Ethnolinguistic Notes, Series 3, Number 6

Links to pages: 102, 103, 105, 106, 108, 110, 113, 114, 116, 117, 119, 120, 121, 127, 130, 131, 132, 133, 135

Comments Welcome
George W. Grace
University of Hawaii
July 30, 1982

ETHNOLINGUISTIC NOTES

Series 3, Number 6

THOUGHTS ON TRANSLATION AND MEANING:
A PROGRESS REPORT

"Equivalence in difference is the cardinal problem of language and the pivotal concern of linguistics." Roman Jakobson (1959: 233).

1. INTRODUCTION............................................................................…………...…2
---1.1. The Concept of Content Form...............................................…….…...…........4
---1.2. The Truth-Conditional Conception of Meaning........................…….........…......7
2. SEPARATING MEANING FROM FORM..................................……….....…......9
---2.1. What is involved in saying something: Sentence-level Signs.....……......….......... 15
---2.2. Word-level Signs.....................................................................…………..…... 19
3. WHAT IS EQUIVALENT IN TRANSLATION?..............................……….......... 21
---3.1. The Meanings of Linguistic Signs ...........................................…….……............24
---3.2. Understanding.......................................................................………………..... 27
---3.3. Text-Meaning versus Speaker's Meaning in Translation......................………..... 31
4. IDIOMATOLOGY: A FURTHER LEVEL OF CONVENTIONALIZATION ........ 32
5. IMPLICATIONS............................................................................………………..36
NOTE...........................................................................................…………………....38
REFERENCES.............................................................................………………........39

1. INTRODUCTION

The subject of this report is what I have thought of as the most intriguing of all examples in language of equivalence in difference: namely, the equivalence that obtains in translations between different languages. That is, if something is said in language A, and you ask me to give you a translation in language B, you are asking me to provide you with something which is necessarily different--since it is in a different language--but which is also equivalent in some significant sense. The question which concerns me is: What requirements must the B expression meet in order to qualify as a satisfactory translation in the everyday world; What kind of equivalence is tacitly assumed in requests of this sort?

The inquiry which I want to report on is something that I got into somewhat casually and without adequate preparation. The inherent importance of the problems involved and their seemingly endless complexity have enticed me along from one step to the next into deeper and deeper involvement. Twice I have thought I had come to a reasonable quitting place, and each time I wrote a book (one of which has been published) describing the point I had reached. In each instance I subsequently found that I could not quit at that point because new aspects of questions kept opening up. The report given here reviews some of the main points of the two books (Grace 1981a, 1981b) and subsequent "Ethnolinguistic Notes", but it also introduces significant changes.

What I am describing here as a separate inquiry actually developed out of a quite different kind of inquiry. The beginning came when my main research task was (as I suppose it still is) to attempt to determine the historical relations among the Austronesian languages. It had quickly become apparent that the Melanesian languages presented a particularly difficult problem. Three different schools of thought, each starting from a quite different hypothesis about the prehistoric linguistic relations of the area, could be distinguished (cf. Grace 1968: 66-70). One hypothesis which I found particularly difficult to deal with was that the Melanesian languages all resulted from the pidginization of Indonesian languages (and, in some cases, from subsequent pidginization of the pidgins). It assumed the convergent developments resulting from language contact to be so great as to obscure the actual genetic relationships of the languages involved.

The reason that it was difficult to deal with was that, as I soon discovered, linguistics had very little idea of convergence phenomena--no theory of the processes involved in the development of linguistic areas or of pidgins--in spite of numerous assertions by prominent linguists that areal phenomena were of great linguistic interest and in spite of excellent work by a few scholars such as Uriel Weinreich and Einar Haugen. For that matter it also had very little idea of any kind of linguistic changes other than those which could be attributed to the structure of the language in the period just preceding the particular changes.

From about 1975 on I have been devoting most of my attention to trying to figure out what would be required in an adequate kit of explanatory devices--that is, what a theory of linguistic diachrony adequate to elucidate the Melanesian situation would have to contain. However, what seemed most important for the Melanesian situation was to understand more about how linguistic convergence comes about and what evidence it leaves.

I was struck by a recurring feature in the literature and informal discussions of linguistic convergence. This feature was the emphasis laid upon intertranslatability of the two or more languages which had converged. Again and again what most struck observers seemed to be the possibility of something approaching morpheme for morpheme translation (or at least a closer approximation of such translation than was presumably possible before the convergence). This was particularly emphasized in Gumperz and Wilson's (1971) account of Kupwar village in India; I have also heard such intertranslatability between Tok Pisin (Melanesian Pidgin English) and local Melanesian languages offered as evidence that Tok Pisin had Melanesian "grammar".

It seemed then that the process of convergence might best be understood by considering the actual linguistic expressions generated in the languages rather than the more abstract grammatical systems which presumably underlay them. In fact, what should be considered were linguistic expressions (e.g., sentences) in each language in conjunction with their translation equivalents in the other language (or languages). The principal object of attention should be how the two differed from simple morpheme for morpheme matches, that is in Louis Hjelmslev's terms, how the form given the content in the two--the content form-- differed.

1.1. The Concept of Content Form

Hjelmslev illustrated content form with the English expression "I do not know" along with expressions having the same meaning (the same "purport") in Danish, French, Finnish, and Eskimo. I will just quote his account of the content form of the Danish equivalent, jeg véd det ikke and of the English. He wrote (1961 : 51),

"In Danish, first jeg ('I'), then véd ('know'--present indicative), then an object, det ('it'), then the negative, ikke ('not');

"In English, first I, then a verbal concept that is not distinctly represented in the Danish sentence, then the negation, and only then the concept 'know' (but nowhere the concept corresponding to the Danish present indicative véd, and no object)"

Thus, the content form of a particular utterance is the way its content (I will use "content" and "meaning" as synomyms throughout) is construed for expression--the analysis represented by the wording (and, in fact, by the very morphemes) used. From everything I have seen so far it seems that a language may profitably be considered to consist of two components which behave quite distinctly in situations of language contact. I call the two "lexification" and "content form" (cf. Grace 1981a, esp. 23ff.). My term "lexification" was inspired by the concept of "relexification". See, for example, the statement by Bickerton and Givón (1976: 12), "It would appear that, in the classic contact situation, the average speaker begins by gradually relexifying his original grammar--slotting newly-acquired vocabulary into surface structures characteristic of his own language." Thus, one might describe the form of Language A spoken by some persons as really a "relexified B", or, in my terminology, one might say that they are speaking A (i.e., using the lexification of A) with an approximation of B content form. I think it probable that bilinguals will almost always experience some tendency to speak each of their languages in a manner which at times suggests a relexification of the other, and that that tendency is at the bottom of both pidginization and linguistic convergence.

Anyway, I hypothesize that much of the convergence which takes place in instances of intensive contact between languages is a matching (making mutually compatible) of content form. (Although in the strict sense, content form is the property of specific utterances, I presumed that it is also reasonable to speak of a language as a whole as having a content form. The content form of a language is to be conceived of as a sum of the content forms of all possible utterances in the language.) The next stage in my undertaking therefore became that of attempting to understand the nature of content form.

I should point out that the significance of content form extends far beyond linguistic diachrony; it is, in fact, fundamental to the role of language in representing reality for us. Therefore, although there has not been much work done on the problem of content form as I conceived that problem, there is a very great quantity of relevant literature. In fact, one body of literature leads to another in a sequence which so far shows no sign of ending.

However, the understanding that I needed to obtain was one that would permit integration in terms of my main questions rather than questions deriving from other interests. One particular constraint on my own approach was that its central concern had to be the differences in content form between translation equivalents in different languages.

Moreover, although there was and is an enormous literature on topics which are in one way or another relevant to the question of content form, there had been little previous study of content form as such. Linguistics had dealt with pertinent aspects of language-- notably syntax and lexicon--but from a quite different point of view. And there had been virtually no attention given to what I (following Murat Roberts 1944) have called "idiomatology"; and idiomatology seemed to play a very prominent role in content form (cf. Grace 1981a: 43ff.).

1.2. The Truth-Conditional Conception of Meaning

To add to the difficulties, that aspect of linguistics which should have been most relevant--semantics--seemed to be dominated by an approach that made it quite irrelevant. I am talking about the approach which has been called "truth-conditional semantics". Its assumptions seem quite irrelevant for the kind of ordinary language functioning which my Melanesian problem presumably involved.

In truth-conditional semantics, the meaning of a text--say, of a sentence-- is assumed to be identical with the class of conditions under which the sentence would be true. That is, the sentence is conceived of as defining a class of real world conditions such that if any condition which is a member of the class actually obtains then the sentence is true. Likewise, the meaning of a word is also taken typically to be a class (its "extension"), or more precisely the criterial attributes by which members of the class can be identified (often called its "intension"). I do not believe that either of these notions has much relation to the way ordinary people use and understand language. In fact, it seems to me that truth-conditional semantics should be regarded as belonging to a kind of applied linguistics, relevant to information processing language, but with little relevance to an understanding of ordinary human language.

The fact that the study of meaning had become so entangled in these assumptions has probably contributed in no small way to the rise of a different approach to language functioning which does not accord any special status to meaning, but makes the functions of language one of the central questions to be investigated. This approach is unquestionably a valuable antidote to the usual assumptions about the nature of language and language functioning. However, it could not have been a satisfactory approach for my purposes because my approach starts from the assumption that functionally equivalent utterances in different languages can be identified by their speakers. That is, it assumes that if a person who knows languages A and B is told something in A, he/she can normally find a way to say something equivalent, to translate with something equivalent, in B.

I have also assumed that any bilingual can be expected to be able to judge roughly what qualifies as equivalent. Finally, I have assumed that what is equivalent in such translation equivalents is meaning--that is, that they have some kind of essential meaning in common. It seemed that the only general basis for equivalence, especially when the cultures do not approach isomorphism, must be equivalence of meaning. Furthermore, it has seemed a respectable assumption (because it has been pronounced by respectable people) that any meaning can be expressed in any language.

2. SEPARATING MEANING FROM FORM

It will probably be helpful if I say a little more about my initial assumptions, not in order to justify them since it does not matter much now whether they are justified or not, but simply to provide a better context for understanding the further course of my inquiry. I did, I now realize, assume that, although the term "translation" is applied to many different kinds of operations, there is a sort of core operation which underlies them all and from which they are all developed in more or less complicated ways. This core operation, which I conceived of as a sort of pure form of translation, was best exemplified by the task of putting simple, short, matter-of-fact utterances into another language. It is what I would be asking for if I said to someone of such an utterance in another language, "Will you say that in English?". (I discussed this question at some length in Grace 1981a: 35ff., cf. also 1981b: 171ff.).

I make much use of the concept "translation equivalent" here. By that I mean roughly whatever serves as a translation in the "pure translation" sense just indicated. I have assumed that people who speak both languages can ordinarily come up with some kind of equivalents (especially in the simple kinds of cases which I considered to be basic). I also assume that such speakers can to some degree reach judgments as to the approximate relative aptness of several proposed equivalents. And I probably assumed further that if some disagreement remained after the judgments were submitted, much of that disagreement would be due to different interpretations of the intention of the original speaker. But since my objective was to throw light on how and why different forms are given in different languages to the same content, the problem of how we arrive at a correct interpretation of what the original speaker meant in the first place was an irrelevant complication (for further discussion, cf. Grace 1981b: 180f.). One might remove this complication by imagining that the person who was to make the translation was the person who had made the source language utterance in the first place. (Note the lack of a clear distinction between the concepts of text-meaning and speaker's meaning at this stage of my inquiry).

Finally, I have assumed throughout that the unit of translation is small--a sentence-sized unit or therebouts. This assumption is in keeping with my notion of the core operation involved. I am still much inclined to believe that larger units mainly add complications--both in making additional strategies available and in involving additional kinds of conventions which impose additional constraints--but in any case I have had enough to deal with as it was.

Given the preceding clarifications, let me now restate three basic assumptions which I was making: (1) that anything can be said in any language, (2) that that assumption entails that for any utterance in any language there can be found in any other language some utterance which somehow qualifies as a translation equivalent, and (3) that what translation equivalence means is equivalence of meaning. The first task to undertake seemed to be that of determining how translation equivalents may be recognized. How does one determine whether or not a particular expression in A adequately translates a particular expression in B; or, given several A expressions, how does one determine which is the closest equivalent of the B expression?

As the term "content form" implies, I was, of course, assuming the separability of meaning (=content) from form. Presumably the form was language specific; it would undergo transformation from one language to another but the content itself could remain constant. Still, from all I had been able to observe it generally seemed to be true that, other things being equal, that expression in one language which most nearly preserved the content form of an expression in another would be regarded as the best translation equivalent. It seemed, also, that where expressions closely approximating the original content form were possible in the target language, it was fairly easy to decide on the most accurate translation. And, finally, where such approximation was possible, it seemed intuitively correct to say that the essential meaning of the original had been preserved; i.e., that although the content form had not been exactly preserved, it was reasonable to say that there was a sense in which the same content was being expressed.

However, in some cases it is not possible to make any very close approximations of the original content form. And in just such cases it seemed less clear what it would mean to say that the original content had been preserved. In other words, the more it became necessary to separate the content from the specific form which it received in a particular linguistic expression, the more difficult it was to see how that content might be identified. As this fact became apparent, I endeavored to find a way to work my way step by step from the solid ground of the translation equivalents which preserved a close approximation of the original content form to those which preserved somewhat less exact approximations. I hoped thereby ultimately to reach those which represented nearly complete transformations. I hoped to be able to specify in what ways the form might be altered while leaving the content recognizably the same. Or differently put, I wanted to know under what conditions expressions in different languages which differ in content form can be perceived as translation equivalents (which I took to mean essentially the same in meaning). I have identified three ways in which the form given the content in the source language can be changed--three kinds of transformations of the content form--which I think can fairly be said to leave the essential content still recognizable. The three kinds of transformations are the following:

(1) There are what we may call grammatical adjustments. We readily accept certain kinds of grammatical constraints on content form as not seriously affecting the meaning. I suggested (cf. Grace 1981a: 44f.; 1981b: 183) that the main grammatical differences may be characterized as differences in the information which is compulsorily specified and differences in the order of the compulsory elements. I assume that where only differences of these kinds exist between translation equivalents, the latter can easily be admitted to be alike in their essential meaning--that is, as expressing the same "content". But we would still be talking about a similar content form, one which conformed to the same model as the original. I am thinking of the situation where each of the major elements of a sentence is translated by an element of like role in the target language--an agent is translated as an comparable agent, a patient as a comparable patient, an action as a comparable action, etc.

(2) An additional kind of transformation of content form which permits one to perceive the translation equivalents as having the same content is what Vinay and Darbelnet (1958) call transposition. In transposition, the model of the original language expression is still recognizable, but some elements of the meaning are carried by grammatically different elements in the target language. One of their examples is English "He limped across the street" vs. French Il a traversé la rue en boitant (Vinay and Darbelnet 1958: 6), where the semantic contribution of "across" corresponds approximately with that of the verb in the French, and en boitant makes a semantic contribution corresponding roughly with that of the verb of the English.

(3) A further kind of transformation is what Vinay and Darbelnet call modulation figée (fixed modulation). Modulation (which is of two kinds: fixed and free) is defined (1958: 11) as "variation obtenue en changeant de point de vue, d'éclairage et très souvent de catégorie de pensée". Fixed modulation is found in translations between analyzable (motivated) lexical items, either analyzable single words or idiomatic expressions. Two of Vinay and Darbelnet's favorite examples of fixed modulation are "fire-boat" and "fireman" vs. bateau-pompe and pompier, respectively (e.g., 1958: 88). In my terminology, (see below) such translation equivalents would be said to have like senses, but unlike characterizations.

[Free modulation, "celle que les dictionnaires n'enregistre nt pas encore, mais à laquelle les traducteurs ont recours lorsque la langue d'arrivée rejette la traduction littérale" (1958: 11), must be treated separately. The kind of conventionalization involved in fixed modulation is relatively well accounted for in our current theories (see discussion below), but that involved in free modulation is much more interesting and problematic. It concerns what I have called (1981a: esp. 43ff.) "idiomatology".]

It is possible in the above cases (not including free modulation) to think of a content--a meaning constant--which has been transformed in ways which can be exactly specified. That is, the transformations leading from one content form to the other can be traced. But in some cases of translation equivalents there does not seem to be any way to relate one content form to the other. What are we to say about such cases, and particularly, what kind of interpretation can be given to the concept "content" in view of such cases?

It seemed that if I was to make any progress toward finding an answer to these questions, I would have to try a different tack. It seemed necessary to consider the nature of the linguistic sign, and in particular, what it is that we do when we say something. My attempt to answer this further question appears in Grace 1981b: 99ff. and 1981c: esp., 13ff. Briefly, my account of saying something is as follows.

2.1. What is involved in saying something: Sentence-level Signs

To say something is to perform a speech act. The linguistic vehicle of that speech act is the means by which it is said. The sentence is the most important and the minimal speech-act-oriented linguistic sign. That is, the sentence may be conceived of as a kind of linguistic sign the main function of which is to serve as the vehicle of a speech act.

A linguistic sign is conventionally assumed to be made up of two parts: a signans (signifier) and a signatum (signified). The sentence itself (by our ambiguous terminology which systematically uses the same term for the sign vehicle and for the sign as a whole, cf. Reddy 1979)--i.e., the sentence as sign vehicle--is the signans; what is the signatum? An account of the signatum of a sentence, then, should be approximately an account of what it is to say something. More precisely, it should be an account of what it is that is being said--of what the "something" that is being said is.

In my analysis, the signatum of a sentence consists of at least three components. First, there is the designation of what I sometimes refer to as an "emic situation". This takes the form of a model--modeling an aspect of reality or what we might call "pseudo-reality". By "pseudo-reality" I mean imagined situations which do not purport to be real but may be hypothetical or may involve fairy- tale-like objects or events; in fact, anything similar enough to what we imagine real reality to be like to be talked about. Such a model is made up of one or more elements ("emic elements") in appropriate roles. To take just one example, the model might specify the performance of a certain kind of act by a certain kind of person upon a certain kind of object with a certain kind of instrument, etc. (for example, a boy hitting a ball with a bat). The model, I believe, constitutes the most complex and generally important component of the meaning of the sentence. However, the model alone does not actually say anything at all; it simply characterizes an abstract emic situation.

The actual saying, in the strictest sense, consists of what I have called "the specification of the conditions of instantiation". This seems more or less to correspond to what is sometimes called a "modality" in linguistics. What I have in mind is that simply putting forward the model does not tell us anything about the existence or non-existence of actual instances which the model would fit. Thus the model consisting of the boy hitting the ball with the bat could be complemented, for example, by an assertion that an actual instance of such an event (or "emic situation") existed: for example, "A boy hit a ball with a bat". However, although assertion is so commonly given as the example of a specification of conditions of instantiation that we may be led to imagine that it is the only one, there is actually a great variety of possible distinctions. Consider, for example, the following sentences made with the same model:

A boy didn't hit a ball with a bat.
Did a boy hit a ball with a bat?
A boy will hit a ball with a bat.
A boy might hit a ball with a bat.
A boy would have hit a ball with a bat if...

It is my contention that this part of the meaning of the sentence is quite distinct from the specification of the model.

However, a well-formed sentence must also include another meaning component, one which we might describe as "indexical". It is the ordinary language counterpart for what truth-conditional semantic theories usually call "reference". It is this component which indicates the connection between the instantiation (whether actual or hypothetical) of the model and what I have called the ken of the audience. The truth-conditional view of meaning would interpret this as a central part of the meaning of the sentence (some people even speak of "referential meaning" as if that were meaning par excellence). I believe that the indexical meaning that is involved here should be thought of as quite distinct from the symbolic meaning of the sentence.

I will say more about this below, but briefly, I believe that the "symbolic meaning" of a sentence (or any linguistic expression) must be thought of as an abstract representation. In order for an act of communication to be completed the audience must provide this abstract representation with an interpretation. This interpretation is very analogous to the "interpretation" by which a mathematical calculus becomes a deductive system or eventually part of a scientific theory. A well-formed sentence provides indications-- hints, we might say--(for example, by means of devices such as definiteness, anaphora, voice) as to how it should be interpreted. The act of understanding, which involves devising this interpretation, is performed by the audience. It is essentially a putting into context of the abstract representation consisting of an emic situation and conditions of instantiation.

As an illustration, the sentences given above about the boy hitting the ball would actually make little sense to an audience unless that audience were given some further clue as to what I was talking about. In fact, if I were simply to walk up to the next person I see and say, "A boy hit a ball with a bat", I could reasonably expect him or her to be extremely perplexed. He/she would probably wonder what the reported situation had to do with anything he/she might possibly be able to think of, and also what I was trying to accomplish in telling him/her this. In such circumstances a great deal more connecting up-- putting into context--would be required in order for the speech act to make sense.

However, that sentence alone might suffice to elicit the desired understanding if it were pronounced in the course of (what the audience knew was) an explanation of how my window got broken yesterday. The implication of this point is that no sentence really says all that the speaker intends to be understood from it. In fact, no sentence says enough to be understood at all without a contribution from the audience.

To summarize, signs of the sentence level (speech-act vehicles) specify an emic situation (a model of a bit of reality or the like) and, further, say that such a situation obtains (or doesn't obtain, or might obtain) or ask whether or not it obtains, etc., with respect to particular persons, objects, etc., or to a particular time and place, etc.

I have so far discussed only sentence-level signs--speech act vehicles, signs that (can be used to) say something. However, the sentence is composed of elements, which are themselves linguistic signs. It is useful, at this point, to distinguish between ad hoc signs and conventional signs. An ad hoc sign is one which is put together ("generated") by the speaker for use on a particular occasion. A conventional sign is one which is known holistically-- i.e, is recognized as a unit--by (a sufficient number of) the speakers of the language. The vocabulary of a language is its inventory of conventional signs. On the other hand, most sentences are ad hoc signs (although some sentences have become conventionalized--for example, "How do you do?" or "A bird in the hand is worth two in the bush").

2.2. Word-level Signs

The crucial difference between sentence-level and word-level signs is that appropriate use of the former requires the responsible intervention of a sayer. The specification of the conditions of instantiation which constitutes part of the sentence defines the responsibility assumed by the sayer. Thus, sentence-level signs actually (may be directly used to) say something. The elements of which they are composed are what I call word-level signs. The elements of which a sentence is immediately composed (its "immediate constituents") may be either ad hoc or conventional signs. However, ad hoc signs, whether sentence-level or word-level, must be made up of other (word-level) signs, and the ultimate constituents of every sign are conventional signs.

Signs which are made up of other signs are called motivated signs. The basis for the motivation is a characterization. The meaning of an ad hoc sign (which, as we saw, must be a motivated sign) is nothing more nor less than the characterization which it provides. On the other hand, conventional signs have another kind of meaning--what I have called a conventional sense. In fact, the meaning of unmotivated conventional signs consists of nothing more than their conventional senses. However, any motivated sign necessarily has its characterization; therefore, any motivated conventional sign which has its own conventional sense must also have its characterization. It is these characterizations which, in translation, necessitate the "fixed modulations" spoken of by Vinay and Darbelnet (see above).

This two-part character of the signata of conventional signs (and, as I have tried to show in Grace 1981b: 57ff., almost all new conventional signs which are introduced into a language are motivated--hence have this two-part meaning) is possibly the single most significant fact for the understanding of the content form of a language. The vocabulary of any language, I propose, consists of a small core vocabulary of unmotivated conventional signs--most very ancient--plus a larger (in languages such as English, enormously larger) vocabulary of motivated conventional signs which characterize in terms of other conventional signs. These characterizations may be made immediately in terms of other motivated signs, but ultimately all characterizations trace back (perhaps through several layers of derivation) to the store of unmotivated signs.

That, briefly, is my current account of what is involved in saying something. Before returning to the question of translation, it might--in view of the current "metaphormania" (as Mark Johnson, 1981: ix, puts it)--be worth pointing out that the characterization which I have just been discussing is metaphoric in nature. Characterization is necessarily made in terms of something else; it is an indicating of what the thing in question is (purportedly) like. In fact, one kind of characterization which I have recognized (Grace 1981b: 62) is the simple metaphoric extension of a single morpheme, such as "head" of a bed.

3. WHAT IS EQUIVALENT IN TRANSLATION?

The account of content form given so far provides a conceptual basis for most of the kinds of transformations which the content form must undergo in translation, that is, it gives a basis for characterizing all of them except for free modulation. However, free modulation is a problem of particular interest which requires particular attention; I will return to it below.

Briefly the account which I have given includes the following means of expression. (1) A traditional vocabulary (mostly quite ancient) of unmotivated conventional signs. (2) Grammatical means for combining conventional signs to produce ad hoc signs. These means must provide especially for characterization, but also for specification of the conditions of instantiation and indicating connection with the ken of the audience. As is well known, the grammatical means are not identical from language to language, and although there are many functions which must be provided for by all grammars, there is also much which is language specific. (3) In addition to the unmotivated conventional signs, there is further vocabulary of motivated conventional signs (mostly reflecting the same grammatical relations as ad hoc signs)--signs which carry both a conventional sense and a characterization. The vocabulary, or more precisely the senses of the conventional signs, constitutes an inventory of "emic elements"--emic objects, states, processes, actions, acts, etc.

The content form of an utterance involves some kind of "saying", i.e., specification of conditions of instantiation of an "emic situation". The emic situation consists of a characterization in terms of word-level signs. Some of these word- level signs may themselves be ad hoc signs representing characterizations in terms of other word-level signs. All of the ad hoc signs eventually reduce to characterizations made from conventional signs. Some (probably most in most languages) conventional signs are themselves motivated and therefore comprise a characterization. However, all conventional signs are ultimately composed of unmotivated (i.e., unanalyzable) conventional signs. The content form of an utterance includes all of this.

The necessity for three of the kinds of transformations of content form recognized above can be explained in these terms. (1) Grammatical adjustments, of course, are explained simply in terms of grammatical requirements. (2) Transpositions are necessitated by the lack of comparable emic elements in the target language. (3) Finally, fixed modulation occurs because where the two languages do have equivalent emic elements, these may be provided by (may be the senses of) motivated signs with different characterizations.

The problem which we want to consider is the basis upon which translation equivalents between languages are identified. So far, we have been assuming that what is equivalent is their meaning. On that assumption, it appeared that if we could specify how translation equivalents are determined in all particular cases, we would also have a definition of meaning. As a first step I proposed that, starting from a fully-formed content in the source language, there were three kinds of transformation of the content form which still left the essential content recognizable in the target language. The three were, of course, those just referred to: grammatical adjustments, transpositions, and fixed modulations.

However, I have not succeeded in going beyond this first step; I have not been able to go further in identifying kinds of transformations of which it can convincingly be said that they do not significantly modify the meaning. In fact, all other transformations seem to fall under the general rubric of "free modulation". But what kinds of constraints govern free modulation? Indeed, how is it possible to determine what target language expression is derived by free modulation from what source language expression. How is it possible to determine what is a translation equivalent of what if the equivalents may be related through free modulation?

My inability to specify the nature of the constraints on free modulation has led me to suspect that there is something wrong with my conceptualization--that it is not always possible to find sentence-level signs in different languages which can be said in any useful sense to have the same meaning. But, if that is so, what is it that we do when we translate in such circumstances? What is it that is equivalent in translation equivalents? If it is not the meaning of the linguistic signs involved, how does it differ from such meaning? Or is there such a thing as the meaning of a linguistic sign at all?

3.1. The Meanings of Linguistic Signs

To begin with, I would say that such meanings do exist, that linguistic signs do have meanings. For example, it seems reasonable to say of unmotivated conventional signs (e.g., a word such as "nose") that they combine an identifiable signans with an identifiable signatum and that the signatum (what I call the "sense") may appropriately be called a "meaning". This meaning, by the way, is not a set of criteria for recognizing a nose, say, when one encounters one. I have tentatively proposed (Grace 1981b: 98) that it is more accurately conceived of as a remembered exemplification or a stereotype.

By the same token, the characterizations of ad hoc signs (including sentences), being nothing more than two or more such senses assembled into a structure, should also qualify as meanings. It seems a perfectly straightforward matter, then, to say that linguistic signs do have meanings.

But is it possible to say that different linguistic signs can have the same meaning? It seems that (if we disregard such contrivances as truth-conditional semantics) no two linguistic signs can possibly have exactly identical meanings. That would be true of the senses of conventional signs, and therefore, since characterizations are composed of senses, of the characterizations of ad hoc signs (or of motivated conventional signs, for that matter).

So, we conclude that no two linguistic signs, and therefore no translation equivalents, can have the same meaning in a strict sense. Is it possible, however, that there is some loose sense in which translation equivalents have the same meaning? My answer at present would have to be that the only possible such loose sense that I can think of encompasses the cases where it is possible to specify the transformations (grammatical, transposition, fixed modulation) by which one sentence (or sentence-level sign) is related to the other. In such cases, there seems a reasonable basis for claiming that the same meaning, in a loose sense, is involved.

But what if that is the most which can be claimed, that in certain limited types of cases, translation equivalents can manifest what can, in a loose sense, be called the same meaning? If that is the most that can be claimed, then it can hardly be put forward as a general principle that what is equivalent in translation equivalents is their meaning.

But on what other basis might translation equivalents be equivalent? A possible lead comes from the fact that an act of communication does not depend on the meaning of the linguistic expression alone. In discussing sentence-level signs I mentioned that in order for an act of communication to be completed, it is necessary for the audience to achieve an understanding of what the speaker intended to have understood--of what the speaker meant by what he/she said. This act of understanding involves an act of interpretation--the audience must put an interpretation upon the abstract representation provided by the linguistic sign. I also pointed out that aspects of the sign are designed to provide indications of the intended interpretation. We may say, then, that underlying any speech act there is some kind of understanding which the speaker has the intention of eliciting by means of the linguistic sign (along, of course, with gestures, anything else in the immediate context, anything he/she knows about the audience, etc.) which he/she employs.

This understanding which the speaker intends to elicit I will call the speaker's meaning. I will define it (with obvious indebtedness to H. P. Grice, e. g., 1968: 230) as the "understanding which the speaker intends to produce in the particular audience by means of the recognition by the audience of that intention".

We have seen that although sentence-level signs can be said to have meanings, except in certain strictly limited categories of cases these meanings cannot be said to be what is equivalent in translation equivalents. Clearly, we must look elsewhere for the basis of that equivalence. I want now to consider the possibility that what is equivalent is the speaker's meaning--the understanding which the speaker is attempting to elicit by means of such a sign.

3.2. Understanding

Having defined the speaker's meaning in terms of the understanding which he/she intends to elicit, we must begin by asking what we mean by "understanding"--what it is to understand something. Jay F. Rosenberg (1981: 33) says that to understand something is to "give it a 'reading' in terms of which it 'fits' into a larger, coherent, unitary picture of the world." He adds (ibid.) that "... it is the lack of a coherent context which, in general, will be what stands extrinsically in the way of understanding something, which makes it difficult to understand."

I said above that the symbolic meaning of a sentence must be thought of as an abstract representation, and that in order for an act of communication to be completed, the audience must provide this abstract representation with an interpretation. To provide the interpretation is to perform what I (1981b, e.g., 99ff.) called the "act of understanding". I take understanding in this sense to be an act of perception--perceiving what we may provisionally call the significance of what was said (and like all perception, of course, it is fallible).

To discuss perception--and therefore understanding in this sense--requires that we turn our attention temporarily to the part played by the hearer rather that that played by the sayer. I take perception to be a matter of information pickup (a term which I take from James J. Gibson [1979]), that is, of the organism (in this case the hearer) gleaning information from the environment. But the speech act is an intervention by the sayer which is ordinarily intended both to stimulate the hearer to undertake perception, i.e., to try to understand, and to stimulate and shape that perception. The design of the utterance--and here I am particularly concerned with its meaning--may be expected to reflect that purpose.

Gibson emphasizes the richness of the array of information available in the "ambient sea of energy" (1979: 57). With regard specifically to visual perception, he emphasizes the nested nature both in space (a nested hierarchy of solid angles which specify an invariant structure--e.g., 1979: 68) and in time (a nesting of episodes within episodes, with "what we take to be a unitary episode...a matter of choice", 1979: 101). The point which I want to make is that whatever the extent in time or space of the unit which the perceiver determines to focus upon, it exists, and is perceived to exist, within a context in time and space. And this context is part of what is perceived.

A speech act is an unusual challenge to perception in that it involves two kinds of contexts. The first is the context of the speech act itself. There is, necessarily, some perception of the performance itself--of what the speaker is doing in performing that act. Now one might want to contend that a speech act has not been completely understood until the speaker's motives have been recognized--until we have figured out what he/she is up to. However, that is clearly not the kind of understanding that is involved in translation; I believe no one would expect a translation to reveal the hidden motives of the speaker. On the other hand, we would expect the translation to attempt to reveal those intentions of the speaker which he/she intended to have recognized by the audience--i.e., what I am calling the speaker's meaning. Those intended-to-be-recognized intentions do not seem to belong exclusively to either the first or the second context.

The second context is that of what I have called the "emic situation" specified in the actual linguistic sign--i.e., the context of that which is being talked about. As would be expected from Gibson's discussion of perception, understanding of a linguistic sign which characterizes such an emic situation might require a context involving more inclusive situations in which it is nested or situations likely to have preceded it or to be consequent upon it.

In this connection, George Lakoff and Mark Johnson suggest that "understanding takes place in terms of entire domains of experience" (1980: 117). These domains of experience are known to us as experiential gestalts. That is, reference to them calls up a whole schema of things which, in our experience, tend to go together. An example of an experiential gestalt given by them (1980: 167) is that of shooting someone. Someone's shooting someone involves a whole array of characteristics which we tend to assume to be present unless there is an indication to the contrary (we might say that they are present in the "unmarked" case of shooting). These include a shooter, a target, a gun, a bullet, the shooter's purpose, the shooter's plan, and a sequence of acts: loading, aiming, firing, etc. All of these and more go together to form a single gestalt, and they are likely to occur together in the individual's perceptual schemata.

There is a great deal of literature dealing with the complex contexts which are invoked in perception, memory, and other psychological processes, although the discussions do not usually focus on the use of language. There is, for example, an extensive literature on concepts such as "schemata" (a concept due to Henry Head, 1920: 605, and Frederic Bartlett 1932: 201 and passim) and "cognitive maps" (due to Edward Tolman 1948). For a more recent account of both of these and other related concepts, see Downs and Stea 1973. I also rather like Julian Jaynes's (1976) term "narratization" for a kind of interpretive process by which we contrive contexts for experiences of all kinds--contexts which provide rationalizations for them and perhaps suggest what is to be expected next.

The information pickup required for understanding varies widely as circumstances differ. I presume that it always requires some putting into context. Saying something in the first place involves (what we will some day recognize as) an impressive feat of abstraction. Understanding what was said involves reversing the abstraction--restoring the context--in some degree. The context which must be restored or supplied may be quite narrow, or it may be very broad.(1)

What determines when enough context has been supplied and understanding has been successfully accomplished? It would seem that from the hearer's point of view it may be considered to have been accomplished when the hearer is satisfied that he/she has understood sufficiently well and turns his/her attention to the next order of business. From the beginning it is the speaker's task to anticipate the strategy which the hearer will employ in seeking to understand and to tailor the utterance so as to exploit that strategy. If the speaker has anticipated inadequately, what satisfies the hearer as understanding will probably not satisfy the speaker that his/her meaning has been understood. In any case, the translator's role is that of a surrogate speaker, and the translation should be designed according to considerations analogous to those which governed the design of the original utterance.

3.3. Text-Meaning versus Speaker's Meaning in Translation

The conclusions tentatively put forward here have included the following: (1) Linguistic signs do have meanings--what I will call "text-meaning". (2) No two signs can be said to have the same meaning in a strict sense. (3) However, where the form given a meaning (content) in two languages is sufficiently similar, it may be possible to describe the two as having loosely the same meaning. (4) When for a particular source language sign there is no target language sign available with even loosely the same meaning, the objective in translation must be to design a sign in the target language which is apt to elicit an understanding equivalent to that which the original source language sign was intended to elicit. To put it differently, the translation must endeavor faithfully to reflect what we are calling the speaker's meaning.

The above principles obviously are not sufficiently precise to serve as a practical guide for translators. One kind of problem which I have not even mentioned arises when a target language expression which seems to qualify as loosely the same in meaning as the original source language expression seems, nevertheless, not apt be understood in the same way. I have discussed a few cases which seem to be of this kind in Grace 1981a: 48-50. There is, in general, some competition between the two kinds of meaning, and this competition has figured prominently in the literature on translation.

4. IDIOMATOLOGY: A FURTHER LEVEL OF CONVENTIONALIZATION

One outstanding puzzle remains: there are cases where the grammar and vocabulary of a target language would seem to make it possible to approximate the content form of a source language expression, but where the resulting expression is not idiomatic ("we wouldn't say it that way") or where, although it is a possible expression in the language, it would be understood in a way quite different from that intended. I assembled some examples of what I took to be the latter problem in Grace 1981a: 46ff. I am talking here of constraints on ad hoc signs, not of fixed idioms, clichés, or catch words, all of which I regard as conventional signs (cf. Grace 1981b: 60ff.).

The reality of the phenomenon of idiomaticity is easy to observe. For example, I have recently made a few sporadic attempts to collect examples of wordings which appear to me to violate English idiomatology. I have taken most of them from writings of professional linguists who are not native speakers of English. My experience indicates that a sizeable number of examples could be collected without great difficulty. I will not present any of those which I have collected now for two reasons. First, there is the possibility that some of the authors might be identified by some of my readers. Second, I find that examples of this kind are difficult to present succinctly. They usually seem to require some discussion of the context and of what the author seems to have intended to convey. Often, also, I feel the need for a discussion of what kind of misinterpretation seems likely to be made and why.

In any case, the phenomenon of idiomaticity appears to establish the existence of conventions which are not grammatical conventions and yet which constrain the formation of sentences. We are in the habit of thinking of constraints on behavior as taking the form of rules. However, I have so far found it difficult to see just how we might formulate such rules. For one thing, it does not seem that the operation of rules of this sort is generally categorical in the same way that the operation of grammatical rules is. That is, the violation of such rules may result in an utterance perceived, not as strictly unacceptable, but rather as somewhat strange. This strangeness is likely to interpreted as a kind of markedness, and as a hearer, one is likely to feel that one has not understood the utterance completely until one has figured out the significance of the marking--i.e., figured out why it was said in that strange way.

It is noteworthy, by the way, in the examples of non-idiomatic writing by foreigners which I was describing above that an important factor in the correct interpretation of such sentences by sophisticated English speakers is precisely the awareness that they were written by non-native speakers. That is, we do not seek the explanation for the particular kind of "markedness" involved in the same directions as we would if we took the author to be a native speaker. However, some of the exceptionally monolingual speakers that can be found in great centers of monolingualism such as the United States seem to have difficulty suspending their normal strategies for interpreting such marking even though the speaker may be obviously a non-native speaker.

I puzzled for a long time over what kinds of constraints could make a sentence which had been produced out of unexceptionable lexical items and in strict accordance with the rules of the grammar uninterpretable, unacceptable for a certain meaning, or strange-sounding. If idiomatology is understood to include all such constraints, then I think it will probably turn out to involve several quite different kinds of conventions. For example, the constraints governing the order of such fixed binomial expressions as "black and white" (where the order "white and black" seems strange) are probably of a different kind than those governing most of the kinds of violations of idiomaticity committed by foreigners. At present I am not prepared to talk about more than one of these kinds. The kind of convention which I have in mind is based on meaning, or subject matter. Such conventions manifest themselves as constraints on how one can talk about particular things. They are conventions about how particular things are to be characterized even in ad hoc signs.

I believe that the basis for this kind of convention lies in the metaphoric nature of language: in the fact that our ways of talking about almost everything are metaphorically based and that there is in each language considerable consistency in the basic metaphors (cf. Lakoff and Johnson 1980; for the term "basic metaphor", cf. Grace 1981b: 152) used for particular subject matters.

The Lakoff and Johnson book, appropriately entitled Metaphors we live by, is a very important work, which should be read by anyone with an interest in this topic. It shows convincingly how thoroughly permeated our ways of talking are with metaphor, and how our metaphors hang together in self-consistent systems.

I have written some highly tentative speculations about ways of talking about things (Grace 1982b). It is noteworthy, for example, that competing ways of talking about essentially the same thing can exist in the same language--such as the ways of talking about their subject matter employed by opposed schools within the same scientific discipline--and that essentially the same way of talking about a particular subject matter can be common to two, or even to many, different languages. It is mainly in cases of the latter kind that we find the possibility of translations which preserve (loosely) equivalence of text-meaning.

This topic also is in need of much further investigation, but if I was right in my speculation (1982b) that it would be possible to analyze languages into collections of ways of talking, I suspect that there will turn out to be something of a hierarchy of ways of talking within each language--a hierarchy extending from those ways of talking designed for very specialized subject matter down to those of greater and greater generality. For example, some ways of talking in English have been developed very recently and to a large extent deliberately, and they obviously draw upon ways of talking that already existed. An example would be the way of talking of a particular school in, say, linguistics. Such a way of talking may seem to involve new vocabulary most prominently but it is likely also to have other quite recognizable stylistic features. The latter often have recognizable sources, moreover (for example, such a source might be the way of talking of mathematics or of anthropology), a fact which constitutes a kind of metaphor in its own right.

It seems possible that at the base of the hierarchy there will be found a "core" way of talking--a way of talking about the most culture-independent things. It would presumably include core or "non-cultural" vocabulary something like that proposed for lexicostatistics, and would provide for certain core grammatical relations. Of course, it is possible to talk about subject matters for which there is no consecrated way of talking in the languages available. It is always possible, supposedly, to talk about anything using the ways of talking established for other things (and probably especially the core), but it will be awkward-- sometimes exceedingly awkward. And it demands extraordinary ingenuity of the speaker.

Anyway, whatever the merit of some of these speculations, it does seem quite clear that conventionalization must be recognized as extending up to the level of ad hoc signs.

5. IMPLICATIONS

The tentative conclusions which have emerged from this inquiry so far appear to have far-reaching implications. I am not qualified to discuss them in detail, and anyway, to do so would expand this report intolerably. I will, therefore, confine myself to calling attention briefly to the nature of just some of them.

There is, first of all, an implication for epistemology and the problem of the acquisition and representation of knowledge of reality. The conclusions presented here seem to add further confirmation to the idea (which seems rapidly to be gaining ground, anyway) that our access to reality is much less immediate than the over-optimistic philosophies of modern times have allowed us to believe. Language is a key instrumentality in epistemology; the knowledge of reality which is possible to species without language is of necessity limited to whatever their systems of cognitive representations can accommodate. It is only with the emergence of language that these limitations can begin to be surpassed. However, once language as we know it is available, there seems to be no limit to the potential for superimposing new ways of talking about things (and thereby creating new "things") on the base of those already there.

I have speculated (cf. e.g., Grace 1981a: 95ff., 1981b: 126ff.) that it was this epistemological role of language which was its main selective advantage, the main causal factor in the evolution leading to language. However (and I have made a crude attempt to discuss this matter in Grace 1981d and 1982a), I think that our civilization has tried to make of language a rather different kind of epistemological tool from that which it was originally designed to be. The attempts to make languages function in this supposedly desirable way have led, I believe, to a strong prescriptive bias which has seriously distorted our conception of the nature of language. In sum, it is my conclusion that a language is a conventionalization of presuppositions about reality to a much greater extent than we have believed, and an open, ontologically uncommitted code to a much lesser extent.

The other implication that I want to mention, and this one is actually just an additional aspect of the first, is that the realities which different languages are capable of representing are different. That is, as I put it in Grace 1981b: 122ff., a somewhat different world may be said to correspond to each language. But that statement, too, reflects our customary bias. Our approach to language has focussed attention so exclusively upon the kind of patterning represented by the langue that we tend to think of language as nothing more than the langue writ large.

But the reality proffered by language does not derive immediately from one or another langue. It is rather a matter of the ways of talking available to the particular individual, particular community, or whatever else we choose to make the focus of our attention. It matters little which langues the individual ways of talking are attributed to.

These implications seem of sufficient significance to justify my putting my conclusions before you in their present unfinished form.