Comments welcome
George W. Grace
University of Hawaii

Ethnolinguistic Notes

Series 4, Number 25

LANGUAGE ADAPTATION

What I want to talk about here is the adaptation of language structures in response to different circumstances of use. What I’ll say is very much an exploratory venture; I know of very little directly relevant research. However, I’m convinced that the matter is very much worth exploring, and I’ll undertake to sketch out some guideposts that might aid further investigation. In doing so, I’m going to focus on what I believe are two examples of adaptive changes. The first is the evolution of compositionality; the second is the pursuit of autonomous expression.

1. Analytic Processing and Compositionality

I’ll talk first about my reasons for believing that compositionality in language was achieved very gradually and was an adaptation to the conditions under which language was being used. This may appear to be at odds with an assumption that compositionality is a defining characteristic of human language. At least, it often seems to be assumed that the prototypical utterance (1) consists of one or more sentences each of which is made up of parts in such a way that the meaning of the whole is deducible from the meanings of the parts. Then, utterances that depart from this norm may be interpreted as derivative in one way or another. For example, some can be explained as abbreviations that rely on information supplied by the context; others as involving stereotyped formulas that are ultimately derived from proper compositional structures.

I originally got interested in this kind of adaptation from reading various reports of how languages or dialects that were spoken mainly or exclusively among members of a society tended to develop in different ways from those that were widely used in speaking with outsiders. Bailey (1982:10) applied the term “connatural” to the kind of developments that occur “when languages are left alone”, and “abnatural” to those that occur “as the result of contact with other systems”. Thurston (e.g., (1987, 1994) distinguishes particularly between languages in northwestern New Britain that are spoken only with other members of the same community and those used as lingua francas. He refers to the former as “esoteric” and the latter as “exoteric”. Others who have written on the same pattern of differentiation include Trudgill (e.g., 1989, 2002) and Henning Andersen (1988).

These writers typically describe exoteric languages as “simple” and esoteric ones as “complex”. Thurston (1987: 41) provides a convenient explanation of what is involved:

The most complex language, on the other hand, would have all the things we use to torture students of introductory linguistics--allophony, allomorphy, unpronounceable consonant clusters, gratuitous morphophonemics, and unpredictable suppletion. It would have an enormous morpheme inventory with many near-synonyms differing in slight shades of meaning with implications of degrees of formality and socioeconomic status. The most complex language would also have a lexicon which relied heavily on a large number of opaque idioms.

Communication under Esoteric and Exoteric Conditions

In retrospect, the fact that languages spoken under these different conditions should develop differently should not have been surprising. The principal difference seems to be in how much pressure there is on speakers to produce utterances that can be understood by means of analytic processing.

It can easily be observed that talk among familiars in which familiar things are said about familiar subjects tends to be fluent and to a great extent formulaic. And it’s also apparent that to the extent that formulas are already lexicalized, they require no analytic processing at all by either speaker or hearer. Where analytic processing isn’t required, the speaker has little incentive to articulate his/her utterances in such a way as to display the structural information that would be required for analytic processing. All that’s needed is for the utterance as a whole to be recognizable. Among other things, conditions where analysis isn’t required are favorable for fast speech with its associated articulatory phenomena (as emphasized, for example, in Trudgill 2002).

To the extent that such conditions predominate throughout a speech community, neither speakers nor hearers will need to resort to, or even be aware of, the analytic structure of utterances(1) (cf. the discussion of “needs-only analysis” in Wray 2002a). I’ll call communication occurring under such conditions, and the conditions themselves, esoteric (see Wray and Grace, in press, as well as Thurston 1987).

On the other hand, to the extent that esoteric conditions are lacking, the analyzability of the utterance structures assumes greater significance. Some factors creating exoteric conditions include: (1) addressees who have limited competence in the particular language or speech variety being used, (2) the need to express unfamiliar content, (3) addressees about whose background the speaker has little or no prior knowledge.

All such exoteric conditions would seem to select for greater compositionality. By “compositionality” I mean utterances or other linguistic expressions that are composed of parts in such a way that the meaning of the expression as a whole can be deduced through analysis. Although compositionality as I’ve defined it doesn’t imply transparency, communication with speakers of limited competence in particular should be expected to exert pressure for transparency, as well.

An expression is “transparent” in the intended sense to the extent that identification of the constituent parts is straightforward. In the ideal case of transparency there would be no idioms or allomorphy at all. Where there were allomorphs, automatic alternation (alternation governed by a general phonological rule) would rank next on the transparency scale followed by phonologically conditioned alternation (governed by morphophonemic rules). Suppletion would, of course, rank last.

Speakers with limited competence in the language might themselves contribute to greater transparency with their own utterances, because when non-native constructions are heard often enough in a speech community, they can sometimes be adopted by native speakers.

In the same way, outsiders may with their own speech be one source of the need to express unfamiliar content. People who are thinking in a certain language may be said to experience reality in terms of that language, or at the very least to interpret their experiences in its terms. Therefore, since these outsiders have surely done some of their thinking in another language, they’re likely to attempt to formulate in the new language content originally conceived in the other. The utterances containing these attempts might, if frequent enough or useful enough, be perceived by native speakers as exemplifying new structural possibilities.

Of course, the need to express unfamiliar content may arise in other ways. It will presumably occur from time to time in any community, no matter how esoteric. In particular, it would presumably increase with any increase in technological complexity or division of labor.

Speaking with outsiders would also contribute to the third exoteric condition mentioned-- unfamiliarity with the background of the addressee. And successful communication under such circumstances, of course, requires the inclusion of a certain amount of background information. In some cases this has itself probably brought pressure to accommodate new kinds of unfamiliar content. However, the need to accommodate background information may still never have been an enormous factor until after the advent of written language.

Esoteric and Exoteric Conditions in Evolutionary Perspective

It seems most likely that compositionality was achieved only very gradually in the course of language evolution. Of course we have no direct evidence of what the structure of language was like in its earliest stages, but there seem to be two good reasons to conclude that it must have been far toward the esoteric pole. The two reasons are: (1) the cultural conditions that must have existed and (2) the primitive stage that could have been reached in the evolution of utterance structure.

For perspective, consider what would constitute “perfectly esoteric” conditions. Under these ideal conditions there would exist a single recognized set of conceptual situations and events, to each of which a particular utterance would be dedicated. Furthermore, most if not all of these would be known by everybody who might potentially be involved as speaker or audience. Although these ideal conditions can serve as a convenient reference point in the esoteric-exoteric spectrum, it seems apparent that the conditions that obtained at the time when language was evolving fell short of this esoteric extreme. Otherwise, analytic processing would have offered no selective advantage and presumably would never have established itself.

However, as Alison Wray and I have pointed out ( Wray and Grace, in press; Grace 2003) , the conditions that existed when language was evolving, while not reaching the hypothetical extreme, must surely have been far toward the esoteric pole of the spectrum. That is, most communication must have taken place in small groups where individuals knew one another well and where the range of subject matters was, from today’s perspective, very limited. The obvious implication of this would be that earliest language was highly “complex”, but such a characterization is likely to be misleading in this case. (2) Therefore, I’ll just refer to languages themselves as “esoteric” when they show a high degree of adaptation to esoteric conditions.

Anyway, it’s hard to doubt that at the stage when it had first evolved far enough to be deserving of that name, language must have been quite esoteric in structure. Of course we can have no direct evidence on that structure, but I believe we do have sufficiently reliable speculations to provide a basis for discussion. In a number of postings on this website I’ve taken the invention of analytic processing as defining the stage that we may think of as “earliest language”. I’ve tried there to identify possible small steps through which this invention might have taken shape, and have come up with several suggestions that don’t appear (at least to me) to be unreasonable. I’ve summarized these in Grace 2005b , where I proposed a series of possible steps that would have led to the gradual recognition of recurring partials in some of the originally holistic utterances. Thereafter, my hypothesis is that these partials gradually came to function more independently, assuming a role similar to the morphemes of today’s languages. (3)

Subsequent Exoteric Pressures

Obviously all languages today reflect a significant amount of evolution away from this hypothesized original structure. The typical utterances of even the most esoteric of them analyze into many more parts. This, I suggest, is evidence of adaptation--at least at some times during their history--to exoteric pressures greater than those present at the first emergence of language.

What kinds of subsequent conditions might have been involved? For one thing, growth in population and in specialization would themselves have been enough to increase exoteric conditions. Population increase would ultimately have led to more contacts with a more diverse selection of people, while specialization would have meant more diversified knowledge and less completely shared experience within the population. And eventually these developments have reached, and in some measure affected, the entire world.

There would accordingly have been an increasing need for more and more background information. The expected result would be a selective advantage for any changes that would contribute to what Paul Kay 1977) described as a more autonomous style of speech. Kay’s description is worth quoting:

Functionally, the style of speech that Bernstein calls elaborate may be characterized as “autonomous”; that is, it is minimally dependent upon simultaneous transmission over other channels, such as paralinguistic, postural, and gestural, and it is minimally dependent on the contribution of background information on the part of the hearer. Characteristically, autonomous speech packs all of the information into the strictly linguistic channel and places minimal reliance on the ability of the hearer to supply items of content necessary either to flesh out the body of the message or to place it in the correct interpretive context. Autonomous speech is suited to the communication of unfamiliar or novel content to someone with whom one has little in common. It is ideal for technical and abstract communication among strangers and inappropriate for the communication of immediate and emotionally laden content between intimates. (Kay 1977: 21-2)

The pressure to produce utterances that were more autonomous--that supply background information that the addressee might not have--must have continued to increase over time in the world as a whole. This of course is not intended to imply uniform increase in all parts of the world at all times. However, such conditions would seem to have been at the forefront of evolution toward language structures that more closely conform to the requirements of modern global culture.

2. The Pursuit of Autonomous Text

The pursuit of greater autonomy of expression is very central to what I want to propose as a second example of adaptive change. However, before introducing this example, I want to comment on the importance of written language to the whole evolution. Written language deprived the addressee (or anyone else who chose to read and attempt to understand a message) of any knowledge that could have been obtained through observation of the speech act. Thus, more background information needed to be included in the message itself.

Of course, linguists have often avoided the designation “written language” in order to emphasize their position that language was inherently an oral phenomenon and that “reducing a language to writing” did not materially alter its essence. I see the merit in that argument, but I want to emphasize something different here--namely that in the longer run, the existence of a written form can radically affect the directions in which the language evolves.

The invention of written language seems first to have come about through the generalization of written notations that were already in use for limited purposes such as account-keeping to provide in the end an instrument for recording a full gamut of linguistic expressions.

It isn’t clear how many times this invention has occurred independently--i.e., without so much as stimulus diffusion. My own suspicion is that such an invention is a much more improbable event than seems usually to be assumed. For one thing writing isolates--for the first time?--what we now recognize as language from its context. What I mean is that a text that persists after an act of writing thereby automatically separates itself from the rest of that act or from anything else that does not persist with it. By this fact alone, it delineates what is to be counted as language. Although there may be no way of knowing, my guess would be that this separation entailed a more radical revision in the speakers’ understanding of how the world worked than seems usually to be assumed.

One significant fact about written language, of course, is that it increased the above-mentioned pressure for autonomous styles. Another significant fact about it is that, unlike its oral counterpart, it produces artifacts that are not ended with the act that creates them--they continue to exist until they are destroyed. This permanence of written texts seems to have opened up a range of new potential uses for language.

In fact, written languages have been called upon to perform a variety of functions and have taken a variety of forms. However, I want to focus here on one particular development that began in Europe but has played a major role in the recent history of the world.

The Modern Standard Language

This development--my second example of adaptive change--is what John E. Joseph in his 1987 book, Eloquence and Power, calls simply “the standard language”. However, since I’ve been concerned that it may be misleading to call languages of this kind simply “standard languages”, which seems to suggest that the reduction of free variation was their most noteworthy characteristic, I’ve resorted to calling them “Modern Standard Languages” (MSLs). I’ve discussed them somewhat further in Grace 2002a. (4)

There was a particular interest in the process of shaping new MSLs (although of course not under that name) in the period of accelerated decolonization following World War II (see, for example, Ferguson 1968). The process of designing an MSL involves essentially two parts: (1) selecting a dialect to serve as the standard and codifying it, and (2) expanding (“elaborating” is the word proposed in Haugen 1966) it to perform all of the functions of MSLs elsewhere. It’s apparent that these steps could hardly be accomplished without some effective mechanism of enforcement. In these post-colonial environments there were usually planning committees accompanied by substantial government support.

According to Joseph (1987), the creation of an MSL requires another language that serves as a model--in a relation that he labels “superposition”. In fact, the process of “elaboration” consists essentially of making whatever adaptations are required to permit the adapting language to translate effectively from the model language. More specifically, as Ferguson (1968: 28) points out, the necessary intertranslatability concerns “a range of topics and forms of discourse characteristic of industrialized, secularized, structurally differentiated, ‘modern’ societies.”

Linguists have tended not to consider the development and spread of the MSL as linguistically important. One reason is that they have regarded the adaptations involved as largely limited to vocabulary, and vocabulary has been regarded as less representative of the essence of a language than its grammar. Another is that an important part of the adaptations involve aspects of language use that linguistics has paid little systematic attention to. For example, Ferguson points out the importance of “the development of new styles and forms of discourse” (1968: 32), mentioning in particular “the structures of nonliterary prose (paragraphing, ordered sequences, transitions, summaries, cross-references, etc.)” (1968: 33).

However, we shouldn’t be too hasty in assuming that there are no syntactic features involved. It’s probably significant that Asians often speak of the innovation or extension of the use of passives in their languages, and I’ve seen various mentions of the importance of nominalizations (e.g., Olson 1994, passim). (In fact, I seem to remember that a couple of centuries ago while the MSL was still being shaped in Europe, Jeremy Bentham went so far as to advocate the replacement of all other verb forms with nominalizations, but I haven’t been able to confirm that).

Be that as it may, it would seem that the experiences of speakers of the languages involved deserve some attention. And there is a common complaint of those whose languages have undergone modernization in the not-too-distant past that the modernization has so altered their languages that they have difficulty reading pre-modernization writings. Specifically, I’ve been told this by Japanese and read it about Turkish (see Gallagher 1969: 65), but it also isn’t difficult to find comments on the strangeness of works in Western European languages that were written earlier than something like the middle of the 17th century. Grace 2006 (actually written in 1978) provides some examples.

Of course these complaints aren’t necessarily telling us that the changes made by modernization significantly implicate syntax, but if not, they may be telling us something more basic. Maybe they’re telling us that we need to find out what they do implicate and modify our conception of the nature of language to assure whatever it is a more prominent place.

But is it really adaptive change that the MSL exemplifies? In what sense does the “modernization” process represent adaptation? I believe there are two perspectives from which it can be described as doing so. First, the modernizing can be seen as adaptation to intercommunication with the existing MSLs with which the modernizing language is becoming increasingly intertranslatable. Second, it’s adapting at the same time to the requirements of the additional functions for which it’s to be used.

On the Origins of the MSL

However, I believe the first MSLs must also have been the product of adaptive changes even though they had no existing MSL to guide the adaptation. Actually, the essential feature of the MSL seems to come down to nonliterary prose styles of a particular kind. However, this kind of style seems quite generally to acquire a special prestige in the MSLs, and in fact to assume what might be called a canonical role. By this I mean that the rules required by such styles are described as being the rules of the language. As I discussed in Grace 1996, utterances that don’t conform to these rules are likely to be categorized as sub-standard or simply incorrect.

It’s important to note that styles of this kind provide possibly the closest approach to truly autonomous text that has been achieved anywhere. It seems that the original development of the MSL must have been a matter of co-adaptation between evolving institutions of Western culture and language structures that facilitated such styles. It’s particularly noteworthy that the rise of Western science seems to have depended heavily on the availability of such prose styles, while at the same time contributing to their development (note that the style of language to be used was a major concern of the founders of the Royal Society in 1660).

However, other institutions also probably played a role. An increasingly autonomous prose style no doubt facilitated the keeping of permanent records and promoted the growth of their multifarious progeny such as constitutions, statutes, contracts, treaties, records of many kinds in a great variety of bureaucratic agencies and businesses, the "information" recorded in reference works of all sorts (including scientific statements), and much else. (Of course, the fact that written texts can be analyzed, negotiated, and redesigned is critical to the proliferation of such functions).

In any case, what I want to propose is that the evolution of this style was an adaptation to such developing institutions at the same time that the latter were adapting to the possibilities presented by the development of the style.

One point that remains to be cleared up is this: If each new MSL requires at least one already- existing MSL as a model, how was the first MSL created? Joseph (my main source for this) traces the tradition back to Alexandrian Greek. It’s easiest to quote him:

With the first alphabet, the first systematic thought about language, grammar, and the koine, Greek in the Alexandrian period became the prototype for the standard language. (1987: 50)

Although Greek itself had no existing MSL to serve as its model, it did serve as a model to Classical Latin, “a language which, having been formed in superposition, comes closer than Greek to being the first ‘standardized’ language”. (1987: 50). To quote him further:

For over a thousand years Latin was the only language employed in what we would identify as standard-language functions throughout Europe and when the modern standardization process got under way with a series of cultural events which also serve to define the Renaissance, Latin was H. It was in every sense the model that the standardizers of the European vernaculars wanted their languages to approach. (1987: 50).

(His reference to the “H” role of Latin is borrowed from the literature on diglossia--in other words Latin served as the model for the emerging MSLs to adapt to).

As I understand Joseph, he considers the European vernaculars that adapted to add the functions previously performed by Latin to be the first true MSLs. And this raises another interesting point. It may be that the reason that it was in Europe that the MSL emerged was because of the unique condition in which Latin had existed for several centuries. During this time it was a secondarily-learned language reserved for what may be described loosely as intellectual functions. For example, I’m told that there was no way to swear in the Latin of that period. In the words of Walter Ong, it was used for:

… more or less abstract, academic, philosophical, scientific subjects or for forensic or legal or administrative or liturgical matters. (Ong 1977: 25)

Ong proposed that the existence of a language under such circumstances played a crucial role in the subsequent history of Europe--in particular by permitting the development of science.

Modern science only gradually became viable in the vernacular atmosphere as it transformed this atmosphere by injecting it with Latin terms and forms of thought. (1977: 36)

One puzzle is what he meant by “Latin forms of thought”. He seems to suggest that something at a deeper level was going on at the same time. There have been other suggestions that at about this time there were major changes in our conception of what language is. For example, a change in the whole conception of language is a central part of Foucault’s “classical episteme”. As I understand it, the key new idea might be summed up as the possibility of definitive statement--statements such that their meaning can be found in the text itself rather than only being revealed through commentary (see esp. Foucault 1973). Olson describes what seems to be the same ideological innovation. He attributes particular importance in the assertion and diffusion of this new conception to Martin Luther (see esp. Olson 1977, 1994). And it seems worth quoting Ong again:

Studies of the matter are nonexistent, but one could argue as an initial hypothesis that the modern intellectual world and the modern state of consciousness could never have come into being without Learned Latin or something like it. (1977: 36)

This ideological innovation clearly persists to the present day. I’ve written elsewhere about my concern that modern culture, including especially contemporary linguistic theory, takes for granted a conception of human language that is far from being universal or from being self- evident. It seems to me that our language ideology has been shaped to a significant extent by written language and particular prescriptions motivated by its envisioned use in producing autonomous text. However, this isn’t the place to pursue the matter. I’ve discussed it most recently in my “Collateral damage from linguistics?” series, maybe most directly in Grace 2002b.

Conclusions

What I’m proposing here is that languages are constantly making, and/or having made to them, adaptations to better fit them to be used under different kinds of conditions. The process I have in mind is the classic one of variation and selection.

I picture the variants as arising in various ways: experimentation (deliberate or unconscious), accidents of various kinds (e. g., blends), and the models provided by other speech varieties (which can be calqued or directly borrowed).

Selection among variants can be more or less conscious and deliberate. This is true both for the earliest adopters of an innovative variant and for those who follow. One kind of consideration (which can also be unconscious) for later adopters is likely to be the identity, authority, and actions of the earlier ones. The effect of these precursors will presumably depend on their status--which might involve anything from passively-exercised authority as role models (negative as well as positive) to the active authority to prescribe and enforce behavior.

Deliberate prescription seems to have played an exceptionally, possibly uniquely, prominent role in the evolution and dissemination of the MSL. The modern world’s formal educational systems alone constitute powerful and prestigious instruments for disseminating and enforcing its rules. I haven’t been able to conceive of any likely institutions in the preliterate world that would have been able to exercise a comparable measure of control over a population of speakers.

On the other hand, we also shouldn’t imagine that the changes were ever carried out by a disembodied language. They go back ultimately to the accumulation of choices (conscious or unconscious) made by individuals, and these choices must often have been influenced (again consciously or unconsciously) by the choices being made by other (often more authoritative) individuals.

NOTES 1. I’m using the term “utterance” here as I defined it in Grace 2005b: to refer to signals appropriate for being used as the vehicle of a single communicative act. Thus, “utterance” here refers to the signal alone, not to the act as a whole. Furthermore, an utterance is repeatable-- i.e., it is a type rather than a token. (Back up)

2. It appears that languages can earn the description “complex” in a variety of ways. One is elaborate and intricate sentence structure, especially where embeddings are involved. But in fact, the conditions that permit this kind of complex sentence tend toward being the diametric opposite of those associated with esoteric adaptation. A particularly important difference is the part played by writing. Whereas writing clearly plays an essential role in facilitating multiply-branched trees, it's more likely to result in a hostile environment for esoteric complexity.

On the other hand, the complexity ascribed to languages that are adapted to esoteric conditions seems mainly to consist of intractability. However, some of the features in Thurston’s above- quoted description of complexity certainly wouldn’t have been present in early evolutionary stages. The morpheme inventory would surely have been the opposite of enormous. It’s unclear how the concept of opaque idiom could apply in a situation where compositionality was still very limited. And it isn’t clear what role “unpronounceability” might have played. In fact, I’m not even confident about how flexible the speech apparatus would have been at this stage or even that communication was yet primarily oral.

However, two kinds of simplicity that might be lacking even in the early evolutionary stage defined by our hypothesis are compositionality and transparency. Thus, a language might be categorized as relatively simple to the extent that its utterances are compositional and transparent, and, of course, complex to the extent that it is not simple. By the first of these criteria, a language at the stage in its evolution at which grammar was just emerging could have qualified as complex. Its utterances were for the most part holistic, which means that they were for the most part not compositional.

The case with transparency is less clear. Would the relatively few utterances that were analyzable at this stage have been transparently so? It might seem natural to expect so. One might expect that in a system that had only a limited repertoire of recombinable parts (call them “morphemes”), these morphemes would not have exhibited any extensive alternation that was not phonologically conditioned. However, I think of at least one factor that might have led to a lack of transparency: the fitting-together of mini-systems to produce a single over-arching system (see Grace 2005a).

What I mean is this: As these fledgling systems progressed through their somewhat parallel evolutionary steps, the sets of recurring partials that emerged would have been different in each. Each in turn would have tended to abstract out quite different sets of referents (of “conceptual elements” [Grace 1987], that is, of things, acts, states, people, etc.). However, it seems most likely that as the different systems converged toward mutual compatibility, similarities between particular conceptual elements in different mini-codes would have become recognizable. As the integration progressed further, these similar conceptual elements sometimes would have come to be more and more identified with each other. The utterance-partials (morphs) that expressed them might then have come to function like morphologically-conditioned--presumably mostly suppletive--allomorphs of the same morpheme.

Thus, it seems at least possible that early emergent language not only was largely non- compositional, but also that such compositionality as existed may not have been very transparent. (Back up)

3. As I explained in Grace 2005b, I’m not sure what terms such as “word” and “sentence” might have been most profitably applied to at an early stage of the sort proposed here. Utterances of the posited type might be thought of as antecedents to our sentences in that they refer to an entire (conceptual) situation or event, and thereby play the role of vehicle of a complete speech act. On the other hand they would have been more like our words in their more morphology-like than syntax-like internal structure.

Of course, a division between syntax and morphology did eventually have to emerge. Alison Wray has made an interesting suggestion of a possible first step. She (Wray 2002b: 123ff) describes how a kind of topic-comment structure might have been improvised by combining complete utterances in sequence. (Back up)

4. Coulmas 1989 is also a useful source on the spread of the MSL, including its early European period. (Back up)

REFERENCES Andersen, Henning. 1988. Center and periphery: Adoption, diffusion, and spread. In Historical dialectology: Regional and social, ed. by Jacek Fisiak. Trends in linguistics, Studies and monographs 37. Mouton de Gruyter, pp. 39-83. (Back up)

Bailey, Charles-James N. 1982. On the yin and yang nature of language. Ann Arbor: Karoma Publishers. (Back up)