Fig. 3: Design and results of Experiment 3 & 4.

a Schematic diagram of the task in Experiment 3. Two conditions were tested: one with events containing four items (Event 4) and the other containing six items (Event 6). Three matched pair types were tested: Pair type 1 marked by the red and yellow square brackets, Pair type 2 marked by green and light green square brackets and Pair type 3, marked by blue and light blue square brackets in the two conditions. b Box plots of group averaged temporal order memory for the three pair types (n = 32). No significant difference was found between the Event 4 condition and the Event 6 condition for pair type 1 (t(31) = 0.06437, p = 0.9491, two-sided, q = 0.3322, FDR corrected for multiple comparisons); however, TOM was significantly better for the Event 4 than for Event 6 condition for both pair type 2 (t(31) = 2.334, p = 0.0262, two-sided, q < 0.05, FDR corrected) and pair type 3 (t(31) = 3.93, p < 0.001, two-sided, q < 0.05, FDR corrected). c Schematic diagram of the task in Experiment 4. Two sequence types were tested containing the same number of event boundaries but segmented differently, e.g., 336336… vs. 633633… items per event. Three pair types were tested: within-event pairs taking earlier and later local positions (matched by absolute list positions) denoted by the red and yellow square brackets (short for Within Early and Within Late respectively). Across-event pairs taking earlier and later local positions (matched by absolute list positions) denoted by the blue and light blue square brackets (short for Across Early and Across Late respectively). Across-event pairs with a longer lag than pairs marked by blue and light blue square brackets, are marked by light grey square brackets. d. Box plots of group averaged temporal order memory for within and across-event pairs taking earlier and later positions shown in Fig. 3c (n = 30). TOM was significantly better for within vs. across-event pairs (F(1,29) = 8.217, p = 0.0076, η² = 4.657%), and for early vs. late pairs (F(1,29) = 8.933, p = 0.0057, η² = 2.503%). Planned contrasts further confirmed that TOM was significantly better for earlier vs. later pairs for both pair types (for within event pairs: t(29) = 2.256, p = 0.0318, two-sided, q < 0.05, FDR corrected; for across-event pairs: t(29) = 3.839, p < 0.001, two-sided, q < 0.05, FDR corrected). The boxes show the inter-quartile range (IQR) and the median. Whiskers in box plots represent the minimum and maximum in the dataset. The asterisk (*) represents statistical significance at p < 0.05. Numbers in red denote event boundaries in Fig. 3a, c. Source data are provided as a Source Data file.