Lexical Factors Affecting Spoken Word Recognition Donald D. Dirks,
Sumiko Takayanagi, Anahita Moshfegh, Stephen Fausti and Douglas Noffsinger Speech recognition testing has been a traditional procedure used to assess hearing aid benefit. Most speech recognition tests are based on processing "bottom-up" sensory information in which phonemic content is considered critical. Evidence, however, has accumulated indicating that cognitive and linguistic capabilities ("top-down" information) also play a major role in speech recognition. Three investigations are reported in which the effects of lexical discrimination on word recognition were examined in normal and hearing-impaired listeners. The conceptual framework for these studies is based on the Neighborhood Activation Model (NAM) developed by Luce and Pisoni during the past decade. This theory suggests that words are recognized relationally in the context of other phonemically similar words. Specifically, three lexical factors influencing word recognition have been identified; (1) neighborhood density, the number of phonemically similar words activated by the stimulus, (2) neighborhood frequency, the frequency of occurrence of a target word in mental lexicon. In the investigations, monosyllables (consonant-vowel-consonant), were partitioned into eight experimental conditions created by combining two levels (high and low) of each of the three lexical factors. The similarity neighborhood structure was estimated computationally from a computer-readable dictionary developed by Pisoni. The results documented the significance of the lexical factors in quiet and noise conditions and for listeners with hearing impairment. Future development of speech recognition tests should allow for the effects of higher level properties (lexical factors) on lower level acoustic-phonemic processing. |