 49 33
push 11 Sol could 186 82 31
butcher 0 1 —- look 140 SP 27
bush 1 2 — would 453 88 16
bull 2 1 - should 54 5.108
pudding l 0 - shook 0 1 -
Totals Totals
. disyllables 3 9 75 disyllables 2 4 44
monosyllables 235 191 45 monosyllables 964 Ly eee

As we have shown elsewhere, however, it is not always the case in

such examples that full merger within the language system has come
about. The unreliability of reported mergers has been argued for
most strongly by Labov and his colleagues (Labov, Yaeger and
Steiner, 1972; Labov, 1975 etc.), and for inner-city Belfast it can be
160 On the Time-depth of Variability in English

argued that meat is not really merged with mate, despite the fact that
people believe that it is a merger. We demonstrated this statistically J.
Milroy and Harris, 1980): the data in table 5.5, were subjected to the
chi-squared test, and the difference between the distributions of meat
items and mate items was found to be very highly significant. This
means that although realizations of these overlap and are sometimes
identical, it does not really make sense to say that this is a merger in
the language system in BV (of course, it may be in other dialects). If it
were a full merger, the variation observed (as in variation in RP
meat/meet, for instance) would be insignificant and random. Thus, at
least one of the systems that children have been learning over the
centuries in the ancestral forms of BV must have incorporated a
three-way distinction of meet/meat/mate.

Table 5.5 Distribution of mate and meat items: vowel-height.

mate meat

[1e] 33 0
[e][ee] 60 20
lell¢ee] 6 38
le] 0 2

5.12 Conclusions

In this chapter, the general underlying issue with which we have been
concerned is the question of what a sound-change actually is. In
observing states of language, at what point can we specify that what
used to be A is now B? From a sociolinguistic perspective, we may
accept that a change is complete when some community agrees that it
is and reflects this change in their usage; it is a change in community
norms and so it is not adequately described if it is presented in the
traditional way as a phonetic movement. However, if we recall the
history of [h]-loss, it will be clear that the change from one phonetic
realization to another may take many centuries and may never be
complete. We must therefore extend our definition of change to
 On the Time-depth of Variability in English 161

include change in the distribution of variants. Thus, if at one time [h]

was categorical before vowels, the change to socially-marked variation
in (h) is itself a linguistic change. Similarly a change in the social
evaluation of a variant is also a change. It seems that questions of this
kind (especially: what do we mean by a linguistic change?) should now
be given more prominence than the more traditional Neogrammarian
questions, as discussed by Labov (1981) and more recently by
Kiparsky (1988). The most relevant Neogrammarian question here is
the question whether sound-change is (1) phonetically gradual and
lexically sudden, or (2) phonetically sudden and lexically gradual. I
end this chapter by turning briefly to this question.
My own studies of dialect divergence have repeatedly encountered
examples that fit with (2), that is, lexical diffusion. The (pu//) and
(meat) sets are examples: the transfer from [a] to [u] in pull, stood etc.
is achieved by a leap across phonetic space, and the process is
lexically gradual in that different lexical items are transferred at
differential rates. Observable change in the BV system is overwhelm-
ingly like this — phonetically sudden and lexically gradual. It is,
however, much less clear that we can demonstrate unequivocally that
sound-change can also be implemented in a phonetically gradual
fashion, although linguists still usually argue that this can happen. To
take an example from Belfast, we might attempt to argue for
phonetically gradual change on the basis of our data on /a/, which in
present-day Belfast is undergoing retraction to [a]. However, within
the /a/ system we repeatedly find that some parts of the backing
process are phonetically sudden and implemented through alternat-
ing sets: initial velar items such as cab, can are realized alternately by a
fully front low-mid vowel and a fully back low to low-mid (sometimes
rounded) vowel. This is plainly phonetically sudden, and not confined
to the initial velar set: it seems that subsets of phonologically defined
items within the system undergo a process of transfer one after the
other. So it is possible that the swing to the back in the /a/ system is
‘triggered in a phonetically sudden manner, and that patterns in the
data that appear to be gradual are secondary patterns dependent on
the primary sudden impetus to change. Thus, in cases where the
evidence available at a given time suggests that the process is
phonetically gradual, we may simply not have looked at the phonology
in sufficient depth and detail.
The Neogrammarian position on phonetically gradual change was,
of course, purely theoretical in the sense that they did not have the
162 On the Time-depth of Variability in English

technology to observe change in progress empirically. They may

therefore quite simply have been wrong. It is possible that sound-
change in the narrow traditional sense (in which A ‘becomes’ B) must
be phonetically sudden (involving movement from one allophone to
another, lexical diffusion and/or, sometimes, phonemic restructur-
ing), and that the apparently phonetically gradual patterns that we are
now able to observe are not sound-changes in progress, but simply
variation. This variation, when subjected to fine-grained analysis,
may at a given time appear to be moving in a particular direction, but
the direction may change, and the realizations may all drift back again
to where they started off. If there is to be a change, it seems that this
will be determined socially. Bearing in mind the points I have made
above about what a sound-change actually is, it may therefore be
useful to consider a possible socially based definition of language
change, dependent on a prior distinction between innovation and
change. In chapter 6, I will have more to say about this, amongst other
things.

Appendix: Etymological Applications of Variation Studies

Whereas relatively standardized varieties suppress alternative pro-

nunciations of the same word (with a few exceptions), vernaculars that
are less affected by the ‘mainstream’ variety are rich in lexical
alternants. In Belfast, for example, post-velar /a/ can be realized even
by the same speaker on different occasions as [¢] or as [a]. T.M.
(Clonard) pronounces can with [e], and then follows it with three
tokens of can with [a]. EModE seems to have been in this respect
similar to Belfast vernacular and dissimilar to post-eighteenth-
century relatively standardized varieties: as Kékeritz points out, it was
more tolerant of variation than PresStE is. It is as residues of the
EModE alternating state that PresStE preserves ‘fossilized’ pairs such
as strop/strap, with semantic differentiation that has rendered the
alternation functional in conveying meaning.
As for variation between /a/ and /e/, etymologists consistently
recognize doublets resulting from post-velar raising of the kind
attested in present-day Belfast (see, for example, Onions, 1966, s.v.
catch/ketch). The pre-velar raising rule, however, is not so consistently
recognized. Margaret/Meg and hackle/heckle are given by Onions as
 On the Time-depth of Variability in English 163

alternants, the second item in each pair being ‘Northern’ or ‘Scot-

tish’. However, the pairs drag/dregs and pack/peck (quarter of a bushel)
are not thought to be etymologically connected. Rather than appeal-
ing to rules that may have played some role within the English
language, scholars have been inclined to look outside of the language
for their inspirations — often to Old Norse, and sometimes (in
desperation?) to Middle Dutch.
For dreg(s), Onions says: ‘prob. of Scand. origin (cf ON pl.
dreggjar...), and he adds that ‘the problem of immed. origin is
complicated by the occurrence in early mod. E. of the forms dragges
and dredges.’ Patterson (1860), however, had no doubt that it was an
alternant of drags: he interpreted dregs as an Ulsterism deriving from
the operation of pre-velar raising, and recommended the hypercor-
rection drags, a hypercorrection which is still current in Ulster. It is
possible, of course, that the ultimate origin of the alternate form was
Old Norse, as that language did have pre-velar (probably palatalized
velar) raising of /a/, but this applied only when /i/ or /j/ followed in
the next syllable. So it is wise to be sceptical about wholesale
explanations based on Old Norse.
As for pack/peck, Onions says that both occur in the thirteenth
century, pack being from Middle Dutch and peck from Anglo-
Norman. Both ultimately are of unknown origin, that is, it is not
known how pek got into Anglo-Norman (could it have been from
English?). It could, of course, be an alternant of pack, a peck being the
measure appropriate to a donkey’s pack. If so, the alternation may be
due either to post-velar raising of /a/ or to pre-voiceless stop
lowering of /e/ at some early stage of English. The etymology of
wrack (cf. wreck) is similarly obscure, according to Onions (possibly
‘Middle Dutch’ — again!). Middle-class Ulster speakers often in-
terpret wrack/wreck as alternants: nerve-wracked is often ‘corrected’ to
nerve-wrecked, presumably on the assumption that wracked manifests
stigmatized lowering of /e/ before the voiceless stop. Surely we must
‘ consider the possibility that many such pairs are merely alternants,
arising from the operation in early English of the rules that we have
discussed. Origins in Old Norse and Middle Dutch in such cases
seem to be far more speculative than explanations based on alternat-
ing pairs, especially when (as in this case) we can support the
explanations by appealing to both historical and contemporary
sources.
 6
Speaker-innovation and Linguistic
Change

6.0 Introduction: Actuation and the Speaker

The discussion so far has been mainly about the embedding of

language in society and about the patterns, including historical
patterns, that can be revealed by systematic and accountable methods.
Much of the discussion has been about patterns of maintenance: it
has not been exclusively focused on linguistic change. In this chapter
I am concerned with change — specifically with the most intractable of
the five problems distinguished by Weinreich, Labov and Herzog
(1968). This is ‘the very heart of the matter’ — the actuation problem
itself — essentially the problem of explaining the causes of language
change. It is appropriate to start by recalling their statement of the
problem (p. 102):

Why do changes in a structural feature take place in a particular language

at a given time, but not in other languages with the same feature, or in the
same language at other times?

This is such a challenging formulation that (as we have seen above)

many linguists do not address it directly. We noted in chapter 2 that,
strictly speaking, the problem is insoluble, but that this is not an
excuse for neglecting it entirely. The point of view adopted here is
that linguistic change originates with speakers and is implemented in
social interactions between speakers, so it is reasonable to suggest
that by systematic observation of language in use we can come closer
 Speaker-innovation and Linguistic Change 165

to understanding actuation. But, recalling Principle 1 (that language

use — except in literary modes—is always observed within social
contexts), we must accept that our analysis and interpretation must
take account of society, situation and speaker.
To clarify this, I would like to take up again the methodological
distinction between speaker and system, which I have mentioned
several times so far. It is important to specify when we are talking
about speakers and when we are talking about systems, and I shall
attempt to show that observing this distinction can lead to important
insights that we would be unlikely to achieve if our approach were
purely language-internal.

6.1 Speaker and System

One clear difference between historical and social linguistics is the

methodological centrality of the speaker in sociolinguistic investiga-
tions. The database is not derived from written records, but from live
speakers in social contexts. It does not follow, of course, that more
direct access to speakers (on the part of field linguists) necessarily
implies direct access to linguistic ‘facts’. As we noted in chapter 2, we
do not observe linguistic patterns directly: we observe people talking.
Similarly, the ‘products’ of our analyses (Kibrik, 1977) are idealiza-
tions in the form of ‘grammars’; in this sense, Labov’s graphs of
linguistic change in New York City are ‘grammars’, just as generative
rule-statements are. Thus, the output of all linguistic investigations
can be regarded as ‘grammars of language’ constructed by linguists,
and not necessarily what any individual speaker of a language uses or
‘knows’.
In the process of data-analysis, the non-congruence of speaker-
based and system-based accounts becomes very obvious. We have
‘seen some examples of this from outer-city Belfast in chapter 4: vowel
realizations that we describe systemically as ‘long’ turn out to be
‘short’? in individual instances, and West (1988) has shown that
Harris’s (1985) characterization of Ulster vowel length does not turn
out to be correct for certain individual instances collected in Bally-
gawley, Co. Tyrone. But that does not invalidate Harris’s description
of the system, and I would be extremely surprised if these systemic
descriptions ever turned out to be correct for all speakers. For these
166 Speaker-innovation and Linguistic Change

reasons, it can be quite misleading not to draw the speaker/system

distinction in studies of language change. The following example may
make the necessity for the distinction clear.
I showed in chapter 4 (Sections 4.6 and 4.7) that /a/ (as in pat, bad)
and /e/ (as in set, bed) have been undergoing change for some time:
/a/ is becoming backed and lengthened, while /e/ is being front-
raised and lengthened. However, both seem to have started off from
much the same (low to low-mid front) position in phonological space,
and these older variants (apparently merged in many cases) are still
common, especially before voiceless stops. Thus, at the level of
language system we are able to make a suitably abstract and correct
statement about these changes. Rule 1 and Rule 2 describe the
changes affecting /a/ and /e/ respectively:

Rule 1 (/a/) —Back > +Back

+Low > +Low
Rales2(/2/9e-Backs = Back
+Low > —Low

This seems to capture what is currently happening to this part of the

vowel system, and it conveys the impression that the two vowels are
moving apart in phonetic space, leaving the low-front region unten-
anted. Viewed at the abstract level, we may interpret this as an
attempt on the part of the /anguage to disentangle a merger of such
items as pat/pet, can’t/Kent. But if we look at the output of speakers, a
very different pattern emerges. We can get a clue to this by noticing
that the implementation of the changes is sex-differentiated (see table
4.9, p. 111). Although the differences between males and females
here do not predict absolutely that females will always behave in one
way and males in another, the figures are nevertheless based on a
close analysis of the speech of individuals. It is clear that in general
those speakers (usually male) who have considerable /a/-backing do
not also have /e/-raising, and that those who have /e/-raising (almost
always female) do not normally have /a/-backing. Therefore, what
appears to be a grammar of change does not predict the manner in
which individual speakers are implementing the change, and it
appears that not all members of the speech community are participat-
ing in the same changes. Thus, it seems that a speaker-based account
does not necessarily support Labov’s view of ‘speech community’, or
the recent characterization by Labov and Harris (1986) in terms of
 Speaker-innovation and Linguistic Change 167

the two Philadelphia ethnic communities not ‘participating’ in the

same changes. It may be that members of the same inner-city Belfast
communities do not always participate in the same changes.
As we wish to use sociolinguistic findings to throw light on more
traditional system-oriented approaches to language change, it is
necessary to clarify further the implications of drawing a distinction
between speaker-activity on the one hand, and grammars of language
or of linguistic change on the other. In so far as linguistic changes are
in some sense brought about by the activities of speakers, it seems
appropriate to distinguish sharply between speaker-activity and the
linguistic system, to which speakers have access and which they can
influence. But here I want to show that whereas the approach to
language change advocated in this book is both speaker-oriented and
system-oriented, the approach of most quantitative social dialectology
has agreed with the tradition in being primarily system-oriented.
Traditional historical linguistics, as we have seen, has generally
presented linguistic change as something that can be described (or,
ultimately, explained) as a language-internal phenomenon, that is,
without primary reference to speakers in social groupings as the
agents of change: it has also tended to impose constraints on the kind
of historical data that can be adduced as evidence, focusing on what
Lass (1980: 87n.) calls ‘native languages of the usual type’. When
speakers have been referred to, it has usually been in post-generative
accounts that have recognized ‘idealized native-speaker competence’,
and the idea of speakers’ internalization of rules and capacity to bring
about rule-change (an example is Andersen, 1973). Sociolinguistics,
on the other hand, has recognized the centrality of speakers in a more
thoroughgoing way — very often in studies that need to take account of
language-mixing and code-switching in contact situations, that is, by
studying situations that do not necessarily focus on nativeness of
languages at all and that therefore do not have any place for the
Chomskyan idealized ‘native’ speaker. But sociolinguists have dif-
' fered amongst themselves in the emphasis they have given to the
centrality of speaker-behaviour in their methodologies.
The position adopted by different sociolinguistic investigators
ultimately depends on the aims of their research. One approach is to
produce an account based primarily on speakers (without obligatorily
attempting to map their output on to a grammar), and to focus on the
discourse strategies they use in interaction. This is associated with the
work of Gumperz, and the introduction to Gumperz (1982) is in fact
168 Speaker-innovation and Linguistic Change

one of the clearest accounts of the differences between speaker-based

and system-based approaches.’ The quantitative paradigm, however,
is primarily system-oriented and in Gumperz’s view devoted to the
production of ‘grammars’. Although it uses social (or speaker)
variables, and so takes account of the speaker, the main aim of
Labov’s work is to discover and describe linguistic patterns of
variation: this is very clear from the fact that the graphs and diagrams
of Labov’s early work (themselves mini-grammars of variation) were
then developed into variable rule descriptions, using an intra-
linguistic generative framework, and subsequently adopted in Kipars-
ky’s (1988) account of phonological change. Labov has replaced
uniform grammars with variable grammars, and this in itself is a
tremendous achievement. In the present perspective, however, it is
fair to say that the quantitative methodology has refined our un-
derstanding of variable language systems much more than our
understanding of speaker-innovations.
Clearly, it is by using the system-oriented approach of Labov that
we have made most progress in looking at the traditional problems of
historical linguistics (which, as I have pointed out, seldom took
account of the speaker’s role in change). As my concerns in this book
are with these traditional problems, I am, like Labov, necessarily
concerned with historical grammars. It seems, however, that a
solution to the traditional question of how changes are actuated
requires that the speaker’s activities in innovating should be in
principle distinguished sharply from the language state affected by
the innovations. For that reason, we must give a principled position to
the speaker in accounts of change, and the distinction between
speaker and system is fundamental in this book as a perspective from
which we interpret our findings. We can, of course, take the argument
somewhat further than this and claim that the speaker’s position in,
and relation to, broader social structures and processes should be
more systematically studied than it has normally been in sociolin-
guistics. We shall look at this in chapter 7.
For the moment, it is sufficient to observe that, as we are using
both system-oriented and speaker-oriented approaches in our work,
we need to draw a clear distinction between them, so that in any
discussion we will know which of these orientations we are actually
adopting. One obvious consequence of making this distinction is that
we can separate speaker-innovation from linguistic change, and we
now turn to this.
 Speaker-innovation and Linguistic Change 169

6.2 Speaker-innovation and Language Change

On the basis of the speaker/system distinction, we can propose an

associated distinction that bears directly on the actuation prob-
lem— the distinction between speaker innovation on the one hand,
and linguistic change on the other. Innovation and change are not
conceptually the same thing: an innovation is an act of the speaker,
whereas a change is observed within the language system. It is
speakers, and not languages, that innovate.
We can therefore approach the actuation problem in the following
way. We can describe speaker-innovation as an act of the speaker which 1s
capable of influencing linguistic structure. The innovation may, or may
not, enter the language system: thus, part of the solution to the
actuation problem will be to explain the conditions in which an
innovation is unsuccessful in addition to those in which it is
successful. This is one reason why it is important to associate our
account of change with a prior account of language maintenance:
incipient changes can be resisted in the speech community, and we
must attempt to understand the conditions under which this happens.
If, however, the innovation is successful, the reflex of this speaker-act
is change in the language system, which of course is always observed
after the speaker-act of innovation has taken place. Once a new
linguistic structure is created — that is, once change has entered the
system — it appears to penetrate in an orderly fashion and to constrain
individual and collective behaviour in the manner that has been
shown in quantitative analysis by Labov (1966) and many others. We
must notice here, however, that what the graphs, diagrams and
variable rule statements of the quantitative paradigm actually model is
not the behaviour of speakers, nor is it the act of actuation: what they
model is the /inguistic system. This quantitative modelling of the
- system is, of course, much more sophisticated than the homogeneous
and uniform systems that are postulated in other branches of
linguistic inquiry, as it incorporates the orderly heterogeneity of the
community ‘grammar’, but it is nonetheless a linguistic system — one
that characterizes the constraints on the linguistic behaviour ofgroups and
individuals. It can be interpreted as modelling the effects of the
linguistic system on speakers rather than the effects of speakers on
the system. This is conceptually quite a different matter from the
170 Speaker-innovation and Linguistic Change

modelling of how speakers introduce an innovation into the system in the

first place. These relationships are expressed graphically in figure 6.1.
It seems that while investigators may observe something quite close
to speaker-innovation, they have no principled way of determining
whether what they have observed is the beginning of a linguistic
change in the system. As figure 6.1 shows, what the quantitative
linguist describes is a change that has already assumed a regular
pattern of social variation in the community. For these reasons,
intra-linguistic approaches generally dismiss actuation as unobserv-
able. It is, for example, discussed by Lass (1980: 95-6) in the
following way. Suppose I have observed that a speaker utters [e] in a
‘word of etymological category X’ on one day, and utters [i] in the
same environment on the next day, all that I have observed (according
to Lass) is a diachronic correspondence: I have no way of determining
whether it is a change. But there are some distinctions that socio-
linguists customarily make that Lass does not make here; the ‘change’
from one day to the next might, for example, be stylistic — an instance
of orderly heterogeneity. But what is noticeable in the present context
is the absence of the speaker/system distinction: it appears from
Lass’s example that a change in the output of a single speaker might

Speaker
innovation

S-curve Development of
of social constraints
diffusion in linguistic system

Orderly heterogeneity
in speaker-behaviour

Observation of variation
by sociolinguist

Figure 6.1 Model of transition from speaker-innovation to linguistic

change.
 Speaker-innovation and Linguistic Change 171

be regarded as the locus of a change in the system, whereas of course

a change is not a change until it has been adopted by more than one
speaker. Therefore, we cannot deal with actuation by positing
examples like this, based on a single speaker. But it does appear that
we can observe a speaker-innovation (perhaps completely acci-
dentally): the problem is that we do not know whether it will be a
successful innovation — we cannot demonstrate systematically that it
leads to a linguistic change until after it has spread. Yet, as we have
already noted, the quantitative methodology does not in itself give us
the means to deal with actuation and the very early stages of a change.
The distinction I am trying to make here can be further clarified by
referring to some recent work by Trudgill (1986b), which strongly
suggests that we need methods quite different from the standard
quantitative ones to study actuation of change (depending on speaker-
behaviour) as distinct from the effect of the system on speakers. Returning
to Norwich eighteen years after his original survey, Trudgill
noticed that one particular linguistic feature (a labio-dental variant of
/t/) was by then firmly embedded as a patterned sociolinguistic
variable. In his 1968 fieldwork, however, he had noticed this feature
sporadically, but had thought that it might be pathological — perhaps a
speech problem experienced by a few younger speakers. Now,
although actuation of this change must already have taken place when
Trudgill observed it in 1968, it had not yet become embedded in
sociolinguistic structure in the way that is accessible by quantitative
methods, but was near the beginning of the S-curve of diffusion (see
figure 6.1). Therefore, using the classic quantitative methods, it was
not possible at that point to show the patterning of this change in the
sociolinguistic system.
In fact, the quantitative methodology predisposes us to overlook
cases like the Norwich one, and this example draws attention in a
clear way to a familiar difficulty. When we were selecting phonolo-
gical variables for quantification in Belfast, there were many variables
that had to be rejected from the quantification precisely because
tokens of them were sporadic or relatively rare. Using the standard
methods, we would not have been able to show large-scale patterns in
these variables. As I pointed out in chapter 3, some of these were
recessive forms (such as the meat/mate distinction). Others, however,
seemed to point to the early stages of change. For example, we
noticed in a pilot survey that the glottal stop for /t/ (as in mater,
what) — which is emphatically not part of traditional Belfast vernacu-
she Speaker-innovation and Linguistic Change

lar — occurred now and again, chiefly among female adolescents. I am

fairly sure that this is the leading edge of a change (the beginning of
the S-curve), because I know that urban central Scots English (which
has glottal stops) can influence Belfast —and because I believe that
there is a general tendency (‘drift’) towards glottalling in Cisatlantic
English — but I cannot convincingly demonstrate this by standard
quantitative methods. This is because (as we noted above) the classic
methods are designed to examine the effect of the system on
speaker-behaviour rather than vice versa, and so it is hard to see how
anyone using these methods could make a systematic attempt to
handle actuation or the very early stages of a change.
Bearing in mind the speaker/system distinction, therefore, we have
tried to approach the problem in a more abstract way by attempting a
social characterization of those persons, or sections of society, who are
responsible for the actuation of change, in terms of the social links
that can exist between speakers. We might call this a characterization
of the idealized speaker-innovator. In the following sections, I am
concerned with how this might be modelled.

6.3 The Speaker-innovator according to Labov

It is useful to begin with the best-known attempt to determine the

social characteristics of the speaker-innovator, which is that of Labov.
Labov’s characterization is based directly on the findings of neigh-
bourhood studies of socially cohesive groups, and his main conclu-
sions can be summarized as follows:
1 Speakers who lead sound change are those with the highest
status in their local communities as measured by a social class index.
2 Among persons of equal status ‘the most advanced speakers are
the persons with the largest number of local contacts within the
neighborhood, yet who have at the same time the highest proportion
of their acquaintances outside the neighborhood.’ Labov then goes on
to comment “Thus we have a portrait of individuals with the highest
local prestige who are responsive to a somewhat broader form of
prestige at the next larger level of social communication’ (1980: 261).
According to this account, linguistic innovation is accomplished by
persons who have many ties within the community but who simulta-
neously have a large number of outside contacts. This raises some
 Speaker-innovation and Linguistic Change 173

issues concerning the nature of social ties between persons, and we

shall return to this point below. But this characterization seems also
to rely on certain presuppositions about the broader social structure:
there is heavy reliance on stratificational social class and the asso-
ciated notion of prestige. Clearly, all this raises a number of general
issues that are central to sociolinguistic explanation, and we shall
discuss these more general issues in chapter 7.
Here, the most immediately relevant point is that Labov uses the
notion of prestige in two different ways: the innovators are said to
have prestige in two dimensions — both inside and outside the
communities. First, there is a superordinate locus of change, with
prestige depending on the broader socio-economic class distribution
in the wider community (this is implicit in measuring the status of the
innovator by a social-class index); within this, however, Labov
presents a more refined or micro-level locus of change within a
neighbourhood group of roughly equal status, in which the innovator
has ‘local prestige’. On the one hand, we have a kind of prestige that
is somehow agreed on by the wider community, but on the other a
more fine-grained kind of local prestige, which is presumably not the
same, but which must depend on the way in which this innovator is
subjectively evaluated by his or her peers in day-to-day encounters
within the local social networks.
These two kinds of prestige belong to two different orders of
conceptualization. In so far as socio-economic class is used in the
Labov paradigm to access the social structure in which change is
embedded, the first kind of prestige is macro-level: it is accessible
through a theory of abstract social structure, it often appears in
practice to be institutionalized, and it is associated with unequal
distribution of power in society at large. The second kind of prestige
is micro-level and subjective: it is predicated on personal attitudes
‘developed in the situations in which speakers interact as individuals,
and it is something of a truism that people who are accorded prestige
in this second sense frequently do not have prestige in the first sense.
In fact, these two kinds of prestige are often in conflict: the social
values. that confer prestige on prominent members of street gangs, for
example — such as skill in street-fighting and playing truant from
school (Labov, 1972a) —are clearly in conflict with the values of
mainstream society.
But there is another aspect of Labov’s characterization that leads to
a difficulty, and this difficulty arises from the nature of social links as
174 Speaker-innovation and Linguistic Change

modelled by a network-based account. In this perspective, we have

pointed out elsewhere (J. Milroy and L. Milroy, 1985b), following
Granovetter (1973), that it is doubtful whether an individual can be a
central member of a close-tie community and at the same time have
large numbers of close-tie outside contacts, or carry both kinds of
prestige (J. Milroy, 1989), and furthermore that even if this type of
person does exist, it is not clear that he or she, as described in Labov’s
account, is actually the innovator (Labov’s innovator in fact looks very
much like Boissevain’s (1974) broker, who by definition is not central
to the network). But this is a difficult point, to which I shall return
below; we have preferred to argue, not so much in terms of
characterizing a type of personality, but more abstractly in terms of
characterizing the kind of social ties that can exist between indi-
viduals. In this argument, it is weak ties and not strong ties that are
crucial.
Before I go on to develop this further, it is relevant to comment
briefly on Labov’s use of the concept of socio-economic class, as this
is a crucial factor in his model of the innovator. This use of
socio-economic class (as I shall further explain in chapter 7) derives
from a particular consensus-based theory of class (Parsons, 1952),
which is not self-evidently the only or correct theory of social class
structure. The relevant point here, however, is not so much the
rightness of the model as the fact that the theoretical category of
social class is basic to Labov’s methodology. As a result of this,
descriptions of ‘speech community’ and interpretations (or explana-
tions) of change tend also to be expressed primarily in terms of social
class, and this is probably why the notion of prestige looms so large in
interpretation. We produce a community grammar that is modelled in
terms of social class, and we then use social class (and the related
notion of prestige) to interpret the patterns that we have revealed.
This has many effects. One of these is that other social categories are
frequently interpreted as subsidiary to social class, and so patterns of
gender-differentiation, for example, are interpreted as being enacted
within this class framework: as female speech patterns usually tend
upwards in the social hierarchy, females are said to be exhibiting
‘prestige’ in their speech. But what is especially important here is that
Labov’s characterization of the speaker-innovator is measured against
the social-class hierarchy and expressed in terms of it. To judge by
other comments (Labov, 1980: 154), it looks as though the innovator
is conceived of within this model as being a lower-middle to upper
 Speaker-innovation and Linguistic Change 175

working-class individual. Taking all these points together, it is not at

all clear from Labov’s account how we could locate the innovator in a
community that is vot stratified in this way. Indeed, the Labov model
may even be interpreted as containing the entailment that change is
possible only in socially stratified societies.
The model of the innovator that I shall propose differs in many
ways from Labov’s characterization. It is not based primarily on the
idea of prestige, or on the operation of prestige in the social class
dimension, but on the rather different model of speech community that
I have been explaining in this book. This model is based on the
strength of social ties that can exist between individuals, and
innovation is conceived of as passing through relatively weak ties. The
model is therefore less personalized than Labov’s model appears to
be: it is not so much about a kind of person as about the kind of links
that exist between persons, and between groups of persons. It is in
this sense more abstract than Labov’s model. It therefore seems to be
applicable to a much broader range of language situations and is
capable of casting light on patterns of change at both the micro- and
macro-levels and at all points between these. In section 6.4, my
purpose is to introduce this model by first explaining the functions of
weak ties in facilitating language change.

6.4 Network Structure and Linguistic Change

In chapters 3 and 4, I attempted to give some idea of how an extensive

array of linguistic variants can be associated with a variety of
intersecting and overlapping social functions in close-tie communi-
ties. I also explained the reasoning behind the extension of our
analysis in the second phase of the Belfast project to outer-city
locations and to a city-wide survey. Essentially, the prediction is that
‘to the extent that the strong localized norm-enforcement of the inner
city is reduced, there will be a movement away from localized
vernacular norms of language. As we have seen, this means not only a
movement away from what are usually known as ‘stigmatized’ forms
in the direction of supra-local forms which are sometimes, but by no
means always, more similar to the ‘standard’: it also means a decline
in the incidence of the variable norms discussed in chapter 4. ‘That is
to say that to the extent that alternant forms (especially the phono-
176 Speaker-innovation and Linguistic Change

lexical ones) marking close-tie situations become less functional in

the outer city, they recede, and we are left with (amongst other things)
a reduction in the alternative choices that mark symbolic functions.
So it is relevant to recall that linguistic change is conceived of here as
being observable in different patterns: not only in the traditional
pattern of sound change from A to B, but more broadly in movements
towards greater or lesser consensus on community norms of usage,
for example movements towards greater or lesser convergence on
uniform norms.
Recall that the Belfast research design here depends on the idea of
norm maintenance, which we have operationalized in terms of seaal
network, and within this model we have distinguished between
relatively weak and strong network links. In any real community
individuals and groups will vary in the relative intensity of ties, and
this is what makes it possible to compare them in these terms. But
behind this there lies an idealization which predicts that in a
community bound by maximally dense and multiplex network ties
linguistic change would not take place at all. No such community can
actually exist, but the idealization is important, because it also implies
that to the extent that relatively weak ties exist in communities (as in
fact they do), the conditions will be present for linguistic change to
take place. This perception was partly borne out even in the
inner-city research. We noted that a very few individuals had
markedly low network strength scores, and furthermore that these
individuals tended to use language much less close to the core Belfast
vernacular, with a much lower use of the ‘close-tie’ variants (such as
[a] in words of the (pud/) class). The idea that relative strength of
network tie is a powerful predictor of language use is thus implicit in
the interpretative model we have used throughout: it predicts,
amongst other things, that to the extent that ties are strong, linguistic
change will be prevented or impeded, whereas to the extent that they {

are weak, they will be more open to external influences, and so

linguistic change will be facilitated.
Weak ties are much more difficult to investigate empirically than
strong ones, and the instinct of the network-based ethnographer is
usually to study relatively self-contained small communities that are
internally bound by strong links and relatively insulated from outside
influences. The ethnographic work reported in Cohen (1982), for
example, focuses on peripheral areas of the British Isles that have a
strong sense of local ‘community’. Although we may surmise that
 Speaker-innovation and Linguistic Change 177

urban situations (such as Belfast) are likely to exhibit lower density

and multiplexity in personal ties than remote rural ones (and are by
the same token also likely to be more open to outside influences),
many studies, both urban and rural, have shown that a close-knit
network structure functions as a conservative force, resisting press-
ures for change originating from outside the network; conversely,
those speakers whose ties to the localized network are weakest
approximate least closely to localized vernacular norms, and are most
exposed to external pressures for change (J. Milroy and L. Milroy,
1985b). This second observation suggests that since strong network
structure seems to be implicated in a rather negative way in linguistic
change, a closer examination of weak network ties might be profitable.
The difficulty in studying weak ties empirically means that the
quantitative variable of network (which can be readily applied to
close-knit communities) cannot be easily operationalized in situations
where the population is socially and/or geographically mobile. The
networks of mobile persons tend to be loose-knit; such persons form
(relatively weak) ties with very large numbers of others, and these are
often open-ended, seldom forming into close-tie clusters. It is
therefore difficult, in studying loose-knit situations, to produce direct
empirical (quantitative) evidence of the kind usually used to support
sociolinguistic theories, and indeed (as we noted above) the speaker-
innovator cannot easily be directly observed and located. However,
we have argued (J. Milroy and L. Milroy, 1985b) that the speaker-
innovator is a necessary theoretical construct if we are to clarify what
is involved in solving the actuation problem. Therefore, as we are
again dealing with an idealization here, we use a mode of argumenta-
tion that differs from the usual inductive mode favoured by quantitat-
ive linguists, and to support the argument we adduce evidence from
various sources.

6.5 Weak Ties

This evidence is of several different kinds. As is the case so often in

network analysis, we find that anthropological and sociological studies
of small-scale communities (as in Cohen, 1982) are illuminating. On
the basis of evidence from a number of such studies, Mewett (1982)
has observed that class differences in small communities begin to
178 Speaker-innovation and Linguistic Change

emerge over time as the proportion of multiplex relationships declines,

(multiplexity being an important characteristic of a close-knit type of
network structure). This observation, in addition to associating social
class stratification with the decline of close-knit networks, suggests a
framework for linking network studies with larger-scale class-based
studies in formulating a more coherent multi-level sociolinguistic
theory than we have at present, and I shall return to this point in
chapter 7. But we have also derived insights from important work by
Granovetter (1973, 1982), who has argued that ‘weak’ and uniplex
interpersonal ties, although they may be subjectively perceived as
unimportant, are in fact important channels through which innovation
and influence flow from one close-knit group to another, linking such
groups to the wider society. This rather larger-scale aspect of the
social function of weak ties has a number of important implications
for a socially accountable theory of linguistic change and diffusion,
some of which I shall briefly outline.
Granovetter’s working definition of ‘weak’ and ‘strong’ ties is as
follows: ‘The strength of a tie is a (probably linear) combination of the
amount of time, the emotional intensity, the intimacy (mutual confid-
ing) and the reciprocal services which characterise a tie’ (1973: 1361).
This is probably sufficient to satisfy most people’s feeling of what
might be meant by a ‘weak’ or ‘strong’ interpersonal tie, and it fits in
fairly well with our indicators for measuring network strength in the
Belfast inner-city communities. It also fits in with the principles
followed in comparing inner-city with outer-city Belfast (on which
see chapter 4): broadly speaking, the former is characterized by
stronger and the latter by weaker ties. Thus, although strength of tie is
a continuous variable, for the purpose of exposition Granovetter
treats it as if it were discrete, and we need always to bear in mind that
we are speaking in relative terms: a tie is ‘weak’ if it is less strong than
the other ties against which it is measured. Granovetter’s basic point
is that weak ties between groups regularly provide bridges through
which information and influence are diffused. Furthermore, these
bridges between groups cannot consist of strong ties: the ties must be
weak (that is, relatively weak when measured against internal ties).
Thus, weak ties may or may not function as bridges, but no strong tie
can. This is shown in figure 6.2.
Strong ties, however, are observed as concentrated within groups.
Thus, they give rise to a local cohesion of the kind that we explored in
 Speaker-innovation and Linguistic Change 179

A ~~

ER aes rl ote at
E B
Nae M
NY i

D .N —
N Ay 3 J iN
\ DS ~
\ p Br
X\

R Vaal
SS a :S V

Nea? x
—~
sS
N T U

Strong tie
—-—-—-—— Weak tie

Figure 6.2 Weak ties as bridges.

inner-city Belfast; yet, at the same time, they lead paradoxically to

overall fragmentation. Clearly, this perception is potentially very
illuminating in accounting for different language states at different
times and places at many levels of generality, ranging from the
interpersonal situations, through dialect-divergent, bilingual and
code-switching communities to the very broadest of language situa-
tions, and it throws light on the question of convergence and
divergence that we discussed in chapters 2 and 3. The model of
strong and weak ties presented graphically in figure 6.3 can be
thought of as an idealized representation of (for example) an urban
community which consists of clumps connected by predominantly
strong ties, which in turn are connected to other clumps by predomi-
nantly weak ties, but it can of course represent other kinds of
language situation that we might conceive of.
The important point (from our perspective) that follows from all
this is that weak inter-group ties are likely to be critical in transmitting
innovations from one group to another, despite the common-sense
assumption that strong ties fulfil this role. For example, Downes
(1984: 155) suggests that the network concept is important in
developing a theory of linguistic diffusion, but assumes that it is
strong ties that will be critical. This assumption seems to be shared by
many linguists who have considered the matter; indeed, as we have
noticed above, Labov (1980: 261) presents a model of the innovator
as an individual with strong ties both inside and outside a local group.
180 Speaker-innovation and Linguistic Change

Middle Peartics
social 8 18
groupings Pe ay

Lower
social
groupings

Strong tie
Sa Weak tie

Figure 6.3 Idealized representation of an urban community in which

weaker ties are more numerous in middle social groupings and between the
groups.

Clearly, this conflicts with the arguments presented here, which

predict that to the extent that ties are strong, linguistic change will be
impeded, not facilitated.
Granovetter’s principle seems at first sight to go against ‘common
sense’, and for this reason I need to expound it a little further. First of
all, it is likely that weak ties are much more numerous than strong
ties, simply because the time and energy invested in the maintenance
of strong ties must place an upper limit on how many it is possible to
have, whereas weak ties require little effort. Second, many more
individuals can be reached through weak ties than through strong
ties; consider for example the bridges set up by participants at
academic conferences, which link cohesive groups associated with
each institution and through which new ideas and information pass.
Conversely, information relayed through strong ties tends not to be
innovatory, since persons linked by strong ties tend to share contacts
(that is, to belong to overlapping networks). So they may, for example,
hear the same rumour several times. This general principle entails
that mobile individuals who have contracted many weak ties, but who
as a consequence of their mobility occupy a position marginal to some
 Speaker-innovation and Linguistic Change 181

cohesive group, are in a particularly strong position to carry informa-

tion across social boundaries and to diffuse innovations of all kinds.
In view of the norm-enforcing capacities of groups built up mainly
of strong ties, it is easy to see why innovators are likely to be persons
weakly linked to the group. Susceptibility to outside influence is likely
to increase in inverse proportion to strength of tie with the group.
Where groups are loose-knit—that is, linked mainly by weak
ties — they are likely to be generally more susceptible to innovation.
We might note that this contention is consistent with the principle
enunciated by Labov and Kroch that innovating groups are located
centrally in the social hierarchy, characterized as upper-working or
lower-middle class (Labov 1980: 254; Kroch 1978). For it is likely
that in British (and probably also North American) society the most
close-knit networks are located at the highest and lowest strata, with a
majority of socially and geographically mobile speakers (whose
networks are relatively loose-knit) falling between these two points.
One apparent difficulty with the proposal that innovators are only
marginally linked to the group is in explaining how these peripheral
people can successfully diffuse innovations to central members of that
group, who are of course resistant to innovation. One part of the
answer here is that central members often do not accept the innova-
tion: hence, for example, the persistence of regional varieties and
minority languages in strong-tie situations (compare here Andersen’s
(1986) idea of endocentric dialect communities). But to the extent that
they do accept innovations, two related points are relevant. First,
since resistance to innovation is likely to be strong in a norm-
conforming group, a large number of persons will have to be exposed
to it and adopt it in the early stages for it to spread successfully.
Now, in a mobile society, weak ties are likely to be very much more
numerous than strong ties (especially in urban communities), and
some of them are likely to function as bridges through which
innovations flow. Thus, an innovation like the London merger
between /0,0/ (as in brother, thin) and /v,f/ reported in Norwich
teenage speech (Trudgill 1986b: 54ff.) is likely to be transmitted
through a great many weak links between Londoners and Norwich
speakers, and Trudgill suggests tourists and football supporters as
individuals who might contract such links. Quite simply, before it
stands any chance of acceptancé by central members of a group, the
links through which it is originally transmitted need to be numerous
(compare Granovetter 1973: 1367). Thus, the existence of numerous
182 Speaker-innovation and Linguistic Change

weak ties is a necessary condition for innovations to spread: it is the

quantity as well as the quality of links between people that is crucial
here.
The second point we need to make in explaining the success of
marginal members of a group as innovators relates more directly to
Labov’s view of the innovating personality type. As Granovetter
suggests, persons central to a close-knit, norm-enforcing group are
likely to find innovation a risky activity (indeed it is probably more in
their interests to maintain and enforce norms than to innovate); but
adopting an innovation that is already widespread on the fringes of
the group is very much less risky. There is of course a time dimension
involved, and in this dimension a point may be reached at which
central members begin to accept that it is in their own interests to
adopt the innovation. Informal observation of cultural and political
innovation suggests that this is generally true. As an example we may
cite the final adoption of a marginal cult (Christianity) in ancient
Rome: it took centuries for this innovation to penetrate to the centre.
Central members of a group diminish the risk of potentially deviant
activity by adopting (after a lapse of time) an innovation from persons
who are already non-peripheral members of the group, rather than by
direct importation from marginals, who tend to be perceived as
deviant. Thus, we can in this way understand how accep-
tance — under certain conditions — can be a rational strategy on the
part of central members of the group.
Within the network model, therefore, the existence of numerous
weak ties is a necessary condition for innovation to be adopted. But
there must be additional conditions, and at least one of these is r
a
a
a
t
e
ie
r
i
E
i
eicee

psycho-social: this is that speakers from the receptor community want

to identify for some reason with speakers from the donor community.
Thus, the Norwich speakers cited by Trudgill in some sense view
London vernacular speakers as persons with whom, in Andersen’s
(1986) terms, they wish to express solidarity. Ultimately, for an
innovation to be adopted, it seems that the adopters must believe that
some benefit to themselves and/or their groups will come about
through the adoption of the innovation. The cost of adopting the
innovation in terms of effort will thus be perceived by the adopters as
less than the benefit received from adopting it. It also seems that an
explanation based on the idea of group identity and solidarity is more
satisfactory than one that relies on prestige in a social class dimen-
sion, and we shall return to this question in chapter 7.
 Speaker-innovation and Linguistic Change 183
Bearing all these points in mind, it is appropriate now to return to
Labov’s account of the innovator and compare it with our own. The
most general difference is that Labov’s account is about a type of
person, whereas ours is abstract and structural, focusing on the
nature of interpersonal links: it is based on relationships rather than
on persons. We might describe Labov’s innovator as a person who is
sociable and outgoing, and who has many friends both inside and
outside the local group. Intuitively, it seems very likely that informa-
tion of all kinds (including linguistic innovation) can be diffused by
such persons, for the reason that they have many contacts. But
according to our account, such individuals could not be near the
centre of a close-knit group and at the same time have many strong
outside ties. More probably, they would have relatively few multiplex
links with others, and many of their links would be open-ended and
hence low on density; they would have a predominance of weak links,
including many that constitute bridges between groups. In class terms
such persons would probably be mobile, and their profile would
therefore fit in with Labov’s view that socially mobile sectors
(upper-working to lower-middle class) are the ones in which lin-
guistic innovation and change are carried. It seems, however, that this
profile is not that of the innovator at all, but that of an early adopter,
and I shall consider this point fully in the next section.
What I have presented here is an abstract model, supported by the
insights of Granovetter, which in effect implies that a community
characterized by maximally strong network ties (and hence maximal
norm-enforcement) will not permit change to take place within it.
Real communities, however, contain varying degrees of internal
cohesion and varying degrees of openness to outside influence
through weak ties. The speaker-innovator within this model is not a
close-tie person, but one who is marginal to more than one (relatively)
close-knit group and who therefore forms a bridge between groups
across which innovations pass. In the next section, I shall adduce
some further support for the model; I shall then go on to look at some
case studies that demonstrate how the model can be used to interpret
patterns of variation, including some that are very difficult to make
sense of in any other way.
184 Speaker-innovation and Linguistic Change

6.6 Innovators and Early Adopters

Empirical support for our modelling of the speaker-innovator is

provided by Rogers and Shoemaker’s (1971) studies of about 1500
cases of innovation in many areas of life, including, for example,
innovations in agricultural, educational and technological methods. In
the present discussion, the most important principle emerging from
this work is the distinction between the innovator and the early adopter.
As the innovator has weak links to more than one group and forms a
bridge between groups, he or she is, in relation to the close-tie
eroups, a marginal individual. Rogers and Shoemaker’s studies
confirm the marginality of innovators and further suggest that
innovators are often perceived as underconforming to the point of
deviance. If this is correct, the innovator does not resemble Labov’s
(1980) characterization (an individual who has ‘prestige’ both inside
and outside the local group), but actually seems to have more in
common with the famous ‘lames’ of the Harlem study (Labov,
1972a). Conversely, Labov’s ‘innovator’ resembles what Rogers and
Shoemaker call the ‘early adopter’.
Early adopters are relatively central to the group and relatively
conforming to the group norms. Once the innovation reaches them, it
diffuses to the group as a whole, and at this stage it moves into the
middle part of the S-curve structure that is associated with the
diffusion of innovations generally. Thus, although linguistic pro-
cesses are much more complex than many of the other processes that
have been studied from this point of view, they share this pattern of
diffusion with other kinds of innovation. Later, once the new forms are
established in the group, they may diffuse from the centre outwards.
At the macro-level, therefore, it is tempting to see these patterns in
broad sweeps of cultural and linguistic history (the history of
Christianity comes again to mind), but we must leave this speculation
aside and return to the matter in hand, because there seems to be no
easy way for empirical studies of change in progress to identify in the
data the crucial distinction between innovators and early adopters.
However, we should again recall that we are not attempting to
describe the characteristics of personality types, but of relations
between groups and individuals, and these may vary considerably
according to different social and cultural conditions. That is to say
that we are not thinking of identifying some individual who lurks
 Speaker-innovation and Linguistic Change 185

around the margins of a group and labelling him or her ‘the

- innovator’. Nor is it a case of ‘once an innovator, always an innovator’,
and it is obviously true that people who are innovative in some ways
may not be innovative in others. We are thinking in structural terms,
and so we are concerned with the kinds of relationship between persons
that determine the conditions in which linguistic innovations can be
accepted or rejected. Thus, the whole question is relative, just as the
definition of the weakness of a tie is relative. What is clear, however,
is that if innovations are transmitted across relatively tenuous and
marginal links in fleeting encounters that are perceived as unimpor-
tant, we are unlikely to observe the actuation of a change. However
sophisticated our methods may be, we are much more likely to
observe the take-up and diffusion of the innovation by the more
socially salient early adopters.
Bearing these difficulties in mind, we now turn to some detailed
examples in order to demonstrate how the model developed here
affects the interpretation of linguistic variation in speech communi-
ties. First, I shall consider two cases of phonological variation from
the Belfast inner-city study, and then move on to suggest a tentative
analysis of some parts of the Philadelphia data, as reported by Labov
and Harris (1986) and Ash and Myhill (1986). Finally, I shall
consider a number of more general patterns of change that may be
illuminated by the model.

6.7 Weak Ties: Crossing the Peace-line in Belfast

It is usual to suppose that the diffusion of linguistic change is

encouraged by relatively open channels of communication and
discouraged by boundaries or weaknesses in lines of communication.
In Belfast, however, there are many patterns that are difficult to
explain in this apparently common-sense way, and we shall consider
two of them here. They are: (1) the social configuration of the spread
of /a/-backing from the Protestant east of the city into the Clonard, a
West Belfast Catholic community; and (2) the city-wide younger
generation consensus on the evaluation of the (pull) variable, as
against conflicting patterns in the older generation. The details of
these variables are more fully discussed in chapters 3, 4 and 5, and in
‘various publications referred to there. Table 4.10 (p. 116) and figure
186 Speaker-innovation and Linguistic Change

6.4 show respectively the stylistic patterning in East and West Belfast
and the crossover pattern in casual conversational style for (a).
We have already noticed that the backing of /a/ is led by East
Belfast males. Figure 6.4 shows this, and it also shows that the
movement of back /a/ into West Belfast is not led by Protestant males
in the Hammer, as might be expected, but by the younger female
group in the Catholic Clonard area. This is the group that exhibits the
crossover pattern and reverses the generally expected ‘stable norm’
patterns (of the kind demonstrated in chapter 4). In this group the
city-wide female movement away from /a/-backing is reversed: the
incidence of /a/-backing in the group is higher than in the other older
and younger female groups, higher than amongst older females in the
same area, and — surprisingly — also higher than amongst their young
male counterparts in the Clonard. When measured against other
groups, these young women are deviant.
When stylistic patterning (table 4.10) is additionally taken into
account, this young female group appears as the only Clonard group
with noticeable stylistic differentiation on the same pattern as in East
Belfast: they favour backing appreciably more in casual style than in
careful style. This pattern is innovatory in West Belfast in that it is not
established in the other Clonard groups: the social value attached to
the variants by the Clonard young women is the same as the social
evaluation evident in East Belfast: they seem to have adopted this
evaluation as a community norm. The linguistic change that they are
carrying is thus manifested as a change of norms in addition to a

(a) Index score

350

te, oe See ed gia ee Clonard

250 6 og NGOS Say I ak See cen
TET ooNOW ae “EE HE co. seeses
ie
——Hammer
.
oe .
. .
oe .
.
. .
..
on .
.
200+ A .
. . Ballymacarrett |
.

spots ban Snstivh ie 2 Cla

Men Women Men Women
40-55 40-55 18-25 18-25

Figure 6.4 Backing of /a/ in Ballymacarrett, the Clonard and the Hammer.
 Speaker-innovation and Linguistic Change 187

change in phonetic realization. Thus, while superficial inspection of

the data in figure 6.4 might suggest that the young Clonard female
pattern is modelled on that of the older Clonard males (who have
fairly high backing scores), such an explanation would fail to account
for the reduction of /a/-backing in other groups, and (more impor-
tantly) it would not explain the adoption of (a) as a stylistic marker by
the young Clonard females.
The social barriers that inhibit contacts between working-class
communities have been well described for many locations throughout
the world (examples are cited by L. Milroy, 1987), and they were
evident in our inner-city fieldwork even inside sectarian boundaries.
Inter-ethnic conflict in Belfast, however, has had the effect of
strengthening the barriers that are present in all such communities
(Boal, 1978). In fact, the major sectarian boundary in West Belfast is
now marked physically by a brick and barbed wire structure, which is
described by the military authorities, apparently without intentional
irony, as the ‘Peace Line’. The puzzle is that an East Belfast pattern
can be carried across these boundaries, evidently by a group of young
women whose movements and face-to-face contacts have been
constrained from a very early age. As we noted in chapter 4, there is a
long term shift in the vowel system towards back /a/, and this
- diffusion pattern from east to west is a continuation it. That this shift
is continuing across the iron barriers (both physical and psycholo-
gical) that separate the Protestant east and Catholic west, is a fact for
which we are obliged to seek a principled explanation.
The most accessible, and possibly the only, explanation is one that
takes account of weak ties and the distinction between the marginal
innovators and the early adopters. It seems that the Clonard young
women are central members of the group, and so they resemble early
adopters rather than innovators. This is quite clear from their
Network Strength score (as reported by L. Milroy, 1987: 204): they
all score extremely high on this— much higher than the young
Clonard males. Their average score is 4.75 out of a possible
maximum of 5.00.
Further personal information about this group points rather clearly
to innovation through multiple weak ties. These young women, unlike
their male counterparts, were in full employment: they all had regular
jobs outside the Clonard community at a rather poor city-centre
store. Here they were very likely to be in weak-tie contact with large
‘numbers of people from all over the city, both Catholic and Protes-
188 Speaker-innovation and Linguistic Change

tant. Thus, they would be well placed to adopt innovations trans-

mitted by persons peripheral to their core networks, and as a result
exposure to innovatory forms would be frequent. Given the large
number of service encounters in the store, it becomes possible for the
weak-tie encounters with back [a] users to exceed greatly the number
of strong-tie encounters with non-back [a] users. Hence the capacity
of innovation-bearing weak ties to compete with, and in this case
overcome, the innovation-resisting strong ties.
If we have a theoretical perspective such as the one developed here,
which explicitly predicts that an innovation will be transmitted
through (frequent and numerous) weak ties, we have a solution to the
problem of explaining how back [a] can diffuse in this way, and we
can present a plausible account of how the innovation can appear to
jump across a barrier of brick and barbed wire. If, however, we make
the usual assumption that innovations are diffused through strong
ties, the pattern is very difficult to explain. Yet, it is only if we make
this strong-tie assumption in the first place that [a] diffusion appears
to be a puzzle at all.
Whereas back [a] diffusion is mainly a change in a phonetic
segment, the change of pattern in the (pu//) variable is a change of
evaluation (or of agreement on norms). This variable is quantified on
the basis of a small phono-lexical set consisting of items such as pull,
push, took, shook, foot, which exhibit vowel alternation between [a] and
[u]; we discussed it in chapter 2. Although the [a] variant is recessive,
it has very strong affective values and is a very salient marker of casual
speech between close acquaintances. But here I wish to point out only
one thing — the change in consensus on norms over the generations.
Whereas the (dag) variable referred to in chapter 4 shows consensus
across the different groups — old and young, male and female — the
(pull) variable (shown in figure 6.5) shows consensus only in the
younger generation, where it has become a marker of gender-
differentiation. The question is: how can this normative consensus
come about in this divided city?
The pattern here is one in which the older groups do not agree on
the gender marking in use of the ‘in-group’ variant [a]. In Ballyma-
carrett, males favour [a], but in the two West Belfast communities,
gender preference is reversed: the [a] variant is favoured by the
females. The younger groups, however, show the same pattern in all
three communities: in all cases [a] preference is stronger among
males and weaker among females.
 Speaker-innovation and Linguistic Change 189

ee .
oe °°
ee °
AGS er 6 Oe EEO
ee ee
oe ee
50 oe?

25 s.Clonard
Ballymacarrett

\ Hammer

Men Women Men Women

40-55 40-55 18-25 18-25

Figure 6.5 Distribution of the (pull) variable (percentage of [a] variants are
shown) by age, sex and area in inner-city Belfast.

Again the puzzle is to explain how young people living in closed

communities, whose outside links are quite tenuous, could reach
cross-community consensus on the social value to be attached to the
two variants of (pull). In their parents’ youth there was greater
freedom of movement, and people frequently formed friendships
across regional and sectarian divisions; however, since the beginning
of the civil disorder in 1969, people have been much less able to form
strong ties outside their communities. Yet, despite this, the absorp-
tion of the (pu//) variable into the regular sociolinguistic structure of
Belfast vernacular has continued unhindered. Again, it is only if we
‘accept that weak ties are the normal channel for the diffusion of
innovations that the apparent paradox is resolved.
In these examples, I have selected instances based on extensive
quantified information which is very fine-grained and which is fully
accountable to the data, but the general pattern here had already
become evident from observation of other cases, and the (pull)
variable can be regarded as testing out a hypothesis that had already
been formed. There are many other examples involving different
190 Speaker-innovation and Linguistic Change

dimensions of variation (including, for example, phonological

mergers) that can be observed fairly easily and that appear to show
this general pattern of consensus in the inner-city younger genera-
tion. Indeed, once you are ‘clued in’ to the possibilities (especially
with regard to gender-differentiation), it is remarkable how readily
you can observe the trends in everyday encounters. Perhaps the most
dramatic of these trends is the progressive loss of localized lexical
items and reduction of phono-lexical variable sets of the (pull) type (J.
Milroy, 1981). Another example is the three-way merger (or apparent
merger) of words of the type fur/fir/fair, which are very close to being
fully merged amongst younger speakers. The few elderly speakers
that we studied (around 70 years old in 1975), however, exhibit a
three-way differentiation, and middle-aged speakers often have a
two-way differentiation. As indications of this greater consensus
amongst younger people had already been observed before we started
our quantitative analysis, we spoke in terms of ‘the rise of an urban
vernacular’: the first research application in 1975 proposed the
hypothesis that we were witnessing in Belfast a fairly early stage in the
development of a focused urban vernacular, in which there is a
generally observable trend towards greater consensus on norms.
However, we have also emphasized in this section the psycho-
social barrier of the sectarian difference in Belfast, which we might
expect to inhibit the trend towards consensus, and from the beginning
of our research we naturally wished to discover whether the ethnic
difference was consistently and reliably reflected in language. In our
pilot research, therefore, we looked at two East Belfast communities,
one Catholic and one Protestant (Catholics being a small minority in
this part of the city). In our analysis of the tapes, however, we could
find no appreciable differences between the two groups: the Catholics
spoke with an East Belfast accent (including back varieties of /a/) just
as the Protestants did, and their speech was more similar to East
Belfast Protestants than it was to West Belfast Catholics. Subse-
quently, after comparing different communities very fully in our
inner-city study, I was able to state, rather cautiously, that ‘there is as
yet no persuasive evidence to show that the two ethnic groups in
Belfast (and Ulster) can be clearly identified by differences in accent’
(Jj. Milroy, 1981: 44). Indeed, it seems that those features of
differentiation that in the past could have been seized upon as ethnic
markers, have been abandoned in favour of greater inner-city
agreement on norms of age, sex and contextual style. In general, the
 Speaker-innovation and Linguistic Change 19]

Catholic immigrants arrived in the city later than the Protestants and
brought from mid and west Ulster a number of features (such as
palatalization of initial [k]) that could have been used to reinforce
differences, but this does not seem to have happened. Both groups
seem to be moving in the same direction in the younger generation
even though there may be divergent movements in small details;
similarly, both groups appear to evaluate variants in much the same
way (and this evaluation is often quite divergent from ‘standard’
evaluations). Thus, whereas our findings indicate a trend towards
city-wide working-class consensus which overrides the ethnic diffe-
rence, the Philadelphia findings (Labov and Harris, 1986) show a
trend towards greater divergence between the two ethnic groups. In
the next section, therefore, I would like to examine the findings
reported by Labov and his colleagues in terms of the theoretical
model of innovation proposed in this chapter.

6.8 Weak Ties: Inter-ethnic Patterns in Philadelphia

The most general finding of the Philadelphia studies (Labov and

Harris, 1986) is that the speech of the two ethnic groups is diverging,
and not converging. The researchers express this in terms of
participation in linguistic changes: the black community is not
participating in the changes that are in progress in the white
community. It can also be expressed in terms of the model I have
been developing in this book. This would emphasize consensus and
conflict: the two communities do not agree on norms of usage, and so
in broad terms we can say that the sociolinguistic situation is one of
conflict rather than consensus.
In the Philadelphia reports, however, the idea of prestige is still
very influential in interpreting the variation revealed, and the idea of
‘social stratification is prominent in the argumentation. Labov and
Harris (1986: 20-1) mention the ‘prestige’ of the localized ‘innovator’
(pointing out, quite rightly, that the activities of this type of person
have much more influence than the ‘mainstream’ norms such as radio
and television); yet, at the same time they speak of the ‘dominant’
dialect as against the ‘dominated’. This dominance model is used in
interpreting their findings (Ash and Myhill, 1986) on four groups of
speakers: a core white group, a core black group and two marginal
sep Speaker-innovation and Linguistic Change

groups — a group of blacks who have considerable contact with whites

(let us call these WBs) and a group of whites who have considerable
contact with blacks (henceforth BWs). The findings are that the WBs
converge towards white morpho-syntactic norms more markedly than
the BWs converge towards black norms (see figure 6.6), and this is
where the idea of dominance comes into the argument. The pattern
discovered here is described in terms of dominance: the white dialect
is seen as dominant and less subject to change in a non-prestige
direction. Although this may be quite plausible in this case, where
power-structures are likely to be relevant, it cannot be generally
applicable as an explanation, because it does not account for the
converse pattern that has been so frequently noticed: historians of
language often point out that dominant (or ‘elite’ or ‘high prestige’)
dialects can be influenced ‘from below’, and Labov has also pointed
this out (Labov, 1972b). This kind of explanation for linguistic
patterning is therefore insufficiently general: it does not take into
account the fact that ‘dominant’ dialects do not always dominate.
Therefore, although Labov and Harris explicitly reject social network
as an interpretative category, it seems that some of these findings are
open to interpretations based on strong/weak ties, the identity
function of linguistic variation, and models of linguistic accommoda-
tion and politeness. Furthermore, it also seems that this type of
interpretation can have greater explanatory power than one based on
‘prestige’, as it is applicable to a wider range of different language
situations.
What is particularly noticeable about the two contact groups
mentioned above (white-oriented blacks and black-oriented whites) is
that on morpho-syntactic variation their scores average about the
same: on copula-deletion and ain’t for didn’t, the whites actually
out-perform the blacks on ‘black’ variants. What is also noticeable
from figure 6.6 is that, whereas the core black group uses these
features quite variably (presumably also using the ‘white’ variants), the
core white group does not use the ‘black’ variants at all. Thus, the core
black vernacular (whether or not it is ‘dominated’) incorporates a
resource not available to mainstream white speakers — the capacity to
alternate between ‘black’ and ‘white’ morpho-syntactic variants
according to occasion of use. To this extent, there is a structural
resemblance to the inner-city Belfast speakers described in chapters
3 and 4, who also have at their disposal alternating forms which carry
different symbolic functions according to occasion of use. I have also
 Speaker-innovation and Linguistic Change 193
Per cent use of ‘black’ variant
100-

90h

80-

40+

30;-

20-

10-

3rdsg Copula Poss _ Ain’t/

—s —s didn’t

— — — — Blacks who have little contact with whites (Index of contact < 7)
Blacks who have considerable contact with whites (Index of contact > 8)
— — — — Whites who have considerable contact with blacks
Whites who have little contact with blacks

Figure 6.6 Average per cent use of ‘black’ morpho-syntactic variants by

‘four groups of speakers. (Adapted from Ash and Myhill, 1986)

suggested in chapter 3 that this kind of situation is actually a normal

situation: it is the more uniform state of language that 1s abnormal.
From this perspective, however, the convergence of WBs towards
‘white’ norms is not so remarkable, as these ‘white’ norms are already
available to them within core black vernacular variation. The reason why
WBs use the ‘white’ norms more often than other blacks is plainly
194 Speaker-innovation and Linguistic Change

accessible through a theory of weak ties, as it is clear from the

authors’ descriptions of these speakers that they are a contact group
and that their contacts with whites are of a classic weak-tie type: they
are described as con-men, hustlers and political activists. Their
contacts through these activities must generally be weak contacts, as it
is hardly possible to pursue these activities within dense and multiplex
networks where everyone knows everyone else. The degree to which
these speakers use the white norms is increased by the range and
number of situations in which they have these weak-tie contacts
outside their core community, and for them the adoption of more
‘white’ usage is functional in their weak-tie contacts. Con-men do not
directly threaten the ‘face’ of their victims, but try to gain their
confidence, and they will normally accommodate to the victims’
norms and expectations. Hence the broader models of accommoda-
tion (Giles and Smith, 1979) and politeness (Brown and Levinson,
1987) are relevant to interpretation here. The suggestion (Ash and
Myhill, 1986: 41) that ‘prestige’ is the explanation for this shift
towards ‘white’ norms seems, on the other hand, to be quite a weak
explanation, which merely begs the question of what ‘prestige’ can
actually mean in this case.
The convergence of BWs to ‘black’ norms is in a sense more
remarkable, as the core white dialect does not possess the new
variants (copula-deletion, etc.) that the BWs adopt (to a certain
extent) in carrying out their act of accommodation: these outside
variants have to be acquired, and so some affirmatory effort (or some
‘cost’) is involved. Although the researchers do not give precise
information as to the strength of these speakers’ participation in black
culture, a theory of innovations (Rogers and Shoemaker, 1971) and
weak ties (Granovetter, 1973) would predict that their ties with both
communities are likely to be relatively weak. It is this group, and not
the WB group, who most resemble the peripheral innovators charac-
terised by Rogers and Shoemaker, and who, as have I pointed out
above, may be ‘underconforming to the point of deviance’. This looks
very much like innovatory behaviour in terms of the model I have
suggested, and these data are particularly relevant to the argument as
we are dealing here with contact groups in which weak ties must be
involved. I think that Labov and his colleagues have succeeded here
in locating an innovating group.
Of course, we may well consider it to be very unlikely that these
particular innovations by marginal whites will penetrate deeply into
 Speaker-innovation and Linguistic Change 195

white society. However, our model is designed to deal with this kind
of situation also. It is based on the idea of language maintenance, and,
as I pointed out above, it must account for situations in which
innovations are unsuccessful as well as for those in which they are
successful. The model predicts that, for these innovations to pene-
trate, they will have to be taken up by early adopters, who in this case
would have to be more central to white society. Thus, whether these
‘black’ norms will ever penetrate will depend on the amount of access
these peripheral white speakers are likely to have to central members,
and the willingness of central members to adopt the innovations.
Unlikely as this eventuality may be, it is worth remembering that
forms originating in black usage have been accepted into ‘mains-
tream’ English in the past, and it is sometimes suggested that these
have been originated by, and diffused through, jazz musicians. These
are normally underconforming people, as the Rogers and Shoemaker
model would predict.
Thus, although it is unlikely that these particular innovations will
ever show up as regular variants in white society, this does not make
the behaviour of these speakers any less innovatory. If we are to
understand the phenomenon of language change, we need to observe
the distinction between innovation and change, and we must recogn-
ize that innovations do not always lead to change: as Tarde (1903:
140) points out, we need to learn why, if 100 innovations are
conceived simultaneously, 10 will spread while 90 will be forgotten.
The data I have used in sections 6.7 and 6.8 are drawn from
extensive fine-grained analysis of present-day speech communities;
there are, however, many language situations in the world in which a
model of strong and weak ties might help to illuminate broader
patterns of change. In section 6.9 therefore, I go on to discuss a
number of more general situations in which the model may be
helpful. At this broader level, the majority of situations that I shall
discuss are those generally regarded as /anguage-contact situations.

6.9 Weak Ties as an Explanation for Change: Some

Examples

There are many patterns of linguistic maintenance and change that

have been widely observed, but that are not easy to explain in a
196 Speaker-innovation and Linguistic Change

consistent way if we depend on the usual assumptions about diffusion

being brought about through strong links. These patterns can be
observed at widely differing levels of generality, and I shall review a
small number of them here. At the macro-level, it can be noted that
while some languages (such as English and Danish) have been
structurally innovative (mainly in the direction of inflexional simplifi-
cation), other related languages (such as Icelandic) remain highly
conservative. In such cases we may be able to show that the
conservative communities are characterized by close-knit networks
and the innovatory ones by the development over history of numerous
loose-knit ties, and we have attempted to demonstrate this in the case
of English and Icelandic (J. Milroy and L. Milroy, 1985b). Although
the relative stability of Icelandic through the centuries may be largely
attributable to its geographical isolation, its uniformity throughout the
country requires further explanation. In a sparsely populated country
in which communities have been traditionally isolated from one
another by climate and terrain, one might expect that dialectal
diversity, rather than uniformity, would result. We attempted to throw
light on this by adducing evidence from the Icelandic Family Sagas of
the twelfth and thirteenth centuries, which show very clearly the high
value placed on the maintenance of strong-tie relationships over long
distances, and we therefore suggested that the maintenance of strong
ties may be crucial in explaining the failure of Icelandic to split up
historically into divergent dialects. English, on the other hand, has
changed radically and exhibits gross dialect divergence: this may be
largely attributable to the development of weak-tie patterns.
It can also happen that within a particular language, some dialects
of the language are innovative, whereas others are conservative. The
conservative dialects are often (but not always) regionally peripheral,
and therefore likely to be relatively strong-tie communities that are
less exposed than centrally located dialects to influence from mains-
tream norms. Relevant here is the case of the Scots vowel systems
(Catford, 1957), in which the regionally peripheral dialects have a
greater number of vowel phonemes than the central ones, the latter
having innovated by merging vowel phonemes that were previously
distinct. In a wide-ranging discussion of innovatory and conservative
patterns in a large number of dialects of different languages,
Andersen (1986) proposes a distinction between open and closed
dialect communities, and further proposes that within these catego-
ries some communities are endocentric, whereas others are exocentric. It
 Speaker-innovation and Linguistic Change 197

seems that endocentrism and exocentrism may have to do with

psycho-social attitudes within the communities towards the mainte-
nance of local identities and resistance to external influence, with the
endocentric communities strong in these respects and the exocentric
communites weak. However, it is not easy to see how we could
undertake an empirical investigation of personal attitudes in such a
wide range of relevant language situations, although it does appear
that once we have supported a theory of weak and strong network ties
by fine-grained sociolinguistic analysis of the kind we have described
in sections 6.7 and 6.8, we may be justified in projecting the model on
to these more general instances. This will predict that the closed
communities will be characterized by strong internal cohesiveness
with relatively few weak external links, and the open ones by more
loose-knit networks, with more external links. It is then a matter of
adducing supporting evidence to assess the degree of network
strength in the relevant communities.
There are other situations which are also clearly amenable to
weak-tie explanations. For example, it has been widely noted that
neighbouring languages which are unrelated or distantly related may
share similar linguistic changes. Trudgill (1983b: 56-9) draws
attention to the wide distribution of uvular /r/ in north-west Euro-
pean languages, and certain Balkan languages are known to share
specific grammatical features even though they are very distantly
related. Trudgill suggests that we come closer to an explanation for
the cross-language distribution of uvular /r/ if we focus on urban
centres which have it: Paris, The Hague, Cologne, Berlin,
Bergen and others (indeed, in French, alveolar /r/ is overtly
perceived as a rustic feature). One of the most important ways
in which this innovation has been implemented is by ‘jumping’ from
one urban centre to another. Within specific languages also, it is
known that innovations tend to spread from one centre of population
to another without immediately affecting the intervening countryside
(which is of course more sparsely populated). Urban development
seems therefore to be implicated in the maintenance and diffusion of
vernacular features within a language. The adoption of London
features by Norwich teenagers (see above, p. 181) is a case in point,
and the historical spread of /h/-dropping and (probably) glottal-
stopping in British English seems also to be urban. There is support
for this in the fact that, despite the long history of /h/ dropping, there
are still rural (but not urban) dialects in southern England which do
198 Speaker-innovation and Linguistic Change

not /h/-drop (see, chapter 5 and Trudgill, 1974). As for slang and
colloquial expressions, it is (for example) a matter of common
observation in Belfast that many of these are diffused from Glasgow,
which is 150 miles away across the sea, but which is associated with
Belfast through numerous weak ties.
In an extended discussion of the process of innovation, Trudgill
(1986a: 54) raises problems rather similar to the ones raised here; he
suggests that it is difficult to account for the London-based Norwich
innovations if close ties are postulated between members of the
innovating and donor communities. It seems to be adolescents who
are implementing the merger of [0,6] with [v,f], and this is an age
group that does not normally contract strong ties over long distances.
We have also carried out informal tests on British adolescents in
Sheffield, who appear to be diffusing the use of well as an intensifier
before certain adjectives, as in well nice, well happy. The indications
are that this is well established in London vernacular: if this is so, its
spread to distant parts amongst adolescents must be due to weak,
rather than strong, ties. Trudgill further cites a report by the
dialectologist, Gary Underwood, that persons who moved from his
home area of Memphis and then returned were influential in
spreading urban speech-forms to their rural friends. These persons
were effectively weakly linked to the rural network. Similarly, Trud-
gill reports the notion of the ‘language missionary’, developed by the
Norwegian social dialectologist, Anders Steinsholt. Discussing the
spread of urban forms from Larvik into the rural dialect of Hedrum,
Steinsholt comments:

The urban dialect spreads into Hedrum partly as a result of the

influence of particular individuals living in different parts of the area.
Such individuals — we can call them ‘language missionaries’ — may be
village people who have been particularly heavily influenced by the
urban dialect. The most important language missionaries are first the
young girls who come home after living for some time in the town, and,
second, the whalers. (Trudgill, 1986a: 57)

All these situations involve relatively weak links, but they are also
often connected with spread from urban centres. Thus, on the basis
of such observations, it might be proposed that urbanization is in itself
the main ‘cause’ of such phenomena. This does not, however, seem
to be satisfactory, as at least some of the phenomena discussed appear
 Speaker-innovation and Linguistic Change 199

to apply in some circumstances to non-urban situations, or to

situations that are rather marginally urban (involving very small towns
and villages). For example, one possible case of the spread of
innovations from one language to another is the occurrence of
pre-aspiration of consonants in two geographically peripheral and very
distantly related languages — Scots Gaelic and Icelandic. While this
development could be independent in these languages, it is neverthe-
less true that they were in prolonged contact in medieval times. Many
Norse speakers settled in the Western and Northern Isles and
northern mainland of Scotland, intermarried with Gaelic speakers
and subsequently emigrated, often with Celtic servants, to Iceland.
We cannot appeal to urbanization in this case, but we can appeal to
language contact in a more general way. Therefore, as urban
situations are a sub-type of contact situation, the correct generaliza-
tion is that contact situations (including urban ones) result in an
increase in the number and frequency of weak ties existing within
populations.
As we have raised the question of language contact, it seems
appropriate to comment here that a theory of weak ties at the level of
speaker may help to illuminate language-contact situations generally.
What we call close contact situations, which lead to linguistic change,
seem to be characterized not by close networks, but by open ones — not
by numerous strong ties, but by the development of numerous meak
ties. Clearly, if we do not distinguish between speaker and system,
this might appear to be contradictory or paradoxical, because the
influence on the language situation concerned is strong, rather than
weak. For example, the influence on Scandinavian and Norman
French on medieval English was strong, and the consequences of the
shift from Hungarian to German speaking in Oberwart, reported by
Gal (1979) are considerable. However, if we bear in mind the fact
that although change is observed in systems, it must be brought about
by speakers, the apparent contradiction is resolved. When linguists
speak of a close contact situation, they are usually thinking of contact
between systems, but what actually occurs is contact between speakers
of different languages: the changes that result and that are then
observed in the system have been brought about by the speakers, who
form weak and uniplex ties when two populations first come into
contact. So, strictly speaking, it is not really language contact at all, but
speaker contact. In such situations the model would predict that the
innovators in close contact situations are those who form weak ties
200 Speaker-innovation and Linguistic Change

both inside and outside their own community, and not the central
members of either community.
We also speak of different types of language contact, which lead to
different results, for example: stable bilingualism, gradual language
shift (Gal, 1979, Gumperz, 1982), survival of one of the languages
after it has been influenced by the other, rapid language-mixing,
pidginization. It would seem that these different situations may all be
approachable in terms of the relative strength of the ties that are
formed between the relevant populations in given cases. For example,
we might predict that the survival of stable bilingualism in a
community will be associated with the maintenance of dense and
multiplex networks within each community in roughly equal balance,
and that language shift (for example, the shift to German in
Oberwart) will be facilitated by the development of numerous and
frequent weak-tie contacts. The model of weak ties gives us a basis
for interpreting the results of many such studies, and it seems to be
illuminating in a wider range of situations than those that have been
approached mainly in terms of hierarchical social stratification and
‘prestige’. In the final section of this chapter, I want to recall some of
the general points about intra-linguistic theories of change that I have
already made (particularly in chapter 2), and consider the general
relevance of the weak-tie model to orthodox theoretical positions.

6.10 Conclusion: On Linguistic Change as a Social

Phenomenon

In this chapter I have attempted to explore the consequences of a

network-based approach for our understanding of how linguistic
change can happen, and I have suggested that the model is more
powerful than sociolinguistic models based primarily on social status
or class, in that it is capable of dealing with a wider range of language
situations. But I do not pretend that by using this model alone we
have solved the actuation problem. What I have tried to do, therefore,
is to approach actuation by first making a distinction between speaker
and system, and within this a distinction between speaker-innovation
and linguistic change. At the very least, this has the merit that it
clarifies what is actually involved in solving the actuation problem. We
are forced to consider carefully what we actually mean by ‘linguistic
 Speaker-innovation and Linguistic Change 201

change’. We are also forced to try to explain why some innovations are
successful, while others are not, and the notion of resistance to
change which is inherent in the network model is crucial here.
As many approaches to linguistic change still appear to assume that
changes pass through strong links between communities, I have also
summarized the arguments of Granovetter (1973) on the importance
of weak ties in the diffusion of change, and I have used the findings of
Rogers and Shoemaker (1971) on innovations. On the basis of these,
I have suggested a model of the speaker-innovator that differs from
Labov’s model, and have further suggested that the individuals or
groups that we identify as carrying linguistic changes are likely to be
early adopters of the change, rather than innovators. I have then
attempted to apply the model to fine-grained language situations of
the kind studied in Belfast and Philadelphia, and have tried to explain
its effect on how we interpret such data. It is fairly clear from all this,
however, that since the model appears to emphasize the notion of
external influences on communities, there are some implications for
orthodox notions of language change.
It is noticeable that, after a period of quiescence, the Neogramma-
rians have returned in recent years to a prominent position in the
theory of sound-change (see especially Labov, 1981; Kiparsky, 1988).
In chapter 5 we briefly considered the distinction between Neogram-
marian phonetically gradual sound-change and phonetically sudden
change, and we have noted from a sociolinguistic point of view that
although changes do not appear to be phonetically gradual, they are
certainly socially gradual. Here, it seems that the Neogrammarian
distinction between blind exceptionless change and linguistic ‘borro-
wing’ is also relevant, as it would be easy to conclude that what I have
proposed accounts mainly for patterns of diffusion and contact (that
is, ‘borrowing’) rather than for internal ‘spontaneous’ change within
speech communities. I do not think that this is quite correct: it seems
that the social model actually cuts across the traditional distinctions
‘and may even call into question the validity of the Neogrammarian
distinction. The distinction between innovation and change is not
merely terminological: it has consequences for what we mean by a
sound-change and for the manner in which questions about sound-
change are put. In the weak-tie model a// sound-change is socially
conditioned, simply because those so-called changes that arise
spontaneously are not actually changes: they are innovations, and they
do not become changes until they have assumed a social pattern in the
202 Speaker-innovation and Linguistic Change

community. If, as often happens, these innovations are not adopted by

a community, however small that community may be, then they do not
become changes at all. Thus, we are not asking how spontaneous
innovations arise, but how we are to specify the conditions under
which some of these innovations, and not others, are admitted into
linguistic systems as linguistic changes.
This interpretation of what is actually meant by a change differs
considerably from what is usually assumed. It is usual to suppose that
sound-change is an internal linguistic phenomenon, perhaps explain-
able with reference to the mental capacities of the idealized speaker,
but not a social phenomenon. It does not become social until it is
observed to show a social pattern in the community, and this comes
about when the community latches on to the variants that have already
arisen and imposes social meanings on them. Andersen’s insightful
account of deductive and abductive change is one of those that differ
in emphasis in this way, in that linguistic constraints are seen as
primary: his ‘implementation rules’ are structurally motivated within
the language system and therefore productive ((1973) 1978: 332). The
present account, on the other hand, would suggest that while all
variants must be constrained by aspects of linguistic structure and
may be produced by deductive and abductive generalizations (as
argued by Andersen), language-internal arguments or mentally-based
idealizations do not account for change (or indeed stable variation),
although they help to account for innovations. What has to be
explained is the manner in which speaker-based variants actually feed
into the system as established changes (Andersen handles this aspect
by postulating ‘adaptive rules’). A linguistic change is a social
phenomenon, and it comes about for reasons of marking social
identity, stylistic difference and so on. If it does not carry these social
meanings, then it is not a linguistic change: it is a random variant
stuck somewhere near the beginning of the S-curve in figure 6.1 (p.
170). Since our perspective differs in these ways, it is appropriate to
examine this point a little more fully.
Suppose we observe at a given time that a spontaneous innovation
(for example, a common one such as palatalization of /k/ before front
vowels) occurs in the speech of a small number of persons who do not
have any social contact with one another, and suppose also that we
can demonstrate that this really is an innovation in the community
(like the glottal stop in Belfast). The question is how we are to
determine whether and in what manner the innovation will feed into
 Speaker-innovation and Linguistic Change 203

the system as a patterned change. It is clear that the structural

conditions exist in many languages for palatalization of initial /k/ to
take place: in some languages (or dialects) we can observe that it has
taken place, whereas in others it has not, but has remained at the
stage of a sporadic innovation that is always a potential change. At
billions of moments throughout history, the change has been possible
in these varieties, but has not been realized. This is the point: the
likelihood of any specific event occurring at any given place or time is
close to zero; therefore, no specific sound-change is ever likely to
happen at any particular time even when favourable structural
conditions exist in the language (for example, when it might be
regarded as ‘natural’), and when there have been innovations which
might favour the initiation of a change. It appears that for the change
to take place it is necessary for the social conditions to be favourable.
Such conditions must have applied to the case discussed by
Andersen (1973)-—the attested change from sharped labials to
dentals in the Tetak dialects of Czech, which Andersen explains as
arising from the acoustic similarity between sharped labials and
dentals. The difficulty that our account raises here is the question
why this change, which was always possible for these acoustic
reasons, took place at the time and place that it did, and not at other
times or places. The acoustic similarity adduced by Andersen defines
the Tetak innovation as one of a class of innovations that must have
been possible candidates for change at any time or place, so long as
the linguistic conditions (in this case, acoustic similarity) existed
within the language system. The historian’s problem, of course, is
that in historically attested, but highly specific, instances like this, we
cannot have sufficiently precise social information to explain the
social conditions that favoured the change and then (in the nineteenth
century) helped to bring about its recession. But it is unlikely that
social identity factors were not involved, as Andersen’s subsequent
work would itself suggest (for example, Andersen, 1986). In socio-
‘linguistic inquiries into present-day states of language, we can pursue
these social questions more fully.
The distinction between innovation and change may have further
relevance to some general assumptions that linguists commonly make
about the origin of language changes, for example the proposal that
the locus of change is in child language-learning phenomena. One
difference between this and a sociolinguistic or discourse-oriented
model is in the nature of the information available. Whereas social
204 Speaker-innovation and Linguistic Change

dialectologists can demonstrate the patterning of change in progress

in adolescent and adult populations by large-scale investigations and
quantitative techniques, mentalist accounts depend on more idealized
data and on theoretical positions that take little note of heterogeneous
databases. It would clearly be very difficult to demonstrate, beyond
reasonable doubt, which of the many innovations observed in child
language (for example) will actually be accepted by speech commun-
ities and become linguistic changes, as most of the innovations
observed in such circumstances (as in others) will never become
changes. We are thinking here of present-day communities and
assuming that innovations can be directly observed in them, but as so
much of our knowledge has depended on history, it should also be
remembered that phonological innovations cannot be observed in
historical data; what historical data normally display are changes at a
late stage of development— usually late enough to be accepted into
writing systems and message-oriented styles. Thus, much historical
linguistic argumentation has depended on the observation of com-
pleted changes and not directly on the origins of innovations. As for
the Neogrammarian axioms, if we accept that a change is not a change
until it assumes a socially regular distribution of some kind, we have
no criteria for determining absolutely that there is an axiomatic
distinction between sound-change and borrowing (or contact change)
because, as we have defined them, all changes must arise from
contact between speakers. But we can certainly save the Neogramma-
rian hypothesis that change is regular, provided we allow social
regularity to be counted as regular: in this perspective it is innovations
that may have an irregular distribution, and change must be regular
by definition. This implies, amongst other things, that phenomena of
the type that appear to be irregular in an intra-linguistic account may
turn out to be regular in relation to a more fully developed
sociolinguistic model of social structures, processes and relationships.
Finally, we need to make a few comments on the social model of
change proposed here and some suggestions as to how it might be
extended. ‘The weak-tie model is not in itself sufficient to provide a
full social explanation of linguistic change. What it proposes is a set of
conditions that are necessary —but not sufficient — for linguistic
change to take place. There are certain things that are important
socially about which social network has nothing to say. It is not about
psycho-social attitudes to language or about language-learning pro-
cesses, and so it has nothing directly to say about these things.
 Speaker-innovation and Linguistic Change 205

Similarly, it is not about social stratification, and so it has nothing

directly to say about that either. In order to make progress towards a
fuller account of the social embedding of language change, we have
turned our attention to exploring the connection between network
and wider patterns of social structure (L. Milroy and J. Milroy,
forthcoming). As we have noted above, ethnographers, such as
Mewett (1982), have suggested that social stratification is observed to
develop in cohesive communities to the extent that the strong ties are
weakened. This is where we also make the link: in the Belfast
outer-city research, we hypothesized that movement away from the
core vernacular would be associated with this decline in close-knit
networks. It may therefore be possible to show a link between the
interpersonal concept of network and the broader dimensions of
social structure that are often spoken of in terms of social class.
However, it also appears that whereas a network approach can be
approached through the idea of shared values and consensus in social
groupings, social class (together with other stratificational concepts
such as rank or status), if it is to be fully understood, involves conflict
and inequalities of power. It is reasonable to point out that the social
models on which sociolinguists have relied have not always been
noticeably sophisticated, and there has been much vagueness in the
use of important socially-based concepts such as social class, lan-
guage standardization and prestige. In order to reach a fuller
understanding of social aspects of language change, therefore, we
need a more sophisticated understanding of social structures and
relationships. In chapter 7 I shall pursue this point: I shall be
concerned there with the relationship between social network and the
phenomena of power, status, class and prestige in society.
 7
Towards an Integrated Social Model
for the Interpretation of Language
Change

7.0 Introduction: Two Scenarios for Language Change

In chapter 6 I proposed a condition, based on the social network

model, which is a necessary condition for linguistic change to take
place: this is that there must be weak ties in a population through
which influence can pass from one close-knit group to another. It is
assumed that (relatively) weak ties exist in all situations and that the
maximal strong-tie scenario, in which density and multiplexity are
100 per cent for all members, is an idealization and does not exist in
reality. There are two broad types of scenario that do actually exist in
reality: in scenario 1, the group has some external contact (through
weak ties) but is resistant to external influence: it is norm-maintaining
and ‘conservative’. The changes that do take place in this scenario,
therefore, may include changes that tend to differentiate the in-group
from external contact groups rather than changes that make it more
similar to them, and the effect of such changes may be to maintain
and assert the separate identity of the in-group. Andersen (1986)
discusses situations that seem to be like this, in which the dialects
undergo ‘exorbitant’ internal changes seemingly as acts of separate
identity. Here, the influence of external groups is negative to the
extent that the in-group moves away from them, and we can speculate
that the number of weak ties in this scenario may well be fewer than in
scenario 2. In scenario 2, the group admits external influence more
readily, and contact-changes result. In Andersen’s terms, these two
 The Interpretation of Language Change 207

scenarios are those of the endocentric and exocentric speech communi-

ties.
In any given case, of course, the two scenarios may be mixed, in
that the group may admit some external influences and resist others,
and a group that is norm-enforcing at one period of time may become
norm-changing at another (and vice versa). What we need to do,
therefore, is to explain the different social conditions under which the
alternative strategies of scenarios 1 and 2 are adopted by different
groups, in different places or at different times. We have attempted to
make progress here by using the social network model, but it appears
that this in itself is not sufficient to account satisfactorily for all
situations: it needs to be supplemented by something else. This final
chapter is intended merely to suggest how we might go about
supplementing the model, and it is based on current work (L. Milroy
and J. Milroy, forthcoming).
The need to extend the model is partly a result of methodological
difficulties in measuring weak ties, on which I commented in chapter
6.4. Social network is most readily operationalized to account for
solidary, strong-tie patterns, which normally contribute to language
maintenance rather than change and which involve consensus on
norms rather than conflict. But even if we could measure weak ties
satisfactorily, there would still be inadequacies, because there are
certain things that are socially important which lie outside a
consensus-based model of this kind. These have to do with the
broader structures of society, involving patterns of social stratifica-
tion, or of social rank, class, status, power or economic wealth. But
although it would be difficult, if not impossible, to include social
network systematically in an integrated model alongside certain other
matters of importance (such as psycho-social attitudinal studies or
language-acquisition studies), it does seem to be feasible to work
towards an integrated sociolinguistic model of social network and
social class.

7.1 Consensus and Conflict

Social network has sometimes been represented in sociolinguistics as

an alternative model to social class and considered to be incompatible
208 The Interpretation of Language Change

with it. Because of this it may seem to be surprising that I should

propose an integrated model here. However, if we are to move
towards a fully-fledged social model, it is clearly necessary to relate in
some way the study of micro-level social relationships (as, for
example, accessed through social network) with the study of the wider
social structures within which these micro-level relationships are
enacted. For it must be made clear that the forms that various
network groupings take are not arbitrary: they are in some sense
dependent on the wider organization of society. The particular type of
solidary network structures that is found in the shanty towns studied
by Lomnitz (1977), for example, does not spring up for no reason: it
is related to socio-economic inequalities at a macro-social level,
which make it necessary for people to form solidary groupings for
mutual support, protection and even survival. The purpose of this
chapter is to suggest what an integrated model might be like, and to
examine the consequences of adopting such a model for interpreting
attested language changes.
We can make a start here by relating our own findings on the
relationship between social network and language variation (as
discussed in chapters 4 and 6) to a broader theory of social class
which can account for larger-scale social structures and processes.
But this does not amount to accepting the conceptualization of social
class that is current in much contemporary sociolinguistics. I think we
can agree with Kathryn Woolard, who has recently commented on the
integration of sociolinguistic and social theory; her view is that much
of the social theory implicitly adopted by sociolinguists is in need of
explicit formulation and critique: ‘. . . sociolinguists have often bor-
rowed social concepts in an ad hoc and unreflecting fashion, not
usually considering critically the implicit theoretical frameworks that
are imported wholesale along with such convenient constructs as
three—, four—, or nine-sector scalings of socioeconomic status’
(1985: 738). Other investigators (such as Rickford, 1986) have
expressed broadly similar views, and of course the avoidance of social
class measures in the Belfast inner-city research was based on a
similar critique of my own (J. Milroy and L. Milroy, 1978, and
elsewhere).
In developing a suitable social model, we start with sociolinguistic
findings and then look for a coherent theory which accounts for them.
Proceeding in this way allows us to make principled decisions about
the kind of framework that seems to be required, without binding us
to a particular social class model that has been superimposed
 The Interpretation of Language Change 209

beforehand as a means of organizing the data. But we do not claim to

have found the ideal social class model; we are simply attempting to
integrate existing findings and suggest the kind of social model that
seems to be practical and capable of giving insights. However, many
sociolinguistic findings suggest that the kind of social class model
required differs quite sharply from that underlying Labov’s work,
which is most clearly manifested in his New York City study, but still
influential in interpreting at least some of the findings of his more
recent work in Philadelphia. Labov’s definition of speech community
(his key sociolinguistic notion) depends, as we have seen, on the idea
of shared values throughout the community at all levels of social
structure. But as the community shape implicit in Labov’s analysis is a
stratified shape with agreement across the different strata, it assumes
a consensus model of the type associated with Parsons (1952), in
which the community is envisaged as fundamentally cohesive, and
where speakers are said to agree on the evaluation of the very
linguistic norms that actually symbolize the divisions between them.
My own sociolinguistic research started with the view that divergence
and conflict are synchronically more salient than consensus, and I
think the findings of much sociolinguistic research now give support
to this view. The reality and persistence of non-standard vernacular
communities uncovered by many researchers, constitute evidence not
primarily of consensus, but of conflict and sharp divisions in society.
This view seems to be shared by some others: for example, Rickford’s
work on Guyanese Creole has led him to call for more attention to
conflict models of class (Rickford, 1986). And despite the character-
istic focus of Labov’s work on shared values, support for a conflict
model of society is provided by some of the Philadelphia findings
showing progressive segregation and linguistic differentiation be-
tween black and white networks (Labov and Harris, 1986). Further-
more, a conflict-based social theory seems to be essential if we are to
account for the phenomenon of linguistic change, because if values
-are fully shared throughout a community there will be no reason to
change them. In fact, pointing to the social conflict associated with
linguistic change, Labov himself has suggested that ‘a thorough-
going structural-functional approach to language could be applied
only if linguistic systems did not undergo internal change and
development’ (Labov 1986: 283).
From these points it seems reasonable to conclude that a social
class model based on conflict, division and inequality will account
best for patterns of language variation at the macro-ievel. This has
210 The Interpretation of Language Change

consequences not only for interpretations based on shared values

(following Labov), but also for the idea of the linguistic market
(Bourdieu, 1977). This has been widely used (see especially Sankoff,
et al., 1989), the argument being that language represents a form of
social and cultural capital, which is convertible (in varying degrees)
into economic capital. Dittmar, Schlobinski and Wachs (1988)
provide a useful exposition of the concept in relation to their analysis
of Berlin vernacular. Woolard (1985), however, has discussed
standard/vernacular opposition in terms of alternative linguistic
markets, rather than adopting Bourdieu’s view of a single dominant
linguistic market where the rule of the ‘legitimate’ language is merely
suspended by the ‘vernacular’ user. This is our view also: we have
argued that the use of the vernacular cannot be seen simply as a
relaxation of tension, a temporary absence of domination of the
standard, as Bourdieu posits. Indeed, much of the argumentation of
this book has suggested that standard languages are not ‘normal’
states of affairs, and that variability is normal and primary. A major
reason for rejecting Bourdieu’s assumptions about the standard is
that much sociolinguistic work (including our own) has shown that,
just as there is strong institutional pressure in formal situations to use
varieties approximating to the standard, so also effective sanctions are
in force to promote ‘vernacular’ use in non-standard domains. The
vernacular in this view is a positive force in itself, and if you study
highly divergent vernaculars, you cannot help being impressed by the
sheer irrelevance of the ‘standard’ or the ‘legitimate’ language in
many situations (for some relevant comments, see chapters 3 and 4).
The orthodox account of the linguistic market appears to offer a
unidimensional analysis of the speech community, in which the
control of the ‘legitimate’ language is basic. Woolard (1985),
however, suggests that much recent sociolinguistic work which has
concentrated on competing social values using a bi-dimensional
analysis — such as one based on contrastive status and solidarity (we
specifically developed such an analysis in inner-city Belfast) — offers a
particularly promising bridge between sociolinguistic and_ social
theory. Within this status (power)/solidarity framework, close-knit
social networks are seen as mechanisms enabling speakers to main-
tain a vernacular code, which itself constitutes an actively con-
structed, symbolic opposition to the ‘dominant’, legitimized code. But
it should also be borne in mind that the dominance of a given
legitimized code is not necessarily maintained through time. It is
 The Interpretation of Language Change 211

quite common for forms and varieties that were at one time
non-standard to take over as the basis of legitimized codes and, on the
other hand, for legitimized forms to recede and/or become vernacu-
lar forms. Unidimensional models (including standard-based model-
ling of the linguistic market) do not offer much help in explaining
these historical changes.
It is these patterns of symbolic opposition (present or past) that
make it necessary for us to to move beyond network and look for an
appropriate model of wider social structure, because, as Gal (1988)
argues, the success, persistence and precise form of the symbolic
opposition will depend not upon internal linguistic or interactional
factors, but upon the relation of the resisting group to the national
economy and to like groups in other states. The outcome in terms of
language survival or shift in Belfast may be different from that in
Brighton; in Catalonia different from Gascony. It will necessarily be
constrained by local historical contingencies.
So far I have tried to outline some general prerequisites for a
socially coherent sociolinguistic theory, constructed to take account of
sociolinguistic findings — particularly those that have emphasized the
strength and persistence of local vernaculars in opposition to legiti-
mized languages. In order to put this together, we need to recall
briefly the chief principles underlying a network analysis of language
variation, looking first at close-knit communities and then at more
loose-knit network situations. I shall attempt to show that the
structure and social function of both types of situation need to be
considered if we want to integrate a network model with a sociolin-
guistically and socially adequate model of social class.

7.2 Close Ties in Rural and Urban Communities

Dialectological research has generally drawn a sharp distinction

between rural and urban dialectology, and in some cases (such as
Andersen, 1986) between ‘central’ and ‘peripheral’ dialect situations.
Before we go on to suggest a way of integrating class and network, it
seems necessary to comment on the relation of social network to these
distinctions. Fundamentally the concept of network is neutral with
regard to such situations, but it is sometimes felt that close-knit
networks are receding under the impact of modern urban life, and
22 The Interpretation of Language Change

that the analysis of close-knit network situations will therefore

become less and less relevant to the explanation of observed linguistic
and social variation. This is probably not correct.
It has been argued (for example, by Wirth, 1938) that urban
conditions give rise to impersonality and social distance. But this
‘urbanist’ thesis does not contradict the model I have suggested: it fits
in with the arguments in chapter 6 on the role of weak ties (urban
communities being a sub-type of contact situation where numerous
weak ties are bound to arise). However, weak ties do not tell the
whole story about urban life. It may well be true that the Italian
American ‘urban villagers’ described by Gans (1962), or the close-
knit mining communities described in Yorkshire by Dennis, Henri-
ques and Slaughter (1957), are less salient in American and British
cities than they once were. But there are still close-tie situations in
cities, as these traditional communities are gradually replaced by
similar types of community created by newer immigrants. More
importantly perhaps, as Giddens (1989) points out, neighbourhoods
involving close kinship and personal ties seem to be actually created by
the conditions of city life. It is easy to understand that when people
settle in a strange (and possibly frightening) environment, they will
seek security by developing ties with people of their own ethnic and
cultural background — people they understand. And as Fischer’s work
(1984) suggests, it is in large cities rather than in small towns that
cultural diversity will be enabled to develop in this manner.
It seems, therefore, that those who form part of urban ethnic
communities will gravitate to form ties with, and often to live with,
others from a similar linguistic or ethnic background. These ethnic
groups use the close-knit network as a means of protecting their
economic and social interests while their community develops the
resources to integrate more fully into urban life. For example,
differences in the network structure of members of the Chinese
community in Newcastle upon Tyne correlate with different patterns
of language choice and with different levels of integration into
non-Chinese domains of urban life (L. Milroy and Li Wei, forthcom-
ing). Bortoni-Ricardo (1985) makes a similar point with regard to
rural migrants into Brasilia. Thus, the type of close-knit community
that is most easily conceptualized in (close-tie) network terms is as
likely to be a product of modern city life as it is to be a residue of an
earlier type of social organization. It is an instance of a recurrent
phenomenon —a phenomenon favoured by the conditions of urban
 The Interpretation of Language Change 213

life and therefore likely to renew itself cyclically as new populations

move in —and of course these close-tie communities are functional
for their members in that they offer security and mutual support.
These close-knit groups are important in providing a focal point
for ‘stigmatized’ urban vernaculars and other non-legitimized lin-
guistic norms, and so need to be accounted for in any sociolinguistic
theory. That is why some form of network analysis that examines the
relationship between the individual and the primary group is so
important. But the indicators of strength and intensity of network tie,
which is the variable relevant to the kind of analysis we are discussing
here, will vary in kind with community organization. For example,
membership of religious groups might be irrelevant in a contempor-
ary northern English coal-mining community, but highly relevant in a
Midlands black community.
The link between network and class, however, is not provided by
strong ties, but by weak ties. As I have suggested in chapter 6
(following J. Milroy and L. Milroy, 1985b), those speakers whose
network structures are least close-knit are also less likely to appro-
ximate closely to conservative vernacular norms of language and are
more exposed to external influences, which in many cases will
be — relatively speaking — standardizing influences. The link between
network and class pointed out by Mewett (1982) suggests a route for
constructing a two-level sociolinguistic theory, linking small-scale
structures such as networks, in which individuals act purposively in
their daily lives, with larger-scale and more abstract social structures
(classes) which determine relationships of power at the institutional
level. The relationship between weakness of ties and the development
of class structure also gives us a consistent explanation for the

Solidarity

Status

Solidarity

Figure 7.1 The anvil-shaped status/solidarity model.

214 The Interpretation of Language Change

openness to innovation of the socially mobile middle groups in the

social hierarchy, as against the conservatism of the uppermost and
lowermost groups. During the Belfast projects I suggested a status/
solidarity model measured against social class, which gives us the
anvil-shaped diagram of figure 7.1. In this perspective, network-
based and class-based analyses are not contradictory (as is sometimes
suggested); rather, they complement each other. Figure 7.2 is a more
elaborate attempt to indicate the pattern of relationships that are
likely to exist in the community in terms of both strong and weak ties.

7.3 Towards an Integrated Model: The Concept of

Life-modes

I commented above that the findings of sociolinguistic research,

particularly small-scale network studies of ethnic minority and
working-class groups, constitute evidence of conflict and sharp
division in society, rather than of cohesion. We also commented
above that we need a social theory that links our network analysis to
an analysis of social structure at the political, institutional and
economic level. Such a link may be provided by ‘life-mode’ analysis
(Hgjrup, 1983), which we develop further below.
What is required here is a model representing various ethnic and
class groups as both internally structured and connected to each other
with varying proportions of strong and weak ties, which we have
attempted to show in figure 7.2. For example, ethnic subgroups in
Britain such as the black speakers studied by Edwards (1985) have a
predominantly strong-tie internal structure, but seem to be linked by
relatively few weak ties to white working-class groups. These white
groups in turn might have a similar internal network structure but
have more weak tie links with other white working-class groups.
Vertical links to middle-class groups might be fewer (this seemed to
be the case in Belfast) and moreover to be frequently institutional to
such persons as doctors, lawyers, teachers, welfare personnel and the
like. Middle-class groups for their part — professional, neighbour-
hood and friendship groups — are characterized by a higher propor-
tion of weak ties internally than working-class groups; hence the
problems of studying them systematically in network terms in Zehlen-
dorf—a middle-class suburb in Berlin (Dittmar, Schlobinski and
 The Interpretation of Language Change JAS
Maintenance of Dominance of
non-legitimized legitimized
linguistic code linguistic code

| |
Community-based ties "

Non-community-
Life-modes 1 —»| based strong ties
(coalitions, power
elites etc.)

Macro level social, political

and economic structure

Figure 7.2 Macro and micro levels of sociolinguistic structure.

Wachs, 1988) — and in outer-city Belfast. Thus, however we interpret

the concept of class, and however we model these localized networks,
we can use Granovetter’s concept of the weak tie to link close-knit
community level groupings to more abstract institutional structures.
In the model we are developing, the behaviour of speakers is
attributed to the constraining effects of the network, or the diminu-
tion of those effects which enables supra-local norms (including the
legitimized language) to permeate networks, rather than to any direct
effect of ‘prestige’ as defined by the perceived attributes of speakers
who are seen to ‘belong’ to different status groups. Social class is
viewed here as a structural concept, not as a set of labels which might
be attached to particular individuals. Thus, local and individual social
behaviour is not seen as directly related to class but as mediated
through these smaller-scale structures. However, while network
analysis can delineate various economic, political and subcultural
groupings in society, it cannot say anything about the varying
potentials of such networks to exercise the economic and political
power that is the source of conflict and inequality in society. In
216 The Interpretation of Language Change

linguistic terms, of course, this means that powerful networks have

the capacity to impose their linguistic and cultural norms on others,
while powerless ones do not, but can merely use the resources of the
network to maintain and at best renew their own linguistic and
cultural norms.
This last comment takes us back to the discussion of norms and
prescription in chapter 1. As we noticed there, Jones’s English
Pronouncing Dictionary (1955) is an account of RP which, although it
claims to be descriptive and not prescriptive, is nonetheless used by
readers as a prescriptive guide. If it is used in this way, however, this
does not arise from Jones’s intentions, but from the social factors I
have just alluded to. If the accent described were not RP, but Belfast
inner-city vernacular, the description would not be used prescript-
ively. Like other accents, RP is a network accent, and like other
accents it is also a badge of identity. But there is one overriding
difference: RP is also the possession of a network that has the greatest
socio-economic power, and that is why some socially mobile persons
(mainly in life-mode 3, below) may wish to acquire it.
We cannot really understand all this without reference to the
broader structures of society, and for this reason we need a social
theory that links a network analysis of subgroups within society to an
analysis of social structure at the political, institutional and economic
level. These two levels correspond to what Gal (1988: 247) describes
as ‘the interactional and socio-political level of analysis’. The most
helpful analysis of this kind we have encountered is that of Hejrup
(1983), who proposes a division of the population into subgroups
which are described in terms of three life-modes.' These life-modes
are seen as necessary and inevitable constituents of the social structure
as a whole. Hojrup’s conception of this larger social structure is
Marxist, and the initial analysis is in terms of modes of production
and consumption. Thus, crucially, these subgroups are not seen as
socially or culturally arbitrary, but as the effect of ‘fundamental
societal structures which split the population into fundamentally
different life-modes’ (Hojrup, 1983: 47). The analysis is particularly
useful from our point of view, since the different types of network
structure that we have distinguished emerge naturally from differ-
ences in the life-modes of different individuals. Although the argu-
mentation in support of his analysis is lengthy and complex, Hojrup
uses a limited number of straightforward concepts to distinguish the
three life-modes, and we shall look at these briefly. Life-mode 1 is
 The Interpretation of Language Change 247

the life-mode of the self-employed, life-modes 2 and 3 of two

different types of wage-earners. Of critical importance is the ideolo-
gical orientation of the three groups to work, leisure and family. We
shall focus a brief description of each of them on points of contact
with our network analysis.
Life-mode 1 Here Hojrup is thinking of any simple commodity
producing unit, be it agriculture, fishing, a corner shop or a
restaurant. In this life-mode, social relationships in the form of family
ties or co-operative relations among colleagues bind the producers
into a cohesive production unit. The primary concern is to keep the
production rolling, and all the family and other affiliated producers
are involved in this. The purpose of the enterprise is to be able to
remain self-employed, a means that is its own end. The concept of
free time has little meaning in this life-mode, since the producer is
not put to work but puts himself to work to gain independence. This
is done to the extent that it is necessary and beneficial for himself and
his family to maintain or expand their enterprise. The concepts of
‘leisure’ and ‘work’ thus have a totally different meaning from that
which they assume for wage-earners, and it is clear that a close-knit
type of network structure and a solidarity ethic will be associated with
this life-mode, which itself follows from the type of economic activity
in which the producers engage. Hgjrup does not see this kind of
life-mode as a relic of an earlier period (see my comments, above, on
close-knit networks in modern cities) but as highly efficient and
competitive given its flexibility of operation and the commitment of
the producers. He uses the Danish fishing industry as an example,
but his description equally well applies to Chinese family restaurant
businesses, Pakistani corner shops, or small painting and decorating
businesses in Britain.
Life-mode 2 Wage-earners are different from life-mode 1 com-
modity producers, in that they are incorporated in a long and complex
process of production which they do not own or control. Life-mode 2
is that of the ordinary wage-earner, the purpose of whose work is to
provide him with an income that will enable him to live a meaningful
life during his free time. The family differs from life-mode 1 families
in being separate from his work activities; it is the framework within
which non-productive leisure activity takes place. The life-mode 2
worker lacks the commitment to his work characteristic of life-mode
1. He is prepared to sell his labour, thereby becoming mobile and
‘severing existing close-knit network ties if there is an adequate
218 The Interpretation of Language Change

inducement to do so. If wages are low, however, he has to demand

enough to survive. Hence, the solidarity that arises amongst workers
who earn little — a solidarity reflected at the institutional level in the
establishment of trades unions. At a neighbourhood level this
solidarity is embodied in the close-knit networks of the traditional
working-class society of the kind we studied in Belfast. Following
through Hogjrup’s analysis, we would surmise that the solidarity ethic
would collapse and network ties become loose-knit if economic and
political conditions allowed workers to feel secure in their future
prospects, if they earned enough to become mobile, to buy better
houses and cars, to take holidays abroad and so on. There do in fact
appear to be differences of this kind in behaviour between different
groups of wage-earner, as we noted in our analysis of the outer-city
versus inner-city areas in Belfast. Lockwood’s (1966) classic investi-
gation of images of class structure in Britain fits in broadly with such
an analysis, particularly in its distinction between the outlooks
described respectively as proletarian traditionalism and privatized
worker (see Giddens 1989: 224). Privatized workers, exemplified by
the Luton car workers studied by Goldthorpe et al. (1968-9), live
apart from traditional working-class areas in the suburbs, and see
work as a way of achieving a satisfactory life style for themselves and
their families. They apparently reject the traditional working-class
solidarity ethic, but given a certain level of grievance it tends to
reinstate itself as does the us/them, insider/outsider imagery charac-
teristic of close-knit communities and traditional proletarian ideology.
The persistence and renewal of this imagery (and associated network
structures) seem to spring from changes in economic and power
structures in society.
Life-mode 3 The life-mode 2 wage-earner performs the routine
tasks of the workforce at a given daily or hourly rate. The life-mode 3
wage-earner is, however, a higher professional or managerial
employee with a high level of skill which is itself a saleable commod-
ity, and he or she is paid to arrange, monitor and control the
production process. Typically, the concept of work and leisure and
the role of the family are in sharp contrast to those of life-mode 2.
The life-mode 3 goal is to rise up through the hierarchy, to obtain
control through managerial and professional roles, to exercise more
and more power and ultimately escape from the control of others so
as to control resources and exercise power on one’s own account.
This process demands an immersion of the individual in work, a
 The Interpretation of Language Change 219

competitive attitude to colleagues and a blurring of the boundaries

between work and leisure. The family and its way of life fulfils a
supportive role in relation to the career. Work therefore is life to a
high degree, and the concept of freedom is not one of free time but is
associated with the work situation and the career perspective. This
will be uncomfortably familiar to those who, like me, are university
teachers.
Just as different types of network structure emerge from the
economic conditions associated with life-modes 1 and 2, so a certain
type of personal network structure seems to follow from life-mode 3.
These wage-earners will be socially and geographically mobile as they
pursue their careers, forming many loose ties, particularly of a
professional kind, through which innovations and influence may be
transmitted. However, they will also form close-knit clusters and
coalitions within their personal networks, through which they control
considerable resources. This seems to fit in with our general
characterization of the differing role of loose-knit and close-knit
network ties; the primarily loose-knit network of the life-mode 3
individual ensures that the dominant linguistic market holds sway
without hindrance from (in Woolard’s (1985) terms) alternative
vernacular markets. Thus, to the extent that it is true that the
- standard language influences speakers, it will be life-mode 3 speakers
who are open to that influence, and not the economically powerless
people that we studied in inner-city Belfast, for whom the standard
language is almost irrelevant. In a totally different dimension, and
recalling my comments on the consequences of standardization in
chapter 5, I find it very tempting to suggest here that orthodox
historical linguistic description has been dominated by the perspect-
ive of life-mode 3 individuals, who have believed their own attitudes
to language to be based on ‘common sense’, and have not been aware
that their perspective is in conflict with that of other sectors of society.
That is why we have the emphasis on dominant linguistic markets or
dominant dialects, the focus on ‘shared values’ throughout the
community, and repeated unanalysed appeals to the concept of
‘prestige’.
It is important to emphasize that the concept of life-mode, like that
of social network, is a structural one. People cannot be neatly slotted
into pre-determined life-modes like letters into pigeon-holes. ‘Their
ideological and cultural characteristics will be determined by their
‘contrast to the other life-modes in the social formation. The
220 The Interpretation of Language Change

interrelationships between the three life-modes and the cultural

practices associated with each one will therefore take different forms
in, for example, Denmark, Ireland, England and Germany. In each of
these countries the three fundamental modes of production which the
life-modes reflect ‘will appear in different variants and in different
combinations of opposition and independence’ (Hagjrup, 1983).
One consequence of this chain of dependence running from
political and socio-economic structure through life-modes to network
structure and ultimately to sociolinguistic structure, is that close-knit
networks will be associated with life-mode 2 individuals in some
states more than in others. This seems to be the case if we compare
Belfast with Copenhagen, where wage-earners are apparently more
mobile and prosperous and less inclined to organize themselves into
close-knit groups of the kind described in Belfast (Gregersen and
Pedersen, forthcoming). This in turn will give rise to locally contingent
sociolinguistic patterns, with urban vernaculars varying in their
degree of ‘focusing’, opposition to mainstream norms, and general
vitality. But to analyse the life-modes we need to look simultaneously
at cultural practices and specify more completely the modes of
production that give rise to them. It seems, however, that Hojrup’s
analysis has the capacity to help us develop a socially sensitive
sociolinguistic theory which links the macro and micro levels of
analysis in a coherent way. It is offered here as a way forward towards
an integrated social model of language variation and change, and it is
discussed more fully in L. Milroy and J. Milroy (forthcoming).

7.4 Conclusion: Towards a Historical Sociolinguistics

In this chapter I have tried to build on our social dialectological

research and further developments in the social network model in
order to suggest how a broader social framework for the interpreta-
tion of language change might be constructed. There are other
conceivable social models, some of which might be complementary to
this and some of which might be alternatives to it. What this model
does is to indicate the kind of relationship that might exist between
micro-sociological concepts and macro-sociological concepts as they
relate to interpreting language variation and change. There can be
little doubt that some form of integrated social model is necessary.
 The Interpretation of Language Change 221

Even from within sociolinguistics there have been constant com-

plaints about the naiveté of the social underpinning of the subject (a
recent one is Cameron, 1990). In interpreting data, such terms as
class, prestige and standard language are frequently used but seldom
adequately defined, with the result that there is much vagueness,
ambiguity and circularity in the use of these terms in explanation.
The view of historical sociolinguistics that I have developed in this
book is, like other approaches (for example, Romaine, 1982a),
variationist. As language is variable at all times, an adequate historical
account must be an account of changing patterns of variation.
However, the approach advocated here is also much more social in
nature than any other sociolinguistic approach (as far as I know). The
changing patterns of language are not seen exclusively as changes in
the linguistic shape of a language variety, but also as changes in social
agreement on the linguistic norms of communities. In this account,
much more attention is given to the social side than is given in the
Labov paradigm, which is mainly system-oriented, as I have tried to
show, despite its strong emphasis on fieldwork methods. Other
approaches also are less social than ours. Romaine (1982a), for
example, proposes a version of ‘socio-historical linguistics’ which is
mainly system-oriented and concerned with philosophy of science
issues of an abstract idealizing kind, rather than the experimental
‘realist’? brand of philosophy of science that has influenced our work
(Plutchik, 1974). Similarly, although John Harris worked with us in
Belfast, his treatment of similar data (Harris, 1985) is also less social
and less speaker-oriented than ours.
The main theme that I have tried to develop in this book is that a
strongly social approach is justified because linguistic change is a
social phenomenon. It is negotiated by speakers in face-to-face
encounters, and an innovation in a speaker’s output is not a linguistic
change until it has been agreed on and adopted by some community
of speakers, however small the community may be. By further
developing our social model of language change we can hope in time
to reach a better understanding of the conditions in which innovations
are sometimes adopted as changes and sometimes rejected, and
models of social identity (Le Page and Tabouret-Keller, 1985),
accommodation (Giles and Smith, 1979) and politeness (Brown and
Levinson, 1987) will not be irrelevant to these developments.
In the meantime we may wish to consider how far we can be
satisfied with mainly mentalist or language-internal explanations for
LLL The Interpretation of Language Change

language change, and within sociolinguistics we may also wish to

consider more carefully than has been usual what is really meant
when we appeal to concepts like class and prestige as explanations for
attested changes. Finally, within the tradition of historical linguistic
theorizing, our social approach seems in certain respects to supersede
some of the axioms of the Neogrammarians. From a social point of
view, linguistic change, when it is carefully observed by accountable
methods, is always gradual.
A major task, therefore, is to link linguistic change with social
change in such a way as to explain the conditions under which
linguistic change takes on particular patterns, including patterns of
rapid and slow change. This is a very different task from that of
orthodox historical linguistics, and I hope that I have shown in this
book that it is a very exciting challenge.
 Notes

Chapter 1 Introduction: Language Change and Variation

1 I have in mind here languages such as ancient Greek, Latin or Sanskrit,

the classical forms of which survived the demise of native speakers. It is
true that they developed into vernacular languages and were used for
ritual purposes, but in these later forms they displayed changes only
when they had speakers or writers.

Chapter 2 Social and Historical Linguistics

There is evidence that both Danish and English had a higher incidence
of palatization of velars in the past than they have now. For English,
McKnight (1928: 456) lists examples (e.g. cyar, gyet for ‘car, get’) from
1653 until 1791, noting that one writer at least regarded palatalization
as ‘essential in polite pronunciation’. The recession of palatalization
here is just as much a sound-change as its implementation, and a
solution to the actuation riddle would involve explaining cases of
recession as well as cases of implementation of the phenomenon. We
have been able to observe a late stage in the recession of palatalization
in Belfast. See J. Milroy (1981) and chapter 3.
This is particularly true of the Chomskyan ‘idealized native speaker/
listener’.
To give a socially-based explanation of the merger of bird/fern/hurt in
seventeenth-century English, we would need detailed information of a
kind that is not normally directly recoverable from historical sources.
See my comments on the limitations of historical inquiry in section 2.7,
below. However, we can observe similar patterns of merger in present-
224 Notes

day English, particularly in urban dialects: see chapter 5 for some

discussion.
Smith (1989: 181) suggests that it is unlikely that a homogeneous input
(to the language-learner) would make language learning impossible.
Perhaps not quite, but a person who had acquired a first language from
a homogeneous model would surely be perceived as a very odd person.
But it is widely believed by linguists that variation and change are
non-functional. Consider Postal (1968: 283): ‘...there is no more
reason for languages to change than there is for automobiles to add fins
one year and remove them the next... the ‘causes’ of sound change
without language contact lie in the general tendency of human cultural
products to undergo ‘nonfunctional’ stylistic change.’ Manufacturers
normally have good reasons to alter designs that they want to sell to the
public. However, beside the strong feelings that speakers express about
language variation, this intra-linguistic view seems to me to be rather
pallid.
mn Jespersen, of course, did not use the term ‘transparent’.
Vennemann (1989) has recently claimed that language change is
‘meliorative’. This seems to be a non-social view: languages become
better in themselves, independently of society. Vennemann links this to
the idea of evolution: languages are seen as evolving to a better state, in
common with biological evolution generally. As for ‘drift’, which has
recently come to the fore again, one problem with this, as possibly with
some of Vennemann’s ideas also, is that it can only be discerned
retrospectively.
I discussed some of these in J. Milroy (1978). They were subsequently
further discussed in the wider context of conversational breakdown by
L. Milroy (1984).
One of the most naive assumptions of early transformational-generative
theory is the idea that different dialects of English can be derived from
the base form by the addition of relatively ‘low-level’ or ‘late’ rules. This
was still being seriously proposed as recently as 1981 (Smith and
Wilson, 1981).
I have used the distinction between message-oriented and listener-
oriented speech as a very general distinction here. There is no
suggestion that it would be possible to define every part of a conversation
as belonging to one category or the other. One might also adduce
Bernsteinian distinctions here — for example, between context-free and
context-sensitive discourse.
10 In J. Milroy (1982a) I have discussed the absence of Verner’s Law
voicing from the Gothic strong verb, where according to the theory it
should occur. There are two alternative traditional explanations for this
(see Wright, 1954: 369-70). In my paper, I suggest a third explanation,
 Notes 225

which fits the facts equally well. Because the phenomena lie deep in
history, we do not have reliable criteria for choosing amongst these
three explanations, and any preference we express is likely to be
influenced by some theoretical consideration. Verner’s own explanation
(analogical levelling), for example, was derived from Neogrammarian
theory.

Chapter 3 Analysing Language in the Community: General Principles

We can say that the conclusion based on this instance of hypercorrec-

tion is ‘warranted’ by the data, very much as conversational analysts use
the term ‘warranted’.
The back /a:/ in RP is so socially salient that people tend to believe far
too readily that backing in other dialects is RP-influenced. For a
rebuttal of this view in the case of Tyneside English, see Beal (1985).
Social network did not form part of the original research design; it was
not mentioned in our application for funding in 1975. Misinterpreta-
tions of the role of social network in the research design have been
common enough to require comment. Sometimes the hypo-
thesis — about language maintenance through normative consen-
sus — with which we started out (before we thought of using the social
network model) is presented as if it were a finding arising after
quantitative analysis. Cameron (1990: 87), for example, says that the
quantitative (network) analysis ‘led Milroy to conclude that people in
her (sic) survey behave linguistically as they did because of the
normative influence of their peer group’, and describes the language/
close-tie hypothesis as a ‘finding’. This is not quite right: our data
analysis supported our hypothesis. Unfortunately, the Belfast work
(which was not mainly a survey) is also misattributed here to Lesley
Milroy, and ‘social networks’ are wrongly equated with ‘peer groups’.
Unlike Labov (1972a) in Harlem and Cheshire (1982) in Reading, we
did not carry out peer-group studies in Belfast. The social network
model covers person-to-person relationships generally, and not solely
within bounded groups. See further Chapter 4.1.
Some additional guidance was in fact available from the prescriptions
made by the elocutionist David Patterson (1860) (see chapters 4 and 5 for
further references to Patterson) and from descriptive work on Ulster
dialect generally by Adams (1964), Gregg (1964, 1972) and others.
It is possible also to quantify variables of a supra-phonemic type: for
example, Rigg (1987) has quantified glottal-stopping and glottal rein-
forcement for three consonants ([p, t, k]) in Tyneside, and it would have
been possible to quantify supra-phonemically in Belfast. Centring
226 Notes

glides, for example, occur under similar conditions for several different
vowels, and several vowels show similarities to one another before [r].
In my judgement at the time, it was possible to deal adequately with
these features descriptively and without quantification.
It is important to note that there is nothing new in using co-variation
arugments in linguistics. Traditional comparative linguistics has been
based on observed co-variation and correlation (without quantification,
of course), and well-known ‘sound-laws’, such as Grimm’s Law, are
based on these perceptions. I have elsewhere discussed the correla-
tional arguments that account for ‘Verner’s Law’ (J. Milroy, 1982a).
The notion of causality tends to get involved here as Germanic accent
shift correlates very closely with Verner’s Law voicing, so there may be
a ‘causal relationship’, but we would be on shaky ground if we claimed
that accent shift was the cause of the voicing. Proposed solutions to
actuation have to recognize that it is multi-causal, and that is why we
look at numerous social and linguistic conditions when we study
sound-change in progress. And, of course, the use of quantification is
irrelevant to any substantive claims that are made on the basis of
co-variation patterns.

Chapter 4 Interpreting Variation in the Speech Community

It is easy to give examples of strongly expressed feelings about language

variants and to show that such views often have a ‘gate-keeping’
function. Edwards (forthcoming) quotes a letter sent to a teacher who had
been rejected for an appointment after an interview in which his use of
‘aspirates’ ([h]) was objected to. She also mentions the stigma against
‘Cockney’ in particular: Walker (1791) describes it as ‘offensive and
disgusting’; Matthews (1938: x) points out that even ‘philologists’ deny
Cockney the status of a dialect and ‘describe it as vulgar speech based
on error and misunderstanding’. In my own experience a university
Dean once objected to the appointment of the best candidate for a
particular post on the grounds that the candidate had a ‘London
accent’. In chapter 2 (note 4), we noticed that despite the strong
feelings that people express about language, linguistic scholars often
believe that variation and change are non-functional (see especially
Postal, 1968: 283).
Horvath (1985) has re-graphed Labov’s New York City data in terms of
sex-difference, and has shown that differentiating in terms of sex works
at least as well as differentiating in terms of class. In fact, Nathan B,
who seems anomalous in terms of class, is ‘normal’ if the data are
presented in terms of sex-difference. Rigg (1987), analysing the
 Notes 227

incidence of glottal-stopping in Tyneside, has presented figures that

show that the use of the glottal stop is related to sex, rather than
class, as reported by L. Milroy (1989):

Working class Middle class

(p) (t) (k) (p) (t) (k)

Males %: os bes rege 4)es Uh ie Be) 20D) Jl | OULD
Females %: 60.0 31.0 28.0 7 Nall eke) evel i wh 8

The term ‘marker’ is used here in a general sense, and not in the more
technical sense used by Labov, which requires that the variable is
marked in terms of both class and style. So it should be borne in mind
that the variables are not selected in the belief that some are indicators,
some markers, and others stereotypes (Labov, 1972b), or that some are
‘low-consciousness’ and others ‘high-consciousness’ variables
(Johnston, 1983). These taxonomies are not involves at this stage, but
we can perhaps regard our procedures as helping to test to what extent
they are valid.
Gregg (1972: 110) speaks of the ‘ubiquitous Ulster “light” [l]’; Jones
(1956: 92) states that ‘in English as spoken in Ireland [I] is always clear’,
and a similar view is expressed by Gimson (1970: 204) and O’Connor
(1973: 149).

Chapter 6 Speaker-innovation and Linguistic Change

The arguments in this chapter about weak ties, speaker-innovation and

diffusion of innovations are largely based on a paper jointly authored by
James and Lesley Milroy, ‘Linguistic change, social network and
speaker-innovation’, which appeared in the Journal of Linguistics in
1985. I am extremely grateful to Lesley for the continuing collaboration
out of which this paper emerged.
Gumperz points out that Labov shares with other linguists an interest in
understanding the character of grammars; a speaker-oriented approach,
on the other hand, ‘focuses directly on the strategies that govern the
actor’s use of lexical, grammatical, sociolinguistic and other knowledge
in the production and interpretation of messages in context’ (Gumperz,
1982: 35). Thus, a display such as the graph of the famous cross-over
pattern for /r/ in New York City is not aimed primarily at accounting
228 Notes

for speaker-behaviour: it is a grammar of change which locates a change

in progress within the system.

Chapter 7 Towards an Integrated Social Model for the Interpretation of

Language Change

1 My thanks to Inge-Lise Pedersen, Copenhagen, for noting the possible

usefulness of Thomas Hogjrup’s life-modes model and for calling our
attention to it in 1987. I am also grateful to Lesley Milroy for the
continuing collaboration on the social aspects of our research and in
particular for her collaboration on the forthcoming paper on which this
chapter is substantially based.
 References

Abdel Jawad 1987. Cross-dialectal variation in Arabic: competing prestigious forms.

Language in Society, 16, 359-68.
Adams, B. (ed.) 1964. Ulster Dialects: a symposium. Holywood, Co. Down: Ulster Folk
Museum.
Adamska-Salaciak, A. 1989. On explaining language change teleologically. Studia
Anglica Posnaniensia, 22, 53-74.
Ahlqvist, A. (ed.) 1982. Papers from the Sth International Conference on Historical
Linguistics (Current Issues in Linguistic Theory, 21). Amsterdam: Benjamins.
Alahdal, H. 1989. Standard and Prestige: a sociolinguistic study ofMakkan Arabic. PhD
thesis: University of Reading.
Alatis, J. (ed.) 1978. International Dimensions of Bilingual Education. Washington DC:
Georgetown University Press.
Andersen, H. 1973. Abductive and deductive change. Language, 49, 765-93.
Reprinted in Baldi and Werth (1978: 313-47).
Andersen, H. 1986. Center and periphery: adoption, diffusion and spread. Paper
delivered to the Conference on Historical Dialectology, Poznan, Poland.
Anderson, J. and Jones, C. (eds) 1974. Historical Linguistics. Amsterdam: North-
Holland Publishing Company.
Ash, S. and Myhill,J.1986. Linguistic correlates of inter-ethnic contact. In Sankoff,
D. (ed.), Diversity and Diachrony, Amsterdam: Benjamins, 33-44.
Bailey, C. J. 1973. Variation and Linguistic Theory. Washington, DC: Center for
- Applied Linguistics.
Baldi, P. and Werth, R. N. (eds) 1978. Readings in Historical Phonology. University
Park: Pennsylvania State University Press.
Beal, J. 1985. Lengthening ofa in Tyneside English. Current issues in Linguistic
Theory, 41. Amsterdam: Benjamins.
Berdan, R. 1977. Polylectal comprehension and the polylectal grammar. In Fasold,
R. and Shuy, R. W. (eds), Studies in Language Variation, Washington DC:
Georgetown University Press, 12-29.
Bertz, S. 1975. Der Dubliner Stadtdialekt. Doctoral dissertation: Albert-Ludwigs-
Universitat, Freiburg in Breisgau.
230 References

Bliss, A. 1979. Spoken English in Ireland. Dublin: Dolmen Press.

Bloomfield L. 1933. Language. London: Allen and Unwin.
Blom, J.-P. and Gumperz, J. J. 1972. Social meaning in linguistic structure:
code-switching in Norway. In Gumperz, J. J. and Hymes, D. (eds), Directions in
Sociolinguistics, New York: Holt, Rinehart and Winston, 407-34.
Boal, F. W. 1978. Territoriality on the Shankill-Falls divide, Belfast: the perspective
from 1976. In Lanegran, D. A. and Palm, R. (eds),An Invitation to Geography, 2nd
edn, New York: McGraw Hill, 58-77.
Boissevain, J. 1974. Friends offriends; networks, manipulators and coalitions. Oxford:
Basil Blackwell.
Bolton, W. F. 1966. The English language: essays by English and American men of letters,
1490-1839. Cambridge: Cambridge University Press.
Bortoni-Ricardo, S. M. 1985. The Urbanisation ofRural Dialect Speakers: a sociolingu-
istic study in Brazil. Cambridge: Cambridge University Press.
Bott, E. 1971. Family and Social Network, revised edn. London: Tavistock.
Bourdieu, P. 1977. The economics of linguistic exchanges. Social Science Information
16, 6, 645-68.
Brown, G. 1982. The spoken language. In Carter, R. (ed.), Linguistics and the Teacher,
London: Routledge and Kegan Paul, 75-87.
Brown, P. and Levinson, S. C. 1987. Politeness. Cambridge: Cambridge University
Press.
Brown, R. and Gilman, A. 1960. Pronouns of power and solidarity. In Sebeok, T.
(ed.), Style in Language, Boston: MIT Press, 253-75.
Brunner, K. 1963. An Outline of Middle English Grammar. Oxford: Basil Blackwell.
Bynon, T. 1977. Historical Linguistics. Cambridge: Cambridge University Press.
Cameron, D. 1990. Demythologizing sociolinguistics. In Joseph,J.E. and Taylor, T.
J. (eds), Ideologies of Language, London: Routledge, 79-93.
Carter, R. (ed.) 1982. Linguistics and the Teacher. London: Routledge and Kegan
Paul.
Catford, J. C. 1957. Vowel-systems of Scots dialects. Transactions of the Philological
Society, 107-17.
Cedergren, H. and Sankoff, D. 1974. Variable rules: performance as a statistical
reflection of competence. Language, 50, 333-55.
Cheshire, J. 1982. Variation in an English dialect: a sociolinguistic study. Cambridge:
Cambridge University Press.
Chomsky, N. 1975. Language and Responsibility. Sussex: Harvester Press.
Clark, C. 1990. The myth of the Anglo-Norman scribe. Paper delivered to the
International Conference on English Historical Linguistics, Helsinki.
Cohen, A. (ed.) 1982. Belonging. Manchester: Manchester University Press.
Crystal, D. (ed.) 1981. Linguistic Controversies: festschrifi for F. R. Palmer. London:
Edward Arnold.
Davenport, M., Hansen, E. and Nielsen, H.-F. (eds) 1983. Current Topics in English
Historical Linguistics. Odense: Odense University Press.
Dennis, N., Henriques, F. M. and Slaughter, C. 1957. Coal is our Life. London: Eyre
and Spottiswood.
Diaconis, P. 1985. Theories of data analysis: from magical thinking through classical
statistics. In Hoaglin, D., Mosteller, M. and Tukey,J. (eds), Exploring Data Tables,
 References px!

Trends and Shapes, New York: Wiley, 1-5.

Dittmar, N. and Schlobinski, P. (eds) 1988. The Sociolinguistics of Urban Vernaculars.
Berlin: Mouton/de Gruyter.
Dittmar, N., Schlobinski, P. and Wachs, I. 1988. The social significance of the
Berlin urban vernacular. In Dittmar, N. and Schlobinski, P. (eds), The Sociolingu-
istics of Urban Vernaculars, Berlin: Mouton/de Gruyter, 19-43.
Dobson, E. J. 1955. Early Modern Standard English. Transactions of the Philological
Society, 25-54.
Dobson, E. J. 1968. English Pronunciation: 1500-1700, 2 vols, 2nd edn. Oxford:
Clarendon Press.
Downes, W. 1984. Language and Society. Bungay, Suffolk: Fontana.
Eckert, P. 1980. The structure of a long-term phonological process: the back
chain-shift in Souletan Gascon. In Labov, W. (ed.), Locating Language in Time and
Space, New York: Academic Press, 179-219.
Edwards, V. 1985. Language in a Black Community. Clevedon, Avon: Multilingual
Matters.
Edwards V. forthcoming. The grammar of Southern British English. In Milroy, J.
and Milroy, L. (eds), The Grammar of English Dialects of the British Isles, London:
Longman.
Ekwall, E. 1975. A History ofModern English Sounds and Morphology. Oxford: Basil
Blackwell.
Fasold, R. and Shuy, R. W. (eds) 1977. Studies in Language Variation. Washington
DC: Georgetown University Press.
Fischer, C. 1984. The Urban Experience, 2nd edn. New York: Harcourt, Brace,
Jovanovitch.
Francis, W. Nelson 1983. Dialectology: an introduction. London: Longman.
Gal, S. 1979. Language Shift: social determinants of linguistic change in bilingual Austria.
New York: Academic Press.
Gal, S. 1988. The political economy of code choice. In Heller, M. (ed.), Code-
switching, Berlin: Mouton/de Gruyter, 245-63.
Gans, H. J. 1962. The Urban Villagers: group and class in the life of Italian-Americans,
2nd edn. New York: Free Press.
Giddens, A. 1989. Sociology. Cambridge: Polity Press.
Giles, H. and St Clair, R. 1979. Language and Social Psychology. Oxford: Basil
Blackwell.
Giles, H. and Smith, P. 1979. Accommodation theory: optimal levels of convergence.
In Giles, H. and St Clair, R. (eds), Language and Social Psychology, Oxford: Basil
. Blackwell, 45-65.
Gimson, A. C. 1970. Introduction to the Pronunciation of English, 2nd edn. London:
Edward Arnold.
Givon, T. (ed.) 1979. Discourse and Syntax (Syntax and Semantics, 12). New York:
Academic Press.
Gleick,J. 1988. Chaos. London: Heinemann.
Goldthorpe, J. H. 1968-9. The Affluent Worker in the Class Structure, 3. vols.
Cambridge: Cambridge University Press.
Granovetter, M. 1973. The strength of weak ties. American Journal of Sociology, 78,
1360-80.
232 References

Granovetter, M. 1982. The strength of weak ties: a network theory revisited. In

Marsden, P. V. and Lin, N. (eds), Social Structure and Network Analysis, London:
Sage.
Gregersen, F. and Pedersen, I. L. forthcoming. The Copenhagen Study in Urban
Sociolinguistics (Universitetsjubilzets Danske Samfund, Serie A). Copenhagen: C.
A. Reizel.
Gregg, R. J. 1964. Scotch-Irish urban speech in Ulster. In Adams, B. (ed.), Ulster
Dialects: a symposium, Hollywood, Co. Down: Ulster Folk Museum, 163-91.
Gregg, R. J. 1972. The Scotch-Irish dialect boundaries in Ulster. In Wakelin, M. F.
(ed.), Patterns in the Folk Speech of the British Isles, London: Athlone Press, 109-39.
Gumperz, J. J. 1982. Discourse Strategies. Cambridge: Cambridge University Press.
Gumperz, J. J. and Hymes, D. 1972. Directions in Sociolinguistics. New York: Holt,
Rinehart and Winston.
Gunn, B. 1982. Prestige lengthening in Belfast English. Paper delivered to the
Sociolinguistic Symposium, Sheffield.
Guy, G. 1980. Variation in the group and the individual: the case of final stop
deletion. In Labov, W. (ed.), Locating Language in Time and Space, New York:
Academic Press, 1-36.
Guy, G. 1988. Language and social class. In Newmeyer, F. (ed.), Linguistics: the
Cambridge Survey, Cambridge: Cambridge University Press, vol. 4, 37-63.
Haas W. (ed.) 1982. Standard Languages: spoken and written. Manchester: Manchester
University Press.
Hall,J. (ed.) 1920. Early Middle English. 2 vols. Oxford: Clarendon Press.
Harris, J. 1984. Syntactic variation and dialect divergence. Journal of Linguistics, 20,
2, 303-27.
Harris,J.1985. Phonological Variation and Change. Cambridge: Cambridge University
Press.
Harris,J.,Little, D. and Singleton, D. (eds) 1986. Perspectives on the English Language
in Ireland. Dublin: Trinity College.
Haugen, E. 1972. Dialect, language, nation. In Pride, J. B. and Holmes, J. (eds),
Sociolinguistics, Harmondsworth: Penguin Books, 97-111.
Heller, M. (ed.) 1988. Code-switching. Berlin: Mouton/de Gruyter.
Henderson, E. 1971. The Indispensable Foundation. Oxford: Oxford University Press.
Hoaglin, D., Mosteller, M. and Tukey, J. (eds). 1985. Exploring Data Tables, Trends
and Shapes. New York: Wiley.
Horvath, B. 1985. Variation in Australian English. Cambridge: Cambridge University
Press.
Horvath, B. and Sankoff, D. 1987. Delimiting the Sydney speech community.
Language in Society, 16, 179-204.
Houston, A. 1987. Continuity and Change in English Morphology: the variable (ING).
PhD dissertation: University of Pennsylvania.
Hojrup, T. 1983. The concept of life-mode: a form-specifying mode of analysis
applied to contemporary western Europe. Ethnologia Scandinavica, 1-50
Jespersen, O. 1922. Language: its nature, development and origin. London: Allen and
Unwin.
Johnston, P. 1983. Irregular style variation patterns in Edinburgh speech. Scottish
Language, 2, 1-19.
 References 233

Jones, D. 1955. An English Pronouncing Dictionary. London: Dent.

Jones, D. 1956. The Pronunciation of English, 4th edn. Cambridge: Cambridge
University Press.
Joseph, J. E. and Taylor, T. J. 1990. Ideologies of Language. London: Routledge.
Kibrik, A. E. 1977. The Methodology of Field Investigations in Linguistics (Janua
Linguarum, Series Minor). The Hague: Mouton.
Kiparsky, P. 1988. Phonological change. In Newmeyer, F. (ed.), Linguistics: the
Cambridge Survey, Cambridge: Cambridge University Press, vol. 1, 363-416.
Knowles, G. 1978. The nature of phonological variables in Scouse. In Trudgill, P.
(ed.), Sociolinguistic Patterns in British English, London: Edward Arnold, 80-90.
Kokeritz, H. 1953. Shakespeare’s Pronunciation. New Haven: Yale University Press.
Kristensson, G. 1967. Survey of Middle English dialects: the six northern counties and
Lincolnshire. Lund: Lund Studies in English.
Kroch, A. 1978. Toward a theory of social dialect variation. Language in Society, 7,
17-36.
Labovy, W. 1966. The Social Stratification ofEnglish in New York City. Washington DC:
Center for Applied Linguistics.
Labov, W. 1972a. Language in the Inner City. Philadelphia: Pennsylvania University
Press.
Labov, W. 1972b. Sociolinguistic Patterns. Philadelphia: Pennsylvania University
Press.
Labov, W. 1973. Where do grammars stop? In Shuy, R. W. (ed.), Sociolinguistics:
current trends and prospects, Washington DC: Georgetown University Press, 43-88.
Labov, W. 1975. On the use of the present to explain the past. In Baldi, P. and
Werth, R. N. (eds), Readings in Historicall Phonology, University Park: Pennsylvania
State University Press, 275-312.
Labov, W. (ed.) 1980. Locating Language in Time and Space. New York: Academic
Press.
Labov, W. 1981. Resolving the Neogrammarian controversy. Language, 57, 267-308.
Labov, W. 1982. Building on empirical foundations. In Lehmann, W. P. and Malkiel,
Y. (eds), Perspectives in Historical Linguistics, Amsterdam: Benjamins, 79-92.
Labov, W. 1986. Language structure and social structure. In Lindenberg, S. et al.
(eds), Approaches to Social Theory, New York: Russell Sage.
Labov, W. 1989. The child as linguistic historian. Language Variation and Change, 1,
85-98.
Labov, W. and Harris, W. 1986. De facto segregation of black and white vernaculars.
In Sankoff, D. (ed.), Diversity and Diachrony, Amsterdam: Benjamins, 1-24.
‘Labov, W., Yaeger, M. and Steiner, R. 1972. A quantitative study of sound change in
progress. Report on National Science Foundation Project no. GS—3287. Philadel-
phia: US Regional Survey.
Laing, M. 1989. Dialectal analysis and linguistically composite texts in Middle
English. In McIntosh, A., Samuels, M. and Laing, M. (eds), Middle English
Dialectology: essays on some principles and problems, Aberdeen: Aberdeen University
Press, 150-169.
Lanegran, D. A. and Palm, R. (eds) 1978.An Invitation to Geography, 2nd edn. New
York: McGraw Hill.
Lass, R. 1976. English Phonology and Phonological Theory. Cambridge: Cambridge
234 References

University Press.
Lass, R. 1980. On Explaining Language Change. Cambridge: Cambridge University
Press.
Lass, R. 1987. The Frame of English. London: Dent.
Lavandera, B. 1978. The variable component in bilingual performance. In Alatis, J.
(ed.), International Dimensions ofBilingual Education, Washington DC: Georgetown
University Press, 391- 411.
Lehmann, W. P. 1962. Historical Linguistics. New York: Holt, Rinehart and Winston.
Lehmann, W. P. and Malkiel, Y. (eds) 1968. Directions for Historical Linguistics.
Austin: University of Texas Press.
Lehmann, W. P. and Malkiel, Y. (eds) 1982. Perspectives on Historical Linguistics.
Amsterdam: Benjamins.
Leith, Dick 1980. The Social History ofEnglish. London: Routledge and Kegan Paul.
Le Page, R. B. 1975. Projection, Focusing, Diffusion. MS: University of York.
Le Page, R. B. and Tabouret-Keller, A. 1985. Acts of Identity. Cambridge:
Cambridge University Press.
Lightfoot, D. 1979. Principles ofDiachronic Syntax. Cambridge: Cambridge University
Press.
Lockwood, D. 1989. The Black-coated Worker: a Study in Class-consciousness, 2nd edn.
Oxford: Clarendon Press.
Lomnitz, L. A. 1977. Networks and Marginality. New York: Academic Press.
Lutz, A. 1985. On the historical phonotactics of English. In Kastovsky, D. and
Bauer, G. (eds). Luick Revisited, Tibingen: Gunter Narr, 221-39.
Macaulay, R. K. S. 1977. Language, Social Class and Education: a Glasgow Study.
Edinburgh: Edinburgh University Press.
McIntosh A. and Wakelin, M. F. 1989. John Mirk’s Festial and Bodleian MS Hatton
96. In McIntosh, A., Samuels, M., and Laing, M. (eds), Middle English Dialecto-
logy: essays on some principles and problems, Aberdeen: Aberdeen University Press,
170-8.
McIntosh, A., Samuels, M. and Benskin, M. 1986. 4 Linguistic Atlas of Late Medieval
English, 4 vols. Aberdeen: Aberdeen University Press.
McIntosh, A., Samuels, M. and Laing, M. (eds) 1989. Middle English Dialectology:
essays on some principles and problems. Aberdeen: Aberdeen University Press.
McKnight, G. 1928. The Evolution of the English Language. New York: Dover
Publications.
Marsden, P. V. and Lin, N. (eds). 1982. Social Structure and Network Analysis.
London: Sage.
Martinet, A. 1955. Economie des changements phonétiques. Bern: Francke.
Matthews, W. 1938. Cockney Past and Present. London: Routledge and Kegan Paul.
Mewett, P. 1982. Associational categories and the social location of relationships in a
Lewis crofting community. In Cohen, A. (ed.), Belonging, Manchester: Manchester
University Press, 101-130.
MillerJ.forthcoming. The grammar of Scottish English. In Milroy,J.and Milroy, L.
(eds), The Grammar of English Dialects of the British Isles, London: Longman.
Mills, C. Wright 1956. The Power Elite. Oxford: Oxford University Press.
Milroy, J. 1976a. Length and height variation in the vowels of Belfast vernacular.
Belfast Working Papers in Language and Linguistics, 1.3, 69-110.
 References 230

Milroy, J. 1976b. Synopsis of Belfast vowels. Belfast Working Papers in Language and
Linguistics, 1.4, 111-15.
Milroy, J. 1977. The Language of Gerard Manley Hopkins. London: André Deutsch.
Milroy, J. 1978. Stability and change in non-standard English in Belfast. Bulletin of
the Northern Ireland Speech and Language Forum, 2, 72-82.
Milroy, J. 1980. Lexical alternation and the history of English. In Traugott, E.,
Labrum, R. and Shepherd, S. (eds), Papers from the Fourth International Conference
on Historical Linguistics, Amsterdam: Benjamins, 355-62.
Milroy, J. 1981 Regional Accents of English: Belfast. Belfast: Blackstaff.
Milroy, J. 1982a. On the problem of historical interpretation: Verner’s Law in
Gothic. In Ahlqvist, A. (ed.), Papers from the Sth International Conference on
Historical Linguistics, Amsterdam: Benjamins, 223-9.
Milroy, J. 1982b. Probing under the tip of the iceberg: phonological normalisation
and the shape of speech communities. In Romaine, S. (ed.), Sociolinguistic
Variation in Speech Communities, London: Edward Arnold, 35-47.
Milroy, J. 1982c. Some connections between Galloway and Ulster speech. Scottish
Language, 1, 9-20.
Milroy, J. 1983. On the sociolinguistic history of /h/-dropping in English. In
Davenport, M., Hansen, E. and Nielsen, H.-F. (eds), Current Topics in English
Historical Linguistics, Odense: Odense University Press, 37-53.
Milroy, J. 1984. Present-day evidence for historical change. In Blake, N. F. and
Jones, C. (eds), Progress in English Historical Linguistics, Sheffield: University of
Sheffield, 173-91.
Milroy,J.1986. The methodology of urban language studies: the example of Belfast.
In Harris, J., Little, D. and Singleton, D. (eds), Perspectives on the English Language
in Ireland, Dublin: Trinity College, 31-48.
Milroy,J.1989. The concept of prestige in sociolinguistic argumentation. York Papers
in Linguistics, 13, 215-26.
Milroy, J. forthcoming. Middle English dialectology. In Blake, N. F. (ed.) The
Cambridge History of the English Language, vol. 2, Cambridge: Cambridge Univers-
ity Press.
Milroy, J. and Harris, J. 1980. When is a merger not a merger?: the MEAT/MATE
problem in a present-day English vernacular. English World Wide, 1, 2, 199-210.
Milroy, J. and Milroy, L. 1978. Belfast: change and variation in an urban vernacular.
In Trudgill, P. (ed.) Sociolinguistic Patterns in British English. London: Edward
Arnold.
Milroy, J. and Milroy, L. 1985a. Authority in Language. London: Routledge and
Kegan Paul.
Milroy, J. and Milroy, L. 1985b. Linguistic change, social network and speaker
innovation. Journal of Linguistics, 21, 339-84.
Milroy, J. and Milroy L. (eds) forthcoming. The Grammar of English Dialects of the
British Isles. London: Longman.
Milroy, J., Milroy, L., Gunn, B., Harris, J., Pitts, A. and Policansky, L. 1983.
Sociolinguistic variation and linguistic change in Belfast. Report to the Social
Science Research Council. (Grant no. HR 5777).
Milroy, L. 1984. Comprehension and context: successful communication and
communicative breakdown. In Trudgill, P. (ed.), Applied Socialinguistics. London:
236 References

Academic Press, 7-31.

Milroy, L. 1987. Language and Social Networks, 2nd edn. Oxford: Basil Blackwell.
Milroy, L. 1989. Gender as a speaker variable ... the glottalised stops in Tyneside.
York Papers in Linguistics, 13, 227-36.
Milroy, L. and Li Wei forthcoming. A social network perspective on code-switching
and language choice: the example of the Tyneside Chinese community. MS.
Milroy, L. and Milroy, J. 1977. Speech and context in an urban setting. Belfast
Working Papers in Language and Linguistics, 2.1, 1-85.
Milroy, L. and Milroy, J. forthcoming. Social network and social class: towards an
integrated sociolinguistic model.
Mitchell, J. C. 1986. Network procedures. In The Quality of Urban Life, Berlin: de
Gruyter.
Morris, R. (ed.) 1873. Genesis and Exodus. Oxford: Early English Texts Society.
Mugglestone, L. 1990. Ladylike accents: female pronunciation and perceptions of
prestige in nineteenth-century England. Notes and Queries, 37, 1, 44-52.
Miller, F. Max 1861. Lectures on the Science of Language, \st series. London:
Longmans, Green.
Miller, F. Max 1881. Selected Essays on Language, Mythology and Religion. London.
Newbrook, M. 1986. Sociolinguistic Reflexes of Dialect Interference in West Wirral.
London: Peter Lang.
Newmeyer, F. (ed.) 1988. Linguistics: the Cambridge Survey, 4 vols. Cambridge:
Cambridge University Press.
Ochs, E. 1979. Planned and unplanned discourse. In Givon, T. (ed.), Discourse and
Syntax, New York: Academic Press, 51-80.
O’Connor, J. D. 1973. Phonetics. Harmondsworth: Penguin.
Oliphant, T. K. 1873. The Sources of Standard English. London: Macmillan.
Onions, C. T. 1966. Oxford Dictionary ofEnglish Etymology. Oxford: Clarendon Press.
Orton, H., Barry, M. V., Halliday, W. J., Tilling, P. M. and Wakelin, M. F. 1963-9.
Survey of English Dialects, 4 vols. Leeds: E. J. Arnold.
Owens, E. 1977. The Distribution of /I/ in Belfast Vernacular English. MA dissertation,
Queen’s University, Belfast.
Palmer, F. 1965. A Linguistic Study of the English Verb. London: Longman.
Parsons, T. 1952. The Social System. London: Tavistock.
Patterson, D. 1860. Provincialisms of Belfast. Belfast: Mayne Boyd.
Pellowe, J., Nixon, G., Strang, B. and McNeany, V. 1972. A dynamic modelling of
linguistic variation: the urban (Tyneside) linguistic survey. Lingua, 30, 1-30.
Phillipps, K. C. 1984. Language and Class in Victorian England. Oxford: Basil
Blackwell.
Pitts, A. 1982. The elusive vernacular: an account of fieldwork techniques in urban
sociolinguistic studies in Northern Ireland. Belfast Working Papers in Language and
Linguistics, 6, 104-21.
Plutchik, R. 1974. Foundations of Experimental Research, 2nd edn. New York: Harper
and Row.
Postal, P. 1968. Aspects of Phonological Theory. New York: Harper.
Pride, J. B. and Holmes, J. (eds) 1972. Sociolinguistics. Harmondsworth: Penguin
Books.
 References 237

Prokosch, E. 1939. A Comparative Germanic Grammar. Philadelphia: University of

Pennsylvania Press.
Rickford,J.1986. The need for new approaches to social class analysis in linguistics.
Language and Communication, 6, 3, 215-21.
Rigg, L. 1987. A Quantitative Study of Sociolinguistic Patterns of Variation in Adult
Tyneside Speakers. Undergraduate dissertation: University of Newcastle upon
Tyne.
Rogers, E. M. and Shoemaker, F. F. 1971. Communication of innovations, 2nd edn.
New York: Free Press.
Romaine, S. 1981. The status of variable rules in sociolinguistic theory. Journal of
Linguistics, 17, 93-119.
Romaine, S. 1982a. Socio-historical Linguistics. Cambridge: Cambridge University
Press.
Romaine, S. (ed.) 1982b. Sociolinguistic Variation in Speech Communities. London:
Edward Arnold.
Romaine, S. 1984a. The sociolinguistic history of t/d deletion. Folia Historica
Linguistica, 5, 2, 221-55.
Romaine, S. 1984b. The status of sociological models and categories in explaining
linguistic variation. Linguistische Berichte, 90, 25-38.
Samuels, M. L. 1972. Linguistic Evolution. Cambridge: Cambridge University Press.
Sankoff, D. (ed.). 1978. Linguistic Variation: models and methods. New York: Academic
Press.
Sankoff, D. (ed.) 1986. Diversity and Diachrony. Amsterdam: Benjamins.
Sankoff, D. and Laberge, S. 1978. The linguistic market and the statistical
explanation of variability. In Sankoff, D. (ed.), Linguistic Variation: models and
methods, New York: Academic Press, 239-50.
Sankoff, D., Cedergren, H., Kemp, W., Thibault, P. and Vincent, D. 1989.
Montreal French: language, class and ideology. In Fasold, W. and Schiffrin, D.
(eds), Language Change and Variation, Amsterdam: Benjamins.
Sapir, E. 1921. Language. New York: Harcourt, Brace.
Schegloff, E. 1979. The relevance of repair to syntax for conversation. In Givén, T.
(ed.), Discourse and Syntax, New York: Academic Press, 261-86.
Scragg, D. 1974. The History of English Spelling. Manchester: Manchester University
Press.
Sebeok, T. (ed.) 1960. Style in Language. Boston: MIT Press.
-Shuy, R. W. (ed.) 1973. Sociolinguistics: current trends and prospects. Washington DC:
Georgetown University Press.
Shuy, R. W., Wolfram, W. A. and Riley, W. K, 1968. Field Techniques in an Urban
Language Study. Washington DC: Center for Applied Linguistics.
Sisam, K. (ed.) 1915. The Lay of Havelok the Dane, edited by W. W. Skeat, 2nd edn
revised by K. Sisam.
Skeat, W. W. 1897. The proverbs of Alfred. Transactions of the Philological Society.
Smith, N. V. 1989. The Twitter Machine. Oxford: Basil Blackwell.
Smith N. V. and Wilson, D. 1981. Modern Linguistics. Harmondsworth: Penguin
Books.
Staples, J. H. 1898. Notes on Ulster English dialect. Transactions of the Philological
238 References

Society, 357-87.
Sturtevant, E. H. 1917. Linguistic Change. Chicago: Chicago University Press.
Tarde, G. 1903. The Laws of Imitation, translated by E. Clews Parsons. New York:
Holt.
Taylor, M. Vaiana 1973. The great southern Scots conspiracy: patterns in the
development of northern English. In Anderson, J. and Jones, C. (eds), Historical
Linguistics, Amsterdam: North-Holland Publishing Company, 403-26.
Traugott, E., Labrum, R. and Shepherd, S. (eds) 1980. Papers from the Fourth
International Conference on Historical Linguistics. Amsterdam: Benjamins.
Trench, R. C. 1888. On the Study of Words, 20th edn. London: Macmillan.
Trudgill, P. 1974. The Social Differentiation of English in Norwich. Cambridge:
Cambridge University Press.
Trudgill, P. (ed.) 1978. Sociolinguistic Patterns in British English. London: Edward
Arnold.
Trudgill, P. 1981. On the limits of passive ‘competence’: sociolinguistics and the
polylectal grammar controversy. In Crystal, D. (ed.), Linguistic Controversies:
festschrifi for F. R. Palmer, London: Edward Arnold.
Trudgill, P. 1983a. Sociolinguistics, 2nd edn. Harmondsworth: Penguin Books.
Trudgill, P. 1983b. On Dialect. Oxford: Basil Blackwell.
Trudgill, P. (ed.) 1984a. Language in the British Isles. Cambridge: Cambridge
University Press.
Trudgill, P. (ed.) 1984b. Applied Sociolinguistics. London: Academic Press.
Trudgill, P. 1986a. Dialects in Contact. Oxford: Basil Blackwell.
Trudgill, P. 1986b. The apparent time paradigm: Norwich revisited. Paper presented
at the 6th Sociolinguistics Symposium, University of Newcastle upon Tyne.
Tukey, J. 1977. Exploratory Data Analysis. Reading, Mass.: Addison Wesley.
Vennemann, T. 1989. Language change as language improvement. In Modelli
esplicativi della diachronia linguistica, Pisa: Giardini editori.
Wakelin, M. F. (ed.) 1972. Patterns in the Folk Speech of the British Isles. London:
Athlone Press.
Wakelin, M. F. and Barry, M. V. 1968. The voicing of initial fricative consonants in
present-day dialectal English. Leeds Studies in English, new series, 2, 47-64.
Walker, J. 1791. A Critical Pronouncing Dictionary.
Wang, W. 1969. Competing changes as a cause of residue. Language, 45, 9-25.
Wardhaugh, R. 1986. An Introduction to Sociolinguistics. Oxford: Basil Blackwell.
Weinreich, U., Labov, W. and Herzog, M. 1968. Empirical foundations for a theory
of language change. In Lehmann, W. P. and Malkiel, Y. (eds), Directions for
Historical Linguistics, Austin: University of Texas Press, 95-189.
Wells, J. C. 1982. Accents of English: an introduction. Cambridge: Cambridge
University Press.
West, S. 1988. Aspects of the Phonology of Ballygawley. MA dissertation: Queen’s
University, Belfast.
Williams, R. A. 1903. Remarks on Northern Irish pronunciation of English. Modern
Language Quarterly, 6, 129-35.
Wirth, L. 1938. Urbanism as a way of life. American Journal of Sociology, 44, 1, 1-24
Woolard, K. 1985. Language variation and cultural hegemony: towards an integra-
tion of linguistic and sociolinguistic theory. American Ethnologist, 738-48.
 References 209

Wright, J. 1954. Grammar of the Gothic Language, 2nd edn, revised by O. L. Sayce.
Oxford: Clarendon Press.
Wyld, H. C. 1927. A Short History of English, 3rd edn. London: John Murray.
Wyld, H. C. 1936. A History of Modern Colloquial English. Oxford: Basil Blackwell.
Young, M. and Wilmott, P. 1962. Family and Kinship in East London. Harmonds-
worth: Penguin Books.
 Index

abductive change, 202-3 conflict, 207-10

accountability to the data, 6, 66 and powerful groups, 215-16
actuation, 20-2, 164-5, 169-72 see also consensus
Andersen, Henning, 181, 182, 196, 202-3, consensus
206-7, 211 norms, 7-8, 82-4
Anglo-Norman, 163 as opposed to conflict, 191, 207-10
scribes, 132-3 see also norms
spellings, 132-3, 142 constraints on linguistic change, 35
Ash, Sharon, 191-4 the constraints problem, 14-15
context (of situation), 5-6
Barry, Michael V., 136 contextual style
Belfast, sociolinguistic research in, 10-11, style-shifting, 43-5
49-50, 55-122, 171-2 stylistic variation, 103, 186-7
door-step survey in, 101-4 convergence, 50-1
inner-city variation in, 87-91, 95-100, conversation, 5
185-91 conversational interaction, 39-42
outer-city variation in, 101, 104-10 planned and unplanned discourse, 39
Belfast vowel-system, the, 57-60, 71-4, correctness, 4
105-9
raising before velar consonants in, 148-9, deductive change, 202-3
151-2 Diary of Henny Machyn, The, 127
Benskin, M., 132 diffuseness in the speech community, 61-2,
Bertz, S., 67 88; see also divergence; speech
Bestiary, The, 132 community
Bourdieu, P., 210 Dittmar, Norbert, 210, 214-15
Brown, Gillian, 39-42 divergence, 50-1
describing divergent states, 55-64
Cameron, Deborah, 77-8 dialect-divergence, 55, 61-4, 68
causation of change, 77-8; see also actuation in Philadelphia, 191
child language as locus of change, 204 Dobson, E. J., 126-9, 154-6
Clark, Cecily, 133 Downes, William, 179
Cockney, 127 drift, 29-30
Cohen, Anthony, 62-3, 96
communication, breakdown of, 31; see also early adopter(s) of innovations, 183-5
function; miscomprehension Early Modern English, 125-9
comprehension, 35; see also constraints; vowel-system, 145-60
miscomprehension embedding of linguistic variation, 15
 Index 241

empirical foundations, 13-16 idealization, 3

ethnic differences idealized speaker-innovator, 172
in Belfast, 187, 190-1 ideology of the standard language, 12-13,
in Philadelphia, 191-5 124-5
ethnography, 62-3, 176-7 indirectness in conversation, 41
etymology, 162-3 innovation
evaluation, 113-14 innovators, 120
change of, 143-5 see also speaker-innovation
the evaluation problem, 15
exploratory science, 62
Jespersen, Otto, 26-7
Johnston, Paul, 55, 62
face in conversational interaction, 41
Jones, Daniel, 9, 216
PischeruG@. 212
flip-flop rules, 151-2
focusing, 61-2 Kiparsky, Paul, 24-5, 161
Francis, W. Nelson, 31 Kristensson, Gillis, 16-17
function(s) Kroch, Anthony, 181
discourse functions, 39-42
information-bearing, 36-9 Labov, William, 13-16, 20-1, 53, 86, 90-1,
listener-oriented, 39-42 159-60, 166-7, 172-5, 181, 183,
message-oriented, 39-40 191-4, 209
of social networks, 85 Laing, Margaret, 132
of variation and change, 29-45, 92-5, language contact, 195-200
98-100, 107-8 contact change, 206-7
functional load, 37 contact groups, 191-5
as speaker-contact, 199-200
Gal, Susan, 199-200, 211 see also linguistic borrowing
gender-differentiation see sex-differentiation language maintenance, 10-12, 13-14
Genesis and Exodus, 141 language missionaries, 198
Giddens, Anthony, 212, 218 Lass, Roger, 22, 24, 26-7, 170-1
Goldthorpe, J. H., 218 legitimization of a language variety, 125
Granovetter, Mark, 178-82 legitimized varieties (v vernaculars),
Gregersen, Frans, 220 210-11
Le Page, Robert B., 61
Hall, Joseph, 132, 133 lexical diffusion, 24-5, 161-2
Harris, John, 159-60, 165, 221 lexical phonology, 25
Harris, Wendell, 166-7, 191-4, 209 life-modes
Havelok the Dane, 134-6 analysis of, 214-20
(h]-dropping in English and conflict-based social theory, 214-16
change of evaluation of, 138-9 life-mode 1, 217
in Early Modern English, 142-3 life-mode 2, 217-18
geographical distribution of, 137-8 life-mode 3, 218-19
in Middle English texts, 140-2 as a structural concept, 219-20
time-depth of, 137-45 Linguistic Atlas of Late Medieval English, 132,
Herzog, Marvin, 13-16, 20-1, 53 136-7, 140
‘Hiberno-English, 33-6, see also Belfast; linguistic borrowing, 201, 204
Ulster dialect linguistic change, 1-19
historical description as change in consensus norms, 17-18, 91
database of, 45-7 conversation as a locus of, 36
of English, 23, 28, 49-53 as improvement, 30-1
internal v external, 23—4 locating ...in progress, 109-22
historical linguistics natural..., 21
explanation in, 11-12 regularity of, 204
as an uncomfortable science, 46 as a social phenomenon, 200-5, 221-2
Hgjrup, Thomas, 214-20 linguistic market, the, 210-11
homonymic clash, 37 linguistic variables, 48
Horvath, Barbara, 61 selection of, 68-71
hypercorrection, 56-7, 163 Li Wei, 212
242 Index

MeIntosh, Angus, 132 Samuels, M. L., 37, 132

Martinet, André, 36-8 Sankoff, David, 61, 79, 210
merger, 37, 74-5, 156-60 Saussure, Ferdinand de, 2-3
methodology, 64-80 Schegloff, Emmanuel, 41
fieldwork method, 65-7 Schlobinski, P., 210, 214-15
quantitative method, 75-80 Scots dialect, 150, 155-6
Mewett, Peter, 177-8, 213 Ulster Scots, 113-14, 116, 117
Middle English Scragg, D. G., 133-4
spelling, 132-5 sex-differentiation in language, 86-7,
structured variability in, 134-6 103-5, 116-17, 118-22, 166, 188-9
as a variable state, 130-4 Shoemaker, F. F., 184
Milroy, Lesley, 85, 187, 207, 212 simplification patterns, 96-7, 101-4
miscomprehension Sisam, Kenneth, 132, 133
cross-dialectal, 33-6 Skeat, Walter W., 133
repair of, 41-2 Smith, N. V., 7-8, 23, 76-8
MitchellsjaGs 3s social class, 84-5, 90-1, 172-5
Miiller, F. Max, 22-3 consensus model of, 208-9
Miyhill, John, 191-4 relationship to social network, 177-8, 181
social network, 84-6, 89-90, 107-8,
Neogrammarian principles, 24-5, 161-2, 118-22, 175-83, 187-9
201-5, 222 and life-modes analysis, 215-20
norms(s) of language, 6-7, 57-60, 66-7, and social class, 207-14, 215-20
216 in urban communities, 2] 1-14
community norms, 81-4, 87 see also weak ties
non-standard norms, 8 socio-economic class see social class
normative consensus, 188-90 socio-historical linguistics, 221]
normative judgements, 7 solidarity, 218
supra-local norms, 108-9 communities, 208
and status, 213-14
observation Southern British English, 124, 126, 149
observer effect, 54 speaker-innovation, 168-72, 177-83, 194-5
observing language in use, 5-6 and Neogrammarian change, 201-5
occasional spelling(s), 127-9, 154-6 speaker innovator, 172—5, 182-3, 184-5,
Ochs, Elinor, 39 191
Onions, C. T., 162-3 speaker-oriented (~ system-oriented)
Ownes, Elizabeth, 93-4 approach, 6, 22, 24, 165-8
speaker-variables, +9, 84-90
palatalization, 21 functions of, 90
Palmer, F. R., 7-8 speakers, role of, +, 26-8, 43-5
Paston Letters, The, 127 speech community, 9
Patterson, David, 99, 112, 113, 115, 116, consensus-based model of, 209-10
148-50 endocentric and exocentric, 181, 196-7,
Pedersen, Inge-Lise, 220 206-7
Perv, Thomas, 127 shape of, 53-5, 59-60, 82-4, 95-109
Philadelphia, community studies in, 166-7, stability in, 86-8
191-5 status/solidarity model of, 213-14
pre-aspiration in Icelandic and Gaelic, Standard English, projection on to the past,
198-9 51-2, 123-9
prescription, 8, 9, 31-2, 216 standardization, consequences of, 124-34
prestige, 100, 129-30, 172-5, 191-2, 194 Staples,J.H., 112, 116
projection, backward, 49-53, 74-5, 145-60 statistics
confirmatory, 78-9
Received Pronunciation, 8, 9, 58-60, exploratory, 78
124-9, 149, 216 Steiner, R., 159-60
Rickford, John, 208, 209 Steinsholt, Anders, 198
Rogers, E. M., 184 stigma, 44, 175
Romaine, Suzanne, 62, 77-8, 221 structuralism, 2—4
 Index 243

Sturtevant, E. H., 26-7 structuredness in, 4, 11

style see contextual style time-depth of, 123-62
system-oriented approach variable rules, 25-6
see speaker-oriented approach vernacular, the, 66-7

Tabouret-Keller, Andrée, 61 Wachs, I., 210, 214-15

teleology in linguistic change, 29 Wakelin, M. F., 132, 136
Tetak dialects (of Czech), 203 weak (network) ties, 104, 107-8, 176-83,
transition problem, 15-17 187-90, 212-13
Trench, Richard Chenevix, 22 and language contact, 195-200
Trudgill, Peter, 7-8, 171, 197-8 in life-mode 2, 217-18
in life-mode 3, 219
Ulster dialect(s), 113-18; see also Scots Weinreich, Uriel, 13-16, 20-1, 53
dialect Williams, R. A., 112
uniformity in language, 1-4, 32-3, 123-4, written language, 5
130 as database for historical linguistics, 45
urban dialect, 100-1, 197-9 as a message-oriented function, 40
in relation to urban community structure, see also function
211-14 Woolard Kathryn, 208, 210
see also Belfast; Philadelphia Wyld, H. C., 5, 51-2, 141-2, 154-7

variability in language, 1-2, + Yaeger, Malcah, 159-60

 « os
a

ere
me

Ny, Anema, 10 pbecsnbouse = Sanraels, Vib.) LATS hw switvoiar®

oh told 3 REE Ao diqeb-sis — Sankoff, Dinkd, 5, Tig MManuinoy “Pee

By Mey
ary1 Mew Stoners; Pevdaganidegtait Lennie -ansieve
ISye atu aldainay Siri AS are He
te4 i , wert Wa baT, : | a? soph iT
.
: ‘ 5
4 j ‘ Spey!

ids tA eas allies)

! , pitas mG al Anlhoist
ot } mah ape}
it ine t wu ry wottlernetl
‘ y
whet Dis y Ae wil
: ty = 1 Pad Yi
ame 3
ie ~~
] ry si

PPsal| Jn osha i:
Song :
“a

iu fis
, Ff

it
ie

é iv ' -

P .
H

. =<

;
-

’ .

WHERE 9 : ;
% ‘ er
AN”
v

.
ri ; ;
-
,

; a
So.
wi
2
c ae,

3 1) ‘Meat ‘ = :
\ fees ia
*
ye
Baty é :
ug ' COS

~
yy A
ae
occur
ttmcepa tone em hy
)

[chet eeteee
" ; i : ' uk
: aaa
: a F oe PRN.
v ¥ ;

a he
' } * 4 nl ; ; th
aie aces cecaaie clade Soraicllbieaipeeetioitcciamea
‘camen’ sesieeemetdadhan
eet tem eeeeh eet aad I
; : ae a 4 | tage .

een
\
= ox ; \ 4
x

Se

.
re rere
 ACPA
This book is concerned with the e
change. Focussing on variation in the English language, it
explores the extent to which language change is a social
phenomenon.
Language, James Milroy holds, cannot adequately be
observed or described independently of society. In analysing
patterns of language use, we must be aware of social and situ-
ational contexts and of the norms of usage in the speech com-
munity. He discusses these methodological issues in relation to
his own sociolinguistic research in Belfast, and argues that in
explaining language variation we need first to understand
these factors which maintain language and resist change.
In contrast to the intra-linguistic approach of traditional histor-
ical work, this book presents a social model of change derived
from the study of social networks and the links between net-
works and social class. Language change, Professor Milroy
suggests, is made possible to the extent that it is passed from
person to person in conversational encounters. It is the relative
strength or weakness of social ties between speakers that
offers the most powerful explanation available for the mystery
of language change.

James Milroy was formerly Professor of Linguistics at the Uni-

versity of Sheffield and Senior Lecturer in English at Queen’s
University, Belfast. He has left full time university teaching and
is now engaged mainly in research and writing.

Cover illustration: Claudius Ptolemy’s map of the British Isles, Mercator, 1578.
Cover design by Workhaus Graphics

E
ISBN O0-631-14347-x

BLACKWELL
Oxford UK & Cambridge USA
AN 780631