Sarah Grey Thomason and Terrence Kaufman. 1988. Language contact, creolization, and genetic linguistics. Berkeley and Los Angeles: University of California Press. 423 pp. US$48.50, cloth; $16.00, paper.

William R. Thurston. 1987. Processes of change in the languages of North-Western New Britain. Pacific Linguistics B-99. Canberra: The Australian National University. 155 pp. Aus$23.90, paper.

"It seems that whenever two languages come into close contact, they will borrow features from one another" (Comrie 1990:10).

For too long, the central questions in historical (apart from comparative) linguistics have been "How do languages change?" and "Why do languages change?" Now, in response to richer data turned up by sociolinguists and by field linguists in a great diversity of sites around the world, more historical linguists seem to be shifting their focus to the agents of change and rephrasing these central questions, asking: "How do people change their languages?" "Why do people change their languages?" Both of the works reviewed here call for a revolutionary emphasis on the social dimension of language change, arguing that human agents play the primary role.

However, the inertia of nineteenth-century determinism in the scholarly discourse of historical linguistics is still with us. Most historical works, even these two, focus on individual languages as they passively undergo a series of changes. The style of discourse is reminiscent of a Kafka novel. Other languages exert vague influences, while the human actors in the drama remain as faceless as bureaucratic statistics. They are invisible. Their dialog is inaudible. Only the narrator’s monolog can be heard tracking the shapes and movements [end page 191] of the starring language as it proceeds through a series of still-life contortions to move across center stage in a silent historical pageant. This kind of linguistics is more than autonomous. It is autistic.

More historical and descriptive linguists need to rewrite such dramas, as sociolinguists have been doing, to put more than one language on stage at a time, to add realistic settings and dialog, to recast the languages as instruments (albeit very complex ones) and to put humans back into the play as agents and experiencers. But that is a tall order. It is easier to paint a bowl of fruit than a horse race, easier to describe the structure of dead plants and animals than a living ecology, and easier to describe the abstract structures of idealized, homogeneous languages in isolation than the actual linguistic behavior of living humans. Nevertheless, perhaps we can improve by making a concerted effort to implement one small, but far-reaching, change—putting the missing agents back into our statements about language change. I have made such an effort in this discussion of two revisionist approaches to understanding how and why languages change, and I find the exercise not only challenging but fruitful. It keeps our models, metaphors, and claims closer to reality. (Crowley 1990 is exemplary in this regard.) After outlining some of the implications of these two works, I would like to rephrase one of the basic questions of historical linguistics in order to put the human agents of change back into the spotlight.

Thomason & Kaufman: Speaking two languages is normal. Thomason and Kaufman (T&K) aim to broaden the scope of genetic linguistics to embrace language contact and creolization. Although many of their observations have been made, acknowledged, and then ignored (as inconvenient) in the past, the copious documentation, carefully crafted arguments, and comprehensiveness of their approach ensures that their conclusions will be much harder to ignore in future work.

T&K see social factors as the primary determinants of change, and linguistic factors (such as markedness and typological distance) as secondary determinants. After devoting chapter 2 (13–34) to presenting evidence of the transfer from one language to another of elements that appear to violate putative linguistic constraints, they provide an analytic framework for contact-induced language change in chapter 3 (35–64). Since they incorporate cases of language shift as well as language maintenance into their framework, they allow for the possibility of mixed languages. However, they argue that only in extreme cases is genetic continuity disrupted enough to disqualify a language from being assigned a position on a family tree. Table 1 summarizes T&K’s framework. Although it is based on their table 3 (50), I have replaced their agentless nominals ("language maintenance," "language shift," "heavy borrowing," etc.) with a series of agentive sentences.

Chapters 4–7 (65–199) explicate these different possibilities and chapter 9 (214–331) presents eight case studies, the longest of which is Kaufman’s contribution, entitled "English and other coastal Germanic languages, or why Eng-[end page 192]lish is not a mixed language" (263–331). Chapter 8 (200–213) notes that the results of borrowing are most apparent in the lexicon, while the results of language shift (or "substratum") are most apparent in the grammar, and suggests that any "significant discrepancy between the degree of lexical correspondence and the degree of grammatical correspondence" between any two languages most likely indicates "abnormal transmission, or at least extensive diffusion," rather than direct inheritance (206). The extent to which transmission can be considered "normal" (and therefore genetic) or "abnormal" (and therefore nongenetic) thus rests on the degree to which such discrepancies are judged to be "significant." Table 2, which tabulates T&K’s list of the most likely conclusions to be drawn from different kinds of matches and mismatches between any two languages (205), gives a false impression of how clear-cut actual cases are likely to be. Much depends on the definition of such qualifiers as "markedly," "widely," "closely," and "partially."

Speakers maintain their NL, but know the OL.

*Few know the OL.

NL speakers borrow OL (nonbasic) words.

*Many know the OL.

NL speakers borrow many OL words and structures.

Everyone wants to or has to know the OL.

NL speakers borrow massively from the OL or abandon their NL.

Speakers shift from their NL to the OL.

*All learn the OL well. (OL teachers overwhelm learners.)

NL speakers do not change the OL, leaving it unadulterated.

*Few learn the OL well. (Learners overwhelm OL teachers.)

NL speakers use NL structures in the OL, giving it an NL cast (or "stratum").

No one wants to or is able to master the OL.

NL speakers use OL words but NL structures, abruptly creating a new creole NL.

Speakers shift away from their NL but not to any one OL, creating a new pidgin.

Many agree on pidgin usage, stabilizing the pidgin.

NL speakers recognize a new OL. Some may shift to make it their NL.

(NL = native language; OL = other language; * = "normal" genetic transmission; "structures" = phonology, syntax, and semantics)

Partly because it accommodates some degree of fuzziness, T&K’s framework seems not just more comprehensive than others, but also refreshingly in touch with the real world, in which humans are agents actively employing often-multiple linguistic instruments, and not simply passive hosts, each inhabited by a single linguistic organism. Unfortunately, T&K’s focus on "two-language contact situations in which the direction of interference can be definitely established" leaves them with little to say about the "notoriously messy" Sprachbund situations (95). However, their framework could easily be extended to cover diglossic situations in which the two varieties in use are considered dialects of a single language rather than two separate languages. But T&K do not intend to provide a unified explanation for all language change [end page 193] under conditions of contact. In Kaufman’s words, "overarching models that use a small number of axioms and try to explain practically everything annoy us. We dislike reductionism (ad absurdum). We are splitters, not lumpers. The purpose of this book is to introduce some subtlety into the thoughts and words of historical linguists" (328). In that, they succeed admirably.

Thurston: Creating new languages is normal. Thurston (T) is a lumper, not a splitter. He would throw out the family-tree model altogether and explain all language change in terms of cycles of pidginization and "indigenization" (150). Instead of expanding genetic linguistics to accommodate pidgins and creoles, he adopts a pidgin-and-creole model for all language change. His work is far more polemical, attacking traditional linguists and anthropologists in sweeping terms without, however, citing much literature to support his uncharitable—though not entirely unfounded—characterizations. (He lists 5 pages of references against T&K’s 21 pages.) Nevertheless, his strength lies in having done the kind of fieldwork that is needed to begin describing how multilingual Melanesian villagers manipulate the languages in their repertoires, rather than how each language looks after it has been immobilized in its own private pickle jar. His weakness lies in extrapolating from the small-scale societies of Melanesia to all the rest of the world, especially the world of huge populations speaking standardized languages with long literary traditions. His claims far transcend his evidence.

In gross outline, the hypothesis ... is that languages change in numerous ways which can be classified as functions of two basic patterns, one gradual, the other rapid. Each kind of change is associated with a set of sociological conditions that motivate language change in the first place; and each kind of change leaves traces in the modern form of the language [end page 194] that reveal the sociolinguistic functions of the language in the most recent generations. In the path leading to modern vernaculars, all languages have been moulded by alternating periods of gradual and rapid change. [T acknowledges Stephen Jay Gould’s writings on punctuated equilibrium—JB.] The most recent period in the prehistory of any language is that most evident in the structure of the language, particularly in communities with no tradition of literacy. (1)

The two basic sociological conditions he refers to are communication with insiders and communication with outsiders. Each function calls for different linguistic resources and thus generates different kinds of language change.

Esoteric languages function primarily as codes of communication among people of the same social group. Esoterogeny is a process that adds structural complexity to a language and makes it more efficient as a medium of communication among people of the same social group, while making it more difficult for outsiders to learn to speak well. Esoteric speciation is a process in which two or more endolexically-related languages are differentiated by copying resources (primarily ectolexical) from different sources. [The difference between the endolexicon and ectolexicon is explained below—JB.] (150)

Exoteric languages have, as one of their important functions, use as a lingua franca between people of different social groups. Exoteric languages tend to be structurally simple, because they must be easily learned by adults with different linguistic backgrounds. Exoterogeny is the process of simplifying an esoteric language to create a register that is more easily acquired by outsiders. Exoteric speciation, like pidginisation, is a process in which endolexical items are taken, without complexity, from one language and mapped onto the phonology, syntax and semantics of a substrate language to create a new language. (150)

In a later publication, T acknowledges that "literacy and formal instructional institutions muddle this relationship" (Thurston 1988:559, n. 3). Many large, national languages serve both esoteric and exoteric functions. In their work, T&K take pains to cite cases of both borrowing and shift that have complicated, as well as simplified, components of the languages affected. Although it seems to be true that the Austronesian coastal languages in Northwest New Britain are morphologically simpler (more regular) and also more exoteric in function than the inland languages, especially the one Non-Austronesian language in T’s sample, perhaps the same languages were "simpler" in the same way before they were adopted for exoteric functions. Or perhaps the coastal languages were adopted for exoteric functions because the coastal people had wider trade contacts. People are more likely to adopt a language for exoteric functions if its speakers are politically or economically dominant, regardless of how easy or [end page 195] hard the language is for outsiders to learn. English has displaced French as an international language in many domains, but not because English is any easier to learn. Once a language has been adopted as a lingua franca, it does not always undergo simplification as a result. Has English morphology become more regular as a result of its exoteric use? The outcome depends on the relative power of those who use the same language esoterically, on the degree of social contact between the two groups of users, and on other social factors.

Since all the languages in the Sprachbund he studied share much the same phonology, syntax, and semantics (despite significant disparities in morphological regularity), T focuses most of his attention on the lexicon. He distinguishes two lexical domains: an endolexicon of lexemes that "have high frequencies of occurrence in mundane conversation" and "are usually among the first forms mastered in a language," and an ectolexicon of lexemes that "have low frequencies of occurrence in mundane conversation" and "are usually learned after most of the endolexicon has been mastered" (149). Unlike the culturally neutral basic vocabulary used in compiling Swadesh lists, the endolexicon includes all the basic vocabulary in everyday use in a particular speech community, including culture-specific categories of food, clothing, and shelter, as well as all the necessary grammatical items to put basic sentences together, including pronouns and inflectional affixes.

T feels it necessary to state that he regards languages as cultural constructs that spread by cultural diffusion, like nautical or automotive technology (102), rather than as living organisms that spread by migration, like birds or insects. I doubt many reasonable linguists these days would disagree with him, so long as the particular manifestations of individual languages (as countable entities) are carefully distinguished from the universal aspects of human language (as an uncountable entity). It is true that culture historians tracking population movements sometimes fail to make careful distinctions between languages and populations, and between migration and diffusion. On this point, T singles out Bellwood for extended criticism (99–103). Unfortunately, it is also true that a great many autonomous linguists concern themselves almost exclusively with a small number of putative structural universals of human language (as an uncountable entity) abstracted from any sociocultural context, and ignore all that is most particular about individual languages and cultures (as countable entities). But T wastes no ammunition on them. Instead, he fires repeated broadsides at traditional comparativists.

It is true that comparativists have tended to think of languages as species (see Goodenough [1992] and McCawley’s [1992] discussion of Jespersen, for example) and to go about classifying language specimens the way nineteenth-century botanists classified dead plants. No doubt the traditional definition of species as reproductively isolated from each other accounts for the dogma that languages can never mix. (At one point, T characterizes Dyen’s attack on Capell’s notion of mixed languages as "mere sniping from the trenches of dogma" [90].) As a practical matter, however, the potential for taxonomic spe-[end page 196]cies to interbreed (or for taxonomic languages to mix) is rarely ever tested—or even testable, given the nature of the specimens used to determine the vast majority of taxonomic categories, and given the fact that all but the terminal nodes on a dendrogram are hypothetical reconstructions. This has led even biologists to question the utility of breeding potential as a criterion for classifying species. (See Sokal & Crovello 1970, for example.) Nevertheless, comparativists tend to use mutual intelligibility as the linguistic equivalent of interbreeding potential, even allowing for the well-known problems of applying that criterion to dialect chains or to stable multilingual societies. But, to my knowledge, very few linguists actually test mutual intelligibility by seeing whether two speakers understand each other. Intelligibility is a lot more difficult to measure than fertility. The former is a gradient phenomenon; the latter all-or-nothing. Many linguists thus extrapolate values for mutual intelligibility by calculating cognate percentages from lists of basic vocabulary. T’s investigation of endolexical and ectolexical relationships among one group of Melanesian communalects casts serious doubt on this procedure.

Several communalects scattered along the coast are "so similar in endolexicon and structure that they have been called dialects of a single language" (93). The speakers of these communalects, however, regard each as a separate language—and for good reason. Each community has built up a good portion of its local ectolexicon by acquiring terminology from different, less closely related (or entirely unrelated) neighboring languages. As one simple example, inland languages often share maritime vocabulary with their neighbors toward the coast, and coastal languages often share forest vocabulary with their neighbors immediately inland. (But not all the borrowings are so utilitarian.) Of course, it can be argued that nonbasic shared vocabulary counts for little in genetic classification, to which I would counter that all shared vocabulary, especially borrowed vocabulary, improves mutual intelligibility, insofar as that counts for anything in genetic classification. The many latinate borrowings into English, combined with a few years of formal instruction in high school, allow me to read French tolerably well, in spite of minimal conversational ability. Similarly, the many sinitic borrowings into Japanese and Korean allow traditionally educated speakers of those two languages and of Chinese to communicate tolerably well with each other in writing, even if they cannot understand a word of each other’s speech. In my own fieldwork along the south coast of Morobe Province in Papua New Guinea, I found that mutual intelligibility seemed to count for very little in local perceptions of linguistic boundaries. Although linguists relying on lexicostatistical surveys recognized the communalects of Kui, Buso, Lababia, and Kela as "dialects" of a "language" they called Kela, local residents recognized no such cover term uniting all the communalects. For them, I suspect that all isoglosses—whether between dialects or language families—could be equally emblematic, depending on the social context. Much the same could be said of the myriad of isoglosses uniting and dividing the various communities considered to be speakers of the sociolinguistic con-[end page 197]struct labeled "English" (or Tok Inglis, or anglais, or eigo, or ‘ôlelo Pelekane, or whatever, each with somewhat different implications).

T is so concerned with issues of simplification (exoterogeny) and complication (esoterogeny) that he fails to consider questions of language maintenance and language shift apart from such issues. (Grace 1994 deals with the issue of "simplification" at greater length.) In T’s view, people create simplified versions of neighboring languages for use in talking with strangers. T calls this pidginization. Later, they (or their descendants) may make the new language their own, building up the esoteric resources by copying ("borrowing") what they need from neighboring languages and by coining new usages. He calls this indigeniza-tion. It is hard to recast this model of pidginization and indigenization in T&K’s mold of language maintenance and language shift. T stresses the creativity of human actors as they constantly restructure, copy, innovate, and synthesize new linguistic instruments, while T&K stress the overall persistence of linguistic traditions despite the freedom of humans to adopt, abandon, or modify each tradition. However, I think it is possible to reinterpret T’s observations, if not his conclusions, in T&K’s framework of language maintenance and language shift.

In a traditional language-maintenance scenario, the new immigrants to the coast and the earlier arrivals farther inland would each have maintained the essentials of their own language traditions, including their endolexicons, but would have copied ectolexical items from each other, with the coastal people borrowing local forest vocabulary and the forest people borrowing maritime vocabulary, for instance. Under this scenario, current linguistic boundaries would largely coincide with older population boundaries. This seems highly unlikely, given the high degree of exogamy and multilingualism, as well as attested instances of language shift, that are found in small Melanesian communities.

In a language-shift scenario, on the other hand, the new immigrants to the coast might possibly have been only a tiny population. If the immigrants established good relations and inlanders began to learn the immigrant language as a lingua franca for trade purposes, the latter could have learned just the basics (the endolexicon) of the new language and retained their own esoterica for ritual and other purposes. One would have to classify this as a full shift if the endolexicon determines genetic classifications, but as only a partial shift if the entire lexicon is taken into account. This is essentially what happened when small populations of European colonizers arrived on the coast in recent times. In contrast, immigrants to larger metropolitan countries typically do the reverse. After a few generations, they abandon most of their ancestral language, but retain more esoteric items unique to their own heritage, such as terms for foods, social relationships, and rituals. People who compare Swadesh lists in an effort to track prehistoric population movements will not see such ectolexical traces of partially abandoned immigrant or indigenous languages. This may lead them to over-estimate the number of immigrants wherever indigenous populations have shifted to immigrant languages and, conversely, to underestimate the number of immigrants wherever immigrant [end page 198] populations have shifted to indigenous languages. In recent cases, where the immigrant and indigenous populations are often physically distinguishable, the evidence of human physiognomy may easily outweigh the endolexical evidence, but no such counterbalance is available in much of coastal Melanesia.

Another possible counterweight is grammar. People who shift to another language en masse are not likely to leave all of their former phonology, syntax, and semantics behind. Their structural alterations thus give clues to the language(s) they shifted away from. T&K see both grammatical and lexical correspondences as necessary to establish a genetic relationship between two languages. Mismatches are indicative of "abnormal" (nongenetic) transmission. (See Table 2.) In T&K’s framework, therefore, New Guinea languages that have an Austronesian endolexicon but decidedly non-Austronesian grammar (with SOV word order and postpositions, for example) would have to be considered the result of abnormal (nongenetic) transmission. T&K would presumably agree with George Grace in calling such languages "aberrant," that is, "relatively intractable by the comparative method" (1992:115). Perhaps respectable linguists will soon be able to call such languages hybrids again.

Although many linguists have accepted the dogma that languages cannot mix, I do not know any who would claim that cultures cannot mix. If individual languages are cultural constructs, no matter how constrained they are by a universal human capacity for language, then it would be very surprising if languages were somehow more impervious to hybridization than, say, political or economic systems. I wonder whether the idea that languages are reproductively isolated organisms can be traced in part back to the emergence of linguistically defined nationalities that considered themselves separate "races" requiring their own national languages, their own governments, and their own borders. (For very brief overviews, see MacNeill 1986 on the emergence of modern nation-states, Newmeyer 1986 on the emergence of modern linguistics, and Mayr 1991 on the emergence of modern evolutionary thought.) In the "postmodern" era of multinational economies, multinational historiography, and multiculturalism, too many linguists remain blinded by "prepostmodern" visions of idealized homogeneity, the curse of the modern nation-state. But that is a subject for further research.

Postscript: Who makes sound change regular and why? To the neogrammarians, sound change was blind, regular, and exceptionless because, like genetic mutation, it was a gradual, mechanical, natural force that humans did not perceive and thus could not control. Analogy, on the other hand, was artificial interference in the natural process of change that humans undertook in order to clean up the grammatical irregularities left behind. It was regularizing but not exceptionless.

Although the neogrammarians never convinced everyone, even in their heyday, the weight of evidence has by now completely overturned several of their major assumptions. In every recognized language, there are exceptions to many sound changes. (The assumption that all exceptions must result from dialect mix-[end page 199]ture or some other "abnormal" process is circular.) The only question is whether the body of exceptions is large enough to require a separate explanation or small enough to be ignored. Nor is sound change phonetically gradual and imperceptible. (See Ohala 1993.) Since the 1960s, William Labov and other linguists studying change in progress have presented massive evidence of the social motivations for sound change and of social regularities in the distribution of competing variants. (See Weinreich, Labov, and Herzog 1968 for a seminal work aimed at historical linguists.) Innovators in a population adopt new ways of speaking. If they are influential enough, their innovations will gradually spread through the more conservative members of the speech community, after which the community’s language as a system can be said to have changed. (See Milroy 1993, who carefully distinguishes innovation from change and characterizes the innovators as those who have strong ties within their speech community, but many weak ties with members of other speech communities.) Meanwhile, Wil-liam Wang and others studying the lexical distribution of changes in progress have produced substantial evidence that phonetically abrupt innovations spread gradually through the lexicon. (See Wang & Lien 1993 for a recent discussion.)

Speakers are thus often cognizant of innovations, but do not change their speech behavior as soon as they become aware of them. Instead, they may try out the innovations in restricted social, stylistic, or lexical environments until the novelty fades, then use them in more and more general environments. In other words, sound change can be a regularizing but not exceptionless process by which humans modify their speech behavior in order to impose order on the heterogeneity around them. Sound change and analogical change can thus be considered essentially the same process, even though one rearranges phonological categories, while the other rearranges morphological categories (and both can rearrange lexical categories).

After earlier asserting that "sound change is not regular" (56), Thurston offers a different explanation for the "systematic sound correspondences that the neogrammarian hypothesis was designed to account for" (88). Speakers of one language who are shifting to another make "fairly regular phonological interpretations of the target language" (88). In other words, where they exist, regular sound correspondences indicate that language learners in natural settings failed to master the phonology of the target language, producing the same kinds of fairly regular substitutions and creations that have long been observed by classroom language-teachers and by researchers in contrastive and interlanguage phonology. This must be true in many, many cases of borrowing and language shift, but it cannot be the only explanation for regular sound correspondences. In his antidogmatic zeal, Thurston overstates the case for irregularity as much as dogmatic neogrammarians have overstated the case for regularity. The fact remains that surprising regularities exist, and I believe they do so because humans abhor chaos and always seek to impose order on the bewildering heterogeneity that eternally surrounds them. [end page 200]

I am grateful to George Grace, Ken Rehg, and Al Schütz for comments on an earlier draft of this work.