In one sentence
In a small Kirk Kerkorian School of Medicine cohort, students with more matured Anki cards and more hours in spaced-repetition practice scored higher on the Comprehensive Basic Science Subject Exam (CBSE).
What the researchers did
Winter and colleagues studied 36 medical students at the Kirk Kerkorian School of Medicine at UNLV during the compressed preclinical phase. Students consented to export deidentified Anki metrics via the Stat Scraper add-on.
Metrics included matured cards (interval > 21 days), total study hours, and unsuspended cards. Researchers correlated these with CBSE scores, stratified students above/below the mean on each metric, and ran independent t-tests, Pearson correlations, and linear regressions.
What they found
- Mean CBSE score was 65.2% (SD 11.6%).
- Linear regression showed a significant positive link between number of mature cards and exam scores.
- CBSE scores also correlated positively with total hours studied (p = 0.013), matured cards (p = 0.002), and unsuspended cards (p = 0.010).
- Students above the mean mature-card count scored 71.5% vs 60.0% for those below.
- Higher total study hours and unsuspended-card counts similarly predicted higher scores in this sample.
What this means for learners and educators
- Track mature cards and consistent review time — not only total cards downloaded — if you use Anki in preclinical years.
- Keeping cards unsuspended (actually in rotation) mattered alongside maturity; a huge suspended library may not help.
- Schools shortening preclinical time can point students to measurable spacing habits rather than vague “study more.”
- Treat associations as planning cues, not proof that Anki alone raises board-style scores for every learner.
Limitations and what we don't know yet
n = 36 at one school limits precision and generalisability. Observational Anki exports cannot prove causation — high performers may both grind cards and score well for other reasons. CBSE is one checkpoint, not Step exams or clerkship grades. Self-selected Anki users who consent to data export may differ from non-users. Exact deck sources and card quality are not captured by count metrics alone.