Zeigarnik Effect
Date: 2026-09-27
The claim that unfinished tasks are remembered better than finished ones. It’s the standard justification for onboarding checklists and “complete your profile” prompts — and the memory claim doesn’t survive a 2025 meta-analysis. What does hold up is a separate effect: people tend to go back and finish what was interrupted.
The Zeigarnik effect is the claim that interrupted or unfinished tasks are recalled better than completed ones.
The Ovsiankina effect is its sibling: the tendency to resume an interrupted task when given the chance. Both come from Kurt Lewin’s Berlin group in the 1920s — Bluma Zeigarnik (1927) and Maria Ovsiankina (1928).
What the evidence now says
A 2025 meta-analysis (Ghibellini and Meier, Humanities and Social Sciences Communications) pooled the replication literature:
CLAIM RESULT
ZEIGARNIK (memory) unfinished tasks recalled no general advantage
better — fewer than a third of
replications found one
OVSIANKINA (resumption) interrupted tasks get holds up: roughly two-thirds
resumed resumed, against a 50% baseline
[CHECK: the paper’s exact pooled figures before quoting them.]
In plain terms: people don’t reliably remember unfinished things better. They do tend to go back to them. The UX literature routinely cites the first to justify designs that actually rely on the second.
Why the distinction matters. If the mechanism were memory, a reminder would be redundant — the unfinished task would already be on their mind. Because it’s resumption, the product has to offer the way back: a saved state and a prompt that returns them to exactly where they stopped.
Where it’s used
- Onboarding checklists — “3 of 5 steps done”. Mostly goal gradient and resumption, not memory — Goal Gradient Effect
- Profile completion meters — LinkedIn’s is the canonical example
- Saved baskets and abandoned-basket emails — the resumption prompt, with the saved state doing the real work — Lifecycle Messaging
- Series and cliffhangers — episodic content ending mid-story
- Multi-step forms with saved progress — “continue your application”
RESUMPTION NEEDS BOTH HALVES
STATE PRESERVED PROMPT BACK TO IT RESULT
yes yes, deep-linked resumes
yes no relies on them remembering ← weak
no (basket cleared) yes prompt to a dead end
no no nothing
Where it goes wrong
- Citing it as settled memory science. It’s a 1927 finding that doesn’t replicate in its original form. Say “people tend to resume interrupted tasks” and you’re on solid ground
- Manufacturing unfinishedness. Inventing incomplete states (“your profile is 40% complete” when nothing’s needed) to create compulsion is a dark pattern in the making — Deceptive Design
- Nagging. A resumption prompt repeated after the person has decided not to finish is a prompt below the action line — Fogg Behaviour Model, Reactance
- Breaking the saved state. Session expiry, cleared baskets, or a login wall between the prompt and the task undo the whole mechanism — Sessions and Tokens