Language Learnability and Language Development (Cognitive Science Series) [Hardcover]

How this process occurs is the topic of Steven Pinker's monograph "Language Learnability and Language Development". He focuses on how a child can learn the grammar of his language. His approach is quite formal and technical, as is fitting for a professor of linguistics writing for an audience of professional researchers. The ultimate goal is to define a set of algorithms that processes the input (the sentences heard by the child) and creates a set of rules that define the grammar of the language the child is hearing.

Pinker's description of this process begins with the question of word order. A child must determine in what order his language puts words. Should he put the subject before the verb (He runs. vs. Runs he.), should an adjective precede a noun (white house in English, casa blanca in Spanish)? Where should an indirect object go? This isn't easy work. All the child has to go on are the sentences she hears and some non-verbal signals (pointing, tone of voice, the context of the utterance). Actually, the child has a little bit more; Pinker argues, as do many linguists, that we are all born with a mental framework for grammar. This can be imagined as a series of rules with some blanks to be filled in. All of the world's grammars can be described by filling in the blanks differently. So the child is not without some guidance.

Once the child has identified the basic meanings of some of the words this framework for grammar and the algorithms Pinker proposes work together to determine how sentences are constructed. As the child learns more words and more of the basic grammar rules, more difficult notions can be tackled. The text proceeds through topics such as noun-verb agreement, verb forms and irregular verbs, auxiliary verbs (including the troublesome word "do" in English) and the formation of passive constructions. For each issue Pinker describes a framework and a series of rules for filling the blanks.

The fundamental constructs and the nature of the rules require the reader to have some understanding of formal grammars (the transformational grammars of Chomsky or the LFG that Pinker bases his arguments on). Your understanding doesn't have to be deep, but it would be difficult to work your way through the book without some familiarity with the subject. Further the algorithms are given reasonably formally (not purely mathematically, but certainly it requires some effort to piece them together). In some sections of the text you may be forced, as I was, to simply skim over the details in an attempt gain some understanding.

This is an older text (1984 for the first edition) and Pinker was trying not only to present his results, but--and perhaps most interestingly--develop a method for analyzing any theory of language acquisition. He even uses these criteria to judge his own theories. He can tell us where his ideas on language acquisition succeed, admits where they fail, and hypothesizes how they can be improved. This methodology defined a successful research program that continues today Pinker makes his case through careful, clear, and compelling arguments. This discipline isn't easy. Most of the experiments one would need to run, in order to validate the theory, would be highly unethical (let's speak to a child only in the passive voice, let's never ask a child a question and see what happens). So the theories have to be verified by the sentences children speak and a few experiments, usually with made up words. Unfortunately, the sentences children speak do not unambiguously tell us the grammar they used to construct them.

I am no expert in this subject matter. I picked up the book because I had enjoyed the books Steven Pinker has written for the interested layman, such as "The Language Instinct", "Words and Rules" and "How the Mind Works". I did have to work hard to try to understand the text and I am sure I didn't fully understand some of the more technical arguments. However, as a soon to be father of a new language learner, I will watch with heightened appreciation my child's fascinating ability to learn English.

"Braine concludes that in every instance in which a child frequently utters both possible orders of a pair of categories, either (a) both orders are found in the adult language, and the child has learned the two orders separately (evidence for this is the fact that one order typically predominates at first and then is supplemented by the alternative order; or (b) there is independent evidence that the utterances reflect a pregrammatical "groping pattern" in which the child wants to communicate a semantic relation, lacks the grammatical means to do so, and strings words together randomly in the hope of being understood."

That one is not so bad, actually. How about,

"When a paradigm is split in response to a violation of the Unique Entry principle for a given cell, the child in effect "expects" to find alternative entries for each of his or her other existing affixes that do not already have alternatives with strong lexical entries".

Having three post graduate degrees and learned to speak a second language fluently as an adult, so I actually know something about the subject of language acquisition and am not afraid of dense prose. But this book defeated me after a few pages.

It is in my view supremely ironic that, given their field of study, linguistic scholars seem to be abominably poor at communication.

However, the argument seems fairly simple:

1. A grammar can be abstracted from speech (in this case, English speech) based on observed regularities.

2. We can observe a change in a child's speech as he/she moves from simpler to more complex grammatical forms. So we can identify each such change in terms of the starting and ending forms.

3. For any such changes, we surmise the child has come up with one or more rules to enable him/her to acquire the more complex form.

The focus seems to be predominantly on the lexical and syntactical forms. Semantics seemed to be mentioned but not much of a concern here. Behaviorists, on the other hand, tend to emphasize the function of speech (although not discounting the importance of syntax).

So, as a probably very naive example, consider a child who can say "Give me the apple". Pinker may have noted that the syntax is correct but omits the use of an adjective. Skinner may be wondering whether the child's statement will be reinforced by being given an apple. Now suppose at some later time there are two apples in sight, a red one and a green one. The child now says "Give me the red apple." Pinker has never heard the child use an adjective before and notes that as a development in language acquisition. Skinner wonders if the child's statement will be reinforced by being given the red apple and may be pleased to see that child was able to request that red one.

The child's syntax, for whatever reason, was correct but it happened in service of the child's request for an apple. It's wonderful he/she was able to say it in that way, but it also seems wonderful that he/she got an apple.

In such a situation, one can really see the red apple. One can really hear the child ask for it and really see the child get it. One really heard the request. How real was the syntax? Well, it may not be explicit in the sense the child or listener are aware of it as syntax. But it may make a big difference if the syntax were wrong in that the child's request could fail. So perhaps we can say the syntax is real (and diagram it if we like). Now can we say that the rule is real that allowed the child to go from the request without the adjective to the request with the adjective. Maybe, but it seems less clear. It also seems less clear how the child made the step. Did he/she create the rule somehow using an innate langauge ability? Or did he/she imitate something he/she had heard? But what if the child had heard a request for a "green apple" and could tell what "red" was but had never heard of a "red apple". Or what if the child had heard of a "red apple".

Now I've gotten myself into trouble. Do I turn to Pinker or Skinner for the answer? Is it an either/or? Does it matter that the child wants something to eat or only that he/she is demonstrating an innate capacity for grammar?

So where did the syntax come from? Where did the rules come from? Who ate the apple?

Grammarians find grammar. Cognitive scientists find mind. One way or another, the child got the apple. Which of these does natural selection seem to favor?