Figure 2. Absolute directional coding in posterior parahippocampal gyrus.

(a) Analysis logic of the one-fold directional analysis for three example trials. High pattern similarity was predicted for pairs of trials sampling similar directions with a maximum angular difference of 30° (red) compared to trials sampling directions 60° or more apart (blue). Note that for illustration purposes the predicted similarity matrix is shown for comparisons across conditions, not single trials. (b) Searchlight results show a significant cluster of voxels in the posterior parahippocampal gyrus (peak voxel MNI coordinates: 34 -34 -10; T₂₃ = 4.82, p = 0.024 corrected for multiple comparisons using small-volume correction) with higher pattern similarity for trials sampling similar directions compared to trials sampling dissimilar directions. Results are shown on the structural MNI template. For display purposes, the statistical map is thresholded at p<0.001 uncorrected. (c) Sagittal and coronal view of the mask used to correct for multiple comparisons (see Materials and methods) displayed on the MNI template brain.

DOI: http://dx.doi.org/10.7554/eLife.17089.008

Figure 2—figure supplement 1. Increased pattern similarity for similar directions after excluding trial pairs sampling a direction with the same combination of buildings.

To control for the possibility of increased pattern similarity in the parahippocampal gyrus for similar direction pairs being due to the imagination of identical scenes, we excluded trial pairs from the analysis in which the same combination of start and target building was used. This revealed a significant cluster of voxels (peak voxel MNI coordinates: 34 -36 -10, T₂₃ = 4.54, p = 0.042 corrected for multiple comparisons using small volume correction) very similar to the one observed in the main analysis (see Figure 2b). For display purposes, the statistical map is thresholded at p<0.001 uncorrected.

DOI: http://dx.doi.org/10.7554/eLife.17089.010

Figure 2—figure supplement 2. Absolute directional coding during imagination is independent of locations and distances in Donderstown.

(a) Pattern similarity difference between trial pairs sampling similar and dissimilar directions in the peak voxel of the main absolute directional coding analysis (Figure 2) after exclusion of comparisons with trials using the same start (bar I) and the same target location (bar II). Both T₂₃ > 3.00, both p<0.007. (b) We considered the distances between start and target locations in a trial pair by controlling for three distance measures, which differed between pairs of trials sampling similar and dissimilar directions. The distance measures are illustrated based on two example trials. For each trial pair, we calculated (I) the mean length of the vectors connecting start and target location, (II) the difference in length of the two vectors and (III) the mean length of the vectors connecting all four relevant locations of a pair. In separate GLMs we used the distance measures as predictors of pairwise pattern similarity. (c) Mean pattern similarity difference in peak voxel of cluster from main absolute directional coding analysis (Figure 2) between trial pairs sampling similar directions and pairs sampling dissimilar directions computed on the residuals of the GLMs. With this approach we controlled for pattern similarity due to (I) the average distance from start to target location in a trial pair, (II) the difference in distance from start to target location in a trial pair and (III) the average distance between all four buildings in a trial pair (see Materials and methods). All T₂₃ > 3.60, all p<0.001. Bars in a and c show mean pattern similarity difference with error bars reflecting S.E.M., dashed line shows mean pattern similarity difference in peak voxel from main analysis.

DOI: http://dx.doi.org/10.7554/eLife.17089.011