SYNTACTIC EFFECTS

OF MORPHOLOGICAL
CHANGE
Edited by
David W. Lightfoot

1
 Contents

Notes on contributors vii

Foreword xi
. Introduction 
David W. Lightfoot
P I. M  
. The History of the Future 
Ian Roberts and Anna Roussou
. Case and Middle English Genitive Noun Phrases 
Cynthia L. Allen
. Split Constituents within NP in the History of English:
Commentary on Allen 
Željko Bošković
. Inflectional Morphology and the Loss of Verb-Second in English 
Eric Haeberli
. The Rise of the To-Dative in Middle English 
Thomas McFadden
. Double Objects and Morphological Triggers for Syntactic Case 
Chiara Polo
. Cue-Based Change: Inflection and Subjects in the History of
Portuguese Infinitives 
Acrisio Pires
. Morphology and Null Subjects in Brazilian Portuguese 
Cilene Rodrigues
. Loss of Overt Wh-Movement in Old Japanese 
Akira Watanabe
. Changes in Subject Case-Marking in Icelandic 
Thórhallur Eythórsson
P II. I  
  
. A Reinterpretation of the Loss of Verb-Second in Welsh 
Dirk Bury
. The Loss of IP-Scrambling in Portuguese: Clause Structure,
Word-Order Variation and Change 
Ana Maria Martins
vi Contents

P III. I    

. Residual V-to-I 
Dianne Jonas
. Syntax and Morphology are Different: Commentary on Jonas 
Stephen R. Anderson
 Verb–Object Order in Old English:
Variation as Grammatical Competition 
Susan Pintzuk
. VO or OV? That’s the Underlying Question: Commentary on Pintzuk 
Jairo Nunes
. Movement, Morphology, and Learnability 
Susana Bejar
. Object Shift and Holmberg’s Generalization 
in the History of Norwegian
John D. Sundquist
P IV. C S
. The Computational Study of Diachronic Linguistics 
Partha Niyogi
. Grammar Competition and Language Change 
Charles D. Yang
References 
Index 
 Foreword

The papers making up this collection were presented in some early form at the
sixth meeting of the Diachronic Generative Syntax series, DIGS VI, held at the
University of Maryland in May . That meeting was supported by grants from
the National Science Foundation (BCS) and from the College of Arts
and Humanities at the University. The funds enabled us to invite certain speakers,
whose papers were circulated in advance of the meeting and drew commentary
from invited respondents. The funds also enabled the meeting to be run in a con-
genial way, maximizing opportunities for discussion in what has become a vibrant
DIGS tradition. The meeting itself constituted not only DIGS VI but also the
annual Mayfest run by the Department of Linguistics and, as usual, the Mayfest
was supported by graduate students who organized registration, accommodation,
coffee, food, and all the accoutrements of successful intellectual events. The meet-
ing owed much in particular to Kleanthes Grohmann, Andrea Gualmini, Luisa
Meroni, Mitsue Motomura, Acrisio Pires, and Cilene Rodrigues.
The meeting also flourished through the work of the referees who read abstracts,
and then that process started again when the participants at the meeting decided
to put together a volume with a specific theme. The refereeing process turned out
to yield not only decisions about what to include but also productive discussions
with authors about revisions. For their work at various stages, I would like to thank
Cindy Allen, Avery Andrews, Pilar Barbosa, David Denison, Paul Elbourne, Olga
Fischer, Charlotte Galves, Mark Hale, Eric Haeberli, Norbert Hornstein, Ans van
Kemenade, Willem Koopman, Tony Kroch, Pino Longobardi, Jairo Nunes, Wayne
O’Neil, Susan Pintzuk, Ian Roberts, Beatrice Santorini, Peter Svenonius, Hoski
Thráinsson, Juan Uriagereka, Barbara Vance, Nigel Vincent, Anthony Warner, John
Whitman, and Wim van der Wurff. Then discussions with Oxford University Press
yielded helpful, anonymous referee reports, leading to more revisions and improve-
ments. As with any scientific endeavor, this volume stands on many shoulders.
The logistics of putting such a book together were made easier by the computa-
tional genius of the peerless Pires: Acrisio worked graceful magic in many ways
and enabled us to get the final manuscript to press almost within a year of the meet-
ing. Jeff Parrott and Eric Lightfoot helped at the copy-editing stage and John Davey
at the press mediated wisely and smoothed the way to an easy and speedy publica-
tion.
What do languages have to lose but their morphemes? Losing morphemes may
have consequences elsewhere and that’s what this volume is about.
David W. Lightfoot
June 
 1
Introduction
DAVID W. LIGHTFOOT

..    

There is much involved in language change, a group phenomenon, but the perspec-
tive taken in this book focuses on changes in grammars, properties of individuals.
Any perspective focuses on certain things, leaving other things in a peripheral blur.
It idealizes, abstracts away from certain data, and deals with only part of the gen-
eral phenomenon, but grammar change is a central part of language change and, as
we shall see, linked, naturally enough, to other aspects of the larger, group phenom-
enon. Grammars are formal characterizations of an individual’s linguistic capacity,
conforming to and exploiting the tools provided by a universal initial state (‘Uni-
versal Grammar’) and developing as a person is exposed to his or her childhood lin-
guistic experience. These grammars (mental organs; see Anderson and Lightfoot,
) develop in accordance with the demands and provisions of UG and may
vary by predefined parameters. And they may change over time: people developed
certain grammars in London in the thirteenth century and different ones in the fif-
teenth century, with some parameters set differently.
Historical change in grammars constitutes a particular form of variation, which
affords special opportunities. Sometimes one can find cases where grammars vary
only minimally: there may be effectively just a single difference in two abstract
systems, even with the lexicon largely unchanged. When one is lucky enough to
isolate points of variation in this way, one can then ask why the new grammar
should have emerged in children at that time. One possibility, arguably the only
possible answer, is that those children heard different things; their experience dif-
fered in some crucial fashion. In that case, the linguist wants to find how the gram-
mars changed and how the relevant childhood experience might have changed just
prior to the change in grammars, in such a way that the new grammar was the only
possible outcome. In this perspective, the study of grammar change is fused with
work on variation and the acquisition of grammars. We explain the emergence of
the new grammar and the explanation illuminates the nature of the child’s trigger-
ing experience and the way in which children acquire their linguistic capacities;
the study of change has implications for grammatical theory and for theories of
language acquisition. This enterprise makes no use of a theory of change as such
and it demonstrates that at least some changes may be explained through language
 Syntactic Effects of Morphological Change

acquisition by children (which is not to deny other kinds of change, e.g. that adults
may have different grammars, but that would be another story, composed under a
different muse).
Grammars differ sharply: a person either has a grammar with a certain property,
or not. People’s speech, on the other hand, is in constant flux and languages, con-
glomerations of the output of people’s grammars, are inherently fluid, unstable,
always changing—rivers of experience. Much of this variation and fluidity is irrele-
vant for the acquisition of grammars. No two children take the same step into the
same river or have the same linguistic experience, but they may converge on the
same grammars, certainly in terms of the available structures and operations. Vari-
ation in language is often so slight as to be imperceptible by any individual survey-
ing her immediate environment in the course of her lifetime. The variation may
be irrelevant for emerging grammars, even imperceptible, or it may reach some
threshold level and trigger a different grammar.
The linguistic experience of young children (their ‘primary linguistic data’) is
always in flux, but sometimes small variations have bigger, structural consequenc-
es, changes in grammars. Grammar change, change in structures, is different; if
grammars are abstract, then they change only occasionally and sometimes with
dramatic consequences for a wide range of constructions and expressions; gram-
mar change tends to be ‘bumpy’, manifested by clusters of phenomena changing
at the same time. Grammatical change is contingent, dependent on the details of
the use of language (for example, changing distribution of words and construc-
tion types), language contact, perhaps even social attitudes and second language
acquisition. Grammar change is linked to changes in people’s speech, and we can
only know about it by studying what people say, often through written texts, and
it must be studied in conjunction with other kinds of change; but grammar change
constitutes a distinct type of change. The contingency of change is like geological
or cosmic change: a volcano erupts or a star explodes because of contingent events
which reach some threshold, but the nature of the eruption is subject to the general
principles of physics. So grammatical change takes place as a reaction to contin-
gent factors of language use, and the new grammars emerge subject to the usual


This bumpiness reflects the fact that grammars are abstract, motivated by the need to be demonstra-
bly acquirable by children and therefore to solve the poverty-of-stimulus problems. The abstractness
of grammars explains the bumpiness of change and different models make very different predictions
according to the abstractness of the structures postulated. For example, lexicalist models have rela-
tively surfacy representations and take change to be relatively piecemeal and gradual, but more abstract
approaches take change to be more abrupt (for some discussion, see Lightfoot : –). To antici-
pate the metaphor a few lines hence in the main text, some lexicalists might view the grammatical
volcano as not subject to periodic eruptions but rather as always just rumbling along. These are differ-
ences of degree, an empirical matter which concerns the right degree of abstraction in grammars. It is
my view that we now have a substantial body of work arguing for structural changes in grammars and
that languages change in fits and starts, in a bumpiness which can be understood in terms of changes in
individuals’ grammars, where the new grammar spreads through a population of speakers, as illustrated
now by interesting computer models (see the chapters here by Niyogi and Yang).
 Introduction 

principles of Universal Grammar. The study of the contingent events is comple-

mentary to the search for the general organizing principles of UG, but there is a
different focus. To focus on grammar change is to attend to one aspect of language
change, one which illuminates the variation and acquisition of grammars by chil-
dren, but one which is dependent on other kinds of language change.
All the writers in this volume take grammars to be properties of individuals, but
they assume different models of grammar. Some adopt lexicalist models, others
some form of government-binding model, and others are more Minimalist in orien-
tation. The volume deals with a matter of interest to theoreticians of all stripes, the
relationship between morphology and syntax. To put matters into the perspective
just discussed, we focus on a particular link between childhood experience and
the form of the grammars which emerge in individuals, dealing with the syntactic
consequences of prior changes in inflectional systems. Sometimes morphological
case systems or the inflectional properties of verbs change. They come to be mani-
fested differently in childhood experience, and it may be that, when that happens,
there are then structural changes in people’s grammars. On the other hand, this
might not be true. That is what this volume is about, the link between morphology
and syntax, and the issue has consequences for all grammatical models and for
theories of acquisition. The book offers interesting correlations between morpho-
logical and syntactic changes and those correlations, if they are principled and not
accidental, need to be expressed by good theories; they are likely to be of more
enduring interest than the particular technologies used to analyze them here.

..     

It is an old intuition, going back to the nineteenth century, that languages with rich
inflection have a different kind of syntax from languages with poor inflection, that
they show freer word order, for example. How to connect morphological proper-
ties accurately with word-order phenomena is a long-standing puzzle. Now work
within the Minimalist Program posits a tight connection between syntactic opera-
tions and morphological features and this led us to adopt the nineteenth-century
intuition as a theme for the volume. In fact, the volume ends up postulating sev-
eral interesting links but being quite skeptical about Minimalist models in this
regard. But a negative result may be illuminating, suggesting alternative hypoth-
eses, and morphology–syntax connections interest all theoreticians. Regardless of
the model in which the correlations are stated here, the enduring value of the vol-
ume will lie in the usefulness of the empirical correlations offered.
Nonetheless, since it was claims of Minimalist analyses which generated the
theme for the volume and since there seems to be a negative result emerging in that
regard, let us track the particular line of thinking which yielded these Minimalist
correlations.
Work of the s was successful in postulating principles and parameters at
the level of Universal Grammar, which helped to solve poverty-of-stimulus acqui-
 Syntactic Effects of Morphological Change

sition problems in a number of ways. Principles like the Binding Theory, X-bar
theory, the Empty Category Principle and others permitted syntacticians to claim
that children attain rich systems of mature knowledge when exposed only to rudi-
mentary experience. Experience showed that elements moved, certain words were
anaphors, other words were proper governors, and so on. Once children glean from
their environment that wh-phrases move, that each other is an anaphor, etc., the
principles of UG do the rest, providing the bridge from fragmentary experience to
the rich capacity that people attain in maturity.
To take one illustration, consider the Binding Theory, which had three princi-
ples ():
() (a) anaphors are bound (co-indexed with a c-commanding DP) in their
binding domain;
(b) pronouns are free in their binding domain;
(c) names are free everywhere.
So nouns come in three classes and children must learn which nouns are anaphors
and which are pronouns; all others are names and require no learning. Under this
view, exposure to a sentence like (a) suffices to show a child that himself is an
anaphor, if she hears it in a context where the recommender and the recomendee
are the same boy (hence co-indexed). That is what is required of experience; the
distribution of himself then follows from the fact that it is construed as an anaphor,
a fact about the individual’s grammar, and from the Binding Theory. The Binding
Theory, part of UG, guarantees that himself, classified as an anaphor, has the distri-
bution that it has, not occurring in structures like *Hei said that Mary washed him-
selfi or *Shei hurt himselfj, etc., where it is not locally bound.
() (a) The boy recommended himself.
(b) The girl waited for her mother.
Similarly, exposure to a sentence like (b) in a context where the waiting girl and
the one whose mother is expected are the same (co-indexed), shows that her is a
pronoun. Her could not be an anaphor, because it would not be co-indexed within
its binding domain (effectively the larger DP [her mother]), nor could her be a
name, because names are never co-indexed with c-commanding antecedents. Then
UG guarantees that her, classified now as a pronoun, has the distribution that it has,
not occurring in structures like *The girli expected heri to win (where her is locally
co-indexed with the girl), among many others.
There is more to be said about this illustration, of course, and one might want to
know more about the definitions of c-command, binding domains, and so on. But
it is a theory of the right type, which meets some basic requirements: given certain
assumptions, one can point to the relevant experience (), which enables the child
to determine the appropriate grammatical properties. Principles of UG then deter-
mine the distribution of the elements.
The principles and parameters postulated in the s may have been success-
 Introduction 

ful in this way but they all looked very different from each other and often con-
tained very specialized information. The Minimalist Program introduced criteria of
economy and elegance in new ways and, as one move, made the specific simplify-
ing hypothesis that there were no distinct levels of D-structure and S-structure of
the kind which featured in government-and-binding models. These were not inter-
face levels meshing with any kind of external performance system and were justi-
fied mostly on theory-internal grounds. The hypothesis that there are no distinct
levels of D- and S-structure bears directly on the relationship between syntactic
operations and morphological features.
S-structure was the level distinguishing overt and covert movement: it incor-
porated the result of all overt movement (What did Jay buy?, where what is pro-
nounced in a position other than where it is understood, as the direct object of buy)
and no covert movement (Jay bought what?, where what is pronounced where it is
understood and moves covertly to the higher position where its scope is defined).
At other levels of representation, at D-structure and at Logical Form, these two sen-
tences, What did Jay buy? and Jay bought what?, have identical structures, but they
differ at S-structure. Wh-phrases move to the Specifier of CP overtly in English
(being pronounced there: What did Jay buy?) but covertly in Chinese (Lisi mai-le
sheme, ‘Lisi bought what?’), and these distinctions are captured in the S-structures.
At D-structure and LF, there is no structural difference between the English and
Chinese sentences. Similarly verbs move to a higher inflection position in all
grammars, and this is overt in the grammars of French speakers but not in those
of English speakers: a verb may be pronounced in higher inflection positions by
French speakers, having moved away from its direct object (Lit-elle toujours les
journaux?), but not by English speakers (*Reads she always the newspapers?).
Also S-structure was the level where Case theory applied: phonetic DP/NPs and
variables needed to have Case at this level of derivation. In the (partial) D-struc-
tures of (a, b) Jay, being the complement of a participle and the subject of an in-
finitive, is not in a position to receive Case, but, after movement and at S-structure
(c, d), Jay is in a position to receive nominative Case, being the subject of a finite
verb. Since Jay and other nouns in English are not marked for morphological case,
the convention was adopted (Chomsky ) of using Case to indicate the abstract
counterpart of nominative (as indicated by the form of pronouns, nominative they
vs. accusative them, etc.).
() (a) was VP[Prtarrested Jay]
(b) seems IP[Jay to be tall]
(c) Jayi was [arrested ei]
(d) Jayi seems [ei to be tall]
Chomsky () argued that these two features of S-structure, representing overt
movement and Case, could be captured without any distinct level of S-structure rep-
resentation if one made the small technical change of replacing feature assignment
with feature checking: words are merged into syntactic structure with all of their
 Syntactic Effects of Morphological Change

morphological features, which must be ‘checked’. Movement of any element (head

or phrasal category) to any functional position takes place only when required to
check those morphological features. Under the bottom-up approach to phrase struc-
ture, Jay would be inserted into structures with its Case already attached. They
might be structures like (a, b), where the accusative Case might be checked more
or less locally (‘more or less’, because it might move to Spec of AgrO), yielding
an eventual IP[the police VP[V arrested Jay]] and IP[we expect IP[Jay to be tall]],
or (c, d), where the nominative Case could only be checked after movement in
structures like IP[Jay was VP[Prt arrested]] and IP[Jay seems IP[to be tall]].
() (a) VP[V arrested Jay]
(b) expect IP[Jay to be tall]
(c) VP[Prt arrested Jay]
(d) IP[Jay to be tall]
Furthermore, features may be ‘strong’, needing to be checked before the point at
which a derivation splits into separate PF and LF branches, requiring overt move-
ment. Alternatively, features may be ‘weak’, checked at LF and requiring only cov-
ert movement. So English verbs carry a weak feature, checked at LF, while French
verbs carry a strong feature, to be checked in overt syntax and requiring overt move-
ment: if the finiteness feature on the verb is strong, it must move to some position
where it can be checked prior to the PF/LF split (French); if it is weak, it does
not require overt movement (English). Now one doesn’t need a distinct level of
S-structure, at least not for these purposes. The general idea here is that the fea-
tures which require movement are uninterpretable at one of the interface levels (PF
or LF) and must be eliminated (‘checked’) before reaching that level.
Wh-movement may be treated in the same way: wh-phrases in English gram-
mars carry a strong feature which must be checked in the Specifier of CP prior to
the PF/LF split, unlike wh-phrases in Chinese, which carry a weak feature which
does not require checking until LF. So wh-phrases are pronounced in a displaced
position in English but not in Chinese. As a result, whether or not this constitutes
an advance (and Chomsky  seeks an alternative analysis), one can capture the
variation whereby elements move overtly not in terms of S-structure, now disposed
of, but in terms of whether the relevant features are strong or weak.
This raises the question of how the strong/weak distinction matches overt
morphological properties and some recent historical work has sought to make sub-

This kind of analysis grew out of an earlier treatment of passives, in which a DP/NP had to move
to some subject position in order to acquire Case. That treatment of passives did not account for the
movement of an understood subject PRO to a non-Case-marked position as in John tried to be arrested,
with a s-style government-binding structure along the lines of John tried [PROi to be arrested ti].
In the main text, I abstract away from the question of whether accusative DPs must move to a higher
Specifier position to be checked, e.g. to Spec of AgrO.

There was a third reason for S-structure, namely that the Binding Theory needed to apply at that
level, but Chomsky disposes of this reason by construing movement operations as instances of Copy
and Delete. See Nunes (a) for extensive discussion.
 Introduction 

stantive claims on this point, sometimes some strong hypotheses. The strongest
hypotheses have postulated a one-to-one relation between morphological proper-
ties and syntactic movement as in the ‘Rich Agreement Hypothesis’ (rich agree-
ment causes V-to-I movement), which Bobaljik () links to ‘its attendant
diachronic corollary’ that if a language loses rich agreement, it will lose V-to-I
movement. Bobaljik offers good critical discussion; he notes the disarming simplic-
ity of the hypothesis and the fact that it emerged during a period of optimism, again
misplaced in retrospect, concerning another proposed link between rich agreement
and syntactic variation, the Null Subject Parameter (null subjects occurred alleg-
edly in grammars with rich subject–verb agreement morphology).
The Rich Agreement Hypothesis was driven by observations like the following.
In Scandinavian languages with rich subject agreement on verbs, Icelandic and the
Ålvdalsmålet dialect of Swedish, finite verbs occur to the left of negative markers
(), moving to an inflection position (a„) (I illustrate this in embedded clauses,
where one can abstract away from the verb-second effects of matrix clauses).
() (a) . . . að hann keypti ekki bókina (Icelandic)
. . . that he bought not the book
(b) . . . að Jón hafa ekki komið
. . . that John has not come
(c) . . . ba fo dye at uir uildum int fy om (Ålvdalsmålet)
. . . just because that we would not follow him
(a„) . . . að hann Ikeyptii ekki VP[ei bókina]
However, in those Scandinavian languages with no subject agreement (Danish,
Norwegian, Swedish), the verb is pronounced after the negative marker, to its right.
If the inflection position in these grammars is also to the left of the negative, as
in Icelandic and Ålvdalsmålet, then finite verbs, not morphologically marked for
agreement with the subject DP/NP, do not move there overtly (); (a„) has some
analysis for (a).
() (a) . . . at hen ikke købte bogen (Danish)
. . . that he not bought the book
(b) *. . . at hen købte ikke bogen
(c) . . . om Johan inte köpte boken (Swedish)
. . . if John not bought the book
(d) *. . . om Johan köpte inte boken
. . . if John bought not the book
(a„) . . . at hen I[ ] ikke VP[købte bogen]
This kind of correlation between the position of elements and their inflectional
properties is extended in Baker’s (, ) Mirror Principle, which not only
keys movement to inflection but keys the order of affixes to the order in which syn-
tactic operations apply. For example, in Chamorro the passive morpheme is closer
to the verbal stem than the subject agreement marker ().
 Syntactic Effects of Morphological Change

() Para-u-fan-s-in-aolak i famagu’un gi as tat-n-niha.

----spank the children  father-their
‘The children are going to be spanked by their father.’

The order of the affixes mirrors the order in which the syntactic operations of pas-
sive and subject agreement take place; the passivizing operation takes the underly-
ing object to the subject position before the agreement operation applies.
Observations like these have led syntacticians to try to correlate morphological
‘strength’ with overt morphology under what Bobaljik () dubs the Rich Agree-
ment Hypothesis. The presence of overt verb movement has been keyed to the
relative morphological strength of the landing site; so the loss of verb-second
phenomena has been related to the loss of morphological strength (Roberts ,
, Rohrbacher , ). Relations have been postulated between abstract,
syntactic Case and morphological case distinctions (Lightfoot , Longobardi
). This is the line of work which led us to the theme for this volume, but the
correlations are controversial and things are not as simple as they suggest. For ex-
ample, there is good reason to believe that Swedish verbs raise to a higher I pos-
ition despite the fact that they are by no means richly inflected (Lightfoot :
, ), similarly for the Kronoby dialect of Swedish under the analysis of Rob-
erts (), and for Chinese, Japanese, and Korean (Otani and Whitman ).

All the chapters in this volume were presented at the sixth meeting of the Diachron-
ic Generative Syntax (DIGS) series, held at the University of Maryland in May
; some were invited in advance and three of those have commentaries. There
were also good papers on other aspects of grammatical change and these papers
are not included here because we chose to collect papers with a cohesive theme.
These biennial meetings focus on research into syntactic change within the theoreti-
cal framework of generative grammar. They have consistently attracted linguists
of the highest international reputation, those who are involved in ground-breaking
research in syntactic theory and language variation and change. They provide fora
(historical linguists sometimes talk like that) where recent research is presented
and discussed, where future research plans are proposed and collaborative inter-
national research is initiated and developed. Much of this collaboration has been
devoted to the development of partially parsed, computer corpora of texts, which
have enabled scholars to use their computers to ask detailed questions about the
structures of, say, Middle English and to bring vast amounts of data quickly to bear
on such questions. This technological innovation, particularly the availability of


This means that the Inflection position is to the right of the negative in (a, c). That seems plausible
because VP Fronting may occur in Swedish with a dummy verb and that verb occurs to the right of the
negative.
(i) [läser boken] kanske Allan inte gör
read the book maybe Allan not does
 Introduction 

the Penn–Helsinki Parsed Corpus of Middle English, has helped to shape papers
presented at DIGS meetings. Progress has been aided by theoretical developments
which have linked work on change more closely to work on variation and acquisi-
tion (see Clark and Roberts , Lightfoot , among others). This, in turn,
has brought work on change into the view of specialists in other areas, for mutual
enrichment. One interesting development lies in the computer simulations of
change developed by Robert Berwick, Ted Briscoe, Partha Niyogi, Charles Yang,
and others, involving population models and the spread of grammars through
speech communities. This work helps us understand the link between individual
and group behavior (grammar change vs. language change) and the volume con-
cludes with two chapters devoted to this embryonic but important work.
Questions about the consequences of morphological changes are the central
focus of this volume. Morphology has consequences and several chapters make
that clear: syntax changes as morphological case endings are lost. The volume is
also concerned with the specific question of the nature of the trigger experience
for movement (copying) operations: do they follow from morphological proper-
ties that the child is exposed to, as argued by Roberts and Rohrbacher, or do they
require more syntactic evidence? That is, can children deduce syntactic operations
from a good morphological analysis or do they need access directly to syntactic
structures in order to posit movement operations?  To put it in the terms of Light-
foot (), what are the cues for movement operations? That question, which
may have a complex answer, bears on a major issue within syntactic theory and
extends the dialogue between historical linguists, theoreticians, and acquisition-
ists, such that their work is mutually enlightening. It is clear that there are connec-
tions between morphology and syntax; Part I of this book discusses some direct
connections and Part II discusses more indirect relations. On the narrower point
about morphology driving movement operations, the chapters are skeptical and
lend little support to the idea that movement operations are morphologically driv-
en, some arguing that, in fact, changes in movement operations are independent of
morphological changes (Part III). The results reported here may be negative to a
large degree with respect to this specific hypothesis, but a negative result can be
useful and the studies present much new material on correlations between morpho-
logical changes and changes in syntactic structures, which will be useful in stimu-
lating and testing other hypotheses.
So, syntactic effects of morphological changes—what might that encompass?
There is a line of work not represented in this volume, perhaps curiously, and that
has to do with morphology and category membership. Part of the task of the child
acquiring a language is to determine the categories to which words belong: which


Under the bottom-up approach to phrase structure, ‘movement’ consists of copying and deletion.
What is merged into a tree in its ‘understood’ position, yielding a structure like VP[buy what], which may
be extended to IP[Jay Ican VP[buy what]]. Then Ican may be copied and deleted; then similarly what, to
yield the structure What can IP[Jay VP[buy]], for What can Jay buy?
 Syntactic Effects of Morphological Change

are the nouns, the verbs, the inflectional elements, and so on? There is good reason
to believe that this is done on the basis not of notional definitions (‘a noun is the
name of a person, place or thing’) but on the basis of distributional and formal
properties. Items which occur in the same positions and undergo the same com-
putational operations belong to the same category. Similarly, items which look
alike and have the same morphological properties belong to the same category. Fur-
thermore, historical linguists have enriched ideas along these lines by identifying
changes in category membership and linking them to morphological changes. For
example, there has been much discussion of the emergence in English of a distinct
class of inflectional items like must, can, will, should, etc., the so-called modal
auxiliaries, which had been more or less normal verbs in earlier forms of English
(Lightfoot a, Plank , Denison , Roberts , Warner , among
others). This category change, complete by the early sixteenth century, has been
attributed to prior morphological changes which distinguished these items from
verbs. The morphological change consisted in the wholesale loss of verbal inflec-
tion in early Middle English, with the exception of the third person singular suffix
in -s. This had the effect of distinguishing the former preterite-present class of
verbs from other verbs, because they never had the third person singular ending.
The modal auxiliaries had all belonged to this class, which now became categori-
ally distinct, identified by the absence of a third person singular suffix in -s, the
only surviving morphological property of verbs, and by the lack of a transparent,
time-based relation between past and present tense forms. The evidence for the
categorial change consists of changes in the distribution of must, can, etc: they
could no longer occur with perfective and progressive markers, in infinitivals, or
in sequences with one true verb. The reason for the change was the prior morpho-
logical simplification, which changed the basis on which children identified verbs.
See Lightfoot (, §.) for a recent account of the details. Similarly Warner
() showed that the verb be underwent what was in effect a categorial change
quite recently, in nineteenth-century grammars, and that this was triggered in part
by the loss of second person singular suffixes in -est. There are many other ex-
amples. So morphological changes may have syntactic effects through changes in
category membership.

..   

If this is the kind of thing that the volume does not include, what does it feature?
In brief, a wide range of hypotheses. As noted, much of the research leading to
these chapters was stimulated by Minimalist ideas about the role of morphology
in requiring syntactic movement operations, but the resulting proposals require
enriching ideas about the role of morphology and have consequences which go
well beyond current Minimalist hypotheses. Some authors posit a more or less
direct link between morphological properties and syntactic operations, arguing
that certain syntactic changes were a direct function of changing morphology,
 Introduction 

although no chapter here makes the specific claim that movement operations (or
copying; see note ) are morphologically driven. For example, Ian Roberts and
Anna Roussou consider three well-known cases of ‘grammaticalization’ regard-
ing the development of future-tense constructions in English, Romance, and Greek.
They argue in each instance that the change consists in the loss of movement of a
verbal element to a higher functional position; in the new grammars the verbal ele-
ment is merged directly in the functional position, not merging there via copying
and subsequent deletion of the lower element (i.e. movement, in earlier formula-
tions). They argue intriguingly that grammaticalization-type changes of this kind
follow a ‘path’, which is structurally defined along Cinque’s () universal hier-
archy of functional categories in the clause structure. An internal property of UG
prefers grammars without movement operations, so there is always a tendency for
grammars to lose movement operations. This historical tendency is prevented if
morphological properties require the movement. Until about  English modal
verbs like may, must, etc. had infinitives marked with an -e(n) ending; as long as
infinitives had this ending, there was evidence for a lower T position, hence a bi-
clausal structure where may, must, etc. were verbs which raise to a higher function-
al position. Once the infinitive ending was lost (which happened only in English
and in no other Germanic language), there was no evidence for bi-clausal structure
and therefore for movement of a modal verb to a higher T position. Consequently
children were free to follow the preferences of UG and to merge elements direct-
ly in their ‘surface’ position, without movement, and the new syntactic structure
resulted from the loss of infinitival morphology. The grammatical change is that ele-
ments like will come to be base-generated in an inflection position and not as verbs,
and the change is due to a prior morphological shift, as in the account given in
the previous paragraph, but the key morphological shift is different. Under this
view, children are equipped with Cinque’s elaborate structures but they need to
learn the position of various adverbs, which requires subtle judgments about the
distribution and semantic scope of adverbs; those judgments must constitute the
primary linguistic data for Roberts and Roussou, the linguistic experience which
determines the shape of the mature grammar.
Old English allowed ‘split genitives’ like Ælfredes godsune cyninges ‘King
Alfred’s godson’, where Ælfredes and cyninges are in apposition to each other,
both marked for genitive case, but ‘split’ in the sense that they occur on either
side of the head noun godsune. Cynthia Allen considers the Middle English intro-
duction of new split genitives, where the rightmost element is not case marked:
the king’s son Alfred, meaning ‘King Alfred’s son’. She argues (contra Lightfoot
) that this is not a direct effect of the loss of genitive case, which continues to
be attested after the introduction of these caseless appositives, but rather an effect
of the loss of another morphological property, agreement. In the Old English ex-
ample, the case-marking on the rightmost cyninges is a kind of agreement with
Ælfredes, but the new caseless appositive appears just at the time when agreement
(e.g. of adjectives with nouns) ceased to be obligatory, a different morphological
 Syntactic Effects of Morphological Change

fact from the loss of case endings but a morphological fact nonetheless. The gen-
eral loss of agreement, she argues, is due to the loss of morphological case distinc-
tions and was triggered by a gradual relaxation of the Morphological Blocking of
Aronoff () and Andrews (), which prevents unspecified forms being used
when a more specific form is available. In his commentary, Željko Bošković asks
how the two genitive elements in these split constructions are composed together
to form a semantic unit. He argues that an overt movement analysis is implausible
and adopts Bošković and Takahashi’s () analysis of scrambling, developed
first for Japanese: the two elements are base-generated in their surface positions
and put together at LF for θ-licensing, because the rightmost noun does not receive
a θ-role in the position in which it is pronounced. This yields some nice predic-
tions and accounts for the observation that the postnominal PP in Middle English
(e.g. the wife’s tale of Bath) has a θ-role of Locative/Source but never of Theme or
Possessor (*the portrait’s painter of Saskia, *the wife’s tale of Jim—see Lightfoot
: §.). Bošković also raises some conceptual questions about what it means
to ‘relax’ a deep principle like Morphological Blocking, which does not have a
natural place within the Minimalist framework and sounds more like a violable
constraint of Optimality Theory.
Eric Haeberli investigates the loss of verb-second effects in English. He offers
a very different analysis from what is usually claimed for the modern Germanic
verb-second languages, that verbs move to a higher C position. He distinguishes
two positions for subjects and argues that non-pronominal subjects may remain
in the lower subject position (below the AGR-head occupied by the finite verb) if
there is an empty expletive in the higher position (above the AGR-head); otherwise
the subject DP must move to that higher position. Early English verb-second with
non-operators results from an empty expletive occupying the higher subject pos-
ition, which precedes the verb. So the loss of verb-second results from the loss
of empty expletives in that higher position. Consequently subjects come to move
to the higher position (a counterexample to the notion that languages tend to lose
movement operations—cf. Roberts and Roussou’s chapter). The new movement
operation, in turn, results from the weakening of verbal agreement, if licensing of
empty pronouns is a function of rich inflectional morphology; therein lies the link
with morphological change.
Two chapters consider ditransitive constructions in English, contrasting the
surviving Give him the book with the obsolete Give the book him, which was
replaced in Middle English by the prepositional Give the book to him. Both chap-
ters begin from traditional observations that rich case-marking permits free word
order and that loss of case-marking entails restricted word order possibilities. Tho-
mas McFadden keys the rise of to-datives to the loss of dative case and views
it as a continuation of the older direct–indirect object order, where the indirect
object was marked for dative case. He adopts Harley’s () adaptation of Lar-
son’s VP-shell analysis and treats a wide range of data from the Penn–Helsinki
Parsed Corpus of Middle English. He found the to-dative first emerging in early
 Introduction 

Middle English, as the morphological case system collapsed in most dialects. He

offers a careful analysis of various movement operations, arguing for a single base
order for to-datives. This permits him to argue that what is constant is the fact that
indirect objects at all stages received oblique case; what changed is that oblique
was once realized by overt case-marking but came later to be realized as a PP.
Chiara Polo refines the correlation between loss of overt case-marking and the
new to-datives by invoking abstract notions of Case and arguing that there was a
distinction between structural accusative Case and inherent dative Case of DP/NPs
as long as the third person pronouns manifested the comparable morphological
distinction. In other words, even though DP/NPs ceased to be marked for dative
case morphologically, they had abstract dative Case until the pronouns became
indistinct somewhat later than nouns. At that point and only at that point was the
abstract accusative–dative distinction lost. This analysis echoes that of Lightfoot
() but Polo argues for it on the basis of ditransitive constructions. Her data,
drawn mostly from the Peterborough Chronicle, are slender but instructive. In the
First Continuation she finds structures like Gæf þone abbotrice of Burch an abbot
‘gave the abbey of Peterborough to an abbot’, where abbot is not overtly marked
for case but is, she claims, abstractly dative. In the Second Continuation, however,
and in late Middle English texts that she examines, such forms are replaced by
the prepositional Underþeden ðat mynstre to Clunie ‘make-subject the monastery
to Cluny’. Polo keys this change in the distribution of DP/NPs, not to their own
inflectional properties but to the inflectional properties of pronouns: as long as
pronouns had distinct forms for dative and accusative cases, there was an abstract
dative Case which determined the distribution of all DP/NPs.
Two chapters investigate the consequences of the reduction of verbal inflection
in Brazilian Portuguese. Acrisio Pires provides a cue-based approach to the devel-
opment of inflected infinitives, shows how the loss of inflected infinitives has led to
the loss of pro subjects in infinitives, and argues that only non-inflected infinitives
occur in obligatory control structures. He follows Hornstein () in treating
control as the result of a movement operation and he predicts two syntactic conse-
quences of the loss of inflected infinitives in Colloquial Brazilian Portuguese: first,
there must be an alternative way to license overt subjects of infinitivals, which does
not depend on inflection; second, uninflected infinitives cease to display non-oblig-
atory-control effects (except of arbitrary PRO interpretations). Cilene Rodrigues
also adopts Hornstein’s treatment of control as involving a movement operation,
motivated by the need to check nominative case. She argues that null subjects in
Brazilian Portuguese are quite different from null subjects in the other Romance
languages. She observes that referential null subjects are permitted in matrix claus-
es if the verb is marked as first person singular; they do not occur in other contexts.
Null subjects occur in embedded clauses with an inflected verb; however, these are
not referential null subjects but rather control structures, resulting from movement
of the subject DP/NP. She correlates the loss of verbal inflection with changes in
the properties of Brazilian Portuguese null subjects: reduced inflection prompts
 Syntactic Effects of Morphological Change

loss of referential null subjects. She also argues that the changing structure of the
φ-feature properties of clauses permits movement from what appear to be finite
clauses.
Akira Watanabe posits strong connections between movement operations and
the inflectional properties of verbs. He examines the loss of overt wh-movement in
the Heian period of Japanese (ninth to twelfth centuries), arguing that this entailed
the loss of ‘Kakarimusubi’, an agreement phenomenon by which a wh-phrase was
accompanied by the focus particle ka, which in turn required that the verb take
on the adnominal ending. He analyses the focus particle as driving wh-movement
and that, once the operation of wh-movement was lost, the particle was reanalyzed
first as a polarity quantificational element in rhetorical questions and then as a
more general existential quantifier. This, in turn, entailed changes in verb inflection,
by which ‘conclusive’ forms were replaced by adnominals, which were no longer
associated with wh-movement.
Thórhallur Eythórsson examines two changes in case assignment with oblique
subjects in Icelandic, so-called Dative Sickness, which substitutes dative case
for earlier accusatives, and Nominative Sickness, by which accusative or dative
DP/NPs occurring with the finite verb in the third person become nominative and
agree with the verb in person and number. He rejects an explanation of Nomina-
tive Sickness as being due to the loss of morphological case, because Icelandic has
preserved its case endings for centuries. He discusses these changes in terms of
changes in the abstract Case system. He argues that the two phenomena are condi-
tioned by different factors relating to the distinction between structural and inher-
ent Case, in combination with a frequency factor determining the directionality
of morphosyntactic leveling. He adopts a cue-based approach to acquisition and
claims that morphological case forms signal the set of abstract structures to which
Case is assigned. So children overgeneralize the use of nominative case for subject
DP/NPs, in much the same way that English-speaking children overgeneralize the
use of -ed past tense forms. Similarly, the morphological dative case on all but a
small number of oblique Experiencers is overgeneralized at the expense of accusa-
tive subjects with these verbs. The result is a grammar with thematic (dative) case
on subject DP/NPs.
All the chapters in this section show syntactic consequences of the loss of
morphology, but in no instance do they discuss operations where a category moves
in order to pick up morphological features.

..       

Two writers follow Bobaljik and Thráinsson (), whereby functional categories
may vary from one grammar to another, and may vary within one grammar from
one clause type to another. Children may assume only those functional categories
that they have evidence for, Thráinsson’s Real Minimalist Principle. Bobaljik and
Thráinsson proposed an indirect relationship between verbal morphology and verb
 Introduction 

movement. A language that has a ‘split’ verbal morphology (e.g. distinct inflec-
tional markers for tense and agreement) will have a ‘split IP’ (distinct functional
projections for Tense and Agreement) and hence obligatory movement of the verb
out of the VP to each of these positions. Since split morphology is not the only evi-
dence for a split IP available to the language learner, there could be a ‘lag’ between
the loss of complex verbal morphology and the loss of a split IP in the historical
development of a given language. So a language might show syntactic evidence for
verb movement after it has lost complex verb morphology.
Dirk Bury examines the loss of verb-second in Welsh, keying the loss not to
morphological changes but to the innovation of a main clause complementizer. He
follows ideas from the Minimalist Program and analyzes verb movement as a copy-
ing operation (‘self-attachment’), which entails that clause structure is not univer-
sal: he follows Bobaljik and Thráinsson in allowing a functional head to occur only
if it is realized overtly, perhaps by a moved verb. This yields a restrictive theory
of constituent structure which predicts the key properties of verb-second construc-
tions and entails some interesting claims about VSO languages, notably that they
must have preverbal particles for the verb to adjoin to. Furthermore, Bury’s system
predicts that verb-second languages lack (projecting) preverbal particles and there-
fore that a verb-second language changing to VSO will develop a projecting VP-
external head. This ambitious chapter then considers analyses which are forced
by this highly restrictive theory of clause structure for Middle Welsh, which is
verb-second, and for Modern Welsh, which is VSO. He reconsiders Willis’s ()
analysis of changes affecting Welsh preverbal particles in the sixteenth century,
which yielded a new main clause complementizer, fe. This precipitated a change to
VSO structures and Bury’s account predicts the simultaneity of the changes.
Ana Maria Martins also adopts Bobaljik and Thráinsson’s approach. She ana-
lyzes apparent optional scrambling in Old Portuguese and its subsequent loss.
Scrambling affected DPs, PPs, APs, and small clauses alike and it was lost for all
these categories more or less simultaneously. Since the distribution of PPs, APs,
and small clauses is not driven by Case, she is skeptical that the diachronic changes
were driven by changes in overt case-marking. Instead, she argues for changes
in phrase structure. In Old Portuguese AgrS allowed multiple specifiers, hosting
scrambled objects as well as the subject. So the diachronic change was one in
which AgrS ceased to allow multiple specifiers. This, in turn, was a function of the
loss of clitic structures where the clitic is separated from the verb by intervening
material, so-called interpolation. Clitics had been reliable indicators of or ‘cues’
for a scrambling position, but they came to occur only adjacent to the verb and
therefore the scrambled SOV option ceased to be an option.

..     

Skeptical authors discuss syntactic changes which have been linked to morpho-
logical change, where it has been suggested that items move (are copied) in order
 Syntactic Effects of Morphological Change

to pick up inflectional features, but they argue against such links, claiming that
specific syntactic changes are not caused by prior morphological changes. For ex-
ample, Dianne Jonas discusses the loss of the V-to-I operation. She adopts the
usual assumption that this operation was lost in the Mainland Scandinavian lan-
guages but not in Icelandic. The mainland languages have lost much of their ver-
bal inflection, but Icelandic has remained resolutely unchanged in its morphology.
This has led scholars to link the loss of V-to-I with the loss of morphological
richness, along the lines suggested in Section  (but note the problems observed
there). Jonas’s chapter investigates Faroese and Shetland Dialect, which have an
intermediate status: there is a partial loss of V-to-I and a moderate richness of
inflection, and she sketches the diachronic developments. She compares Shetland
Dialect with (Older) Scots in this regard and concludes that, when one examines
the details, the loss of V-to-I diachronically does not follow directly from the loss
or reduction in verbal inflection. Languages with the same morphology may differ
in whether they have the V-to-I operation. Therefore, whether or not a grammar
has the V-to-I operation results not from exposure to a certain kind of morphology
but rather from exposure directly to syntactic triggers, where the verb has clearly
moved out of the VP, to the left of the subject DP/NP or to the left of a negative
or adverb. In his commentary Stephen Anderson comments on the implausibly
‘legalistic’ correlations in the literature, whereby grammars are said to have V-to-I
raising if verbal paradigms distinguish first from second person in at least one tense
(Rohrbacher : ) or if person morphology is found in all tenses (Vikner
: ), etc. Under this view, morphology drives syntax but the driving force
is not the morphology of individual verbs but a property of the whole paradigm.
He accepts Jonas’s conclusions that presence of the V-to-I operation does not cor-
relate with morphological distinctions, and he emphasizes that syntacticians can-
not expect morphology to indicate syntactic structure in some isomorphic fashion.
Susan Pintzuk makes an argument similar to Jonas’s. She notes the traditional
observation that languages with overt case-marking have freer word order than
those with little or no nominal inflection and she considers modern work linking
change in verb–complement order in the history of English to change in the
strength of the case feature forcing objects to move to a preverbal position. For
example, Roberts () postulated a Kayne-style universal underlying subject–
verb–object order; in an object–verb grammar the object DP/NP moves to the
left of the verb, attracted there by a case feature, an instance of morphologically
driven movement. Roberts argued that the strength of this feature weakened in
Middle English, leaving objects in their base-generated, postverbal position and
yielding SVO order. Pintzuk rejects the movement analysis in favor of a direction-
ality parameter involving two possible base orders, object–verb and verb–object,
and she argues that the morphological case system plays no role in determining
the position of the complement: postverbal complements increase steadily through
the Old English period, despite the fact that the case system remains robust, and
furthermore there were no special changes in word order in early Middle English,
 Introduction 

when there was a significant loss of morphology. She treats the change from object–
verb to verb–object as a function of competing grammars, not driven by mor-
phological changes in any way. In his commentary, Jairo Nunes discusses Pint-
zuk’s argument for analyzing Old English in terms of a directionality parameter
rather than in terms of a Kaynean universal head–complement order. He accepts
Pintzuk’s point that morphological case does not seem to motivate the relevant
DP/NP-movement but he modifies Roberts’s analysis of Old English as underly-
ingly subject–verb–object with a leftward-movement operation (moving the direct
object DP to the specifier of a higher functional category) and argues that the modi-
fied movement analysis fares at least as well as the directionality parameter and
that therefore it might be productive to search for something else requiring move-
ment of the object DP/NP.
Susana Bejar adopts a cue-based approach to acquisition and argues that
A-movement is triggered not by morphological properties but by syntactic cues,
and indeed that A-movement and morphology are orthogonal to each other. If
A-movement is triggered by abstract Case properties, then the question arises to
what extent abstract Case properties can be determined by morphological proper-
ties. Bejar argues that morphology is neither necessary nor sufficient for deter-
mining whether Case is thematically or structurally licensed and that the crucial
cue for determining whether a Case is structurally or thematically licensed is the
(in)ability to undergo A-movement, a syntactic matter. She examines the loss of
dative experiencers in English and argues that loss of morphology correlates with
the emergence of A-movement in these constructions and that the relation between
overt morphology and movement in such instances is closer to being disjunctive
rather than conjunctive. Movement and overt morphology are independent trig-
gers for abstract Case; Case may trigger movement and therefore be acquired inde-
pendently of overt morphological properties.
The last of our skeptics, John Sundquist, investigates Object Shift in the Scandi-
navian languages, where Danish, Swedish, and Norwegian have undergone signifi-
cant changes in nominal and verbal inflection, but at different rates, quite different
from the far more conservative Icelandic. He finds no syntactic changes in Middle
Norwegian despite the loss of case endings on nouns, which can be tracked in a
large corpus of diplomatic letters, and he argues that Object Shift operates regard-
less of case inflection. It is driven instead by semantic forces, that move unfocused
elements out of the VP. He argues that Middle Norwegian had both Object Shift
and scrambling (an unusual state of affairs among the modern VO languages, prob-
lematic for Holmberg’s Generalization), and competing OV and VO structures (cf.
Pintzuk), until OV died out toward the end of the fifteenth century.

..  

The volume concludes with two chapters on computer models of change, both
inspired by models of population genetics, but taking different approaches to lan-
 Syntactic Effects of Morphological Change

guage acquisition. They do not deal with the relationship between morphological
and syntactic change, but the computer models represent an important develop-
ment in diachronic studies, as noted earlier, linking changes in individual gram-
mars to the group phenomenon of language change. Partha Niyogi introduces the
basic framework for the computational study of diachronic phenomena in human
languages and takes stock of the explosion of activity in computational models
of language change and evolution. He focuses on the subtle interplay between
language acquisition and change, building a population model which helps us to
reason our way through the complex explanatory possibilities for language popula-
tions changing in time, analogously to recent practice in evolutionary biology. He
uses the Triggering Learning Algorithm approach to acquisition, in which children
are error-driven, changing their grammars when they confront a sentence which
cannot be analyzed by the present grammar (Gibson and Wexler ). Charles
Yang also develops a formal population model characterizing the interplay of lan-
guage acquisition and change. He focuses on notions of grammar competition
driving change, providing a link between the statistical properties of linguistic
evidence and the different grammars that children attain at different stages of the
language; the distribution of grammars changes as an adaptive response to the lin-
guistic evidence in the environment. He offers a ‘fundamental theorem’, that a
grammar wins out if its advantage is greater. Similarly, variation observed during
historical change is attributed to distributional changes in grammars over succes-
sive generations of speakers. He illustrates the model with an analysis of the loss
of verb-second in Old French.

.. 
Many years ago Stephen Anderson () invoked the image of a deer in a forest:
one may not see that there is a deer present until it moves. Change can reveal
structure and language change can sometimes reveal the properties of grammars.
Recently linguists have restored to the research agenda questions about the relation
between morphology and syntax, particularly movement (copying) operations.
This volume offers a set of case studies on the effects of morphological change,
with a range of hypotheses about the relationship between morphology and syn-
tax. Changing morphology certainly has consequences for syntax; several chapters
make specific connections and other work, not represented here, has emphasized
the role of morphology in determining category membership (e.g. Lightfoot :
–) . However, there is no simple isomorphic relationship, as far as one can
tell. Furthermore, no chapter here argues for a connection whereby morphology
drives movement operations; indeed, our skeptical authors argue against such for-
mulations. That suggests the possibility of a one-way relationship, that rich morph-
ology entails movement operations, but not a two-way relationship, whereby, in
addition, poor morphology entails lack of movement operations (Lightfoot :
–, Bobaljik ). If the relationship is indeed one-way, interesting questions
 Introduction 

arise about how it may be captured, but that goes beyond where the chapters of this
volume take us. These detailed diachronic studies will be useful for theoreticians
investigating morphology–syntax connections, I hope, long after their theoretical
hypotheses and the technology in which they are couched have been revised out of
recognition.