What the research on body doubling supports, and where it stops

People were body doubling¹1 Body doubling means working in the presence of another person, in the same room, over a call, or even alongside a recording; the other person does not have to be doing the same task, and does not have to know they are doing anything at all (pp. 1 to 2). ↩ long before it had a name: of the 220 adults who answered a 2023 survey about it, 53 learned the word from the survey itself and 52 said they had always done it (p. 2). So does it help you work? The closest measured answer in this post comes from 241 experiments on working near other people, where the effects are small and run both ways (p. 265; Table 1, p. 273). The survey that named the practice asked what people do: it set no task, ran no timer and made no comparison (pp. 1 to 2).

What body doubling means, and who was asked

The term was coined in 1996 by Linda Anderson, in a popular article rather than a study, and spread on TikTok (p. 1). Researchers at UC Santa Cruz put an open survey online; of 410 responses, the 220 complete ones from adults were hand-coded, 193 (87.7 percent) neurodivergent (p. 2).

Asked what the words mean, they described someone in the room (n = 127) or on a call or chat (n = 27), to get a task done (n = 65), the other person usually on something different (n = 65) and rarely a similar one (n = 13) (p. 2). Those are counts of volunteers, not rates.

What another person in the room does to your work

Two psychologists pooled 241 experiments, 23,970 subjects, in which people did a task alone or in company²2 A meta-analysis pools many separate studies into one estimate, so the answer rests on a whole literature rather than on whichever study you happened to read. ↩ (pp. 269 to 271). People were allocated to those conditions, so this is the evidence here that can speak about cause, and what it says depends on the task.

On simple repetitive work people were faster, d = .32, the paper's most robust effect (Table 1, p. 273). On complex accuracy-sensitive work they were less accurate, the largest effect in the review: the average person working with others present was more accurate than 39 percent of the people working alone (p. 272; Table 1, p. 273).

And it is small: the presence of others accounted for .3 to 3 percent of the variance in the typical experiment (p. 265). Everything else is the person, the task and the day. The tables count effects, 742 of them, not studies (p. 269).

Being watched is not the ingredient

Whether the other person could evaluate the work did not reliably magnify the effect on performance, and reduced it nearly as often (p. 278). Disregarding significance, familiar others produced smaller effects than unfamiliar ones in all six comparisons in the paper (p. 280), so a friend is not the stronger version of this. Both are moderator comparisons inside that corpus, not one assigned contrast.

The randomised number is about tracking, not company

In 138 studies, adults were randomly allocated to something designed to get them monitoring their own goal progress, or to a control condition, 18,398 people across both arms, and monitoring raised goal attainment by d+ = 0.40, a small to medium amount (pp. 198, 208, 213). It cannot travel alone: 0.33 among the 70 adequately powered studies and 0.19 after adjusting for missing small ones, so the range is about 0.2 to 0.4 (p. 213).

Studies where progress was reported to at least one other person came out at 0.47 across 95 studies; where it stayed with the person alone, 0.19 across 14 (Table 3, p. 210; p. 214). Nobody was allocated to share, so that contrast compares groups of studies, uncorrected for multiple tests and confounded with other moderators (p. 218), and the authors ask for the experiment nobody has run (p. 217).

One of those 138 studies monitored anything outside health, a 61-person time-management study (pp. 211, 218), so reading it as evidence about focus sessions is our extrapolation.

Finishing and coming back are outside these three papers

What a focus app most wants to claim is the thing none of these three looked at. The people surveyed say they are more likely to finish a task alongside someone else, in the paper's words, "[o]ver half (n = 186)" of the 220 (p. 2), which is self-reported perceived likelihood, not measured completion. The experiments timed speed, accuracy and sweat over minutes, and nothing about starting, continuing or coming back.

They reach for it on the work they dread, one of them writing that it is for tasks "particularly unpleasant, difficult, or emotionally stressful" (p. 3), which is the kind of work where the experiments expect company to cost accuracy. Nobody has reconciled that, and this post does not either.

What it suggests trying

Two moves, neither a promise. Match the company to the work: in these experiments the gain showed up on simple repetitive work and the cost on hard accurate work (Table 1, p. 273), so admin and filing are the sessions to put beside someone, and the thing you are still working out is worth one quiet run alone. Those tasks were things like winding a reel and memorising syllables, minutes long and mostly undergraduates, so carrying them to your afternoon is ours too. The other move is the record rather than the room: write the progress down, and tell one person what it says. The randomised part of Harkin is being prompted to monitor at all; writing it down and telling someone are both contrasts between groups of studies, uncorrected for multiple tests and, by the authors' own account, confounded with each other (pp. 210, 214, 218).

Where Yappie fits

We make Yappie, a focus app with a pet beside the timer, so read this as an interested party. We build sessions you sit down to alongside someone else: a friend, a study group, or a stranger the app matches you with. The evidence does not reach that far. Everyone in these papers was in the same room and ours is over a screen, and whether sitting down beside someone brings you back tomorrow is the thing none of them measured. We cannot tell you that either, and this is an honest answer.

References

1. Eagle, T., Baltaxe-Admony, L. B., & Ringland, K. E. (2023). Proposing body doubling as a continuum of space/time and mutuality: An investigation with neurodivergent participants. Proceedings of the 25th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '23), 1 to 4. https://doi.org/10.1145/3597638.3614486↩
2. Bond, C. F., Jr., & Titus, L. J. (1983). Social facilitation: A meta-analysis of 241 studies. Psychological Bulletin, 94(2), 265 to 292. https://doi.org/10.1037/0033-2909.94.2.265↩
3. Harkin, B., Webb, T. L., Chang, B. P. I., Benn, Y., Prestwich, A., Conner, M., Kellar, I., & Sheeran, P. (2016). Does monitoring goal progress promote goal attainment? A meta-analysis of the experimental evidence. Psychological Bulletin, 142(2), 198 to 229. https://doi.org/10.1037/bul0000025↩