Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Irregular click trains reveal the oscillatory nature of auditory neural responses
Poster Session B, Wednesday, September 30, 4:30 - 6:30 pm, Wangari Maathai
Peter Donhauser1, David Poeppel2; 1Ernst Strüngmann Institut, Frankfurt, Germany, 2New York University
Neural oscillations have been suggested as a putative neural mechanism supporting segmentation at different linguistic scales. Critics argue that the apparent oscillatory nature of auditory neural responses reduces to sequences of evoked responses. We argue that the focus on either natural speech or isochronous auditory sequences results in sampling a restricted range of neural dynamics and limits progress on this debate. We therefore recorded MEG while participants (n=19) were presented with aperiodic, Poisson-distributed auditory click trains. The spectrum of these stimuli is uniform over a wide range of frequencies, allowing an unbiased analysis of the nature of resulting neural responses. We compute coherence of neural responses and click trains and parameterize the resulting frequency spectra. All participants’ spectra show oscillatory peaks despite flat stimulus spectra. These cluster into gamma (50 Hz), alpha (10 Hz) and theta (5 Hz) peaks, with the gamma band being particularly pronounced and consistent across participants. To characterize these gamma band responses, we isolate spatial components displaying gamma peaks using canonical correlation analysis. Modeling the continuous neural response to click trains using TRFs, we find that they display a sequence of peaks and troughs resembling a damped harmonic oscillation. The latency of these corresponds to the well-known mid-latency auditory evoked responses. Comparing click pairs with different ISIs, we find that the linear approximation (TRF) only fits the observed data for ISIs of roughly one gamma cycle (~20ms), whereas it overestimates response amplitudes for shorter and underestimates response amplitudes for longer ISIs. We therefore suggest that the oscillation and evoked response view are compatible with each other, potentially describing the same physiological process. However, we argue not that ‘the oscillation is an evoked response’ but rather that ‘the evoked response is an oscillation’: sequences of peaks and troughs - like the mid-latency evoked responses - can be parsimoniously described as the impulse response of an underlying gamma oscillator. Although the responses of this oscillator can be approximated by a linear convolution of a damped oscillatory impulse response, probing the oscillator with an irregular stimulus reveals clear deviations from a linear convolution model.
Topic Areas: Methods,