Royal Institute of Philosophy

The Theory of Translation

Author(s): W. Haas
Source: Philosophy, Vol. 37, No. 141 (Jul., 1962), pp. 208-228
Published by: Cambridge University Press on behalf of Royal Institute of Philosophy
Stable URL: https://www.jstor.org/stable/3748438
Accessed: 11-02-2020 09:42 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Royal Institute of Philosophy, Cambridge University Press are collaborating with JSTOR
to digitize, preserve and extend access to Philosophy

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 THE THEORY OF TRANSLATION

W. HAAS

To translate is one thing; to say how we do it, is another. The

practice is familiar enough, and there are familiar theories of it.
But when we try to look more closely, theory tends to obscure rather
than explain, and the familiar practice-an ancient practice,
without which Western civilisation is unthinkable-appears to be
just baffling, its very possibility a mystery.
To translate, Dr Johnson tells us, is 'to change into another
language, retaining the sense'; and it is easy to agree with him. But
can we think it out? How do we effect this exchange of languages?
Is it like changing horses or carriages? And what, exactly, is it that
we retain? Images are powerful instruments of interpretation. But
this one, the image of something carried across from one vehicle to
another, can it bear the weight we put upon it?
At first sight, this is what we are tempted to make of translation
-an operation with three terms: two expressions, and a meaning
they share. When we translate, we seem to establish a relation of
three distinct entities, each separately apprehended: the two
expressions seen on paper or heard in the air, and the meaning in
the translator's mind. The meaning, presumably, we 'retain' and
translate; we 'transfer' it from one expression to the other. Strictly,
then, when a sentence or speech or novel is translated, say, from
French into English, what is supposed to be translated or transferred
is not the French sentence or speech or novel at all; it is something
utterly different, something inaudible and invisible-'the meaning'
itself, which is not in French nor in English nor in any language
whatever.
This interpretation of translation as a triadic relation does, of
course, accord with some deeply ingrained habits of thought. It
conforms with a model which we are inclined to apply to all con-
scious and voluntary manifestations of human life. As a human
being might be thought of as the temporary embodiment of an
independent soul, so a spoken sentence is regarded as the temporary
expression of an independent meaning. The translator might then
be said to effect a migration of meanings. Translation is supposed
to be possible on account of a twofold relation of an entity, called
'meaning'; two expressions are viewed as 'vehicles' of the same
meaning. Thus:

208

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

Sign 1

Expression 1 ->.Meaning - Expression 2

( )

Si
What is cardinal, here, is a theory of meaning, which interprets
'significant expression' ('sign', in linguistic terminology) as con-
stituted by a relation of two distinct entities: an expression and a
meaning. The relation itself is mysterious. The vehicle has a ghostly
passenger. Inevitably, a 'triadic' theory of translation implies some
form of a 'dualist', and therefore mysterious, theory of meaning.
There can be no doubt that we should be hard put to it, if having
done some particular piece of translation, we were asked to explain
it in terms of this theory. We should only be aware of having operated
with expressions, and we could say something about such operations.
We should be able to explain the difference between a good and a
bad translation in terms of their respective expressions; we should
refer to their occurrences among other expressions, and among
persons and things. To take an example, we might argue that a
famous poem by Goethe about Italy which begins with the line:
'Kennst du das Land, wo die Zitronen bl/hn?' has been translate
badly, when in English it was made to begin: 'Knowst thou the
land . . .'. Our argument would be that 'Kennst du . . .' is a straight-
forward piece of colloquial language, as one might say "Do you
know the shop where the sale is on ?" ' (Kennst du den Laden/ . . .
Herrn Schinidt/ . . . meinen Bruder, etc.), whereas ' knowest thou'
would be unusual in corresponding contexts. Again, we should say
that 'Land' occurs in very many contexts ('Stadt und Land, Ausland,
von Land zu Land', die sfidlichen Ldnder Europas, etc.), where the
corresponding English word would be 'country' ('town and country,
foreign country, from country to country, southern countries', etc.),
rather than 'land'.1 At the same time, we shall reject equally 'Do you
know the country': not only because at times German Land will also
correspond to English land, rather than country-in contexts such as
'landscape' (Landschft) 'land of promise' (gelobtes Land), 'land
of dreams' (Land der Trdume)-but mainly because in such a
translation, the important and far from common rhythm of the
poem, which is established with this its first line, would be lost. In
fact, we should not be able, here, to translate word for word, or even
sentence for sentence; the line we have to find will have to be
internally very different from the original if it is to preserve a
comparable role within the poem as a whole. And having chosen,
11 borrow the example from Professor L. Forster's Translation, in 'Aspects of
Translation' (Studies in Communication 2, 1958).

B 209

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

rejected and accepted, we should explain what makes a better or a

worse translation by saying to what extent there is a correspondence
between (i) the habitual contexts (verbal and situational) of the
given original expressions, and (ii) the contextual relations of the
expressions used in the translation. But what, in any particular case,
could we say, or be expected to say, about operations with pure
ideas? A triadic theory of the craft of translation, if it were accepted
at all-as, for want of a better, it might be-could not be regarded
as a working-theory, not as a general account of how we do what we
do. We should have to look upon it as some kind of 'ulterior explana-
tion' of the finished work-'mere' theory, 'pure' theory, free from
empirical tests, and devoid of technical implications.
It would seem to have fallen to the philosopher to deal with the
difficulties of disembodied meanings. If there are such things,
where do we find them? How observe them-those naked ideas
under their changeable verbal clothing? (Another favourite
metaphor, this, besides the 'vehicle of sense' !) Do we ever find them
without their verbal clothes, just in their natural state ?
We are familiar with various attempts to deal with these diffi-
culties. Generally, the dualist scheme of the linguistic sign is pre-
served, and with it the basic triadic scheme of translation. It is
within these limits that the attempt is made to rescue 'pure meanings'
from their shadowy existence; generally, by tying them to, or even
replacing them by, 'pure' physical facts. Meanings are then said
to be 'references' to such facts, 'denotations'; and one expression
is supposed to be a translation of another, if both have the same
denotation. It is true that this is rarely considered to be quite
enough. The two expressions, in addition to their denotation, would
be required to share an aptitude for calling forth certain responses,
certain 'emotive overtones', and they would have to incorporate a
number of purely syntactical operations. There would be a large
bag of tricks-some more, some less important-but all of them,
mere accessories to communication. They either presuppose
'reference', or, like a sigh or a smile, can dispense with language
altogether. The core of meaning is supposed to be denotation, factual
reference.
There seem to be two main variants of the reference-theory of
meaning. There are, firstly, those following Ogden and Richards'
who would try to tie 'ideas' or 'meanings' (or, as they call them,
'references') to external things ('referents'), giving us a three-term
elaboration of the 'dualist' theory of linguistic signs:

Expression -* (Reference -* Referent)

1C. K. Ogden and I. A. Richards, The Meaning of Meaning, Ch. 1.

210

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

Expressions, here, refer indirectly. We may call this the theory

of indirect reference. It gives us a five-term elaboration of the triadic
scheme for translation:

Expression 1 -?[Reference 1 --Referent*- Reference 2]*-<Expression 2

The middle term is presented as being, itself, a relation of three
terms: namely, of two 'ideas' or 'meanings' or 'references' (each
peculiar to the speakers of a particular language) to the same
external things (the same 'referent').'
Others have persuaded themselves that-in theory at any rate-
we can do away with specifically mental facts; that 'reference' need
not involve any 'thing in the- mind'. This gives us again a simple
dyadic relation for the linguistic sign:
Expression --->Referent,
the reference of the expression (the arrow) being explained as a
physiological disposition or habit we have of using that expression
for denoting certain facts. We might call this a theory of direct
reference. It presents translation as a simple relational scheme of
just three physical terms:

Expression 1 ---Referent ---Expression 2

This, it might seem at first sight, dispels the mystery. We seem to
have succeeded in avoiding the puzzle of psycho-physical relations.
For we are supposed to operate with ordinary external facts only
-with expressions amongst persons and things. On closer examina-
tion, however, the puzzles turn out to be still with us. The crucial
relation we are said to establish when using or translating an
expression is still a relation of correspondence between two distinct
orders of thing, linguistic and extralingual. This is why, even in the
case of a theory of direct reference, we may still speak of a dualist
view of linguistic signs; and also, why we find that we are still
mystified about their meaning, and about translation.
I am not here concerned with the relative merits of different
dualist theories of meaning; and I have no interpretation of my own
to offer of that opaque and puzzling something which is supposed to
'correspond' to linguistic expressions. I have nothing to say of that
extralingual second term of the alleged sign-relation, the middle
term of the alleged triadic scheme of translation. Rather, I am
concerned to show that the 'of' in 'meaning of' cannot be interpreted
as a relation of correspondence between two orders of fact, and that
translation is not an operation with three terms. If there are such
entities as are postulated in a dualist theory of sign and a triadic
theory of translation-if there are pure meanings or pure external
'Cf. the interesting paper by C. Rabin on 'The Linguistics of Translation' in
Aspects of Translation, p. 125.

211

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

facts, there is certainly nothing we can say about them. We cannot

rescue the former from their occult state by tying them to, or
replacing them by, the latter. The facts, or referents so-called, if
supposed to be grasped independently of any and every language,
are themselves as shadowy and nebulous as the naked ideas they are
meant to reinforce or to replace.
Expressions 'have meanings', and they are, themselves, external
things amongst others. But both external things and meanings
dissolve in a dualist interpretation of linguistic signs. They are
assigned positions which they cannot occupy; they are placed
beyond the reach of the conceptual tools of language.

II.

That meanings cannot survive in a dualist theory of signs has been

shown-and shown with admirable clarity-by some of the more
recent studies in the philosophy of language.' 'What an expression
means' cannot be found as a separate entity beside the expression.
If we insist on having it this way, expression will have nothing to
express, and reference nothing to refer to. Meanings, we have
learned, are not entities or objects corresponding to expressions;
they are the uses of expressions; they are the work expressions do.
It remains true that the meaning of an utterance is not in it. But
neither is it an object beside it. It includes and transcends the utter-
ance, as my walking includes and transcends my legs. What an
expression means, is not an object confronting it, any more than my
walking is an object confronting my legs.
The instrumental view of language allows us to discover what is
discoverable about meanings, ideas, concepts, propositions; and it
delivers us from some very common temptations to pursue chimeras.
It is important, though, to observe, where exactly the line is
drawn between fancy and fact. Errors of mistaken identity are not
uncommon. Especially, there seems to be some inclination to assume
(i) that mental facts are among the chimeras, and (ii) that 'bare
facts' confronting expressions are not. I shall try to deal with these
two misconceptions.

II (I). First, an instrumental interpretation of meaning does not

entail any denial of mental events. On the contrary, it seems to
imply that events which we commonly describe as mental, rather
than physical, do occur. There must be memories, organised
memories. For no single use of a word can establish it as significant.

'Cf. L. Wittgenstein, The Blue and Brown Books, Philosophical Investigations.

G. Ryle, Ordinary Language (Phil. Review, 1953), 'The Theory of Meaning' in
British Philosophy in the Mid-century. J. R. Firth, Papers in Linguistics (chs. 3, 14-16).

212

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

The meaning of a word is a collection, an organised recollection, of

many individual uses of it, i.e. of various occurrences of it: in verbal
and non-verbal contexts, and in positions in which it contrasts with
other words. Meaning (like skill) is an 'acquired property'. Whenever
a word is being used significantly, another use is added to remembered
uses of it; a present context joins the previous ones. Clearly, the
organised memories of a word's uses are what would ordinarily be
described as mental events. So would its present choice from a number
of contrasting words. There are relations, even relations of cor-
respondence, between a word's present employment and those
other employments which we remember. Moreover, any word I am
using now may be abstracted from its context and be treated as a
physical fact (for instance, by acousticians or phoneticians or by
lexicographers making an entry), while, without its being presently
used, its past uses may be 'recalled to mind' (e.g. in a dictionary
paraphrase or definition). Signs are souvenirs. When active and
actually employed in a new context, they act as reminders of
contexts past, or they could mean nothing.
Have we come full circle? Are we back at physical symbols as
'vehicles' of 'bare ideas'? Not quite; and the difference is important.
What an inert physical expression may remind us of is not any
unverbalised pure idea.. What we remember is that same expression
in past employments: both among other expressions and in contrast
to other expressions. Our memories or ideas are not extralingual, not
'without language'.
The difference between this articulate account of meaning and
the traditional 'vehicle' theory appears most clearly, when we
consider translation. Even if 'the meaning' of an expression is
identified with a recollection of its past uses: such a collection of
previous occurrences of an expression-in a variety of verbal as
well as non-verbal contexts, and in contrast with a variety of other
expressions-could not possibly qualify as that kind of 'pure idea',
which is supposed to be indifferent to its linguistic setting, and,
therefore, transportable from one linguistic vehicle to another.
Here, meaning is not an entity beside the expression; it is a particular
expression at work, actual work and remembered work. Such work
is not a piece of transferable freight. It cannot be transported to
another expression in another language, any more than the 'goodwill'
of a shop can be transported 'to another shop in another town. It can
be transferred to another user of the same shop in the same town.
Another shop in another town can only parallel its 'goodwill'. What
we have found is no 'bare idea'. We have found expressions which
recall expressions in use. All the facts we are dealing with-the
expressions and their environments-are on the same plane: they

213

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

may be actually perceived as external things, and they may be

remembered.
Memory is, indeed, a problem. So is choice. But neither is a mere
enigma. Our problem-and this is our gain-has become articulated,
familiar, and manageable. We are familiar with relating our habits
or memories to present experience. We know how to trace acquired
skill in the carpenter's use of a hammer; and we know how to trace
acquired significance in a man's use of a word. The meaning of a
word has a history; we may have records of its past uses. But we
cannot know even what it could be like to perceive a carpenter's skill,
as another thing beside his tools, or a word-less idea as a separate
thing beside a word. We have no way of relating present things (such
as hammers being handled, or words being uttered) to objects on
'another plane'. What an expression 'conveys' is not a passenger from
another world. Its meaning, a bequest from its past, is related to a
given word in some such way as yesterday's walk is related to my
legs here and now. To be sure, I may recall it to mind-yesterday's
walk or the meaning. But a walk, whether present or remembered,
is not a legless affair; and 'what a word means' is not a word-less
idea. Nor is it a word-less physical fact. 'Bare facts' are as diaphanous
as 'bare ideas': this is my second point about the dividing line
between fancy and fact.

II (2). Some, who profess to accept the view that 'the meaning of
an expression is its use', seem to claim that they are only making
this view more articulate by telling us, more specifically about the
'use' of expressions, that it consists in referring them to extra-lingual
facts in our physical environment. Essentially, they say, meaning
(or 'the use' of an expression) is denotation. In this way, the instru-
mental approach to language appears to be assimilated to a theory
of reference. This appearance of an amalgamation of the two theories
seems to me utterly deceptive. Expressions cannot be 'used' for
referring to bare and neutral facts.
Denotation has of course been queried recently. It has been argued
that though there are some expressions which do denote, 'there is
not one basic mould, such as the "Fido"-Fido mould, into which
all significant expressions are to be forced.' Even such as may be
said to 'denote' are found to do a good many other things besides.
(This is why, for instance, 'the Morning-Star' would not pass as an
adequate translation of 'l'Etoile du Soir', though the two have the
same denotation).' These denials seem to be wholly justified. But
must we not go further? It is hard to see how denotation, as generally
understood, can be credited even with so much as a partial explana-
2G. Ryle, 'The Theory of Meaning' in British Philosophy in the A/lid-Century, 256.

214

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

tion of some meanings. Correspondence to bare extralingual facts

seems to be a mere fiction. It cannot account even for the meaning
of 'Fido', either all or part of it.
I do not deny that there is a genuine operation with expressions
which one may choose to describe as 'denoting' or 'referring'. We
do use expressions for the purpose of referring to things other than
expressions. Our stock of significant expressions may be augmented
by this operation; but only by assigning both the new expression
and the new thing places among other expressions, never by merely
referring one to the other. 'The use of an expression' cannot consist
in referring it to 'bare' extralingual things. It is true that, in the
present climate of opinion, we might feel safer in attaching expres-
sions to extralingual things in physical space than we would in
associating them with extralingual things in a Geistesraim.l But, we
cannot do either.
I am not trying to advocate some kind of Neo-Berkeleyan meta-
physics-some 'to be is to be spoken of'. It would seem to be absurd
to deny the existence of things 'without' language, whether things
physical or things mental-even though, naturally, we could say
nothing about such existence. Affinities with Kantian epistemology
would appear to be far more plausible. Having persuaded ourselves
that conceptual thought, for all we can say about it, consists in
'operating with words',2 it appears that things outside language-
i.e. things unaffected by our operations with words-are something
as opaque, and unprofitable to 'refer to', as the Kantian 'thing in
itself'. Our world-remembered, imagined, or perceived-is organ-
ised by the language we speak.
What are we to make, then, of the notion of 'extralingual refer-
ence' ?-There is a perfectly sensible interpretation of it. But this
cannot tell us how meanings are created in pre- or extra-lingual
space, physical or spiritual. We are familiar with the experience of
having an idea, image, or concept, as yet lacking a word for it, or
the experience of discovering some thing in our physical environ-
ment, without having a name for it. We might ask 'What shall we
call it?' or 'What is it called?' But such experiences are not enough
for extra-lingual reference. Those things we seek a name for are not
extralingual in the required sense. We can always say a great deal
about them; indeed, we may be able to describe them quite ade-
quately, and entirely with the help of words already at our disposal.
The fact that we may want another word for a thing, besides the

1Geistesraum being the most serious drawback of some inquiries into 'semantic
fields' which are otherwise of considerable interest (cf. J. Trier's works, e.g.
Deutsche Bedeutungsforschung in 'Germanische Philologie', 1934).
2G. Ryle, Ordinary Language (Philosophical Review, 1953), p. 185.

215

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

many which are already involved with it, is nothing to establish a

reference-theory of meaning. The least that such a theory requires
is that the thing, which lacks a name, should be capable of being
singled out for reference, without there being any other expressions at al
none except the one by which we want to 'refer' to it. Of this one
we should be able to say what Russell says of the words in his 'object-
language': that it has 'meaning in isolation', or that it has been
learnt 'without its being necessary to have previously learnt other
words.'" What is it that is being asked for, here? Fact or chimera?
What evidence could we possibly have of a word having meaning or
having been learnt in isolation? There might be just one word that
could provide such evidence:-the first word I ever learned. But
what a dubious fiction that would be! Could it be a word at all?
The theory of isolated reference is clearly not meant to be put to
the test of observation. It is in principle unverifiable. Whatever
experience we have of referring to external things, or to ideas, is not
an experience of isolated reference. So far from explaining the mean-
ings of expressions-even of the referring type-reference presupposes
a language of significant expressions.
Even very young children asking 'What is it called?' do not
merely refer to, they can tell us a lot about, 'it' ! For this very reason
it interests them. They have already rejected a large number of
words as inappropriate to the thing, or they would not ask its name.
They have prepared a large number of utterance-frames which the
new name will fit into; and where it will join, and contrast with, a
large number of other words which already fit those frames.-An
'Alsatian'? ... runs, ... barks, . . . is big, . . . has a thick fur, I don't
like . . .s, etc. The question 'What is it?' or 'What is it called?' is
a request to fill in a blank in an indefinite number of incomplete expres-
sions; it is a request, we might say, for notational help in giving new
values to a number of prepared utterance-functions-help in fixing
an organisation of utterances about a new focal term. Long before it
is named, the new thing has already been placed; and it has been
contrasted with other things that run or bark, or are big or have a
thick fur, are not liked, etc. It has been so placed and contrasted by
the help of expressions which were already dealing with it. When I
am looking for a word, 'have it on the tip on my tongue', this is
never a case of some 'pure idea' or 'brute fact' begging a name; it is
always a case of fragmentary utterances seeking completion. The
blank is a variable in a large number of determinate functions; it
has a determinate range, and I can already give it many determinate
and contrasting values. The variable is the 'unknown' in given
expressions. But there is no need to interpret it as an impercep-
2B. Russell, An Inquiry into Meaning and Truth, 65.

216

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 'RANSLATION

tible 'soul' searching for its body, or a 'thing in itself' wanting a

label.
Nobody will deny the existence of extralingual things. But do we
ever come across one requiring a name, without there being a large
number of expressions already engaged with it ?1 Is it not with such
'mere things' as with Wm. James's 'pure experience'? 'Only new-
born babes', he said, 'or men in semi-coma from sleep, drugs, ill-
nesses, or blows, may be assumed to have an experience pure in the
literal sense of a that which is not yet any definite what.'2 Reference
to such 'pure experience' is required by a 'reference theory of
meaning'. This is what Russell has brought out so clearly. If we wish
to explain meaning by 'extralingual reference', then we must insist
on a complete and permanent dualism of two orders of thing-
linguistic expressions on the one hand, and extralingual things on
the other. We must accept the fiction of isolated references. This
dualism cannot be substantiated in any way; and it makes no dif-
ference whether the extralingual things are ideas in the mind or
physical facts. For 'reference' to make its sign-producing link, words
and other things must be supposed to be permanently divided. My
argument is that there is no such division, hence no such link, hence
no theory of meaning in terms of such a link.
We can have no conception of what it might be like to confront
the general blur of a world that is not already prepared and organ-
ised by the use of some signs. Learning a language, or extending it,
is not like drawing a map by putting in one line or one colour at a
time. This we can do; but only because we have another language
at our disposal, which tells us where to draw the lines and what
colours to put in. The map-maker does not confront a world without
language. He is, in fact, a translator-from the language of words
into a restricted language of lines and colours.
Those who have tried seriously to construct a language by means
of operations of 'reference' have in fact usually proceeded in analog-
ous fashion. Like the cartographer, they worked on the basis of a
given language, selecting 'things' already circumscribed. They
selected what was suitable for 'reference', as a cartographer selects
what is suitable for a map. They picked out what their language
allowed them to describe as 'external things', of various kinds. They
did not try to refer to 'bare facts'; and the fiction, often upheld, that
they might have done so, or even that ordinary language might have
done so in the first place, is neither here nor there.
The construction of an 'object-language' can be of considerable

1Cf. L. Wittgenstein, Philosophical Investigations on Ostensive Definition, par.

28 if, especially par. 30/31.
2Wm. James, Essays in Radical Empiricism ('The Thing and Its Relations'), 93 f.

217

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

interest. It may be important to know what can be translated into

a special language which would restrict the meanings of expressions
by rules of reference: e.g. by the rule that the significant use of every
accepted expression should be capable of being accompanied by a
significant pointing-gesture. In this way, rules of reference can restrict
the meanings of expressions within a given language, but they
cannot, by themselves, establish a language. Nothing can come of
mere pointing into a world of bare 'thats'. For (i) pointing itself, if
it is to be of any use as an operation of reference, must have a mean-
ing, which cannot be established by pointing, and (ii) it can have
meaning only as part of a language.
(i) If it is of any use, pointing is no mere gesture, any more than
a word is a mere noise. Like a word, a pointing finger 'has meaning';
and it is a conventional sign. Some communities point with their
chins, others with eyes and brows; we do it with hands and fingers;
pointing is clearly not like laughter or tears: it is not a natural
physiological symptom. For a dog (and presumably for a child) to
learn the meaning of a pointing finger is as difficult as to learn the
meaning of an uttered noise. The gesture-language of pointing may
include a variety of signs: rigid ('beam') pointing, sweeping ('area')
pointing, scanning, pointing at various angles, forward, sideways,
upward, etc. Each of these can acquire meaning only as part of a
language. That is,
(ii) whatever actual experience we have of the gesture-language
of pointing, shows it in the role of an auxiliary language. By itself
it would be hopelessly ambiguous. (At a race, how do we point at
the track, at a horse, its rider, his number, his cap, his whip, his
skill, at the horse's breed, its colour, its speed?) 'Such ambiguity is
commonly resolved by accompanying the pointing with . . . words'
such as 'this track', 'this horse', 'this colour', etc., assuming that the
words 'track', 'horse', etc., are already intelligible.' This assumption
is made, even when we establish or explain the meaning of a new
word by pointing. We cannot establish it by mere pointing. We might
say, for instance, 'This colour is jonquil'. 'The ostensive definition
explains the use-the meaning-of the word when the overall role
of the word in the language is clear',2 i.e. when we know it is some
individual thing, or a colour, or a shape, etc., that is being pointed
at. It appears that I 'must already be master of a language in order
to understand an ostensive definition'.3
Meanings, then, which are established by ostensive definition
cannot be neutral between different languages. Jonquil, for instance,

'W. V. 0. Quine, From a Logical Point of View, 67.

2L. Wittgenstein, Philosophical Investigation, ? 30.
3lbidem, ? 33.

218

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

will be definitely a kind of yellow and no kind of red in English; but

it need be nothing of the sort for someone like a speaker of Bassa,
a language of Liberia, which has only one word to correspond to
both 'red' and 'yellow', even if it were defined for him by apparently
the same pointing-gesture and in the same situation.1
One might, perhaps, try to distinguish between 'what is meant'
and 'what is pointed at', by saying that we may 'point at' the same
thing, even though 'meaning' different things. 'What we point at'
would then be neutral between different languages, though what we
mean when pointing might not be. A reference-theory of meaning,
and of translation, must insist that this is so. But how could we know
this? How could we know that 'what we point at' is neutral between
different languages, if we can never find it except within some
language or other? Strictly speaking, even when I say of two persons,
as I did, that they appear to be witnessing the same pointing-gesture
in the same situation, 'the same' means 'the same in the language in
which I describe it'-e.g. when I say, 'Three people together, one
of them raising his arm'. Of situations so described, I can claim to
know that often different things are 'meant' or 'pointed at' by the
same gesture, for speakers of different languages. I do not know this
by comparing 'what they mean' with something neutral and extra-
lingual that is 'just there'-for I have no access to things outside
language; I know it by a process of translation. I compare the uses of
expressions (including pointing-gestures) which belong to different
languages. I find, for instance, an Eskimo distributing three different
words over the 'same situations' (including the same paintings), in
which an Englishman utters the one word 'snow'. The two, I con-
clude, cannot mean and cannot 'point at' the same thing. There is
no puzzle here: each points at facts which are circumscribed by his
language.
However, the ghost of bare and neutral facts is not an easy one to
lay. Reference theories of meaning have made an effort to rescue
ostensive definitions from their emplacement in particular languages.
A universally valid, 'logical' procedure of generalisation seemed
capable of replacing the caprice of varying linguistic directives.
The operations of 'isolated reference'-pointing, for instance-will
be supplemented: but instead of submitting to linguistic guidance,
we are to rely on something like Mill's Canons of Induction. No
single pointing-gesture or apposite utterance is then supposed to
be sufficient for establishing the meaning of an expression; but a

1It is on record that the botanists required two general colour-terms which
would correspond to the only two colour words of Bassa. They created, but with
reference to their own languages, xanthic and cyanic. Cf. H. A. Gleason, An Intro-
duction to Descriptive Linguistics, 4 f.

219

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHIlLOSOPIIY

number of such would be assumed to accomplish it. Collectively,

they would be credited with establishing a 'habitual association'
between, on the one hand, a set of similar extralingual facts, and,
on the other, a set of similar utterances. In this way, we are supposed
to make utterances significant, indeed to construct a whole language
-a restricted language of reference, a 'primary' or 'object' language
of 'ostensive predicates'.'
It is clear at once that we cannot expect to find examples of
such a language being developed from scratch; it will always
be constructed by someone who is already master of another. This
is why it would be better to make clear how a restricted language
such as that of reference is derived from some given ordinary language,
rather than try to do the reverse.2 However, anyone subscribing to a
'correspondence theory' of meaning has a vested interest in postu-
lating for his object-language 'possible independence' or 'possible
priority', whatever that may mean. Otherwise, his referential
meanings might be derived from other meanings, when they are
supposed to be our original stock-the offspring purely from an
intercourse of human utterances with bare and neutral facts.
Reference alone cannot generate meaning. Can it do so with the
help of induction? Can the repetition and assembly of isolated
references give us the meaning of an expression? This might seem
plausible. After all, one might say, was there not a first significant
expression? And is it not the point of isolated reference simply to
assume of every expression of an object-language that it might have
been the first?
We may well boggle at the pseudo-empiricism of a theory which
requires us to view every significant expression as hidden in some
mythological pre-history of the language to which it belongs-
every expression as capable of having gathered its meaning in that
one dramatic moment, long ago and unremembered, when it might
have been uttered as the first. However, we need waste no time on
these problems of verification, formidable though they are. For
even if we assume that, somehow, empirical sense could be made of
the notion of isolated reference, we should still have to ask whether
any expression, by itself, could possibly be deemed to be significant;
and it seems clear that such an expression, even if repeated and
applied a hundred times, could never be said to have acquired
'Cf. e.g. B. Russell, op. cit., pp. 67, 76; W. V. 0. Quine, op. cit., p. 68. S. K6rner,
Conceptual Thinking, p. 7.
2A special language of reference, constructed for the purposes of logical or
epistemological inquiry, may of course be compared with ordinary language,
without being made its core or source. Of the authors just mentioned, neither
Quine nor Kdrner seems to be interested in 'deriving' ordinary language from. the
referential. Indeed, Professor Quine seems to repudiate the idea (op. cit., p. 78).

220

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

meaning. It could not be viewed as 'having meaning' until, on some

occasion, we found it inappropriate, and said so: 'This and this and
this is a cat. But that is not a cat. It's a dog.'1 A language, one might
say, requires at least two words. Language and meaning take their
origin from difference of meaning. A language could conceivably
be born with two words: and of as many as two, each refers to a
world involved in language. Even a single 'other' word presupposes
language, no matter whether this word, like 'not', is classed as
belonging to a secondary logical language, or whether, like 'dog',
it is considered to belong to the primary one-the 'object-language'.
Similar considerations apply to the presumed 'set of similar
expressions'. There can be no such set, in isolation. Modern
phonology has made it abundantly clear that we cannot even make
up such a set without regard to other and contrasting sets. Unless
there are other words, no word could have so much as determinate
phonetic shape. We cannot say, for any particular language, what
counts as repetition, as occurrence of a 'similar' expression, unless
we know what counts as occurrence of a different one. The recurrent
shape of 'cat' is determined by contrasts such as 'cat/pat/mat;
cat/cot; cat/cap/can'. For a Japanese, wrong is a repetition of long,
right of light and grammar of glamour. This is simply because, in his
language, though he does make use of both 1 and r, he has no need
of their difference for distinguishing meanings. To him, they are in
fact indistinguishably 'similar'. The l/r-difference is no more a bare
fact, it is as much part of the English language as is the difference
between what light and right 'mean'.
Nor could there be neutral facts in that hypothetical language-
of-reference. Ostension or reference, when it is supplemented by
rules of assembly and classification-no less than when it is supple-
mented by different languages-is free to generate a variety of
different meanings. If we are supposed to refer to what is similar in
a number of facts, the question must immediately be: 'Similar in
what respect?' 'With what degree of similarity?' And the answer
to these decisive questions can certainly not be found simply by
referring to the facts. Is not what we call 'high' in English similar,
in some respect, to what we call 'deep'? and therefore deserving of
being compassed by just one word, as in Latin, where 'altus'
corresponds to both? Is what we call 'blue' similar enough to what we
'Professor Korner would require of ostensive rules that they contain a 'com-
paring clause': this and this and 'everything like it'. But until we have said: 'That
is not a cat' or 'That is a dog', everything is like a cat, in some sense. Professor Korner
says that we have understood an ostensive rule, 'when we are competent to give
further instances or to give "anti-examples" ' (Conceptual Thinking, pp. 7, 33). If
'or', here, were replaced by 'and', my point would be made. (Professor Korner tells
me that he would accept the conjunction.)

221

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

call 'green' to allow us to be satisfied with just one word, such as the
Celtic 'glas' which corresponds to both English 'blue' and English
'green'? Which things are similar, and which are similar enough,
will be decided by our interests, most of which are not imperative
even for us; still less for all climates and all communities. Where,
then, are the required neutral facts? If we cannot find them in our
talk about colours, what can we expect of plants, animals, work
human relations? 'What there is' is different facts, picked out and
ordered by different languages-even by different languages-of-
reference. If it were possible to establish a language by reference plus
induction, the same 'rules of reference' would result in a variety of
languages.
It might be suggested of course that, in addition to Mill's Canons
of Induction, we should supplement ostension by the whole body of
scientific theory; and say, for example, that what colour-words in
any language 'refer to' is what the language of Optics refers to by
'light-spectrum'. But this unfortunately would merely be to claim a
privilege for the particular language of science, and for the 'facts'
within it (e.g. for a continuous spectrum, without divisions). It would
do nothing to establish extralingual, neutral facts.
Of course, there must be facts which permit the distinctions we
choose to make. And no one will deny that, human lives and interests
being what they are, some distinctions are all but imperative. We
know this by translation. It is also true that we shall find it easier to
translate from one language into another, if the two are restricted by
similar 'rules of reference' (e.g. if our translation is of scientific texts) .
But human lives and interests are still so varied, and linguistic
instruments so subtle, that, again and again, what appears as one
and the same fact in one language, corresponds to a number of
different facts in another. The range of permissible choices, which
we have no way of surveying, must be tremendously wide. By switching
to different languages, or to different times in the history of the
same language, we constantly alter the fact of 'what there is'. Pliant
facts far outnumber the stubborn. One has to ignore this great variety
and continual change in languages if one is to find plausibility in
the familiar assumption of a pervasive extra-lingual order of 'natural
kinds'-this assumption that everything is clearly set out before us,
ready to be mapped in more or less uniform fashion, by every
language.' A doctrine of 'natural kinds' is the last refuge of a
denotational theory of meaning. Only within the assumption of such
a doctrine can the alleged inductive accumulation of isolated
references be supposed to do its work. What this amounts to, on
closer examination, is just a naive belief in the divinity of one's own
'See Russell: 'fortunately, many occurrences fit into natural kinds' (Inquiry, p. 76).

222

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

language; God or Nature is supposed to speak it. The more sophisti-

cated may reserve such divinity for scientific discourse. But whatever
the privileged language, it is facts circumscribed by it that are spoken
of as extralingual, as bare and neutral.
It will be acknowledged, then, that we can endow expressions
with meanings, and even construct languages, by submitting to
what we may continue to call 'rules of reference'. We can insist,
if we wish, that the significance of every expression be vouched
for by some regular concomitance with other things-things other
than expressions. But we cannot insist on simply 'finding' them (the
expressions or the other things) just 'there' ready to be matched.
What expressions there are, and what other things, is determined by
what we do with them in developing and speaking a language.
It is of course, ultimately, some relation of linguistic expressions
to other things that constitutes their meanings. The question is:
What sort of relation? My point is that it is not, and cannot be, a
relation between two distinct orders of thing. The alleged con-
frontation of language with facts, the alleged reference of expressions
to things un-involved in language-this we cannot make sense of.
If we divide language from other things in this dualist fashion, both
are dissolved in a general blur. It is only in their active interplay
with one another that either assumes determinate shape; and it is
this interplay-this active co-operation of utterances with things-that
constitutes the meanings of the utterances.
One way of using an expression is to use it as 'a name for a thing'.
But before an expression and a thing can be so used, both must have
found their places among others. This they cannot do by way of mere
'naming'. Only when it is clear that an expression can be used in
many and various ways, and a thing be spoken of in many and
various ways, are the two sufficiently established for the one to be
used as 'name' for the other.
The generation of meaning is not a naming-ritual. Primary
are the meanings of whole utterances-utterances as part of our
active lives. We retain a word as a token, a souvenir, a keepsake of
the utterances and situations in which it occured. The active and
organised memory of these constitutes the meaning of the word-
the meaning with which it enters new utterances and new situations,
adding these in turn to its potential for future use. Every situation,
old and new, is organised by the continuous commemorative power
of words. Words do not confront situations; they make them what
they are.
To be sure, not everything is permitted. But hardly anything
is predetermined. Even under restriction of the same rules of
reference, we are free to construct different languages, different

223

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

worlds. There are no rules for a unique matching of mere vocal

noises with linguistically neutral facts. Outside a linguistic system,
which assigns them their places and roles, there can be neither
'expressions referring' nor 'things referred to'. With an inductively
constructed language of reference, as with any other, for a word to
'have meaning' is for it to play a distinctive role, among other
words, and among persons and things; our world takes shape in the
evolution of our language.

III

It is when we think of translation that we are most liable to

become confused about meanings-and tempted to locate them in
extra-lingual entities. After all, when we judge different expressions
to be equivalent by translation, do we not, in fact, abstract something
from them? And this something which we call 'the same sense',
is it not some separate entity-idea or physical fact-which is some-
how related to the different linguistic expressions? The answer to
the first question is: Yes, something-call it with Dr Johnson 'the
same sense'-is abstracted. But the answer to the second question is:
No, the same sense is not some separate entity related to the
expressions. There are abstractions and abstractions.
I may abstract an apple from its branch. The apple is an object
distinct from its branch; I can observe the two in different places
and at different times. Again, I may abstract the shape of the apple
from its colour. I can feel its shape in the dark, or I can see its colour
with some of the shape obscured. Shape and colour may still be
regarded as different 'objects' (though of another sort than apples
and branches); I can observe them at different times, and by
different senses. Furthermore, I may abstract the shape of the apple
from its size, though I cannot observe these two in different places
or at different times, and I have no sense which could let in the one
without the other. There might be things I can do to take in size
without shape-say, measuring the circumference; but there is
nothing I can do to observe the shape of anything without its size.
When I say of two apples that they are the same round shape, only
the one big and the other small, I do not take myself to be dis-
tinguishing three distinct objects: one shape and two sizes. The
round shape may be abstracted from the different sizes, but not as a
third object. Geometers, when dealing with figures of different sizes,
do not define 'similarity' of shape as the recurrence of some object
distinct from size. Shape does not accompany size. They abstract
'shape' from size by establishing a correspondence between differently
sized figures-a correspondence of points, angles and lines. Similarly,
we may abstract his dance from a dancer, though we cannot observe

224

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

it in another place or at a different time, and though we have no

special sense to take in the dance without the dancer. The same
dance, performed by two different dancers, is not a dancer-less dance
recurring. The dance is not in the two dancers, but neither is it a
third thing related to them. The dance of the two is the same, if
there is a correspondence between their two performances. This,
also, is how its meaning may be abstracted from an expression. We
can examine the expression without attending to its meaning, as
we might examine a dancer without attending to a dance. (Phonetics,
like anatomy, is a respectable discipline. It deals with expressions
in the context, and the language, of physics or physiology.) But we
cannot observe the meaning without the expression: it is never in
another place or at another time, and we have no special sense which
would let it in by itself. (This is why present-day Semantics is so
largely a dubious discipline.) The meaning of different expressions
is the same if, and only if, there is a correspondence between their
uses. What we abstract from different expressions as 'similarity of
sense' is a correspondence between their functions. Unless we succeed
in thus explaining translation, the mystery of bare and neutral fact
will continue to haunt us.
Why-one might ask-are we so strongly inclined to postulate
some separate and word-less meaning-entity, in order to account
for similarity between the performances of two utterances, while
yet we are not so tempted to postulate anything like a third object,
a dancer-less dance, in order to explain similarity between the
performances of two dancers? Does not this point to an important
difference between the two cases? It does. The difference is that it is
more difficult to establish correspondence of performances in the
case of two utterances than it is in the case of two dancers. We tend
to evade the greater difficulty by taking refuge in a myth. Yet, we
don't do so always. It depends on the degree of difficulty. This seems
worth examining.
When a speaker of what is described as Standard English and a
speaker of Cockney English converse with one another, they perform
some kind of translation. They establish certain correspondences:
e.g. 'a good bay' (for bathing) corresponds in Cockney to something
that sounds much like Standard English 'a good buy'; Standard 'a
good buy', on the other hand, corresponds very nearly to Cockney
'a good boy'; and Standard 'a good boy' to something with a vowel
(a closer one) which is unfamiliar to Standard English-'a good
body'. Translation here is easy: the correspondences concern
generally a few recurrent sounds. Though the expressions are
different in the two systems, we soon discover that on the whole
there is a one-to-one correspondence between their constituent

c 225

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

elements: /ei-ai, ai-oi, oi-o-i/. Translation then appears to be

sufficiently explained by this similarity of structure, and we have
little inclination to postulate a half-way house of pure meanings.
The situation is similar when we translate from speech into writing
or print, especially when the writing or print is phonetically regular.
Marks on paper share nothing with sounds in the air; but the
expressions in the two media are of similar structure. We say
that they have the same meaning. With modern English spelling, and
even more so with, say, Chinese logograms, the difficulties of such
translation between writing and speech are greater. But there is no
difference of principle. The complication arises simply from the
fact that the items which correspond to one another in the two
systems are so numerous. Assume that they are words, and there are
thousands of them. Nevertheless, since the number of words, though
large, is limited, we should still find it natural here to explain
translation by similarity of structure, i.e. by a broad one-to-one
correspondence between words in the written utterances with words
in the spoken. Also, if we discovered two communities actually
speaking languages which were related in this way, we should say
that they spoke closely related languages, though we should not go
so far as to describe these as dialects of the same language.' Difference
of language is a matter of degree; and the degree of difficulty we
find in translation (and in explaining translation) is a measure of
the difference.
Two linguistic systems are said to be different languages, i.e.
not just different dialects of the same language, if the sounds they
employ, though possibly identical, do not on the whole, utterance
for utterance, occur in a relation of one-to-one correspondence.
Commonly, difference of structure will extend further. It is a common
experience of translators that they cannot even rely on being able to
match words with words. Generally, the only kind of unit which
on the whole permits interlingual matching is the whole sentence.
But sentences are unlimited in number. There are no finite classes
of them, to be mapped on one another, in the way in which two
alphabets or two dictionaries might be. It is here that we tend to
despair of the task of explaining the actual operation of translation,
and are inclined to fall back upon the intervention of mythological
entities and processes to help us out.
'In fact, the Chinese logographic script is even less closely related to the spoken
languages of those who use it. Such spoken utterances as might accompany the
reading of the written words are rarely intelligible to anybody by ear alone. The
script is a visual language on its own; to have learned to write and read it is to have
made oneself bilingual. It is not surprising then that speakers of different languages
can understand one another by means of this script; they have learned the same
third language.

226

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 TRANSLATION

But what of the translator himself? He is well aware of having

no list of correspondences to refer to; but his task remains to establish
correspondences between expressions of the different languages. He
can do nothing else. He operates with expressions, not with wordless
ideas.
He will try to reduce his difficulties by limiting his range of choice.
He will, first of all, determine the required 'style of speech', i.e.
confine his range to a type of context: scientific description, report-
age, love-story, advertisement, religious tract, poetry, conversation,
etc. Within such a style, he may admit units smaller than the sentence
for his one-to-one mapping operation-for instance, technical
terms; or he may have to choose units larger than the sentence-for
instance, the stanzas of a poem, or even a whole poem.' But whatever
he does, he will work on the assumption that there is some type of
unit which permits a fair measure of one-to-one correspondence
between utterances in the two languages. In other words, he will
assume that, if chosen correctly, such units will show comparable
possibilities of combination (comparable 'mutual expectancies', as
the ancient Indian grammarians would have said) as well as com-
parable contrasts in the two languages.2 Essential about the chosen
units will be only these powers of combination and contrast; what
happens within any such unit will be less important (much as in
translating from speech into alphabetic writing, the shape of a letter
is less important than its 'distribution' and its contrasts).
The discipline of translation consists very largely in choosing the
smallest possible unit that will admit of adequate matching. But
it may well be impossible even to find a normal sentence in one of
the two languages to match a normal sentence in the other. In that
case, a difficult and unusual sentence will have to do the new job.
It must be difficult and unusual, or else it would do what is in that
language an old and normal job, instead of the new. In the Kikuyu
Bible, for instance, 'the Holy Ghost' is rendered by words which,
if they were matched with English words, would correspond to
something like 'white liver'.3 But they are not so matched. There is
no bilingual dictionary of metaphors. If the powers of combination
and contrast of the Kikuyu metaphor, in its Kikuyu context, are
parallel to those of the English expression 'the Holy Ghost', in its
place amongst other English expressions, then the internal difficulty
of the corresponding Kikuyu phrase stands to be resolved in the
required way by those who hear it. The language will have been

'Cf. L. Forster, Translation, pp. 11 ff. in 'Aspects of Translation' (Studies in

Communication, 2).
2Cf. p. 209, above.
3C. Rabin, op. cit., 136.

227

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms
 PHILOSOPHY

made to provide the required correspondences, as Old English once

was, when missionaries introduced the strange expression 'haleg gast'
into their sentences. They, and the translators of the Kikuyu Bible,
might have done worse. Instead of constructing a metaphor or a
'loan translation', they might have left a virtual blank in their
difficult sentence; i.e. they might have put a new 'borrowed' word,
relying on the rest of their text to determine its role.'
We should say, then-the translator chooses what units to translate,
and he chooses such units as correspond or can be made to correspond
to one another. He tries to keep the size of his translation units to a
minimum. But he cannot, generally, avoid having to deal with units
larger than the word. It follows that he will operate with open
classes and have no ready map to follow. But he has a compensating
advantage which alone makes his task feasible: the classes of match-
ing units being open, he is able to create expressions for his one-to-one
mapping. This is how languages are fashioned and re-fashioned by
translation. The translator, dealing with 'free constructions', con-
structs freely. He is not changing vehicles or clothing. He is not
transferring wine from one bottle to another. Language is no
receptacle, and there is nothing to transfer. To produce a likeness is
to follow a model's lines. The language he works in is the translator's
clay.

University of Manchester.

1A functional or instrumental theory of meaning, when fully worked out,

should be able to explain in detail how this happens: i.e. how expressions acquire
their meanings from their contexts, and how their meanings continually and
continuously change. This is a task for linguistic studies. The present discussion
can do no more than try to discern the general direction which such studies
would take.

228

This content downloaded from 5.23.123.100 on Tue, 11 Feb 2020 09:42:30 UTC
All use subject to https://about.jstor.org/terms