Figure 5
From: Robust encoding of scene anticipation during human spatial navigation
Sequential decoding in two motion decision task (MD) blocks.
(a,b) Scene reconstruction from single-scan fMRI activity (2 s) over a single MD block, but before this participant (participant 1) saw the actual scene. The left panel indicates the trajectory that was taken by the participant, starting from an initial state (red arrow) to a goal (blue circle). A digit on the map shows the scan index corresponding to the horizontal axis in the middle panel. The middle panels indicate the series of correct scene views (upper) and predicted scene views (lower). The hemodynamic response delay was not compensated for in this decoding analysis. L, forward-left; C, forward-centre; R, forward-right. (c) Time shift Hamming distances, which signify the discrepancy between a real scene (not necessarily the present one) and the predicted one decoded from current brain activity. The top and bottom panels correspond to the MD blocks in (a) and (b), respectively. (d) Distance between the code words predicted for the scenes around the goal position based on the initial three scans and the code words of the corresponding true scenes. The optimal encoding model was used. In the blocks with detour trajectories, the distance was longer than the median of the other trajectories (white line). The bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. A point located beyond 1.5 times the box height was defined as an outlier.
