Self-regulated spacing in a massive open online course is related to better learning

The study investigates whether the well-established spacing effect—distributing learning across multiple sessions—extends to real-world, self-regulated learning in an educationally authentic context. Traditional laboratory research shows that spaced study enhances long-term retention compared to massed study, and that optimal spacing depends on the retention interval. However, less is known about how spacing operates when learners control their own study in complex courses, where content is related rather than identical and timelines span weeks. The research questions are: (1) Is self-regulated spacing associated with better learning outcomes in a MOOC? (2) Which learners are more likely to space their study? (3) Do benefits of spacing depend on learner ability and engagement in active practice? The purpose is to evaluate spacing at classroom-relevant timescales under learner control, and to understand individual differences that moderate spacing’s benefits, thereby informing recommendations for study strategies in authentic learning environments.

The spacing effect, first documented over a century ago, is robust across domains and ages, with benefits particularly evident at delayed tests. In classic paradigms, spacing is operationalized as temporal lags between repeated exposures to the same content, and optimal spacing scales with the retention interval (longer delays benefit from longer spacing gaps). Prior work (e.g., Cepeda et al.) shows a temporal ridge of optimal spacing relative to test delay. In educational contexts, imposed or scheduled spacing improves outcomes (e.g., personalized review schedules in language courses). Yet, self-directed learning can alter how and when strategies like spacing are used and how effective they are. Some studies indicate differences between imposed versus self-paced spacing conditions, suggesting that learner control may moderate spacing benefits. Thus, there is a need to test spacing under naturalistic, self-regulated conditions with complex materials and longer timescales, while considering individual differences (e.g., ability, engagement) and the interplay between spacing and other strategies such as active practice.

Design and data source: Observational analysis of an online Introductory Psychology MOOC (Coursera, Georgia Institute of Technology; Spring 2013). Data were retrieved from CMU’s DataShop repository (dataset 863). IRB approvals covered data archiving and analysis; data were anonymous. Participants and course: 615 students consented; 437 completed the course (final exam) and were included in analyses. The 12-week course comprised 12 units plus an initial learning strategies unit. Materials included an online textbook (214 pages), 645 embedded activities (e.g., fill-in-the-blank, multiple-choice, drag-and-drop), 127 images, and 43 videos (video access data not analyzed). Each week a multiple-choice quiz was released (11 total; unit 12 assessed only on the final exam). Quizzes counted for 30% of the final grade (lowest dropped); the final exam (released June 10–15, 2013) was 40%; written assignments comprised 30%. Measures: - Prior knowledge: Pretest of 20 true/false questions before course start (M = 10.93, SD = 3.46, range 2–20). - Outcomes: 11 unit quiz scores; final exam score (M = 27.72, SD = 5.50, range 14–35). Average unit quiz and final exam scores correlated (r = 0.69, p < 0.001). Pretest correlated positively with average quiz (r = 0.19, p < 0.001) and final exam (r = 0.20, p < 0.001). - Study behaviors from clickstream logs: Every page load and activity response recorded with timestamps. • Spacing: Number of sessions used to complete a unit (greater sessions = more spacing). Descriptives: reported mean 40.8 sessions per unit (SD = 3.17; median = 3). • Time spent: Sum of durations of all sessions within a unit. Students took on average 2 days to complete each unit (SD = 3 days; median = 1 day). • Retention interval (RI): Time between the timestamp of the last studied page for a quiz-related activity and the start of the unit quiz (M = 45.73 h; SD = 1464.89; median = 26.12 min). • Activity completion rate: Count of embedded activities completed per unit. Statistical analyses: Mixed-effects regressions with unit quiz grade (or final exam grade) as the dependent variable. Fixed effects included spacing (sessions), retention interval, and where relevant, interactions (spacing × retention interval; spacing × ability; spacing × activities). Pretest score and total time spent in unit were included as covariates. Random intercepts for student and unit accounted for between-student and between-unit variability, allowing within-student inference on spacing effects. Predictors were z-scored; two-tailed significance tests. Supplementary Tables contain full model summaries.

- Spacing predicts better unit quiz performance: β = 0.10, SE = 0.02, t(6593.84) = 6.41, p < 0.001, controlling for pretest, time-on-unit, student and unit random effects. - Retention interval main effect: Shorter RI associated with higher quiz grades, β = -0.07, SE = 0.01, t(6666.91) = -5.19, p < 0.001. Spacing × RI interaction not significant, β = 0.02, SE = 0.01, t(16512) = 1.28, p > 0.20. - Spacing predicts higher final exam performance: β = 0.13, SE = 0.07, t(727) = 2.76, p < 0.01, controlling for total time, RI, and pretest. - Who spaces more? Students with higher final exam scores spaced more: spacing predicted by exam grade, β = 0.06, SE = 0.02, t(787.7) = 2.82, p = 0.005 (controlling for pretest and time-on-unit). - Ability moderates spacing benefits: Spacing still beneficial overall, β = 0.08, SE = 0.01, t(644) = 5.24, p < 0.001, but spacing × final exam grade interaction significant, β = -0.04, SE = 0.01, t(607) = -1.18, p = 0.02; benefits of spacing were larger for lower-ability students. - Activity engagement and spacing: Completing more activities predicted more spacing, β = 0.35, SE = 0.01, t(428) = 28.06, p < 0.001. - Spacing × activities interaction for quiz performance: Significant interaction, β = -0.03, SE = 0.01, t(6415) = -2.71, p = 0.007. Spacing’s positive relation to quiz performance was stronger when students completed fewer activities (i.e., spacing buffered against lower engagement in active practice). - Descriptives: Average time to complete a unit ~2 days; median retention interval ~26 minutes; pretest positively but modestly correlated with course outcomes.

The findings demonstrate that when learners self-regulate their study in a semester-long MOOC, greater spacing—operationalized as completing a unit across more study sessions—is associated with better performance on unit quizzes and the final exam, above and beyond prior knowledge and time-on-task. This extends laboratory evidence of the spacing effect to an authentic educational context where learners control pacing and scheduling. Importantly, individual differences shape both the use and impact of spacing: higher-ability students tended to space more, yet the benefits of spacing were more pronounced for lower-ability students, suggesting that spacing can serve as a compensatory strategy that buffers lower-ability learners. Spacing also interacted with active practice: while completing more embedded activities generally improved quiz performance, the benefit of spacing was greatest when activity completion was low, implying partially overlapping mechanisms (e.g., retrieval) and non-additive effects. Collectively, the results indicate that self-regulated spacing is a valuable strategy in real-world learning and that encouraging spacing may be especially impactful for learners less inclined to engage in other active learning behaviors.

This study provides evidence that self-regulated spacing during a real MOOC is positively related to learning outcomes at both unit and course levels. It shows that: (a) learners who space more perform better; (b) higher-ability students are more likely to space; and (c) spacing particularly benefits lower-ability students and those who complete fewer practice activities, potentially by leveraging retrieval-based mechanisms. Contributions include translating the spacing effect to educationally meaningful timescales under learner control and elucidating individual differences in spacing use and efficacy. Future research could experimentally promote spacing in authentic courses to test causality, investigate optimal spacing schedules relative to real-world retention intervals, explore personalized spacing recommendations, and examine how spacing combines with other active strategies (e.g., practice testing) across learner profiles.

- Causality and third variables: As an observational study, unmeasured factors (e.g., motivation, time management, life events) may influence both spacing and performance, and cannot be fully ruled out despite within-student controls and covariates. - Ability proxy: Final exam score used as a proxy for ability may be confounded with study behaviors during the course; pretest provided converging but less sensitive evidence due to lower variability and scope. - Selection of completers: Analyses focused on students who completed the final exam, potentially limiting generalizability to all enrollees. - Measurement constraints: Retention interval estimation from clickstream timestamps may be noisy (very large SD). Spacing defined as number of sessions may conflate spacing with session granularity choices. - Content coverage: Analyses focused on interactions with the online textbook and embedded activities; video engagement data were not analyzed, which may omit relevant study behaviors. - Course/context specificity: Single MOOC and subject area (introductory psychology) may limit generalizability to other courses or platforms.