THE REPRESENTATION AND PROCESSING OF INFLECTED FORMS IN A MULTILEVEL LEXICAL SYSTEM by

Five unmasked-and four masked-priming lexical decision experiments were carried out in Spanish in order to probe the data structures and processing routines that are exploited during the comprehension of inflected words. In unmasked priming, responses to targets (e.g., mor-os 'Moors') were inhibited when they were preceded by stem-homograph primes (mor-ir 'to die') compared to unrelated control primes (sill-a 'chair'), and this inhibitory effect was over and above a weak inhibitory effect found for orthographic relative primes (moral 'moral'). Stem-homograph inhibition is interpreted as a consequence of morphological decomposition in lexical access. In additional experiments, this inhibitory effect was observed when targets were preceded by primes that were not themselves stem-homographic with the target, but rather were stem-allomorphic variants of the stem-homograph words (muer-e 'she/he/it dies'). Because target inhibition was found for primes whose inflectional stems are not strictly ambiguous in terms of form, it suggests that the inhibitory effect arises at a level where form-neutral, morphologically abstract (lemma) representations are encoded. The results observed in masked (subliminal) priming contrast with those observed in unmasked priming. When the primes were presented subliminally, stem-homograph priming (mor-ía —mor-os) was facilitative, while orthographic relative priming (moral —mor-os), was neither facilitative nor inhibitory. This was taken as further evidence that stems are decomposed at the form level, and that ambiguous decomposed stems, such as mor-, activate multiple (compatible) lemma representations during the initial stages of lexical access. Moreover, the facilitative effect observed in masked stem-homograph priming suggests that the lemma-level inhibitory mechanism implicated by stem-homograph allomorph priming should not be characterized in terms of built-in inhibitory links between certain lemmas, but rather as a mechanism that is invoked only when mutually exclusive lemma representations compete for selection as the interpretation for an ambiguous form. Furthermore, because stem-homograph priming was inhibitory when participants were aware of the primes, but facilitative when they were not, these priming results suggest that a single lemma interpretation for an ambiguous lexical form must be selected before that form can be consciously perceived as a meaningful word.