Fig. 4: Computational model.

a Schematic of the model. The context signal was defined as a vector of binary elements, which drifts at rate p when there is no event boundary and resets at rate λ at event boundaries. b, c shows representational dissimilarity (1-R) for each time point relative to all the other time points averaged over 1000 iterations of the simulation of the model containing no event boundaries (b) and the model containing 5 event boundaries (i.e., event boundary occurred every 6 time points, c), respectively. d Schematic of the memory index (MI) used to quantify recency judgments. To make a recency judgment, the learning contexts of each probed object is reinstated, and compared to the contextual representation of the first time point during encoding (i.e., d1 and d2, which refer to the representation dissimilarity between each of the probed items and the reference item). The one more dissimilar with the reference point (quantified by d2-d1) is judged as the item presented later. arb. units refers to arbitrary units. Source data are provided as a Source Data file.