Objective: Expectations can strongly influence perceptual processes, with unexpected stimuli being detected slower than expected ones. However, the neural mechanisms underlying these behavioral effects remain controversial with some neuroimaging studies implicating facilitatory, others suppressive influences. In this regard, predictive processing models can offer a currently underutilized framework.

Methods: We implemented a three-layered recurrent cortical model (Heeger, 2017) and modeled our previous human data (n= 8, Ürgen & Boyacı, 2021) to investigate the computational mechanisms underlying the behavioral effects of expectation. Using three previous fMRI studies (1: Egner et al. 2010; 2: Kok et al. 2011; 3: Aitken et al. 2020), we also tested whether model predictions match neural evidence. We simulated efMRI experiments with our optimized parameters and obtained BOLD responses with GLM. Results were analyzed with 2 (trial type: expected, unexpected) X 2 (expectation validity: 75, 50) repeated- measures ANOVAs and post-hoc t-tests.

Results: Model-fitting results revealed a main effect of expectation (F(1,7)=18.511, p=0.004) with significantly higher iteration numbers for unexpected trials (t(7)=3.220, p=0.015). Thus, when actual input differs from expectations, the sensory process requires prolonged computations. Simulated BOLD responses also mostly paralleled empirical data. Consistent with studies on V1 (studies 2 & 3), our model predicted expectation facilitation in lower layers (F(1,7)=5.811, p=0.047; F(1,7)=6.019, p=0.044). On higher-order layers, our model predicted marginally significant expectation facilitation for both stimuli (houses and faces) (F(1,7)=4.973, p=0.061) and overall higher responses to the preferred stimulus type (e.g. FFA: faces). Empirical data on FFA (study 1) corroborates these except showing a trend toward expectation suppression with face stimuli.

Conclusion: Overall, our findings demonstrate that a parsimonious recurrent cortical model can explain both behavioral and neural effects of expectation on sensory processes.

Keywords: Computational modeling, cortical model, expectation, prediction, predictive processing, visual perception