I'm currently reading Stephen Pinker's The Stuff of Thought: Language as a Window into Human Nature on audiobook. Audio is not a very good medium for this book, since there are lots of examples comparing constructions of sentences, like:

Mary poured the water into the glass. (acceptable)
Mary poured the glass with water. (unacceptable)

Being able to look at contrasting examples side-by-side in print is easy. Holding them in memory for comparison is a lot more difficult.

That said, the book is pretty interesting. I'm not enjoying it as much as How the Mind Works (even though that book didn't really explain how the mind works).

I'm only about halfway through, so I'll probably have more to say about later chapters, but so far he's talked about language acquisition (particularly as it relates to verbs in English) as a way of understanding language and cognition as a whole.

He talks about the inherent puzzle of how we learn language in the first place. Yes, every language has regularities, but every language is also rife with exceptions. We only hear certain constructions a few times, and if we tried to generalize based on those sparse examples, we'd end up making lots of mistakes (which is what children tend to do). This is related to the Poverty of the Stimulus argument, which basically says that kids just don't get enough good information to learn languages as fast and as well as they do. I've never really bought the POTS argument, though I think it would be damned interesting to try to record everything that a child could hear from birth to 10 years old. I think if anything, the stimulus is overly rich.

But some people argue that we can't possibly have learned the rules of grammar and syntax by extracting the statistical regularities out of the language we're exposed to, so they argue to a greater or lesser degree that aspects of language are innate. Pinker is labeled one of these nativists, but for a good early portion of the book he goes to great lengths to show that regularities in learning particular types of verbs (like the example above) do exist, where they didn't seem to.

Constructions we can use for verbs like load and spray are not suitable for other verbs like pour or drip. Pinker argues that there are differences in the usage of the verbs because there are differences in the qualitative physics in the scenarios where those verbs are used. Pouring has to do with allowing gravity to work on a liquid, while loading implies that less of the work is being done by gravity than by an agent. So he's basically making the argument that where there didn't seem to be much regularity, there actually is. Kind of a strange argument for a nativist to make.

He then goes on to use this as evidence that we have a "language of thought" based on our conception of how things move and change over time, and that the languages we speak are based on it.

This seems perfectly reasonable to me. Having concepts of motion and change in the world seem like necessary prerequisites for evolving language in the first place. Even though chimpanzees don't have words for grab or drop, I would tend to think they have concepts for these actions and be able to make conceptual distinctions between them. In other words, I'd guess that many other mammals have various levels of this "language of thought" that Pinker talks about, wordless concepts for actions and events in the world.

Anyway, I'll review later parts of the book when I get to them. Apparently he's got a whole chapter on swear words. He's already previewed his argument (that we substitute words dealing with expelling bodily waste or sexual acts for the yelps and cries of pain that non-verbal animals use), but it should be to hear him elaborate.