Schedule of Events | Symposia

Predictions do not modulate the perception and time-resolved representation of objects

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Phuong Dang¹ (phuong.dang@uqconnect.edu.au), Margaret Moore, Amanda Robinson, Jason Mattingley; ¹The University of Queensland, ²Queensland Brain Institute, ³School of Psychology, ⁴School of Psychology and Queensland Brain Institute

According to predictive coding accounts of brain function, efficient processing of sensory information is achieved by modelling current and future perceptual states. Previous research has shown that neural representations of low-level visual features, such as orientations, are modulated by predictions. It is unclear, however, whether the same is true for more complex, real-world objects. To address this gap, we recorded brain activity using EEG while participants (N = 40) viewed RSVP sequences containing high-fidelity (intact) and degraded (diffeomorphically warped) object images which were placed in the sequence such that objects were statistically equally likely to occur, expected or unexpected based on a preceding leading image. Participants were asked to report whether target object images were ‘intact’ or ‘warped’. Multivariate pattern analysis was used to quantify the degree of object information represented in neural activity. Decoding accuracy and discrimination task performance were compared across equal-likelihood, expected, and unexpected objects. There were no significant differences in decoding accuracy, discrimination accuracy or discrimination response times amongst the three probability conditions. One potential explanation for these null results is that participants did not have sufficient exposure to the embedded statistical structure; another is that the behavioural task might have interfered with participants’ learning of the image statistics. In a follow-up experiment, we explored these possibilities by increasing exposure to the structured sequences, and employing a modified object detection task that was undertaken separately from the neural recordings. We will discuss the implications of these findings for predictive coding theories of visual perception.

Topic Area: PERCEPTION & ACTION: Vision