The Zeigarnik Effect: Why Unfinished Tasks Won’t Leave You Alone

That last point reverses the practical advice usually attached to this effect, and it is the part worth carrying away.

The original finding

Bluma Zeigarnik was working in Kurt Lewin's group in Berlin in the 1920s. The origin story — that Lewin noticed a waiter who could recall unpaid orders in detail and forgot them the moment the bill was settled — is repeated everywhere and should be treated as anecdote rather than data.

The study itself was real. Participants were given a series of short tasks — puzzles, modelling clay, arithmetic — and interrupted partway through roughly half of them. Later, without warning, they were asked to list the tasks they could remember.

Interrupted tasks were recalled substantially more often than completed ones, by a ratio in the region of two to one. Zeigarnik's interpretation, following Lewin's field theory, was that starting a task creates a state of tension that persists until the task is finished, and that this tension keeps the task accessible.

What happened to it afterwards

This is where the popular account stops and the actual literature gets complicated.

Attempts to reproduce the effect over the following decades produced inconsistent results. A critical review by van Bergen in 1968 examined the accumulated evidence and concluded that the effect was considerably less reliable than its textbook status implied — sometimes appearing, sometimes absent, sometimes reversed.

The moderators that emerged are informative. The effect tends to appear when participants are personally invested in the task and expect to complete it, and to weaken or vanish when the task is trivial, when interruption is interpreted as failure, or when the participant has disengaged from the goal. Motivation is doing much of the work that "tension from incompletion" was supposed to do on its own.

So the honest position on the classic effect is: a real phenomenon under some conditions, not a general law of memory, and considerably less robust than the confident version in productivity writing.

The modern reframing

The more interesting recent work does not test recall of interrupted tasks at all. It tests what unfinished goals do to you while they are unfinished.

E. J. Masicampo and Roy Baumeister ran a series of studies showing that activated but unfulfilled goals produced intrusive thoughts during an unrelated reading task, raised the mental accessibility of goal-related words, and impaired performance on unrelated tasks requiring executive function — anagrams, logic problems, impulse control.

The specificity is what makes it convincing as a set. The interference appeared only for goals participants personally held, disappeared when the goal was fulfilled, and was not produced by mere frustration or by reflecting on someone else's unfinished goal.

Then the finding that matters most: allowing participants to write a specific plan for how they would complete the unfulfilled goal eliminated the intrusions and the interference — without the goal being completed. The reduction was mediated by how earnest the plans were, and those who made serious plans were also the ones who later carried them out.

The implication is that the persistent activation is not a demand for completion. It is a demand for resolution of uncertainty about completion. Once the mind has a concrete plan, it releases the resources it was spending on keeping the goal accessible. This is also the more defensible version of the frequently-made parallel to songs that will not leave your head — the unresolved-loop framing is intuitive, but the earworm literature is separate and the shared mechanism is assumed rather than demonstrated.

The forgotten sibling

Zeigarnik's colleague Maria Ovsiankina reported a companion finding a year later, and it is arguably the more robust of the two. Given the chance, people spontaneously resume interrupted tasks — even trivial ones with no reward attached, and even when nobody has asked them to.

The resumption effect matters because it is behavioural rather than a memory measure. It does not depend on what participants can recall under questioning, which is where much of the original effect's fragility lives. Someone quietly going back to finish a jigsaw they were interrupted on is harder to explain away than a recall score.

How much to trust the reframing

This needs saying, because the plan-making result is genuinely useful and it comes from a specific place.

The work emerged from Baumeister's laboratory during the period when that lab's best-known finding — ego depletion — was reported extensively and then failed to replicate in large multi-site registered efforts. That does not make this particular set of studies wrong. It does mean the surrounding literature was produced under methodological norms that have since been substantially revised, and no large registered replication of the plan-making effect has been published.

Treat it as a promising and internally coherent finding rather than settled fact. Its practical recommendation happens to be free and harmless, which is a reasonable basis for trying it regardless of how the replication question eventually resolves.

What this predicts about everyday life

Several familiar things fall out of it. Cliffhangers work because an unresolved narrative goal stays active. Progress bars and partially completed loyalty cards exploit the same accessibility. Lying awake mentally circling an unfinished task is the interference effect running at a moment when you have nothing to distract you from it.

And the standard advice — write it down before bed — is better supported by the plan-making result than by the classic Zeigarnik effect. Writing "deal with the invoice" is not the operative move. Writing when, where and how you will deal with it is.

It is worth noting the contrast with the other memory quirk in this hub. Here, an open goal is held onto tenaciously and intrudes where it is not wanted. In the doorway effect, an open goal is discarded the moment you cross a threshold. Both are true, and reconciling them is an open question rather than a contradiction: goal accessibility and working-memory contents are not the same store, and a boundary appears to flush one while leaving the other running.