Postersession 1
Poster #: 115
Topic: Speech and language (incl. deficits)
Wednesday, Sep 9, 2015
1st floor

The generation of speech-specific MMN: solutions from dynamic causal modelling

Faith Chiu & Jyrki Tuomainen

Speech, Hearing and Phonetic Sciences, University College London, London, United Kingdom

To further elucidate the neural basis of speech perception – specifically whether the MMN response reflects speech-specific processing – we used pure tones, speech, and spectrally rotated speech as stimuli. Rotated speech is an acoustic manipulation which preserves the acoustic complexity of the original speech signal but renders it unintelligible. The main focus of the study was on the potential difference between speech and rotated speech in the effective connectivity of auditory areas as revealed by Dynamic Causal Modelling using the same modelling approach as Schofield et al. (2009).

Speech and pure tone stimuli used in the passive listening task were obtained from Schofield et al. Rotated speech stimuli were produced by low-pass filtering the speech stimuli at 3500 Hz before performing spectral rotation.

Preliminary results indicate that the DCM solution to spectrally rotated speech differs from both the speech and pure tone conditions, particularly between speech and rotated speech conditions. Especially, the strength in backward connections between left Heschl's gyrus and left superior temporal gyrus is lower in the rotated speech condition. Backward connections are top-down and connect infragranular to agranular layers (cf. Kiebel et al., 2007). We speculate that predictive coding in the rotated speech condition does not access speech categories. The results suggest that acoustic complexity is a necessary but not a sufficient condition to trigger pre-attentive speech-specific (phonetic) processing. These results are in line with studies which have previously investigated the effect of intelligibility in speech processing (Rosen et al., 2011).