Do Study Groups Work? What the Research Says

Do Study Groups Work? The Short Answer

Yes, if each person does the remembering alone and the group does the checking. In a meta-analysis of 39 studies of college science, math, engineering and technology courses, students who learned in small groups did better on achievement, stayed in their courses more often, and came away with better attitudes toward the subject.

Groups also have a specific weakness, and it happens to be what many study groups spend their time doing. When people try to remember material together, the group recalls less than its members would have recalled working alone and pooling their answers. Psychologists call this collaborative inhibition, and a 2016 meta-analysis found it across 64 studies.

The two findings fit once you separate recalling from checking. Recalling is a solo act: other people's voices knock you off your own path through memory. Checking, explaining and arguing are where a group earns its keep. Other people catch your errors, fill your gaps and push you through problems that are too big for one head. So make each person recall alone first, and save the group for everything after that.

---

The Hidden Cost: Groups Remember Less

The standard experiment is simple. People study a list of words, a set of pictures or a short story. Later, some recall it alone and some recall it in groups of two to four. To judge the groups fairly, researchers build a nominal group: the same number of people who recalled separately, with their answers pooled and duplicates counted once. A nominal group shows what the members would have produced if they'd never talked.

Mary Susan Weldon and Krystal Bellinger ran this comparison in 1997. In their first experiment, three-person groups recalled 48% of the material together. Nominal groups of three reached 57%, and individuals working alone managed 28%. The group easily beat the average member, which is why studying together feels productive. It still lost about a sixth of what its members could have produced on their own.

This isn't one lab's quirk. Stéphanie Marion and Craig Thorley's 2016 meta-analysis in Psychological Bulletin pooled 75 effects from 64 studies and found a sizable average penalty (d = 0.78). The penalty was bigger:

• In groups of three or four than in pairs (d = 0.86 against 0.55).
• Among strangers than among people who already knew each other (0.77 against 0.44).
• When people could recall in any order than when the test set the order for them, as in cued recall (0.76 against 0.40).

A larger 2024 meta-analysis by Xiaochun Luo and colleagues, covering 140 studies and more than 19,000 participants, found the penalty again (d = 0.54).

Why groups lose memories

The best-supported explanation is retrieval strategy disruption. Everyone stores and retrieves material in their own order. You might recall a story scene by scene, while your friend jumps straight to the parts that surprised them. When you hear their items, your own route gets interrupted, and things you'd have reached later never come up. Barbara Basden and colleagues showed in 1997 that the penalty disappeared when group members recalled separate parts of a list, or when recall was organized category by category. Both setups stop people from tripping over each other's routes.

Brainstorming groups run into a close cousin of this problem. A 1991 meta-analysis by Brian Mullen, Craig Johnson and Eduardo Salas found that people brainstorming together came up with fewer ideas than the same number of people brainstorming alone, and the loss grew with group size.

Michael Diehl and Wolfgang Stroebe had traced the main cause in 1987: when they made people wait for their turn to speak, as they would in a group, those people produced about half as many ideas as people who could speak freely. While one person talks, everyone else waits, forgets, or decides their idea isn't worth saying anymore. Diehl and Stroebe called this production blocking.

The lesson for study groups is blunt. If your session opens with everyone calling out what they remember, you're running the exact condition that produces the penalty.

---

What Groups Do Better Than You Can Alone

The same research that found the penalty also found a payoff, and it shows up after the group breaks up. In Marion and Thorley's meta-analysis, people who had recalled with a group later remembered more on their own than people who had only ever recalled alone (27 effects from 22 studies, d = 0.59). A 2024 meta-analysis in the British Journal of Psychology, pooling 64 studies and 13,398 participants, reached the same conclusion: recalling with others moderately improved what people could later recall alone. Group recall costs you during the session and pays you back afterward.

Suparna Rajaram and Luciane Pereira-Pasarin's 2010 review lays out three reasons for that benefit:

• Re-exposure. Hearing a partner recall something you'd forgotten works as a free second study session for that item.
• Error pruning. Partners catch each other's wrong answers before they get repeated.
• Relearning through retrieval. Each item you recall out loud in the group is another round of retrieval practice for you.

Recall alone first, and the penalty can disappear

The most useful finding for students came out of Rajaram's lab. In a 2011 study, Adam Congleton and Suparna Rajaram had some people study a word list three times, while others studied it once and then recalled it twice on their own. Then everyone recalled the list either in a group of three or alone. When that recall came a few minutes after learning:

• After repeated studying, the usual penalty appeared: groups recalled 71% of the list, against 84% for nominal groups.
• After repeated solo recall, the penalty was gone: 68% for groups against 70% for nominal groups, a difference too small to count.
• Later, tested alone, people who'd recalled on their own and then worked in a group remembered 52% of the list, against 42% for people who'd kept recalling alone.

Practicing retrieval had organized their memories well enough that a partner's recall order no longer threw them off. One caution: these people had recalled the list twice before the group met. In an earlier study by Helena Blumen and Rajaram, a single solo recall before the group didn't prevent the penalty (65% for groups against 74% for nominal groups), so one quick attempt may not buy the same protection.

That's the logic to build a study group around. Build your own memory first with solo retrieval, then let the group add what you missed.

Discussion can produce answers nobody had

Groups can also build understanding that the members didn't have going in. In a 2009 Science study, Michelle Smith and colleagues followed 350 students in an undergraduate genetics course. Students answered a clicker question alone, discussed it with their neighbors, and voted again. Then each student answered a new question that tested the same concept, without discussion and before seeing the answer to the first.

Correct answers on the new question were 21 percentage points higher than on the first question before discussion, and 38 points higher on the hardest questions. Of the students who switched from a wrong answer to the right one after talking it over, 77% then got the new question right on their own, which suggests they had understood the idea rather than copied a neighbor. The authors also calculated that many of the groups were unlikely to have contained anyone who knew the answer at the start. Their conclusion: discussion helps students work out the concept together, not just find the classmate who already knows.

Hard problems favor groups

Femke Kirschner, Fred Paas and Paul Kirschner tested when collaboration pays off. In a 2011 experiment, 83 Dutch high school students learned to solve heredity problems either alone or in groups of three, with problems that were either simple or complex. On complex problems, groups performed better during learning while reporting less mental effort, and later tests showed that group learning was the more efficient route. On simple problems, learning alone was more efficient. (Final test scores on their own didn't differ much, so think of this as a finding about effort and efficiency.)

The authors call it the collective working memory effect: a hard problem can be split across several heads, while an easy one only adds the cost of coordinating.

The catch: groups can spread errors too

Error pruning is real. When Michael Ross and colleagues had married couples recall six household scenes, couples recalling together reported 8.77 false items of their own making on average, against 19.30 for pairs of people recalling separately (though they also recalled fewer correct items).

It has a dark twin, though. In the "social contagion" experiments of Henry Roediger, Michelle Meade and Erik Bergman (2001), participants recalled household scenes with a partner who was secretly working for the researchers and who mentioned objects that hadn't been there. On a later test taken alone, participants falsely recalled 22% of the suggested objects, against 6% of comparable objects nobody had suggested. A confident wrong answer in a study group can travel the same way. The defense is cheap: settle disputed points against the textbook, the slides or the original source, not against whoever sounds surest.

---

Is Group Study Effective? What Classroom Studies Show

Lab studies use word lists and short stories. Classroom studies ask a broader question: do students learn more over a whole course when they work in groups? Here the evidence is mostly positive.

• Small-group learning in college STEM. Leonard Springer, Mary Elizabeth Stanne and Samuel Donovan's 1999 meta-analysis in the Review of Educational Research combined 39 studies. Students who learned in small groups showed higher achievement (d = 0.51), greater persistence (0.46) and more favorable attitudes (0.55). The achievement effect was larger when the group work happened outside class (0.65) than in class (0.44), though that comparison is across different studies rather than a controlled test.
• Cooperative learning. David and Roger Johnson spent decades comparing cooperative, competitive and individualistic learning. A 1998 summary with Karl Smith covered more than 168 studies of college-age learners and found cooperation ahead on achievement by about half a standard deviation (0.53 against individual work, 0.49 against competition). The Johnsons are clear about the conditions: groups need positive interdependence (members need each other to succeed) and individual accountability (each person's own learning gets checked). Putting students at the same table isn't enough.
• Active learning. Scott Freeman and colleagues' 2014 meta-analysis in PNAS covered 225 studies of undergraduate science, engineering and math courses. Under active learning, which often involves group work, exam scores were about 6% higher and the average failure rate was 21.8%, against 33.8% under traditional lecturing. Not every active-learning method in it involved peers, so treat this as support for working with the material, not proof about study groups.

The calculus workshop that put groups on the map

The best-known study group story starts at UC Berkeley in the mid-1970s. Black students there were failing first-year calculus at high rates, and Uri Treisman set out to understand why. In 1975 and 1976 he followed students through their study routines. The Black students he observed mostly studied alone. The Chinese American students he followed worked through problems on their own first, then got together to check each other's answers and talk through the problems. They also put in more hours, roughly 14 or more a week against 6 to 8.

Treisman's response was the Mathematics Workshop Program, which later became the model for the Emerging Scholars Program at other universities: small groups working on problems harder than the homework, with a facilitator and high expectations. In an evaluation by Robert Fullilove and Treisman, 3% of Black workshop students (5 of 188) earned a D+ or lower in calculus from 1978 to 1982, against 40% (57 of 143) of Black students who didn't take part.

It wasn't a randomized trial, and the workshop bundled several things at once: hard problems, extra practice time, a facilitator and a community. Still, the study habit that inspired it, solo work followed by group checking, is the same one the lab research points to.

Richard Light's interviews with Harvard undergraduates, collected for his 2001 book Making the Most of College, pointed the same way. The Harvard Gazette summed up the finding as students learning more when they study together in small groups outside class than when they study alone. Those are students' own accounts, not an experiment, but it lines up with the outside-class result in the Springer meta-analysis.

Exams can be collaborative too

Some instructors now run two-stage exams: students take the exam alone, then take it again in small groups. Brett Gilley and Bridgette Clarkston tested the idea in a controlled study in a 98-student earth and ocean science course. Three days after a midterm, a surprise individual test showed bigger gains on the questions students had worked through in a group (about 13 points) than on questions they had simply answered a second time alone (about 4). Notice the order again: alone first, then together.

One caveat applies to all of this. In a classroom, an instructor designs the task, forms the groups and checks each person's learning. A study group you set up with friends has none of that by default, so it has to supply its own structure.

---

Study Alone or in a Group? What Each Is Good For

The research gives a fairly clean division of labor. Use it to decide what to do on your own and what to bring to the group.

The pattern holds across the table. Anything that depends on your own retrieval starts alone. Anything that depends on checking, explaining or untangling a hard problem benefits from other people.

---

Why Most Study Groups Fail

Many students have sat through a study group that turned into an hour of snacks and gossip, or one where a single person explained everything while everyone else nodded along. Research explains both.

Social loafing. People put in less effort when their work is pooled with other people's. Steven Karau and Kipling Williams's 1993 meta-analysis of 78 studies found the effect was reliable (d = 0.44) and grew with group size. It all but disappeared when each member's own contribution could be identified and evaluated, when the task felt highly meaningful, or when the group's overall performance would be evaluated. A study group where nobody is responsible for anything in particular is built for loafing.

The lecture trap. When one member explains and the others listen, the explainer gets the learning benefit (that's the protégé effect) and the listeners get the feeling of understanding. Following an explanation is easy. Producing one yourself is hard. That gap is the illusion of competence, and a passive study group can hand it to everyone at the table.

Recalling together from the start. As the lab work shows, a group brain dump at the top of the session turns up less than the same people would have recalled alone and pooled.

Unequal preparation. If half the group hasn't done the reading, the session becomes a lecture for them and a slow review for everyone else.

Contagious errors. A wrong answer said with confidence can end up in other people's memories, not just the speaker's.

---

How to Run a Study Group That Works

Here's a structure that builds the research into a session. It works for exam review, problem sets and reading groups.

1. Keep it small: three is a good default, four at most. Physics education researchers Patricia Heller and Mark Hollabaugh found that groups of three worked a little better than pairs or groups of four for problem solving. Pairs sometimes didn't know enough between them, and groups of four tended to leave one person out. Bigger groups also invite more loafing and side conversations.
2. Prepare alone, every time. Agree on the material in advance. Everyone reads, watches or works the problems before the meeting and tests themselves on it at least once. Arriving unprepared turns you into an audience.
3. Open with solo recall. Spend the first ten minutes writing down everything you remember, alone and in silence. This is the blurting method, and it gets everyone's own recall on paper before anyone else's can interfere. Together with the self-test from your preparation, that's two rounds of solo recall before the group starts, an adaptation of the repeated solo recall used in the Congleton and Rajaram study.
4. Compare, then argue. Go around the table and add items one at a time. Mark disagreements and settle them by checking the source, not by vote.
5. Assign teach-backs. Give each person one concept to explain without notes, and rotate who teaches what each week.
6. Quiz each other. Bring questions you wrote in advance. Everyone answers on paper first, then you discuss. That's the same answer-then-discuss order as the clicker study.
7. Close with solo recall. End with five minutes of individual recall. It's one more round of retrieval practice, and it shows each person what they still can't produce without the group.
8. Meet often, not long. Several shorter sessions spread across the term give everyone spaced practice. One marathon the night before is just cramming with company.

A one-hour session might look like this:

---

How to Study With Others Online

Plenty of group study now happens without a meeting. Classmates share slides in a group chat, split up readings and swap notes. The same rule carries over online: do your own retrieval before you look at anyone else's work.

A simple routine for a group that studies asynchronously:

1. Read and highlight alone. With Glasp's web highlighter, each person highlights the assigned web pages and PDFs and adds notes. Every highlight has its own URL, so you can share the exact passage you mean instead of "the part about the second experiment."
2. Recall before you compare. Before you look at anyone else's highlights, write a short summary from memory. Reading other people's highlights first is the online version of everyone shouting answers at once.
3. Compare with the group. Follow each other's Glasp profiles. When classmates have publicly highlighted the same web page, the Community tab in Glasp's browser extension shows their passages, so you can see what they stopped at and what you skipped. (PDFs you upload to Glasp stay private, so paste those highlights into the group chat instead.) Disagreements about what mattered make good discussion prompts.
4. Work from lecture transcripts. For lectures and explainers on YouTube, YouTube Summary gives each person a summary and the full transcript to read and highlight, with timestamps that make it easy to point at the exact moment you're asking about.
5. Settle arguments with sources. Glasp's AI chat answers questions from everything you've highlighted and points to the highlights it drew on, so you can open the original and check who was right.

The same idea scales past your own group. On a page many people have read, the Popular view of the Community tab shows the passages more than one reader highlighted, and Glasp's community shows what other readers are reading and highlighting. It's a bit like a study group of strangers who read the material before you did.

---

Frequently Asked Questions

Is it better to study alone or in a group?

Both, in sequence. Recalling as a group from the very start usually turns up fewer items than the same people would produce by recalling alone and pooling their answers, so do your first rounds of recall alone. Checking answers, explaining concepts and working through hard problems go better in a group, and recalling with others after you've practiced alone can improve what you remember later. A good default is to recall alone first, compare with others, and recall alone once more at the end.

How many people should be in a study group?

Three is a good default, and four is a sensible maximum. Physics education researchers found that groups of three solved problems a little better than pairs or groups of four: pairs sometimes lacked the knowledge between them, and fours tended to leave someone out. Larger groups also show more social loafing.

Are study groups good for memorization?

Yes, once you've done the first rounds alone. Recalling together from the start usually turns up fewer items than pooled solo recall, but in experiments, recalling with others after practicing alone improved what people later remembered on their own. Learn facts and terms with solo retrieval practice, such as flashcards or writing out what you remember, then use the group to quiz each other, with everyone answering on paper before anyone speaks.

What is collaborative inhibition?

It's the finding that a group recalling together remembers less than the same people would recall separately with their answers pooled. A 2016 meta-analysis of 64 studies found it consistently. The leading explanation is that hearing other people's recall disrupts the order in which you'd retrieve things yourself.

Do online study groups work?

Direct research is thin: we didn't find a large trial or meta-analysis comparing online study groups with in-person ones. A lab study from Rajaram's group ran group recall online and found that whether the group penalty appeared depended on how the session was set up. Asynchronous groups even have an edge: it's easy for each person to recall alone before seeing anyone else's notes. Video-call groups face the same problems as in-person ones, plus more distractions.

How often should a study group meet?

Once or twice a week through the term works better than one long session before the exam, because spreading study out beats packing it into one sitting. Keep each session to about an hour with a clear agenda.

---

Study Alone First, Then Together

Study groups work, just not the way many students run them. A group is a poor place to dig memories out and an excellent place to check them, fix them and stretch them. Recall on your own, bring what you've got to the group, argue about the gaps, and recall on your own again before you leave.

Glasp fits that rhythm. Highlight your readings and YouTube lectures on your own, compare notes with classmates who highlighted the same pages, and check disputed points against your sources with Glasp's AI chat.

The best study partner is the one who makes you do the remembering.