Neural signatures of perceptual inference

Abstract

Generative models, such as predictive coding, posit that perception results from a combination of sensory input and prior prediction, each weighted by its precision (inverse variance), with incongruence between these termed prediction error (deviation from prediction) or surprise (negative log probability of the sensory input). However, direct evidence for such a system, and the physiological basis of its computations, is lacking. Using an auditory stimulus whose pitch value changed according to specific rules, we controlled and separated the three key computational variables underlying perception, and discovered, using direct recordings from human auditory cortex, that surprise due to prediction violations is encoded by local field potential oscillations in the gamma band (>30 Hz), changes to predictions in the beta band (12-30 Hz), and that the precision of predictions appears to quantitatively relate to alpha band oscillations (8-12 Hz). These results confirm oscillatory codes for critical aspects of generative models of perception.DOI: http://dx.doi.org/10.7554/eLife.11476.001

Oscillatory Mechanisms of Stimulus Processing and Selection in the Visual and Auditory Systems: State-of-the-Art, Speculations and Suggestions

“…We emphasize that the switch between “entrainment-mode” and “alpha-mode,” as described above (Lakatos et al, 2016), has so far only been demonstrated during rhythmic stimulation. It was speculated that the “alpha-mode” can be activated—despite the regular stimulation—due to lapses in attention to external stimuli, leading to an increase of internal attention (an idea that was formulated already by Ray and Cole, 1985), in agreement with the other studies cited above (Leske et al, 2014; Pachitariu et al, 2015; Keitel and Gross, 2016; Sedley et al, 2016). However, in principle, a dominance of the alpha band when external input is (supposedly) ignored—and therefore virtually “ absent ” for the brain—might also mean that the alpha band dominates in the “true” absence of regular input.…”

“…Moreover, (reduced) alpha power in the auditory system has been linked with the perception of illusionary phenomena, such as the Zwicker tone, an illusionary tone that is perceived for several seconds after the offset of broadband noise with a spectral gap (Leske et al, 2014). Finally, using intracranial recordings in human auditory cortex and an experimental protocol during which expectations had to be updated continuously, Sedley et al (2016) showed that alpha power is related to the confidence (or precision) of their listeners' predictions (and thus related to internal processes) but not necessarily to the stimulus input itself.…”

An Integrative Tinnitus Model Based on Sensory Precision

“…Based on known neuronal microcircuits [39], simulations, and empirical observations 40, 41, 42, 43, 44, 45, 46, predictions and prediction errors have been linked to oscillations in specific frequency bands. State units are thought to be located in infragranular (deep) layers, and generate predictions using beta and other low-frequency oscillations.…”

“…This phenomenon is termed synchronous gain, and is particularly sensitive to synchrony within the gamma band 51, 52. A role for lower-frequency oscillations is much less well established; recent empirical evidence has linked low-frequency oscillation magnitude to the precision of internal representations [41] including, in the case of alpha oscillations, the precision of the prediction that a sensory change will not occur [53]. This latter claim has challenged the widely-held view that alpha oscillations simply modulate cortical excitability.…”

Distinct roles of theta and alpha oscillations in the involuntary capture of goal-directed attention

“…This alpha lateralization is maintained as long as attention is directed to one side of space (Kelly et al 2006), the magnitude of lateralization tracks the likelihood of targets appearing at the cued location (Bauer et al 2014;Gould et al 2011), and predicts subsequent perceptual outcomes (e.g., Händel et al 2011;Thut et al 2006). The frequency of lateralized activity can spread up to ~25 Hz (the beta range) in visual attention tasks (e.g., Bauer et al 2014), but it is currently unclear whether this is simply an extension of the alpha range (Michalareas et al 2016), or whether this beta activity reflects distinct processes (e.g., Sedley et al 2016).…”