The Language Instinct, How the Mind Creates Language
Steven Pinker, 1994

Does Pinker tend to write half-books?
Once again, I am struck by the difference between the first half of this book and the last. I first read
How The Mind Works and got bogged down when he got into the details of family life. At least, he was still talking. Here, beginning with chapter 4 , he seems to lose his train of thought. There are indeed nuggets scattered about, but he has lost his earlier razor-thin story-telling edge.

To What Audience?
It is not clear to me just who the intended readers of this book are. The discussion seems a bit on the technical side to be of interest to the general lay reader. But there is not enough detail to convert a critical psychologist or sociologist (I doubt if any living neurosurgeon doubts the story told here). The book seems to be geared toward the sceptical Hilary Putnams of this world. But even there, it doesn't have the force of Pinker's later works, such as
The Architecture of the Language Faculty or How The Mind Works .

That brings me to what I believe is (was) the book's real purpose. It was a precursor to the thoughts Pinker was coming to when he wrote those two best-sellers.

Excessive Praise?
I have the 1995 Harper Perennial paperback edition of this book. I find the long list of rave testamonials preceding to frontispiece to be a bit excessive. Both the front and the back cover are similarly draped with quoted praise. It gets a bit hard to swallow.

Embedding Problems Compatible with Place Memory
The type of clause embedding limitations such as Pinker describes on pages 205 & 206 is compatible with a place-storage type of short-term memory, such as Edelman's TNGS theory would involve.

To What Audience?
Again, I am puzzled about Pinker's audience. The chapter
Big Bang seems clearly written to those who do not believe in evolution. Or maybe they consider the possiblility of evolution, but deny that language could have evolved. Is Pinker surrounded by such people? I know a few perhaps, but not enough to justify a chapter. Those people wouldn't be reading this book anyway.

On p 352 (middle) Pinker uses the term in what seems to be the same sense as Bickerton's usage. Of course, Pinker has mentioned Bickerton several times earlier and lists a reference to "Language and Species" (1990) in which Bickerton uses the term in his evolutionary/developmental sense rather than with the original historical linguistics meaning.

The Evolution of Language
pp 367..369 present a plausible scenario for the "upgrade" from proto-language to early syntax, although Pinker doesn't quite present it that way. He doesn't really consider proto-language as the likely "starting" point. While he was being so wordy about it, why not explore how the grammar could be pieced together structure-by-structure and still remain useful at all steps? p 367 bot and 368 top is a good start in this respect.

Chapter 12?
Just starting ch 12, it appears to be the high point of wandering, repetitious blah blah.

Innateness of Similarity Judgements?
I agree with Pinker that the cosmos is heterogeneous (p 418) and therefore we must be able to make a large number of different kinds of similarity judgements. However, I suspect that these similarities can be broken down into a much smaller number of features. It would be the mechanisms to detect the features, rather than the full set of possible similarities, that are innate. The same would be true of grammaticality judgements. Here, I believe Pinker overworks the syntactic issues while neglecting the semantic issues. I suspect that many of the required similarity features are common to those required for general survival in the jungle. (I have not yet read
Learnability and Cognition . Perhaps he will clear up such issues there).

For example, following up on the discussion pp 423..424, we might suppose innate mechanisms to classify objects, such as

Most general: recognize classes of similar objects More specific: animate (living) vs. inanimate (not living)

inaninate: non-hierarchical, objects move only if pushed

animate: classes are hierarchical, objects are self propelled

I would argue that at least the subcategories of animate and inanimate objects are composite classes which can be learned, given a few lower-level feature detectors, such as cause and effect (intuitive physics).

Top of Page | LIMCL Summary | Sort by Topic | Sort by Title | Sort by Author