Poster E19, Saturday, August 18, 3:00 – 4:45 pm, Room 2000AB
Phonetic content of auditory representations
Ryan Rhodes1, Chao Han1;1University of Delaware
Past studies have used the varying standards MMN approach to investigate the content of auditory representations (e.g. Dehaene-Lambertz & Pena, 2001; Shestakova et al., 2002). Different tokens introduce variance in many dimensions simultaneously, such as pitch, timbre, intensity, and VOT. Studies using this paradigm claim that this variance – most of which is orthogonal to phoneme category membership – enforces a phoneme representation. The variance in these dimensions (pitch, intensity, etc.) prevents the auditory processing system from creating a representation that contains phonetic detail. Instead, a phoneme category representation (e.g. /t/) is retrieved from long term memory and used to make predictions about incoming sounds. This experiment investigates the limits of the variance that can be used to enforce a phoneme representation. We used a varying standards MMN design to elicit a within-category MMN effect, which indicates that the representation used by the auditory processing system contains detailed phonetic information about the sound on the dimension that determines category membership. We used a varying standards oddball design, presented in two blocks: a variable-standard condition, and a static-variable condition. Standards and deviants appeared in a ratio of 9:1. All stimuli were single syllable CV tokens (ta). The standards in the variable-standard block varied on a specified acoustic dimension (syllable pitch) while the oddball differed (within-category) on a different dimension (onset VOT). In both blocks, the consonant had a VOT of 95ms for standards and 55ms for deviants. The baseline standard and the deviant had a pitch ranging from 116Hz-97Hz over the syllable. Four other standard tokens shifted this pitch contour ±10Hz or ±20Hz. EEG data was collected from 20 subjects. Two temporal and spatial regions of interest were identified via PCA: an early component peaking at 200ms (corresponding to P2), and a later component peaking at 300ms (corresponding to N2). A repeated measures ANOVA found a significant main effect of mismatch in both time windows (p = .023 at 200ms; p = .001 at 300ms), but no significant effect of pitch condition or interaction. In the early time window, a paired t-test found a significant within-category MMN in the static pitch condition (p = .027; one-tailed) and no significant MMN in the variable pitch condition (p = .6; one-tailed). In the later time window, t-tests found a highly significant effect of mismatch in both the static pitch (p = .006) and variable pitch condition (p = .003). These results indicate that the auditory processing system was sensitive to a within-category VOT difference despite variance on the unrelated dimension of pitch contour. The presence of pitch variance did not enforce a phoneme representation and cause the auditory processing system to lose phonetic information. References Dehaene-Lambertz, G., & Pena, M. (2001). Electrophysiological evidence for automatic phonetic processing in neonates. Cognitive Neuroscience and Neuropsychology, 12(14), 3155–3158. Shestakova, A., Brattico, C. A. E., Huotilainen, M., Galunov, V., Soloviev, A., Sams, M., Nììtìnen, R. (2002). Abstract phoneme representations in the left temporal cortex: magnetic mismatch negativity study. Cognitive Neuroscience and Neuropsychology, 13(0), 1–5.
Topic Area: Perception: Auditory