The Unified Language Hypothesis

Today in Bilingualism we talked about one of those things that seems so obvious after you've heard it that you kick yourself for not having thought of it on your own: the Unitary Language Hypothesis.

The question is a simple one: if you have a child growing up in a bilingual environment, and this child is exposed to both languages from day 1, when does the child begin to recognize that it is being exposed to more than one linguistic system?

The way I've phrased the question exposes what I expect would be a natural bias in everyone: we're inclined to think that there is some stage, even if a short one, where the child can't tell the difference at all. As far as it's concerned, both languages are one. It's well-known (and not at all surprising) that a child's first words are all (or mostly) nouns. This makes sense: the child learns to pair meaningful sounds first with concrete objects that are immediately present. At this level, it's hard to see how a child would need to know or even be capable of knowing that the labels it's learning come in two flavors. Thus, the Unitary Language Hypothesis would seem intuitive.

However, the professor doesn't seem to believe in it, and upcoming discussions are to be about some counterevidence.

What kind of things would constitute counterevidence?

Apparently, the Unitary Language Hypothesis makes some testable predictions. First, children apparently don't acquire synonyms until later - well after age 2. So - if the Unitary Language Hypothesis is true, we wouldn't expect a child to acquire words in one language that mean roughly the same thing as words in the other. For example, a child being raised bilingual German-English shouldn't learn both "cookie" and "Keks." For another thing, there should be a correspondence between the amount of time the child is exposed to each language and the percentages that items from each language occupy in its vocabulary. So, if our German-English bilingual hears German 70% of the time and makes no distinction between the two languages, we would expect that 70% of its words would be German, the remaining 30% English.

What's interesting to me is what it would mean to say that a child can distinguish between languages "from the get-go." Surely this can't be literally true. Surely at at least some stage in the development there really is no distinction. I don't see how it could be otherwise without a an extremely far-reaching nativist appeal.

I think what is meant here must be that by the given benchmarks (age 2 in this case) the child has learned to distinguish the languages.

But OK, let's see if we can puzzle this out. Perhaps the ability to distinguish between languages starts from the moment a child begins to grasp the difference between linguistic and non-linguistic sounds (an ability which must be partly innate - at the minimum on the level of a sensitivity to human voices in general, its mother's in particular). The claim would then be that there were several such processes happening in parallel. There might be two ways to look at this: (1) that the child takes a bottom-up approach to phonology, identifying (and repeating) linguistic sounds in small chunks. Not whole words, just chunks of things it heard someone say. As these chunks become stable, it begins to associate them in sequences. It will notice, at this point, that certain of the chunks go with certain other of the chunks, but not all of them. There would come to be two clusters, maybe - them that goes with German, and them that goes with English. (The idea here is that the child is only initially able to pay attention to parts of words, never the whole thing - or at least not able to remember the whole thing.) Another picture might be that the child takes a top-down approach - storing whole utterances that it hears. Of course, it cannot yet repeat these utterances exactly, but it may begin to section them off, to learn that they are made up of separable parts. And again, separating out the parts in this way it will recognize automatically that some such segments do not go with certain others. It forms two sets of combinatorial sounds.

Although there may be some meaningful sense in which these are the same process, the second one sounds more plausible to me. In either case, though, counterintuitive as it may have originally seemed, I think it's easy to imagine how it might happen that a child in a truly bilingual environment could get two languages from the get-go. Before sound-meaning pairings even occur, in other words, the child will have stored all its (pre-)linguistic knowledge in two "boxes." This now seems, indeed, intuitively more likely than the Unitary Language approach.

Of course, this all assumes that Phonology of some rudimentary kind comes before initial learning of vocabulary items. However, this seems likely, as children even in early learning do not repeat the same word in exactly the same way each time. There are variations in pitch, loudness, tone, etc. that seem to indicate that the pattern is stable on some abstract level.

The two predictions mentioned earlier still don't necessarily follow for me, though. In particular, the synonym prediction does not. Just because a child has separate systems for linguistic storage, it doesn't seem to automatically follow that it will store separate words for the same item. All the child really is aware of at this point is that there are two classes of building blocks for communicative sounds. That fact alone doesn't seem strong enough to carry the conclusion (to it) that it might need separate sound-meaning pairings for the same meaning for each of these classes. To take the example of "cookie" and "Keks" again - a child exposed to "Keks" first might just assume that this item happens to be named in the "German" class. Or - put more carefully - it seems that the phenomenon that children at age 2 don't form synonyms (or at least not many) is that it doesn't yet see the need. If "Keks" gets it what it wants, there is no need to acquire "cookie." The fact that it has two groupings of linguistic building blocks and can correctly determine that "Keks" belongs to the "German" one has no bearing on this. It simply stores "Keks" and gets the cookies it needs, mission accomplished.

Now, one might object that in situations where it has a parent or grandparent who simply doesn't understand German it would also learn (bother to store) "cookie." And that's fine, as far as it goes. That would indeed show that the children that exhibit these translational equivalents (prior to having acquired synonyms in either of the two languages - a crucial point) do not have a Unified Language system. However, it doesn't cover all cases. It seems that it would be possible that some such children (children without a Unified Language) would show no translational equivalents at all. If all the people who communicate with it are able to understand both languages (at least to the limited extent that they understand what the child says), then a child that, in fact, has two language groupings in its head might not bother to learn any translational equivalents - might not exhibit the predicted behavior.

The interesting question is whether such a child really does, in fact, have two languages. If phonological development proceeds the way I speculated, then I suppose it would. However, it's equally possible that phonological development is still far from complete at this stage (and I don't know, this is my first real acquisition class). A child that hasn't had a need to develop translational equivalents might could be said to be slower to grasp, then, that the sound units it is learning to identify as meaningful belong to two groups. Such a child might simply have a collection of them and only the vague notion that there are separate groups. Since it has never had to learn two words for the same item, the notion that these systems are separate is less important.

That said, the child that does have to learn translational equivalents by this stage constitutes an existence proof that the Unified Language Hypothesis cannot be true in all cases. I'm not convinced, however, that there are not some children for whom it is true. It seems plausible that the specifics of the learning environment will play a role in determining whether a child goes through a stage in which it has a Unified Language, that some children will and others will not.

In any event, the topic is fascinating. It's not something I had ever really thought about in detail, but now that it's come up in class, it seems like the kind of thing that should have occured to me before now. I'll be interested to see what the studies we're about to read show. In principle, I have no problem with the idea that language learning can be different in the ways outlined for bilinguals, however.