f. I wonder if Susan knows that Fred assured Lois that Clark would remind Pat to buy food for dinner.
g. We havea youngchildand our chef has twins so we know howdifficultitis tofind afirst rate restaurant that
doesn't shudder when you show up at the door with kids in tow.The length and complexity of utterances involves not only the number of words and their syntactic organization. The
messages that utterances convey—and their topics—are equally unlimited. In this respect human utterances contrast
sharply with the long and complex songs of certain species of whales and birds, which,as far as can be told at present
(Hauser 1996; Payne 2000; Slater 2000), basically convey only the message“Here I am, everyone!”
In principle, a communication system might construct arbitrarily long mes-sages just by adding more and more new
elements to the end of utterances, along the lines of a shopping list. But human language doesn't work like that: it
builds up large utterances according to structural principles orrules. What made generative grammar in the modern
sense possible was the development of formal techniques for describing rules and systems of rules, deriving from
work in the foundations of mathematics during thefirst half of the twentieth century. (The very same techniques, of
course, led to the development of the digital computer.) And, although there were precursors in post-Bloomfieldian
structuralism, in particular Chomsky's teacher Zellig Harris (Harris 1951), it was Chomsky who developed and made
clear the connections between this mathematical work and linguistic description.
Putting theissue of combinatorialityintoa mentalist frameworkadds an importanttwist. Since thenumber of possible
utterances of a human language is unlimited, language users cannot store them all in their heads. Rather, f-knowledge
of language requires twocomponents. One is afinitelist of structural elementsthat are availableto be combined. This
listis traditionallycalledthe“lexicon,”and itselements are called“lexical items”; for themomentletus suppose lexical
items are words or morphemes (we will alter this substantially in Chapter 6. The other component is afinite set of
combinatorial principles, or agrammar.^12 To the extent that speakers of a language (or a dialect) are consistent withone
another (see section 2.5), we can speak of the“grammar of the language”as a useful approximation to what all its
speakers have in their heads.
The task of a theory of linguistic competence, then, is to model the lexicon
COMBINATORIALITY 39
(^12) A terminological point:Sometimes“grammar”is taken to encompass rulespluslexicon, particularly when, as will be seen in section3.3, the lexicon is thought to contain
rules as well. Fro mforce of habit, I will no doubt fall into this inconsistency, but I will atte mpt to make clear what is intended when it makes a difference.