510 ROBERT D. VAN VALIN, JR. & DAVID P. WILKINS
Two of the complement types that remember can take are are a that-clause
and an infinitive, as in John remembered that he had left the faucet on and
John remembered to turn the faucet off. Semantically, the that-clause may
be termed a cognition complement, the infinitive a psych-action comple
ment. Claudia Brugman (personal communication) has pointed out that
this same semantic constrast can be found with simple NP objects, as in
John remembers the answer vs. John remembered his checkbook. The
former has a cognition reading only, and while the latter could have a
cognition interpretation, it is more likely to have a psych-action reading,
e.g. he remembered to do something with his checkbook. The use of the
verb in the simple present tense is possible only with the cognition inter
pretation, which is consonant with the Stative nature of cognition pred
ications. These facts suggest that the cognition vs. psych-action contrast
is a feature to be attributed to the meaning of the verb itself and is not
simply a reflection of the different complement types. Given that re
member is represented semantically as remember' in this system, it would
be difficult to represent this semantic difference except by positing two
remember verbs, remember 1 (cognition) and remember 2 (psych-action).
This captures the fact that remember is involved in at least these two
interpretations, but it does not in any way represent the different meanings
involved. It is not even clear whether polysemy or homophony is being
suggested. Moreover, the cost of this analysis is tremendous in terms of
multiplying the number of verbs in the lexicon, since any verb with semanti
cally contrasting complement types will be represented with separate
entries for each meaning. Furthermore, this analysis is counterintuitive in
claiming that there are two (or more) different verbs remember, as it fails to
capture all of what is common to the two interpretations. In sum, it is an
extremely problematic analysis, and it is our contention that these and
other similar problems indicate the necessity for a deeper semantic rep
resentation for complement-taking predicates. It should be noted that this
verb has a third interpretation which relates to a different complement
choice from the ones discussed above; this one will be labelled "direct per
ception" and is exemplified in sentences like I remember turning off the
faucet and Fred remembered Jim signing the check.
We now turn to the semantic decomposition of remember. The seman
tic metalanguage to be employed is based on natural language, following
the approaches of Wierzbicka and Dixon as outlined above, and using basic
operators from the Dowty system, e.g. BECOME, NOT, CAUSE. The