Peter van Inwagen's now-familiar strategy to refute Frankfurt's thesis involves (among others) two claims.Foot note 4_2 The first general claim is that Frankfurt's framing of the issue, in a principle called PAP (for Principle of Alternative Possibilities) is overly simple. Thus, instead of Frankfurt's PAP:

(PAP) One is morally responsible for one's acts only if one could have done otherwise than one did.

Van Inwagen wishes to substitute his own PPA (Principle of Possible Action) specifically for instances of moral failure:

(PPA) One is morally responsible for failing to perform an act only if one could have performed that act.

Secondly, van Inwagen proposes an example of PPA that results in an agent's lacking responsibility. He hypothesizes that an apathetic agent who knows of a crime in progress fails to call the police. Coincidentally, it turns out that the only phone he could have used is broken. Therefore, that agent is not fully responsible by PPA for the fact that the police were not called. In Frankfurt's analysis, this is so because while van Inwagen's agent's behavior was sufficient for the fact of his failure to call the police, it was not necessary in virtue of the further fact that the phone was broken. (I should note that Frankfurt resists the conclusion that the apathetic agent was not fully responsible for trying to call the police. Nevertheless, he continues to analyze conditions of responsibility here for the objective state of affairs that the police were not called.)

Thus, Frankfurt proposes the following revision to reconcile PAP with PPA. That is that one must distinguish personal from impersonal failure. Van Inwagen's case of failure was of the latter kind -- his agent in some sense inevitably failed because of states of affairs or events beyond his control. Then Frankfurt contrasts this case with another of his own construction, namely a personal failure of an agent Q who, because he voluntarily looks to the left at some crucial moment while driving, fails to keep his eyes on the road despite the fact that there are present (though actually otiose) overdetermining conditions which would have otherwise kept Q's eyes looking left. Unlike van Inwagen's case, Frankfurt holds Q personally and thus fully responsible for this failure because it is fully his act, even though there are overdetermining conditions which did not in fact influence Q's actions.

Frankfurt then provides a defense of his distinction between his case of personal failure and van Inwagen's impersonal case. To reinforce his claim that Q is indeed fully responsible for his failure, Frankfurt states:

Notice that Q is fully responsible for his failure. Failure to keep one's eyes straight ahead is exclusively a matter of what movements a person makes; it is constituted by what the person does, and what the person does is therefore both a sufficient and a necessary condition for it. It cannot be said, then, that Q's failure would have occurred no matter what he had done -- i. e., regardless of what bodily movements he made. If he had not moved his eyes to the left at all he would not have failed. (Frankfurt, 101)

One can sympathize with the sense of Frankfurt's claim here. His argument is an attempt to show that proper counterfactual analysis cannot remove or absolve Q's role in originating and causing the act of which he is accused. I wish to argue, however, that Frankfurt's use of this counterfactual claim as a basis to refute (his clarified version of) PAP cannot stand close scrutiny, because such a use of that claim relies upon equivocations on key terms.

It will be convenient to label Frankfurt's key claim as:

(F) If he [Q] had not moved his eyes to the left at all he would not have failed.

Since this is a counterfactual statement containing key moral terms, interpreting it requires explicit semantics, modally and otherwise. Perhaps the most convenient here for the modality of the statement is a Lewisian-style (L) account, since this allows an accounting of individuals in terms of closest possible worlds (to the actual one) by reference to those individuals' counterparts. (I should note here my confidence that any appropriate semantics will yield the same critical insights I offer below.)

Placing on (F) one such (L) account, and further elucidating the moral content of «failed» we have:

(RC/Remote Counterpart) If Q* had not (in that world actually) turned his eyes to the left at all, (then) he would not have failed (not have turned his eyes left and thus be not morally responsible for doing so).

(RC) posits a possible world in which an individual counterpart of Q, Q*, does nothing himself to move his eyes left (or anywhere else than on the road), and thus the state of affairs of his eyes-turning-left (or anywhere else) does not obtain. Note that the fact that the state of affairs of eyes-turned-left (E-T-L) does not obtain in Q*'s world entails that there is no overdetermination of that state of affairs (at that time) in that world. But, since Q*'s case is invoked precisely to justify Q's responsibility under overdetermined conditions in the actual world, the plausible relevance of PAP or PPA to Q*'s situation as one involving free action/free will is of no use in evaluating the responsibility of Q in the actual world where PAP or PPA doesn't apply. The very question is whether Q's action is a free and responsible one in spite of the inapplicability of PAP or PPA.

However, another (L) account of (F) is available:

(CC/Close Counterpart) If Q** had not (in that world actually) turned (or tried to turn) his eyes to the left at all, but nonetheless had done so because of overdetermination, then he would not have failed (would have not been responsible for what happened).

(CC) posits a possible world in which an individual counterpart of Q, Q**, does nothing himself (personally) to move his eyes to the left (or anywhere else), but overdetermining (impersonal) forces make him do just that, so the state of affairs of E-T-L does obtain. Q**'s world, like Q's, is an overdetermined one, but also one in which the actual functioning of the overdetermination results in the state of affairs of E-T-L. Like Q, Q** is thus also subject to conditions precluding PAP or PPA, although here that coincides with an evaluation of Q**'s lack of responsibility. Hence, it is plausible that this judgment of Q**'s lack of responsibility in part relies on the fact that the overdetermining conditions violate PAP and/or PPA. If so, then (CC) is of no use in supporting Frankfurt's apparent reading of (F). (It should be obvious that Frankfurt himself would challenge the relevance of PAP/ PPA to Q**'s lack of responsibility, since this would be a case of impersonal failure. That is quite aside the point, however, that (CC) cannot itself be used to leverage (F) against PAP or PPA.)

No doubt Frankfurt would greet my analysis thus far with something akin to an incredulous stare. What of the fact that neither (RC) nor (CC) backs (F) against PAP or PPA? That's not the job (F) was meant to do! Rather, (F) is only to uphold the claim that Q and Q alone produced the state of affairs of E-T-L, which is then described as a failure. Recall that Frankfurt says above that «[f]ailure to keep one's eyes straight ahead is exclusively a matter of what movements a person makes; it is constituted by what the person does, and what the person does is therefore both a sufficient and a necessary condition for it.» Clearly this language is metaphysical, as particularly evident in its assertion that the failure is «constituted» by Q's movements. So here «failure» refers merely to the physical movements of Q as described by the state of affairs E-T-L (or looking anywhere else than on the road). On the basis of such a purely metaphysical meaning, Frankfurt's subsequent statements that culminate in (F) are consistent and true -- though in that case resulting on the triviality of (F) as meaning only

(F*) If he [Q] had not moved his eyes to the left at all, he would not have moved his eyes to the left at all.

However, should Frankfurt protest that (F) means more than (F*), and specifically that Q (or more precisely, some counterpart of Q) «would not have failed» by avoiding E-T-L in the relevant world, then he explicitly invokes «fail» in an evaluative sense -- in the sense connotative of moral responsibility. But in that case (F) then reads as (RC), which cannot support the same evaluative sense of «fail» as originally occurs in (F) because of the plausible evaluative role of PAP/PPA in Q*'s world, where overdetermination does not occur. Although (CC) likewise involves an evaluative meaning of «fail», clearly it fares no better in supporting such a sense in (F) that favors Frankfurt's use of the term against PAP or PPA in overdetermined circumstances similar to Q's where overdetermination does not function.

There is only one conclusion to be drawn from this analysis: «failure» and all its cognates are used in Frankfurt's above quote in two senses, sometimes separately, sometimes combined. There is a clear purely metaphysical use, putatively evaluative uses, and arguably mixed uses. Labeling these respectively (M), (E) and (ME) we have:

Notice that Q is fully responsible for his failure (E)/(ME). Failure (M) to keep one's eyes straight ahead is exclusively a matter of what movements a person makes; it is constituted by what the person does, and what the person does is therefore both a sufficient and a necessary condition for it. It cannot be said, then, that Q's failure (M)/(E)/(ME) would have occurred no matter what he had done -- i. e., regardless of what bodily movements he made. If he had not moved his eyes to the left at all he would not have failed (E)/(ME). (Frankfurt, 101; my additions)

Note that the penultimate cognate of «fail» as «failure» is the vaguest -- although its use as (M) both yields the strongest and the most trivial reading of (F) as (F*) simultaneously. Excluding that, a pure (E) reading of (F) evokes the controversy of (RC) or the irrelevance of (CC), and a (ME) reading does as well.

Quo vadimus? The first conclusion is that Frankfurt has work to do to show that Q is fully responsible for his failure E/(ME). There is no doubt that Q caused his failure (M) as it so happened, but it remains for Frankfurt to supply an account that sweepingly supports Q's complete failure (E)/(ME). After all, one could plausibly argue that Q «had the deck stacked against him» -- that although he did in fact fail (M)/(ME) due to his own efforts, surrounding circumstances required him to fail (M)/(ME) come what may, pace Frankfurt. As well, one may plausibly reduce the responsibility of Q in direct proportion to what is felt to be the conspiratorial nature of the overdetermination. (Accidental entrapment by these circumstances versus God's deistic enforcement of them, say.) The point is that any such overdetermination is a factor that is not easily morally disregarded.

The second conclusion is that such parallel vagaries might contaminate Frankfurt's more familiar arguments against PAP. The general scenario of an effete entrapment is the staple of Frankfurt-style counterexamples, after all, and the problems that besiege (F) similarly plague analogous defenses of those counterexamples. Typically Frankfurt's strategy emphasizes the actual ineffectuality of the overdetermination, which subtly evokes similar counterfactual circumstances in which, like (RC), overdetermination doesn't apply. But such implied comparisons may well rely on an intuitive sense that PAP is tacitly applicable in those insinuated circumstances. Thus a silent appeal to PAP may well be used to explicitly dispose of PAP. And that, in other words, would be tantamount to an equivocation on key terms of responsibility in the compared cases.