14th International Congress of Phonetic Sciences (ICPhS-14)

San Francisco, CA, USA
August 1-7, 1999


Invariance of Lexical Context Use in Spoken Word Recognition

Theresa Hnath-Chisolm

University of South Florida, USA

Determining a sensory aid increases access to acoustic speech patterns typically involves measuring percent correct wholeword recognition for open-set monosyllabic word lists. Thus, not only is access to acoustic/phonetic information assessed, but also lexical knowledge. These effects can be separated by application of simple probability theory, in which the recognition probabilities for the wholes (Pw) is compared to those for the parts (Pp). The result is a j-value indicating the number of independent perceptual units necessary for correct word recognition. An underlying assumption of j is that it is a constant - that it is not affected by the underlying recognition probabilities of the parts. This assumption was examined by varying type and availability of sensory information through: (1) stimuli presentation via three input-modalities; and, (2) presentation by multiple speakers. Results confirmed j-values were independent of speaker but not modality manipulations, suggesting the speech recognition task may be approached differently as a function of input-modality.

Full Paper

Bibliographic reference.  Hnath-Chisolm, Theresa (1999): "Invariance of lexical context use in spoken word recognition", In ICPhS-14, 579-582.