Poster C76, Thursday, November 9, 10:00 – 11:15 am, Harborview and Loch Raven Ballrooms

Cortical entrainment depends on temporal predictability, not periodicity

Geoffrey Brookshire1, Daniel Casasanto1,2;1University of Chicago, 2Cornell University

When people perceive language, low-frequency cortical oscillations (< 8 Hz) entrain to rhythms in the stimulus. Entrainment to language may be a neural strategy to boost perceptual sensitivity to informational peaks in a time-varying signal. What neural processes support cortical entrainment to language? Here we test two possible answers to this question. First, neural entrainment may occur when intrinsic cortical oscillations become synchronized to rhythmic stimuli, analogous to mechanical resonance between coupled oscillators. Second, entrainment may reflect active neural predictions: The brain may form predictions of upcoming events, and adjust the phase of ongoing broadband oscillations to ensure that perceptual systems are maximally sensitive when important information occurs. To test these hypotheses, we recorded electroencephalography (EEG) while participants watched sequences that varied in their temporal predictability and periodicity. If entrainment relies on coupling of intrinsic oscillations, then cerebral cortex should only entrain to periodic stimuli. Alternatively, if entrainment depends on active, flexible temporal predictions, then cerebral cortex should entrain to any stimuli that are temporally predictable. We recorded EEG from human subjects (N=11) while they watched sequences of flashing abstract images. Participants performed a vigilance task in which they pressed a button whenever a specific image appeared. Inter-stimulus intervals (ISIs) comprised 4 approximately logarithmically-spaced values (100, 167, 300, 500 ms) in a similar frequency band as variability in the volume envelope of speech (2–10 Hz). Stimuli appeared in three conditions. In the periodic and predictable (Periodic-Predictable) condition, all stimuli within a block were separated by the same ISI, giving uniform and perfectly predictable timing. In the aperiodic and unpredictable (Aperiodic-Unpredictable) condition, the ISIs were randomly shuffled. In the final condition, stimulus onsets were not periodic at the frequency of the individual stimuli, but remained perfectly predictable (Aperiodic-Predictable). To construct sequences in this condition, we created a group of the 4 distinct ISIs. This ISI group was then repeated back-to-back (e.g. ABCDABCD…). Because every ISI uniquely predicts the following ISI, this condition is perfectly predictable despite between aperiodic on the level of individual items. We measured entrainment by computing the cross-correlation between stimulus onsets and EEG activity in each electrode. Stimulus onsets were coded as impulse responses and matched with EEG recordings using a photodiode. The magnitude of the cross-correlation at any given time-lag indicates how strongly the EEG signal predicts a stimulus onset. Cross-correlations were stronger for Periodic-Predictable sequences than for Aperiodic-Unpredictable sequences. Furthermore, cross-correlations were stronger for Aperiodic-Predictable sequences than for Aperiodic-Unpredictable sequences. Predictability boosted entrainment at both positive and negative lags, indicating that cortical activity reflects predictable stimuli before they occur. However, item-wise periodicity did not have any effect on entrainment beyond that of temporal predictability; differences between Periodic-Predictable sequences and Aperiodic-Predictable sequences were weak and varied by timing-lag. People entrain more strongly to predictable sequences than to random sequences, both before and after the stimulus appears, even when those sequences are not periodic at the item frequency. We conclude that cortical entrainment does not rely on coupled oscillators. Instead, entrainment reflects active and flexible cortical predictions.

Topic Area: Perception: Speech Perception and Audiovisual Integration

Back to Poster Schedule