OK, Mrr: you asked for it, so here it is. My review of "Words and rules" by Steven Pinker.

First of all, it's an entertaining read, although reading a few pages each evening before going to sleep is not the best way to read it - it really needs just a few concentrated sessions. Had I more time, I'd re-read it, but queuing up to be read are Age of Reason and Enchanted Looms at the top of the pile and many more below.

I've little experience in the subject outside of popular texts so am happy to be corrected on detail, but Pinker makes, it seems to me, a good argument for his thesis and seems to be even-handed in discussing not only evidence in support, but also evidence which may cast doubt on it. The thesis is that language-processing in the brain is a combination of retrieving words from a stored dictionary and modifying and combining them with rules. So, for example, my mind stores a word, walk, and I retrieve it and describe my current action, I walk. To put it into the past tense, I apply a rule, add "-ed", and describe yesterday's action as I walked. Thus it is not necessary for the brain to store a past-tense for every verb in its dictionary: storing the root of the verb only and one rule for making regular past-tenses is all that's necessary. Now, that doesn't seem so surprising, does it?

But what about irregular verbs: Today I go, yesterday I went; today I write, yesterday I wrote? These, says Pinker, are stored in the dictionary, and, for example, lookup of the past-tense of I am follows the path "Is there a stored past-tense? Yes, was, therefore retrieve it and block the regular rule, add "-ed", to yield I was. Intuitively appealing, to my mind.

Having proposed this thesis, Pinker takes us on a guided tour of current linguistics theories discussing as we go such interesting points as why the final consonant of "cats" sounds different from that of "dogs" (caution, all you Mudcatters reading this at work: your colleagues will think you mad, hearing you muttering CAH-TSSSSS and DOH-GZZZZ to yourself), why we say mice-infested (more than one mouse) but rat-infested (only one rat), why Chomsky's Deep Structure theories are likely to be mistaken, why neural-network computer-models of forming the past tense are unsatisfactory, how children acquire language and how the behaviour of people with brain abnormalities lend credence to the basic thesis. And a lot more. Fascinating stuff. All in all, a book to be recommended; I certainly enjoyed it, anyway.