(a) Single trial parameter estimates were generated by building a single model with a separate regressor for each trial. (b) Subfields were demarcated manually to create separate ROIs for CA3/DG, CA1, Subiculum, and PHG. (c) The searchlight classifier was trained using single trial estimates from half of the retrieval blocks and tested on the remaining retrieval data. Training/testing was repeated for all searchlight spheres in each subjects MTLs, creating subject specific statistical maps. (d) Within-city similarity was assessed for each ROI by extracting the trial parameter estimates from the subfields and correlating between matched trials of a city’s “A” and “B” retrieval blocks. (e) Between-city similarity was calculated consistent with within-city similarity.
DOI:
http://dx.doi.org/10.7554/eLife.10499.006