Intelligibility and simulated deaf-like segmental and timing errors

An experiment tested the effect on intelligibility of modifying high-quality synthetic speech in two independent ways, each of which made the speech more like the speech of one particular deaf child. This child´s phonology is described and two component errors from his speech pattern are used as examples and tested. The two component errors are (1) consonantal omissions and substitutions and (2) suprasegmental timing distortions. Sentences with and without errors in consonant articulation were synthesized with and without timing distortions. Listeners identified the stimuli resulting from these deaf-like speech components, alone and in combination. Four different indices of intelligibility were extracted from the listeners´ responses.