Bermúdez claims that agents think about framed outcomes, not outcomes themselves. As he puts it: “We cannot help but see the objects of choice as framed, or described, or conceptualized in certain ways” (target article, sect. 3.1, para. 5). Accordingly, he argues that our preferences range over framed outcomes, not outcomes themselves.

We are supposed to apply this lesson to Bermúdez's literary case studies of Aeschylus's Agamemnon and Shakespeare's Macbeth. Agamemnon is dealing with two distinct framed outcomes, Following Artemis's Will and Murdering his Daughter, which simply happen to share an extension in his particular context. Whether Agamemnon prefers Following Artemis's Will to the alternative of Failing his Ships and People is one question. An independent question is whether Agamemnon prefers the framed outcome Murdering his Daughter to the alternative of Failing his Ships and People. Agamemnon prefers Following Artemis's Will to Failing his Ships and People while also preferring Failing his Ships and People to Murdering his Daughter. This is despite the fact that following Artemis's will involves murdering his daughter. It is because Agamemnon's preferences range over framed outcomes, rather than outcomes themselves, that he is supposed to escape the charge of having inconsistent preferences.

Similarly, Macbeth is dealing with two distinct framed outcomes, Murdering the King and Bravely Taking the Throne, which happen to share an extension in his context. In this case, there are also two distinct framed alternatives, Fulfilling his Double Duty to Duncan and Backing Away from his Resolution to Make the Prophecy come True. These too are co-extensive. Macbeth's preference between Murdering the King and Fulfilling his Double Duty to Duncan is supposed to be independent of his preference between Bravely Taking the Throne and Backing Away from his Resolution to Make the Prophecy come True. Macbeth prefers Fulfilling his Double Duty to Duncan to Murdering the King, while also preferring Bravely Taking the Throne to Backing Away from his Resolution to Make the Prophecy come True. Again, since Macbeth's preferences range over framed outcomes, rather than outcomes themselves, it is argued that his preferences are consistent.

Whatever the other merits of defining preferences over framed outcomes, I deny that doing so can justify patterns of preferences like Agamemnon's or Macbeth's. This is because both agents are stipulated to know that they are dealing with pairs of frames which describe the same outcome. In each case, then, that common outcome, must be an object of the agent's intentional state of knowledge. Whether it is an extensional phenomenon or just another framed outcome, it must at least be neutral between the two target frames; otherwise it would be impossible to know that it is shared by each. So, knowing that two frames have the same outcome involves having an intentional state, the object of which is a relevantly frame-neutral outcome. And, once we have accepted that agents are able to think about relevantly frame-neutral outcomes, there is no reason to suppose that their thoughts and preferences should be restricted only to the two framed outcomes.

We can unpack this a little by re-examining Agamemnon's dilemma. As Bermúdez explicitly asserts, Agamemnon knows full well that Following Artemis's Will and Murdering his Daughter are the same outcome, differently framed. Yet if Agamemnon knows this, then an object of his intentional state of knowing is an outcome that is necessarily independent of – or neutral between – these two alternative ways of framing it. In other words, Agamemnon can and does think about the outcome independently of the two frames in question. Given this, there is no reason to suppose that his preferences should range only over the framed outcomes Following Artemis's Will and Murdering his Daughter. Instead, Agamemnon's preferences can concern the frame-neutral outcome – say, Killing Iphigenia.

In fact, it seems entirely right that Agamemnon's preferences do concern the frame-neutral outcome. He fully appreciates that there are strong reasons for and against killing Iphigenia. These competing reasons do not remain frame-relative, even if they are initially made more salient by one or other frame. Instead, Agamemnon recognises that the reasons pertain simultaneously to the single shared outcome. The great difficulty he faces is in how to weigh them up and decide which should take precedence. Thus, Agamemnon's dilemma is substantive, not merely linguistic. Indeed, it may be precisely Agamemnon's ability to reason beyond the two frames – and not remain bound by them – that makes his dilemma so acute.

Note that, if this analysis is correct, Agamemnon's preferences are straightforwardly cyclical. He oscillates between two diametrically opposed preference orderings, sometimes preferring the frame-neutral outcome Killing Iphigenia to the alternative of Failing His Ships and People, and sometimes the reverse. By Bermúdez's own lights, such cyclical preferences cannot be rationally maintained.

A parallel analysis can be run for Macbeth, and for each of Bermúdez's examples concerning self-control, strategic coordination, and discursive deadlock. This is no accident, as the criticism generalises across the set of framing effects Bermúdez is interested in. After all, he explicitly focuses on situations where agents are well aware that they are framing a single outcome in different ways. As I have argued, such awareness requires agents to be conceptualising the outcome in a relevantly frame-neutral way. It is then no longer clear why they cannot or should not have preferences about the frame-neutral outcome. On the contrary, it seems absolutely right that they can and should.

Bermúdez's argument, then, does not allow us to conclude that there are rational quasi-cyclical preferences. Instead, I believe Bermúdez must acknowledge that agents can and do conceptualise outcomes in relevantly frame-neutral ways; and that their preferences can and do range over these frame-neutral outcomes.

There might still be some other route to the conclusion that preferences like Agamemnon's and Macbeth's are ultimately rational. However, I believe this would require an entirely different line of argument. For a sketch of how it could look, see Fisher (Reference Fisher2022). In the meantime, the jury remains out on such cases.