goingfrom simplecasestorather complexones. (Toward theend,I'veadded contexts insquare bracketstohelpmake
sense of the examples.)
(3) a/the/this/that/what/which/whichever/every/each/some house
these/those/such/some/which/what houses
both/all/enough houses
such a/many a house
{so/too/how} {tall/old/...} a house
{much/far} too {tall/old/...} a house
{nearly/almost/practically/not quite} every house
{nearly/almost/practically/not quite} {all/enough} houses
how much more expensive a house [did Bill buy?]
[Bill bought]five times as expensive a car [as I did]
[you can get] {more than/almost}five times as fast a computer [for cheaper]
...Nobody would clai mthat when one learns a new noun, one adds to one's lexicon versions of that noun with all the
dozens of possible determiner combinations. Rather, every theory assumes these are built“in the syntax,”i.e. by
principles of free combination that assemble complex determiner structures and prefix the mto nouns, in accordance
with syntactic and semantic restrictions on the individual parts.
The same approach seems altogether appropriate for productive morphology. Words can be composed by combining
stems and affixes according to principles of free combination, in accordance with the phonological, syntactic, and
semanticrestrictionson the individual parts. For instance, Englishprogressive -ingcan be specified phonologicallyas a
suffix, syntactically as something added to a verb to form a participle, and semantically as something that (roughly)
operates on an eventto for ma process. Averb for mlikecutting, then, is the product of free combination of this suffix
with the verb stemcut.
For our purposes, the important question is: What does it mean to say this free combination is the result of a“lexical
rule”thattakes place“inthelexicon”? Taking toheart thecases ofexpletiveinfixationand Turkishverbinflection, and
bearing in mind the parallelto the determiner system, it seems implausible that“in the lexicon”here means“stored in
long-term memory.”Rather, like the examples in (3), the outputs of productive lexical rules are more likely to be
constructed online in working memory. (And like remembered phrases or song lyrics, theycanbe memorized and
stored redundantly, buttheyneed notbe.)From whatmaterialstoredinlong-term memoryare theyconstructed, then?
Obviously, the individual stems and affixes. In other words,