Similarly, the requirement that the meaning of a pseudo-complement must mesh with the meaning conveyed by the verb it modifies can account for the contrasts in resalt16.
The river froze solid/*icy. The toast burned black/*hard.
Such unaccusative verbs all inherently express changes of state -- e.g. if the river froze, then the river wasn't frozen beforehand but is now. The meaning of to freeze could be captured as shown in figfreeze, expressing a change of state to a state of being frozen (ch-st-result-rel is a subtype of ch-st-rel which licenses a specific stative effect).
The lexical entry for the standard use of freeze.
Any AP which modifies such a change of state verb must express a
property which unifies with the final state expressed by the verb
(e.g. frozen-rel in the case of freeze). So the entry in
figfreeze could be combined with an entry analogous to that
introduced in resalt25 on page 
. A
pseudo-complement lexical rule for APs would pull out the semantic
contribution of the AP (the first element of the RESTR set)
and require it to unify with the state expressed in the verb
semantics, thereby ruling out APs which bear no relation to the
semantics expressed independently by the verb.
The lexical entry for the use of freeze
with an AP pseudo-complement. The state expressed by the AP,
represented by 
, must unify with the state expressed
internally in the verb semantics (in the thematic effect).
In particular, it seems that these APs are constrained by pragmatic
factors to express a state more specific than the property encoded in
the lexical semantics of the verb.
This constraint
is met when an AP expresses how the state expressed by the verb is
manifested in the referent of the subject. So, when water is frozen
it is solid, and when toast is burned it is black, etc. In fact, the
state expressed by the verb seemingly needs to be entailed by the
state expressed in the adjective phrase, relative to the referent of
the subject (by default, a solid river must be frozen and a black
toast must be burned while an icy river may not be entirely frozen and
a hard toast may not be burned).
World knowledge therefore seems to interact with the ability of an AP
to act as a pseudo-complement in these cases. Goldberg's (1995)
account cannot explain these differences as her Intransitive
Resultative Construction does not
require the result phrase to fill a role in the basic predicate
expressed by the verb.