A to-do list is a pile of unpaid loans: the Zeigarnik maths

Open tasks do charge interest, paid in intrusive thoughts and lost focus. The famous memory effect behind the idea barely replicates, but the interest is real, it can spiral, and a written plan refinances it to almost zero.

An old hand-written account book ledger lying open, its pages ruled into columns of entries
Photo: John Carlin, CC0, via Wikimedia Commons

A to-do list really does behave like a pile of unpaid loans: each open task charges a small daily interest in attention, in the form of intrusive thoughts, slower work on other things and rumination after hours. A 2026 meta-analysis of 31 samples found unfinished work tasks reliably linked to off-job work thoughts (ρ = 0.25 to 0.38). But the famous claim that we remember unfinished tasks better barely survives: across 59 publications the memory advantage was a ratio of 0.99, essentially none. And unlike a bank loan, this interest can be cut to almost zero by writing a specific plan.

Key takeaways
  • 1.9 to 2.0: Zeigarnik's 1927 ratio of unfinished to finished tasks recalled; the 2025 meta-analysis puts it at 0.99.
  • 67%: how often people resume an interrupted task when given the chance (the Ovsiankina effect), against 50% by chance.
  • ρ = 0.38: link between unfinished work and after-hours work thoughts across people (Wendsche 2026).
  • 6 minutes: in the queue model below, the daily attention tax per open task above which a to-do list spirals instead of settling.
  • Plan making removed the intrusive-thought cost of unfinished goals in Masicampo and Baumeister's experiments.

This is a collision post: a finance idea and a psychology idea, tested for a real link. The test has three parts. Is there really a principal and an interest? Does the interest compound? And can you refinance? Each gets a number.

Where did the idea of open loops come from?

From a Berlin laboratory in the 1920s, where Bluma Zeigarnik, a student in Kurt Lewin's group, tested whether interrupted tasks stay in the mind. Lewin's theory was that forming an intention creates a tension that only discharges when the task is done. An unfinished task is a tension still waiting to be paid off.

The story

Zeigarnik gave each person about twenty small tasks in one sitting: continue a honeycomb pattern, string beads, punch holes, solve a jigsaw, name a German philosopher, actor and city that start with the same letter. She let them finish half and interrupted the other half by handing over a new task just as they were close to a solution. After each task, the materials went into a drawer "to keep the table tidy". At the end of the hour she simply asked what they had been doing. Thirty-two people in the first series named unfinished tasks 1.9 times as often as finished ones; fourteen in a second series, 2.0 times. Kurt Koffka, writing in 1935, praised it as one of the finest experiments from Lewin's school.

Koffka K.: Principles of Gestalt Psychology (1935), pages 334 to 336
Portrait photograph of Bluma Zeigarnik as a young woman in 1921
Bluma Zeigarnik in 1921, a few years before the experiment that made her name a household word in productivity writing.Photo: Andrey Zeigarnik, Public domain, via Wikimedia Commons

A popular version of the origin story has Lewin noticing that a waiter remembered unpaid orders until they were settled. Wikipedia attributes it to Koffka's book, but I could not find the anecdote in the text of that book, so treat it as folklore. The experiment itself is documented.

Do we really remember unfinished tasks better?

Mostly not. The memory half of the Zeigarnik effect has been hard to replicate for nearly a century, and the first meta-analysis, published in 2025, found no general memory advantage for unfinished tasks.

Interruption, recall and resumption: a meta-analysis of the Zeigarnik and Ovsiankina effectsRomain Ghibellini, Beat Meier · Humanities and Social Sciences Communications · 2025

The authors screened 1,349 publications and kept 59 that used the classic interrupted-task design: 38 on recall (Zeigarnik), 20 on resumption (Ovsiankina) and one on both. Excluding Zeigarnik's own study, the weighted ratio of interrupted to completed tasks recalled was 0.99, and interrupted tasks made up 49.16% of what people recalled. The small set of studies that allowed an effect size gave d = 0.15. By contrast, people resumed interrupted tasks 67% of the time, well above the 50% chance level. Limit: many old studies reported too little to compute effect sizes, so the authors used weighted averages rather than a full random-effects model.

Read the paper (free)
Unfinished vs finished tasks: memory and the urge to resume
See the data as a table
Recall ratio, unfinished ÷ finished
Zeigarnik 1927, series 11.9
Zeigarnik 1927, series 22.0
Meta-analysis 2025 (excl. Zeigarnik)0.99

Source: Koffka 1935; Ghibellini and Meier 2025 · A ratio of 1.0 means no memory advantage. The same meta-analysis found interrupted tasks were resumed 67% of the time against 50% by chance.

So the original "loan ledger in memory" is weak. What survives is the urge to pay: the Ovsiankina effect, named after Maria Ovsiankina, who showed in 1928 that people tend to go back to interrupted tasks unprompted. The meta-analysis authors suggest the memory effect needed special conditions (an authoritative experimenter, high involvement) that were more common in 1920s labs than today.

So what is the interest an open task charges?

Not better memory but intrusion: an unfinished goal keeps pulling attention while you try to do something else. This is the part with strong, modern evidence.

  • Intrusive thoughts. In Masicampo and Baumeister's 2011 experiments, unfinished goals produced more intrusive thoughts during an unrelated reading task, made goal words more mentally accessible and worsened performance on an unrelated anagram task.
  • Attention residue. When you switch away from an unfinished task, part of your mind stays with it. The idea is covered in depth in attention residue and task switching.
  • After-hours rumination. The 2026 meta-analysis by Wendsche, Weigelt and Syrek found unfinished work linked to work thoughts in off-job time at ρ = 0.38 between people and 0.25 within the same person day to day, strongest for affective rumination.
Managing daily work intrusions: An intervention to reduce attention residue and exhaustionWard van Zoonen, Yuri S. Scharp · International Journal of Stress Management · 2025

A field experiment with a daily diary design at the Dutch branch of a technology consulting firm: 50 employees over 4 days (200 day-records). On days with more interruptions, employees reported more attention residue and more exhaustion. Writing short "ready-to-resume" plans, noting where they were and what came next before switching, reduced attention residue and weakened its link to exhaustion. Limit: a small, single-company sample, self-reported measures and only four days.

Read the paper (free accepted version)
The interest on an open task is not paid in memory. It is paid in the moments it steals from everything else.The author, summarising the evidence above
Bronze bust sculpture of the psychologist Kurt Lewin
A sculpture of Kurt Lewin, whose tension theory predicted that an unfinished intention keeps pulling at the mind until it is discharged.Photo: Pko (photographer), CC BY-SA 4.0, via Wikimedia Commons

Does the interest compound?

It can, through a feedback loop: open tasks tax attention, the tax slows your work, slower work leaves more tasks open, and more open tasks raise the tax. Queueing theory lets us test when that loop settles and when it runs away.

The theory, in plain words

Little's law says that in any steady flow, the average number of items waiting equals the arrival rate times the average time each item waits: \( L = \lambda W \). It holds for bank queues, factory lines and to-do lists alike. The simplest queue model, with tasks arriving at rate \( \lambda \) a day and being finished at capacity \( \mu \) a day, gives the average number of open tasks:

\[ L_0 = \frac{\lambda}{\mu - \lambda} \]

Worked derivation: adding the attention tax

Assumptions. You get 8 new tasks a day (\( \lambda = 8 \)) and can finish 10 on a clear day (\( \mu = 10 \)). You have \( T = 480 \) working minutes. Each open task costs \( c \) minutes of attention a day in intrusions and residue; I use \( c = 5 \) as an illustrative estimate, since no study has measured a per-task minute cost. Lost minutes reduce capacity in proportion.

  1. No tax. \( L_0 = 8/(10-8) = 4 \) open tasks; by Little's law each waits \( W = 4/8 = 0.5 \) day.
  2. Capacity after tax. \( \mu_{\text{eff}} = \mu\left(1 - \dfrac{cL}{T}\right) = \mu - aL \), where \( a = \dfrac{c\mu}{T} = \dfrac{5 \times 10}{480} = 0.104 \).
  3. Self-consistent queue. Substitute into the queue formula:
    \[ L = \frac{\lambda}{\mu - aL - \lambda} \;\Rightarrow\; aL^2 - (\mu-\lambda)L + \lambda = 0 \]
  4. Solve. \( L = \dfrac{2 - \sqrt{4 - 4(0.104)(8)}}{2(0.104)} = \dfrac{2 - 0.816}{0.208} = 5.7 \) open tasks, costing \( 5 \times 5.7 = 28 \) minutes a day.
  5. Find the cliff. A real solution exists only while \( (\mu-\lambda)^2 \ge 4a\lambda \). Solving for \( c \):
    \[ c^{*} = \frac{(\mu-\lambda)^2\,T}{4\lambda\mu} = \frac{4 \times 480}{4 \times 8 \times 10} = 6\ \text{minutes} \]
    Above 6 minutes of attention per open task per day, there is no steady state. The list grows without limit: a debt spiral.

That is the compounding. A loan at a fixed rate grows as \( A = P(1+r)^n \); for example ₹10,000 at an illustrative 3% a month becomes about ₹14,258 after 12 months. The to-do list compounds differently, through the feedback in step 3, and it has a sharp threshold rather than a smooth curve.

The to-do list as a queue with a feedback leakopen tasks Lnew tasks λ = 8/daydone μ = 10/dayattention tax c × L slows the outflowplan made:tax ≈ 0
Tasks flow in and out. Every open task leaks a small attention tax that slows the outflow, which leaves more tasks open. A specific plan closes the leak. Model by the author; arrival and capacity numbers are illustrative.

Sensitivity check

ScenarioTax per open taskOpen tasks LAttention lost a day
No attention tax0 min4.00 min
Plans written for most tasks1 min4.24 min
Mild tax3 min4.714 min
Base case5 min5.728 min
At the cliff6 min8.048 min
Over the cliff8 minno steady stategrows daily
One fewer task a day (λ = 7), 5 min tax5 min2.613 min
Open tasks vs attention tax (8 in, 10 out a day)attention tax per open task (min/day)0356848cliff: no steady statethe list grows dailyplans keep you here
Average open tasks from the queue model as the attention tax per task rises. The curve stays flat, then climbs steeply and ends at 6 minutes, beyond which no steady state exists. Calculated by the author; inputs are illustrative.

Two levers stand out. Cutting the tax per task (by planning) and cutting the inflow (by saying no to one task a day) both move you far from the cliff. With seven tasks a day instead of eight, the cliff moves from 6 to about 15 minutes per task.

Use this to check what you just read. Put in your own rough numbers.

Is your to-do list settling or spiralling?

Where does the loan analogy break?

In three places, and they matter more than the places it holds.

  • The interest can be cancelled. Masicampo and Baumeister found that letting people write a specific plan for an unfinished goal eliminated the intrusions, the word-accessibility effect and the anagram penalty. No bank forgives interest because you wrote a repayment schedule. The mind does.
  • Not every open loop charges. Ghibellini and Meier found the Zeigarnik memory effect depended on involvement and setting; low-stakes tasks seem to charge little. A loan charges whether you care about it or not.
  • Default is allowed. Deleting a task you no longer value closes the loop cleanly. That is the one thing you cannot do with a real debt.
A page from a nineteenth-century merchant's business ledger listing customer accounts and amounts owed
A merchant's ledger of customer accounts. Every open line is a debt someone has to think about; every settled line can be forgotten.Photo: Unknown, Public domain, via Wikimedia Commons

How do you refinance a to-do list?

By converting each open loop into a specific plan, which is what cut the cost in every study above. Here is the procedure, built from those findings.

  1. Write it, don't hold it. Get every open task out of your head and onto one list. Holding them in mind is where the tax is charged; see the working memory limit.
  2. Add the next action and a when. "Call the insurer, Tuesday 10 am" is a plan; "insurance" is a loan. Masicampo and Baumeister's effect came from specific plans that people actually carried out.
  3. Leave a ready-to-resume note. When you are interrupted, write one line on where you stopped and what comes next, as in van Zoonen and Scharp's intervention.
  4. Write tomorrow's list at bedtime. In Scullin's sleep-lab study, five minutes of writing a specific to-do list helped people fall asleep faster than writing about what they had done. More in the CBT-I sleep guide.
  5. Close or delete small items first. Each closed task stops charging. Then protect a block for the big ones, as in the deep work routine.
  6. Cut the inflow. One fewer new commitment a day moved the spiral threshold from 6 to 15 minutes in the model above.
The effects of bedtime writing on difficulty falling asleep: A polysomnographic study comparing to-do lists and completed activity listsMichael K. Scullin, Madison L. Krueger, Hannah K. Ballard, Natalya Pruett, Donald L. Bliwise · Journal of Experimental Psychology: General · 2018

Fifty-seven healthy adults aged 18 to 30 wrote for 5 minutes before a night of polysomnography in a sleep lab, randomly assigned to write a to-do list for the coming days or a list of tasks already completed. The to-do group fell asleep significantly faster, and the more specific their list, the faster they fell asleep; the reverse trend appeared for the completed list. Limit: one night, young healthy adults, and no group of people who wrote nothing.

Read the paper (free)

Turning these into a lasting habit takes repetition rather than willpower; the habit guide covers how.

Loan termTo-do equivalentEvidence
PrincipalThe work the task needsFixed; planning does not shrink it
InterestIntrusions, residue, after-hours ruminationMasicampo 2011; van Zoonen 2025; Wendsche 2026
Urge to repayOvsiankina effect, 67% resumptionGhibellini and Meier 2025
Ledger memoryZeigarnik memory effectRatio 0.99: weak or absent
RefinancingA specific, dated planIntrusions eliminated (Masicampo 2011)
Debt spiralTax per task above the queue cliffAuthor's queue model; not measured
Verdict: true about the interest, false about the ledger, and better than a loan

Open tasks do charge interest: unfinished goals cause intrusions and lost performance, and unfinished work tracks after-hours rumination at ρ = 0.25 to 0.38. The interest can compound; in a simple queue with 8 tasks in and 10 out a day, an attention tax above about 6 minutes per open task per day has no steady state. But the "we remember what we owe" part of Zeigarnik's effect fails replication (ratio 0.99 across 59 publications), and unlike any bank, the mind forgives the interest the moment you write a specific plan.

What if next?

Questions people ask

What is the Zeigarnik effect in simple words?

The claim that people remember interrupted or unfinished tasks better than finished ones. Zeigarnik reported a ratio of about 2 to 1 in 1927, but a 2025 meta-analysis of 59 publications found a ratio of 0.99, meaning no general memory advantage.

Why can't I stop thinking about unfinished work?

Unfinished goals stay active and intrude on other activities until they are completed or planned. The 2026 meta-analysis found this link strongest for affective rumination. Writing a specific plan for when and how you will finish removed the intrusions in Masicampo and Baumeister's experiments.

Does writing a to-do list help you sleep?

In Scullin's 2018 sleep-lab study, people who wrote a specific to-do list for five minutes at bedtime fell asleep faster than people who wrote about tasks they had completed.

Is the Zeigarnik effect real or a myth?

Half and half. The memory effect is weak or absent in modern data. The related Ovsiankina effect, the urge to go back and finish an interrupted task, is reliable at about 67% resumption.

Sources

  1. Ghibellini R., Meier B.: Interruption, recall and resumption: a meta-analysis of the Zeigarnik and Ovsiankina effects, Humanities and Social Sciences Communications (2025) (source for 59 publications, ratio 0.99, 49.16%, d = 0.15, 67% resumption)
  2. Koffka K.: Principles of Gestalt Psychology (1935) (source for Zeigarnik's tasks and the 1.9 and 2.0 ratios)
  3. Masicampo E.J., Baumeister R.F.: Consider it done! Plan making can eliminate the cognitive effects of unfulfilled goals, Journal of Personality and Social Psychology (2011) (source for intrusions and the plan-making effect)
  4. Wendsche J., Weigelt O., Syrek C.J.: Unfinished work tasks and work-related thoughts during off-job time: meta-analysis, Anxiety, Stress and Coping (2026) (source for ρ = 0.382 and 0.247)
  5. van Zoonen W., Scharp Y.S.: Managing daily work intrusions, International Journal of Stress Management (2025) (source for the ready-to-resume intervention)
  6. Scullin M.K. et al.: The effects of bedtime writing on difficulty falling asleep, Journal of Experimental Psychology: General (2018) (source for the bedtime to-do list result)
  7. Brown A.L., Lahey J.N.: Small Victories: Creating Intrinsic Motivation in Savings and Debt Reduction, NBER Working Paper 20125 (2014) (source for the smallest-first ordering result)
  8. Wikipedia: Little's law (the queueing law used in the derivation)
  9. Wikipedia: Zeigarnik effect (origin of the waiter anecdote, attributed to Koffka)

More on the science of the mind on the Manthan world page.

Zeigarnik effectOvsiankina effectto-do listopen loopsattention residueLittle's lawplan makingcompound interest
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Mindblog.

Keep reading in Mindblog

From the other blogs