How Necessary is the Past? Reply to Campbell MATTHEW H. SLATER

How Necessary is the Past? Reply to Campbell MATTHEW H. SLATER e best known argument for Incompatibilism van Inwagen s (1983, 56) third Consequence Argument trades on our ordinary assent to the past s necessity. Recently, Joseph Keim Campbell has challenged this assumption: he argues that the Consequence Argument does not provide us with general reason for accepting Incompatibilism. Unfortunately, I believe that it is of little consequence for the traditional problem about free will (i.e., whether we have any) and more supple forms of Incompatibilism. e past is necessary enough to pose a worry for friends of free will. His approach does, however, suggest an intriguing line of response to free will skeptics. In what sense is the past necessary? Ordinarily, we suppose certain moments of the past to have been open to our influence, though they are now unchangeable. Incompatibilists (about free will) argue for collapsing this distinction in deterministic worlds: past necessities extends deterministically into the future, robbing us (in those worlds) of freedom. e best known argument for Incompatibilism van Inwagen s (1983, 56) third Consequence Argument trades on our ordinary assent to the past s necessity. But should we assent so easily? In Free Will and the Necessity of the Past (Analysis 67.2:105 11), Joseph Keim Campbell offers reason for thinking that the Consequence Argument does not provide us with general reason for accepting Incompatibilism. For the Consequence Argument depends on the thesis that the past is in some sense necessary but its necessity doesn t seem necessary. And if Incompatibilism is a thesis not just about this world, but about any possible world, the Consequence Argument does not establish it. is is an interesting development. Unfortunately, I believe that it is of little consequence for the traditional problem about free will (i.e., whether we have any) and more supple forms of Incompatibilism. But while not a clear victory for Compatibilism or human freedom, Campbell s argument does suggest that a shift in how we approach these arguments may be in order. DEJA VU ALL OVER AGAIN Campbell considers the following formulation of the Consequence Argument (which scarcely needs review): Take some arbitrarily chosen proposition P that s supposedly subject to my will: determinism implies that, necessarily, the laws of nature (L) together with some proposition about the past (P 0): (1) ((P 0 & L) P ), or equivalently: (2) (P 0 (L P)). Now, clearly, if these are necessary truths, then I m not responsible for them. Van Inwagen proposes that in general: (α) P NP, where N is a modal operator such that NP means P and no one has, or ever had, any choice about whether P (van Inwagen 1983, 93). So we infer from (2) and (α): (3) N (P 0 (L P)). Now very plausibly, if we choose our P 0 right say, a proposition describing the state of the universe long before humans existed it follows that no one has, or ever had, any choice about whether P 0: 1

(4) NP 0 ings start rolling from here with the help of rule (β): NP, N (P Q) NQ (5) N (L P) from (3), (4) by (β) (6) NL premise (the laws are not subject to human will) (7) NP from (5), (6) by (β) It turns out that on the assumption of determinism, no one has, or ever had, any choice about whether P a proposition whose truth was supposedly subject to my control. CAMPBELL S ARGUMENT Campbell focuses on the fourth premise ( NP 0 ) regarding the necessity of the past. What justifies it? Campbell identifies and rejects two lines of support: (a) at no one can change the past (and P 0 is a proposition concerning the past); and (b) at NP 0 follows from the remoteness of the past P 0 describes (2007: 106). Let s take (a) the pastness justification of NP 0 first. Campbell rightly notes that the fact that no one can change the past is irrelevant to whether anyone ever had a choice about whether some true proposition about the past is true (107). While I agree with this, I don t find his reasons convincing. He asks us to consider the following example: Drunk Driver: Smith begins drinking at t 1 and a few hours later he gets in his car and drives home down a dark road in the middle of the night. At t 2 he is driving well over the speed limit and he is so drunk that both his vision and his reaction time are greatly impaired. A moment later, at t 3 he strikes and kills Jones who is walking in a crosswalk. (ibid.) Campbell claims that while Smith might be correct that he had no choice at t 2 whether he killed Jones at t 3 but this does not mean that he had no choice about whether he killed Jones at t 3 (107). For Smith could have decided not to drink that night or, once drunk, not to drive (or perhaps also whether to drive on that very road, &c.). Several strands of thought need untangling here. For one, we should set aside at once the question of Smith s responsibility for Jones death. Plausibly, I can be held responsible for events that are strictly out of my control if they resulted from my negligence (if I should have known that the event would follow from others that were in my control). For two, Smith s having a choice at t 3 about whether he killed Jones may well differ from his having a choice (punkt) about whether he killed Jones (at a certain time, t 3). In other words, we might identify a scope ambiguity in Campbell s use of the choiceless operator. Consider the difference between these claims: and (wide) N(Smith kills Jones @ t 3) (narrow) N(Smith kills Jones) @ t 3. Intuitively, the wide scope reading encourages us to view the action (Smith s killing Jones at some time) in the context of a larger set of events (his freely choosing to drink at t 1, for example). e narrow scope reading (appropriately) narrows the scope of our consideration to the events occurring at t 3. 1 1 is distinction seems to me completely uncontroversial for cases of responsibility: we may wish to distinguish between an agent s responsibility for a series of events some of which the agent is not responsible for. So perhaps it s right to say of Smith at t3 that he s not responsible for killing Jones (in the narrow sense), but that he is responsible in the wide sense for the fact that Jones is killed 2

To see the difference, let s change the example a bit. Imagine that instead of ale, Smith knowingly and willingly orders a specialty mind-control cocktail that renders his actions for the next twelve hours totally out of his control (never mind whose control they are under). Curiously, this tonic leaves its imbibers conscious of their actions they merely become fleshy puppets. Now it seems plausible that at t 3, when Smith finds himself hurtling down the road without apparent care for his or others safety, Smith has no choice about whether he kills Jones. Indeed, he feels great regret for this very fact, wishing that he had never foolishly imbibed that mind control cocktail (what was I thinking!?, he asks himself). As Smith bears down on Jones, he struggles to shake the effects of the cocktail and apply the breaks, but THUD! to no avail. Smith cannot control what happens at t 3 because he cannot undrink the cocktail the damage is done, his control is lost. But as Campbell points out, Smith can hardly use the fact that he is at t 3 unfree to undrink the cocktail as defense for his actions at t 3 for they followed from earlier (by hypothesis) free actions. In that weak and wide sense, perhaps, we can reckon Smith as freely killing Jones at t 3 though, in the narrow sense, he is unfree at t 3 (just then). Nevertheless, one could reasonably resist granting Smith even wide freedom regarding the events at t 3. And clearly, Compatibilists can share this intuition. e example is red herring. It seems to me that the point Campbell really wants to make about the pastness defense of the N operator is that it mixes up tenses. e past is necessary in the sense that I cannot now change the truth-values of (genuine 2 ) past propositions. But the fact that I now cannot change the truth-value of propositions regarding some past moment does not imply that the proposition s truth was never in my power. Part of the trouble stems from van Inwagen s chosen interpretation of the N operator: that no one has, or ever had, any choice about whether the operated-upon proposition is true. We need no philosophical tale to show us that Np is false (or question-begging) concerning past propositions ordinarily reckoned as having been in my control. Does anyone have, or has anyone ever had, a choice about whether, say, Smith had scotch for breakfast this morning (a single-minded-malt, perhaps)? Well, though he now lacks such a choice, we can certainly regard him as having had a choice. And if his having scotch set into motion events that ineluctably led to the death of Jones, he may be held responsible without it being held that he had a choice concerning these subsequent events. e Incompatibilist merely needs to interpret Np differently for example, as expressing the unfreedom of a particular subject to affect p now to regain a plausible sense of past necessity. If my present actions follow deterministically from past events now beyond my control, then we are inclined to see those present actions as unfree irrespective of my once having them in my control. e response that Campbell clearly sees simply shifts the choice of the P 0 in premise four (that NP 0) back further into the past so that the truth of NP 0 owes to the non-existence of humans at that time. He writes: Consequently, it is not the pastness of our past that gives it its necessity.... Rather it is the remoteness of some of our past the fact that our past continues back to a time prior to our existence (6 7). Campbell thus employs a special sense of remote in which an event is remote (to S) if it occurs at a time prior to S s existence. But though I seem to have a remote past in this sense indeed, though it seems that every creature had one there might have been possible beings which lacked a remote past. ese philosophical Adams would seem to be immune to at t3. Perhaps when assigning responsibility we tend to think expansively (though, pretty clearly, not as expansively as we could). It s less clear to me whether the analogous scope distinction doesn t immediately collapse on van Inwagen s interpretation of the choiceless operator. 2 We must take care to distinguish between sentences genuinely about the past (made true or false by past states of aff airs) and those merely "written" in the past tense (e.g., 'It was true ten thousand years ago that there was going to be a sea-battle tomorrow') (see Kenny 2004: 134; Anscombe 1956). 3

van Inwagen s ird Argument. For there would be no moment at which Compatibilist and Incompatibilist alike could agree that events were then out of this person s control. ONE FREE CHOICE: USE WISELY! So what?! the incompatibilist might demand. At most, what you ve shown is that there are possible worlds in which determinism is true and yet there is some choice (perhaps just one). But this, urges Campbell, is precisely the denial of Incompatibilism. e conclusion is sociologically startling: neither of the two most popular versions of the argument van Inwagen s First and ird Arguments offer general reasons for endorsing Incompatibilism (110; emphasis added). We discover instead that the incompatibility of freedom and determinism follows only when some contingent facts are assumed. Incompatibilism isn t necessarily true (if true at all). Of course, this will seem like rather cold comfort for those of us wondering whether we have any freedom. Is it any comfort? First, does Campbell s argument indeed show that van Inwagen s Consequence Arguments offer no general reasons for endorsing Incompatibilism? at depends on how general we regard the Incompatibilist thesis. Is it a thesis about our world or about a greater number of worlds (perhaps all of them)? Campbell mentions Ted Warfield s assertion that the conclusion that the claim If determinism is true, there is no freedom is strictly weaker... than the proper incompatibilist conclusion that necessarily, if determinism is true, there is no freedom (Warfield 2000: 170). But we might fairly wonder why this is the Incompatibilist s proper conclusion. Indeed, being an a certain ist plausibly requires a corresponding ism does Warfield just beg the question in claiming that an Incompatibilist (read: a real Incompatibilist ) should go beyond the weak, mere truth of Incompatibilism? One might even be tempted (uncharitably, I think) to compare this assertion to a favored move of creationists (old and new): make your opponent a straw man by saddling them with an impossibly strong theoretical burden (point out that evolution has not been proved, for example). Clearly such a charge cannot be made against Warfield, as he offers an argument for the stronger Incompatibilist conclusion. But the question remains: why think that the strong, necessitarian reading of Incompatibilism is proper? Perhaps van Inwagen merely misstated the conclusion of the Consequence Argument: it shows that were determinism true in this world, there would be no freedom. 3 And yet weaker Incompatibilist theses are possible: that only one person has freedom in deterministic worlds (or that each person has but one free choice to make, and so on). In any case, if we do regard the necessitarian reading as the right one, van Inwagen s argument is the wrong one to target as it has no pretensions of establishing this thesis. Second, consider the use to which Campbell puts his philosophical Adams. We cannot establish premise four by the (b) justification (that NP 0 because P 0 is a true proposition about the remote past ) if not everyone has a remote past. Campbell claims that the Consequence Argument is unsound at worlds where there are conscious, rational beings at every instant. But it s not plain to me that the fourth premise is false in such worlds. Perhaps we cannot use the consequence argument to establish it without begging the question. But that s not surprising we clearly cannot expect arguments to establish their own premises. Campbell is clearly correct that the (b) justification will not work in every world, but again: have we any reason to think that the fourth premise is false at those worlds? How would these worlds look? Let s say that time can be either bounded or unbounded; temporally unbounded worlds can be finite if time is circular. Temporally circular worlds seem of little help here. Say we pick up the action at t 3 when Smith (one 3 Assuming that the assumption of determinism (counterfactually or not) has no relevant bearing on our past. 4

of Campbell s philosophical Adams ), sloshed on his single-minded-malt, hits Jones. Was this action free? I suggested that it is not (in the narrow sense). Whether it is in some wider sense depends on the freedom of its causal antecedents: are any of them free in the narrow sense? But we can see that these antecedents must eventually loop back to this very act. Only by maintaining, quite implausibly, that Adam was free at t 3 can we escape the loop. Nor does anything much turn on the strangeness of Adam s situation. 4 e existence of rational agents at every moment of time (at least in this case) does not seem to establish the falsity of NP 0 Suppose Adam s world is like ours 5 and has a beginning a first moment of time. Note that not only must Adam exist at this first instant; he must make a choice (P 0) his one free choice. How ironic and disappointing it may seem then that he has apparently no reason to choose one way or another. Is it then even correct to call this kind of ur-choice free? Is it clear, in other words, that NP 0 is false? It is by hypothesis undetermined; but this is another matter. Some may regard Adam s lack of reasons as precluding his freedom. But perhaps that s too quick. 6 Anyway, Adam s one free choice seems more imperiled by an temporally-evenhanded view of determinism: we needn t view it, for example, as a temporally-asymmetrical thesis. Temporally-earlier events are as much determined by later as earlier determine later. In general, any total state of the world at a time (if such things there be) determines all of the others. Now if we suppose that unfreedom transfers, Adam has a problem. Considering the deterministic world of circular time, we found it impossible to locate one moment that was undetermined, as no moment is prior to all others. Any candidate for Adam s one free action gets undermined by it s being determined by other events even if those events follow from this very action. If we construe determinism as a temporally-symmetrical thesis in a world with a beginning, we must again (apparently) grant that no moment of time is undetermined, notwithstanding one moment s being prior to others. If we cannot make sense of Adam s purportedly one free action in the world of circular time, it looks hard to make sense of it in the world of linear time (construing determinism in the manner I ve suggested). Even if there s nothing wrong with the above reasoning (I m certainly not sure there isn t), perhaps there s something strange about viewing a free action as undermining its own freedom via a sort of deterministic feedback. For would there not be a sense in which such events (though they follow deterministically from later events) are self-determined insofar as the later events were determined by the first one? Why regard selfdetermined actions as unfree? I confess to being somewhat torn on these questions no doubt partly in virtue of my arguing from Incompatibilist intuitions I don t in fact share. In any case, it hardly seems obvious that the fourth premise is false (and thus the Consequence Argument unsound) even in worlds where the proffered justification for the fourth premise fails. CONSEQUENCES So what?! Even supposing we grant Adam his one free choice, how does bear on our conception of our freedom? Well, in a way, it doesn t and perhaps we never should have expected it to. e debates about Incompatibilism 4 I don t think I m trying to pull a fast one on anyone. e single-minded-malt merely colorfully illustrates the fact of determinism. 5 Let's speak naively here: perhaps it's not sensible to speak of a "first moment" of time. Since I'm out of my depth here, I'll stay silent. 6 Perhaps the world came to be, Bertrand Russell-style, complete with an extensive apparent past. Adam remembers waking up, rubbing his eyes, and, sauntering into the kitchen for breakfast, decides to have eggs because though he s in the mood for cereal he fears that the milk has gone bad. Now in fact, Adam has never been asleep nor has he stepped foot outside his kitchen: for the world came into existence at the precise moment when he decided to have eggs rather than risk cereal. Adam s reasons in this case are purely psychological. Perhaps this is all one needs for freedom in a deterministic world: as artificial light is light illusory reasons are reasons. 5

have at best an incidental connection with the Traditional Problem of free will (of whether we are free)? For as Incompatibilism does not bear on the truth of determinism, it clearly leaves this issue alone. Still, the Traditional Problem (of whether we are free) is hardly indifferent to the truth or falsity of Incompatibilism. If my freedom depends on the truth of some relatively obscure philosophical thesis (determinism) about which I suspend belief, should I not also suspend belief about whether I am free? Incompatibilism tends to engender skepticism about our freedom and this can be almost as damaging to our conception of ourselves as the discovery that we in fact are unfree. e Compatibilist denies this move: Freedom and Determinism are in fact compatible. Freedom skepticism is unwarranted. One worries that Campbell s argument does not aide the Compatibilist in offering this comfort in our world. 7 I suspect that the best strategy for defusing freedom skepticism lies in attacking not the soundness of the Consequence Argument, but it s validity. 8 Still, Compatibilists might seek progress on multiple fronts. Perhaps Campbell s point can help defang freedom-skepticism. If for all I know, the world popped into existence just now with merely apparent deterministic causal antecedents, then for all I know, my present choice is free (determinism notwithstanding). More would, of course, have to be said to address the concerns I raised above. (Maybe the best that could be said for my subsequent choices is that they exhibit wide freedom. Perhaps the skeptical trade-off isn t worth making.) In the meantime, however, I for one will enjoy the prospect of using one pesky skeptical scenario to defuse another. 9 References Anscombe, G. E. M. 1954. Aristotle and the sea-battle. In From Parmenides to Wittgenstein. Oxford: Blackwell, 1981. Campbell, J. 2007. Free will and the necessity of the past. Analysis 67.2: 105 11. Kenny, A. 2004. Ancient Philosophy. Oxford: Oxford University Press. Author s paper. van Inwagen, P. 1983. An Essay on Free Will. Oxford: Clarendon Press. Warfield, T. 2000. Causal determinism and human freedom are incompatible: a new argument for incompatibilism. Philosophical Perspectives 14: 167 80. 7 Consider a schematic analogy. Suppose a Cartesian skeptic proclaims that it follows from a certain eminently plausible account of belief-acquisition that no one has any knowledge about the external world. Inroads are made when it is discovered that there are certain logically-possible conditions in which these belief-forming-mechanisms do not succumb to the skeptical argument. Unfortunately, while logically possible, such conditions do not seem to obtain in this world (nor would they obtain very widely in the worlds in which they are true). So again, they have very little bearing on my knowledge. Indeed, the extent to which this style of retort has any bearing on its true target obviously depends on what that target is. 8 As I noted in my (2005), while the crucial transfer principle (β) NP, N(P Q) NQ is a theorem of "normal" modal systems (unlike more contentious theorems, it follows simply from the possible worlds modeling), there's independent reason for thinking that normal modal systems are too strong for deontic modalities. For example, deontic systems which construe P as obligatorily P might assert ' ( P & P)'. at's controversial, though: plausibly, moral dilemmas are cases where I am obligated to both P and not-p. It's much less credible that one is ever obligated to a contradiction. But normal models have as a theorem ( P & P)', wrongly collapsing the distinction. Hence normal models are inappropriate modal systems for deontic logics. is suggests precedent, at the very least, for rejecting (β). In any case, it illustrates that we cannot construe (β) as a principle of logic. 9 anks to Joseph Campbell, Chris Haufe, Michael O Rourke, Mitch Stokes, and Achille Varzi for comments on previous drafts of this essay. 6