Comments Welcome
George W. Grace
University of Hawaii

Ethnolinguistic Notes

Series 4, Number 11

THE GENETIC HYPOTHESIS

2. THE CONCEPT OF LANGUAGE "MIXING"

"Mixing" scenarios
"Lexification" and "content form" (again)
The emblematic role of lexification
Lexification as the criterion of ancestry
More scenarios

QUESTION: We ended the last Note (Grace 1998) with this question: Under what conditions does normal continuation from one point in time to a later one fail to occur? (By "normal continuation", of course, we mean continuity between the KOLs of the people who count as speakers of a particular "language" at one particular time and either the KOLs of their descendants or their own KOLs at a particular later time.)

ANSWER: Well for one thing, the conditions for such failure would apparently have to include the intervention of another language or dialect, as we saw earlier (in Grace 1998).

QUESTION: Maybe you should explain again what it was that we saw there.

ANSWER: All right. I agree that at first blush it would seem that there are three possible kinds of ancestries for any given language (that is, for what's represented by any given language-state description): (1) The language is a normal continuation of a previously-existing language, (2) it's a continuation of more than one such language (i.e., has mixed ancestry), and (3) it's a continuation of no earlier language(s). However, what we said there was that the third of these has generally not been regarded as a realistic possibility at all.

QUESTION: All right, if we rule out possibility (3), we're left with only one kind of possible failure of what we've called normal continuity, namely that the language in question is a continuation of more than one language--that is, that it has mixed ancestry.

ANSWER: Yes, although some linguists who recognize the existence of mixed languages might prefer not to use a word like "continuation" for their relation to the languages that supplied the components in the mixture.

QUESTION: All right. Now what is actually assumed to happen in these cases? How are languages supposed to go about mixing?

"Mixing" scenarios

ANSWER: That's not a simple question to answer. A good place to begin would probably be with Thomason and Kaufman 1988 (henceforth, TK). It's the most influential work on the subject in recent years.

QUESTION: And what do they say?

ANSWER: They say (TK:8) that "genetic relationship in the traditional sense of one parent per language can only be posited when systematic correspondences can be found in all linguistic subsystems--vocabulary, phonology, morphology, and (we would add) syntax as well".

QUESTION: Can you explain what's meant by "systematic correspondences" in such "subsystems" as morphology and syntax?

ANSWER: Well, I'm not aware of any concept of systematic morphological or syntactic correspondences that would be at all comparable to the concept of regular sound correspondences (unless they themselves had regular sound correspondences as their basis). However, I don't think the general idea is hard to grasp: the inflectional systems of the older Indo-European languages would probably be the most representative example of what they have in mind.

QUESTION: How is the mixing in question supposed to come about?

ANSWER: According to them the kinds of mixing that result in languages to which the genetic hypothesis isn't applicable come about in two principal ways: interference through shift (more familiarly known as "substratum" influence) and borrowing. Both "have as their most extreme outcome the emergence of a language whose lexicon is not from the same source as the bulk of its grammar." (TK: 48).

QUESTION: You'll need to explain how each of these actually works. Why don't you start with how interference through shift results in a mixed language.

ANSWER: I'll quote from TK:

"If a whole population acquires a new language within possibly as little as a single lifetime, therefore necessarily other than by parental or peer-group enculturation, the linguistic system which results may have massive interference from the structure(s) of the language(s) originally spoken by the group. If this population is not integrated into the group that provided it with a new language, this deviant form of speech may crystallize into a new language." (TK: 10).

QUESTION: I'll try to formulate that in a way that accords better with the view of language that we're developing here. A population comes to be exposed to an exotic set of role models who (the role models) have themselves shared a common set of role models. (In more familiar terms, the population is exposed to a set of role models who speak the same language, and that language is different from the language or languages previously spoken by the population in question). Furthermore, the conditions under which this exposure occurs are such as to coerce the population in question into using the new language.

The members of this population therefore hastily add a new component to their KOLs, basing it as far as possible on the examples provided by the new role models, but also (and perhaps increasingly) on examples provided by their peers--fellow learners.

The conditions that require the use of the new language persist into the next generation, but the principal role models available to this new generation are the second-language learners of the preceding generation (and their peers).

ANSWER: That sounds about right.

QUESTION: I think we need to examine this in more detail. We begin with what we'll call Population A, who speak Language a. Then, they encounter speakers of Language b under circumstances that oblige them to learn and use b. They add a b-component to their KOLs as best they can, but they're obliged to use the language before they've been able to build up an adequate memory-store of experiences involving b utterances. Therefore, they're often obliged to resort to calquing a expressions in b (as we discussed in Grace 1997b). Contact with B people is limited, and therefore, as A people add experiences involving b utterances to their memory stores, a large proportion of the additions involve utterances made by A people which are calques of a expressions.

As time goes on, the members of Population A continue to speak b (and perhaps abandon a entirely). However, they continue to have only limited contact (or perhaps eventually, no contact at all) with members of Population B. Therefore, the examples of b that children learning the language have available to them are heavily influenced by calquing from a.

If Population B continues to exist and to speak b, its version of b (Call it Bb) presents (as far as this account is concerned) no problem for the genetic hypothesis. However, Ab (which, if Ab is to constitute an example of a mixed language, must be different enough from Bb to count as a distinct language) shows heavy interference through shift. A conventional linguistic description of Ab would show a grammar with many affinities to that of a. Therefore, it's a mixed language.

ANSWER: Again, that sounds to me like an accurate account.

QUESTION: O.K., that's how interference through shift can lead to a mixed language. Now, how can a mixed language arise through borrowing?

ANSWER: Again I'll quote from TK:

"Or, to take another extreme possibility, a population may come under such heavy cultural (including sociopolitical and economic) pressure from another group that the entire pressured population becomes bilingual in the dominant group's language. The bilingual population may then actually shift completely to the second language, while retaining the lexicon of the original language for use, with the other language's grammar, as an ethnic-group code or jargon. Children born to members of that group might then begin to learn, not the whole original ethnic-group language, but only its vocabulary, in addition to the entire language of the dominant group." (TK: 10).

QUESTION: What would be an example of that?

ANSWER: The example that they particularly refer to is AngloRomani, and if I interpret rightly, Ma'a is considered to be another example.

QUESTION: Now if we try to reformulate that in terms of speakers' KOLs, we'd have population A encountering B and learning their language, b. But then what?

ANSWER: The only scenario that I imagine would go something like this: the A community's knowledge of their original language begins to be lost. At one stage, they still retain a repertoire of useful utterances which prove to be quite helpful when they want to avoid being understood by members of the B community (and possibly also by their children). Subsequent generations know even less of a. They've probably managed to pick up very few complete utterances, but in particular they've managed to acquire enough vocabulary to produce a kind of relexified b that they use as a secret language. It's this relexified b that's counted as the mixed language.

That's my best guess as to the scenario, although I can't guarantee that Thomason and Kaufman would find it acceptable.

QUESTION: Let me attempt to review. In the case of interference through shift, the lexicon of the mixed language, Ab, has b lexicon, but the bulk of its grammar comes from a. In the case where borrowing is the cause of the mixture, the mixed language is Aa, which has a lexicon, but the bulk (or even all) of its grammar is from b. Is that right?

ANSWER: That's the way I understand it.

QUESTION: Do you find their scenarios satisfactory?

ANSWER: Not really. I'm not happy with the talk about sources of such things as "lexicons" and "grammars".

QUESTION: Why?

ANSWER: I think it distorts the facts to conform to the terms of conventional linguistic description.

QUESTION: Explain what you mean.

ANSWER: You may remember that our 1981 book (Grace 1981) dealt with this kind of language mixing (if that's what it's to be called) in some detail. You may remember that we had a chapter entitled, "That a language must not be assumed to have just the characteristics of its linguistic description". The point was, of course, that any description (linguistic or other) must report only selected facts about its object, and that the basis on which these facts are selected is their perceived relevance for particular purposes--relevance from a particular point of view. Therefore, no description can be adequate for all purposes. Furthermore it's obvious--and that was the main immediate point--that the purposes (whatever they may have been) for which the standard grammar-lexicon (or syntax-lexicon) model of language description was designed didn't include that of elucidating the phenomena that result from so-called "language contact".

QUESTION: Nor (as we hope has been shown in Grace 1995, 1997a, and various other places) does it appear to be at all well-designed for discussions of the knowledge of language possessed by speakers, even though it's that that most of us assume to be what we're ultimately talking about when we refer to "language".

ANSWER: That's very true. And it's really also very relevant. However, what we're particularly concerned with in the present discussion are the phenomena that result from "language contact", and I think it'll probably be best to focus pretty narrowly on them.

QUESTION: All right, what's to be said about the phenomena that result from language contact?

ANSWER: Well for one thing, the very next chapter in Grace 1981 (a chapter entitled, "…the dotted line") attempted to show that in a wide variety of cases where languages were purported to have their lexicon and grammar partly or entirely from different sources, the components involved weren't lexicon and grammar in the usual sense of these terms--in fact they weren't accurately identifiable with any of the components of a standard linguistic description.

"Lexification" and "content form" (again)

QUESTION: I think you'd better remind us what we concluded that the components involved in fact were.

ANSWER: We called them "lexification" and "content form".

QUESTION: You probably need to remind us further of what those terms mean.

ANSWER: The "lexification" of a language (or of a word, or any linguistic expression) is its signans (or signantia). We chose the term "lexification" because the word "relexification" was already in use to refer to the replacement of one lexification with another.

The term "content form" comes from Louis Hjelmslev (see Hjelmslev 1963: 50ff). The content form of a language as a whole would include everything reported in its dictionary except the spelling and pronunciation of the words, plus everything reported in its grammatical description except the spelling and pronunciation of grammatical morphemes, plus something that isn't dealt with in any systematic way in standard linguistic descriptions--what we called (in Grace 1981) its "idiomatology".(1)

QUESTION: What is "idiomatology"?

ANSWER: We might describe it as what accounts for what Pawley and Syder (1983) call "nativelike selection".

QUESTION: Then, are you saying that when the contributions from the two sources are different, the two will sort out, respectively, into what we've been calling "lexification" and "content form"?

ANSWER: I don't want to make any such universal claim. I certainly don't pretend to understand what's going on in all of the mixing situations that have been reported (including several discussed in TK).

That said, it does appear that the tendency for lexification and content form to develop differently in situations of contact between unequal communities is very widespread--I would almost say that it's expectable.

QUESTION: What do you mean by "unequal communities"?

ANSWER: I mean that the community of speakers of one of the languages brings more power or prestige to the situation than does the other one--more generally, I have in mind any situation where the functions performed by the two languages aren't essentially interchangeable. In each of the two scenarios that we've just discussed the community speaking one of the languages obviously had greater clout than the other.

QUESTION: Why do you suppose lexification and content form would tend to develop differently in such situations?

ANSWER: Well, the two seem to present themselves to us differently. Lexification is what's most easily graspable by one trying to learn a language. The lexification may be thought of as that which identifies what language one is speaking (or attempting to speak). (The content form, on the other hand, reflects the linguistic experience that one brings to the task).

QUESTION: Let me review. [Recall that the designation "strong form of the genetic hypothesis" is understood to refer to the hypothesis that there are no mixed languages--the hypothesis that any language must be a continuation of a single language of any particular prior period].

First of all we've said that the only cases where this strong form of the hypothesis has seriously been challenged concern languages which (according to the claim) are simultaneously continuations of more than one language--so-called "mixed languages".

Secondly, we've seen that the putative mixing that's received the greatest attention is the kind in which some linguistic "subsystems" could be described as coming from one source and others from another.

ANSWER: Yes, where the contributions of the two source languages or groups of languages differ in such a way as to be reported in different parts of the linguistic description of the resulting language.

QUESTION: Thirdly, we've said that where the contributions from the two sources are different, it's very usual for the two to sort out, respectively, into what we've been calling "lexification" and "content form".

And finally, we've said that the lexification seems generally to be seen as emblematic of what language is being spoken.

ANSWER: That seems a fair summary.

The emblematic role of lexification

QUESTION: Then, wouldn't there be good grounds for claiming that the two hypothetical languages in the TK scenarios are in fact not mixed?

ANSWER: What do you mean?

QUESTION: The putatively mixed language in the first case has b-lexification. Furthermore, the case is labeled "interference through shift". The use of the word "shift" makes it clear that the speakers are seen as having shifted to b--that it's b that they're assumed to be speaking, in however substandard a form.

ANSWER: That would seem to be true--at least up until the moment that the "deviant form of speech" "crystallizes into a new language" (to quote from TK:10).

QUESTION: The question is: What would distinguish the "new language" from the deviant form of b? By what sign could an observer on the scene tell when the crystallization had occurred? It seems to be the speakers' intention to be speaking b, it seems to be what they think they're doing. Who's to say that's not what they are doing even though a linguist (or a member of the B community) may think they don't do it very well? In short, what's to keep me from arguing that what we've called Ab is nothing other than a (deviant) form of b?

ANSWER: There'd certainly be nothing unique about such an argument.

QUESTION: The "mixed" language in the second case has a-lexification. And this case is labeled "borrowing", which suggests that a (as we've labeled it) is continuing--that a is the language that's being borrowed into. Furthermore, TK's (p. 10) description of the process refers to their "retaining" the lexicon of the original language--which again suggests continuity. It seems likely that what the speakers are in fact doing is speaking a--as they understand it.

ANSWER: That may well be right.

QUESTION: Now, the question that I want to ask is how those historical linguists who claim that the so-called "comparative method" provides the only reliable evidence of genetic relationship can fail to acknowledge lexification as criterial for (one might even say as "emblematic of") linguistic ancestry.

ANSWER: What do you mean?

Lexification as the criterion of ancestry

QUESTION: What I have in mind is that the comparative method can only be applied where there are regular sound correspondences. Is that right?

ANSWER: Yes.

QUESTION: And the comparative method is a method for "reconstructing" the language ancestral to the languages being compared?

ANSWER: Yes.

QUESTION: However, aside from the proto-phoneme system, of course, all that can be reconstructed by this method is the (partial) lexification of the proto-language. Am I right?

ANSWER: Well, of course historical linguists do "reconstruct" more of the proto-languages than that.

QUESTION: But that isn't done by means of the comparative method (in any strict sense), is it?

ANSWER: True enough. However, I should also point out that what you're calling reconstructed "lexification" is really vocabulary with associated meanings.

QUESTION: Yes. However, most of the meanings associated with the lexifications are quite general--closer to what Hjelmslev 1963 would call "purport" than to "content form". Furthermore, they themselves aren't reconstructed by the comparative method.

ANSWER: I suppose that's right.

QUESTION: All right. The point that I was trying to make was that it's inconsistent for linguists to take the position that the comparative method provides the only reliable evidence of genetic relationship, and not acknowledge that lexification--which is what the comparative method compares--is criterial for linguistic ancestry. That is, their appeal to the comparative method seems to say that the kind of evidence that (as we've already agreed) is relevant to deciding what language is being spoken (i.e., lexification) is also the kind that's relevant to deciding what language is being continued by that language (and that would seem eminently sensible since the two are pretty much the same question).

ANSWER: And supposing that point is granted, what then?

QUESTION: Well, it would seem from that point that a language that's described as having its "lexicon" (which, you've suggested, in most cases really means its lexification) from one source and its "grammar" (which, you've suggested, in most cases means its content form) from another doesn't have mixed ancestry at all--that according to the criteria we've just defined, it can only be regarded as a continuation of the language that's the source of its lexification (what's been called the "lexifier" language).

ANSWER: And exactly that solution has been proposed by some linguists. However, I think there's more that should be said on the subject of scenarios for language mixing before we go further with the question of what criteria should be decisive in questions of genetic classification.

QUESTION: All right. Say it.

More scenarios

ANSWER: First of all, I'm not sure we've interpreted TK's idea of the "borrowing" process correctly.

QUESTION: You're talking about the process by which borrowing leads to a language that's mixed in such a way as not to "fit within the genetic model"--that is, as not to qualify as a continuation of any single language?

ANSWER: Yes.

QUESTION: What's the problem?

ANSWER: It's the part played by lexical borrowing. They say on the one hand (TK: 37), "Invariably, in a borrowing situation the first foreign elements to enter the borrowing language are words." They go on to say (TK: 38) that in cases of moderate to heavy borrowing, phonological and syntactic interference precede at a comparable pace, with morphological interference lagging a bit behind those two.

However on the other hand, they say (on TK: 7) that AngloRomani has a Romani lexicon and a "completely English grammar" and that the entire grammar has been borrowed from English. This seems to be saying that the borrowing did not begin with word borrowing--that in fact words (which I think we can take to mean lexifications) were precisely what were not borrowed.

Furthermore, this interpretation of what happened in the AngloRomani case seems to agree with the description of the borrowing scenario that we quoted originally. To review, what was emphasized in that scenario was a population coming under such pressure from another group that it became bilingual in the dominant group's language, and subsequently actually shifted "completely to the second language, while retaining the lexicon of the original language for use, with the other language's grammar, as an ethnic-group code or jargon. Children born to members of that group might then begin to learn, not the whole original ethnic-group language, but only its vocabulary, in addition to the entire language of the dominant group." (the quoted material is from TK: 10. Italicization is added.)

QUESTION: In sum, if cases of heavy borrowing always begin with lexical borrowing, how can it be possible for the borrowing language to wind up with a borrowed grammar and a retained lexicon?

ANSWER: That's what I haven't succeeded in understanding. What I'd like to do next is turn to a discussion by Malcolm Ross that may further our understanding. Ross illustrates both processes, borrowing and interference through shift, with the primary concern of showing how prior historical processes can be inferred from resulting language states. However, I think it'll be better not to start that in this Note.