II. Frankfurt's Argument and the Main Moves Against it

Frankfurt’s attack on the PAP is a modern version of an attack made by John Locke in An Essay Concerning Human Understanding. Locke contends that someone who is (unwittingly) locked in a room might stay there voluntarily if she is engaged in pleasant conversation.Footnote ³ On the basis of this, Locke distinguishes between whether an action is done voluntarily and whether an action is done freely: the latter requires the ability to do otherwise but the former does not, and responsibility tracks the former:

In Frankfurt’s example, Locke’s door is replaced by an agent, Black.Footnote ⁴ Black wants Jones to perform some action, A, and Black “is prepared to go to considerable lengths to get his way, but he prefers to avoid showing his hand unnecessarily” (Frankfurt, Reference Frankfurt1969, p. 835). Thus, Black waits to see whether Jones is going to A on his own, prepared to intervene if, but only if, not. When Jones performs A of his own volition, Frankfurt concludes that, because Black did not actually intervene, Jones is responsible for A despite his inability to do otherwise.

As noted above, some respondents to Frankfurt have conceded this attack on the PAP but have contested Frankfurt’s attempt to use it to motivate compatibilism.Footnote ⁵ Many who do so have proposed alternative principles that are not falsified by Frankfurt cases but that preserve the connection between responsibility and the ability to do otherwise. For example, Peter van Inwagen proposes three alternate principles having to do with omission and event responsibility; James Lamb proposes a weak version of the PAP; Keith Wyma proposes a principle about passing the buck; and Michael Otsuka proposes a principle about avoiding blame (Lamb, Reference Lamb1993; Otsuka, Reference Otsuka1998; van Inwagen, Reference van Inwagen1978; Wyma Reference Wyma1997).Footnote ⁶

Others maintain that agents in Frankfurt cases are able to do otherwise (pace Frankfurt). Probably the best-known version of this is called the “dilemma defence,” articulated independently by Carl Ginet, Robert Heinaman, Robert Kane, and David Widerker (Ginet, Reference Ginet1996; Heinaman, Reference Heinaman1986, p. 276; Kane, Reference Kane1985, especially p. 51, note 25; Widerker, Reference Widerker1995). For present purposes, the part of this defence that is relevant is that if indeterminism holds in Frankfurt cases, then the agents will have morally relevant alternate possibilities.Footnote ⁷ This has led to an arms race of ever more sophisticated Frankfurt cases and responses.Footnote ⁸

A third approach, popular more recently, is to attack the idea that the agents in Frankfurt cases are morally responsible for their actions. For example, some have pointed to a connection between the PAP and other principles frequently appealed to in moral theorizing, like “ought implies can” (OIC); others have argued for an analogy between knowledge and moral responsibility, creating mashups of Gettier casesFootnote ⁹ with Frankfurt-style counterfactual interveners.Footnote ¹⁰ This has led to denials of the connection between the PAP and OIC, denials of OIC directly, and more.Footnote ¹¹

In the next section of this article, I am going to begin a novel kind of attack. I propose to look at Frankfurt cases involving merely permissible actions. I argue that mere permissibility entails the ability to do otherwise and, thus, that there cannot be a successful Frankfurt case in which an agent is responsible for performing a merely permissible action notwithstanding her inability to do otherwise. I then argue that this exposes a deep problem for Frankfurters and their attempt to ground super/compatibilism by disengaging moral responsibility from the ability to do otherwise.Footnote ¹²

III. My Argument: Mere Permissibility and the Ability to Do Otherwise

Let me define an mp-Frankfurt case as a thought experiment involving a counterfactual intervener waiting in the wings to ensure that an agent performs an action that is merely permissible. The counterfactual intervener in an mp-Frankfurt case is, as in a normal Frankfurt case, only counterfactual: she does not get involved in the actual sequence of events; in the actual sequence of events, the agent performs the action of her own volition. The point of defining this as a special category is that it deals with a merely permissible action rather than an impermissible one.Footnote ¹³

For example, suppose that J goes out to dinner and is trying to decide what to order. We may stipulate, so as to ensure that there are no confounding moral factors in the case, that the restaurant serves only gluten-free, vegan food sourced from local, organic farms that pay their workers a living wage; that the restaurant employs a diverse staff and pays them all a living wage; that the restaurant is not part of a chain or conglomerate; that the restaurant does not have prices that make it out of reach for any but the ultrarich (or even the somewhat rich); and so on. Sadly, it might be difficult to imagine such a restaurant. But I do not think that anybody would say that this is so because of the logical, nomological, physical, metaphysical, social, or even economic impossibility of such an establishment.Footnote ¹⁴ Now back to J. J is vacillating between the gnocchi and the tagliatelle (both made in-house).Footnote ¹⁵ After a few minutes of deliberation, J decides on the former.

The setup of this thought experiment is exasperatingly deliberate for such a trivial decision. But that is the point: the decision is trivial. That is, it is morally indifferent which pasta dish J chooses. Choosing what to have for dinner is mundane, and most of us, when we take ourselves to have live options, take our choices to be morally indifferent in the sense that anything we select would be permissible. J’s choices are constrained (hopefully sufficiently) so as to avoid any objection on this front.

The example could have been rigged up to make J’s decision morally loaded. The gnocchi might have been served with freshly butchered baby legs, browned and then drizzled with a cream sauce made from their mothers’ milk. Think Procne or Tantalus, only worse. But the point is that the example has not been rigged up in this way. The restaurant is unusual; it is not macabre or mythic.

Importantly, for present purposes, moral responsibility for an action is not in general precluded by virtue of the fact that the action is merely permissible. Indeed, in this particular thought experiment, at least as I am envisioning it, J is morally responsible for choosing the gnocchi. To see this, suppose that the waiter brings out the dish and places it in front of J. J stares at it in such a way that the waiter is prompted to ask whether something is wrong, whereupon J begins to berate the waiter for bringing out the gnocchi rather than the tagliatelle. J demands to speak to the manager and refuses to pay if the tagliatelle is not brought out at once. The waiter apologizes for the mistake and fetches the manager. The manager thereupon apologizes: the meal will be on the house, of course; the waiter must have mixed up the orders. But then J says, “No, I ordered the gnocchi. There was no mistake. The point is that I want the tagliatelle.” I think the absurdity of this situation is on the surface: it is J who is at fault, not the waiter or the restaurant. Although J’s decision to order the gnocchi rather than the tagliatelle is merely permissible, J’s behaviour in this extended sequence is not, and the reason starts with J’s responsibility for the original order.

So far, this thought experiment has two components that are crucial to an mp-Frankfurt case: the agent is responsible for an action, and that action is merely permissible. But nothing has been said yet about J’s ability to do otherwise; that is the third ingredient that must be added in order to complete the recipe. So, now suppose that, at a nearby table, but unbeknownst to J, B was watching and monitoring from the moment that J sat down. B very much wants J to order the gnocchi, not the tagliatelle. Moreover, B would prefer that J order the gnocchi without having to intervene. But B is prepared to intervene, if necessary, using a small, handheld neuroscope.

Why does B want J to order the gnocchi without intervention? Maybe B wants the manager to remove the tagliatelle from the menu because B thinks that tagliatelle is an abomination, and so B does not want anybody to order the tagliatelle (B monitors all the customers, not just J). And maybe intervention in J’s decision-making process would be costly (neuroscopes run on diamonds (or baby legs), and B must keep this up for a month or so if the manager is going to take the hint). Or maybe B would feel validated if J’s unvarnished preferences coincided with B’s own. The details are unimportant.

What is important is that B would have intervened if J had been about to order the tagliatelle. But that turned out to be unnecessary. J ordered the gnocchi of her own volition.

Now, at this point in the story, the question is usually: does the addition of B to the thought experiment (the third ingredient) flip the intuition about J’s responsibility for performing a merely permissible action? Or perhaps: would B’s presence in any way mitigate J’s bad behaviour when she begins complaining to the waiter?

A Frankfurter would answer both of these questions in the negative and thence infer that the PAP is false. But that is not the route that I want to take. To illuminate the route I want to take, I want to put this thought experiment down for a moment. I want to argue, instead, for the following thesis:

The relevance of the MP-thesis is, I hope, clear: if it is correct, then there are no successful mp-Frankfurt cases. I am going to make a preliminary argument for this thesis here and then explain what I think we should say, in light of this thesis, about J.

The argument I want to make for the MP-thesis is based on the following three premises:

• A. If an action is merely permissible, it is not obligatory.

• B. An action is obligatory if and only if, if an agent omits the action, she is behaving impermissibly.

• C. If A and B, then MP-thesis.

Premise A says that no action is both merely permissible and obligatory. I think that premise A is easy to motivate. I think that premise A may be motivated by appeal to the meaning of the modifier “mere.” Some obligatory actions might be permissible and, if there are moral dilemmas, some obligatory actions might be impermissible. But, no merely permissible action is anything more than permissible: no merely permissible action is meritorious, owed, impermissible, or anything else. Premise A, on my account, is an analytic truth. There is support for this in the Oxford English Dictionary, which defines “obligatory” as “compulsory and not merely permissive” (OED online, 2020, definition 2a for “obligatory,” my emphasis).

Premise B says that it is impermissible to omit an obligatory action and, conversely, that the omission of obligatory actions is impermissible. In other words, it says that positive obligations (actions that agents have an obligation to perform) can be accounted for (without residue) in terms of omission and impermissibility.

Some might argue for premise B on the grounds that (i) an action is obligatory if and only if its omission would be blameworthy, and (ii) the omission of an action would be blameworthy if and only if this omission would be impermissible. But this kind of argument would be blocked by those who assert that justifications and excuses are distinct and that there are excuses: if there are excuses that are not justifications, then, although the omission of some obligatory actions is blameworthy, some omissions of obligatory actions are not blameworthy.

That is, an excuse, in this context, is an exculpatory condition that does not change the deontic status of an action.Footnote ¹⁶ So, if an agent has an excuse for performing an impermissible action, then she is not blameworthy for performing that impermissible action. Similarly, if an agent has an excuse for omitting an obligatory action, then she is not blameworthy for that omission. For this reason, I do not want to ground premise B using the argument in the previous paragraph: I do not want to take a stand on excuses here.Footnote ¹⁷

However, I think that premise B can be grounded without appeal to the concept of blame. I think that premise B is, like premise A, analytic. One piece of evidence for this can be found in standard deontic logics. For example, according to the Stanford Encyclopedia of Philosophy entry for Deontic Logic, the “Traditional Definitional Scheme” includes the following two definitions: “p is impermissible if and only if (not-p) is obligatory” and “p is omissible if and only if it is not the case that p is obligatory” (McNamara, Reference McNamara and Zalta2021, Section 1.2).Footnote ¹⁸ Similarly suggestive is the fact that the Oxford English Dictionary defines “obligation” in terms of liability for omission (OED online, 2020, definitions 1 and 7 for “obligation”).

Some might point out that premise B is incomplete. In particular, premise B makes no provision for negative obligations. Premise B, thus, must be supplemented with the following: the omission of an action is obligatory if and only if, if an agent performs the action, she is behaving impermissibly.

But I do not think that premise B is incomplete. The supplement just suggested would be important if my end goal here was to engage in an analysis of deontic categories. But that is not my end goal: my end goal is to ground the MP-thesis, and I can do that without this supplement. It suffices to deal with positive obligations because Frankfurt cases, like the mp-Frankfurt case above, traditionally involve action performances rather than omissions.

Premise C is the glue that connects premises A and B with the MP-thesis. To see why premise C is true, suppose that premises A and B are true but that the MP-thesis is false. If the MP-thesis is false, then there is an agent whose action is merely permissible but who is unable to do otherwise. But from the fact that the agent’s action is merely permissible conjoined with premise A, it follows that the agent’s action is not obligatory; and from the fact that the agent is unable to do otherwise conjoined with premise B, it follows that the agent’s action is obligatory. So, the agent’s action is not obligatory and the agent’s action is obligatory. This is a contradiction. Thus, premise C is true.Footnote ¹⁹

Some might object to the way in which I have used premise B in this argument. But further support both for the analyticity of premise B and for the way in which I deploy it here is that this kind of move seems to be baked into many discussions in normative ethics, even when these discussions are from philosophers from entirely distinct camps and with entirely distinct purposes. For example, in a discussion of Kantian moral worth and acting from duty, Marcia Baron makes the following remark in passing:

For current purposes, what is important about this passage is the inference contained in the parenthetical remark: if there is only one permissible action open to an agent, then the action is morally required. In other words, if a permissible action cannot be omitted, then it is morally required. Similarly, in David Lyons’ now classic discussion of utilitarianism, he remarks that utilitarian principles about the rightness of an action should contain the following condition:

This concludes my preliminary argument for the MP-thesis.Footnote ²⁰ I shall continue this argument in the next two sections of this article when I consider objections to it. For now, however, I want to return to the thought experiment with which I opened this section. Above, I asked two questions concerning J’s responsibility: does the addition of B to the thought experiment (the third ingredient) flip the intuition about J’s responsibility for performing a merely permissible action? And, would B’s presence in any way mitigate J’s bad behaviour when she begins complaining to the waiter? I shall answer these questions in turn.

I think it should be clear by now that I want to answer the first question in the affirmative. But some care must be taken in explaining why I think the first question should be answered in the affirmative.

I think that the MP-thesis shows that there are no successful mp-Frankfurt cases, and in the particular case at hand, I think that the MP-thesis reveals that the addition of a counterfactual intervener to the thought experiment changes the deontic status of J’s action: with the addition of a counterfactual intervener, if the intervener really can block J from doing otherwise, J’s action is no longer merely permissible. If J’s ordering the gnocchi is still an action (an issue I shall not explore in this article), then the action is, from premise B, obligatory.

The reason that I think this order of operations is important in answering the first question above is that it enables me to offer an explanation for why some might think (in disagreement with me) that the first question should be answered in the negative. In particular, those who think that Frankfurt cases are generally successful might say that the addition of a counterfactual intervener flips the deontic status of J’s action but not J’s responsibility for this action: they might concede that J’s action is no longer merely permissible but they may nonetheless maintain that J is responsible for ordering the gnocchi. In other words, a Frankfurter might concede that there are no successful mp-Frankfurt cases but nonetheless maintain that this is only because the addition of a counterfactual intervener transforms the thought experiment into a successful Frankfurt case involving an obligatory action.Footnote ²¹

Further explanation for the temptation to answer the first question in the negative might come from the second question I asked about J (the question about J’s complaining to the waiter). So, I want to try to tackle that question now.

It seems to me that the second question, unlike the first, should be answered in the negative: the presence of a counterfactual intervener does not, in itself, mitigate J’s bad behaviour when J begins complaining to the waiter. This, as just pointed out, might pull some toward saying that J is responsible for J’s initial action of ordering the gnocchi even in the presence of the counterfactual intervener. But, it should not pull anyone in that direction. To see why not, rewind the thought experiment to the point just before the counterfactual intervener was added and construct an alternate state of affairs. Suppose that J has an identical twin (think Twelfth Night or The Comedy of Errors). Suppose, further, that, unbeknownst to the two of them, they happen to be dining out that night in restaurants that are immediately adjacent and that are identical in every way, even employing identical twins in identical positions. Immediately after ordering, J and her twin get phone calls. They step out of their respective twin-restaurants; and, while on their twin phones, they inadvertently walk past one another and then enter into the other restaurant thinking that it is their own. They sit down at their twin-tables and wait for their twin-orders to be brought by their twin-waiters, whereupon the twin-complaining begins. In this modified case, neither twin ordered the food they are complaining about, but both of them think that they did, and their blameworthiness for the complaining is not, to my mind, mitigated. If this is correct, then the two questions I asked about J can be unhitched: J might not be responsible for ordering the gnocchi but nonetheless be blameworthy for complaining about the order.Footnote ²²

At this point, an error theory might be demanded to explain why these two questions seem to be coupled even though actually they are not. Perhaps the following will do: J’s blameworthiness for the complaining rests on J’s understanding of the situation, and in the twin thought experiment, that understanding includes false (but nonetheless justified) beliefs about past actions, including having ordered the gnocchi now in front of her. In the normal course of life, these beliefs will be true, justified in run-of-the-mill ways by what actually took place. But in Gettier- (and Shakespeare-) inspired fake barn and twin country, things go a bit haywire.

I am not sure whether the idea in the previous paragraph will withstand critical scrutiny. But I do not want to get distracted by it: my main focus is to put forth a novel argument about Frankfurt cases, not to put forth a full-blown theory about responsibility. So, I turn now to some objections that might be raised against the MP-thesis.

IV. Objection 1: Moral Ability and Metaphysical Ability

Some might reject the MP-thesis, the thesis that if an agent’s action is merely permissible, then she is able to do otherwise. I suspect that most who would reject the MP-thesis would do so by rejecting premise B, the premise that states that an action is obligatory if and only if, if an agent omits it, she is behaving impermissibly. In particular, some might argue that the counterfactual analysis of obligation found in premise B can be used to undermine Frankfurt cases in the absence of the other machinery deployed in Section II of this article, and it might be urged on these grounds that premise B (and, thus, the MP-thesis) is, as such, question-begging in this context.

The idea behind this objection is something like the following. As noted above, standard Frankfurt cases involve impermissible actions. But if such cases are successful, then the agents in them are unable to do otherwise. Premise B, then, implies that such impermissible actions are obligatory. It is absurd for an action to be both impermissible and obligatory. Therefore, premise B suffices to undermine standard Frankfurt cases on its own. As such, no Frankfurter should accept premise B, and my argument for the MP-thesis comes tumbling down.Footnote ²³

I would like to say three things about this.

First, those who affirm moral dilemmas will not accept the claim that it is absurd for an action to be both impermissible and obligatory. That is, in moral dilemmas, all options are both obligatory and impermissible. So, if there are moral dilemmas, some actions are obligatory and impermissible, whence it follows that <obligatory> is not a subcategory of <permissible>, and the reductio in the previous paragraph fails.Footnote ²⁴ Thus, premise B does not suffice, on its own, to undermine Frankfurt cases. Indeed, one advantage of the argumentative strategy employed in the previous section of this article, then, is that it does not require me to take a stand on the (vexed) question of whether there are genuine moral dilemmas involving conflicting (ultima facie) obligations (or whether, in such cases, there are merely conflicting grounds of obligation).

Second, even for those who reject moral dilemmas and who think that it would be absurd for an action to be both impermissible and obligatory, premise B would not be question-begging in the context of a dialogue with a Frankfurter. Obvious cases of question-begging assertions involve the mere repetition of a claim as a defence against an attack on that same claim. But if appealing to a premise that implies that an interlocutor’s position is misguided is question-begging, then all philosophical dispute is fallacious (including, of course, the attempt to undermine mine). It is not a problem for me if there is an independent argument against Frankfurt cases more broadly that can be motivated using, in part, some of the premises to which I have appealed in Section II of this article. Again, I do not want to take a stand here on the reality of moral dilemmas. The point I want to make is merely that it would be strange if additional arguments for one’s conclusion were taken as countermanding that conclusion.

Third, there is an easy modification of premise B that simultaneously (i) highlights the importance of my focus in this article on merely permissible actions, and (ii) neutralizes this line of objection completely. Consider:

• B′. If an action is permissible, then it is obligatory if and only if, if an agent omits it, she is behaving impermissibly.

B′ differs from B in that B′ limits the scope of B to permissible actions. Because the MP-thesis is only about merely permissible actions, B′ can still play the same argumentative role as B. That is, the conjunction of premise B′ with premises A and C still entails the MP-thesis.Footnote ²⁵ Furthermore, B′ neutralizes the Frankfurter’s objection precisely because Frankfurters concentrate on impermissible actions (i.e., actions to which B′, unlike B, does not apply), thereby making perspicuous why it is so important that we consider also the category of the merely permissible in this context.

V. Objection 2: Asymmetry in Deontic Categories

Some might accept the MP-thesis but argue that it is irrelevant. Some might point out that if we accept the analysis leading to the MP-thesis, then (i) mere permissibility entails the ability to do otherwise; but (ii) the ability to do otherwise is obviously relevant in this context because it can ground an asymmetry between mp-Frankfurt cases and other kinds of Frankfurt cases; so (iii) there is a relevant difference between mp-Frankfurt cases and other kinds of Frankfurt cases; whence it may be concluded that (iv) the MP-thesis does not impugn Frankfurt’s reasoning. Further support for this objection can be garnered from early Frankfurt literature that argues for an asymmetry between impermissible and meritorious actions (Wolf, Reference Wolf1980). This calls into question the relevance of whether there is a special category of Frankfurt cases, those involving merely permissible actions, that cannot be made to work: if my analysis leaves untouched Frankfurt cases involving impermissible actions, and if the majority of the discussion to date has been about such cases, then who cares whether the MP-thesis is true?

To understand why the MP-thesis is relevant, recall the first wave of response to Frankfurt described in Section I of this article. The first wave granted Frankfurt's attack on the PAP but contested the connection between the PAP and compatibilism. I am not granting Frankfurt’s attack on the PAP in general. But, if the MP-thesis is true, then there is a version of the PAP for merely permissible actions:

If the Frankfurter grants that there are no successful mp-Frankfurt cases and, thus, the mp-PAP, that is sufficient to block Frankfurt’s move to compatibilism (and his followers’ move to what they call “supercompatibilism”), especially when one considers that most of everyday life concerns the merely permissible. So, any victory on the part of the Frankfurter that is based on an alleged asymmetry in deontic categories would be pyrrhic: this is not a good route for the Frankfurter to take.