roughly‘put N on something,’where N is the homophonous noun. But for each individual noun in the language, the
lexicon must specify whether it givesrise to such a denominal verb. For example, oneonly asks someonetomustard the
sandwichordoor the cabinet(‘put a door on the cabinet’) as a metalinguistic joke; speakers know that these forms do not
exist.
Derivational morphology raises two further complications that can be illustrated with deverbal nouns. First, deverbal
nouns can have a variety of meanings, for instance those in (4).
(4) ‘put N on’: butter, water, paint, roof
‘take N off’: dust (the shelves), scale (afish), skin (a cat)
‘put on N’: shelve (books), plate (food)
‘put in N’: pocket (the money), bottle (the wine)
‘fasten with N’: glue, staple, nail, tapeAnd just as there are homophonous verbswith differentpast tenses, such ashang–hungvs.hang–hangedandring–rangvs.
wring–wrung, there are homophonous denominal verbs with different meanings. For example, in addition todust (the
shelves)‘take dust off’, there isdust (the cake with sugar)‘put dust-like substance on’; in addition toroof (the house)‘put a
roof on’, my daughters' dialect hasroof (a frisbee),‘put/throw onto a roof’. Thus one must know not justwhethera
particular noun has a correspondingdenominal verb, but also whichvariety or varietiesof denominalverb meanings it
has.
The second complication is that, even if one knows this much, one cannot predict the complete meaning of the
denominal verb. To put a clockona shelfis nottoshelveit;tojust pour wine intoa bottleis nottobottleit;tospillwater
on a table is not towaterit. One cannotsaddlea table by putting a saddle on it; one cannotbutterone's toast by laying a
stick of butter on it. The verbsto motherandto fathermean very roughly‘actas amother/father toward someone’, but
are entirely different in the exact actions that count as relevant. In short, many denominal verbs have semantic
pecularities that are not predicted by the general lexical rule. (See section 11.8 for treatment of the regular part of
denominal verb meanings.)
The upshot is that the outputs of semiproductive rules must be listed (at least in part) in long-term memory; they
cannot be a product of free combination. Hence in this case the notion of a“lexical rule,”a rule applying“in the
lexicon,”makes sense. Such a rule expresses a generalization in the relationship among pairs of lexical entries. But it
doesnotapplytoallentriesthatsatisfyitsconditions, nordoesitfullypredicttheinformationinonemember ofa pair,
given the other. Thus, unlike fully productive morphology, this sort of rule captures the original intention behind
“lexical redundancy rules”(section 3.3.2.).