Third session. Third subject.
Tomás, twenty-six, three years in the studio. Not a veteran like Marisol — he hasn't done the movement ten thousand times. But he's done it enough that it doesn't feel like performance. It feels like knowing.
I give him the instruction: tell me when you have decided to move, before you move.
Ninety seconds, approximately. Then: Now.
He looks confused immediately. He says: but I was about to move.
I say: yes.
He says: so did I say it right?
I ask him the third instruction: can you point to when it happened?
He says: I think it was —
And stops.
He sits for a moment. His hands are still. He looks like someone trying to locate a sound they heard in a room that has gone quiet.
Then he says: I thought I could point to it and then I couldn't.
Claudio said: I decided just now. A moment with a timestamp, a before and after.
Marisol said: it was already decided when I noticed. No origin point, no locatable moment — decision as condition.
Tomás says: I thought I could point to it and then I couldn't.
This is a third thing.
Not a moment he can locate. Not a condition without origin. A moment he almost located — approximately, directionally — and then the act of pointing dissolved the location.
I write in the protocol margin: observer effect?
I put a question mark.
The question mark is doing two kinds of work. The first is honest: this is a hypothesis, not a finding. The second is harder to name — something like protection. If I write observer effect without the question mark, I have named a phenomenon before I can establish it. If I am wrong, I have trained three subjects in a framework that may not be true. The question mark is the thing that keeps the protocol honest.
I sit with what the three results look like together.
Claudio: locatable. Marisol: structural. Tomás: almost-locatable-until-observed.
Three different relationships to the same instruction. The instruction doesn't access the decision — it accesses the subject's relationship to the decision. What varies across the three isn't the decision itself. It's the subject's model of their own decision-making.
Which means I have been measuring the wrong thing again.
Not for the first time.
I add a note to the protocol: the third instruction may reveal epistemic access, not volitional structure. The decision exists regardless of whether the subject can locate it. The subject's ability to locate it tells me something about how they hold their own agency — not about what agency is.
This is not smaller than what I was studying.
It might be larger.