Multimodal evidence challenges the effectiveness of probabilistic cueing for establishing sensory expectations.

Abstract

Predictive coding theories posit a reduction in error-signaling neural activity when incoming sensory input matches existing expectations-a phenomenon termed expectation suppression. However, the empirical evidence for expectation suppression, as well as its underlying neural mechanism, is contentious. A further aspect of predictive coding that remains untested is how predictions are integrated across sensorimotor domains. To investigate these two questions, we employed a novel cross-domain probabilistic cueing paradigm, where participants were presented with both visual and motor cues within a single trial. These cues manipulated the orientation and temporal expectancy of target stimuli with 75% validity. Participants completed a reproduction task where they rotated a bar to match the orientation of the target stimulus while their neural and pupil responses were respectively measured via electroencephalography and eye tracking. Our results showed a consistent, feature-unspecific effect of motor expectancy across multiple measures, while evidence for visual expectancy was limited. However, neither motor nor visual expectancy modulated the fidelity of sensory representations. These results indicate that violations of temporal expectancy in the current study may reveal the brain's intrinsic sensitivity to temporal regularities in the natural settings, rather than feature-specific predictions. In contrast, the absence of visual expectancy effects in both neural and pupillometry results adds to a growing body of evidence questioning the effectiveness of probabilistic cueing paradigms for establishing expectations capable of altering sensory representations. Due to null findings in the visual and sensory representation analyses, we did not further investigate cross-domain prediction integration.