November 24, 2020

Differential neural encoding of sound under visual attention is shaped by audiovisual precision and unimodal uncertainty priors

Surroundings continually propagate audiovisual (AV) signals, and by attending we make clear and precise sense of those that matter at any given time. In such cases, parallel visual and auditory contributions may jointly serve as a basis for selection. It is unclear what hierarchical effects arise when initial selection criteria are unimodal, or involve uncertainty. Uncertainty in sensory information is a factor considered in computational models of attention proposing precision weighting as a primary mechanism for selection. The effects of visuospatial selection on auditory processing were investigated here with electroencephalography (EEG). We examined the encoding of random tone pips probabilistically associated to spatially-attended visual changes, via a temporal response function model (TRF) of the auditory EEG timeseries. AV precision, or temporal uncertainty, was manipulated across stimuli while participants sustained endogenous visuospatial attention. TRF data showed that cross-modal modulations were dominated by AV precision between auditory and visual onset times. The roles of unimodal (visuospatial and auditory) uncertainties, each a consequence of non-synchronous AV presentations, were further investigated. The TRF data demonstrated that visuospatial uncertainty in attended sector size determines transfer effects by enabling the visual priming of tones when relevant for auditory segregation, in line with top-down processing timescales. Auditory uncertainty in distractor proportion, on the other hand, determined susceptibility of early tone encoding to automatic change by incoming visual update processing. The findings provide a hierarchical account of the role of uni- and cross-modal sources of uncertainty on the neural encoding of sound dynamics in a multimodal attention task.

 bioRxiv Subject Collection: Neuroscience

 Read More

Leave a Reply

%d bloggers like this: