Comments Welcome
George W. Grace
University of Hawaii

Ethnolinguistic Notes

Series 4, Number 10

THE GENETIC HYPOTHESIS

1. THE NOTION OF CONTINUITY OF A LANGUAGE OVER TIME

The fictional status of "languages"
The genetic hypothesis
Continuity across generations

QUESTION: Some years ago, you wrote (Grace 1965:1), "A fundamental assumption is that any language spoken today represents a point in a tradition which is continuous in the past back to the (or an) origin of language." Is that "assumption" what people have in mind when they speak of "the genetic hypothesis"?

ANSWER: Yes, that's essentially what the hypothesis is.

QUESTION: There are some points about the precise content of the hypothesis that aren't clear to me, but there's a more fundamental problem that makes me wonder if it's worth even considering them.

ANSWER: What do you mean?

The fictional status of "languages"

QUESTION: I mean that we've been saying that languages are fictional constructs, that what we call a language is in reality nothing but a similarity in the KOLs (sc. knowledge of language) of those people who count as its speakers. Why doesn't that simply mean that the genetic hypothesis is meaningless?

ANSWER: The fact that what we call a "language" is only a similarity among KOLs will certainly be one of the main problems that we'll be grappling with here.

However, we want to make every effort possible not to throw the baby out with the bath water. We certainly don't want to discard the impressive accomplishments that have resulted from the assumption that the hypothesis does have meaning. The question is how to go on from here--how the results already achieved by means of the genetic hypothesis should be interpreted, and what questions we should seek to answer in the future.

We should think of the concept of individual languages as a simplifying assumption that serves a number of purposes. It serves, in particular, to simplify discussion of matters dealing with the fact that Person A may speak in a way which is very similar to that of Person B, and very dissimilar to that of Person C.(1)

Moreover, since the existence of languages is taken for granted in the literature of linguistics as well as by the general public of the Western world, and since many specific "languages" are recognized and known by name, it's virtually impossible to talk about language at any length without tacitly accepting the notion that it consists of individual languages.

I might also remind you that, as we've noted before, many of these "languages" appear to be quite well defined.

QUESTION: Explain "well defined".

ANSWER: By saying that they're "well defined" I mean that the KOLs of all of the people who count as their speakers are very similar to one another, while none is very similar to the KOL of anyone who doesn't count as a speaker.

The genetic hypothesis

QUESTION: All right, but it's also clear from what you say that there are people who don't speak a well-defined language. We'll need to remember that ultimately these can't be ignored.

However, first I want to get back to a more precise understanding of the content of the genetic hypothesis. We were using as our definition your statement: "A fundamental assumption is that any language spoken today represents a point in a tradition which is continuous in the past back to the (or an) origin of language."

It seems to me that that actually entails two separate propositions.

ANSWER: What do you mean?

QUESTION: Doesn't it imply (1) that every language is a continuation of earlier languages, and (2) that any language has had only one such ancestral language at any given point in time?

ANSWER: Yes, the definition we're using does appear to have both of those implications. We might call the hypothesis defined so as to include those two propositions "the strong form of the genetic hypothesis".

QUESTION: What other forms are there?

ANSWER: Well, some linguists have proposed the possibility of a concept of "mixed" language--that is, a language that simultaneously continues more than one language from the same point in time. However, I don't know of any mixed language hypothesis that's been formulated precisely enough to be usable.(2)

Other linguists have simply taken the position that there are languages that have arisen in such a way that the genetic hypothesis can't be usefully applied to them.

QUESTION: Doesn't that mean that they're rejecting the first proposition as well--that is, rejecting the hypothesis in toto?

ANSWER: I don't think it need go that far. At least, I don't know anyone who claims that languages with no linguistic antecedents at all--languages made up from scratch with nothing taken from any previously existing languages--can exist.

QUESTION: Why not? Aren't there languages here and there in the world that have no generally-recognized genetic relatives? Why aren't these "language isolates" counted as not continuing any prior language(s)?

ANSWER: Well, the general assumption always seems to be that they doubtless do in fact have relatives, but that insufficient evidence of the relationship has survived--presumably because the relationship with these relatives is so distant (i.e., the period in which they've been separated from them has been so long).

QUESTION: And you consider this assumption valid?

ANSWER: I think the assumption that there could be no full-blown language without linguistic antecedents must be true. Every normal language represents a rich and detailed conceptual world (in the sense of Grace 1987) that couldn't have been assembled from scratch in a short time.

QUESTION: All right. I understand you to be suggesting two views that we might call "weak forms of the genetic hypothesis". Both would accept the assumption that a language may be the continuation of an earlier language. In fact, both would assume that every language must owe something to previously-existing languages, but that there are cases in which no one such language at any particular point in the past could be singled out as the sole ancestor. They would differ in that one would propose that the languages in these cases (except the cases of "language isolates") be classified as of mixed ancestry while the other would propose that they be regarded as special cases to which the genetic hypothesis can't usefully be applied.

ANSWER: That seems an adequate summary, except that it suggests a sharper division into two camps than actually exists. Think of the two "views" as ideal positions rather than as being representative of any particular real linguists.

Linguistic continuity across generations

QUESTION: We probably should leave the question of language mixing until after we've considered the problem of the continuation of languages over time, and to keep the problem as simple as possible, let's begin by considering only the "well-defined" languages.

ANSWER: All right. What is the problem?

QUESTION: The problem that I see is this: We've said that a language is in reality nothing but a similarity in the KOLs of those people who count as its speakers, and yet humans, and therefore the "speakers" of any given "language", have only a limited life span. Nevertheless, we assume languages to persist over millennia. How can that be?

ANSWER: It's assumed that the members of each successive generation wind up qualifying as speakers of the same languages as the members of their parents' generation. Recall that we said (in Grace 1997) that two individuals whose KOLs are sufficiently similar will be counted as speakers of the same language. This principle obviously can be applied to individuals of different generations as well as to those of the same generation.

QUESTION: Then, an assumption required by the genetic hypothesis is that the members of what we may call Generation G-1 acquire KOLs that are similar in the required manner and degree [whatever the required manner and degree might be determined to be] to those of the members of Generation G?

ANSWER: Yes, although we should mention somewhere in this discussion that a community as a whole isn't normally stratified into generations. Therefore, while your Generations G, G-1, and so on, may be well defined in the genealogy of a particular family, they normally won't be for a larger community. However, if we keep that fact in mind, there seems to be no harm in continuing to talk about generations as you've begun.

QUESTION: All right, I'll proceed with caution. Now what about Generations G-2, G-3, and so on to G-n? Does the genetic hypothesis assume that their KOLs also would be similar in the required degree to those of Generation G?

ANSWER: Not necessarily. In fact, a language as it's continued over a series of generations is a very close diachronic analog of a (synchronic) dialect chain such as we described in Grace 1997). There's no limit--not even a theoretical limit--to how different the KOLs at the ends of such a chain may be from each other.

QUESTION: All right, the genetic hypothesis requires that the members of each successive generation wind up with KOLs sufficiently similar to those of their elders to qualify them as speakers of the same language. What are we assuming if we assume that to be the case--that is, what conditions would be required in order to produce this result?

ANSWER: Well, we said before (in Grace 1997) that for two speakers to acquire sufficiently similar KOLs it would presumably be necessary for them to share (largely) the same role models (or for their role models to share the same role models, etc.). For clarity in the situation we're now considering, we might add that this should be interpreted to include one being the role model of the other.

QUESTION: Wouldn't it be the normal thing for successive generations to share essentially the same role models? Shouldn't we expect something like the following?

1. The members of Generation G, during their initial process of language learning, have certain people (call them "the original elders") as their role models.

2. By the time Generation G people begin having children, some of the original elders have died, and some additional individuals--many of them younger people who themselves grew up with the original elders as role models--have been added to the set of role models of Generation G people.

3. When the children--Generation G-1--begin language learning, their role models are essentially the same as the set of role models of their parents as that set has now evolved.

ANSWER: The same set of role models with the addition of other people who have that set as their role models. That is, I assume we'll want to allow the parents to serve as role models for their children, for example.

QUESTION: Yes. Shall we say "essentially the same set of role models together with their successors"?

ANSWER: That sounds all right. It seems to allow also for the children's peers--other children--to serve as role models, and they do. (These peer models apparently provide somewhat greater possibilities for innovation at the time of first-language learning than might otherwise be expected from your account).

I'd say that to the extent that the statement is intended to apply only to communities in which multilingualism (or multidialectalism) isn't a significant factor, it's a good description of the way we'd expect what Thomason and Kaufman (1988) call "normal transmission" (which, according to them, is required for the existence of genetic relationship) to work.

QUESTION: If I understand right, we're saying that where we have "normal transmission" (as we've just described it) of a "well-defined language" (as we've defined that), it seems quite reasonable to think of what's happened as "the continuation of the language". In short, it seems justified to think of the genetic hypothesis as having proved successful in such a case.

ANSWER: Yes.

QUESTION: Now what about cases where the "language" is not "well-defined" or where the transmission is not "normal"? What kinds of such cases are there?

ANSWER: Let's dispose of "well-definedness" first. We defined it as depending on two conditions: (1) that the KOLs of all of the people to be counted as speakers of the language be very similar to one another, and (2) that none of them be very similar to the KOL of anyone who doesn't count as a speaker.

QUESTION: Therefore, there are two ways for a "language" to fail to qualify as well-defined: the KOLs of some of its speakers being insufficiently similar to one another, or the KOLs of some of its speakers being too similar to the KOLs of some people who don't count as speakers?

ANSWER: Yes.

QUESTION: Under what conditions are non-well-defined "languages" found?

ANSWER: You're asking about cases where what some persons are speaking--that is, what would ordinarily be referred to as "the language" that they're speaking--is not a well-defined language. But cases of that kind are so common that I wouldn't want even to try to identify all of the different sets of circumstances in which they occur. However, I can illustrate some of the considerations that arise by using dialect chains as an example.

QUESTION: Which of the two conditions for well-definedness does a dialect chain violate?

ANSWER: That depends on the perspective we choose; it can be seen as violating either. Since there's no non-arbitrary way of dividing such a chain up into languages, we have two alternatives: (1) considering the whole chain to be a single dialectally-diverse language, or (2) dividing it up into several languages by drawing (at least partially) arbitrary boundaries between sets of speakers with very similar (i.e., too similar by whatever criteria we're using) KOLs (for example, drawing a boundary on the map that falls between two villages or two islands with very similar dialects).

The first alternative would violate the requirement that the KOLs of all speakers be very similar (since those from the ends of the chain might be extremely dissimilar), while the second would violate the requirement that none of the KOLs of people who count as speakers of a particular language can be very similar to the KOL of anyone who doesn't count as a speaker (because the boundaries would necessarily have to be drawn between some people with quite similar KOLs).

QUESTION: Are you going to say that the genetic hypothesis is not valid except for well-defined languages?

ANSWER: Not necessarily. Consider, for example, a locality in which all of the KOLs of the population are similar to one another but not sufficiently different from those of any other locality to make the language/dialect qualify as well-defined. There's no reason in theory why Generation G-1 in such a locality (and I intend here for "locality" to serve as a cover term for the smallest linguistically-relevant unit: the speech of a single village, a single island, or whatever it may be) can't have the same role models as the Generation G of the same locality. For example, normal transmission could be said to have occurred in such a locality no less than in a well-defined language.

QUESTION: Then the necessity for dealing with non-well-defined "languages" presents no problem for the genetic hypothesis?

ANSWER: The absence of well-defined languages in an area doesn't mean that genetic continuity (e.g., what we're calling "normal transmission") can't occur at all. However, it does result in some practical problems.

QUESTION: What kind of problems?

ANSWER: For one thing, where the "language" whose continuity we're positing isn't intrinsically well defined, there's the problem of being sure that our (in part arbitrary) definitions of its earlier and later versions are equivalent--for instance, that the pre-change and post-change language boundaries are drawn at the same places ("place" may not be the right word; what concerns us is to ensure continuity of role models). Note, however, that often it simply isn't clear what the boundaries of the "language" whose synchronic state was represented by a particular description or reconstruction is or was.

QUESTION: Explain.

ANSWER: I mean it often isn't clear exactly which varieties a linguistic description is supposed to account for and which it isn't.

This problem is more serious, of course, when the "language" in question is more extensive than the smallest linguistically-relevant unit.

QUESTION: Can you expand on the point about the problem of getting the boundaries drawn "at the same place"?

ANSWER: What I mean is that if there were linguists working in the area before the period in which the changes occurred, they would likely have been confronted with the task of making arbitrary decisions about how many languages there were and where the boundaries of each were. These decisions would then have led to the presumption of that number of "language states" represented--or in principle representable--by linguistic descriptions.

Of course, if there were no linguistic descriptions from that time, the assumed "language states" of the time would be those identified by means of subsequent reconstruction.

For the time following the period of change, they'd be confronted with the same kind of decisions. As long as they made their decisions in an exactly parallel way (i.e., drew their boundaries in the same places--for example, on the map), the resulting language states could (all other things being equal) as validly be considered to be continuations of the respective antecedent states as in the case of well-defined languages that we discussed above.

QUESTION: All right, being sure that the "languages" identified with the "before" and "after" états de langue are equivalent entities is one problem that's intensified when the "languages" aren't well-defined. You implied that there were others.

ANSWER: I just had one other problem in mind. It's that situations of non-well-definedness generally are such as to favor "contact" between dialects; that is, to make a more varied assortment of potential role models available.

However, in principle the absence of well-defined languages in itself doesn't seem to make the genetic hypothesis inapplicable (and I believe that will prove to be true generally, not just in the particular case of dialect chains).(3)

QUESTION: All right, we can apparently conclude that the absence of well-defined languages in a particular case doesn't necessarily mean that the genetic hypothesis can't validly be applied there. The other problem that we were going to consider was the failure of "normal transmission". Under what circumstances do linguistic changes that occur from one generation to the next fail to qualify as normal transmission of "a language" (or two or more "languages")?

ANSWER: Before we go on, let me say that I don't think we should assume that changes that might threaten the hypothesis of genetic continuity necessarily occur in transmission between generations.

QUESTION: I presume that what we mean by "transmission between generations" is the child's initial language learning (what--unfortunately--is frequently referred to as "acquisition"). Is that right?

ANSWER: Yes. The term "normal transmission" seems to presuppose that any discontinuity should be sought only between parents and children (and note our warning above that communities as a whole aren't normally stratified into generations).

QUESTION: What do you have in mind as alternatives?

ANSWER: I don't think we should rule out a priori the possibility of genetic discontinuity resulting from changes occurring within (the KOLs of) individual speakers in the course of their own lifetimes. And, on the other hand, I'd think we might want to give similar consideration to changes that for one reason or another could only be documented as occurring between speakers who were more than a single generation apart (as when we compare descriptive data from two separate periods that were more than a generation apart in time).

Why don't we just talk about "normal continuation" or "normal continuity" rather than "normal transmission"?

QUESTION: All right, then, under what conditions does normal continuation from one point in time to a later one (by which I presume we mean continuity between the KOLs of the people in a locality such as we've talked about at one particular time and their own KOLs or those of their descendants at a particular later time) fail to occur?

ANSWER: Our attempt to answer that will require a fair bit of discussion. Since we're trying to keep these Notes relatively short, let's leave that topic for the next Note in the series.