A map of abstract relational knowledge in the human hippocampal–entorhinal cortex
- Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, United Kingdom
- University of Oxford, United Kingdom
- Max Planck-UCL Centre for Computational Psychiatry and Ageing Research, United Kingdom
Figures
Figure 1 with 1 supplement
Experimental design.
(A) Graph structure used to generate stimulus sequences on day 1. Trial transitions were drawn from random walks along the graph. (B) Objects on reduced graph presented to subjects in the scanner on …
Figure 1—figure supplement 1
Task performance.
(A) Response times (ms) and (B) performance on the orientation judgment cover task performed during training on day 1 for each of the 12 blocks. (C) Graph structure indicating the object position. (D…
Figure 2 with 3 supplements
Functional magnetic resonance imaging adaptation in the hippocampal–entorhinal system decreases with distance on the graph.
(A) Whole-brain analysis showing a decrease in functional magnetic resonance imaging adaptation with link distance in the hippocampal–entorhinal system, thresholded at p<0.01, uncorrected for …
Figure 2—figure supplement 1
Anatomically defined regions of interest used for small-volume correction.
(A) Mask comprising the bilateral entorhinal cortex and subiculum, received with thanks from Chadwick et al. (2015). (B) Mask comprising the bilateral entorhinal cortex, hippocampus, and …
Figure 2—figure supplement 2
Distance-dependent scaling of neural activity is specific to the hippocampal–entorhinal system.
(A) All areas displaying a decrease in functional magnetic resonance imaging adaptation with graph distance. A cluster in the subgenual cortex did not survive whole-brain correction for multiple …
Figure 2—figure supplement 3
Effects of object familiarity.
(A) Number of times each object was presented during training, averaged across participants. Note that this measure is directly related to the number of neighbours an object has on the graph. During …
Figure 3 with 3 supplements
Relational information is organised as a map.
(A) Linear regression on neural activity with number of links and average time between two objects during training as regressors (t22 = 3.29, p=0.003 and t22 = 1.27, p=0.22). (B) Absolute difference …
Figure 3—figure supplement 1
The distance-dependent scaling cannot be driven by a main effect of object position.
Position-specific activity in the region of interest (ROI) defined according to (A) a connected < non-connected contrast (ROI 1, Figure 2B, green, F6,132 = 0.88, p=0.5), (C) a link 2 < link 3 …
Figure 3—figure supplement 2
Map characteristics in a null distribution generated by permuting the links making up the graph structure.
(A) Correlation of link distances for graphs making up the null distribution with distances resulting from performing multidimensional scaling on the neural data. The null distribution was …
Figure 3—figure supplement 3
Distance effects in an anatomically defined region of interest comprising the entorhinal cortex and the subiculum.
(A) Activity scales linearly with link distance (F2,44 = 8.41, p=0.0008). Post-hoc pairwise comparisons revealed a significant difference between distances of lengths 1 and 2 (t22 = 2.36, p=0.03), …
Figure 4 with 1 supplement
Functional magnetic resonance imaging adaptation in the hippocampal–entorhinal system is consistent with predictive representations of relational knowledge.
(A) Visualisation of communicability coordinates for the graph structure by performing multidimensional scaling on the communicability matrix. (B) Whole-brain regression of communicability onto …
Figure 4—figure supplement 1
Activity in the hippocampal–entorhinal system is consistent with the successor representation.
(A) Visualisation of successor representation coordinates for the graph structure by performing multidimensional scaling on the negative of the successor representation, with the free parameter …
Figure 5
Response times reflect graph structure.
(A) Graph structure used to generate stimulus sequences on day 1. Trial transitions were drawn from random walks along the graph structure. (B) Objects on reduced graph presented to subjects on day …
Author response image 1
Visualisation of communicability coordinates for the graph structure by performing multi- dimensional scaling on the communicability matrix.
https://doi.org/10.7554/eLife.17086.016
Author response image 2
