By contrast, UVPs allow a very limited range of NP complements:
Indeed, many and perhaps most UVPs appear to allow only NP complements that designate propositions or quantify over them ('logicism', 'the proposition that mathematics reduces to logic', 'what John said', 'every theorem of Peano arithmetic', etc.).
This data, it seems to me, is quite suggestive, If UVPs really are univocal and express relations between individuals and propositions, then this would explain why the NP complements allowed by such verbs would be restricted to those that designate propositions (or designate things that give expression to or contain expressions expressing propositions).
The third bit of evidence that AVPs are ambiguous and that UVPs are not concerns data involving gapping.
The claim that AVPs are ambiguous and UVPs are not would explain this.
Still, AVPs allow a wider range of NP complements than UVPs.
(17) There still is a question as to the precise semantics of sentences containing UVPs like 'believes' or 'doubts', where the NP complement designates something that can give expression to propositions (or contains expressions expressing propositions, etc.).