Fig. 4: Spatial and decisional uncertainty models.
From: Audiovisual adaptation is expressed in spatial and decisional codes
a Neural models: The spatial (hemifield) model encodes spatial location in the relative activity of two subpopulations of neurons each broadly tuned either to the ipsi- or contra-lateral hemifield. The ratio of the ipsi- and contralaterally tuned neurons was set to 30%/70% consistent with prior research8. The decisional uncertainty model encodes observers’ decisional uncertainty as a non-linear function of the distance between the spatial estimates and the spatial classification boundary. b Predicted mean BOLD-response as a function of sound location along the azimuth in a left hemisphere region for pre-, postVA- and postAV-adaptation. The spatial model predicts a BOLD-response that increases linearly for sound locations along the azimuth. The decisional uncertainty model predicts BOLD-response that decays with the distance from the decision boundary in an inverted U-shaped function. Further, this inverted U-shaped function shifts along the azimuth when spatial estimates are recalibrated. The model predictions were obtained by averaging the simulated neural activities across 360 neurons in a left hemisphere region. c Predicted representational dissimilarity matrices (RDM) based on the individual model neural activity profiles across spatial locations (−12° to 12°) and experimental phases (pre, postVA and postAV). We simulated RDMs from the spatial (left) and the decisional (right) model for (i) top row: no recalibration, i.e. without a representational shift and (ii) bottom row: with recalibration, i.e. with a representational shift. Solid white lines delineate the sub-RDM matrices that show the representational dissimilarities for different stimulus locations within and between different experimental phases. Diagonal dashed white lines highlight the RDM dissimilarity values for two identical physical locations of the postAV- and the postVA-adaptation phases. Comparing the RDMs with and without recalibration along those dashed white lines shows how the shift in spatial representations towards the previously presented visual stimulus alters the representational dissimilarity of corresponding stimulus locations in postAV- and postVA-adaptation phases, while the off-diagonals show the dissimilarity values for neighbouring spatial locations. Source data are provided as a Source Data file.
