Interaction of acoustic and neighborhood density parameters on spoken word recognition.

Abstract

The neighborhood activation model (NAM) [Luce and Pisoni (1998)] predicts the speed and accuracy of word retrieval in the mental lexicon based on word frequency and neighborhood density. Data from NAM experiments have demonstrated processing advantages of dense words over sparse words. Little is known about how acoustic, versus phonetic, variability influences processing strategies used by listeners of acoustical‐degraded speech such as those using cochlear implants and hearing instruments. In this experiment, neighborhood density was investigated using speech‐like sentences which preserved detailed frequency information but removed most amplitude information. Results yielded no difference between dense and sparse words when presented in this time‐varying sinusoidal format [Remez et al. (1980)]. Consistent with prior research, amplitude modulation increased overall intelligibility of the stimuli; however, the effect of neighborhood density was in the opposite direction from that found in natural speech. We hypothesize that this may be indicative of the neighborhood density effect, becoming secondary to a phonotactic probability effect which has previously been associated with sublexical processing tasks. These results indicate that interactions between acoustic and phonetic factors must be accounted for when designing hearing instruments and cochlear implants. Follow‐up investigation with other acoustic variables is currently underway in our laboratory.