Mother, mother, have pity on your sick child! And do you know that the Bey of Algiers has a wart under his nose?
– Nikolai Gogol
That A acts freely only if he could act otherwise has become the default position in a field not known for its strong competition. Faute de mieux. There is nothing better, Professor Robert Sapolsky is persuaded, than nothing better. “Acting on something and knowing you could have done otherwise is often necessary and sufficient to decide that free will has just happened.”1 If Sapolsky is prepared to defer to the default, it is chiefly to show that it is wrong. What has just happened has just not happened. Free will is an illusion. In reaching this conclusion, Sapolsky is not alone in thinking that he is all alone. The rejection of free will is a community endeavor.2 What is common to members of the Sapolsky clan, rather an Albanian designation, now that I think about it, is their commitment to arguments of the locked-and-loaded variety, in which what is loaded is intended to provide a devastating rebuttal to what is locked. “Do I pull the trigger and kill this person or let them live?” 3 The load now follows. “How did you turn out to be the sort of person who would tend to do that at that moment?” The Albanian default notwithstanding, there are circumstances in which the reverse seems more nearly true. “Hier stehe ich, ich kann nicht anders,” Martin Luther remarked at the Diet of Worms. “Gott helfe mir,” he added soberly. These are dramatic words, at odds with everything that has come before. An agent is most free when he is least able to do otherwise. Vos ordres sont mes souhaits, as a courtier might murmur. They are also words commonly connected to the most familiar of circumstances, the physiological commingling of urgency and ecstasy. Seized by passion, Nabokov’s Humbert Humbert rejoices in “the full consciousness of his freedom.” Yes, his freedom.
1 If the Albanian default is divided in the allegiances that it provokes, this is all the more reason to take a closer look at a closed book. There is first the de facto default: 1.1A does x freely → A could do otherwise.4 In quick succession, 1.2A does x freely → A does x and 1.3A does x → A could do x now follow, not as logical inferences – nothing of the sort – but as unavoidable ancillaries, their justification a matter of the incredulity their denial would prompt. There remains the matter of otherwise, which appears clinging limpet-like to the right and rear of 1.1. What is otherwise is plainly other than x: 1.4A could do other than x → A could do y ∧ y ≠ x, purification extending both fore and aft. If 1.4 is a step in the right direction, it is a step that stops too soon. Doing something other than x cannot mean doing anything other than x. A is not acting freely in March in virtue of what he could do in November. What is needed is some y undertaken instead of x: 1.5A could do other than x → A could do y instead of x ∧ y ≠ x. Inasmuch as A could do y, what A could do must coincide temporally with what A is about to do. Where A approaches the temporal sticking point – he is about to pull Sapolsky’s tempting trigger – things must switch around so that A could begin to do otherwise: 1.6A could do other than x at t → A could do y at t ∧ y ≠ x. If x and y are now coordinated in time, it is A himself who coordinates them at a point P in space. It hardly hurts to make this plain. One can never be too fussy about these details. 1.7A could do other than x at t and P → A could do y at t and P ∧ y ≠ x. Nothing better.
2 The Albanian default does what defaults do: it states the obvious. In doing what it does, it remains immured in the English modal system Σ by virtue of its appeal to could as a condition.5 On getting into Σ philosophers have always found it difficult getting out. Could A do otherwise? Under an interpretation that he assigns both to G.E. Moore and A.J. Ayer, Christopher List remarks that “if the agent were to try (or choose) to do otherwise, he or she would succeed in doing otherwise,” [emphasis added].6 And vice versa, I suppose. In both English and French, conditional modals very often undergo reification, emerging in the case of could or pouvoir as an ability, a capacity, or a power: 2.1A could do x → A has the capacity to do x.7 Having gained a capacity at 2.1, A is in danger of losing it at 2.2A has the capacity to do x → A could do x, the ensuing back and forth characteristic of commerce within Σ. Improvements lie elsewhere. In the capacity to do x, 2.1 introduces a definite description and with the description, a referential connection to the real world. That A could do x offers the spectator nothing beyond the fact that A could do x: that A has the capacity to do x locates couldness in things beyond facts. Questions that at 1.4 remain cloaked now become clear. 2.3(The capacity to do x = The capacity to do y) → x = y ? Or is it the other way around? Given the excitement the question provokes, it is easy to become confused.
3 If A does x freely, it follows from 1.2 and 1.3 that 3.1A could do x, and that 3.2A could do y. With could now firing in two directions, an apparent dispersion of authority is in prospect. The dispersion is real enough but confinable, as nuclear technicians so often say. 3.3[(P → Q) ∧ (P → R)] → [P → (Q ∧ R)] is a tautology and by modus ponens on 3.3 3.4A does x freely → A could do x ∧ A could do y. What yet remains odd about 3.4 is the unavoidable asymmetry between what A could do in doing x, and what presumptively, he could do in doing y. In doing x freely, A topples into action. But if 3.5A could do y ∧ y ≠ x follows from 1.1, not so 3.6A does y. Doing y is nothing that A does. And no wonder. He is busy doing x. If 2.1 assigns A the capacity to do y, the capacity remains real even as its object hovers ectoplasmically, like Ra, lost in the spirit world.
4 The Albanian default encompasses events as a matter of course and capacities as a matter of choice. Actions are inescapable, capacities, not so much. Whatever the connection between them, y must fit the vacancy yielded by x; and the fit cannot be cut so finely as to amount to an identity. This is, after all, the meaning of 1.4. The question whether y might fit the facts while remaining true to itself turns on when and whether two events are the same. Always if they are identical and never if they are not is one answer. And correct in its severity. What is now wanted is some series of conditional steps showing that if two events are the same in some respects, they are the same in all of them. The demand has not provoked an embarrassment of riches. Two events are the same, Donald Davidson once argued, if and only if they have the same causes and the same effects.8 4.1e1 = e2 ↔ ∀z (z caused e1 ↔ z caused e2) ∧ ∀z (e1 caused z ↔ e2 caused z). And the reverse, when events fail to coincide: 4.2e1 ≠ e2 ↔ ~ [∀z (z caused e1 ↔ z caused e2) ∧ ∀z (e1 caused z ↔ e2 caused z)]. It hardly matters argumentatively, where things go wrong. Suppose, then, that 4.3∃z (z caused e1) ∧ ~ (z caused e2). Does every event have a cause? What a question! It follows that 4.4∃w (w caused e2), and from 4.2 and 4.3 4.5z ≠ w follows, too. But e2 is simply a stand-in for the y of 1.4: 4.6x ≠ y ↔ e1 ≠ e2, these notational comforts serving both to preserve a link to Davidson’s notation and to accommodate the forthcoming proliferation of events. It follows from modus ponens on 1.4 and from 4.4 and 4.6 that 4.7(A could do y ∧ y ≠ x) ↔ ∃w (w caused y). But 4.7 affirms only that if causes had been different, so, too, their effects; and this is not a proposition calculated to elicit a firestorm of controversy. 4.7 reduces 1.1 to an otherwise trite immersion in the world’s causal order. Had things been different things would have been different. Another suggestion, this one of even longer-standing, confines x and y to the common coffin of identity if they coincide in the same spatiotemporal region T. Same place, same time, same all around. Whereupon, the long-standing suggestion undertakes its own reincarnation in 4.8x = y ↔ (x occupies T ∧ y occupies T). In its various effects, 4.8 is all sibylline suggestiveness, and slinky, too, but it is about to find itself straddling various lapses. Following Sapolsky’s criminal encouragement, A is determined to act. These peevish academic disputes have a way of getting out of hand. This is well known. The moment in which A hands off his decision to himself is t; and at t, A leaves off being an A decently disposed to do y and with the froth of irritation now crusting on his lips becomes an A about to do x. If y is now seen receding rapidly, it remains the y that A could have done; and by 1.5, it must remain the y that A could have done instead of doing x. Having handed off to himself, A is free, at last, to pull that trigger, the gun bursting into life with a thrilling plop. From 1.1 and its various ball boys down to 1.7, it follows that 4.9x ≠ y. But from the consequent of 1.7 – a conjunction, cela va sans dire – it follows that 4.10A could do y at t and P. The joint occupancy figuring in 4.8 now implies that 4.11x = y, a lapse long lost among lapses.
5 In an essay about mental events, Donald Davidson introduced the assassination of the Archduke Franz Ferdinand as the event that started the First World War. The date of the assassination is known precisely: June 28, 1914; and the time roughly because by 11:30 AM, both the Archduke and his morganatic wife, Sophie, the Duchess of Hohenberg, were dead. If the assassination was the event that started the First World War, the start of the war was itself an event, the connection between events both temporal and in some sense causal. The causal connection is very real, although almost impossible properly to parse.9 The temporal connection remains within reach, an instance of the general scheme in which the end of one event is the beginning of another. Retaining Davidson’s notation once again, there is 5.1∃t(e1 at t = e2 at t), even though in the larger sense in which e1 designates the assassination of Franz Ferdinand and e2, the First World War 5.2e1 ≠ e2. This recalls 1.6, where 5.3∃t(x at t = y at t) even though 5.4x ≠ y. But while both the Archduke and his wife were dead by 11:30 AM, they were shot at roughly 10:50 AM. Given the confusion and horror of the event, the time of the assassination inevitably undergoes a demotion from 10:50 AM to ±10:50 AM. And with the demotion, an undeniable uncertainty about when things began and when they ended.10 It is for this reason that the time becomes a temporal interval, a bloody dot at t expanding to a blood-stained splotch at [t±]. Two sequences of events S1 and S2 are now embedded both in history and its analysis. The assassination proceeds. Short, dimwitted and tubercular, Gavrilo Princip raises his heavy Fabrique Nationale 380 calibre pistol and fires. The bullet flies. The Archduke is struck. And slumps over. Princip fires again. Another bullet flies. The Duchess of Hohenberg is struck: 5.5S1 = s1, s2, …, st. And, thereafter, another sequence of events begins, ominous, dreadful, inexorable. The Archduke begs his wife to live for their children. He remarks to Count Harrach that his wound is nothing. Princip is seized. A crowd gathers. The heavy imperial carriage turns and begins sedately to move toward the Governor’s Mansion. And as Winston Churchill would later write, a strange light begins to play across the map of Europe: 5.6S2 = st, st+1, …, st+n. Whereupon, their union in S: 5.7S = S1 ⋃ S2. And in the interval I[t] centered at t, their intersection: 5.8I[t] = S1 ⋂ S2, even though 5.9S1 ≠ S2. The I[t] of 5.8, it is worth remembering, admits of two quite different descriptions: either as the end of the assassination or the beginning of the First World War. Although different, these descriptions must describe one and the same event. If not, then 5.10~ I[t] ⊂ S1 ∨ ~ I[t] ⊂ S2 by 5.8. If 5.8 defines an interval in which one event begins to fade out even as another event begins to flare up, it is not an interval defined by any property of events more palpable than the plus or minus embedded in I[t] = [t±]. Both the time t and the interval [t±] are arbitrary. One historian might argue that t = 11:30 AM, the time of the Archduke’s death; and another, that t = 10:50 AM, the time that he was shot. Disagreeing over this, they may disagree over I[t] and by expanding (or contracting) the window of opportunity, come to different conclusions about its extent. If a proliferation of times is inevitable, so, too, a proliferation of intervals. And, so too, an expanding sequence of nested intervals 5.11I = I[t] ⊂ I[t+1] ⊂ …⊂ I[t+ k], until in the end, 5.12I = S. But 5.12 is the last thing the historian had hoped to see. By 5.8, it follows that 5.13∃!x(x occupies I). And by 5.2 and 5.4, that 5.14∃x ∃y (x occupies S ∧ y occupies S ∧ x ≠ y). From 5.12 and 5.14, a contradiction ensues: 5.15x = y ∧ x ≠ y, circumstances that should provoke an all-Albanian alarm. The conflict between 5.12, 5.13, and 5.14 may be attributed to the arbitrariness involved in the expanding temporal intervals at 5.11. But when things end, and when they begin – these are arbitrary in life and so in history.
6 This argument ramifies upward into the spirit world when capacities come under consideration. There is no need now to specify identity conditions for capacities. In that way lies madness. No more is needed than an inexorable disjunction. Either 6.1(The capacity to do x = The capacity to do y) → x = y or 6.2~ [(The capacity to do x = The capacity to do y) → x = y]. One or the other. If 6.1 is taken as it stands, then 1.4 must go. Uphold 6.1 at your peril. If 6.2, then in virtue of the tautology 6.3~ (P → Q) ↔ (P ∧ ~ Q), it follows that 6.4The capacity to do x = The capacity to do y even though 6.5x ≠ y. But 2.1, now repeated, affirms that 6.6A could do x → A has the capacity to do x. It would seem that 6.7A does x freely → A has the capacity to do x and 6.8A does x freely → A has the capacity to do y are indistinguishable: true alike, false alike, and so linked in and locked together. It was the capacity to do something other than x that at 1.4 was the very key to grasping that A does x freely. Now it is gone, that key.
7 The conditional modalities figuring in Σ have done what they could do. They have made a mess. It may well be that the mess is of my own making. “We are always the last to learn of evil in our own home,” as Saint Jerome observes. But no matter the author, the mess remains a mess, good reason, one might think, to ask for an analysis more severe than anything contingent on Σ. The conditional modalities in Σ are now reductively dismissed, along with powers and capacities. There remains only an uninflected account of action together with a single alethic modality: 7.1A could do x → ◇ (A does x). Minus its misplaced modals, 1.4 goes over to 7.2A does x freely → A does x ∧ ◇(A does y), with y ≠ x taken for granted, here and in what follows. The correlative case in which A does not do x freely follows from 7.2 by means of 7.3P → (Q ∧ S) ∧ ~ P → ~ Q ∨ ~ S. Whereupon, there is 7.4~ (A does x freely) → ~ (A does x) ∨ ~◇(A does y). If philosophers have repaired to 7.4 from a sense that they could do worse, they are aware, as well, that they could do better. Scruples arise, along with second thoughts, because given 7.4, the road to a modal collapse, while not yet open to traffic, is nonetheless open to doubt. There is, for example, 7.5A does x freely → A does x, which follows from 1.4. What is now sous enquête policière is its converse 7.6A does x → A does x freely. And for every good reason: 7.5 and 7.6 serve to obliterate the distinction between acting and acting freely. Given the Albanian default, it must be possible to remark of A that if sometimes he acts freely, sometimes he does not: 7.7A does x ∧ ~ (A does x freely). Inasmuch as 7.8~ (A does x) ∨ ~◇(A does y) ↔ ~ [(A does x) ∧ ◇(A does y)] exemplifies the tautology 7.9(~P ∨~ Q) ↔ ~ (P ∧ Q) it follows from 7.4, 7.7 and 7.8 that 7.10~ (A does x freely) → ~ [(A does x) ∧ ◇(A does y)]. By 7.8 and contraposition on 7.10, 7.11[~ (A does x) ∨ ~◇(A does y)] → A does x freely, now emerges. And from 7.7, that 7.12~◇(A does y) → A does x freely, as well. By 7.2 7.13A does x freely → ◇(A does y). And by contraposition 7.14~ ◇(A does y ) → ~A does x freely. It follows by modus ponens on 7.12 and 7.14 that 7.15A does x freely ∧ ~A does x freely. From 7.15, everything follows and so anything goes.
8 The doctrine that anything goes so exceeds in its excess that redemptive relief, if often unavailing, is never unwelcome. The identical triplets 8.1∀x ~ (A does x → A does x freely); 8.2∀x ~ [~(A does x) ∨ (A does x freely)]; and 8.3∀x (A does x ∧ ~ (A does x freely) now emerge as prophylactic imperatives. If 8.1 offers 7.1 its protection, as so often happens, it comes at a high price. There is 8.4A does x ∧~◇∃y( A does y ∧ x ≠ y), for example, which follows from 8.3. By an exchange of modal operators in which ◇ is replaced by ~□~, there is also 8.5∀x[(~◇∃y(A does y ∧ x ≠ y)) ↔□~∃y(A does y∧ x ≠ y)] and 8.6∀x[(A does x) ∧ □~∃y(A does y∧ x ≠ y)], as well. But 8.6 initiates a metaphysical doom loop all its own. Given 8.7□~∃y(A does y∧ x ≠ y), A is fated to fall into fatalism. In doing x, but not doing x freely, A must do what he does if he does anything at all. His choices have narrowed to what he is about to do, and that is no choice at all. If the Albanian default at 1.4 succeeds in stating the obvious, it is obvious, as well, why it has become the default. It expresses the instinctive judgment that A acts freely only if he could act otherwise. Whatever the judgment, the default remains defective. The amplitude of the alethic modalities is either too large or too small to do much good; and the conditional modalities of old, although there for the asking, remain too rebarbative for the taking. It hardly follows that free will is an illusion. In embracing this conclusion, the Albanian clan is rather in the position of a biologist persuaded that whales are fish and on learning better concludes that there are no whales. Just look around. There are plenty of whales. And plenty of things that A does freely. He has just shot someone, for heaven’s sake. The news is all over the internet.
