Fig. 7
From: The neural dynamics of hierarchical Bayesian causal inference in multisensory perception
The effect of prior numeric disparity on the BCI model’s causal prior (pcommon) and prestimulus alpha power. a Effect of previous numeric disparity on causal prior pcommon (across-participants’ mean ± within-participants’ SEM, n = 23) as a function of the numeric disparity (small: ≤1 vs. large: ≥2) of the previous trial of order 1–5. The effect of previous numeric disparity on causal prior decays with increasing trial order. Asterisks denote statistical significance in two-sided one-sample t22 tests (**: p < 0.01, *: p < 0.05, (*): p < 0.1, n.s.: not significant). b Mediation and moderation: Previous numeric disparity and pre-stimulus alpha power significantly predict pcommon. However, previous numeric disparity does not predict pre-stimulus alpha power. Hence, pre-stimulus alpha power does not mediate the effect of numeric disparity on the causal prior. Instead, we observed a marginally significant interaction between previous numeric disparity and alpha power suggesting that previous numeric disparity moderates the effect of pre-stimulus alpha power on pcommon. c Mediation: Effect of previous numeric disparity on alpha power: Time-frequency t-value map (n = 23, averaged over occipital electrodes) reveals no significant difference in prestimulus alpha power between small vs. large previous numeric disparity (i.e. trial order 1). All clusters p > 0.05 (two-sided cluster-based corrected randomization t22 test). d Moderation: Interaction effect between previous numeric disparity and alpha power on causal prior: Time-frequency t-value maps (n = 23) for the correlation between pcommon and the alpha power averaged across occipital electrodes over deciles for trials with large (lower panel) or small (middle panel) numeric disparity of the previous trial and the difference in correlations between small and large disparity of the previous trial (top panel). We observed only a marginally significant interaction between previous disparity and alpha power on pcommon (p < 0.1; two-sided cluster-based corrected randomization t22 test demarcated by dashed line in top panel). Prestimulus alpha power correlated significantly with observers’ causal prior only if the previous trial was of a small numeric disparity (p < 0.05; demarcated by a solid line, middle panel). Source data are provided as a Source Data file
