Article

Theories of morphological processing differ on the issue of how lexical and
grammatical information are stored and accessed. A key point of contention is
whether complex forms are decomposed during recognition (e.g., establish+ment),
compared to forms that cannot be analyzed into constituent morphemes (e.g.,
apartment). In the present study, we examined these issues with respect to
English derivational morphology by measuring ERP responses during a cross-modal
priming lexical decision task. ERP priming effects for semantically and
phonologically transparent derived words (government-govern) were compared to
those of semantically opaque derived words (apartment-apart) as well as
"quasi-regular" items that represent intermediate cases of morphological
transparency (dresser-dress). Additional conditions independently manipulated
semantic and phonological relatedness in non-derived words (semantics:
couch-sofa; phonology: panel-pan). The degree of N400 ERP priming to
morphological forms varied depending on the amount of semantic and phonological
overlap between word types, rather than respecting a bivariate distinction
between derived and opaque forms. Moreover, these effects could not be accounted
for by semantic or phonological relatedness alone. The findings support the
theory that morphological relatedness is graded rather than absolute, and depend
on the joint contribution of form and meaning overlap.
CI - Copyright A(c) 2010 Elsevier Ltd. All rights reserved.

Langue : ANGLAIS

Tiré à part : OUI