Bermudez criticizes decision theorists' “almost unanimous acceptance” (target article, sect. 2.1, para. 1) of the extensionality principle and argues instead that choosing within frames can rationally make different valuations of a given outcome incommensurable. But the “reason-giving aspects of different possible outcomes and possible actions” (target article, sect. 3.4, para. 3) are all based on reward: The prospect-theoretical and behavioral approaches he cites assume that people form cognitions in the service of getting reward, a view that cognitive theorists are also now accepting in the form of “predictive processing” (Gilead, Trope, & Liberman, Reference Gilead, Trope and Liberman2020) – the venerable behaviorist axiom that all learnable processes are actions. People's frames are then tools for organizing their hypotheses to test options by vicarious trial and error (Redish, Reference Redish2016), the outcomes of which are rewards. Rewards must in principle be commensurable (Ainslie, Reference Ainslie1992, pp. 28–32; Montague & Berns, Reference Montague and Berns2002; Shizgal & Conover, Reference Shizgal and Conover1996), which validates the norm of extensionality.

It is true that different frames can lead to fiercely conflicting conclusions, but Macbeth and Agamemnon manage to reach decisions. One frame finally displaces the other. After extreme failures competing frames may take turns running the whole self, as in alcoholic blackouts where “the alcohol talks,” or where real-life Jekylls give way to Hydes. Even such dissociation of identity is a motivated process, as shown by the struggle the normal self often experiences with impending takeover by an alter (Dell, Reference Dell, Dell and O'Neil2009).

An analysis of frames needs to keep the motivation for them in view. In discussing self control (sect. 4) Bermúdez cites me and the late Howard Rachlin, and starts out largely as we did, with hyperbolic delay discount curves, doubt about the effectiveness of effortful willpower, and a model of choice-framing. However, he presents framing as an unmotivated process, as if Mischel's children, for instance, would not have been tempted to think of marshmallows as something yummy instead of puffy clouds. This leads to trouble after he makes the unnecessary assumption that “an agent exercising self-control must change the shape of her discount function,” as opposed to recruiting additional larger-later (LL) incentive by framing her choices more broadly. To explain “how counter-preferential choice can actually be made [during temptation]” he has to posit a “hot” motive of “successfully resisting SS,” to which the subject “assigns more utility” than to the smaller-sooner (SS) reward itself. The notion that someone can make self-control hot – in effect an occasion for aroused emotion – by assigning it utility would seem to violate the laws of motivational gravity. If you could assign utility just anywhere, why not assign it to the LL alternative in the first place?

Admittedly, the laws of motivational gravity are still in some flux. There is reason to believe that much human reward is endogenous – a fiat currency not backed by external primary rewards (Ainslie, Reference Ainslie2013, Reference Ainslie2017, Reference Ainsliein press). Thus people do assign, or even create, utility, rather than just finding it. But self-created utility still has constraints. We seem to work up “hot” thought processes to indulge temptations, not to dampen them (e.g., Mischel, Reference Mischel2014). Bermúdez's other proposal for the value of a current LL choice makes more sense, and would not involve learning to bend what is probably an inborn discount curve (Ainslie, Reference Ainslie1992, pp. 123–125; Reference Ainslie2001, pp. 36–38): Resisting SS “might be taken as a signal of how the agent will react in the future (if I resist temptation here, then I will be more likely to do so in the future)” (target article, sect. 4.3, para. 5).

This is a solid reward-predicting perception, which would in fact motivate the mechanism Bermúdez did not consider, framing the choice broadly. Here is my choice bundling (Reference Ainslie1975, Reference Ainslie2021), or Rachlin's molar choice (Reference Rachlin1995, Reference Rachlin2016), or the broad bracketing of Read, Lowenstein, and Rabin (Reference Read, Lowenstein and Rabin1999), or, in effect, the high-level construal of Trope and Liberman (Reference Trope and Liberman2010). Essentially, if you make a current choice as part of a whole category of later choices, the relatively higher tails of the hyperbolic discount curves from the LL options in those choices will sum together to exceed the summed lower tails of the SS options plus the upward spike of the current SS option (Ainslie, Reference Ainslie2001, pp. 78–85). However, a person could just as easily frame the choice at hand as standing alone, unless an SS choice would confront her with evidence that she would keep choosing SS in similar situations, and would thus lose the prospect of the wider category of rewards (Ainslie, Reference Ainslie2001, pp. 90–100; see also Fujita, Reference Fujita2011, p. 360). There is always incentive to frame the choice at hand narrowly – “just this once,” and this incentive fuels the search for rationalizations – ways of distinguishing the choice at hand from the broader category of choices at stake. What obliges the person to include her present choice in a broad frame is its implication as a test case. And the present LL choice is not just a signal – an indicator of intent – but a sign – behavioral evidence – of how the agent will react in the future. Given this kind of motivated framing I can again agree with Bermúdez's model.

Incentives for frames are important in Bermúdez's other applications as well. The “I” frame and the “we” frame may be easily interchangeable in experimental games, but not when they lead to major consequences in the real world (sect. 5). In section 6 he proposes cognitive re-framing as a technique for conflict resolution, by analogy to the distancing technique in cognitive behavior therapy. But while it is undoubtedly helpful for negotiators to “appreciate how different frames bring different reasons into play” (target article, sect. 6.2.4, para. 1), he neglects the possibility that framing is itself a motivated choice. The “cognitive distancing” that is the starting place for his re-framing exercises involves withdrawing what may be highly valued emotional investment in a person's stance. Drama more than deduction commits political partisans to their positions – the plights of helpless babies-to-be, for instance, versus the injustice of entrapped pregnant women. The antagonists can probably imagine the others' points of view well enough, but do not want to, and may even feel that not doing so is good impulse control.