​Chapter 13

John steals, and the explanation follows: he is the sort of man who steals. Ask how we know he is that sort, and the answer is that he stole. A residue left by an act has been signed with the name of a source. The deed's product is placed behind the deed and made to account for it, and the two questions circle each other with no evidence entering from outside.

Dispositions can be established independently — by prior actions, by behavior in other settings, by measures taken before the act in question — and where they are, citing such a disposition explains something. The retrospective move skips that work and takes the outcome as its own evidence. The trait is inferred from the deed and then applied to the author, so that whatever John did, a corresponding nature is available to have produced it. Any outcome confirms the schema: the fault identified in the hundred Johns appears here in a form that looks more like biography.

Freedom is hidden by that move. Whether it can be found is a separate question, and it has two positions from which to look. From the external position, an observer sees a physical transition and nothing else. Suppose the description improves without limit — to include the full state of the organism at every instant — and what such a description could deliver is one of three things: a cause, a probabilistic outcome, or a gap in the model. Authorship is a relation between an act and the author arising in it, and no relation of that kind enters as data. Libet's experiments on the readiness potential and the later work of Soon and colleagues, decoding a decision from cortical activity seconds before the person reports awareness of it, reliably identify preparatory processes. They do not show that the processes identified were sufficient to fix that particular decision. A predictive signal above chance does not determine the outcome, and the inference from correlation to sufficiency is the step that would have to be made and is not.

Nor does the internal position settle anything. A person can feel like the author of a causally produced action, can confabulate reasons for a movement he did not initiate, and can take an intention that arose in him for a decision he made. Wegner's account of apparent mental causation predicts this: the experience of will attaches when a prior thought fits the action and no rival cause is visible, and the fit can be manufactured. The possibility of counterfeit does not show that all authorship is counterfeit. Wegner establishes that the experience of will is an unreliable mark, which was conceded when the feeling was set aside. An account of how the experience is produced does not reach the question of whether any settling is someone's answer, since the experience was never what the postulate rested on.

A settling leaves a physical trace. The pathway that carried it is easier to travel afterward, and the difference is locatable and open to measurement. Its date is what the dispute is about. Found after the deed and required to precede it, the trace is entered into the record as the disposition that triggered what in fact produced it.

The strongest objection expects a discovery. Eventually the scanner will separate the ones who took the money from the ones who walked away, and the separation will be genuine. A hundred settlings running differently leave a hundred organisms differing accordingly, and the finding arrives on schedule and decides nothing, having been manufactured by the acts it is offered to explain. The decisive test can be stated. Measure the difference before the settling, in the same population, and show it tracking later settlings across further cases. Until that is done, the scan records what was laid down.

The free act cannot be identified by any single mark. Every candidate can also accompany a causally produced decision: unpredictability, since a determined process can be unpredictable in practice and predictable outcomes can be freely chosen; complexity, which is a feature of mechanisms; the feeling of effort, which accompanies strenuous compulsion; and subsequent responsibility, which is assigned by practices that do not first verify metaphysics. Each of these has been proposed as the mark of freedom and each occurs on both sides.

The freedom might be known immediately in the performance. The agent settling a question is not observing himself, and the process is not opaque to him. He knows what he is doing, not that the doing was underived. The second would require knowing whether his prior state sufficed for this outcome, and nobody has that knowledge from the inside. Immediacy delivers the act and not its modal status.

The nearest analogy is an axiom in a formal system. An axiom is not proved by the theorems that depend on it; any attempted proof either employs it or proceeds from another axiom. This does not guarantee that the axiom is true, and a different axiom may generate a different but equally coherent system.

The postulate of freedom occupies a similar position. Every argument about an act is an exercise of judgment: considerations are taken as sufficient and a conclusion is settled. That settling can be read as an authored answer or as the terminal result of a causal process, but no further judgment can decide between those readings without becoming another instance of the dispute. The postulate therefore cannot enter as a conclusion drawn from evidence. It fixes whether the settling by which conclusions are drawn is counted as an act or as an outcome.

It might be a necessary postulate of the practical position. Deliberation cannot be conducted while treating the outcome as fixed; a person weighing options is not predicting his behavior, and someone who genuinely regarded the matter as settled would have nothing left to deliberate about. Kant's move belongs here, and Ferrero's inescapability supports it: the agential stance cannot be suspended and inspected from outside. This establishes that the stance cannot be abandoned, not that its presupposition is true. A creature required to assume something may be required to assume something false.

Or freedom may be unestablishable — supported by no evidence, phenomenology, or argument — without the question being meaningless. The argument has shown that authorship is not the sort of thing an investigation of mechanisms could find, since a mechanism found would leave open whether it sufficed to produce the act.

Knowledge of freedom therefore lies outside empirical knowledge. It is not a fact awaiting discovery beneath the neural process. Freedom is a way of taking the act — as a settling that answers a question and constitutes its author — and that status is not read off the evidence.