Hofstadter’s law is the rule that “it always takes longer than you expect, even when you take into account Hofstadter’s Law.” Douglas Hofstadter wrote it in his 1979 book Gödel, Escher, Bach. The joke is the second half: knowing you underestimate, and padding your plan for it, still does not save you. Psychologists study the same pattern under the name planning fallacy, and the research shows what actually helps.
Key Takeaways
- Hofstadter’s law comes from Gödel, Escher, Bach (1979), in a passage on chess computers that stayed “ten years away” from beating the world champion.
- The planning fallacy is the research name for it. Thesis students in a 1994 study predicted 33.9 days and took 55.5; fewer than a third finished by their own estimate.
- Padding is not enough: even by their worst-case date, more than half of the students had not finished.
- What helps: how long similar tasks really took you, explicitly linked to this one, and the task split into its steps before you estimate.
- Track actual durations, because memory makes past tasks look shorter than they were.
What is Hofstadter’s law?
Hofstadter’s law is a self-referential saying about time estimates: every task takes longer than you expect, and that stays true even after you adjust your expectation for the law itself. It describes a bias that survives being known about. You can add a buffer, and the task will still overrun the buffer.
Douglas Hofstadter, a cognitive scientist, introduced it in Chapter V of Gödel, Escher, Bach: An Eternal Golden Braid, “Recursive Structures and Processes”. The context was chess programs. In the early days of computer chess, he wrote, people estimated a computer would be world champion in ten years, “but after ten years had passed, it seemed that the day a computer would become world champion was still more than ten years away”. He called that “one more piece of evidence for the rather recursive Hofstadter’s Law”. A computer finally won in 1997, when IBM’s Deep Blue beat world champion Garry Kasparov in a match.
The law is often quoted by programmers, because software estimates miss in exactly this way, but it fits any plan: a home renovation, exam revision, a product launch, a Saturday of errands.
Why does everything take longer than expected?
Things take longer than planned mainly because people estimate from a picture of the plan going well, not from how similar tasks went before. Psychologists call this the planning fallacy, a term Daniel Kahneman and Amos Tversky coined in 1979: a confident belief that your own project will run on schedule, even while knowing most similar projects ran late.
The best-known evidence comes from Roger Buehler, Dale Griffin and Michael Ross in 1994. They phoned 37 psychology students in the final semester of an honours thesis and asked when they would hand it in. The students predicted, on average, 33.9 days. They took 55.5 days. Only 29.7% finished by their own best guess.
Their other studies explain why. When students thought aloud while estimating, they talked mostly about the future: the steps ahead and how they would get through them. Past experience made up about 7% of their thoughts, and possible problems about 3%. Asked about a time they had finished late, they blamed causes that were more external, temporary and specific than the ones they gave for friends who ran late. In a fifth study, a separate group of observers estimating other people’s tasks erred the other way: they overestimated how long the tasks would take, and leaned more on what had happened before.
A second cause sits in memory itself. A 2005 review by Michael Roy, Nicholas Christenfeld and Craig McKenzie argues that people do base predictions on past durations, but they remember those past tasks as shorter than they were. If your memory of “last time” is too short, an estimate built honestly on it is too short as well.
Why doesn’t adding a buffer fix it?
A buffer helps less than it should because the padding comes from the same optimistic picture as the estimate. That is the recursive part of Hofstadter’s law. When you widen an estimate by imagining things going badly, you still imagine a version of the plan, and people rarely think about what could get in the way: in one study, possible problems made up 3% of participants’ thoughts while they estimated.
The thesis study measured this directly. Besides a best guess, the students gave a date “if everything went as poorly as it possibly could”. That worst-case estimate averaged 48.6 days, below the actual 55.5 (a gap too small to be statistically significant), and fewer than half of the students (48.7%) finished by it. Their best-case estimate, 27.4 days, was met by 10.8%.
So a buffer is still worth having, but it works better when its size comes from data than from a feeling. A buffer of “20% more than my estimate” inherits every blind spot of the estimate. A buffer sized from how much your last ten similar tasks overran does not.
Hofstadter’s law vs Parkinson’s law
Hofstadter’s law and Parkinson’s law describe opposite failures of the same estimate. Hofstadter’s law says work overruns the time you planned for it. Parkinson’s law, from C. Northcote Parkinson’s 1955 essay in The Economist, says “work expands so as to fill the time available for its completion”: give a task too much time and it grows to use it.
| Idea | What it says | Where it bites | What helps |
|---|---|---|---|
| Hofstadter’s law | Work takes longer than expected, even with a buffer | Short, optimistic estimates | Estimates based on past actuals |
| Planning fallacy | People underestimate their own task times | Your own plans more than other people’s | The outside view: similar past tasks |
| Parkinson’s law | Work expands to fill the time available | Generous, open-ended deadlines | A firm time box for each task |
Both can be true of the same person in the same week. The cure for one can feed the other: pad every estimate heavily and Parkinson’s law fills the padding. Our Parkinson’s law glossary entry covers that side, and the usual way through both is a realistic estimate inside a fixed time box.
How do you estimate time more accurately?
To estimate more accurately, start from how long similar tasks actually took, not from how this one should go. Then break the task into its steps, compare your number with an outside opinion, and record what really happens so the next estimate starts from real data instead of memory.
1. Start from similar past tasks
Before you picture the plan, find two or three tasks like this one and check how long each took. Kahneman and Tversky called this using distributional information; Bent Flyvbjerg later built a forecasting method on it for large infrastructure projects, called reference class forecasting, which bases a forecast on the actual results of comparable projects. For a personal task, the reference class is your own history: last month’s report, the previous move, the last three times you said “a quick fix”.
2. Connect the past to this task
Remembering is not enough. In Buehler’s fourth study, 123 students predicted when they would finish a computer assignment. Students who only recalled past assignments were no more likely to finish by their predicted time than a control group (38.1% of them did, against 29.3% of the control group). Students who recalled them and then explained how those experiences applied to this assignment did much better: 60.0% finished by their predicted time. The authors note those predictions were less biased but no more accurate: their predictions leaned less toward optimism, but the typical miss, early or late, was just as large.
3. Unpack the task into steps
A task held in mind as one thing (“write the proposal”) hides its parts. Justin Kruger and Matt Evans found in 2004 that people prompted to list a task’s steps gave longer estimates, and in some experiments less biased ones, than people who estimated the task whole. The effect grew with complexity: the more parts a task had, the more unpacking helped. A brain dump is a quick way to get those steps onto paper.
4. Get a second estimate from someone else
In the same 1994 paper, observers predicting other people’s tasks did not share the task-doers’ optimism, but they were not more accurate either: they erred long by about as much as the people doing the tasks erred short. So treat a colleague’s estimate as the other end of a range, not as the right answer. When the two numbers are far apart, look at what your plan leaves out.
5. Record actual durations
Because memory shortens past tasks, the only reliable reference class is one you wrote down. Note the estimate when you start and the actual time when you finish. After a few weeks you will know your own overrun ratio for different kinds of work, and you can size buffers from that instead of from a guess. Reviewing the day at a shutdown ritual is a natural moment to log it.
6. Budget the buffer at the plan level
Put spare time into the day as a whole, not into every task, and protect it. A day planned to the minute has no room for the overrun Hofstadter’s law predicts; a day with unplanned time left in it does. If the day cannot hold its tasks plus that spare time, the problem is the number of tasks, not the estimates. Our article on overcommitting covers that side.
How FocusBox helps you estimate
FocusBox is a timeboxing app, on the web and on iPhone and iPad, and two of its planning features work from your own history, which is the outside view the research recommends.
Plan My Day, one of the daily planning rituals, gives each task “a time estimate from your own history” and shows a capacity meter that warns you before you plan more than the day can hold. AI Time Blocking schedules today’s tasks into the free gaps around your calendar, with durations that adapt to how long tasks actually take you, and it lists anything that does not fit separately instead of squeezing it in. AI Time Blocking runs in the web app and syncs the planned day to the iOS app.
Frequently Asked Questions
What is Hofstadter’s law in simple terms?
Hofstadter’s law says that every task takes longer than you expect, even when you already know about the law and allow extra time for it. Douglas Hofstadter coined it in his 1979 book Gödel, Escher, Bach. It is a joke with a real point: knowing you underestimate is not enough to stop you underestimating.
Who came up with Hofstadter’s law?
Douglas Hofstadter, an American cognitive scientist, wrote it in Gödel, Escher, Bach: An Eternal Golden Braid, published in 1979. It appears in Chapter V, in a passage about chess programs that people kept predicting would beat the world champion within ten years. A computer, IBM’s Deep Blue, finally won a match against world champion Garry Kasparov in 1997.
What is the planning fallacy?
The planning fallacy is the tendency to underestimate how long your own tasks will take, even when you know similar tasks took longer in the past. Daniel Kahneman and Amos Tversky named it in 1979. In a 1994 study, students predicted their theses would take 33.9 days on average; they took 55.5 days.
What is the difference between Hofstadter’s law and Parkinson’s law?
Hofstadter’s law says work takes longer than the time you planned for it. Parkinson’s law says work expands to fill whatever time you give it. One describes estimates that are too short, the other deadlines that are too generous. A realistic estimate inside a firm time limit addresses both at once.
How do I stop underestimating how long tasks take?
Base each estimate on how long similar tasks actually took you, not on how this one should go, and write down real durations so you are not relying on memory, which tends to shorten past tasks. Break bigger tasks into steps before estimating them, and treat your own number and a colleague’s as the two ends of a likely range.
Sources
- Hofstadter. Gödel, Escher, Bach: An Eternal Golden Braid, Chapter V “Recursive Structures and Processes” (Basic Books, 1979; 20th anniversary edition 1999, Internet Archive; accessed October 2026)
- Buehler, Griffin and Ross. Exploring the “planning fallacy”: Why people underestimate their task completion times (Journal of Personality and Social Psychology, 1994; accessed October 2026)
- Kahneman and Tversky. Intuitive prediction: Biases and corrective procedures (TIMS Studies in Management Science 12, 1979, 313-327; as cited by Buehler, Griffin and Ross, 1994)
- Roy, Christenfeld and McKenzie. Underestimating the duration of future events: memory incorrectly used or memory bias? (Psychological Bulletin, 2005; accessed October 2026)
- Kruger and Evans. If you don’t want to be late, enumerate: Unpacking reduces the planning fallacy (Journal of Experimental Social Psychology, 2004; accessed October 2026)
- Flyvbjerg. From Nobel Prize to project management: getting risks right (Project Management Journal, 2006; accessed October 2026)
- Parkinson. Parkinson’s Law (The Economist, 19 November 1955)
