Effects of Clustering Coefficient on Spoken Word Recognition

Since the late 1960’s, researchers have explored how the structure of the mental lexicon affects spoken word recognition. The proposal that words are recognized relationally in the context of other words in lexical memory has encouraged the use of complex systems to describe connectivity in both the phonological and semantic lexicon. The present study assessed the role of the graph theoretical measure of Clustering Coefficient (CC) using two experimental paradigms: same-different discrimination, and perceptual identification to explore how global lexical variables affect spoken word recognition. In Experiment 1, listeners judged whether two words were the same or different. Longer response latencies were obtained for high CC words than low CC words. In Experiment 2, the stimuli were processed with an 8-channel noise vocoder to degrade the signal and a new group of listeners also performed the same-different task. In contrast to findings obtained in Experiment 1, listeners discriminated high CC words more accurately than low CC words. In Experiment 3, an open-set perceptual identification task was carried out to examine correct and incorrect responses using 8, 10 and 12 channels. Listeners identified low CC words more accurately than high CC words in the 10 and 12 channel conditions, but not in the 8-channel condition. Detailed analysis of the incorrect responses revealed that listeners used different perceptual strategies as the number of channels increased. These results suggest that global, emergent factors reflecting the structural organization and connectivity of words in the mental lexicon affect spoken word recognition and should be included in current models of spoken word recognition and lexical access.