Effects of interleaved and blocked study on delayed test of category learning generalization

Abstract

Studying different concepts by frequently alternating between them (i.e., interleaving), improves discriminative contrast between different categories, while studying each concept in separate blocks emphasizes the similarities within each category. Interleaved study has been shown to improve learning of high similarity categories by increasing between-category comparison, while blocked study improves learning of low similarity categories by increasing within-category comparison. In addition, interleaved study presents greater temporal spacing between repetitions of each category compared to blocked study, which might present long-term memory benefits. In this study we asked if the benefits of temporal spacing would interact with the benefits of sequencing for making comparisons when testing was delayed, particularly for low similarity categories. Blocked study might be predicted to promote noticing similarities across members of the same category and result in short-term benefits. However, the increase in temporal delay between repetitions inherent to interleaved study might benefit both types of categories when tested after a longer retention interval. Participants studied categories either interleaved or blocked and were tested immediately and 24 h after study. We found an interaction between schedule of study and the type of category studied, which is consistent with the differential emphasis promoted by each sequential schedule. However, increasing the retention interval did not modulate this interaction or resulted in improved performance for interleaved study. Overall, this indicates that the benefit of interleaving is not primarily due to temporal spacing during study, but rather due to the cross-category comparisons that interleaving facilitates. We discuss the benefits of temporal spacing of repetitions in the context of sequential study and how it can be integrated with the attentional bias hypothesis proposed by Carvalho and Goldstone (2014a).

Keywords: blocking; category learning; comparison; interleaving; spacing.

FIGURE 1

Examples of stimuli used in the experiments presented here. All stimuli were created by randomly generating curvilinear segments that were then added together. Each blob was constituted by eight features (each feature was a specific spatial position in the blob).

FIGURE 2

Performance during the study session and the refresher component for Experiment 1A. The left panel shows performance for interleaved and blocked study of high similarity categories. The right panel shows performance for interleaved and blocked study for low similarity categories. Error bars indicate standard errors of the means. Chance-level performance in this task was 0.33. The vertical dashed line represents session break.

FIGURE 3

Performance in the transfer tests of Experiment 1A. The left panel depicts results for items studied during the study phase while the right panel presents results for items not seen during the study phase. Results for the high similarity categories are presented in red while results for the low similarity categories are presented in blue. For each of these, the dashed lines represent blocked study while the solid lines represent interleaved study. Error bars indicate standard errors of the means. Chance-level performance in this task was 0.33 and is represented in the graphs by the black dashed line.

FIGURE 4

Performance during the Study Phase of Experiment 1B for high and low similarity categories (left panel). The right panel presents data from the second day refresher only presented in Figure 2 and is depicted here for comparison purposes. Solid lines indicate interleaved study, while dashed lines indicate blocked study. Error bars indicate standard errors of the means. Chance-level performance in this task was 0.33 and is represented in the graphs by the horizontal dashed line.

FIGURE 5

Performance in the transfer task of Experiment 1B (left panel). The right panel depicts results for the delayed transfer of Experiment 1A for comparison purposes. Results for the high similarity categories are presented in red while results for the low similarity categories are presented in blue. For each of these, the dashed lines represent blocked study while the solid lines represent interleaved study. Error bars indicate standard errors of the means. Chance-level performance in this task was 0.33 and is represented in the graphs by the black dashed line.