Responses obtained in consonant percepBon experiments typically show a large variability across sBmuli of the same phoneBc idenBty (Phatak at al., 2008; Sing & Allen, 2012; Toscano & Allen, 2014). The present study invesBgated the influence of different potenBal sources of this response variability. It was disBnguished between source-induced variability, referring to perceptual differences caused by acousBcal differences in the speech tokens and/or the masking noise tokens, and receiverrelated variability, referring to perceptual differences caused by within- and across-listener uncertainty. It can be demonstrated that any physical change in the sBmuli had a measurable effect. This holds even for slight Bme-shics in the steady-state masking-noise waveform. Furthermore, responses obtained with idenBcal sBmuli differed substanBally across different normal-hearing listeners, while individual listeners were able to reproduce their responses fairly reliably. To determine how well the source-induced variability is reflec te d i n diffe re n t a u di to ry -i n s pi re d i n te r nal representaBons (IRs), the corresponding perceptual distances were compared to the distances between the IRs of the sBmuli. Several variants of an energy-based IR and a modulaBon-based IR were considered. The results suggest that a normalized modulaBon-based representaBon provides the best match to the perceptual data.
|Number of pages||1|
|Publication status||Published - 2015|
|Event||169th Meeting of the Acoustical Society of America - Wyndham Grand Pittsburgh Downtown Hotel, Pittsburgh, Pa, United States|
Duration: 18 May 2015 → 22 May 2015
|Conference||169th Meeting of the Acoustical Society of America|
|Location||Wyndham Grand Pittsburgh Downtown Hotel|
|Period||18/05/2015 → 22/05/2015|