Minor Correction Made 1 July 2006
Links to pages: 650, 651, 652, 654, 655, 656, 659, 660, 661, 664, 665, 666, 667, 668, 669
Comments Welcome
George W. Grace
University of Hawaii
30 November 1992
In the general scientific and academic division of labor linguistics is (or at least claims the right to be) granted authority over language as a subject matter. There is authority over research. This entails a responsibility, I believe, to make a continuing effort to answer the ultimate questions about language--questions such as: What is language? How does it work? What does it mean to the species that has it? What does it enable them to do? How does it affect the conditions of their lives--their relation to the world they live in? What is there about language that has made the history of the human species so different from that of all other species?
But the authority extends beyond research to the academic curriculum--to how and by whom the subject matter is taught and how it is represented to the public. I understand it to entail primary responsibility to be a source of the best answers available at any particular time not only to those questions but also to any others that concern society.
Although it is our teachings that are my main concern here, I must admit that I'm not entirely satisfied with the strategy guiding our research. However, I recognize that sciences must work within what Thomas Kuhn called "paradigms", and that accordingly they are likely to make assumptions about their subject matter that may seem objectively implausible. Therefore, although mainstream linguistic theory does make assumptions about language that in my judgment are biased and wrong, it might be possible to argue that making these assumptions--even though they may be clearly wrong--is (for some reason) the most effective way to arrive (at some time in the distant future, of course) at the best answers to the most basic questions about language.
However, even if such an argument were valid, there would still remain the problem of what we do in the interval before this eventual understanding is achieved. How do we meet the responsibility of telling the best possible approximation to the truth in the meantime? There's nothing about the activities of normal science (as Kuhn calls it) that prepares scientists for such questions. What seems too often to happen in practice is that, when asked for the best current answers to basic questions of the kind I cited above, the scientists wind up by referring to the assumptions of the paradigm of their discipline. Therefore, even though my main concern here is with what we tell our students and the public in general about language, I am brought back to a consideration of the research paradigm.
Questions such as what language is and how it affects the human condition don't appear to have had much effect in shaping the paradigm of the discipline. The linguistic mainstream, it seems, has largely either ignored such questions or presupposed answers to them. In fact, it appears to me that it has, maybe no more by design than by inadvertence (or intellectual diffidence), largely contented itself with pursuing the interests of those who are concerned with language as an epistemological instrument--most particularly, as a tool of scientists.
Scientists--or at least philosophers of science--have for centuries been concerned with the formulation of unambiguous statements. To produce statements of the kind envisaged would require what has been called "autonomous text", that is, linguistic expressions whose meaning can be specified independently of (any possible) context of use. What was envisaged was a system that would govern the construction and interpretation of sentences, controlling what sentences were authorized within the system and specifying the meaning of each authorized sentence (i.e., permitting it to be worked out). Under such a governing system--that is, in such a language--a word or a sentence would have a meaning in an absolute sense.
[It may be seen that in the case of such a language, anyone who knew the language--that is, knew the governing system--would be able to determine the meaning of any linguistic expression authorized by that system. Conversely, the governing system might be thought of as consisting of just whatever a person would have to know in order to be able to construct, and understand the meanings of, the linguistic expressions permitted in the language. Thus, the governing system and what a hypothetical ideal speaker would know about the language define each other. Now if one imagines such a language (or something close enough to such a language) to exist and imagines such ideal speakers to be somehow approximated by its real speakers, it is just a small step further to the assumption that the desired governing system actually exists in the form of knowledge possessed by speakers. In making this assumption, one attributes an actual locus in the real world to the governing system, to wit, the nervous systems of the speakers.]
It was believed in the past that the governing system would be mainly concerned with word meanings. Or, to look at it from the perspective of knowledge required of speakers--the knowledge that would be required to determine the meanings in such autonomous text would consist primarily of knowledge of the meanings of the words used. To know the meaning of a word was, it was believed, to know how what it may refer to is related to the evidence of the senses (to observables). It was believed that a language could be so regularized that such securely grounded meanings could be assigned to all of the words of its vocabulary.
Later, with--among other things--the development of mathematical logic, the linguistic knowledge assumed to be necessary has been expanded to include knowledge of the logical relations among the (referents of the) words in a sentence--relations that are specified by the syntactic structure of that sentence.
Thus, we have arrived at the idea of individual systems with (1) their store of vocabulary items, each of which would "pick out" aspects of the real world, and (2) their logical syntax--that is, their means of specifying logical relations among such real world phenomena. Each such system would govern the construction and interpretation of sentences. Through the grammar and dictionary of which it would consist, the governing system would authorize the construction of certain sentences and warrant the meaning(s) of each sentence so authorized.
Some philosophers of science, then, pursued the ideal of making such governing systems available to provide science with a means of recording (indeed, of fixing for all eternity) its observations and conclusions. (Some have imagined beings from another world having access to one of the great libraries of earth, obtaining a grammar and dictionary of (say) English, and with them being able to recover all of the knowledge accumulated by earth's science.)
Such philosophers have long hoped that some of the clues as to how to construct such a system for producing autonomous text might be found in human language as it already existed. This hope was encouraged by the idea (which has been around for a long time, at least in Western culture) that human language includes systems--what we call (individual) "languages". (On the other hand, I'm not sure how long the idea has been around that human language is divided up exhaustively into individual languages: that is, that all speech acts--even those of the most unschooled--are assignable to something that deserves to be thought of as a language).
Anyway, the association of existing human languages with epistemology has been around a long time in Western culture, and it is not at all surprising that clues for the design of the kind of language that the philosophers wanted were sought in human language, or that human languages were often considered as providing the base on which the philosophers' language might be built. That is all right for philosophers, but as far as I can see it has nothing to do with the proper concerns of linguistics. What I am concerned about is that after human language came to be recognized as a legitimate object of study, and after an autonomous science emerged to study it, that science has permitted itself to be distracted by the ancient pursuit of the philosophers' language (even while realizing that it's a chimera). What has happened is that linguistics has come to look upon ordinary human language as just a somewhat imperfect version of the philosopher's language.
We have been left (or so it seems to me) with a tacit assumption that the prototypical function of human language is to encode propositions that have a determinate truth value. But we have also been left with a set of assumptions from philosophy about how the performance of this function is pursued: about what strategies are employed in using language and how language is designed so as to facilitate these strategies--in short, about how language works.
More specifically, the assumed strategy for performing (or making some approximation to performing) this supposed prototypical function is that imagined for the philosophers' language. First, human language is assumed to be made up entirely of individual governing systems of the kind we have been discussing, that is, to be divided up into individual systems, called languages. The individual language (like the philosophers' language) is supposed to constitute a system which governs what expressions can be produced and what meaning each of the expressions can have (i.e., what propositions each can express). To perform its role of governing system, each language (in this conception) provides a vocabulary of lexical items that prototypically refer to objects, acts, etc. in the real world, and syntactic rules that permit us to specify relations among lexical items that we've selected in such a way as to produce propositions that, in the prototypical case, are either true or false.
I've referred several times to assumptions about what language or some aspect of it is like, not invariably or absolutely, but "prototypically". I need to explain.
In the final analysis, what I really mean when I say that some function or the like is assumed to be prototypical is that it is treated as the key one--as one which can serve as a key to the others. It is assumed that understanding it first will open the door to understanding the others, whereas understanding one of the others first would offer no such advantage.
But there is more that is implied. Saying it is the key to the understanding of the others implies that it is logically prior to the others--that the others are derived from it and most likely presuppose it. Furthermore, saying that it is logically prior implies that it is chronologically prior, that the others have actually evolved from it.
Therefore, when I said that there is a tacit assumption that the prototypical function of language is to encode propositions that have a determinate truth value, I meant that linguistic theory treats this function as if it were the most representative one, the best example to start with. And I think it is clearly implied that it must be the original function of language--the main function which guided the evolution of language.
It should be pointed out that assuming this to be the prototypical function of language in no way denies that we can and do use the mechanism which (by this assumption) constitutes the governing system for other purposes: to ask questions, give orders, make statements whose truth value is not determinate. But these would be seen as derivative uses. The main function of the governing system (according to the view that I'm describing) is still to permit the formulation of propositions with determinate truth value.
It seems further to be assumed that it is because it has this governing system--this design which makes the language a suitable instrument for precisely encoding propositions--that communication by means of language is possible at all.
Of course, certain points would surely be conceded: first, that (as we saw) people use language for other purposes than expressing propositions. Second, speakers also--even when they are formulating propositions--often (not to say "constantly") use the language without following the rules exactly (i.e., they use it sloppily), and they can often still understand each other to a considerable extent. However, that doesn't destroy the assumption that the prototypical language-use is not sloppy. And (it can be argued) the explanation for whatever understanding does occur in sloppy (i.e., ordinary) language use lies in the fact that the utterances do conform partially to the rules of the governing system. In fact, the degree of understanding that is possible might be claimed roughly to approximate the degree to which the utterances do still conform to the rules.
I hope that most people will agree that the foregoing description fairly approximates the standard assumptions.
I think the governing-system view as I've described it is an important part of what linguistics teaches about the essential nature of language. It amounts to a picture of what language is--of the core principles on which language is supposed to be organized. It provides what purports to be an explanation of how language is able to work, but an explanation that only makes sense if we assume such a highly structured and standardized governing system as the reference point.
My problem is that I think that this picture is entirely wrong: that the nearest approximation to the philosophers' language that can be found in actual use today are the standard languages which have been developed in the last few centuries, and these are far from being an accurate reflection of what is prototypical (in any of the senses suggested above) for human language. I think, on the contrary, that they are a recent invention of intellectuals. But if that is true, we should expect the language faculty--the biological capacity that makes human language possible--to be adapted to a kind of language use that is somewhere near the polar opposite of the standard languages that have arisen in Western culture.
Of course, truly autonomous text is clearly impossible. (1) But consider further that one of the conditions required for it even to be conceived of is writing. It is all but impossible to imagine such a thing having been conceived of until after writing was invented. At least it is hard to imagine what use such texts could reasonably be imagined to have if they were not written.
But linguistics has a tradition of playing down the innovativeness of writing. We have argued against such expressions as "a written language", on the grounds that for any language to become a written language requires only a small, and natural, step. But, once again, I think that we (especially linguists) have projected our own perception of language onto everyone including speakers, past and present, of languages which have never been written. We assume all languages to be precisely designed for writing--in particular, alphabetic writing. We analyze them as having a sound system based on letter-sized segments, which only need to have orthographic symbols provided for them to permit them to be written. And the needed orthographic symbols, as it has turned out, were developed in other attempts at writing by people who somehow had not quite grasped how language was really intended to be written. Writing then, and above all alphabetic writing when people finally did get it, was just the realization of the manifest destiny of languages.
I find all of this very suspicious. I suspect that the design of language as it first evolved could in no way fairly be interpreted as anticipating the requirements of writing. I suspect that the gradual evolution of writing systems required a considerable reorientation on the part of the speakers.
Anyway, the point that I have been trying to make here is that we linguists work from, and lend our authority to, a quite fundamental misrepresentation of the basis of language--of how language is able to function--and therefore of the language faculty.
But there is another aspect to this point. This whole conception of how language works assumes it is organized into individual systems. However, I would argue that individual languages don't really exist, and furthermore that the nearest approximations to such systems that we find result from the intervention of extra-linguistic factors. And I believe that that fact has more serious implications than linguists are accustomed to assuming. The particular implication that I want to call attention to here has to do with the governing-system assumption. I find it hard to see the governing-system picture of how language works as being compatible with the actual state of uncertainty about what is included in any given language.
Maybe what I'm trying to say will be clearer if we imagine a case where a linguist has magically come to know exactly what linguistic expressions each person who could possibly count as a speaker of human language would ever use and when. In short, I want to suppose that this linguist has ready access to any and all linguistic data that could be thought of as relevant to exhaustive and correct linguistic description, but has not been given instructions as to how to partition the data into individual languages. And I want to suppose further that s/he is not permitted to ask the speakers to identify particular utterances or particular speakers as belonging to particular languages or to ask about extra-linguistic considerations such as political boundaries. In short, the linguist is supposed to determine what the languages are on linguistic grounds alone.
To make the situation more concrete, we can imagine that the information is provided by a sort of oracle that will provide specific (and infallibly correct) answers to questions of the form: If individual X (and here the particular person in question is to be identified unambiguously) were placed in a situation with the following characteristics (and here would follow as highly specific a definition of the situation as the questioner desired), is linguistic expression Y something that X would be capable of uttering? If not, what might s/he say?
Presumably the linguist will want to make this information which s/he has obtained available in the form of a published report or reports. What form should these reports take? I think the expected answer would be that s/he should make it available in the form of linguistic descriptions. But how many languages? With such complete information available, linguistic descriptions would become procrustean beds; to fit such information into any specific number of languages would require arbitrary decisions as to precisely which linguistic expressions should be attributed to any given language.
Of course, we would presumably assume that some expressions in some situations (and even some expressions without regard to situation) should ultimately be judged as not being attributable to any language--i.e., simply as errors (something a particular person has mislearned or something misspoken). But presumably these judgments can't be made until the individual languages have been defined.
From the outset, then, there would be the problem of defining the languages--of deciding which expressions are to be attributed to the same governing system (and therefore, of deciding how many languages to set up). The problem that I'm trying to call attention to is that human language isn't exactly organized into languages. We act as if it were: partly because of the influence of philosophy and partly (presumably) because we haven't been able to think of any better way to report the facts. Of course, to act as if it were would be perfectly all right (at least up to a point) if we were careful to see that everyone understood the as-if nature of our statements about this or that "language". But we generally aren't.
I don't mean to deny that language does seem to have a tendency to occur in the form of individual languages. In the real world there are factors that make language focus (to use Robert Le Page's term) into some approximation of the kind of individual systems that our assumptions require. What I do want to deny is (1) the completeness of this focusing--the linguistic facts don't partition exhaustively into individual languages, and the individual languages that can be identified are far from being completely homogeneous systems, and (2) that the factors that lead to such focusing are linguistic factors--in fact, as far as I can see, such focusing as does occur is due entirely to extra-linguistic factors!
No doubt modern standard languages in the European tradition are more suitable objects for description than languages for which standardization has not been attempted or completed. In fact, where there has been codification, that codification itself becomes an apt object for description (indeed, it essentially is that description). If people get together and say "we're regularizing the English [or whatever] language; here are the rules of its grammar, here is a dictionary giving its words with their spelling, pronunciation, meaning, etc.", then they obviously have provided us with an apt object for description. In fact, except for the difference between prescriptive and descriptive tones, what they have done is precisely to produce a description. (Linguists are now pushing these descriptions to much greater detail, of course).
In any case, it would seem to be generally true that the governing-system conception of how language works requires that it occur in the form of individual languages. Indeed, that conception seems to require quite sharply-focused (in Le Page's sense) systems. But we speak of "natural" languages. On what basis could we conceivably claim that any focusing at all is "natural"--at least in any strictly linguistic sense? It seems that a more reasonable hypothesis would be that the linguistically natural state of affairs would be one in which all of human language would take the form of a single giant chain. The description "chain of dialects" suggests itself, but on reflection that suggests a chain of more or less focused systems. So, even, does "chain of idiolects" in that it suggests the idiolect as a locus of systematicity. What I'm proposing as the natural state of affairs as far as strictly linguistic factors are concerned is probably best conceived of as a single continuum with no foci at all.
In any case it would appear that all of the focusing that occurs is due to the intervention of extra-linguistic factors: topographic, political, cultural. [In fact, in an article entitled "The patrilocal band: A linguistically and culturally hybrid social unit" (American Anthropologist 67: 675-90, 1965) Roger C. Owen has suggested that societies that were relatively homogeneous in language (and culture) may not have existed anywhere before the Neolithic.] Therefore, as far as I can see, there is no reason at all to postulate that focused language systems were somehow anticipated in the evolution of the human language faculty.
It's important to remember that when we talk about the language faculty, we are talking about something that developed during the evolution--the biological evolution--of the human species, or more precisely, of whatever ancestor of ours first had language. We therefore can't attribute to the language faculty any capacities that look ahead to cultural institutions that lay in the future, for example, writing or the pursuit of autonomous text. Or at least to do so would require one or another further assumption, such as:
(1) that biological evolution has continued (so that people without written languages would be biologically different from the citizens of modern civilizations) [Such an assumption would be politically unacceptable, of course, but all of the evidence seems to be against it anyway], or
(2) that the original evolution involved massive "exaptation" (to use the term proposed by Gould and Vrba), i.e., traits evolved which were selected for on the basis of an entirely unrelated advantage or advantages, but which turned out to be remarkably relevant after writing and other subsequent developments had emerged, or
(3) that something other than evolution by natural selection is involved--that there was some "looking ahead" (most likely some supernatural intervention--the usual version of this assumption invokes a creator).
As far as I can see, the results of (2) and (3) would be indistinguishable. In fact, (2) might easily be interpreted as crypto-creationist: proposing what, if there were no guiding hand, would have to be a remarkable coincidence of coincidences, and then (coyly?) leaving it to others to draw the conclusion that there must in fact have been a guiding hand.
I think that we must conclude that the language faculty does not contain a language acquisition device that is designed to acquire a governing system of the philosopher's-language type.
What I have been saying is that by permitting our view of human language to be influenced by the philosophers' vision of language, we linguists have come to work from, and lend our authority to, a fundamental misunderstanding of the basis of language--of how language is able to function. That is, we have understood language to consist of individual systems which, by means of lexical items and a logical syntax, govern the construction and interpretation of sentences which prototypically have determinate truth values.
This misunderstanding is supported by other ones. First, there is our projection of written language onto all language in which we conceive of preliterate language as essentially written language using letters-in-sound instead of visible letters. I think that is seriously misleading, and that the point that particularly needs emphasizing is that written language could not possibly have been anticipated in the evolution of the language faculty. But a governing system designed to produce the nearest-possible approximation to autonomous text would be effectively useless if it didn't employ writing.
Second, individual languages are necessary to the kind of governing systems required for the philosophers' language, and we have come to understand human language to consist of individual systems--individual languages. However, as I pointed out, individual languages don't exist (except, at best, in rough approximation) under natural circumstances, and I tried to show that that fact has more fundamental implications than linguists are accustomed to acknowledging. Furthermore, such focusing into clearly-defined systems as does occur is entirely accidental from a linguistic point of view--it is always due to extra-linguistic factors.
I'm proposing further that in addition to the misunderstanding of the basis of language, there is also a serious misunderstanding of the part played by the speaker. The governing-system conception of how language works requires that the individual speaker know the syntactic rules (regularities, generalizations) of a governing system in detail. I believe that this is a fundamental misconception--that it is not true even for Western intellectuals--and that making this assumption about what it is to know and use language directs us away from the basic questions that we should be asking.
This misunderstanding of what the language user has to know has led to a corollary misunderstanding--a misunderstanding of what must be attributed to the innate language faculty. The current assumption is that a detailed universal grammar must be innate in each human infant. This assumption, of course, would not be necessary at all if it were not for the other assumptions which lead back ultimately to the idea of languages which
The assumption that language works on the basis of governing systems is an obstacle to figuring out how people prepare their utterances and understand those of others.
The assumption that a universal syntax is innate--that that's what the language faculty essentially consists of--is an obstacle to any kind of progress toward understanding the origin of language.
I won't even include guesses, here, about the way such assumptions, well implanted in our society, have reshaped the ways in which we conduct our public and private business.
People have told me that if I'm going to write about problems I see in what we're doing, I should balance them with some suggestions about what we can and should do differently. I can certainly understand what they're saying and why, but I haven't any very conclusive answers.
It's easiest to talk about what to do about research. Kuhn has said that scientists could not reject paradigms when faced with anomalies or counterinstances and still remain scientists. He said, "To reject one paradigm without simultaneously substituting another is to reject science itself. That act reflects not on the paradigm but on the man. Inevitably he will be seen by his colleagues as "the carpenter who blames his tools."" (The quote is on p. 79 of the 1970 edition of The structure of scientific revolutions ).
In short, I think that for research, the best thing is for everyone pretty much to go on with what they're doing. The present research is obviously making progress. It's accomplishing something even if that's only gathering new observations about language-relevant phenomena. For one thing, current research approaches have turned up extensive (universal?) regularities of structure in texts. These surely cry out for explanation (even though present linguistic theory dismisses the problem by explaining everything as innate). Maybe with better assumptions we'd make more progress, but for the moment that alternative isn't available. That doesn't mean, however, that we shouldn't keep on the look-out for alternatives.
1. Andy Pawley has pointed out to me that the impossibility of autonomous text has probably been established most clearly by Harold Garfinkel and the ethnomethodologists. Back up
| Home Page | The Ethnolinguistic Notes | The Ethnolinguistic Notes, Series 1 and 2 | Ethnolinguistic Notes, Series 3 | The Ethnolinguistic Notes, Series 4 | Reflections: Language Evolution |
| Reflections: Knowledge of Language | Personal Page | The Human Predicament | Why Write Unpublishable Things? | Modest Proposals | Odds and Ends | Pictures |
Last updated 20 January 1996
Minor Correction Made 1 July 2006