Is conversion a syntactic or a lexical process of word formation?

Resumo

Conversion is sometimes described as a syntactic phenomenon by which a lexical item changes its lexical category according to the syntactic environment where it is inserted. This syntactic-ordered approach comes from theoretical fields that conceive the lexicon as the domain of irregularity, whilst regular patterns are treated in syntax (Chomsky 1995).
However, Portuguese converted deverbal nouns (remendo ‘event of mending’, curte ‘event of having fun’, trinca ‘event of biting’) manifest a structural behaviour that permits us to situate their formation in the lexicon instead of in the syntax. According to the theoretical allusion we made above, this would characterise converted deverbal nouns as lacking a regular pattern. However, what we mean is that the lexicon is not the field of irregularity. Apart from the irregular material that must be stored in long term memory as to be used by speakers, such as inherited lexemes (e.g. rato ‘mouse’, cão ‘dog’, rir ‘to laugh’,
verde ‘green’, etc.), the lexicon is the domain of word formation, which is constraint-based (Rodrigues 2008, 2009). This means the word formation part of the lexicon is constrained by regular patterns that are neither directional in principle, nor syntactic in nature.
We follow Jackendoff (2002) conception on the lexicon, conceiving it an interface of syntax, phonology and semantics. Converted deverbal nouns formation seems to agree with this conception, since it depends on phonological, semantic and syntactical constraints (Rodrigues 2004, 2009). Portuguese verb-into-noun conversion is not a simple case of syntactic environment. This is specially visible when we confront this lexical conversion with a purely syntactic type of nominalisation (Kerleroux 1996), such as the one that occurs in O estudar matemática traz-me vantagens. ‘Studying maths brings me advantages’ or O remendar roupa é um recurso nesta época. ‘Mending cloths is a good resource nowadays’. Apart from the syntactic constraints that converted deverbal nouns must follow (e.g. as real nouns they cannot directly theta-mark, whilst syntactic nominalisations can), these nouns display phonological constraints in their formation and a semantic variability that is not observable in syntactic nominalisations. This semantic variability is not exclusively explainable by
the syntactic environment. Moreover, the formation of converted deverbal nouns obeys to structural constraints that do not act upon syntactic nominalisation. All this means that conversion is not syntactic in nature. Yet it is not irregular either. This evidences that the mainstream generative grammar conception on the lexicon should be abandoned.