Lexical stress representation in spoken word recognition

According to a popular model of speech production, stress is underspecified in the lexicon, that is, it is specified only for words with stress patterns other than the default, termed the “default metrics” assumption. Alternatively, stress may be fully specified in the lexicon as part of every lexical representation. In the current study the two accounts are tested in the perceptual domain using behavioral and eye-tracking data in Greek. In a first experiment, cross-modal fragment priming was used in a lexical decision task. According to default metrics, priming should occur for targets with antepenultimate- or final-syllable stress but not for targets with the default penultimate-syllable stress. The same word pairs were used in two subsequent visual world experiments. Default metrics predicts an asymmetric pattern of results, namely that incoming spoken words with the default stress pattern should inhibit the activation of lexical representations with nondefault stress, whereas the converse should not be observed; that is, spoken words with nondefault stress should not inhibit representations of words with the default stress. None of the results provided support for the idea of default metrics, leading to alternative conceptualizations regarding the representation of stress.