Do we choose what we do?

A slow rise in electrical activity over motor areas, the readiness potential, precedes voluntary movements. In 1983 Benjamin Libet and colleagues timed it against people’s own reports of when they first intended to move, and found the brain signal earlier than the reported intention by a few hundred milliseconds.

In 2008 a different group went further, using brain imaging and asking which of two buttons a person would press. Patterns in prefrontal and parietal cortex carried information about the upcoming choice, the authors report, up to ten seconds before it entered awareness. The prediction was only slightly better than chance: a tendency across many trials rather than a reading of any one of them.

The tasks are decisions with nothing at stake and no reason to prefer one moment, or one button, over another. Whether the results carry over to decisions with content is not something these experiments test. Later studies that compared the two disagree: one in 2019 could not detect the readiness potential before choices about a real donation to a charity, and one in 2025 found it stronger before choices that earned the participants money.

The timing of a reported urge is also a report, produced by the same brain, about an event with no sharp edges. It is evidence, and it is not a stopwatch.

The standard reading is that it is the brain preparing the movement, which would mean the preparation starts before the awareness. A 2012 model proposes something else: that spontaneous neural activity drifts up and down all the time, that a movement happens when the drift crosses a threshold, and that averaging many trials backwards from the movement necessarily produces a rising curve whether or not anything was being prepared.

On that account the readiness potential is partly an artefact of how the data are aligned, and the decision is the threshold crossing rather than the slope leading to it.

Both readings fit the averaged electrical data. Experiments that interrupt people mid-task, or track how the signal varies from one trial to the next, have been used to test them, with results claimed by each side. A 2021 review led by the model’s first author calls the question open and argues that the model is the simpler account.

People will explain choices they did not make. In a 2005 study where the photograph a participant had chosen was swapped for the one they rejected, most of the swaps went unnoticed, and people gave reasons for preferring the face they had just turned down.

That does not show that choosing is an illusion. It suggests that the explanation of a choice can be produced separately from the choice, so an introspective report is not a direct readout of the process, in this experiment or in Libet’s.

A finding that brain activity precedes awareness is most troubling if you assume the deciding self is something other than the brain doing it. On the view that a person is what their brain is doing, an earlier neural cause of a choice is not a rival to the choice. What survives on that view is a narrower question, whether the conscious part does any of the work, and it is the one Libet himself raised.

So part of this is a question about what would count as choosing, and no experiment settles that. What the experiments do show is that our reports of when we decided shift with the clock and the wording used to collect them, and a 2023 review of the method concludes that they are not accurate. That was not obvious.