Practice schedules affect how learners correct their errors

Secondary analysis from a contextual interference study

Abstract

Contextual interference is one of the most established effects in motor learning research; random practice schedules are associated with poorer performance (in the short-term) but superior learning (in the longer-term) when compared to block practice schedules. However, the way this interference affects learners on a trial-to-trial basis remains poorly understood. We present a secondary data analysis of N=84 healthy young adults, replicating the contextual interference effect in a time estimation task. We used the determinant of a correlation matrix to measure the amount of order in participants’ responses. The determinant is conceptually equivalent to the unexplained variance (1-r²) but applies to higher dimensional spaces. We calculated this determinant in two different phase spaces: (1) Trial Space, which was the determinant of the previous 5 trials (lagged constant error 0-4); and (2) Target Space, the determinant of the previous 5 trials of the same target. The distinction in phase space is critical because for blocked practice the previous trial is almost always the same target, but for random practice the previous trial is almost never the same target. In Trial Space, there was no significant difference between groups (p=0.98) and no Group x Lag interaction (p=0.54), although there was an effect of Lag (p<0.01). In Target Space, there were effects of Group (p=0.02), Lag (p<0.01), and a Group x Lag interaction (p=0.03). Participants who practiced using random schedules showed smaller determinants overall, which got smaller as more past trials were included (i.e., increasingly correlated responses). This increase in orderliness was due to the random group having positively correlated errors from trial-to-trial in Target Space. We argue this “response inertia” in the random practice group suggests a greater reliance on the retrieval of the target time from memory. Data from the novel analyses presented herein support the reconstruction account of the contextual interference effect and help integrate the effect with other learning principles in psychology (e.g., retrieval practice being beneficial for long-term recall).