Source
- McKenna & Coates, SEP Compatibilism, LIT-291, §§4.2.2, 4.3 and supplement §A, as read in NOTE-270.
- Dryden, IEP Autonomy (2010), LIT-297, §2.a.ii, as read in NOTE-266.
- Buss & Westlund, SEP Personal Autonomy (2018 revision), LIT-296, §3, as read in NOTE-267.
- Pettigrew (2026), LIT-202, as read in NOTE-144. This is a worked historical condition, in a neighbouring question.
What was actually shown
A “mesh” or “snapshot” account makes free agency a relation among the agent’s present attitudes. Frankfurt’s is one: the effective desire issues from the will the agent wants. So is Watson’s valuational account, and so are reasons-responsive mechanisms described without their history. Such a relation says nothing about how the attitudes came to be there. A manipulator who installs the right mesh therefore produces an agent the account must call free.
LIT-291 argues that this holds for every mesh compatibilism it surveys: Frankfurt, Wolf, Fischer & Ravizza, Haji, Wallace, Bok and Mele (C12). Each must either say what is wrong with manipulation cases or “bite the bullet”.
Frankfurt bites it. LIT-291 quotes him from 2002 (p. 27, “Reply to Fischer”): a manipulator “may succeed … in providing a person … with a new character. That person is then morally responsible”. Whether the causes are natural or “the deliberate manipulative designs of other human agents” is “irrelevant”. This quotation is second-hand. It is the only verbatim Frankfurt passage in the entry apart from one phrase.
LIT-297 states the same objection against hierarchical autonomy (the hypnotist case, citing Mele 2001). It reports Christman’s historical model (1991) as the standard repair, and reports that the repair mishandles freely chosen loss of autonomy.
LIT-296 argues (§3) that coherence is not sufficient for minimal autonomy because the endorsement itself can be manipulated. It adds, in one sentence, that historical supplements have trouble telling autonomy-conferring reasoning from autonomy-undermining reasoning “without implicitly appealing to the very phenomenon one is trying to explain”.
LIT-291 reports a further step: manipulation “from a temporal distance” can load in a historical requirement too (Pereboom; Mele’s zygote argument). On this account Bratman’s externalist view is also exposed.
LIT-202 shows what a historical condition looks like in use. Whether a preference grounds effective consent depends on “the space and resources” its formation mechanism left “for the free exercise of agency”. Its sorting criterion uses free agency without analysing it (NOTE-144, C6). That is the circularity LIT-296 names.
The first half of the claim follows from the form of a snapshot condition, and all three surveys agree on it. The second half, that no historical condition in the record escapes, is the record’s synthesis. It rests on reports and one-sentence arguments.
What this does not say
- It does not say manipulated agents are unfree. That is the premise the manipulation argument needs, and Frankfurt denies it. The claim is a forced choice. It does not settle which option is right.
- It does not say historical conditions fail. LIT-296 asserts the circularity worry and does not argue it. Pereboom’s and Mele’s arguments are known only through LIT-291. Christman’s counterfactual non-repudiation test is named and not assessed.
- It is not the regress objection. The regress (THEORY-029) asks what gives a present attitude authority. This objection grants the authority and asks about its origin. An account could answer one and fail the other.
- LIT-202 is not about free will. It is used only as an example of a historical, mechanism-based condition. Its “Frankfurt-style” point concerns the 1969 alternate-possibilities cases, not the hierarchy.
Connections
-
LIT-294 (Robb): the 1969 Frankfurt cases leave identification as one of two compatibilist accounts of responsibility. That makes this objection’s reach wider than the hierarchy.
-
The curation entry of 2026-09-29 on Frankfurt records that the resigned-addict amendment of 1994 is reported second-hand and is contradicted in part by Frankfurt’s own 2004 text (LIT-293, p. 173).
-
Self-determination theory supplies a historical condition with experimental support: the same regulation is introjected when taken in under control and can be integrated under autonomy support (Deci et al. 1994, as reported in LIT-559, pp. 73–74, and LIT-558, pp. 237–238; THEORY-047). Ryan and Deci’s own test of autonomy is counterfactual reflective endorsement (LIT-560, p. 1572), close to Christman’s. They do not apply either to manipulation, and their sorting of histories by “autonomy support” uses the notion of autonomy it serves, which is the circularity this account records from LIT-296.
-
Korsgaard takes the first option for activity and not for deep responsibility (LIT-587, read first-hand in NOTE-455; THEORY-065). Her account is structural by design. What makes an agent active is conformity to the constitutive norms: “It is the content of the principles we follow, not the metaphysical story about how we came to follow them, that explains the activity implication. It does not matter how it comes about that you do it” (ms. p. 17). A manipulated agent who follows the right principles is therefore active. But history is kept for responsibility of the deepest kind: “For that, it does matter how (and whether) it comes about that we think thoughts about self-determination”, and “we do not all have the same opportunities for self-constitution; responsibility is not that deep” (ms. p. 19, n. 15). That splits this account’s forced choice by what is being analysed, a possibility its title does not mention: free agency in the sense of activity bites the bullet, and deep responsibility adds history.
-
Bratman, by Korsgaard’s report, bites the bullet. Korsgaard asked him whether an intention acquired by post-hypnotic suggestion would be any less the person’s own, and reports that he said no, “this would be just as much the person’s own intention, and the person’s own action, as any other” (ms. p. 21). That is second-hand, from a seminar exchange of 2009. It fits his model of self-governance over time (LIT-567, THEORY-076), which is wholly structural: present standpoints and the links of intention between them, with no condition on how the plans were acquired. That essay does not discuss manipulation.
-
Durability, a third kind of condition. Korsgaard then grants that her own objection to the hypnotic intention “is not after all wholly based on its origin”: we picture its effects “as ephemeral”, and might think differently of an installed intention with “long-term consequences for the person’s identity”, as with Demetrius in A Midsummer Night’s Dream, “changed forever by the outside intervention”. Of such cases “we are rightly puzzled” (ms. pp. 22–23). Durability, or integration over time, is a condition that is neither a snapshot nor a history. It does not escape this account: a manipulator can install a durable state, which is Demetrius’s case, and Korsgaard leaves that case puzzling rather than settled.
-
Note of 2026-10-14, after two surveys of the four-case argument (LIT-1047, NOTE-862; LIT-1045, NOTE-856; THEORY-328). They bear on both halves. Neither is a first-hand reading of Pereboom or Mele.
- First half. Pereboom’s Cases 1 and 2, quoted, stipulate a manipulated agent who meets reasons-responsiveness, first- and second-order desires and a general ability to respond to moral reasons. De Marco and Cyr quote Frankfurt’s hard line, that freedom does “not depend on the origin of the conditions” (1988, 45–6).
- Second half. Both surveys state it as the field’s main worry about soft-line replies: any condition named invites a revised case. Mele’s original-design case (Ernie) defeats bypassing conditions, among which De Marco and Cyr list Fischer and Ravizza’s (n. 15). A natural-force case defeats conditions that turn on the manipulator. Caruso reports Levy’s luck argument: the acts by which an agent takes responsibility for her endowment “either express that endowment … or reflect the agent’s present luck”. That is an argument against a named historical condition, not the one-sentence circularity worry, but it is reported, not read.
Note of 2026-10-29: what holds it back
Kept Proposed by the tenth status review. The first clause of promote_when is met: Frankfurt 1971 is read first-hand (LIT-692, NOTE-545), and its note 10 accepts the designed addict as fully responsible; the 2004 lectures (NOTE-264, p. 172) count “not the story of how we came to be”. The second is not: no first-hand reading of Pereboom’s four-case argument or Mele’s zygote argument shows a manipulation meeting a named historical condition, and the record holds both only through surveys (NOTE-270, NOTE-856, NOTE-862; THEORY-328). The argument’s first half is now THEORY-1118, Active.