The Hawthorne effect, and the data that turned up eighty years late. · published 21 September 2026
The Hawthorne illumination study — the origin of “people work harder when watched” — was described for eighty years by people who had never seen its data; when Levitt and List found the original records in 2011 they concluded the famous rising-after-every-change pattern was entirely fictional, with output tracking the day of the week and the pay cycle instead, leaving only a much smaller reactivity effect and a much bigger lesson about how a tidy conclusion gets named, taught and never rechecked.
These are the working notes, unedited. Everything the video
states as fact comes from here, and nothing else may be stated as fact.
The sections on limits and on what the finding does not show are
part of the brief, not an afterthought — see
the method.
The studies: at the Hawthorne Works, a Western Electric plant in Cicero, Illinois, a series of experiments ran from 1924. The first set varied the electric lighting on the factory floor. A later one, the Relay Assembly Test Room, began in 1927 and followed a small group of women assembling telephone relays while rest breaks, hours and refreshments were changed around them. Output was counted mechanically, by the finished relays each worker dropped down a chute.
The claim that travelled. That productivity rose after every change — brighter light, dimmer light, more breaks, fewer breaks, and again when the original conditions came back. From which: it was not the lighting or the breaks, it was being watched. The effect carries the plant’s name and is taught as the reason a control group is needed at all.
Two things about the name. Nobody at Hawthorne called it that. “Hawthorne effect” was coined in 1953 by John R. P. French, and given wide currency by Henry Landsberger’s 1958 re-analysis — decades after the work ended. The name came from the interpretation, not from the experimenters.
The reversal, in 2011. Steven Levitt and John List found the original illumination data, which had been thought lost, and analysed it properly for the first time. Their conclusion on the famous pattern is blunt: existing descriptions of the supposedly remarkable data patterns prove to be entirely fictional. Output tracked the day of the week and the pay-period cycle far more closely than it tracked any change in the lighting. The paper is “Was There Really a Hawthorne Effect at the Hawthorne Plant?”, American Economic Journal: Applied Economics.
And the relay room. Stephen Jones had already gone back to those records in 1992. Once a new incentive pay scheme and an ordinary learning curve were accounted for, there was very little variation left to attribute to being observed. The women were being paid differently and getting better at the job.
Limits that must be stated, in both directions. Levitt and List did not find nothing: they report weaker, more circumscribed evidence consistent with some reactivity — output somewhat higher while a manipulation was actively running than in the gaps between manipulations. So something is there; it is just far smaller and far less dramatic than the story. Reported details differ between secondary accounts even on the direction of the weekly pattern, which is a reason to state the shape of the finding rather than the arithmetic.
What this does not show
Not that people behave identically whether or not they are observed — reactivity and demand characteristics are real, are well documented elsewhere, and are why blinding exists. Not that the Hawthorne researchers faked anything. What collapsed is narrower and more embarrassing: the single most cited demonstration of the effect does not demonstrate it. A conclusion was written up, repeated, simplified, named, and taught for eighty years while the numbers underneath it sat unexamined.