Finished
- Invoice paid
- Appointment confirmed
- Parcel collected
Said to fade
Interrupted
- Puzzle half solved
- Form up to step 2
- Article half read
Said to stay present
Evidence
Zeigarnik (1927) gave 164 participants 18 to 22 small tasks each — building cardboard boxes, moulding clay figures, puzzles, arithmetic — interrupted half of them and then asked which ones they remembered. In the main experiment with 32 people the interrupted tasks were named 90 percent more often (ratio 1.9 — a ratio of 1.0 would mean no difference); a repetition with 15 new people gave 2.0, and two group experiments gave 1.9 for 47 adults and 2.1 for 45 children. The well-known waiter story comes from the circle of Kurt Lewin and is an anecdote, not part of the experiment.
Her own paper already states the limits. With tired participants the finding reversed (0.61), with recall on the following day a 14 percent advantage remained, and a distraction of 10 to 30 minutes often erased it altogether. Among schoolchildren who worked through the tasks merely as instructed the ratio was 1.03. And when participants were told all the tasks in advance, the advantage disappeared (0.97). These side conditions rest on 3 to 17 people each, however.
The 2025 overall picture is unambiguous. Ghibellini and Meier conclude that the memory advantage for unfinished tasks does not hold generally. Across 38 reports the ratio of recalled interrupted to recalled completed tasks is 0.99 — that is, no difference. Taking interrupted tasks as a share of all recalled tasks gives 49.2 percent, if anything a small disadvantage. The mean effect size is d = 0.15; 0.2 already counts as small.
The figure the meta-analysis reports for Zeigarnik’s own work stands alone. The authors also trace the 1.61 to a computation that weights individual outliers heavily. Zeigarnik herself reports that number nowhere: her main conditions give 1.9 to 2.1, her side conditions values around 1.0 and below.
Where the effect does appear, it depends on the situation: in a deliberately relaxed experimental atmosphere the ratio was 1.07, in a neutral one 0.96, in a performance-focused one 0.88. The authors suspect that Zeigarnik’s 1927 participants were more deeply involved in the tasks, partly because of the authority professors held at the time. That is an explanation, not a tested finding.
Limits
The effect concerns memory, not buying urge. Its use as a justification for progress bars is not supported by the meta-analysis. The related Ovsiankina effect is the better explanation. Cited here is the abridged English translation of the 1927 paper; two sections of the 85-page original appear there only as the editor’s summary.