On "universal truths" -- part 2

In a recent posting, I discussed the research by Michael Dunn and his colleagues that questions there being any universals about human language. Another recent study, this one conducted by Jennifer Culbertson, who worked as a doctoral student in Johns Hopkins' Krieger School of Arts and Sciences under the guidance of professors Geraldine Legendre and Paul Smolensky from the Department of Cognitive Science at Johns Hopkins, argues exactly the opposite. According to Jennifer Culbertson,

"this research shows clearly that learners are not blank slates; rather, their inherent biases, or preferences, influence what they will learn. Understanding how language is acquired is really the holy grail in linguistics."

While I applaud this attempt to show evidence for innate universals of human language, this study is not without problems either. But before we get to the problems, a quick description of the study itself.

The study was designed as an alien-language-learning experiment, with a small, green, cartoonish "alien informant" named Glermi teaching participants, all of them English-speaking adults, an artificial "language" named Verblog via a video game interface. In one experiment, for instance, Glermi displayed an unusual-looking blue alien object called a "slergena" on the screen and instructed the participants to say "geej slergena," which in Verblog means "blue slergena." Then participants saw three of those objects on the screen and were instructed to say "slergena glawb," which means "slergenas three."

While both adjective-noun and noun-numeral orders are found in many languages, the combination of both is not very common cross-linguistically (as it turns out, only about 4% of the world's language exhibit the Verblog word order pattern). As a control, other groups were taught different made-up languages that matched Verblog in every way but used word order combinations that are commonly found in human languages. As you would expect, the adult learners who had had little to no exposure to languages with word orders different from those in English quite easily learned the artificial languages that had word orders commonly found in the world's languages but failed to learn Verblog. Good thing they were not children trying to acquire Hausa or Paiwan as their native language, but more on this below.

Now, to problem one: was the preference exhibited by the experimental subjects specific to our language faculty, or did they employ analogical thinking common to other cognitive tasks? To me it seems that nothing in this study shows that we are dealing with a language-specific property.

Secondly, is it evidence for innateness? Yes and no. At best, what we have here is indirect evidence that linguistic universals are innate. The logic of the argument is this: if indeed what has been shown here is a language-faculty-specific tendency for analoguous word orders (rather than a more general cognitive property), speakers must have somehow "known" that the Verblog word order was extremely unlikely (note that this knowledge is implicit, rather than explicit; we know it without knowing that we know it). But how would they come to possess this "knowledge"? It can't be deduced from taxonomic data (the same way a linguist may confirm the cross-linguistic validity of this observation about word orders), for the simple reason that people do not know enough languages to be able to make such a deduction. This type of information is not part of the general education either. So the only way that speakers of English may "know" this is if this information is innate.

Finally and most crucially, the type of universal -- a statistical, implicational universal -- that this study is all about is problematic as well. What use is it for a person to know that a language with the Verblog word order patterns is very unlikely? If we consider an absolute universal (e.g., "All languages operate with structurally-defined notions and not with linearly defined positions"), such a universal is useful for a child learner: she may simply disregard logically possible hypotheses about her language that are excluded from the realm of the possible human language. A child acquiring any human language would not even entertain a possibility that any given rule involves attaching a suffix to the third word, or moving the fifth word to the beginning of the sentence, or anything of the sort.

But how would the knowledge that a certain word order (e.g., Adj-N-Num) is rare, even extremely rare help our child learner? Crucially, she may not exclude the possibility that her language employs such rare/unlikely word order. Because if she did, she'd never learn her language if it happens to be Hausa or Paiwan, or any one of the other 35 languages in the WALS sample that have exactly that order.

Similarly, a combination of Num-N order with N-Dem order may be rare, with only about 11% of the languages in the WALS sample exhibiting this pattern, but if you are a child learning Basque, Berber, Breton, Fijian, Hebrew, Hmong, Mixtec, Rapanui, Vietnamese or any one of the other 87 languages in the WALS sample that have this Num-N-Dem order, you want to be able to entertain this pattern as a possibility.

Note that the universal being implicational is not a problem: implicational universals based on a hierarchy of parameters (see Mark C. Baker's Atoms of Language) help guide the acquisition of a child's target syntax. It is the statistical universal that is the problem.

So are statistical universals completely hopeless? A waste of our time? I will come back to this question in the next posting.