Is the Zeigarnik Effect Real? The Short Answer
The Zeigarnik effect is the claim that you remember interrupted tasks better than completed ones. The best current evidence says that specific claim is wrong. A 2025 meta-analysis in Humanities and Social Sciences Communications pooled the 37 studies published after Zeigarnik's original and found interrupted and completed tasks are recalled at essentially the same rate: a ratio of 0.99, where 1.00 means no difference at all.
What did survive is the finding next door. Given a chance to go back to something they were interrupted in the middle of, people take it around two-thirds of the time. That's the Ovsiankina effect, named for Maria Ovsiankina, who ran her study a year after Zeigarnik in the same Berlin lab. It replicates. It just never got famous.
Those two results point in different directions, and the difference matters if you read for a living or study for a degree. The popular version of Zeigarnik implies that a half-read article is quietly working on you, filed somewhere accessible, waiting. It isn't. What's documented is that it keeps tugging. Nearly a century of data says the tug is real and the memory benefit is imaginary, which is roughly the worst possible combination: you pay the attention cost and get nothing back.
Romain Ghibellini and Beat Meier put it flatly at the end of their review: "the Ovsiankina effect may represent a general tendency. In contrast, the replicability of the Zeigarnik effect remains questionable."
What Bluma Zeigarnik Actually Did in 1927
Bluma Zeigarnik was a Lithuanian-born graduate student in Kurt Lewin's group in Berlin. Her dissertation, published in 1927 as "Das Behalten erledigter und unerledigter Handlungen," ran a long series of experiments. Almost every article you'll read about her cites only the first.
In that main experiment, 32 people were given 22 short tasks: winding thread, folding paper, multiplication, drawing a vase, counting backwards. Each was meant to take 3 to 5 minutes, with instructions to work "as rapidly and correctly as possible." Half were allowed to run to completion. The other half were cut off, and the timing was deliberate. Zeigarnik interrupted people at the moment they were "most engrossed," simply presenting the next task and saying, "Now do this, please." Once all 22 had been handed out, she asked participants to recall every task they could.
About 81% of them recalled more of the interrupted tasks than the completed ones. Her preferred statistic was the ratio of interrupted to completed tasks recalled, computed per person, and it landed near 1.9. She replicated it three more times: 15 adults with a different task set, 47 adults tested together in one room, and 45 adolescents with a mean age of 14. All four came in around 0.8 for the proportion of participants favoring interrupted tasks, and near 2.0 for her ratio.
She also anticipated the obvious objections. In one follow-up she interrupted every task and let people immediately resume half of them, in case interruption was simply startling. In another she told participants which interrupted tasks they'd get to finish later, in case they were holding on to what they expected to resume. Neither manipulation moved her result. For a 1927 dissertation, it was careful work.
One correction is worth making, because it shows up on a handful of heavily copied pages: the study was not "138 children." That number appears nowhere in Zeigarnik's own report. Her main experiment was 32 adults tested one at a time, and the only young participants anywhere in the four experiments above were the 45 adolescents in a group session. Colin MacLeod's 2020 history of the effect in Memory & Cognition lays out the numbers experiment by experiment, and they don't match the version that circulates.
The Waiter Story Behind the Zeigarnik Effect
The origin story is nearly always told the same way. Lewin and his friends are in a Berlin restaurant. The waiter recalls everyone's order perfectly without writing anything down. After the bill is paid, Lewin calls him back and asks him to repeat it, and the waiter has no idea.
There are two published versions, and they disagree about what the waiter forgot.
Edwin Boring's 1957 account has Lewin asking what he owed, being told instantly, paying, and then a while later asking again how much he'd paid. The waiter no longer knew. In Boring's telling the memory in question is the total on the bill. Donald MacKinnon, who says he was at the table, tells it differently in Alfred Marrow's biography of Lewin. The waiter knew "just what everyone had ordered" without a written reckoning. Half an hour later, asked to write the check again, he was indignant: "I don't know any longer what you ordered. You paid your bill." In MacKinnon's version it's the orders, not the total.
MacLeod's observation about this is a nice piece of irony. Two accounts of the same evening, remembered differently by different people, is a small demonstration of Frederic Bartlett's argument that memory is reconstructive rather than a recording. The anecdote used to launch a memory effect is itself an example of memory going soft at the edges.
What both versions share is the shape Lewin cared about. He'd proposed that forming an intention creates a kind of tension, a quasi-need, that persists until the intention is discharged. Zeigarnik's job was to turn that idea into an experiment, and she did. It's the tension theory, not the experiment, that gave the effect its intuitive grip.
Has the Zeigarnik Effect Been Replicated? 37 Studies in One Table
Ghibellini and Meier searched PsycInfo and PSYINDEX in September 2023, screened 1,349 unique publications, and pulled out the studies their search terms reached that interrupted some tasks, completed others, and measured what people remembered. Then they weighted the results by sample size. They're upfront that a broader set of search terms might have surfaced more.
Here's the shape of the literature, using the measure the meta-analysis reports in its headline: mean interrupted tasks recalled divided by mean completed tasks recalled. Above 1.00 favors interrupted tasks. Below 1.00 favors completed ones.
| Study | Participants | Interrupted-to-completed recall ratio |
|---|---|---|
| Zeigarnik, 1927 (main experiment) | 32 adults | 1.61 |
| Rösler, 1955 | 224 children, mixed sample | 0.97 |
| Butterfield, 1965 | 128 children | 0.88 |
| Van Bergen, 1968 | 220 students, children and adults | 0.89 |
| Raffini and Rosemier, 1972 | 624 students | 0.81 |
| Hofstaetter, 1985 | 144 students | 1.15 |
| Weighted average, all later studies | 37 studies pooled | 0.99 |
Individual studies scatter from 0.70 to nearly 2.0. Weighted together, they cancel. Including Zeigarnik's own data doesn't move the pooled figure off 0.99, because 32 people can't outvote several thousand. On the alternative measure the authors say they actually prefer, interrupted tasks make up 49.16% of everything recalled across the 13 studies that reported enough to compute it, which is a hair under half rather than a hair over. Across the eight studies detailed enough for a standardized effect size, the pooled value is dz = 0.15, small enough that you'd need hundreds of participants to see it reliably.
There's a technical reason Zeigarnik's own number ran high, and the meta-analysis spells it out in a footnote. She averaged each person's individual ratio. Imagine two participants: one recalls 8 interrupted and 4 completed, the other recalls 4 interrupted and 8 completed. Their ratios are 2.0 and 0.5, which average to 1.25, apparently a solid advantage for interruption. But the group recalled 12 interrupted and 12 completed. It's a tie. Averaging ratios systematically inflates the result, which is why her 1.9 and the meta-analysis's 1.61 describe the same data.
None of this is new information. Van Bergen's 1968 doctoral monograph at the University of Amsterdam was a serious attempt at reproducing Zeigarnik's design, and in the closest of those replications, with 34 participants, it came back at 0.88 on Zeigarnik's own preferred statistic. The 0.89 in the table above is a different calculation pooled across her whole monograph. MacLeod's verdict in 2020 was blunt: "At best, it would appear to hinge on certain individual difference characteristics; at worst, it is simply not replicable." The meta-analysis notes that despite this, the effect "is still freely cited, taken as a given, and used as an explanation for a plethora of research findings."
The same arc runs through the learning styles myth: an intuitive idea, an early positive result, and decades of quieter contradiction that never catch up to the popular version.
The Half Nobody Cites: The Ovsiankina Effect
Maria Ovsiankina published her study in 1928, one year after Zeigarnik and from the same lab, and she asked a different question. Not what do people remember, but what do people do. She interrupted participants mid-task, then engineered an opportunity to go back to it and watched whether they took it.
They did. Pooling 20 studies published since, Ghibellini and Meier put the resumption rate at 66.79%, against the 50% they treat as a neutral baseline. Include Ovsiankina's own data and it's 67.00%. It turns up in a 1941 study of 180 students given a single puzzle, at 70.5%, and a 1938 study of 177 children, at 88.7%. It turns up in 2020 with 875 people playing an online game, where 73.85% went back to the interrupted round.
Two honest qualifications. Individual studies range from about 36% to 100%, so this is a tendency and not a law. And the authors say the pooled figure "should not be interpreted as an absolute measure, but rather should be seen as indicative," because the studies measured resumption in incompatible ways. What holds up is the direction: interrupted tasks get resumed more often than they get abandoned.
That's an odd result if you believe Lewin's tension theory, which predicted the memory effect and the resumption effect from the same mechanism. Modern accounts have moved on. Thomas Goschke and Julius Kuhl proposed in 1993 that a pending intention stays activated in memory rather than discharging, what later work calls a heightened subthreshold activation. They demonstrated it with the intention superiority effect: people recognize words from a script they intend to perform faster than words from a script they only memorized. That activation makes you act when the opportunity appears. It doesn't hand you a better conscious memory for the thing.
Put in the language you actually use: your unread queue isn't a memory aid. It's a set of primed triggers. That's true whether the queue is 12 tabs, a stack of books, or 400 saved articles you'll never open. We wrote about the queue itself in Save Now, Read Never; this is the mechanism underneath it.
What an Open Loop Costs You While You Read
The most useful experiment on this topic isn't about memory at all. In 2011, E. J. Masicampo and Roy Baumeister ran a study whose dependent measure was reading.
Seventy-three undergraduates came into the lab believing they were doing two unrelated things: a survey about daily tasks and a reading comprehension test. They were split across three conditions, so roughly two dozen people per cell. One group wrote about two important errands they still had to do, specifically choosing tasks where they didn't yet know when, where or how they'd handle them. A second group wrote a specific plan for those errands. A control group wrote about two errands they'd recently completed. Then everyone read the first 3,200 words of Erle Stanley Gardner's The Case of the Velvet Claws, one word at a time, with occasional prompts asking whether their attention had drifted. Afterward they answered eight questions about the plot.
| Measure | Unfinished errands | Unfinished, with a plan | Completed (control) |
|---|---|---|---|
| Intrusive task thoughts (1 to 7) | 3.00 | 1.77 | 1.82 |
| Mind wandered at least once | 65.2% | 33.3% | 50.0% |
| Reading comprehension (out of 8) | 6.13 | 6.94 | 6.93 |
The comprehension and intrusion gaps against the control group are statistically significant. The mind-wandering row is the weakest of the three: 65.2% against the control's 50% did not reach significance, and only the gap against the plan group did. Take that row as suggestive and the other two as the result.
Nobody in that study was interrupted mid-sentence or handed a cliffhanger. They wrote a paragraph about two errands. That was enough to cost them most of a point on an eight-question test, on material that had nothing to do with the errands.
Two other lines of work point the same way. Sophie Leroy's 2009 paper in Organizational Behavior and Human Decision Processes introduced attention residue: when you switch away from an unfinished task, part of your attention stays behind and degrades your performance on whatever you moved to. Her result is sharper than it's usually quoted, and less comfortable. Simply finishing wasn't enough. Only participants who finished and had been working under time pressure showed less residue and better performance on the next task. People who finished at their own pace did no better than people left hanging. And Cornelia Syrek and colleagues, in a 12-week diary study of 59 employees with 357 matched Friday and Monday observations, found that unfinished tasks at the end of a work week predicted worse weekend sleep, specifically through affective rumination rather than through problem-solving thought.
Caveat, and it's a real one: the Masicampo and Baumeister study is one lab in 2011, from a group whose other headline findings have not all survived replication, with about two dozen people per condition. Leroy's and Syrek's work points the same direction, but three findings sharing a theoretical prior and a publication-bias exposure aren't three independent confirmations. This is the best evidence available on the question, not settled evidence.
If you've ever finished a page and realized you absorbed none of it, this is one of the mechanisms, and it stacks with the extraneous demands described in cognitive load theory.
Plan Making Closes the Loop Without Finishing
The plan condition is the reason that paper matters.
Those participants wrote about two unfinished errands, exactly like the first group, and then wrote down a specific plan for each one: when, where and how they'd do it. They did not do the errands. Nothing got completed. Then they read the same novel excerpt.
They scored 6.94 out of 8. The control group, whose errands were genuinely finished, scored 6.93. Intrusive thoughts dropped to 1.77 against the control's 1.82. On the measures that reached significance, writing a plan did what actually finishing the task did.
The commitment is doing the work, not the thinking about it. In a later study in the same paper, one group listed several possible ways they could get their task done without settling on any of them. That group still reported significantly more intrusive thoughts than the group that committed to one concrete plan.
That fits a much larger literature. Peter Gollwitzer and Paschal Sheeran's 2006 meta-analysis pooled 94 independent tests and more than 8,000 participants and found d = 0.65 for specifying when, where and how you'll act. That's a far bigger evidence base than Zeigarnik's, though it's pooled the same way and carries the same exposure to publication bias. Their proposed mechanism is that a specific plan hands control of the behavior over to situational cues, so you stop having to hold it consciously.
Here's the part that gets left out of every productivity post citing this study. The plan group in the follow-up experiment were still just as anxious about their unfinished tasks as the group with no plan, and significantly more anxious and less content than people whose tasks were actually done. Planning cleared the cognitive interference. It did nothing for the feeling. If a plan makes you calm, that's a bonus, not the finding.
How to Use the Zeigarnik Effect When You Read
The usable half is the resumption half, and the practical move is to stop generating loops you have no intention of resuming.
Reading generates them faster than almost anything else you do. Every article you start and abandon, every video you scrub through, every book with a receipt still in it at page 60 is a pending intention with no plan attached. One caution before the protocol: the lab version of that pile was two errands people had written about ten minutes earlier, not a 400-item reading queue, so what follows is a reasonable extension of the evidence rather than a measured result.
The fix isn't finishing more. It's discharging the intention deliberately, in three steps and one sentence of writing.
1. Name what you came for before you start. One sentence. "I want to know whether this actually replicated." Now the loop has a defined closing condition instead of "read the whole thing," which most articles don't deserve anyway.
2. Capture the answer, not the article. When you hit the part that answers your sentence, mark it. Glasp's web highlighter keeps the passage attached to the page it came from, so what you extracted stays connected to its source rather than becoming an orphaned quote in a notes app. Two or three colors is enough structure for most reading.
3. Write the plan in the note field. This is the step people skip, and the research says it's the one that matters. Not a summary. A commitment: "use this in the Thursday review," "send to Kei when we revisit the pricing page," "check the 2025 meta-analysis before citing 1.9 anywhere." When, where, how.
Then close the tab. The article isn't going to teach you anything from the queue, and the 0.99 ratio is the whole argument for that step.
Video is the harder case, because you can't skim it. A 40-minute talk you stopped at minute 12 is an unusually loud open loop. YouTube Summary collapses it into a transcript with timestamps you can scan, highlight and close in a couple of minutes, so the open loop ends with a decision instead of a bookmark. Books have the same problem in slow motion, which is what Kindle highlights import is for: the highlights come across even from books you abandoned, so the useful part survives the unfinished part.
Later, when you want the thing you extracted, you search your own highlights rather than rereading the source. Glasp's AI chat answers questions against what you've actually marked, which is a much shorter path than reopening six tabs to find one sentence.
Scoring the Popular Zeigarnik Advice
The effect has spawned a lot of tactics. Some are defensible on the evidence and some are folklore.
| Popular advice | Verdict |
|---|---|
| Stop studying mid-sentence so your brain keeps working on it | Doesn't hold. The recall advantage pools to 0.99 across 37 studies. |
| Stop at a point you're eager to return to, so restarting is easier | Plausible. Resumption runs near 67%, but every study used an experimenter-imposed interruption, not a stopping point you chose. |
| Leave tasks unfinished to remember them better | Doesn't hold, and costs you. The pooled ratio is 0.99, and Leroy's work says the residue damages your next task. |
| Write down when, where and how you'll finish, then move on | Strongest evidence here. Matched the completed-task control on comprehension and on intrusive thoughts. |
| Cliffhangers keep audiences hooked | Unproven as stated. The underlying studies used puzzles and errands, not narrative. |
| An unfinished book is quietly teaching you | Doesn't hold. No mechanism survives the meta-analysis. |
The second row is the one genuinely useful tactic, and it's usually stated wrongly. Stopping at a compelling point makes you more likely to come back. It does not make you remember more. If your problem is starting again tomorrow, use it. If your problem is retention, reach for active recall and spacing instead.
Frequently Asked Questions
What is the Zeigarnik effect?
The Zeigarnik effect is the claim that people remember interrupted or unfinished tasks better than completed ones. It's named for Bluma Zeigarnik, who reported it in her 1927 Berlin dissertation. The origin story is an observation by her supervisor Kurt Lewin about a waiter who remembered an order until the bill was paid, though the two published versions of that anecdote disagree on what he forgot. The memory claim does not survive replication. The related tendency to go back to an interrupted task does.
Has the Zeigarnik effect been replicated?
Not successfully. A 2025 meta-analysis of the 37 studies published after Zeigarnik's original found a weighted interrupted-to-completed recall ratio of 0.99, meaning no difference. Butterfield got 0.88 with 128 children in 1965, and Van Bergen's 1968 monograph, the most careful attempt at reproducing the original design, also came back below 1.0. Individual studies land anywhere between 0.70 and 2.0 and average out to essentially nothing.
What is the Ovsiankina effect?
It's the tendency to return to an interrupted task when given the chance, named for Maria Ovsiankina, who published it in 1928 from the same Berlin lab as Zeigarnik. Across 20 later studies the pooled resumption rate is 66.79%, against a 50% neutral baseline. It's the half of Lewin's original prediction that survived.
Does the Zeigarnik effect improve memory?
No. The pooled recall ratio is 0.99, meaning no difference between interrupted and completed tasks. The one effect-size estimate available, dz = 0.15 across the eight studies detailed enough to compute it, is small enough that you'd need several hundred participants to detect it at all. For actual retention, use spaced retrieval.
Why do unfinished tasks keep popping into my head?
The current explanation, from Goschke and Kuhl's 1993 work, is that a pending intention stays in a state of heightened activation so you'll act on it when the opportunity appears. That's an action prompt rather than a memory boost. It's also measurably expensive: in Masicampo and Baumeister's experiment, people with two unresolved errands on their minds scored 6.13 out of 8 on a reading test against 6.93 for controls.
How do you stop thinking about unfinished tasks?
Write down when, where and how you'll deal with each one. In the lab, doing that eliminated the interference as completely as finishing the task did, even though nothing got finished. Vaguely considering your options doesn't work; the people who listed several possible approaches without committing to one still reported significantly more intrusive thoughts than the people who committed.
How do you use the Zeigarnik effect when studying?
Use the resumption half and ignore the memory half. Stopping a session at a point you're keen to pick up can make restarting easier, which is a motivational benefit. It will not improve what you retain, and switching away from unfinished work leaves residue that degrades whatever you do next, so don't leave loops open across a whole study week.
The Bottom Line
Bluma Zeigarnik ran a careful set of experiments in 1927 and found something real in her data. Nearly a century of follow-up says the memory effect she named doesn't generalize, and the honest summary today is that interrupted and completed tasks are remembered about equally well. Meanwhile the finding published a year later by her colleague down the hall, the one about going back, has quietly held up the whole time.
For anyone who reads to learn, that reshuffles the advice. Your unfinished pile is a set of primed intentions, and the available evidence says they cost you something while they sit there: comprehension in one reading experiment, sleep quality in one diary study, performance on the next task in Leroy's switching work. What it does not do is teach you anything. The way out isn't more discipline about finishing. It's a habit of ending: decide what you wanted, take it, write down what it's for, and let the rest go.
That's most of what a good reading system does. Start highlighting with Glasp, attach a one-line intention to each highlight, and the queue stops being a debt you owe your future self.
References: Ghibellini, R., and Meier, B. (2025). Interruption, recall and resumption: a meta-analysis of the Zeigarnik and Ovsiankina effects. Humanities and Social Sciences Communications, 12, 962. Goschke, T., and Kuhl, J. (1993). Representation of intentions: Persisting activation in memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 1211-1226. Gollwitzer, P. M., and Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69-119. Leroy, S. (2009). Why is it so hard to do my work? The challenge of attention residue when switching between work tasks. Organizational Behavior and Human Decision Processes, 109, 168-181. MacLeod, C. M. (2020). Zeigarnik and von Restorff: The memory effects and the stories behind them. Memory and Cognition, 48, 1073-1088. Masicampo, E. J., and Baumeister, R. F. (2011). Consider it done! Plan making can eliminate the cognitive effects of unfulfilled goals. Journal of Personality and Social Psychology, 101, 667-683. Ovsiankina, M. (1928). Die Wiederaufnahme unterbrochener Handlungen. Psychologische Forschung, 11, 302-379. Syrek, C. J., Weigelt, O., Peifer, C., and Antoni, C. H. (2017). Zeigarnik's sleepless nights: How unfinished tasks at the end of the week impair employee sleep on the weekend through rumination. Journal of Occupational Health Psychology, 22, 225-238. Van Bergen, A. (1968). Task interruption. Doctoral dissertation, University of Amsterdam. Zeigarnik, B. (1927). Das Behalten erledigter und unerledigter Handlungen. Psychologische Forschung, 9, 1-85.
