Learning flashcards with friends: what really helps

"Let's study together" is one of the nicest sentences of exam season and one of the most dangerous. Sometimes it turns into a productive afternoon. Often it turns into three hours of coffee, two slides and a very good feeling of having accomplished a lot.

The research on studying together is more honest than most learning apps admit. Here is what is established, what is not, and how to take the benefits without falling into the traps.

The uncomfortable news first

The most direct study on the topic comes from Imundo, Zung, Whatley and Pan. In three experiments with 152, 141 and 403 students, they compared whether word-definition pairs are better learned with flashcards alone or in pairs. Result: partner learning did not lead to better recall overall. Neither after five minutes nor after 24 hours.

There was one difference, though: people who studied alone overestimated their test performance more (d = 0.36). In pairs, people had a more realistic picture of what they could do. When the solo learners in a further experiment also had to answer out loud, this advantage disappeared. The gain therefore probably comes from answering out loud, not from the friendship.

This is a short laboratory study, not a semester of spaced repetition. But it clears up one myth: together does not automatically mean better.

Where groups actually help

Broader research looks friendlier but has caveats. A meta-analysis by Springer, Stanne and Donovan (1999) on small-group learning in science and engineering courses found a performance advantage of d = 0.51, along with better retention in the program and more positive attitudes. However, the analysis combined very different study designs, and later methodological criticism warned about confounding factors. It is not clean causal evidence for "a study group with friends."

A more recent observational study from an introductory biology course (Williams and Denaro, 2026) found no significant association between study groups and exam grades at all. Without randomization, this is neither proof of benefit nor of harm.

What remains is a clear finding from a different corner: explaining helps the one who explains. Fiorella and Mayer (2013) had students learn material and then explain it on video. Those who actually taught performed significantly better a week later than the control group (d = 0.79). Those who only prepared to teach no longer had a significant advantage after a week. A meta-analysis by Kobayashi (2019) confirms the direction: preparing plus actually teaching yielded g = 0.56, preparing alone g = 0.35.

The trap: remembering together

Now the part that constantly happens to you in study groups without you noticing. When three people try to remember something together, they do come up with more than any single person. But they come up with less than if each person thinks alone and the answers are combined afterward. This is called collaborative inhibition and was already described by Weldon and Bellinger (1997).

The meta-analysis by Marion and Thorley (2016) across 64 studies found the effect clearly: after correcting for possible publication bias, d = −0.56. The common explanation: other people's answers interfere with your own retrieval strategy. You hear what someone says, and your own train of thought breaks off.

The good news from the same analysis: those who had first remembered together then remembered better alone (d = 0.59). The group is therefore not bad, only the order matters.

Leaderboards: motivation with side effects

Gamification has measurable effects on average. Sailer and Homner (2020) found in their meta-analysis g = 0.49 for cognitive learning outcomes and g = 0.36 for motivation. Sounds good. But these figures refer to very different packages of game elements, not to leaderboards alone.

Looking at leaderboards specifically, it becomes more mixed. A review by Li et al. (2024) across 22 studies found no consistent effect; everything depended on design. One documented intervention also saw extrinsic motivation rise while intrinsic motivation fell. Unclear point systems led to confusion and disappointment. And Hanus and Fox (2015) observed, in a gamified course with 71 students over one semester, even less favorable motivation and worse final exam results. However, several game elements were combined there.

Our honest reading: a leaderboard is a tool for people who enjoy competition. For everyone else it can be annoying or shaming. Voluntary participation and clear rules are the design principles the research suggests. Even this is not proven.

Accountability and streaks

What is fairly well established: making progress visible helps. Harkin et al. (2016) found across 138 randomized studies that monitoring your own progress improves goal attainment (d₊ = 0.40). The effect was stronger when results were reported or made public. This is not proof of a "study buddy effect," but it fits with the idea that a friend who sees your progress is a small incentive.

Be careful with classics: the often-cited deadline study by Ariely and Wertenbroch (2002) was officially retracted in September 2026. Anyone arguing with it is arguing with nothing.

And streaks? In an experiment with 602 participants, Silverman and Barasch showed that an intact series clearly increased continued participation. By contrast, displaying a broken series decreased it. What was measured was short-term continuation, not retention. The famous 66 days until a habit forms come from Lally et al. (2010): that was the median, the range was between 18 and 254 days, and a single skipped day did not substantially harm habit formation. A missed day is therefore no reason to give up.

How to study with friends without wasting time

1. Repeat alone, understand together. Do the daily card routine for yourself. Meetings are for things you need to explain or discuss.
2. Answer alone first, then compare. Against collaborative inhibition.
3. Take turns explaining. Whoever explains learns the most. So not always the same person.
4. One deck for everyone. Whoever writes good cards shares them. But everyone checks the cards before learning them. What good cards look like is covered in Writing Good Flashcards.
5. Progress visible, leaderboards optional. Keep what motivates you. Ignore what stresses you.

Study groups in checkst

checkst has study groups built in. You create a group and invite friends via link or QR code. In the group you share decks, and each person learns them with their own FSRS schedule, based on what they forget, not the group's pace. Groups are free, including uploading decks. Pro is only needed for syncing via your checkst account. You need a checkst account to join and study.

There is also a leaderboard by XP, showing each member's level and streak. As described above: for some it is a spur, for others not. Streaks, achievements and daily challenges work without a group too. And if someone in the group has a bad day: you can gift a Frost, a streak protection for a missed day, to a group member once a week. Frost can only be earned by studying, not bought.

An honest limitation: checkst is only available for iPhone and iPad. Groups only work between checkst users. If part of your study group uses Android, they cannot join. Then you export the deck as an .apkg file, and the others open it in Anki or AnkiDroid.