Source
Haggard (2017), LIT-1009, Nature Reviews Neuroscience 18(4), read in full from the accepted manuscript in NOTE-827: “Action selection” (manuscript pp. 7–9), “The comparator model of agency” (pp. 9–11), “Prospective versus retrospective agency” (pp. 11–12), “Brain mechanisms” (pp. 12–14), “Pathological sense of agency” (pp. 14–16) and Fig. 4.
What was actually shown
The review reports three kinds of evidence. Sense of agency here means the experience of controlling a specific motor act and its outcome, measured mostly by intentional binding, the compression of the perceived interval between action and outcome.
Prediction without the outcome. A voluntary action produced a tone on some trials. On trials where the tone came, the action was perceived as shifted further towards it, which is a retrospective effect. On trials with no tone, the shift was larger in blocks where a tone was likely (75%) than where it was less likely (50%). So “an additional, prospective component of the sense of agency is present even if the outcome does not in fact occur” (pp. 11–12). This could have come out otherwise: no-tone trials could have shown no block difference.
Its loss in psychosis. In a larger patient study, action binding in schizophrenia “depended entirely on the actual occurrence of the tone”, while in healthy volunteers it “depended on the probability that the tone might occur” (p. 16, Fig. 4).
Selection fluency. Masked primes that eased or hindered the choice of a keypress raised or lowered explicit judgements of control over a colour outcome that the primes did not predict. Further experiments ruled out reaction time as the cue (pp. 8–9). Angular-gyrus activity tracked low agency ratings only after incompatible primes, and TMS over it at prime time removed the effect (p. 13). A meta-analysis of seven tDCS studies found that stimulation of left DLPFC enhanced binding only where participants chose between actions (p. 8).
Against this the review sets the comparator model, on which agency is the absence of a prediction error and is computed after delayed feedback arrives, so that “people experience a sense of agency only retrospectively” (p. 11). Its conclusion: the circuit “operates both prospectively … as well as retrospectively”, and the post hoc inference view “has perhaps been overstated” (p. 18).
What this does not say
- It does not say the experience tracks actual control. The review calls the selection-fluency contribution “in a sense, illusory” (p. 9). A prospective signal is by its nature present before anything has been controlled. The claim is about what the experience is computed from, not about its accuracy.
- It does not say retrospective inference is absent. Both components are reported, and the claim is “not only”.
- It does not locate the computation. The frontal–parietal connectivity proposal (p. 14) is an argument from where activations fall, and is not filed.
- It is a review. The prospective component rests on one binding design and its patient extension, reported, not read here. The review’s author ran or co-authored most of the studies cited for it.
- It is about the motor-instrumental sense of agency. The review sets aside self-efficacy and the social attribution of authorship (p. 2, p. 6). Nothing here bears on whether a person experiences herself as the author of her life or values.
Connections
No relation is declared: no open question asks how the experience of agency is computed.
- QUESTION-048. Deficit evidence (schizophrenia) dissociates the two components, and the review finds explicit and implicit measures only weakly correlated (p. 5). Both bear on what different kinds of evidence establish about an experience.
- THEORY-300, filed from the same reading, uses the same implicit measures on coercion. If the coercion effect works through constrained selection, as the review suggests (p. 17), it is an instance of this account’s selection signal.