Ovsiankina effect (resumption tendency)

Evidence

Between 1924 and 1926, Ovsiankina (1928) had 124 people — 108 adults, mostly students, and 16 children aged 3 to 16 — carry out small tasks and interrupted them while they worked: taking apart puzzle rings, solving riddles, laying a mosaic, threading beads, untangling a ball of yarn. Her summary gives two figures. Where the interruption was experienced as the experimenter getting in the way, participants resumed the work in 74 percent of cases — 83 percent if one counts the cases where only the urge was noticeable. Where it looked like an accident instead, such as a box falling to the floor or a colleague asking for help, they resumed without exception.

The meta-analysis by Ghibellini and Meier (2025) confirms the tendency in general, in contrast to the memory advantage of the Zeigarnik effect: 67 percent resumption across 21 reports, well above chance. The figure it attributes to Ovsiankina’s own work — 91 percent — is not one she gives in her summary. The number appears in her paper at one place only, as a sub-finding: participants who returned to the unfinished work without having caught sight of it again did so in 91 percent of cases; those who had seen it, in 63 percent. Ovsiankina reads the paradox as meaning that a strong need makes people go looking for the work themselves.

Her most telling control concerns not the size of the rate but its cause. She placed other people’s unfinished work on the table — a chequerboard pattern hatched three quarters of the way, a postcard half reassembled. Of 11 people a single one completed anything, and only after 40 minutes. A second version used work of a kind the person had done before, just not that particular copy: 1 of 11, and even there it stopped at the mere tendency. Her conclusion: the tension only arises once the work has become “one’s own work” through doing it. The sight of something unfinished is not enough by itself.

One finding from it is awkward for product design: in McGraw and Fiala (1982) an announced payment lowered the resumption rate. The authors read this as an external incentive replacing one’s own goal of finishing the thing. That individual study is reported here at second hand.

Limits

What was measured is return within 20 seconds. After the interrupting task was finished, Ovsiankina gave participants no more than 15 to 20 seconds in most trials; whoever had not returned by then counted as a non-resumption and the session moved on. 24 percent reached for the work immediately, 86 percent within 20 seconds. The rate therefore describes a reflex in the laboratory, not a return hours later. She mentions individual cases after an hour, ten hours and three days, without counting them.

Without a surveyable goal the tendency collapses. In one of her own arrangements Ovsiankina varied how clearly the end of a task was fixed. Where material and instruction bounded it — a short thread and thirty beads — resumption ran at 70 percent. Where the goal was set by the instruction alone, 44 percent; where it was missing entirely, 46 percent. The arrangement covered only 15 people, and she herself calls its value “roughly orientating”. As far as it carries, it speaks for a display that names a clear end.

The paper contains no statistics. There are no significance tests, no measures of spread and no control group in today’s sense, but percentages from subgroups that sometimes number only six to fifteen cases. How far someone goes into the task also matters: among participants showing “devotion to the thing” the rate was 87 percent, among those with no inner relation to the task 51 percent.

The finding comes from laboratory settings using puzzles, drawing tasks and bead strings. Whether incomplete checkouts or profiles have the same effect is not directly supported, hence the low transfer. The closest case is a 2020 online study with 875 participants in which an interrupted browser game was resumed 73.9 percent of the time; it too is reported here at second hand. Do not exploit it with fake progress.