Why Starting Is Harder Than Continuing

Starting a task and continuing one are different problems that respond to different interventions, and the evidence for that distinction is considerably better than the evidence for most of the advice built on top of it. Initiation fails for want of a cue. Persistence fails for want of a payoff. Confusing the two is why people apply motivational fixes to problems that are structural, and structural fixes to problems that are motivational.

The distinction is not folk wisdom

Alexander Rothman made the case formally in Health Psychology in 2000, in a short paper titled "Toward a theory-based analysis of behavioral maintenance" (volume 19, supplement, pages 64 to 69). His argument was that the dominant models of health behavior change had quietly assumed one set of psychological processes governs both phases, and that this assumption is wrong.

The decision to initiate a behavior, he argued, rests on expected outcomes: you start because you believe something good will follow. The decision to maintain a behavior rests on received outcomes: you continue because you are satisfied with what actually happened. Those are different criteria, and they can point in opposite directions. Someone can be highly motivated to begin, begin successfully, find the reality underwhelming, and stop, without their initial motivation ever having weakened.

That reframing has practical teeth. If you have never once done the thing, adding reasons will not help you, because you already have reasons. If you have done it twenty times and stopped, adding a better cue will not help you, because the cue was working.

Activation energy is a metaphor, not a mechanism

The popular framing of initiation is borrowed from chemistry: a behavior has an activation energy, and lowering it makes the reaction go. It is a good metaphor and it should be labeled as one. No study measures a quantity of activation energy. Nothing is being conserved. The framing is a restatement of the observation, not an explanation of it.

The most-cited formal version, B. J. Fogg's behavior model, holds that behavior occurs when motivation, ability and a prompt converge, and that raising ability is usually cheaper than raising motivation. That is almost certainly directionally right, and it is worth knowing where it comes from: a 2009 conference paper at Persuasive Technology, later expanded into a trade book. It has not been tested as a model through randomized trials of its specific predictions. Treat it as a useful organizing scheme rather than a finding.

What is a finding, and a clean one, comes from Peter Gollwitzer and Veronika Brandstaetter in the Journal of Personality and Social Psychology in 1997 (volume 73, pages 186 to 199). Students named projects they intended to complete over the Christmas break, some easy and some difficult, and half were prompted to specify when and where they would carry each one out. Among the difficult projects, completion rose from 22 percent to 62 percent. Among the easy ones, the plan made no difference at all.

That asymmetry is the whole point. The plan did nothing where initiation was not the bottleneck, and tripled completion where it was. Naming a moment is an initiation intervention, and it only pays where initiation is what failed.

The resumption effect, which is the part that held up

There is a well-known claim that unfinished tasks stay lodged in memory. Romain Ghibellini and Beat Meier pooled that literature in 2025 and it did not survive. Across 37 publications, the ratio of recalled interrupted tasks to recalled completed tasks came out at 0.99, which is no effect at all. The full account of that collapse is documented elsewhere in this cluster.

Buried in the same meta-analysis in Humanities and Social Sciences Communications is a second, much less famous effect that came through intact. The Ovsiankina effect, named for Maria Ovsiankina's 1928 work, is the tendency to resume an interrupted task when given the opportunity. Pooled across 20 publications, the resumption rate was 66.79 percent.

Read against the initiation literature, that number is the most interesting thing in this article. Roughly two out of three people spontaneously pick an interrupted task back up. Nothing in the initiation literature suggests two out of three people spontaneously begin an untouched one. A partially completed task appears to carry its own prompt in a way an unstarted task does not.

The obvious tactic that follows, deliberately stopping mid-sentence so the return is easier, has been advocated by writers for a century and, as far as I can find, has never been tested directly. It is an extrapolation from a resumption rate to an intervention, and it should be described that way.

The two-minute rule and what actually holds it up

There are two different rules with this name. David Allen's, from Getting Things Done, says that if a task takes under two minutes you should do it immediately rather than track it. James Clear's, from Atomic Habits, says a new habit should be scaled down until it takes under two minutes.

Neither has been tested. There is no study comparing a two-minute threshold against one minute or five, and no trial of either rule as stated. Anyone presenting the number as research-backed is not describing the literature.

What does support the direction of the advice is indirect but real. Piers Steel's meta-analysis in Psychological Bulletin in 2007 (volume 133, issue 1, pages 65 to 94) pooled 691 correlations and identified the strongest and most consistent predictors of procrastination: task aversiveness, task delay, self-efficacy and impulsiveness. Shrinking a task attacks two of those four directly. It makes the task less aversive and it shortens the delay between starting and finishing. That is a coherent mechanism, and it is the honest version of the claim.

The other honest note is that shrinking is an initiation intervention specifically. Rothman's framework predicts it will do nothing for maintenance, because a two-minute version of a behavior produces a two-minute version of the received outcome, which is exactly the satisfaction signal maintenance depends on.

How to tell which problem you have, ranked

1. Count, do not introspect. Never started once is an initiation failure. Started and stopped is a maintenance failure. The two feel identical from the inside and require opposite responses.
2. For initiation, name a situation rather than a time. Gollwitzer and Brandstaetter's plans specified when and where. An event you already perform without deciding to is a more reliable trigger than a clock time you intend to respect.
3. For initiation, shrink until refusing costs more than complying. Not because two minutes is special, but because aversiveness and delay are two of Steel's four predictors.
4. For maintenance, change the payoff, not the plan. If the received outcome is unsatisfying, a better cue delivers you more efficiently to something you do not want.
5. Exploit resumption rather than fighting initiation. Two thirds of people return to an interrupted task. Almost nothing produces that rate for an unstarted one.
6. Change the environment before changing yourself. Structural interventions outperform motivational ones on both phases, which is the argument developed at length in the case for treating friction as the variable.

The honest summary

The distinction between starting and continuing is well founded. The mechanism for initiation, that naming a situation converts an intention into a trigger, has strong experimental support. The resumption effect is real and underused. Almost everything else in this territory, including the activation-energy framing, the specific two-minute threshold and the stop-mid-sentence tactic, is plausible extrapolation rather than tested intervention.

That is not a reason to ignore it. It is a reason to describe it accurately, and to notice that the interventions with the strongest evidence are also the least interesting: decide the moment in advance, make the first action small, and stop asking whether you feel like it. The same principle governs the question of when in the day difficult work belongs, where the popular advice again outruns what has actually been shown.