Network dynamics underlying alterations in apparent object size

Abstract

A target circle surrounded by small circles looks larger than an identical circle surrounded by large circles (termed as the Ebbinghaus illusion). While previous research has shown that both early and high-level visual regions are involved in the generation of the illusion, it remains unclear how these regions work together to modulate the illusion effect. Here, we used functional MRI and dynamic causal modelling to investigate the neural networks underlying the illusion in conditions where the focus of attention was manipulated via participants directing their attention to and maintain fixation on only one of the two illusory configurations at a time. Behavioural findings confirmed the presence of the illusion. Accordingly, functional MRI activity in the extrastriate cortex accounted for the illusory effects: apparently larger circles elicited greater activation than apparently smaller circles. Interestingly, this spread of activity for size overestimation was accompanied by a decrease in the inhibitory self-connection in the extrastriate region, and an increase in the feedback connectivity from the precuneus to the extrastriate region. These findings demonstrate that the representation of apparent object size relies on feedback projections from higher- to lower-level visual areas, highlighting the crucial role of top-down signals in conscious visual perception.

Keywords: DCM; Ebbinghaus illusion; fMRI; feedback projection.

Graphical Abstract

Figure 1

Schematic representation of the experimental procedures. (A) In the behavioural experiment, the stimuli consisted of four large and four small circles as well as a central target simultaneously presented in one display. A comparison circle was presented below the illusory configurations. (B) In the fMRI experiment, the illusory configurations and a white fixation cross were simultaneously presented for 10 s, then the fixation was replaced by the central target for 12 s.

Figure 2

Results of behavioural performance, group level GLM and PPI analysis. (A) The perceived size of the central target (%) when presented among large inducers and small inducers (n = 16). (B) Brain activations in response to underestimation and overestimation conditions, and (C) BOLD response in the occipital region of interest (ROI) as a function of illusory conditions (n = 20, paired t-test). (D) Results of PPI analysis with the seed of the occipital ROI for the underestimation and overestimationconditions (P < 0.001, uncorrected, cluster size = 20 voxels). Underestimation is in red, and overestimation is in yellow. Error bars represent one standard error of the mean. Asterisks indicate a significance level of *P < 0.05, and ***P < 0.001. The white arrow indicates a common region (i.e. right precuneus) that showed correlation with the occipital ROI for both illusory conditions.

Figure 3

Model space and DCM results. (A) The model space consisted of three models, which had the same basic architecture and differed with respect to the driving input region. (B) The connection strength of the winning model was derived by comparing overestimation with underestimation condition (n = 20, paired t-test). The values indicate connectivity strength and asterisks indicate a significance level of ***P < 0.001.