Fig 1.
Quasi-experimental study design.
Blue represents the cumulative section, and green represents the non-cumulative section.
Table 1.
Sections were generally equivalent.
Fig 2.
Midterms were equivalent in terms of cognitive skill, but topics could not be covered equally.
Panel A: All cumulative (blue) and non-cumulative (green) midterm exam items were categorized by Bloom’s taxonomy, and percent of total is shown on the y axis. Counts were indistinguishable by section (X2(5, n = 144) = 0.46, p = 0.99). Panel B: All midterm exam items categorized by LOCS and HOCS are shown as a percentage of the whole on the y axis. Counts were indistinguishable by section (X2(1, n = 144) < 0.001, p > 0.99). Panel C: Total number of midterm items covering each course topic is shown for each section. Course topics are shown in the order in which they were covered in the course.
Fig 3.
Effect of midterm type on timing of past exam review.
In at attitudinal survey at the end of the semester, students self-reported the number (none, some, or all) of midterm exams that they reviewed prior to the next midterm exam (Panel A; Mann-Whitney U test, U = 12531.5, p = 0.037, n = 299), and in preparation for the final exam (Panel B; Mann-Whitney U test, U = 9241.5, p = 0.003, n = 299).
Fig 4.
Effect of cumulative midterms on exam scores.
Panel A: The y axis shows percent correct for midterms one through seven (calculated using only test items that were shared between cumulative, blue, and non-cumulative, green, sections), the final exam (F), and the retention exam (R) five months after the semester ended. Panel B: Retention exam scores are shown on the y axis and LCTSR scores are on the x axis. Lines are from a simple linear regression and show different relationships between scientific reasoning and retention for the non-cumulative (slope = 3.48, R2 = 0.27, p = 0.003) and cumulative (slope = -0.001, R2 = 0.00, p = 0.99) sections.
Table 2.
Final results multiple linear regression to target student performance (% correct) on exams.
Fig 5.
Effect of cumulative midterms by cognitive skill.
All test items were categorized as requiring LOCS or HOCS and performance is shown by treatment (cumulative midterms = blue, non-cumulative midterms = green). The y axes show scores on shared midterm items that both sections saw the same day (Panel A, n = 344), shared midterm items that the cumulative section saw later (Panel B, n = 310), the final exam (Panel C, n = 344), and the retention exam (Panel D, n = 62).
Fig 6.
Effect of cumulative midterms on early versus late course topics.
Student performance on shared midterm items (Panel A), final exam (Panel B), and retention exam (Panel C) are shown on the y axes by treatment (blue: cumulative midterms, green: non-cumulative midterms). Scores are grouped by topic, with topics displayed in the order in which they were covered in the course.
Table 3.
Final linear mixed models to predict performance on shared midterm items, final exam, and retention exam by topic.