Talker information influences spectral contrast effects in speech categorization

Abstract

Spectral contrast effects, the perceptual magnification of spectral differences between sounds, have been widely shown to influence speech categorization. However, whether talker information alters spectral contrast effects was recently debated [Laing, Liu, Lotto, and Holt, Front. Psychol. 3, 1-9 (2012)]. Here, contributions of reliable spectral properties, between-talker and within-talker variability to spectral contrast effects in vowel categorization were investigated. Listeners heard sentences in three conditions (One Talker/One Sentence, One Talker/200 Sentences, 200 Talkers/200 Sentences) followed by a target vowel (varying from /ɪ/-/ɛ/ in F1, spoken by a single talker). Low-F1 or high-F1 frequency regions in the sentences were amplified to encourage /ɛ/ or /ɪ/ responses, respectively. When sentences contained large reliable spectral peaks (+20 dB; experiment 1), all contrast effect magnitudes were comparable. Talker information did not alter contrast effects following large spectral peaks, which were likely attributed to an external source (e.g., communication channel) rather than talkers. When sentences contained modest reliable spectral peaks (+5 dB; experiment 2), contrast effects were smaller following 200 Talkers/200 Sentences compared to single-talker conditions. Constant recalibration to new talkers reduced listeners' sensitivity to modest spectral peaks, diminishing contrast effects. Results bridge conflicting reports of whether talker information influences spectral contrast effects in speech categorization.

Consonant categorization exhibits a graded influence of surrounding spectral context

“…This reaffirms arguments that low-level acoustic properties play the primary role in producing SCEs and not higher-level linguistic properties of the precursor sentence (Mitterer, 2006;Laing et al, 2012). This point is also supported by SCEs being produced by a different sentence on every trial (Assgari and Stilp, 2015), unintelligible time-reversed speech (Watkins, 1991), and nonspeech contexts (Holt, 2005(Holt, , 2006Laing et al, 2012).…”

“…An exclusionary criterion of failing to achieve at least 80% correct on CV series endpoints was applied (Assgari and Stilp, 2015). This resulted in the removal of one listener's data, so analyses were based on responses from the remaining 18 listeners.…”

Cognitive load makes speech sound fast, but does not modulate acoustic context effects

“…This would involve a comparison between a target sound and its expected realization given a certain context (see, e.g., Assgari & Stilp, 2015;Barreda, 2012;Glidden & Assmann, 2004;Johnson, 1990;Johnson, Strand, & D'Imperio, 1999;Nusbaum & Magnuson, 1997;Nusbaum & Morin, 1992). Support for this interpretation comes from the finding that foreign-accented speech as well as speech that contains fast-speech processes such as segmental reductions and deletions is implicitly perceived as faster (Bosker & Reinisch, 2015;Reinisch, 2016a), possibly due to greater listening effort.…”

Effects of talker variability on spectral contrast effects.

“…This way of inducing talker variability may not fully represent differences found in actual talkers (Pollack et al, 1954;Watkins, 1991). A more recent study showed that a change in talker can significantly diminish spectral contrast effects and thus challenges this notion that talker information plays no role in spectral contrast effects (Assgari & Stilp, 2015). These experiments EFFECTS OF TALKER ON SPECTRAL CONTRAST EFFECTS 6 added either a +5 dB or a +20 dB spectral peak to sentences from one talker versus 200 different talkers.…”

“…Even simple compilations of sine tones (or "tone histories") that vary in duration and number of tones can be used to produce the effect (Holt, 2005;Holt, 2006 Stilp et al, 2015). Such large peaks may not be common in natural settings (Assgari & Stilp, 2015). In experiments done by Stilp et al (2015), a significant contrast effect was observed with as small as +5 dB filter gain.…”