Fig. 1
From: The neural dynamics of hierarchical Bayesian causal inference in multisensory perception
Example trial, experimental design and behavioral as well as EEG data. a Example trial of the flash-beep paradigm (e.g. two flashes and four beeps are shown) in which participants either report the number of flashes or beeps. b The experimental design factorially manipulated the number of beeps (i.e. one to four), number of flashes (i.e. one to four) and the task relevance of the sensory modality (report number of visual flashes vs. auditory beeps). To characterize the computational principles of the Bayesian Causal Inference (BCI) model, we reorganized these conditions into a two (task relevance: auditory vs. visual report) × two (numeric disparity: high vs. low) factorial design for the GLM-based analysis of the behavioral and EEG data (e.g. audiovisual weight index). c The behavioral audiovisual weight index wAV (across-participants circular mean and bootstrapped 68% CI; n = 23) is shown as a function of numeric disparity (small: ≤1 vs. large: ≥2) and task relevance (auditory vs. visual report). wAV was computed for participants’ numeric reports (solid) and the BCI model’s predicted reports (dashed). wAV = 90° for purely visual and wAV = 0° for purely auditory influence. d Event-related potentials (ERPs; across-participants mean; n = 23) elicited by one to four stimuli in audiovisual congruent conditions averaged across parietal electrodes and the ERP topography at the peak (averaged over conditions with one to four stimuli). The x axis shows the stimulus onsets. e Difference between ERPs elicited by two beeps and one flash and the sum of the corresponding unisensory ERPs (i.e. V1A2- (A2 + V1), black dotted), the unisensory auditory ‘two beeps’ (A2, green), visual ‘one flash’ (V1, red), audiovisual (A2V1, black solid) and the sum of the unisensory (A2 + V1, pink) averaged across occipital electrodes. A positive component was significant from 65–150 ms (p = 0.040, two-sided cluster-based corrected randomization t22 test; see horizontal dashed line) and a negative component was significant 335–730 ms after stimulus onset (p < 0.001). The ERP topography at the peak of the positive component is shown. Source data are provided as a Source Data file
