An efficient Bayesian observer model of attractive and repulsive temporal context effects when perceiving multistable...

Area of Science:

• Visual perception
• Computational neuroscience
• Psychophysics

Background:

• Multistable dot lattices exhibit context effects, where perception is influenced by previous stimuli.
• Existing models explain these effects but lack a detailed underlying process.
• Temporal context effects, both attractive and repulsive, are observed in visual perception tasks.

Purpose of the Study:

• To test if an efficient Bayesian observer model can explain concurrent attractive and repulsive context effects.
• To investigate the role of efficient encoding and perceptual priors in these effects.
• To model the correlation between individual differences in attractive and repulsive effects.

Main Methods:

• A simulation study using an efficient Bayesian observer model.
• The model incorporates variable encoding precision based on stimulus frequency (efficient encoding).
• The model considers the dissimilarity between stimulus and sensory spaces and includes stimulus and perceptual levels.

Main Results:

• An efficient Bayesian observer model with both stimulus and perceptual levels explained the co-occurrence of attractive and repulsive temporal context effects.
• The model reproduced the positive correlation between individuals' attractive and repulsive effects.
• Efficient encoding and likelihood repulsion at the stimulus level explained repulsive effects, while perceptual prior attraction explained attractive effects.

Conclusions:

• Efficient encoding and likelihood repulsion on the stimulus level explain repulsive context effects in multistable dot lattices.
• Perceptual prior attraction explains attractive temporal context effects in these tasks.
• The findings support an integrated model of visual perception accounting for both stimulus-driven and internal factors.