Books: When irregular verbs can be exciting

Words and Rules is a demanding, full-length book about irregular verbs. Yet this description gives little sense of its flavour and excitement. Better to say that it develops the arguments of Pinker's ambitious "language instinct" and "how the mind works" by homing in on a few minor linguistic quirks and then using them to illuminate some essential aspects of what makes us human. Written with characteristic energy and enthusiasm, it sweeps across some pretty dry terrain but remains gripping throughout.

The central point can be made briefly. A dirty house may be "rat-infested" or "mice-infested" but never "rats-infested" (though an infestation requires more than one rat). This, Pinker claims, is because regular and irregular plurals are stored differently in our memory. For irregulars, we have a number of mouse-mice, ox-oxen, sing-sang-sung networks hot-linked in our brains, giving us instant access to all the correct forms. With regulars, we just learn the head word "rat" along with a rule stating roughly: "When in doubt add an s to a noun to create a plural."

Because the word "rats" does not appear in our mental dictionaries, it is unavailable for use in compound expressions. The s and ed default settings for plurals and past tenses can override seemingly tempting alternatives. Cookery-writer Julia Child and her family are "the Childs", not "the Children". Cities are "ringed" by artillery, beans need to be "stringed" (not "strung"), and so on. A bad actor in Hamlet "out-Herods Herod", so theoretically we can create a new verb by adding out to any name, as in "Clinton tried to out-Kennedy JFK". Until a terrible riot at the Attica Correctional Facility in 1971, Sing Sing was the most infamous prison in New York. Pinker exuberantly describes this event as the time when "Attica out-Sing Singed Sing Sing"! No one could conceivably say "out-Sing-Sang" or "out- Sang-Sang". Yet when "outsing" is derived from the verb "to sing", we tap into a different part of our internal lexicon to produce "Domingo outsang Pavarotti". All this is lively and stimulating, but how does Pinker go about testing his theory about two kinds of memory?

One approach is to look at different languages. Since 98 per cent of English nouns have a plural in s, it is hardly surprising that this is established as the default setting. Astonishingly enough, we find the same thing in German, even though only one per cent of common nouns normally do so. If we import Renault Elfs, they export Opel Kadetts. Thomas and his wife are die Manns, films about the Caped Crusader are Batmans, and a feminist might look for Manns in a piece of sexist writing. The two languages have a common origin, of course, and a common tendency to impose a regular pattern on imported foreign verbs. The key difference is that the English lost the Battle of Hastings, so Norman French became the language of the elite and about 60 per cent of our verb roots now come from French or Latin. It is not psychological factors but accidents of history which account for the dominance of a single form of regular verbs and plurals in English.

Default rules can also be found in Chinese. In English, one has to talk about "a blade of grass" and "a slice of bread" rather than "a grass" or "a bread". In Chinese, such classifiers are used with all nouns; one for groups of people, one for animals, one for small things, and so on. Exceptions are mopped up with the classifier ge. This is the word which speakers fall back on when they have a memory lapse and which young children overuse.

In trying to prove his words-and-rules thesis for regular and irregular verbs, Pinker examines language learning, speed-of-response tests, the errors of different kinds of brain-damaged patients and the evidence slowly emerging from brain scans. The result is a fascinating survey of many key areas of linguistics.

The final chapter broadens the focus. We all make frequent use of two different kinds of category: logically watertight classes and far more informal groups based on resemblances. A single concept often slides between the two. A strict definition of the term "grandmother" is appropriate for a medical researcher tracing a defective gene. Yet usually, Pinker suggests, "When people think of a grandmother, they think of grey hair and chicken soup, not of nodes in a genealogical tree" - and of somebody who looks quite unlike Tina Turner. In a context like a paternity suit, it makes sense to define "sex" to exclude fellatio; when Clinton used such legalistic logic with respect to Monica Lewinsky, he was cunningly offending against ordinary usage. Classical categories have an obvious similarity to rule-based regular verbs. Examples include "numbers, ranks, kinship terms, life stages, legal and illegal acts, and scientific theories". "Fuzzy" categories are more like irregular verbs, which mainly come in broad but unreliable family sets (bend-bent parallels send-sent, but mend-mended spoils the pattern). Cutting straight through a centuries- old philosophical dispute, Pinker argues that our brains have evolved to rely on both kinds of thinking in our struggles with the material and social worlds.

Natural languages differ greatly. Standard English verbs come in four forms (open, opens, opened and opening); Spanish and Italian verbs require about 50. Meanwhile, one Bantu language attaches prefixes and suffixes which mount up to about 500,000 combinations. Yet what they all have in common, unlike the streamlined artefact of Esperanto, tells us much about the differences between human beings and computers.