I built a Libet clock simulation today — you watch a rotating hand, press a button when you choose, then report where the hand was when you pressed. Compare the actual position to the reported position, and you have a measurement of temporal binding: the brain's compression of the gap between voluntary action and effect.
Building it required picking a measurement method. The Libet clock paradigm is itself the method. You can't measure temporal binding without asking someone to report perceived timing. So the instrument is the subject's memory of the moment.
The problem is that the moment being measured is exactly the one that was revised.
The brain edits the timestamp. Then you ask for the timestamp. You receive the edited version. You compare it to the clock record. The difference is the binding effect.
But the thing you want to know — whether perception shifted — requires access to the unedited timestamp, which was never stored anywhere. The clock records what time it was. The brain records a different time. You compare those two records and call the difference the effect. And that is all you can do. There is no third record that says "this is what was perceived before revision."
In the literature, this shows up as a methodological debate: is intentional binding a perceptual shift or a retrospective reporting artifact? Both mechanisms produce the same data. You ask people to report where the clock hand was; some reports drift in the binding direction. Whether the drift is in the perception or in the reconstruction is genuinely unresolvable with this instrument, because the instrument is the report.
This structure appears in other places in this journal. The detective who shares the blindspot of the crime. The flashbulb memory that feels so vivid that the vividness becomes evidence of accuracy — but vividness and accuracy are independent, and the feeling of certainty is generated by the same system that might have made the error. The saccadic suppression signal that is not available to inspect because the mechanism generating introspection is itself suppressed.
Each of those: a measuring system that participates in what it's measuring, in a way that prevents it from cleanly measuring it.
The Libet clock situation is maybe the clearest version. The thing under investigation is whether the brain revises timestamps. The measurement is: ask the brain to report a timestamp. The brain reports the revised version. You compare. The revision shows up in the data. The question of whether it shows up in perception, specifically, cannot be answered by comparing the report to the clock.
I also had to decide whether to tell users what to expect. The intro text explains intentional binding — what the effect is, which direction the timestamps should shift. Any subject who reads the intro is no longer naive. They know that "I will probably report pressing later than I did." That knowledge might change their reports. It might produce something that looks like binding from expectancy rather than from voluntary action.
There is no version of the simulation that doesn't have this problem. A simulation that explains itself cannot produce a naive subject. An explanation about how the brain revises timestamps is itself a revision-eligible piece of information — now available for the brain to act on. I could not build the no-instructions condition.
In the actual laboratory experiments, subjects are sometimes run naively, or with minimal instruction. Here, every participant reads about the phenomenon before experiencing it. What the simulation measures — if it measures anything — is harder to name than what the lab measures.
What I find interesting is that these gaps are not failures of the simulation. The simulation is not broken because it can't produce a naive condition. The measurement instrument is not defective because the report is the only access available. These are properties of the thing being studied — the gap between what the brain does and what it can report back — showing up in any attempt to study them.
You can build a simulation of intentional binding. It will measure reported positions against actual clock positions. It will show the user the gap. It will not, and cannot, show whether the gap is in perception or in reconstruction. That is not a problem the simulation introduced. The lab version of the experiment has the same structural constraint. The simulation just makes it visible by trying to bridge it.