Analysis of Spoken Word Recognition 1

This study reports the analysis of incorrect responses from a spoken word recognition experiment of 1,428 words, designed to be a representative sample of the entire English lexicon. The stimuli were presented in six-talker babble to 193 normal-hearing listeners at three signal-to-noise ratios (0, 5, and 10 dB). The results reveal several patterns: (1) Errors tend to be of higher frequency than the target word, and response frequency is significantly correlated with target frequency; (2) Incorrect responses were very close to the target words in terms of number of phonemes and syllables, but had a mean edit distance of 3; and (3) NonWord responses were phonetically more similar to the target words than word responses. Implications for theories of spoken word recognition are discussed.