Abstract
The cause of spontaneous perceptual reversals of visually ambiguous stimuli (e.g., Rubin’s faces-vase) has been attributed to several different mechanisms. One longstanding hypothesis suggests that adaptation progressively reduces the strength of the dominant interpretation’s neural representation until a perceptual reversal occurs and an alternative interpretation becomes dominant. Although there is behavioural and computational evidence that adaptation to an unambiguous stimulus (i.e., adaptor) can bias subsequent perception of an ambiguous one, it is unclear whether adaptation plays a role in spontaneous reversals where no explicit adaptor has been presented. To address this issue, we presented participants with ambiguous stimuli under different levels of adaptation to a preceding unambiguous stimulus (e.g., faces/vase preceded by unambiguous face stimulus). Using multivariate pattern analysis of human EEG data, we tested whether the occurence of spontaneous reversals in sequences of fully ambiguous stimuli could be decoded better by a classifier trained on perceptual reversals that happened in a high or low adaptation state after exposure to an unambiguous adaptor. Our results show that spontaneous reversals can be decoded only by the high adaptation state classifier, presumably because of the representational similarity between spontaneous perceptual reversals and the adapted state of the visual system. These results present persuasive neural evidence in favour of the hypothesis that “neural fatigue” or adaptation plays a role in spontaneous perceptual reversals in the absence of a clear unambiguous adaptor stimulus.