Frankfurt Cases and the Principle of Alternative Possibilities

The single most influential move in the modern free-will debate is Harry Frankfurt's 1969 attack on the Principle of Alternative Possibilities (PAP) — the deeply intuitive idea that an agent is morally responsible for an action only if she could have done otherwise. With a now-famous thought experiment, Frankfurt argued that PAP is false: an agent can be fully responsible even when she had no alternative whatsoever. If he is right, the consequences are enormous, because the most powerful argument that determinism threatens moral responsibility — the Consequence Argument — works by showing determinism removes alternatives. Knock out PAP, and that threat is defused: responsibility could survive even in a world with a single possible future.

This page states PAP and its motivation, presents the Frankfurt cases, and works through the major replies — the flicker of freedom, the Kane–Widerker–Ginet dilemma, and the buffered/blockage refinements — before drawing the upshot: the great shift from leeway to sourcehood conceptions of freedom and responsibility.

The Principle of Alternative Possibilities

PAP is the formal regimentation of one of the two ingredients of free will — the leeway or "could have done otherwise" condition:

(PAP) A person is morally responsible for what she has done only if she could have done otherwise.

Its appeal is immediate. We excuse people precisely by showing they had no choice: the cashier who hands over the money at gunpoint, the driver whose brakes failed, the addict in the grip of overwhelming compulsion. In each, "there was nothing else he could have done" functions as a release from blame. PAP simply generalizes this: no alternatives, no responsibility. It is also the bridge that makes the Consequence Argument bite — if determinism leaves exactly one possible future, and PAP is true, then determinism abolishes responsibility.

Frankfurt's target is exactly this principle. His strategy is to construct a case in which an agent manifestly cannot do otherwise yet just as manifestly is responsible, so that the two conditions PAP welds together come apart.

The Frankfurt case

The canonical version involves a would-be manipulator with a fail-safe device:

Jones and Black. Jones is about to decide whether to do $A$ (say, vote a certain way, or kill someone). Black, a neurosurgeon, wants Jones to do $A$ and has secretly implanted a device in Jones's brain that lets Black monitor Jones's neural states. If Black sees Jones beginning to decide not to do $A$, the device will activate and cause Jones to decide on $A$ anyway. But if Jones shows signs of deciding on $A$ on his own, Black does nothing — the device simply watches.

As it happens, Jones decides on $A$ entirely on his own, for his own reasons. Black never intervenes; the device never fires.

Now the two judgments Frankfurt wants:

1. Jones could not have done otherwise. Had he started to choose differently, Black's device would have stepped in and produced the choice for $A$. The alternative was sealed off; there was no genuinely available path to not-$A$.
2. Jones is nonetheless morally responsible. The counterfactual intervener played no role at all in what actually happened. Jones deliberated, chose, and acted exactly as he would have in a world with no Black at all. The mere presence of an unused fail-safe cannot make a difference to his responsibility.

If both judgments hold, PAP is false: Jones is responsible and could not have done otherwise. What matters to responsibility, Frankfurt concludes, is not the availability of alternatives but the actual sequence — how the action actually came about, whether it flowed from the agent's own deliberation rather than from the device. Responsibility tracks the source, not the leeway.

Why this matters

Frankfurt cases are a hinge for the entire field:

• They decouple responsibility from determinism-via-leeway. If responsibility never required alternatives, then determinism's removal of alternatives is beside the point, and the Consequence Argument — however technically sound about leeway — loses its grip on responsibility. This is the foundation of Fischer's semicompatibilism: grant that we cannot do otherwise, deny that this is what responsibility needs.
• They redirect the theory of freedom from leeway (regulative control, the garden of forking paths) to sourcehood (guidance control, actual-sequence). The question becomes not "were other options open?" but "did the action issue from the agent in the right way?"
• They reshape incompatibilism. Source incompatibilists accept Frankfurt's lesson — responsibility is about the actual source — but argue determinism still poisons the source (the manipulation arguments). So Frankfurt cases push serious incompatibilists from the leeway road onto the source road.

Replies and the ensuing debate

Frankfurt cases are among the most-discussed thought experiments in philosophy, and the defense of PAP has been ingenious.

1. The flicker of freedom

The first response (Naylor, van Inwagen, Fischer's early work): Jones is not literally without alternatives. Something must signal to Black whether to intervene — the "prior sign" by which the device detects which way Jones is tending. So Jones retains some alternative: he could have shown the other sign, begun to incline the other way, even if the device would then have overridden him. There is a residual "flicker of freedom" — a sliver of leeway — and PAP, defenders claim, requires only some alternative, however small.

Frankfurtian rejoinder: the flicker is too "thin and flickering" to ground responsibility. The available alternative — involuntarily beginning to show a different neural sign, only to be immediately overridden — is not a robust alternative: it is not an alternative in which the agent does otherwise voluntarily, or avoids the outcome, or could be the basis of a different appraisal of her. Robustness objection: an alternative grounds responsibility only if exercising it would have made the agent not responsible for the actual outcome (it must be a real exercise of agency that changes the moral situation). A mere flicker fails this test, so it cannot be what PAP is really about.

2. The Kane–Widerker–Ginet dilemma (the indeterminism objection)

The most serious objection (Robert Kane, David Widerker, Carl Ginet) targets the prior sign and presses a dilemma. Black's device must read some prior sign to know whether to intervene. Ask: does that sign deterministically guarantee how Jones will choose?

• Horn 1 — the sign is deterministically sufficient for Jones's choice. Then the case simply assumes causal determinism in the actual sequence. But against a libertarian (who holds responsibility requires indeterminism), that begs the question: the libertarian already denies that a determined Jones is responsible, so a case stipulating determinism cannot show him responsible. The example presupposes exactly what is at issue.
• Horn 2 — the sign is not deterministically sufficient. Then, even given the sign, Jones might still have chosen otherwise — the indeterminism leaves a genuine gap — so Black cannot intervene in time with certainty, and a real alternative remains open before the device could act. PAP is not violated after all.

Either way, the objection runs, a Frankfurt case cannot both be dialectically fair against the incompatibilist and genuinely eliminate alternatives. The argument is sharpest precisely where it matters most — for the libertarian who thinks free action is undetermined.

3. Buffered and blockage cases (the Frankfurt-defender's reply)

Frankfurt's defenders responded with redesigned cases meant to escape the dilemma by removing the robust prior sign while still being compatible with indeterminism.

• Buffer cases (David Hunt; Derk Pereboom's "Tax Evasion"). Pereboom builds a case in which the only way Jones could even begin to choose otherwise is for a specific mental event — a sufficiently attentive moral consideration — to occur, and the device is triggered by the absence of that "buffer" event. The agent's available "alternative" is reduced to the involuntary occurrence of a single necessary precursor, which is not robust (it is not itself a voluntary doing of otherwise, and Jones is not blameworthy for it). Pereboom argues such cases preserve indeterminism and leave no robust alternative — so responsibility without robust alternatives stands.
• Blockage cases (Hunt, following a suggestion of Frankfurt's). Imagine that all neural pathways to any alternative decision are physically blocked (without any need for a triggering sign), so the actual-sequence deliberation is left untouched but every other route is closed. If Jones is still responsible, alternatives are irrelevant — though critics worry blockage either reintroduces a flicker or covertly determines the outcome.

The dialectic continues: Widerker, Ginet, and others contest whether the buffer is truly non-robust or whether some responsibility-grounding alternative always survives; defenders refine the cases further. There is no consensus that Frankfurt cases decisively refute PAP — but they have convinced many that, if PAP survives, it does so only in a weakened, flicker-dependent form too thin to carry the Consequence Argument's weight.

4. Revising PAP rather than abandoning it

Some defenders retreat to a weakened principle rather than give up leeway entirely — e.g. responsibility requires only that the agent could have avoided being the unimpeded source of the act, or could have made some relevant difference. Others (the leeway loyalists) hold a robust PAP and reject Frankfurt's responsibility intuition outright, insisting that if Jones truly could not have done otherwise, our confident attribution of responsibility is itself the thing to question.

The upshot: from leeway to sourcehood

Whatever one's verdict on whether PAP is strictly refuted, Frankfurt cases produced a lasting reorientation. They drove a wedge between two things the tradition had run together — alternative possibilities and moral responsibility — and shifted the center of gravity from:

• leeway / regulative control (the metaphysics of open alternatives, the "garden of forking paths") to
• sourcehood / guidance control (the actual-sequence question of whether the action expressed the agent's own reasons-responsive agency).

This realignment is why so much of the contemporary debate is about the actual sequence: Fischer–Ravizza's guidance control, the deep-self and reasons-responsiveness theories, and the source incompatibilist reply that even a flawless actual-sequence source is poisoned if it was itself determined or manipulated into existence. In the architecture of the field, Frankfurt cases are the pivot on which leeway incompatibilism (which keeps PAP and fights the Consequence Argument on its own ground) gives way to source incompatibilism and to semicompatibilism (which drop PAP and relocate the whole dispute to the nature of the source).