Stress is not underspecified: Evidence from fragment priming in Greek

Presented at the 19th Conference of the European Society for Cognitive Psychology; Paphos, Cyprus, September 17–20, 2015

Angeliki Andrikopoulou,1 Athanassios Protopapas,1 & Amalia Arvaniti2
1University of Athens, Greece
2University of Kent, UK

Lexical stress is known to affect word recognition. Is it fully specified in the lexicon or is it underspecified and assigned upon words through metrical frames? Although previous findings seem to endorse an underspecified representation of stress, the evidence is not clear-cut. The current study addressed this issue using the fragment priming paradigm. The purpose was twofold: the replication of stress priming as it has been observed in Spanish, Dutch, and Italian; as well as the investigation of the nature of stress representation in accordance with theories of underspecification and metrical frames. To achieve this, 72 pairs of three-syllable words were selected and used as primes and targets. Word pairs shared all phonemes of their first two syllables and the first phoneme of their third syllable. Two-syllable fragments of the primes were presented auditorily in the end of neutral carrier sentences; targets appeared visually on a computer screen immediately thereafter. Stress pairs were grouped by stress contrast with primes and targets either matching or mismatching. In an additional neutral condition prime and target differed both segmentally and suprasegmentally. To test theories of stress representation, pairs were grouped by stress position, in three combinations of stress patterns. In particular, 24 pairs had stress on their first and third syllable, 24 had stress on their first and second syllable and 24 on their second and third syllable. If stress is underspecified, there should be no priming on penultimate-stress targets, because the default stress position is not specified in the lexicon; however, there should be priming with antepenultimate- and final-stress targets. 75 adult native speakers of Greek were tested. The results replicated stress priming effects, revealing both facilitation for matching pairs and inhibition for mismatching pairs. However, there were no priming differences among different combinations of stress patterns, raising doubts regarding stress underspecification.