Using error-estimation to probe the psychological processes underlying contextual interference effects.

Abstract

Although the learning benefits of interleaved practice schedules relative to blocked schedules are well-reported, the mechanisms underlying these effects are not fully understood. Researchers have generally suggested that random schedules of practice increase task-related information processing which arises due to switching between variations of the same task (or switching between different tasks). Thus, one potentially useful way to both probe and manipulate contextual interference is to pair it with error estimation during practice. Forced error estimation increases task-related information processing and recording these estimates provides insight into learners' self-awareness of their errors. In the present study, 84 participants were randomly allocated to four groups. Participants practiced a timing task under blocked or random schedules, with and without error estimations prior to feedback. During the acquisition phase, three target times were trained (1500, 1700, 1900ms), with feedback delivered after every trial. We used delayed post-tests (24 hrs later) to evaluate the retention of these target times and their transfer to two new target times (1600, 1800ms). Participants who practiced with a random schedule performed worse (i.e., greater absolute error) than those with a blocked schedule during acquisition (p=.006); however, randomly scheduled participants also showed reduced error (p=.004) on the retention and transfer tests. Although prompting error estimations led to greater self-reported mental effort being invested on the task (p=.001), error estimation was not reliably associated with superior learning (p=.133). The accuracy of error estimations did not differ as a function of practice structure (p=.070), although the accuracy of error estimations improved during acquisition (p=.006). Findings highlight the robustness of the contextual interference effect, but we did not find evidence that error estimations moderated the effect on this task. It is in some ways surprising that we found an effect of contextual interference, as past-work suggests that interference effects are attenuated (or eliminated) when participants switch between different parameters of the same task. We speculate that this might be due to the difficulty of the task; even though participants switched between parametric variations of the same task, the distinction between parameters was subtle (i.e., tenths of a second).