Lexicalist models of morphosyntax need to have a way of individuating lexemes. I adopt Spencer's (2013) Generalized Paradigm Function Morphology (GPFM) model of lexical relatedness, in which lexemes/lexical entries are individuated by means of an arbitrary index, distinct from phonology, syntax or semantics. I sketch an LFG interpretation of the GPFM model of argument structure which crucially appeals to semantic function (SF) roles to specify morphosyntactic category. I focus on transpositions, deploying the notion of complex SF role, (which can be used to label corresponding c-structure nodes). I distinguish true transpositions and meaning-bearing transpositions from transpositional lexemes, distinct lexemes which do not involve an added semantic predicate. I illustrate the latter with deverbal adjectives such as English interesting. The analysis crucially appeals to GPFM's LI attribute. I show that the LI attribute is virtually coextensive with LFG's PRED attribute. GPFM and LFG are therefore compatible.