Figure 2.

Illustration of RSA. A, Simplified conceptual RDM. LDC dissimilarities are computed between every possible pair of events (six cued tasks × four steps), generating a 24 × 24 RDM. Diagonal cells are zero by definition as they reflect dissimilarity between identical events. Off-diagonal cells reflect pattern dissimilarity between events that always differ in search item, with varying additional differences in room, cued task, and step. These included event pairs that shared the same cued task (red cells), shared the same room but different cued task (purple cells), or differed in both room and cued task (orange cells). Saturation indicates the difference in steps between event pairs. B, Hypothetical pattern dissimilarities resulting from room, task, and item representation across step differences. Item representation can be estimated as the intercept, i.e., estimated LDC dissimilarity in in the absence of room, task, or step differences. C, Full model RDM. LDC dissimilarities were computed between every possible pair of event types (six cued tasks × four steps × three distractor tasks), generating a 72 × 72 RDM. Colors as in A. Saturation indicates the difference in steps between event pairs, and brightness indicates the difference in the possible stimuli presented in the visual search arrays (see main text). The six tasks are labeled K1, K2, K3, B1, B2, B3 (with K indicating a kitchen task and B indicating a bathroom task). The asterisk indicates the only case where item is matched; this was excluded from the model fit, so that the intercept could be interpreted as item representation.