Previous
Next
Emily Herman, Matt Carlson, Angelica Brill and Anne Olmstead – Did you say what I thought you said? Probing illusions in speech perception under different conditions
September 3, 2021
9:00 am

Emily Herman, Matt Carlson, Angelica Brill and Anne Olmstead – Did you say what I thought you said? Probing illusions in speech perception under different conditions

Did you say what I thought you said? Probing illusions in speech perception under different conditions

Restrictions on how speech sound sequencing lead to well-documented "misperceptions" of non-native sound sequences. Listeners may appear to perceive additional sounds that are absent from the acoustic signal, to ignore certain features that are present acoustically, or to interpret some features differently, as compared to listeners whose native languages allow more transparent representations of the acoustic signal. Focusing on so-called perceptual illusions, where listeners seem to perceive an extra segment, we explore the hypothesis that these illusions are dependent on the context of processing. We take our cue from an early study by Miyawake, et al (1975) in which Japanese listeners discriminated formant contours encoding the American English [r-l] contrast easily when presented as non-speech sounds, but exhibited the now well-documented difficulty in discriminating the same contrast when presented in a speech context. In this work in progress, we will describe three experiments on Spanish speakers' perception of word-initial [s]-consonant sequences, which do not occur in Spanish, and which are perceived by listeners from this population as beginning with [e], despite the absence of any vocalic material in the acoustic signal (e.g. acoustic [spona] is perceived as [espona], Carlson, et al., 2015; Carlson, 2018; 2019, inter alios). In the first experiment, we use a discrimination task using Spanish-sounding nonwords to ask how much phonological material is necessary to trigger the illusion, showing that it was not enough to present phonetic evidence for the sC cluster, but the illusion only emerged when the stimuli presented could constitute a word in Spanish. In the second experiment, we used lexical decision to test English words such as school to which an initial vowel was sometimes added (e.g. eschooloschool), and the items were presented in either a native American English accented voice or a Spanish accented voice. This tested whether listeners could distinguish whether an intrusive vowel had been produced, at least by a speaker who would be otherwise unlikely to produce one (i.e. the native AE talker). This yielded no evidence that Spanish speaking listeners applied this kind of sociolinguistic knowledge in this task: they appeared to ignore the intrusive [e], but not an intrusive [o], regardless of the talker. Nevertheless, the third experiment showed that they discriminated tokens with intrusive [e] vs. no intrusive vowel above chance. Experiment 3 was similar to Experiment 1, suggesting that discrimination may be better when listening to English speech than when listening to Spanish, and we are conducting follow-up experiments to confirm this result. Our findings provide evidence that perceptual illusions such as that experienced by Spanish listeners hearing sC clusters depend on the processing evnvironment, including task demands and, more interestingly, the language being listened to and sociolinguistic information about the talker.