Olya Bulatova¹, Keisuke Fukuda^1,2; ¹University of Toronto, ²University of Toronto Mississauga

Predictions influence perception by biasing the percept in line with the expectation (Kok et al., 2012; 2018). However, this prediction-induced bias has primarily been observed by eliciting predictions through learned statistical regularities. Therefore, it is not clear whether arbitrary and explicit predictions would also bias our perception. To test this, we had participants perform a categorization task after making an arbitrary and explicit prediction on what they were about to see. More precisely, on each trial, participants first made an explicit binary prediction as to whether they would see one of the two objects (e.g., “Dog” or “Boar”) by clicking either a top or bottom buttons (e.g., “Dog” and “Boar” buttons) displayed on the computer screen. Subsequently, a morph of the two objects was briefly presented (50ms), and participants categorized the object as one of the two objects by clicking either the left or right buttons (e.g., “Dog” and “Boar” buttons) displayed on the computer screen. To avoid motor priming, the response options for predictions and categorizations were orthogonalized and fully counter-balanced on a trial-by-trial basis. Here, we found that participants’ categorization was indeed biased in the direction of their predictions. For instance, 50% morph objects (i.e., a morph of 50% Dog and 50% Boar) were more likely to be categorized as the predicted object than the counterpart, and we replicated this finding in two other object pairs (i.e., “Face and Tree” and “Gecko and Branch”, Stöttinger et al., 2015). Our results extend the prediction-induced biases to arbitrary and explicit predictions, and the results of an ongoing EEG study examining the origin of prediction-induced biases (i.e., do predictions bias perception or categorization decisions?) will be discussed.

Acknowledgements: This research was supported by the Natural Sciences and Engineering Research Council (5009170).