How to finish what you start when the middle goes quiet

Sixty-nine students each proofread nine essays for typos. The ones who could see only which essay they were on caught 0.122 typos a second on the second essay and 0.124 on the eighth. On the fifth, halfway down the pile, the rate fell to 0.092. The essays were rotated between positions beforehand. The fifth one wasn't the hard one.

Every guide on how to finish what you start names the middle as the place things die. Not one of the pages ranking for that phrase today mentions this experiment, or explains the slump they all warn about. Effort sags halfway through because that's the point where you stop measuring from the start line and haven't yet begun measuring from the finish.

The fix that follows is dull. Count what's left instead of what you've done, and keep counting that way through the middle. In the same experiment, students whose screen showed how many essays remained had no dip at all, and their hit rate climbed as the pile shrank, from 0.088 to 0.098 to 0.13 typos a second.

Why you stop caring halfway through

Andrea Bonezzi, C. Miguel Brendl and Matteo De Angelis ran three experiments for that 2011 paper in Psychological Science. In the first, 126 students played a word game across nine rounds with $50 on the line for the top scorers. They found 7.13 words on the fifth round against 8.39 on the second and 8.45 on the eighth, with the target words rotated between positions so the content couldn't explain it.

The second experiment shows the mechanism. A hundred and thirty-seven students got an envelope holding $15 and a note about a charity trying to raise $300. Some read how much had come in already, others how much was still needed, and both were looking at the same drive at the same stage. The money-still-needed group gave $2.86 near the end and $1.10 in the middle. The money-collected group did the opposite, $2.68 at the start and $1.12 in the middle.

So there isn't one motivation curve. There are two, running opposite ways, and which one you're on depends on the number you happen to be looking at. In the third experiment, 75 percent of students given no progress frame switched from counting essays done to essays remaining partway through, and the slump sits in that handover.

Read the limits first. These were afternoon tasks of nine parts rather than a thesis, and the authors grant that ordinary fatigue could also explain the first result. The charity experiment is what rules fatigue out as a requirement, since nobody gets tired reading one line about a fundraiser.

Count what's left, unless you're not sure you care

The to-go count isn't a universal upgrade. Minjung Koo and Ayelet Fishbach held goal progress at roughly 50 percent across four studies and varied only which half people saw. For a core course exam, where commitment was already settled, seeing what was left to cover raised the time students set aside for it. For an elective, where commitment was shaky, the reverse worked, and seeing the ground already covered was what made the exam feel worth an evening.

That's a diagnostic rather than a slogan. If you can say in a sentence why the thing matters, count down the remainder. If you can't, counting your finished work is what keeps you in, which is the case we made for keeping a record of what closed. One refinement: read that record when you're wavering about a goal, not while grinding through the middle of one you already chose.

The same mechanism has a cruder version, tested on 300 car wash customers in April 2004. Joseph Nunes and Xavier Drèze handed out two loyalty cards that both required exactly eight paid washes, a plain eight-stamp card or a ten-stamp card arriving with two stamps already on it. Completion ran at 34 percent against 19 percent, and the head-start customers finished sooner as well. The work didn't change. Only the apparent start line did.

How to finish what you start, sorted by what has been measured

Here is the standard advice with the evidence attached, and the thin rows marked as thin.

Your finish date was wrong before you started

Roger Buehler, Dale Griffin and Michael Ross phoned 37 psychology students in the final semester of an honors thesis course and asked when they'd be done. The average prediction was 33.9 days and the average thesis took 55.5. Fewer than a third, 29.7 percent, finished inside their own best estimate. Asked for a worst case, assuming everything went as badly as it possibly could, they said 48.6 days and were still short.

The reason turned up when students thought aloud while predicting. Seventy-four percent of what they said was about how this job would go from here, and 7 percent touched anything they had done before. Three percent named an obstacle.

In a fourth study, students predicting when they'd finish a computer assignment were pushed to tie specific past assignments to this one, and the share finishing by their predicted date went from 29.3 percent to 60.0. Their forecasts got honest. The days they actually finished on barely moved, 6.3 to 7.0 across conditions, while the deadline itself moved behavior hard: a one-week deadline produced finished work in 4.3 days on average, a two-week deadline in 9.1.

So a better estimate buys you a plan that's true, and a shorter deadline is what buys you the work.

The pile of unfinished things is not eating your memory

One page ranking for this question explains the weight of unfinished work through the Zeigarnik effect, cited to a psychology explainer rather than to any study, and concludes that "unfinished projects reduce the energy needed to complete them." As a piece of history it is solid. As evidence it has been coming apart for sixty years.

Romain Ghibellini and Beat Meier pooled the literature in Humanities and Social Sciences Communications last year. Across 38 publications, the ratio of interrupted tasks recalled to completed tasks recalled came out at 0.99, with and without Bluma Zeigarnik's own 1927 data. Interrupted tasks made up 49.16 percent of what people remembered, which is what a coin would give you, and the average effect size was dz = 0.15. Van Bergen had got there by hand in 1968, running seven experiments of her own and calling the phenomenon a "non-problem" that should be "discarded", and fewer than a third of the 44 replication attempts she reviewed had found it.

What survived that meta-analysis is the half nobody quotes. Offer people the chance to pick an interrupted task back up and roughly 67 percent take it, a result that holds across 21 publications and carries a different name, Ovsiankina's rather than Zeigarnik's. Your half-finished work isn't haunting you. It's waiting, and the evidence says you'll resume it when resuming is easy.

The move that quiets an open loop is smaller than finishing it. E. J. Masicampo and Roy Baumeister had 71 students write about two important unfinished tasks, then read the first 3,200 words of a detective novel. Those students scored 6.13 of 8 on comprehension and reported being distracted by the tasks, while the ones who first wrote a specific plan for those same tasks scored 6.94 and reported no more intrusions than people thinking about finished work. A later study in the same paper found that the plan left anxiety untouched while still clearing the intrusions. So writing down where and when you'll do the thing buys back the attention, and listing the ways you might get to it does not.

Some of it is worth closing rather than finishing

There's a name for the state you're in when you honestly can't tell whether to carry on. Veronika Brandstätter and Marcel Herrmann call it an action crisis, the phase of being torn between pursuit and quitting. Following 207 first-year students for 18 months, they found that an action crisis about a goal predicted later disengagement from it, predicted how soon that came, and for the degree itself predicted the credits the student ended up with. Those effects held after controls for how desirable and how attainable the goal was rated. In an earlier study, marathon runners in an action crisis two weeks out secreted more cortisol during the race and ran it slower.

Not delivering is more common than the guilt suggests. Ethan Mollick surveyed 47,188 backers across 30,323 funded Kickstarter projects for a report Kickstarter co-published in 2015, and around 9 percent of those projects never delivered what they promised, every one of them publicly announced and paid for in advance. Mollick counted rewards rather than finished work, so some of those creators did finish and still left backers empty-handed.

Watch what the research actually licenses, though. Carsten Wrosch and colleagues found that people who let unreachable goals go and put the effort into new ones report higher well-being, across samples of 115 undergraduates, 120 adults and 45 parents, and all of it is correlational. The defensible claim is narrow: the torn feeling is a phase with a name, it predicts the outcome well enough to act on, and deciding it deliberately beats letting it drag.

What to do on the evening it stalls

Four moves, each attached to something above.

• Turn the remainder into countable units and count those down. Rows left on the sleeve, pages left in the chapter, invoices left in the pile. If you can't name the unit, that is the problem to fix first.
• If you can't say why the thing matters, read your finished work instead and then choose, either commit or close it. Don't leave it at 40 percent for another month.
• Write the next action with a time and a place before you shut the laptop. "Draft the opening paragraph after Tuesday's standup" is a plan, and "finish report" is the thing you've been not doing.
• Shorten the deadline rather than sharpening the estimate. If the shortened version needs somewhere to live, the weekly plan is where it survives contact with Wednesday.

The middle stays boring, and that is survivable

None of this makes the middle interesting. It stays boring, and the whole point of counting down is that boring is survivable for as long as the number keeps visibly moving. On the worst evenings the honest move is smaller still, which is the ten-minute version of the step.