Links to pages: 416, 422, 424 Comments welcome
George W. Grace
University of Hawaii
April 30, 1987
In recent years, we historical linguists have had it pointed out to us that our work involves idealization in that we assume a state of language to be fairly represented by a linguistic description, even though such a description generally ignores the heterogeneity which presumably existed in the speech community. In this Note I want to point out other idealizations involved in our work.(1)
As Thomas S. Kuhn has told us, the activities of "normal science" are based on what he calls a "paradigm". The paradigm includes "theory, methods, and standards together, usually in an inextricable mixture" (Kuhn 1970: 109). The paradigm includes assumptions about the nature of the world and especially of that part of the world which is the object of study of the particular science.
There can be no doubt that historical linguistics does depend upon its own set of assumptions. As is to be expected, these assumptions present an oversimplified, an idealized, view of the actual reality. Such idealizations provide a useful foundation for research. However, although it is important to take advantage of the foundation which they provide, it is important at the same time to remain aware that they represent an oversimplification.
I see the following as four key assumptions characterizing the current approach to historical linguistics:(2)
ASSUMPTION 1. At any point in time, human language exists in the form of a number (an unspecified number) of distinct systems. In most discursive contexts, these systems are ordinarily called "languages". Mainstream synchronic linguistics makes the same assumption, of course; in fact, for many linguists these systems (the "langues" of Ferdinand de Saussure) appear to constitute the entire subject matter of linguistics. Historical linguistics conceives of these systems as objects undergoing change--as synchronic manifestations of traditions extending into the past and in most cases the future.
The basic assumption in synchronic linguistics could probably be stated as follows: any utterance must be in some language--must in fact be governed by the grammar of some language. However, this statement should not be interpreted in such a way as to exclude code-switching; few would deny that it is possible for parts of a sentence to be governed by the grammar of some language other than the language whose grammar governs other parts of the sentence, or governs the sentence as a whole.
In fact, this assumption does not of itself even exclude language mixing in the most extreme sense. All that the assumption requires is that the utterance can be said to be governed by a finite number of languages, the identity of all of which can be specified.
However, upon closer examination this assumption can be seen to present problems. We really have no satisfactory way of deciding which utterances are governed by the same grammar. For example, what do we do about dialect chains such as the Micronesian island chain extending from, say, the Truk Lagoon to Tobi? If we assemble a large collection of utterances from every island in the chain, which do we assign to the same language, and how many languages do we wind up with in the process? The same question can arise within a language such as English. If we have one utterance in "mainstream" English and another with distinctive dialectal characteristics which would not be permissible in mainstream English, are we to say that the two are governed by the same grammar? If so, how can we account for the differences in what is permissible? If not, how many grammars will be necessary to account for all of the utterances which are generally accepted as representing some kind of English? Because of problems such as this one we can be certain that no definitive answer is possible to the question of how many languages there are in the world (except, of course, by resorting to the trick of inventing an arbitrary definition of "(a) language" just in order to produce an exact figure).
However, something else becomes apparent in a discussion such as this: we have been considering the definition of the language as a problem in synchronic linguistics. But (especially since synchronic linguistics seems to have no very clear idea of how the scope of a language is to be determined for its purposes) it would seem that historical linguistics would be well advised to attempt to decide what kind of unit--what kind of "language"--would best meet its own needs. That is, if historical linguistics is to continue to conceive of its object in terms of changes of form in continuing linguistic traditions of this kind, it might attempt to determine for itself how such traditions manifest themselves on the synchronic scene. This pretty much boils down to asking what kind of entity on the synchronic scene meets the conditions set by the assumptions which we are discussing here (particularly Assumptions 3 and 4, below). We will return to that question.
ASSUMPTION 2. A language is adequately characterized by its synchronic linguistic description. Therefore, for all intents and purposes when we speak of a change in the language, what we really mean is nothing more and also nothing less than a change in its linguistic description. That is, a language is assumed to have undergone a change when something has happened which calls for a change in its linguistic description (according to whatever school of descriptive linguistics one belongs to).
Although this assumption seems evident in much of the literature, it is rarely made explicit (see, however, the chapter entitled "Kinds of phylogenetic change" in Hockett 1958). It does have certain consequences. One rather conspicuous one is that when a new synchronic theory gains ascendancy, bringing historical linguistics into line is usually one of the first concerns of its advocates. As a consequence, our way of thinking and talking about linguistic change generally undergoes an abrupt transformation whenever synchronic linguistics undergoes a paradigm change.
A problem which arises is this: If, as is often assumed, the most realistic goal for a theory of linguistic change is to state the constraints on single changes--on what states of language can directly follow a given state of language, then it is important to be able to determine the moment at which a change occurred. That is, in order to be able to examine the state of affairs which most immediately preceded a change, we need to know when the change occurred. For this purpose it is important to be able to determine just what the linguistic description of a language is at any given time. But synchronic linguistics is largely silent on how the moment is determined at which the description of a language must be altered. It is not clear what conditions determine how such a decision should be made even in the case of a completely homogeneous speech community. It is even less clear how it should be made in a community of normal heterogeneity. Thus, the conception of change as equivalent to a change in the linguistic description leaves us with a large margin of indeterminacy.
Another (to me) uncomfortable point about this conception of what change is, is the fact that sometimes where speakers of a language feel that that language has undergone very rapid and striking changes--as in the case of China, East and West Germany, or North and South Korea in the post-World War II period--the changes which seem so conspicuous to the speakers appear to affect the linguistic description of the language scarcely if at all. Therefore, these changes are largely imperceptible to the perspective of historical linguistics.
I believe that we must ask ourselves whether the changes which these language users are observing are of an entirely different kind from the changes in which historical linguistics has interested itself. Or whether they, on the contrary, present a close-up view of those very changes; whether they represent precisely the process of change which ultimately leads to the changed linguistic description. In the latter event, it would seem that historical linguistics should be interested in them, whatever their relation to linguistic descriptions.
ASSUMPTION 3. These systems (languages) maintain their (separate) identity through time even while undergoing changes. That is, a language remains the same language (just a different historical stage of it) even though its linguistic description changes significantly. In fact, there is no limit to how much the description may change with the language remaining the same language.
According to this assumption, a language can split into more than one, but different languages can never merge. The boundary separating different languages is therefore assumed to be of an entirely different kind from the boundary separating dialects. A language boundary is supposed to be largely impermeable; in some accounts the only things linguistic which can diffuse across a language boundary are vocabulary items, while any kind of linguistic feature whatsoever apparently can diffuse across dialect boundaries. It is this assumption which underlies the principle that there can be no mixed language in the sense of a language which is the continuation of more than one language of any earlier point in time, and that, therefore, every language represents a single tradition that goes back to the (or an) origin of language.
I will consider this assumption further in conjunction with assumption 4 below.
ASSUMPTION 4. Linguistic change is systematic. The language system is the entity which changes, and the changes are constrained by (and to a very considerable extent motivated by) the system. (Thus, the theory of linguistic change as usually envisaged is concerned with how the linguistic description of the post-change state of a language may be explained in terms of the description representing the status quo ante ).(3)
According to this assumption, a linguistic change should be conceived of as a change in a linguistic system. That is, changes should be conceived of, and reported as, the replacement of one systemic state with another. Likewise, the explanation for a change is to be sought in the systemic state which immediately preceded the change.
Probably few linguists would explicitly subscribe to this assumption as stated. My statement of the assumption is based on what I take to be a general consensus that the theory of linguistic change is to be concerned with specifying possible relations between descriptions of earlier and later stages of the same language and that the specification is to be formulated in terms of the earlier description.
Certain implications of this assumption are worth noting. First, it acknowledges only what Henning Andersen (cf. Andersen 1974) calls "evolutive" changes (those changes which are entirely explainable in terms of the linguistic system that gave rise to them), and ignores all varieties of what he calls "adaptive" changes. In so doing, the assumption clearly implies that everything external to a language is irrelevant to anything in the language--to anything which would be reported in its linguistic description--and conversely, presumably, that all of the features of the language (at least those which would be reported in a linguistic description) are irrelevant to anything outside the language. I criticized the assumption on these grounds in Grace 1981 (esp. pp. 31-32, 92-93).
THE LANGUAGE OF HISTORICAL LINGUISTICS. But perhaps the most intractable question of all is the question of what the language of historical linguistics--the synchronic manifestation of the kind of continuing traditions that we are looking for--is. We saw above that, although the individual language is generally understood to be the locus of linguistic change, synchronic linguistics has provided us with no clear idea as to what a language is--how inclusive a unit it is.(4) Thus, historical linguistics is essentially left to its own devices in defining the entity upon which all of its attention is focused--the synchronic manifestation of the diachronic entities with whose vicissitudes it is concerned. It seems clear that the characteristics of the entity which we need are specified in the preceding assumptions--especially assumptions 3 and 4.
How, then, are we to recognize on the synchronic scene the entity which will maintain its separate identity over time and which will change as a single unified system? Where can we find an entity which behaves like the language of historical linguistics is supposed to behave?
The answer is not easy. Assumption 4 requires a structured unit which changes as an unit. But such units appear to be very narrowly defined. Studies of sound change in progress have generally identified relatively limited communities (e.g., Martha's Vineyard, New York City, the Clonard community in Belfast) as the loci of such changes.
Similarly, in the past history of English, different regional dialects seem clearly to have constituted different traditions in the sense of Assumption 4--each undergoing its own changes and having its own history for extended periods of time. In fact, contemporary standard English includes a number of words whose reflexes demonstrate that some of their history has been separate from that of the standard dialect. The presence of such words is therefore attributed to borrowing even though the source and recipient are both English.
On the other hand, it is impossible to construct a family tree using dialects as the elements because dialects cannot be counted on to maintain their separate identity indefinitely (i.e., they do not meet the conditions of Assumption 3). To continue with the example of English, it seems likely that English as a whole (i.e., the entire range of dialects and varieties which are called English) is fairly sure to maintain its separateness into the indefinite future, but there seems to be no narrower system which presents such prospects. We thus are confronted with a dilemma: English as a whole is certainly not a system of the kind which serves as the locus of linguistic changes. If the changes which have been observed in the English of various communities of the English-speaking world are explainable from the information which would be made available by a linguistic description of the status quo ante of the linguistic system, the system in question--the system whose description is required--is surely not English as a whole (if indeed a description of English as a whole is even possible in the current conception of linguistic description). On the other hand, the local varieties which behave as units in linguistic change (thereby satisfying Assumption 4) can not be counted on to preserve their separate identities over time (thereby to satisfy Assumption 3). It seems, in sum, that it is purely coincidental if the kind of system (if indeed there is one) which meets the requirements set by Assumption 3 and the kind of system (if there is one) which meets the requirements set by Assumption 4 happen actually to coincide in a particular instance. But why should we expect that it would be otherwise?
From the foregoing, it seems apparent that the process of linguistic change as it is construed by the assumptions described above represents an idealization. It is noteworthy that the qualities attributed to the languages posited by the assumptions of historical linguistics are somewhat analogous to those of living organisms. The organic world may be said (1) to be constituted in its entirety of individual organisms, (2) each of which has a definite structure which is subject to analysis and description, (3) each of which evolves over time but maintains its separate identity throughout its life (so that organs from different organisms can't be combined [or couldn't before organ transplant] to produce organisms of mixed history), and (4) each of which evolves largely in response to evolving requirements from within itself. (However, an analogy can also be found to mechanical devices designed of a large number of interacting parts such as an engine in the early "running-in" period in which mutual adjustment of the moving parts is going on).
These analogies are all very well, and no doubt there is much to be gained by identifying and exploiting analogies (or "metaphors"), but languages are not organisms or machines. That much seems quite certain even though we are not able to say what they (and in particular, the languages of historical linguistics) are.
One further thing which is quite certain is that they exist only through the intermediary of the human beings who speak them. It seems, then, that it must be in terms of its speakers that the language is defined. And it seems clear that the definition of the language will be more precise to the extent that the definition of the set of speakers is precise--to the extent that an exact partition of people into speakers and non-speakers of the language in question is possible.
The ideal situation (I presume we are to disregard the modern phenomenon of language standardization where the rules of correct usage have been codified and where the authority of model speakers has nominally been replaced by superorganic authority) would be one in which all speakers agreed as to which speakers were suitable to serve as models of correct use of the language (so that anyone who did not accept them would be clearly a speaker of some other language). We might speak of a language in this ideal situation as "monodialectal". A monodialectal language in this sense would approximately satisfy Assumption 1.
I am not able to conceive of any circumstances under which Assumption 2 (which requires an all-purpose description which is adequate to serve as a surrogate for its object in all situations) could truly be satisfied (for further discussion, cf. Grace 1981: 13-22).
As for Assumption 3, we begin with the fact that the language has been defined in terms of a set of speakers. No doubt there would be some changes in the speech of many (even all?) speakers during the course of their lives. Let us assume (although I am not sure what ensures it) that these changes conform to the requirements of Assumption 4. However, the changes which occur during the lifetime of a single individual are not likely to be very great. The time spans with which historical linguistics ordinarily works (we might think of a "linguistic time" on the analogy of geological time) are generally much greater than the life span of an individual. So we must ask how a language can maintain its separate identity over millennia when its identity depends on individuals who normally live for less than a century. Well, it is obviously impossible for it to do so unless there is turnover in the defining set of individuals-- unless, in other words, there is recruitment to compensate for the losses through death.
One might imagine that this turnover would present no problem. Wouldn't we simply have children replacing their parents, and wouldn't it be their parents from whom they had learned the language in the first place? If so, why shouldn't they be expected to continue the tradition just as faithfully as their parents would have?
Such a process is possible, of course, but it cannot be counted on to represent the whole picture. Presumably some speakers will have no children, while others will have a great many, but that probably will not represent a serious distortion in the next generation. Anyhow, we know that children do not very precisely conform to their parents speech patterns, but are much influenced by others, especially others in their own age group.
But there are other possible complications. Perhaps a child's father and mother do not speak the same language, or at least do not have the same first language. In some parts of Melanesia, virilocal exogamous communities in which the wives all came from communities with other native languages appear to have been fairly common. (A German writing on such communities noted that they gave a new meaning to the word Muttersprache ). And, of course, there may be people who were not born in a community who nevertheless learn its language (not necessarily from members of the community)--that is certainly the case with English, which is being studied daily by people all over the world.
But maybe the recruitment of new speakers is not the key issue. Recall that the original defining set of speakers (of this ideal monodialectal language) was selected on the criterion that they all accepted the same other speakers as models of correct usage. And, of course, here again we encounter the problem posed by the limited human life-span--the problem that the model speakers, who were really the source of the unity of the set of speakers (and therefore of the language), would themselves also die and have to be replaced. How could we select their replacements in such a way as to ensure that there would be no changes between the usage of these replacements and that of the original model speakers (or that, if there were any changes, they would be of the sort required by Assumption 4)?
The key would seem to be that the replacements should be people who had the existing set of model speakers as their own models during their original language acquisition, and who, for that matter, continue to recognize them as their models. But beyond that, it would seem that the chances of avoiding non-conforming changes will be better in the measure that the environment in which the language is spoken is shielded from new ideas and influences. Thus, as a rough rule of thumb we might also say that Assumption 3 would be better satisfied in the measure that (1) the community is monolingual and endogamous and without internal social mobility, and (2) its culture is stable.
I am not sure what conditions are necessary to satisfy Assumption 4, but it would seem that the conditions which favor 1 and 3 would also generally be favorable for 4.
To conclude, then, historical linguistics makes assumptions which have the effect of creating a highly idealized model of diachronic linguistic processes. It seems that the expectations created by this model are likely to be realized in the places that, and to the extent that the actual conditions approximate the assumed conditions. However, it also seems clear that much of the time these assumed conditions are not closely approximated.
The foregoing assumptions suggest that the evolution of linguistic systems, as they are reported by linguistic descriptions, will not be affected by anything external to the language. It will not be affected if the speakers think many of their thoughts in other languages; it will not be affected by changes, no matter how radical, in what its speakers talk about or in changes of other kinds in their ways of talking.
In fact, they carry the implication that monolingualism is the normal human state. I believe that that assumption is questionable, that it is entirely possible that there have been more multilingual than monolingual individuals in the history--even the recent history--of humankind. But we have tended to think of multilingual communities as border phenomena, as abnormal communities resulting from an unclear boundary between normal (i.e., monolingual) communities, and of course a person intending to do a linguistic or ethnographic description would not normally choose such a community for his/her research, but some languages exist in such communities, and we know little of the form which linguistic diachrony takes there.
Our idealizations have limited our horizons. We actually know little of diachronic linguistic processes as such and almost nothing of changes other than those which affect those features of individual languages which would be reported in linguistic descriptions.
1. For the last several years I have given lectures under the heading "the assumptions of historical linguistics" to my classes in comparative method at the University of Hawaii. This note is based mainly on those lectures and on Grace 1978. (Back up)
2. Although I believe that all of the assumptions described may justifiably be characterized as assumptions of the field, their status within the field varies somewhat. Assumptions 1 and 3 are probably most solidly established. Probably most linguists who would think of questioning them are people whom other linguists might characterize as "sociolinguists".
I expect that most linguists might want to qualify Assumption 2 to say that a language is adequately characterized only by an adequate linguistic description. However, to the extent that linguists believe in the descriptive frameworks in which they are working, they are likely to behave as if they believed in Assumption 2. Therefore, even though it is not generally acknowledged in theory, it is very largely adhered to in practice.
Assumption 4 exhibits still more prominently the contrast between assumptions and beliefs (a contrast which I have discussed in considerably more detail in Grace, forthcoming). Presumably no linguist would be prepared actually to assert that all linguistic change is motivated by the system, and therefore entirely disregard what Henning Andersen (e.g., Andersen 1974) calls "adaptive" as opposed to "evolutive" change. However, most approaches to a systematic treatment of the processes of linguistic change regard adaptation to changing socio-cultural conditions or to other languages with which they are in "contact" as something analogous to "noise" in the system, with the truly linguistic phenomena being those which I have described. (Back up)
3. Note, first of all, that assumptions 1 and 2 are not assumptions specific to historical linguistics but rather assumptions of contemporary linguistics generally.
Assumption 3 (together with Assumption 1) underlies the notions of genetic relatedness and of the family tree. Assumption 4 (together with 1 and 2) underlies the notion of regularity of change.
The notion (assumption) of regularity of change underlies the comparative method, and is basic to the notion of reconstruction generally.
Assumption 4 together with Assumption 2 and a model of linguistic description in which phonology is autonomous underlies the assumption that the only possible conditioning factors in conditioned sound changes are phonetic factors. (Back up)
4. As I have tried to show elsewhere (cf. Grace 1985), synchronic linguistics has been much more concerned with the question of what form linguistic descriptions are to take than with the question of what they could be descriptions of. (Back up)
Andersen, Henning. 1974. Towards a typology of change: Bifurcating changes and binary relations. In John M. Anderson and Charles Jones (eds.). Historical linguistics: Proceedings of the First International Conference on Historical Linguistics, Edinburgh 2-7 September 1973. Amsterdam: North-Holland, Vol. II, pp. 16-70. (Back up)
Grace, George W. 1978. On the foundations of diachronic linguistics: What can go right? Ethnolinguistic Notes, New Series, No. 3. Mimeo. (Back up)
Grace, George W. 1981. An essay on language. Columbia SC: Hornbeam Press. (Back up)
Grace, George W. 1985. On the notion "linguistic description". Ethnolinguistic Notes, Series 3, No. 24. Printout. (Back up)
Grace, George W. forthcoming. The linguistic construction of reality. Beckenham, Kent: Croom Helm. (Back up)
Hockett, Charles F. 1958. A course in modern linguistics. New York: Macmillan. (Back up)
Kuhn, Thomas S. 1970, The structure of scientific revolutions. 2nd edition. Chicago: University of Chicago Press. (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|
First put on the Web on 31 March 1997