Why Unfinished Things Won't Leave You Alone

In the mid-1920s a student in Kurt Lewin’s psychology group in Berlin, Bluma Zeigarnik, gave 164 people a run of small jobs to do one after another. Build a box out of cardboard or a figure out of clay, then a puzzle or some arithmetic. There were eighteen to twenty-two tasks each, and about half the time she stopped the person before they finished and moved them on to the next one. Afterwards she asked them, without warning, to list the tasks they had worked on.

They named the interrupted ones far more often. Across her main experiment the ratio came out at 1.9 interrupted tasks recalled for every completed one, a memory advantage of roughly 90 per cent for whatever had been left undone. She published in 1927, and the result went on to carry her name. The Zeigarnik effect is now one of the most retold findings in psychology, borrowed by screenwriters to justify the cliffhanger and by productivity writers to explain why the half-finished job follows you home.

The memory part does not hold up.

What a century of replications found

In 2025 Romain Ghibellini and Beat Meier at the University of Bern pooled the studies that had tried to reproduce Zeigarnik’s result, 38 publications on the memory effect alone. Across them, interrupted tasks were recalled about as often as finished ones. The average ratio was 0.99. Where a standardised effect could be calculated, it was small, and on one measure the interrupted tasks came out slightly worse.

The same meta-analysis pooled a second, less famous finding from Lewin’s group. In 1928 Maria Ovsiankina, another of his students, interrupted people mid-task and then gave them a spare moment with the unfinished work still in reach. Many went back to it without being asked. Across 21 publications, Ghibellini and Meier found that interrupted tasks were resumed roughly two thirds of the time, well above what chance would predict. They concluded that the Ovsiankina effect looks like a general human tendency. The Zeigarnik effect does not.

Their explanation for the gap is historical. A 1927 laboratory was a place where the experimenter’s authority was high and a task handed out by a professor mattered to the person doing it. Later researchers often had to manipulate the atmosphere to get anything like Zeigarnik’s result. Ghibellini and Meier also point to something more recent: people are now interrupted so routinely, by phones and by working across several things at once, that being stopped halfway is no longer an event.

The unfinished thing does not stay in the mind because the mind keeps a better record of it. What survives the replications is something closer to a pull. The interrupted task wants to be gone back to.

What unfinished goals actually do

In 2011 E. J. Masicampo and Roy Baumeister set out to measure that pull directly. They had people bring an unfinished goal to mind and then do something unrelated. The unfinished goal leaked into the unrelated work: it turned up as intrusive thoughts during a reading task, it made goal-related words quicker to recognise, and it dragged down performance on anagrams. It behaved less like a memory and more like a program still running in the background.

Then they let some people make a specific plan for when and how they would finish the goal. The plan did not finish anything. It still switched the effects off. Intrusions stopped, and so did the interference. The detail that matters most is in their fourth study: the plans that worked were the ones people meant. The participants who went on to carry out their plans were the ones whose intrusions had disappeared.

Lewin’s own explanation for all of this was tension. Setting out to do something, he argued, creates a pull that persists until the thing is done. Masicampo and Baumeister’s result suggests the pull will accept a credible promise in place of completion. It is not satisfied by the task being forgotten, and it is not satisfied by a plan nobody intends to keep.

Outside the laboratory the pull has measurable costs. Christine Syrek and colleagues followed 59 employees for twelve weeks, matching their Friday reports with their Monday ones. The more tasks people left unfinished at the end of the working week, the worse they slept that weekend, and the link ran through affective rumination, the kind of turning a problem over that changes nothing about it. A 2026 meta-analysis led by Johannes Wendsche, pooling studies of thousands of workers, found the same thing across workplaces: unfinished work goes home with people as thoughts about work in their time off, and the strongest link is to rumination.

The things that cannot be planned

The laboratory tasks were small and could always be picked up again. Many of the unfinished things a life accumulates cannot.

In 1997 Kenneth Savitsky, Victoria Medvec and Thomas Gilovich asked people about their regrets, and they used Zeigarnik’s idea to explain what they found. People reported thinking more often about the things they had failed to do than about the things they had done and wished they had not. When asked to list their biggest regrets of each kind and then recall them weeks later, they remembered more of the omissions, even when the regrets were matched for how serious they were. That is one set of studies, and the wider pattern it belongs to, that inaction looms larger in regret as time passes, has held up only partly in a large replication run through a Chicago science museum. The intuition behind it is easier to state. A regret of action has an ending, however bad. A regret of inaction is a task that never started and so never stopped.

Some of those can still be done. A letter can be written late. A conversation can be had years after it should have been. For those, the laboratory result applies as it stands: what switched off the pull in Masicampo and Baumeister’s studies was a specific plan that the person meant to keep.

Others belong to a situation that no longer exists. Grief researchers have a name for the clearest case, unfinished business with the dead, and they have measured it. Jason Holland and colleagues, building a scale for it in 2018, found it splits in two: unfulfilled wishes, the things never said and the chances missed, which tend to arise in loving relationships, and unresolved conflict, which tends to arise in difficult ones and carries more risk of prolonged grief. An earlier study of 224 bereaved people led by Kara Klingspon found that simply having unfinished business with the person who died was not the robust predictor of how grief went. The distress it caused was.

That distinction is the most interesting result in this whole literature. Two people can leave the same thing unsaid to the same person, and only one of them is pulled back to it night after night. A plan cannot be the difference, because there is no longer anyone to carry a plan out with. The research can show that the incompleteness is not what predicts the harm. It cannot yet say what the other person has done with theirs.

What the evidence settles is narrower than the legend and more useful. Unfinished things are not remembered better. People go back to them, and they intrude on unrelated thought at some cost to sleep. A specific plan someone means to keep can switch that off. All of that is solid. What it leaves open is the case the old finding was usually borrowed to describe, the thing that can no longer be finished, and why it holds some people and lets others go.