You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

 
Poster A73, Wednesday, November 8, 10:30 – 11:45 am, Harborview and Loch Raven Ballrooms

Stress-timing via Oscillatory Phase-locking in Naturalistic Language

Phillip M. Alday1,2, Andrea E. Martin1,3;1Max-Planck-Institute for Psycholinguistics, 2University of South Australia, 3University of Edinburgh

Linking hypotheses between cortical oscillations and the hierarchical structure of speech and language posit a correspondence across multiple timescales and levels of representation: fine speech structure is represented in the gamma band, while the speech envelope, i.e. syllables and words, in the alpha and theta bands (Ding et al., 2016; Giraud & Poeppel, 2012; Peelle & Davis, 2012). However, these studies have largely focused on oscillatory power as well as auditory power (i.e. the speech envelope) and have ignored oscillatory phase. Oscillatory phase has also been proposed to entrain to speech rhythms, leading to optimised sampling of the speech signal (Gross et al. 2013). However, these proposals fail to account for acoutic-phonological realizations of basic linguistic structure, such as phonological stress. Here we used pitch and intensity information as automatically measurable acoustic proxies for phonological stress (using Praat) and compared them to oscillatory phase across the frequency spectrum from 1 to 30 Hz. The EEG data were reanalyzed from Alday et al. (biorXiv) with 50 participants listening to a 23 minute long story in German read by a trained native speaker. Correlating the sine of the oscillatory phase with pitch and intensity revealed a strong peaks around 3--4 Hz and secondary peak around 7--8 Hz. The 4 Hz peak provides converging evidence that linguistic stress serves as a time-locking event for speech-tracking in the theta band, i.e. that linguistic stress serves as a phase-resetting event to (re-)align the stimulus rhythm with cortical rhythms, which are unable to change their period. The peak in the low alpha band maches previous studies which have shown that data stress is reliably detected in the time domain, i.e. ERPs, which are thought to reflect alpha phase resetting at least in part (cf. Knaus 2013, Min et al. 2007, Hanslmayr et al. 2007). Linguistic stress thus reflects the fundamental organizational unit in spoken language, working as a pacemaker and providing necessary structure for the frequency domain. In natural speech, asynchronous speech, syllables are irregular in length and timing and provide a poor fundamental unit. Instead stress provides the timing foundation upon which further phonological structure, e.g. syllables, and then morphosyntactic structure can built. Here, we demonstrated that the acoustic realizations of stress match the patterns predicted for syllables in oscillatory dynamics.

Topic Area: Perception: Auditory

Back to Poster Schedule