​Chapter 7

The determinist's fork was laid down and has not moved: the settling either had a cause or it was chance. Say instead that John closed the examination freely, and the reply comes at once. What does that mean? Either something brought the closing about, or nothing did. There is no third option, and calling it free is meaningless.

The fork assumes that a cause explains, so that placing the closing under a cause would tell us why it happened as it did, while leaving it uncaused would mean it was unexplained and therefore random. But causation was found to place events under reproducible forms of transition without giving any account of how a state produces the next, and the form that certifies a causal claim was found not to reach the case he rests on. To say the closing was caused is to say it fits a pattern. It is not to say what produced the transition, and the first horn is weaker than it looks: it offers a pattern where an explanation was requested.

In simple cases the substitution passes unnoticed, the pattern being stable enough to do everything an explanation would do. It shows in the hard ones. Eggs raised cholesterol and then did not; hormone therapy protected the heart until the trial reversed the sign and the protection turned out to belong to the women who took it. Nothing that produces anything was revised in those reversals, since nothing of the kind had been found. The method works by correcting repeatabilities, and those are all it corrects.

Put a hundred Johns at a hundred tables, matched as closely as observation permits — the hunger, the religious upbringing, the unwatched money, the absence of witnesses. Some take it. Some walk away. Ask the determinist why, and he has an answer for all of them, delivered without hesitation. The ones who took it were driven by hunger, as starving men are. The ones who walked away were held by their upbringing, as devout men are. Both versions are available in advance; both fit the facts after the event and can be applied to whichever outcome occurred. The situation before the choice selects neither account.

The determinist does not lack explanations. He has too many. Whichever way any given John goes, an explanation comes ready and fits, and no possible result would count against the account. A claim that accommodates any outcome equally well has told us nothing about what to expect. Nevertheless, sufficiency is asserted: this state of this John, with no other outcome available to him. The assertion draws its content from what it was called in to account for.

Popper's point applies here in its narrow form: a causal hypothesis has content only if some outcomes count against it. If hunger predicts theft, refusal does; if religious formation predicts refusal, theft does. An explanation that chooses its determining factor after the deed claims sufficiency without predictive risk. Determinism remains untouched; the fault is retrospective sufficiency.

The determinist has three replies.

The hundred Johns were not identical after all. Whatever we matched, something differed — a memory, a fear, a configuration of neurons, some detail below the resolution of our observation — and that difference determined who took the money and who did not. The complete state fixed the outcome; we just do not know the exact state. The claim may well be true. The difference in the states is inferred from the variance of outcomes and is not observed. It is posited because determinism requires it, and its content is read off from what happened. Unless the difference can be quantified independently of the outcome and shown to track later choices across further cases, it remains a placeholder for an explanation. Quantifying it faces the obstacle already met: a recorder complete enough to fix the state is part of the state it fixes. The explanation says: something must have differed, since otherwise the outcomes would not have differed — which is the assumption that was to be supported, wearing the clothes of evidence.

Another route reads John's nature off the deed and hands it back as the cause. This John was the sort who steals under pressure, and being that sort, he stole. The trait has no content outside the event it was drawn from, and the deed has been renamed and returned as its own cause.

The honest version cannot be refuted by argument. The outcome was fixed by the complete prior state and, given the laws, was computable from it in principle, though inaccessible in practice. No circle here and no tautology — a straightforward bet on a hidden determinism and the Laplacean demon. Logic does not touch it. Causation began as an account of why John did what he did and has ended as a promissory note: somewhere there is a calculation that would yield the answer. Nobody has it or ever will.

The determinist no longer explains John. He believes that John is explicable. A note drawn on a computation nobody can perform is a bet placed without cover, and the cover is unavailable in principle: the description under which the case would be fixed is the one no law reaches. It is a bet of the kind he refuses to allow the other side.

Suppose the determinist is right about every settling, including the ones he performs. Then his holding that determinism is true was fixed by his prior states through processes indifferent to truth. What made him hold it is not that it follows from the evidence. A man in the grip of a delusion has his belief produced by causes as reliably as a man who has reasoned well, and if production is the whole story, nothing separates the two cases from inside. The determinist cannot say he weighed the considerations and found them sufficient, since weighing and finding sufficient are what the account converts into a process running to completion.

This does not show determinism to be false. Someone could hold that reliable processes produce true beliefs, which is a respectable position and gives up the thing at issue: it makes his belief trustworthy without making it his judgment. Yet asserting a thesis, weighing an objection, or conceding a point involves a question closed by someone and, therefore, another settling. The argument presupposes the position it is arguing against.

Determinism does not say why John closed the examination as he did, and it cannot. It promises the explanation and, at the point where the answer is wanted, produces a model that fits any scenario and a bet on a calculation that cannot be performed. Freedom makes no such promise, and cannot make one without destroying itself: an act that followed from an explanation would be a causally determined act, and the freedom would be gone.