Figure 1
An illustration of a trial sequence execution that was chunked differently across four different iterations of execution. In the first iteration, trials 1 to 6 form chunk 1 and trials 7 to 10 form chunk 2. In the second iteration, the execution is parsed into 3 chunks (trials 1 to 4, trials 5 to 8 and trials 9 and 10). In the third iteration, trials 1 to 3, 4 to 7 and 8 to 10 form chunks 1 to 3. In the fourth iteration, trials 1 to 2, 3 to 8 and 9 to 10 form chunks 1 to 3. If RT peaks at the beginning of new chunks, then the RT pattern will vary across executions that are chunked differently, e.g., trial 7 will show a peak in iteration 1 but not in iteration 2 whereas trial 5 will show a peak in iteration 2 but not in iteration 1. Averaging such RT patterns will smudge the peaks. If the RT patterns are very variable, averaging will show a gradually increasing RT across trials 2 to 10 because the probability of having a peak increases with trial position.
Figure 2
(a) If the entire sequence could be executed by the program assembled at trial 1, then trial 1 RT will be high but RTs across trials 2 to 10 would not differ. (b) If the sequence was always parsed into 3-trial chunks then RTs will additionally peak on trials 4, 7 and 10. But as chunk sizes become variable – 3 to 5 (c), 3 to 6 (d) and 3 to 9 (e), the pattern increasing becomes one of gradual increase across trials 2 to 10.
Figure 3
(Experiment 1) Participants executed 10 trial long sequences. Trials were organized into sequences by having low inter-trial intervals (0.5s) within a sequence and high inter-trial interval (2s) across sequence boundaries. This was further reinforced by a faded digit in the background that went from 10-9-8-…-1 across the 10 trials. (a) In Rule-Switch sessions, they executed sequences made of rule-switch trials. Trial rule was cued by the color of the outer margins: blue – choose the smaller value, green – choose the smaller font. (b) In Stroop sessions participants executed sequences made of Stroop trials where they chose the color of the print of the word from the two choices below.
Figure 4
(Experiment 1) Reaction times, Costs (rule-switching and Stroop) and Error rates from rule-switching (left column) and Stroop (right column) sessions. Dashed lines represent RTs and error rates on rule-switch or Incongruent Stroop-trials. Dotted lines represent these during rule-repeat and Congruent Stroop-trials. The gray dashed and dotted line in top graphs represent averaged reaction times. Error bars in this and all subsequent graphs depict 95% confidence intervals.
Figure 5
(Experiment 2) RT patterns across the seven participants. In each participant, RTs across trials 2 to 15 were significantly different (Table). The RT pattern after averaging across all participants (inset) was identical to that in Experiment 1. Error bars represent 95% confidence intervals.
Figure 6
(Experiment 3) Participants were biased to chunk the larger task episode, consisting of up to 20 trials, into specific sized sub-tasks by making them keep a count of the trials in 3s or 5s (informed by the cue screen at the beginning).
Figure 7
(Experiment 3) RT, switch cost and error rates across the trials of the sequence (rule-switch trials: dashed lines; rule-repeat trials: dotted lines). Error bars: 95% confidence intervals.
Figure 8
(Experiment 4) Two kinds of trial-sequences were executed. (a) Difficult sequences consisted of rule-switch trials whereby rules could change across trials. Number stimuli here were of black color. (b) Easy sequences consisted of only rule-repeat trials. Here the number stimuli were crimson in color.
Figure 9
(Experiment 4) Continuous lines depict performance during difficult sequences made of rule-switch trials, dashed lines depict performance during easy sequences made exclusively of rule-repeat trials. As predicted, trial 1 RT was paradoxically higher for the easy sequences, while RTs on trials 2 to 8 were expectedly higher for difficult sequences.