Source
- Velleman (1992), LIT-295, pp. 110–122 of the 2015 reprint, as read in NOTE-265. This is the primary source for the dilemma.
- Buss & Westlund, SEP Personal Autonomy (2018 revision), LIT-296, §3, as read in NOTE-267.
- McKenna & Coates, SEP Compatibilism, LIT-291, §4.2 and supplement §A, as read in NOTE-270.
- Dryden, IEP Autonomy (2010), LIT-297, §2.a.ii, as read in NOTE-266.
- Frankfurt (2004), LIT-293, the Tanner Lectures, as read in NOTE-264. Frankfurt (1971), LIT-692, is also read first-hand, in NOTE-545, and filed as THEORY-120.
What was actually shown
Frankfurt’s 1971 account, first held through LIT-291, is now read first-hand (LIT-692, NOTE-545; THEORY-120): one acts “of his own free will” when the will that moved him “was his will because it was the will he wanted” (p. 19). The will is the effective first-order desire, and the second-order volition endorses it. The objection is that nothing about being second-order makes the endorsing attitude the agent’s. She can be alienated from it just as from the first-order desire, so the question of authority recurs one level up.
Velleman (LIT-295) gives this its sharpest form, and the record holds it first-hand. He runs the objection against each stopping point in turn:
- Second-order desires fall to the regress. He reports this as Watson’s objection, which Frankfurt conceded (IWWCA 65–6). Both halves of that attribution are second-hand.
- Watson’s values fall to the materialism-and-sin case: one can recoil from values that still grip one.
- Frankfurt’s “decisive commitments” fall to the unwitting decision to end a friendship.
Velleman then generalises to a dilemma (pp. 115–116). A state specified without presupposing the agent’s participation is one “from which the agent can in principle dissociate himself”. A state specified so as to guarantee participation “beg[s] the question”. So it is “a choice between begging the question and not answering it at all”.
Buss & Westlund (LIT-296 §3) reach the same place independently, without citing Watson. The regress can stop only at an attitude “no agent can possibly be alienated” from. The one candidate, the desire to be self-governing, cannot be the criterion, because it is at work even when self-government fails.
Frankfurt’s late view (LIT-293) does not answer this. It never mentions the regress. Its stopping point is wholehearted love, whose necessities are “irresistible, but … not coercive” (p. 197). That holds only because lovers are stipulated never to be ambivalent, and any wavering shows the love “had never been truly wholehearted” (p. 198). That is the question-begging horn, reached by definition.
The claim is the record’s synthesis across these five readings. No single one of them states it in this form.
What this does not say
- It does not say higher-order structure is irrelevant to agency. Velleman calls Frankfurt’s work “the best sustained attempt” (p. 110), keeps his adjudication framing, and targets only the reductive strategy.
- The dilemma is not proved. Velleman argues its first horn only by examples, and gives no principle for why every independently specified state must be disownable (NOTE-265, C5). His own stopping point, the desire to act in accordance with reasons, escapes only “while remaining an agent”. Whether that escape is fair is unexamined, and Buss & Westlund’s argument would reject it as a criterion.
- Frankfurt’s replies are not assessed. Satisfaction (1992), active identification (1994) and Bratman’s policies are all held second-hand or not at all. LIT-295 expressly excludes “The Faintest Passion”.
- The middle case is argued against a footnote. Velleman’s decisive commitment is quoted from the 1976 “Identification and Externality” (IWWCA p. 68 n. 3), where decisions “do not seem to be susceptible both to internality and to externality”. “Identification and Wholeheartedness” (LIT-575), which Velleman lists and never quotes (NOTE-265), is reported by the IEP (LIT-297) to add that the commitment is made “without reservation”, with no reason felt to deliberate further, and that stopping there is “hardly arbitrary”. An unwitting decision plausibly fails those conditions. If it does, the counterexample does not reach the 1987/88 view, and the question becomes whether those conditions, which are conditions on the agent’s own stance, can be stated without presupposing her participation. That is still this account’s dilemma, but the record has not run it against the essay. Holton’s lecture handout on the essay (LIT-1856, read in NOTE-1455) reports at second hand that its stopping point is a decision, which Frankfurt does not analyse, in two forms: in checking a sum it carries the anticipation that accurate reviews will confirm it (pp. 168–169); between conflicting desires it makes the loser an “outlaw” with no such anticipation (p. 170). An unwitting decision plausibly fails the first and not obviously the second, so if the report is right the counterexample may still reach the conflict case. Only the essay can settle it.
- Who first stated the objection is unsettled in the record. The IEP and SEP Compatibilism credit Watson 1975. SEP Personal Autonomy treats Watson as a rival account and states the regress as its own. Watson’s valuing-versus-desiring distinction is reported in detail by none of them.
Connections
- The curation entry of 2026-09-29, “Frankfurt’s account of higher-order volitions, assembled from seven texts”, maps which claims are first-hand.
- LIT-292 (Velleman 2002) is the critique of the late view: an essence of volitional necessities would give authenticity, not autonomy.
- LIT-294 (Robb): the 1969 Frankfurt cases are a separate argument. They bear on this one only because identification is one of the two compatibilist accounts left after them.
- The history-based objection to the same accounts is THEORY-040, on manipulation.
THEORY-056 holds that self-control is the interaction of first-order motive states, so a hierarchical account of agency cannot take its structure from the psychology of control. It does not decide whose a motive is.
Three stopping points from the batch of 2026-10-03 are candidates for this account’s refutation condition. Each lands on the second horn, for a different reason.
- Bratman’s reflexive end (LIT-567; THEORY-076). Diachronic self-governance is backed by a “conditional, reflexive end” of one’s own diachronic self-governance, held in the agent’s standpoint. The end is specified by its object, self-governance, which is the agent’s governing participation under another name. So it is specified so as to include what it was meant to explain. Bratman does not claim more for it: other ends can “trump” it when it is not “sufficiently important”, so it is one weight in the standpoint and not a guarantor of ownership. The essay takes the end as given and does not address the regress. His synchronic stopping point, the self-governing policies, is in LIT-596 and unread.
- Korsgaard’s wanton argument (LIT-564, Lecture III, n. 4; THEORY-075). A reflective person cannot be a wanton: acting on the desire of the moment treats it as a reason, and that commitment “counts as a second-order volition”. Taken alone, that makes every reflective action issue from a second-order volition, the wanton’s included, so it cannot tell self-governed action from unself-governed action; it does not answer the question. What does the discriminating, for Korsgaard, is which law the agent is to herself, and that is fixed by her practical identity, “a description under which you value yourself”, which she chooses or constitutes. A stopping point specified by the agent’s own valuing of herself presupposes her participation: second horn.
- Korsgaard’s constitutive norms (LIT-587; THEORY-065). The stopping point is not an attitude but conformity to the categorical and hypothetical imperatives, which give action “the form of self-determined efficacy”. That looks like an escape from both horns, since conformity is fixed by the content of the norms and is indifferent to origin. It is not. The conformity that counts is being “motivated by the thought that you are an agent and so bound by the laws of agency” (ms. p. 17), and the norm conformed to is the norm of self-determination. Both specify the agent’s participation in advance. Korsgaard does not present this as a reply to Velleman; the placement is the record’s.
None of the three discriminates self-governed from unself-governed action without presupposing the agent’s participation, so none meets this account’s refutation condition.
Self-determination theory offers an empirical account of ownership that does not rest on order: a regulation is autonomous to the degree it is integrated with the person’s other values and needs, measured on a graded continuum (LIT-559; LIT-560, p. 1562; THEORY-054). Ryan and Deci set the regress aside as practically finite (LIT-560, p. 1561), which does not answer Velleman. Their stopping point, the “nascent self” of intrinsic activity and organismic integration (LIT-558, p. 247), meets this account’s second horn unless that self can be specified without the agent’s endorsement.
The account this one argues against is now filed as THEORY-120, from the 1971 paper read first-hand (LIT-692). The two are rivals. Frankfurt’s own stopping point there is the decisive commitment, which “resounds” through the higher orders (p. 16). On this account it is either disownable or specified so as to include the agent’s participation, so it lands on the dilemma like the others.
Note of 2026-10-30: what holds it back
“Identification and Wholeheartedness” (LIT-575) is still Unavailable, and the promote_when asks for it first-hand. Holton’s handout (NOTE-1455) reports a resonance condition in the checking case and none in the conflict case, so whether Velleman’s unwitting decision reaches the 1987/88 view turns on the conflict case, which has not been run against the essay’s text. The other replies the promote_when names (LIT-695, LIT-596, Watson 1975) are unread.