Effect of stimulus complexity on the phonemic restoration effect

Abstract

The phonemic restoration effect is the increase in speech intelligibility when the silent portions of the interrupted speech were replaced by a noisethat is louder than the speech signal itself. Previous research has demonstrated that the phonemic restoration effect is maximum when the interruption rate is between 1.5 and 3 Hz with a 50% duty cycle. Here we present phonemic restoration results obtained using five different speech corpora varying in stimulus complexity (least complex – most complex): (1) Callsign Acquisition Test (CAT); (2) Co-ordinate Response Measure (CRM); (3) Boston University Corpus (BUC); (4) IEEE sentences; and (5) R-PRESTO obtained from a group of young normal-hearing, listeners. Interruption rates of 2 Hz and 3 Hz were used and speech shaped noise was used to fill the silent interruptions. Initial analyses of the data revealed that the phonemic restoration effect was absent for the least complex stimuli (CAT) because of ceiling effects and the amount of restoration increased as the complexity of the stimulus increased. The effect of interruption rate and signal-to-noise ratio between the speech and noise on the phonemic restoration effect will also be discussed.