How prior knowledge prepares perception: Alpha-band oscillations carry perceptual expectations and influence early visual responses

Abstract

Perceptual experience results from a complex interplay of bottom-up input and prior knowledge about the world, yet the extent to which knowledge affects perception, the neural mechanisms underlying these effects, and the stages of processing at which these two sources of information converge, are still unclear. In several experiments we show that language, in the form of verbal cues, both aids recognition of ambiguous "Mooney" images and improves objective visual discrimination performance in a match/non-match task. We then used electroencephalography (EEG) to better understand the mechanisms of this effect. The improved discrimination of images previously labeled was accompanied by a larger occipital-parietal P1 evoked response to the meaningful versus meaningless target stimuli.Time-frequency analysis of the interval between the cue and the target stimulus revealed increases in the power of posterior alpha-band (8-14 Hz) oscillations when the meaning of the stimuli to be compared was trained. The magnitude of the pre-target alpha difference and the P1 amplitude difference were positively correlated across individuals. These results suggest that prior knowledge prepares the brain for upcoming perception via the modulation of alpha-band oscillations, and that this preparatory state influences early (~120 ms) stages of visual processing.Knowledge impacts early perceptual processing appear are reflected in alpha dynamics [36][37][38] . Recently, Mayer and colleagues demonstrated that when the identity of a target letter could be predicted, pre-target alpha power increased over left-lateralized posterior sensors 39 . These findings suggest that alpha-band dynamics are involved in establishing perceptual predictions in anticipation of perception.Here, we examined whether verbal cues that offered no direct perceptual hints can improve visual recognition of indeterminate two-tone Mooney images (Experiment 1). We then measured whether such verbally ascribed meaning affected an objective visual discrimination task (Experiments 2-3). Finally, we recorded electroencephalography (EEG) during the visual discrimination task (Experiment 4) to better understand the locus at which knowledge influenced perception. Our findings suggest that using language to ascribe meaning to ambiguous images impacts early visual processing by biasing pre-target neural activity in the alpha-band. Materials and Method Experiment 1Materials. We constructed 71 Mooney images by superimposing familiar images of easily nameable and common artefacts and animals onto patterned background. These superimposed images were then blurred (Gaussian Blur) and then thresholded to a black-and-white bitmap. Materials are available at https://osf.io/stvgy/.Participants. All participants for Experiments 1A-1C were recruited from Amazon Mechanical Turk and were paid $1 (Experiments 1A and 1B), or $0.50 (Experiment 1C) for participating. Demographic information was not collected. All studies were approved by the University of Wisconsin-Madison Institutional Review Boar...

Distinct oscillatory frequencies underlie excitability of human occipital and parietal cortex

“…For instance, visual-evoked multiunit and gamma-band activity in macaque inferotemporal cortex were found to be positively correlated with prestimulus alpha power in the local-field potential, which was also found to increase when attention was paid to visual, as opposed to auditory, input (Mo et al, 2011). Similarly, in humans, predictions about an upcoming stimulus have been shown to increase prestimulus alpha power yet result in larger evoked responses -with the two processes being positively correlated (Mayer et al, 2015;Samaha et al, 2016a). A possible explanation for our finding that prestimulus beta power, rather than alpha, was found to predict PPC TMS phosphenes is suggested by recent work investigating oscillatory responses to TMS of different cortical regions.…”

Trial‐by‐trial co‐variation of pre‐stimulus EEG alpha power and visuospatial bias reflects a mixture of stochastic and deterministic effects

“…() recently identified a correlation between pre‐stimulus oscillatory power (in alpha, beta and low gamma) and subsequent saccadic response time which was explained by both short‐term, stochastic (trial‐by‐trial) and long‐term, deterministic (trial order, fatigue) sources of variance with the relative degree of each contributor differing across brain regions. Additionally, recent studies provide evidence that pre‐stimulus alpha oscillations encode biases of upcoming sensory decisions induced by top‐down predictions (Mayer et al ., ; Samaha et al ., ) and decisions on preceding trials (De Lange et al ., ). Hence, because some short and long‐term changes in EEG characteristics systematically correlate with changes in psychophysical performance, these should be taken into consideration when interpreting trial‐by‐trial oscillatory predictors of perception.…”

(Not so) Great Expectations: Listening to Foreign-Accented Speech Reduces the Brain’s Anticipatory Processes

“…Rather, we believe that higher-level conceptual features are likely to have been activated from the highly predictive sentential context at a later stage, yielding a "false" memory trace. Furthermore, we do not exclude the possibility that the activation of higher-level conceptual features can lead to the activation of lower-level perceptual feature as demonstrated in cuing/priming paradigms Lupyan, 2013, 2017;Boutonnet and Lupyan, 2015;Samaha et al, 2016). Our results are therefore in line with previous reports of such effects on memory by Foucart et al (2015) and compatible with the evidence showing that degraded speech or phonemes can be restored to the extent that their perceptual experience might not differ from that in optimal conditions (Warren, 1970;Groppe et al, 2010;Sohoglu et al, 2012;Bendixen et al, 2014) and fit very well into theories of perception which allow for higher-level representation to feedback (top-down) information to lower levels.…”

“…In language processing, but also in a wide array of other cognitive domains (Kok et al, 2012b(Kok et al, , 2014Boutonnet and Lupyan, 2015;Samaha et al, 2016;Vandenbroucke et al, 2016), comprehenders actively anticipate upcoming information thereby (pre-)activating specific features of such linguistic information, ranging from basic acoustic features to high-level conceptualsemantic ones (Federmeier and Kutas, 1999;DeLong et al, 2005;Van Berkum et al, 2005;Obleser et al, 2007;Schiller et al, 2009;Vinck et al, 2011;McGettigan et al, 2012;Van Berkum, 2013;Foucart et al, 2014Foucart et al, , 2015DeLong and Kutas, 2016), in order to facilitate and reduce the processing load (Pickering and Garrod, 2007). For instance, Schiller et al (2009) present electrophysiological data from two experiments demonstrating that listeners make predictions for upcoming words using a speech-error detection task.…”