Candidates for the first thinkers to form these views, as well as the idea of a non-physical "agent-causal" libertarianism, include Democritus (460-370), Aristotle (384-322), Epicurus (341-270), Chrysippus (280-207), and Carneades (214-129).

After a brief review of the history, we will also look at the arguments of modern classicists and historians of philosophy who have scrutinized the textual evidence for each of these philosophers.

These modern thinkers include Carlo Giussani, Cyril Bailey, Carlo Giussani, David Furley, Pamela Huby, A. A. Long and D. N. Sedley, Don Fowler, Julia Annas, Jeffrey S. Purinton, Susanne Bobzien, Tim O'Keefe, and Ricardo Salles.

All of these modern analyses make implicit or explicit comparisons to modern ideas of determinism and libertarianism. Since these ideas may be flawed, we need to identify and separate the original problems from their modern counterparts. And we need to separate the distinct ideas of logical necessity, physical determinism, fatalism, future contingency (the principle of bivalence), and foreknowledge that are often scrambled together in the work of the ancient thinkers.

Before there was anything called philosophy, religious accounts of man's fate explored the degree of human freedom permitted by superhuman gods. Creation myths often end in adventures of the first humans making choices and being held responsible. But a strong fatalism is present in those tales that foretell the future, based on the idea that the gods have foreknowledge of future events. Anxious not to annoy the gods, the myth-makers rarely challenged the implausible view that the gods' foreknowledge is compatible with human freedom. This was an early form of today's compatibilism, the idea that causal determinism and logical necessity are compatible with free will.

The complete conception of the will according to Epicurus comprises two elements, a complex atomic movement which has the characteristic of spontaneity, that is, is withdrawn from the necessity of mechanical causation: and then the sensus, or self-consciousness in virtue of which the will, illuminated by previous movements of sensation, thought, and emotion, profits by the peculiar liberty or spontaneity of the atomic motions, to direct or not to direct these in a direction seen or selected. (Cyril Bailey translation)

It is a commonplace to state that Epicurus, like his follower Lucretius, intended primarily to combat the 'myths' of the orthodox religion, to show by his demonstration of the unfailing laws of nature the falseness of the old notions of the arbitrary action of the gods and so to relieve humanity from the terrors of superstition. But it is sometimes forgotten that Epicurus viewed with almost greater horror the conception of irresistible 'destiny' or 'necessity', which is the logical outcome of the notion of natural law pressed to its conclusion. This conclusion had been accepted in its fulness by Democritus, but Epicurus conspicuously broke away from him: 'it were better to follow the myths about the gods than to become a slave to the "destiny" of the natural philosophers: for the former suggests a hope of placating the gods by worship, whereas the latter involves a necessity which knows no placation'. Diogenes of Oenoanda brings out the close connexion with moral teaching: 'if destiny be believed in, then all advice and rebuke is annihilated'. If any ethical system is to be effective it must postulate the freedom of the will. If in the sphere of human action too 'destiny' is master, if every action is the direct and inevitable outcome of all preceding conditions and man's belief in his own freedom of choice is a mere delusion, then a moral system is useless: it is futile to tell a man what he ought or ought not to do, if he is not at liberty to do it. Here at all events 'destiny' must be eliminated. It is a more fatal enemy than superstition, for it means complete paralysis: spontaneity — voluntas — must be at all costs maintained.

But why, in order to secure this very remote object, should a protest against 'inexorable necessity' be made at this point in the physical system? It would have been easy, one might think, to accomplish the immediate purpose of securing the meeting of the atoms in their fall through space by some device, such as the Stoic notion that all things tend to the centre,' which should not be a breach of the fundamental law of causality, instead of this sporadic spontaneous deviation. And in what sense can this 'swerve' be said to be vital for the freedom of the will, with which Lucretius so emphatically connects it? The answer must be looked for in the very material notions of Epicurus' psychology, which may be briefly anticipated here. The mind (νοῦς) is a concentration in the breast of an aggregate of very fine atoms, the same in character as those which, distributed all over the body and intermingling with the body atoms, form the vital principle (ψυχή). This aggregation of atoms may be set in motion by images, whether coming directly from external things or stored up as an 'anticipation' (πρόληχις) in the mind itself. Suppose, for instance, that in this way there comes before my mind the image of myself walking: ultimately the atoms of the mind being themselves stirred, will set in motion the atoms of the vital principle: they in turn will stir the atoms of body, the limbs will be moved and I shall walk. But before this can happen another process must take place, the process of volitional choice.

We can choose to do otherwise

When the image is presented to the mind it does not of itself immediately and inevitably start the chain of motions which results in the physical movement; I can at will either accept or reject the idea which it suggests, I can decide either to walk or not to walk. This is a matter of universal experience and it must I not be denied or rejected.

Bailey identifies one swerve with volition

But how is this process of choice to be explained on purely material lines? It is due, said Epicurus, to the spontaneous swerving of the atoms: the act of volition is neither more nor less than the 'swerve' of the fine atoms which compose the mind. The fortuitous indeterminate movement of the individual atoms in the void 'is in the conscious complex (concilium) of the mind transformed into an act of deliberate will. The vital connexion, indeed the identity of the two processes is clearly brought out by Lucretius at the close of his exposition of the theory: 'but that the very mind feels not some necessity within in doing all things, and is not constrained like a conquered thing to bear and suffer, this is brought about by the tiny swerve of the first-beginnings in no determined direction of place and at no determined time'. It is not merely, as has been suggested, that Epicurus decided to get over two difficult problems in his system economically by adopting a single solution, but that he perceived an essential connexion between them: if freedom is to be preserved, it must be asserted at the very basis of the physical world.

The 'swerve' of the atoms is, no doubt, as the critics have always pointed out, a breach of the fundamental laws of cause and effect, for it is the assertion of a force for which no cause can be given and no explanation offered. For if it be said that the atom swerves because it is its nature to do so, that is merely to put 'nature' as a deus ex machina on a level with 'necessity' as it was conceived by some of the early physicists, a force which came in to do what could not otherwise be explained. But it was no slip or oversight on Epicurus' part which a more careful consideration of his principles might have rectified. On the contrary it was a very deliberate breach in the creed of 'necessity' and is in a sense the hinge on which the whole of his system turns. He wished to secure 'freedom' as an occasional breach of 'natural law'. If criticism is to be brought against him, it must not be on the technical ground of inconsistency in this detail, but on the broader ground that in his system as a whole he was attempting the impossible. To escape from the old notion of the divine guidance of the world, the Atomists had set up a materialist philosophy directed solely by uniform laws of cause and effect. Democritus saw that this, if pursued to its logical conclusion, must lead to an unflinching determinism, which with more scientific insight perhaps, but less care for his ethical precepts, he had wholly accepted. Epicurus, unwilling in this way to risk his moral system, tried to escape from the impasse without abandoning a materialist position.

Bailey says some metaphysical agency is necessary to explain freedom

Such a compromise is in reality impossible: a wholly materialist view of the world, which excludes altogether the spiritual and the supernatural, must lead to determinism, and there is no real path of escape, except in the acknowledgement of other than material conditions and causes. From the point of view of ultimate consistency, the 'swerve' is a flaw in Epicureanism, but it is not to be treated as a petty expedient to get over a temporary difficulty, or an unintelligent mistake which betrays the superficial thinker.

It may not be uninteresting to notice that a parallel difficulty arises for modern thinkers and that a solution not unlike that of Epicurus' atomic swerve has sometimes been propounded.
(Bailey, The Greek Atomists and Epicurus, pp. 318-321)

In a single atom the swerve is merely random chance

Of what nature then is this self-initiated movement? In the individual atom it is automatic, spontaneous, and wholly undetermined in occasion or direction. Is the movement of the mind in will merely the result of such a movement in one of its component atoms, or even the sum of many such movements? If so it too must be automatic and undetermined. When the image of action is presented to the mind, it is impossible to foretell in what way the movement will occur, or even whether it will occur at all. In other words the mind is not really self-determined, but is at the mercy of wholly undetermined movements inside itself, and freewill after all its careful preservation turns out to be nothing better than chance. This is indeed the conclusion reached by one modern critic, and it is not to be wondered at that he is unwilling to believe that Epicurus himself can have rested the claim for freewill on the atomic 'swerve'. But the solution of this difficulty lies once again to the Epicurean conception of a compound body (concilium, conciliatus).

In a large number of atoms compounded as a "mind," the swerve of many atoms becomes the free volition of an undetermined consciousness.

The compound is more than a mere aggregate of independent atoms: it is their union in a complex, which has a new individuality of its own in which it may acquire qualities and even powers which are not possessed by the individual component atoms. The soul or mind is a compound body of such peculiar constitution in the nature of its component atoms and their motions among themselves, that it acquires the power of sensation or consciousness. The automatic swerve of the individual atoms then is translated in the complex of the mind into a consciously spontaneous movement, in other words into a movement of volition.

Giussani's two elements look like a temporal sequence - free spontaneous thoughts illuminate the subsequent decision of the will to act

'The complete conception of the will according to Epicurus, Giussani argues in an admirable summary of his position, 'comprises two elements, a complex atomic movement which has the characteristic of spontaneity, that is, is withdrawn from the necessity of mechanical causation: and then the sensus, or self-consciousness in virtue of which the will, illuminated by previous movements of sensation, thought, and emotion, profits by the peculiar liberty or spontaneity of the atomic motions, to direct or not to direct these in a direction seen or selected.' In other words the blind primitive 'swerve' of the atom has become the conscious psychic act. It may be that this account presses the Epicurean doctrine slightly beyond the point to which the master had thought it out for himself, but it is a direct deduction from undoubted Epicurean conceptions and is a satisfactory explanation of what Epicurus meant:

Epicurus did not identify freedom of the will with chance

that he should have thought that the freedom of the will was chance, and fought hard to maintain it as chance and no more, is inconceivable.

And if the further question is asked how can a complex of blind spontaneous movements of atoms become the conscious act of volition of the mind, we are only thrown back once more on the ultimate difficulty, which has made itself felt all through this account of the soul. For indeed, if we look back over it, we find that here and there crudities of thought or incoherences in the connexion of ideas have been noted, yet as a whole the general theory is self-consistent and complete; but at the back of it always lies the difficulty which must beset Epicureanism or any other form of materialism: can the movement of insensible particles produce or account for consciousness? That all forms of consciousness have their physical counterpart, that sensation, thought, will are accompanied by material movements of parts of the physical organism is credible, and indeed scientific investigation seems to be revealing this parallelism more and more clearly to us. The more material thinkers of our own time are content to say that consciousness 'supervenes' as an 'epiphenomenon' on the movements of matter: Epicurus went the step farther and was prepared to say that consciousness, sensation, thought, and will are the movements of the soul-atoms. Such an idea is to most modern minds, as it was to the majority of philosophers in Epicurus' day, unthinkable: between the one set of facts and the other there is a great gulf fixed: nothing can bridge the gulf that lies between the most elementary sensation and the atomic vibrations which accompany and condition it. If we accept a purely materialistic system in any form, its conclusions will have to be mutatis mutandis something like those of Epicurus: but he has done nothing to bridge over the abyss or to make the gulf seem less wide. Consequitur sensus, inde voluntas fit, his pupil says glibly, but each time rouses in us the same feeling that this is just what can never be understood.

And if it is impossible to accept his account of the nature of the soul and its workings, so the inference from it cannot be admitted. If the soul is a mere atomic complex, a 'body', then no doubt like the body it perishes and cannot have any sort of existence after death. But if that account be unsatisfactory, then the problem of survival remains open: the soul may or may not survive bodily death, but the question cannot be decided on the basis of a purely material analysis.

It is impossible in dealing with a material system to refrain from pointing out its fundamental weakness, but in an attempt to estimate Epicurus as a thinker, it is less profitable to quarrel with his base-principles than to think of the superstructure he has built upon them. And once again in examining the account of the soul, for all its weaknesses, we are conscious of the workings of a great mind, capable of grasping alike broad ideas and minute details of elaboration. We are certainly not left with the picture of a moral teacher, who merely patched together any kind of physics and metaphysics to back up his ethical preaching.
(Bailey, The Greek Atomists and Epicurus, pp. 435-37)

If we now put together the introduction to Lucretius' passage on voluntas and Aristotle's theory of the voluntary, we can see how the swerve of atoms was supposed to do its work. Aristotle's criterion of the voluntary was a negative one: the source of the voluntary action is in the agent himself, in the sense that it cannot be traced back beyond or outside the agent himself. Lucretius says that voluntas must be saved from a succession of causes which can be traced back to infinity. All he needs to satisfy the Aristotelian criterion is a break in the succession of causes, so that the source of an action cannot be traced back to something occurring before the birth of the agent. A single swerve of a single atom in the individual's psyche would be enough for this purpose, if all actions are to be referred to the whole of the psyche.

Multiple random events can average out to produce an adequate determinism

But there is no evidence about the number of swerves. One would be enough, and there must not be so many that the psyche exhibits no order at all; between these limits any number would satisfy the requirements of the theory.

The swerve, then, plays a purely negative part in Epicurean psychology. It saves voluntas from necessity, as Lucretius says it does, but it does not feature in every act of voluntas. There is no need to scrutinize the psychology of a voluntary action to find an uncaused or spontaneous element in it. The peculiar vulnerability of Epicurean freedom — that it seemed to fit random actions, rather than deliberate and purposive ones — is a myth, if this explanation is correct.

We can now understand why the swerve gets no mention in Lucretius' account of voluntary action. It gets no mention because it plays no direct part in it. The theory of the swerve asserts merely that our actions are not caused conjointly by the environment and our parentage. There was no need for Lucretius to mention this in his account of the psychology of action, any more than there was for Aristotle to insist on his negative criterion of the voluntary in De Motu Animalium.

It may be objected that a swerve in the psyche must have been supposed to produce some observable effect. But not even this is true. We have already glanced at Lucretius' doctrine that the mind has before it innumerable simulacra which never reach the level of consciousness, because the time interval during which they are present is imperceptibly small. But if the impact of those complicated atomic configurations which constitute simulacra could have no observable effect, it is a safe inference that the minute swerve of a single atom would be undetectable. So we can, after all, make use of the Epicurean concept of the concilium in our explanation. I argued previously against Bailey's use of it in saying that "what in the individual atom is a matter of chance, in the conscious complex of the animus is 'conscious chance.'" It is impossible to see how the random motion of an individual atom can by itself account for the end-directed motions of the complex of which it is a part. It is perfectly reasonable, however, that the random motion of a single atom should be concealed by the fact that it is just one element in a complex.

The Epicurean psychology of action, if I am right, was in outline as follows.

Each person is born with a psyche of a particular character, determined by the proportions of atoms of the four different kinds which constitute a psyche. From the beginning of life, reactions occur between the psyche and the external world, through the medium of atomic eidola which flow from all objects and may reach the psyche through the sense organs and the mind. From the beginning, the child experiences feelings of pleasure and pain; in atomic terms, pain is a disturbance of the motions of the psyche atoms caused by a lack of something, and pleasure is either the restoration of the undisturbed motions which constitute tranquillity, or else the state of tranquillity itself. The child learns to associate external objects with one or other of these feelings. A feeling of something lacking constitutes a motive to make good the lack, and so creates an impulse towards an object in the external world which the child has learned will supply the deficiency.

A person's feelings, and therefore his motives and his behavior, are to some extent determined by his genetic inheritance of a psyche of such and such a constitution.

Swerves allow psychological character development (cf. Robert Kane's "self-forming actions")

But the motions of the psyche (and it is in its motions that all its character and action consists) are not determined ab initio, because a discontinuity is brought about by the atomic swerve. The swerve of an atom or atoms in the psyche means that the inherited motions are disturbed, and this allows new patterns of motion to be established which cannot be explained by the initial constitution of the psyche.

There is both continuity and discontinuity. The character of the person is to some extent still determined by the initial constitution of his psyche, because the proportions of atoms of different types in it remain the same. But to a much greater extent his character is adaptable, because the motions of the atoms are not determined and can be changed by learning.

A person learns by experience. He learns what desires must be satisfied, and what objects satisfy them, simply by constant repetition of the experience of desire and satisfaction. He can learn by individual trial and error, or by precept and example from others. If he is indoctrinated in the Epicurean philosophy, he learns to distinguish desires which arise from nature and must be satisfied from those which arise from nature but need not be satisfied and from those which do not arise from nature and are best eliminated. He learns that the limit of pleasure is the absence of pain, and so ceases to feel pain through desire for some extra pleasure. His feelings become disciplined, so that an improper object—one that brings more pain than pleasure in the long run—no longer arouses desire in him. He learns not so much to reject some of the things he desires as to cease to desire the things he ought to reject.

The wise Epicurean is not to be pictured as asserting himself by repeated "acts of volition" against the temptations of the world, but as having learned not to be tempted. His "freedom" does not consist in being presented with possible alternatives, and in choosing one when he might have chosen the other. It consists rather in the fact that his psyche is the product of his own actions and is not unalterably shaped by some "destiny" from the time before his birth.

The weakness of this theory of "freedom," both in its Epicurean and in its Aristotelian form, is to be found chiefly in its refusal to consider the processes of character formation. When Aristotle says that children should be brought up from the beginning to feel pleasure and pain in the right objects, he obviously does not consider such education to be equivalent to compulsion. He stresses that educators and lawgivers use punishments and other incentives to make people behave in the right way, and at the same time insists that the acts which create virtuous dispositions are not to be referred to causes outside ourselves." It is curious that he does not see this as a problem, since it was clearly raised by Gorgias in his Praise of Helen, almost a century before, when he offered as one of his excuses for Helen's behavior the possibility that she was persuaded by argument. It might well have arisen, too, from a consideration of Democritus' ethical opinions. Part of the explanation is probably that persuasion was commonly seen as an antithesis to compulsion? But Aristotle should have seen the need to reestablish this antithesis, since he had to some extent broken it down himself in talking of a class of actions which were a mixture of the voluntary and the involuntary.

If Aristotle had seriously examined the reasons why he took the results of education to be "in our own power," he would have been compelled to specify more exactly what he meant by saying "the source is in us." He might then have been led to say that the criterion of morality (that is to say, the criterion that determines whether an action is liable to moral appraisal or not) is to be found precisely in our ability to be influenced by persuasion as opposed to force. If he had stressed this, then I think Epicurus might after all have thought the swerve unnecessary (unnecessary, that is to say, in his psychology; it was still needed in his cosmology). For in his theory, the effects of persuasion would be similarly explained whether the swerve were there or not. Persuasion is by words, and words, in the crude atomism of the time, do their work by collisions, through the medium of the sense organs. The swerve is not needed for them to have this effect.

In his conclusion, Furley seems comfortable with modern compatibilism

I leave it to others to decide whether the Epicurean theory, without the swerve, would have been "determinist" as opposed to "libertarian," because I do not yet see how to define this particular antithesis. But if it would be determinist, I think it would be a sort of determinism that is compatible with morality.
(Furley, Two Studies in the Greek Atomists, pp.232-236)

It is unfortunate that our knowledge of the early history of the Stoics is so fragmentary, and that we have no agreed account of the relations between them and Epicurus. On the evidence we have, however, it seems to me more probable that Epicurus was the originator of the freewill controversy, and that it was only taken up with enthusiasm among the Stoics by Chrysippus, the third head of the school.

The outlines of Epicurus' approach are familiar enough. He took over the atomic theory of Democritus almost unchanged, but introduced one significant new point, the swerve of the atoms, a slight change of direction that could occur without any cause. According to tradition this was to solve two problems for him: the change of direction would enable atoms otherwise falling all in the same direction and at the same speed to collide and so enter into larger combinations, and the fact that it occurred without cause would break the otherwise continuous chain of causation and so allow room for freedom of action by men, whose minds were composed of atoms and therefore subject to the same laws as everything else.

In spite of the poverty of our evidence, it is quite clear that one main reason Epicurus had for introducing the swerve, or rather the swerve as a random, uncaused event, was as a solution to the problem of freewill. Unlike Aristotle, he fully appreciated that there was a problem. He believed in free will, because it seemed to him manifestly clear that men could originate action, but he could not, like Aristotle, regard this as the end of the matter. We may not think much of the solution he offers, but he deserves full credit for appreciating the problem.

There are now two main points to be cleared up: (1) was Epicurus the first to appreciate the problem, or was he anticipated by the Stoics or someone else? (2) If he was the first, how did he come to do so, and what exactly was the nature of the problem as he saw it?

...we have to explain why Aristotle was so resistant to determinism, and Epicurus so impressed by it. The answer must surely lie, in part at least, in their differing attitudes to Democritus. Aristotle was indeed steeped in Democritus, and had a considerable admiration for him, but at the same time found his system quite unacceptable. We can see why this was so. Aristotle's thought was dominated by a teleological view of causality, in which the paradigm of what guides change is the tendency of an organism to develop into a certain kind of thing. This made the idea of a causal chain in which the future is entirely determined by the past strange and irrelevant.

...in Book K (1064b 35) Aristotle takes his stand on the point that we know very well that some things happen kata symbebekos, which is in opposition to ex anankes, and that, in this context, means causally determined in our sense. What happens kata symbebekos is, then, undetermined. Aristotle then had two reasons for rejecting determinism, (i) that some things obviously happened kata symbebekos, and (ii) that men had free will [Aristotle only says some actions are "up to us."] At the same time it is putting it too strongly to say that he rejected determinism: rather it seems that it was for him a non-starter. This is clearly in sharp contrast to the views of Epicurus and the Stoics, both of whom made valiant if unsuccessful attempts to reconcile freedom and determinism.

...the fact remains, on the evidence of Cicero and Lucretius, that Epicurus still ultimately traced the freedom of the will to the swerve of the atoms. How exactly he did this remains a mystery.

The philosophical, as distinct from the historical, conclusion of my argument is twofold, first that it was possible for men like Plato and Aristotle to hold many educational and psychological beliefs in common with us without being aware of any freewill problem because they had no notion of thorough-going psychological determinism, and, second, that once the problem had been formulated it was appreciated by philosophers of many different schools throughout later antiquity as if it were indeed a natural problem.
(Pamela Huby, "The First Discovery of the Freewill Problem", Philosophy, 42 (1867), pp.353-62)

Long and Sedley agree with Pamela Huby that Epicurus was first to see the free will problem

Epicurus' problem is this: if it has been necessary all along that we should act as we do, it cannot be up to us, with the result that we would not be morally responsible for our actions at all (especially A, E 3, F 1, G). Thus posing the problem of determinism he becomes arguably the first philosopher to recognize the philosophical centrality of what we know as the Free Will Question. His strongly libertarian approach to it can be usefully contrasted with the Stoics' acceptance of determinism (see 62).

Epicurus certainly saw the Democritean atomism which he had inherited as vulnerable to such a challenge, since it made all phenomena, including human behaviour, fully accountable in terms of rigid physical laws of atomic motion, and hence necessary: see A 2, C 13-14, E 3, G. It is perhaps the most widely known fact about Epicurus that he for this reason modified the deterministic Democritean system by introducing a slight element of indeterminacy to atomic motion, the 'swerve' (on which see also 11H with commentary): E 2-3, F, G. But taken in isolation such a solution is notoriously unsatisfactory.

Randomness is no better than necessity in the
standard argument against free will

It promises to liberate us from rigid necessity only to substitute an alternative human mechanism, perhaps more undependable and eccentric but hardly more autonomous. Epicurus' remarks in A 1, where 'that which depends on us' (or 'that which is up to us') is contrasted with unstable fortune as well as with necessity, suggest that he meant to avoid this trap. In order to see how, we must defer discussion of the swerve for now..

The swerve is not even mentioned in the surviving papyrus fragments [B,C] of Epicurus' book on the issue of responsibility from which B and C are drawn But the book still sheds abundant light on the question. In C he conducts a running debate with a Democritean determinist. Democritus himself, we are told, simply failed to see the implications of his determinism for human action (C 13-14). Epicurus' principal target in C 2-12, on the other hand, is someone who consciously applies mechanistic determinism to all human behaviour, including his own. He probably has in mind such fourth-century Democriteans as his own reviled teacher Nausiphanes — the heirs of Democritus derided C 13, as perhaps also implicitly in G. (The early Stoics have sometimes been identified as his target, but cf. 62 with commentary; 'natural philosophers', A 2. would not normally be used of Stoics, in any case.)

In C 1 Epicurus is arguing that since we start with a wide range of potentials ('seeds') for character development our actual direction of development is not physically predetermined but 'up to us'. There are physical influences, but we can control them (cf. 15D 7-8). If it were they that controlled us, our moral and critical attitudes to each other would make no sense (C 2). This leads him into his anti-determinist digression, which continues until its express conclusion at C 15. The determinist may simply regard these attitudes as themselves necessitated (C 3). But this does not save him from the charge of self-refutation (C 5, and perhaps already in the very fragmentary C 4): his own critical attitude in this very debate still implies what he wishes to deny, that the parties to the debate are responsible for their own views. The determinist will resort to the defence that he is compelled to behave in this way; when challenged once again for continuing to argue, will repeat the defence; and so on ad infinitum. Epicurus' objection to this infinite regress (C 6) is not that it is in itself vicious, but rather that it leaves the inconsistency untouched: at every stage of the regress the determinist's behaviour in continuing to argue his case as if with a responsible agent contradicts his thesis that everything, including our beliefs, is mechanically necessitated.

In the second stage of the digression, C 8-12, Epicurus suggests that determinism cannot amount to a substantive thesis about the world, and that its application of 'necessity' to human agency will turn out to be no more than a change of terminology. First (C 8) comes an appeal to 'preconception' (on which as a criterion, see 17 above). We all share a preconception of our own agency as that which is responsible for our behaviour: to defuse the evidential force of this, the determinist would have to show how the alleged preconception has come to embody a faulty 'delineation' (cf. 17E 2, 5) of the facts. (Compare Epicurus' own grounds for dismissing the alleged preconception of the gods as provident, 23B—C below.) If he cannot, the preconception remains valid and the determinist's contribution is merely a new name for it. Second (C 9), his thesis is pragmatically empty. Since he denies us an internal source of self-determination (an 'auxiliary element or impulse in us') he can never expect his arguments to dissuade us from any action. In this Epicurus contrasts him with someone who has a proper grasp (as recommended in A 1) of the difference between the necessitated and the unnecessitated, and who consequently can expect to dissuade us from actions which would involve resisting necessity (C to) perhaps, for example, dissuade us from a vain desire to evade the inevitability of death, because unlike the determinist he can appreciate that while death is necessary our wishes are up to us. Third (C II), the determinist leaves himself no tools for analysing 'mixed' actions (as they are called by Aristotle, Nicomachean ethics in. I), those performed freely but reluctantly in avoidance of a greater evil, since he is unable to distinguish the voluntary from the necessitated elements in them.

The final stage of the argument, C 13-14, is pragmatic, appealing to the disastrous practical consequences that would have ensued had Democritus remembered to apply his thesis of universal necessitation to himself. No illustration is given, but one easy example would be the abandonment of decision-making (cf. 55S). It is remarkable how closely the internal structure of this anti-determinist argument matches that of 16A's anti-sceptic argument, with the sequence of a self-refutation challenge (C 3-7; cf. 16A 1), an appeal to preconception and word-meaning (C 8-12; cf. 16A 2-3), and a pragmatic argument (C 13-14; cf. 16A 9-10). So too its function as a digression added late in the book to justify the preceding positive account of psychological causation matches the role of 16A in relation to Lucretius' preceding positive account of sense-perception. None of this is likely to be mere coincidence. For scepticism and the kind of mechanistic determinism envisaged here were seen as joint consequences of Democritus' reductionist atomism. If phenomenal properties were reducible to mere configurations of atoms and void, it seemed to follow that the atoms and void alone were real while the sensible properties were arbitrary constructions placed upon them by our cognitive organs. The result was scepticism about the sensible world, which had become the characteristic stance of most fourth-century Democriteans (see further, 1 and 16). Similarly, if the 'self' and its volitions were reducible to mere sequences of atomic motion in the soul, human action would easily appear to be mechanistic, fully explicable in terms of primary physical laws, with no additional explanatory or descriptive role left for such psychological entities as belief and volition. And that is just the kind of theory under attack in C (cf. especially C 2, 9).

Given the extent of this parallelism between scepticism and determinism, and between Epicurus' respective refutations of them, we might expect his own positive alternatives to them to be similarly comparable. And so they are.

Epicurus's reaction to skepticism is similar to David Hume's "naturalism" or "realism."

Just as his answer to scepticism is to affirm the reality of phenomenal properties and the truth of sense-impressions of them (see on 7 and 16), so too his answer to mechanism is to affirm the reality and causal efficacy of the self and its volitions as something over and above the underlying patterns of atomic motion. This plainly emerges from B, despite the lack of context and certain difficulties of interpretation. Epicurus is speaking of self-determining animals. (Volitional autonomy is not restricted to human beings, cf. F 1-2; but elsewhere in the book, j in vol. 2, wild animals seem to be excluded, as lacking self-determination and hence as exempt from moral criticism, though not from hate.) Their misbehaviour is quite explicitly said (B 1-4) to be attributable not to their atoms but to their selves and their 'developments'. The latter term, which is crucial to the entire book's discussion, is explicated at B 5. The kind of 'development' which contributes psychological autonomy is one which is distinct from the underlying atoms in a 'differential' way ('transcendent' would be a tempting translation of the Greek word) — a way more radical than 'the way which is like viewing from a different distance'. The point is apparently that all bodies have certain properties, e.g. colour, over and above their constituent atoms, but that there the main difference is one of scale, one between macroscopic and microscopic analysis; whereas the 'developments' which supply autonomy differ from the atoms in a much more fundamental way. The fragmentary state of the text leaves us to guess at the nature of this difference, although it is hard to doubt that it includes the intentional properties associated with consciousness. How do these psychological entities relate metaphysically and causally to the mind's atoms? They can only be, technically speaking, 'accidental attributes' of those atoms (cf. 7). But they are not mere epiphenomena, supervenient on atomic motions and causally determined by them. For Epicurus is quite explicit in attributing to them a causal efficacy distinct from that of the atoms. Hence, although atomic make-up may be responsible for disorderly motions of the mind-atoms (B 4), it does not follow that we cannot make decisions which override those motions, and according to B 6 psychological causation actually operates on our component atoms. This throws immediate light on Lucretius' insistence at 14D 5 that although atomic composition of the soul determines our natural temperament, we can learn through reason to overcome that temperament. Perhaps, for instance, a natural coward can learn courage through rational reflection. His disorderly motions of soul atoms may then be stabilized, so that he ceases to suffer even the physical sensations of fear.

By now the familiar 'materialist' label is beginning to fit Epicurus less neatly. Although he holds prima facie an Identity Theory of mind (see 14), he does not regard mental states as capable of straightforward physical analysis, for although properties of the corporeal mind they are not mere physical states of it. We have here, then, an interactionist dualism of the mental and the physical. But there is no hint of Cartesian dualism. A better comparison would be with the modern notion of Emergence. In Epicurus' view, matter in certain complex states can take on non-physical properties, which in turn bring entirely new causal laws into operation.

B 7 emphasizes that the distinction between physical and psychological causation is crucial to an understanding of responsibility. And certainly it does constitute at least the beginning of an answer to determinism. The 'self' which is responsible for our actions is, Epicurus will say, more than a mere bundle of atoms, and therefore is not reducible to a link in a physical causal chain. Indeed Carneades, in defending Epicurean libertarianism for his own dialectical purposes (see 70G and commentary), suggested that this was already a sufficient answer to determinism: E 4-7. But how, it will be asked, can this emergent property of the corporeal mind so effectively take control of the soul, and through it of the body, as to move their atoms in ways in which according to the laws of physics alone they should not have moved? If the laws of physics are sufficient to determine the precise trajectory of every atom in us, how can the self be anything more than a helpless spectator of the body's actions?

Here at last a significant role for the swerve leaps to the eye. For it is to answer just this question, according to Cicero at E 3, that the swerve was introduced. The evident power of the self and its volitions to intervene in the physical processes of soul and body would be inexplicable if physical laws alone were sufficient to determine the precise trajectory of every atom. Therefore physical laws are not sufficient to determine the precise trajectory of every atom. There is a minimal degree of physical indeterminism — the swerve. An unimpeded atom may at any given moment continue its present trajectory, but equally may `swerve' into one of the adjacent parallel trajectories (see commentary on 11H).

As far as physics is concerned there is simply no reason for its following one rather than another of these trajectories. Normally, then, the result will be, in this minimal degree, random. But in the special case of the mind there is also a non-physical cause, volition, which can affect the atoms of which it is a property.

Long and Sedley here arrive at our Cogito model, speculating that randomness provides the alternative possibilities from which an adequately determined volition can choose

It does so, we may speculate, not by overriding the laws of physics, but by choosing between the alternative possibilities which the laws of physics leave open. In this way a large group of soul atoms might simultaneously be diverted into a new pattern of motion, and thus radically redirect the motion of the body. Such an event, requiring as it does the coincidence of numerous swerves, would be statistically most improbable according to the laws of physics alone. But it is still, on the swerve theory, an intrinsically possible one, which volition might therefore be held to bring about. For a very similar thesis relating free will to modern quantum indeterminism, see A. S. Eddington, The nature of the physical world (1928). (It may be objected that swerves are meant to be entirely uncaused; but, as E 2 shows, that was only an inference by Epicurus' critics, made plausible by concentrating on the swerve's cosmogonic function, cf. 11H, for there it must indeed occur at random and without the intervention of volition.)

Lucretius' evidence in F does not explicitly state the swerve's relation to volition, although numerous attempts have been made to discover it there. But if the above account of Epicurus' theory is justified by the other testimonia, it becomes clear that F is, at least, fully consistent with it. For the dominant theme of F 1-3 is precisely the evident power of volition to redirect the bodily mass in defiance of its purely mechanical patterns of motion. This is said, in F 1 and 4, to be explicable only if there is an undetermined swerve of atoms, since if impact and weight were the only causes of atomic motion the mind's behaviour would be rigidly mechanistic. Some have also seen in F 1 the further implication that the initiation of every new course of action directly involves the swerve. All this fits the above account comfortably enough. What is missing, of course, is an explanation of the non-physical character of psychological causation — not surprisingly, given that Lucretius' poem is about physics and that his sole object in the context is to complete his account of the laws of atomic motion (cf. 11).

One further dimension to the debate emerges from E 1, H and I. Epicurus saw the threat of universal necessitation not only in unbreakable chains of physical causation, but also in the logical principle of bivalence according to which every proposition is either true or false, including those about the future. His solution of denying the principle as far as certain future-tensed propositions are concerned (the denial is slightly garbled in I's version, where 'one or the other is necessary' ought to read 'one or the other is true'; but the example is clearly authentic — Hermarchus was Epicurus' pupil and successor) was essentially that of Aristotle, according to the traditional reading of his celebrated Sea Battle discussion at De interpretatione 9. But Epicurus, like the Stoic with whom he is contrasted in E I (see further, 38G), saw physical and logical determinism as two aspects of a single thesis. The two formulations of determinism tend to be treated as interchangeable, as do the two respective solutions, the swerve and the denial of bivalence (cf. Cicero, On fate 18-19, and perhaps E 1-3). This conflation seems to rest on the assumed equivalence of 'true in advance' with 'determined by pre-existing causes'; cf. also the telling comment at the end of I.

The interpretation of the swerve theory adopted above may help explain how it could be thought interchangeable with the denial of bivalance. Neither doctrine is involved in analysing the nature of volition itself (as many have thought the swerve to be). Their shared function is to guarantee the efficacy of volition, by keeping alternative possibilities genuinely open.
(Long and Sedley, The Hellenistic Philosophers, section 20, "Free Will," pp.107-112)

[The discussion of the swerve in Book II of De rerum natura] has received brilliant treatment from D. J. Furley in a work which is in many ways a model for the analysis of ancient philosophical texts. Yet it still seems to me that there is more to be said. I want here to try briefly to offer a fresh analysis of the argument of the vital paragraph 251-93, and to situate it within an Epicurean context. Inevitably this will involve criticism of Furley; let me state again at the outset my admiration of his work.
(p.330)

I turn to the overall interpretation. Lucretius is arguing from the existence of voluntas to the existence of the clinamen; nothing comes to be out of nothing, therefore voluntas must have a cause at the atomic level, viz. the clinamen. The most natural interpretation of this is that every act of voluntas is caused by a swerve in the atoms of the animal's mind. The σημείωσις of L. 2. 125-41 is exactly parallel; the visible motions of the dust-particles are a σημεῖον [ἀπὸ τῶν φαινομένον] (128 significant) for the invisible atomic motions which are their cause. There is a close causal, physical relationship between the macroscopic and the atomic. Furley, however, argued that the relationship between voluntas and the clinamen was very different; not every act of volition was accompanied by a swerve in the soul-atoms, but the clinamen was only an occasional event which broke the chain of causation between the σύστασις of our mind at birth and the 'engendered' state (τὸ ἀπογεγεννημένον) which determines our actions. Its role in Epicureanism is merely to make a formal break with physical determinism, and it has no real effect on the outcome of particular actions.
(p.338)

For Furley, both of these accounts are essentially ones of stimulus and response; action follows automatically upon perception, and the nature of the action is determined by our constitution, the sort of person we are. In accordance with this, he analyses the passage from De rerum natura Book 4 as follows:

(1) Simulacra meandi must strike our minds, among the innumerable other simulacra which are always abroad in the air (881-885).

(2) The mind must be focussed, as it were, on walking, so that these simulacra form an image while others do not (882-886).

(3) Voluntas fit . . . animus sese ita commovet ut velit ire (883, 886). (4) The mind transmits motion to the limbs, bit by bit (887-891).

Here the occurrence of voluntas is consequent on the focusing of the mind. But that is not what Lucretius says; a more accurate analysis of the paragraph would be:

(1) 881-2. First simulacra strike the mind, as explained previously.

(2) 883-5. Next voluntas occurs; for the mind does not begin any action before the process of 'prevision' has taken place. An imago is formed of what the mind anticipates.

(3) 886-90. Therefore, when the animus moves itself in such a way as to want to go, straight away it transmits its motion to the anima. Then the anima strikes the body . . .

Lucretius is concerned in this passage with how we move when we wish to, not with how we come to wish to move; hence there is no explanation of how voluntas occurs. But there is certainly no evidence for the idea that voluntas is caused by sense perception directly, and hence that there is no room for the occurrence of a clinamen in the soul-atoms. Simulacra are striking our mind all the time, but we do not 'see' them unless we concentrate on them in an ἐπιβολή τῆσ διανοίας, as Lucretius explains in 4. 802-17. What we concentrate on depends on our voluntas. Once the image is clearly visualized — once we have a φαντασία — then indeed the bodily reactions proceed from that automatically. But voluntas comes before, not after, the production of the image; as K. Kleve remarks, 'wir können selbst wählen, welche Bilder wir bemerken wollen, d.h. auf welche Bilder wir unsere Aufmerksamkeit (ἐπιβολή) richten wollen'. Furley argues that we cannot situate voluntas at this stage 'because Lucretius goes to great lengths to give a causal explanation of why the mind focuses on some things rather than others'. The passage referred to is 4. 962-1036, and in particular 973-83. But Lucretius is clearly there describing an exceptional and involuntary experience which offers an analogy for the phenomenon of dreaming. There is no suggestion that that is what ordinary perception and thought, still less action, are like. There is therefore no reason to doubt that in 4. 881-90 Lucretius situates voluntas before the act of ἐπιβολή and therefore no reason to see voluntas as causally conditioned by perception. Ample room is left for the clinamen to fill; and indeed what else could fill it?
(p.341)

For Lucretius, voluntas takes place in the mind, the animus, but it is also a purely physical occurrence. There is no disembodied faculty of the will separate from the physical constitution of the animal." Voluntas is not, moreover, in Lucretius' view merely the object of introspection; we can see it occurring in others. It takes place when the mind decides to focus on certain simulacra in an ἐπιβολή τῆσ διανοίας, and is thus situated between sense perception and the formation of a specific φαντασία which leads to action. It is caused by a random swerve in the downward motion of an atom or atoms in the 'fourth substance' of the animus, which causes an alteration in the atomic motions which eventually leads to a specific action. What action, if any, a swerve issues in is determined by which atoms swerve and by the constitution of the animus. On any particular occasion, what action the animal will take is unpredictable, but over a series of actions his reactions to the external world will be broadly consistent with the sort of being he is. This theory has usually been greeted with contempt, in ancient and modern times. And its special problems are undoubtedly immense, quite apart from those which face any traditional account of the will as a distinct psychological phenomenon. But it is also a bold imaginative scheme, and an attempt to produce a precise physical account of puzzling psychological problems; it is surely, other considerations apart, a more interesting theory than a mere rehash of Aristotelianism would have been, however philosophically more respectable. It was not the whole of Epicurus' answer to the problems of human freedom; I have not touched at all on Epicurus' denial of a truth-value to statements about the future, which was designed to refute logical determinism as the clinamen did physical. The relationship between this move and the introduction of the clinamen is not clear, and requires further study. But I hope I have shown that the theory of the clinamen as presented by Lucretius is a self-consistent, reasoned theory in itself, firmly embedded in the Epicurean system as a whole and designed to answer real philosophical problems, rather than merely an awkward embarrassment.
(p.351-2)
(Don Fowler, "Lucretius on the Clinamen and 'Free Will'", Συζήτησισ: Studi sull'epicureismo greco e romano offerti a Marcello Gigante, (Naples, 1983) 329-52)

...since swerves are random, it is hard to see how they help to explain free action. We can scarcely expect there to be a random swerve before every free action. Free actions are frequent, and (fairly) reliable. Random swerves cannot account for either of these features. This problem would be lessened if we could assume that swerves are very frequent, so that there is always likely to be one around before an action. However, if swerves are frequent, we face the problem that stones and trees ought to be enabled to act freely. And even in the case of humans random swerves would seem to produce, if anything, random actions; we still lack any clue as to how they could produce actions which are free.

An influential modern line of thought avoids these problems by arguing that our evidence does not demand that there be a swerve for each free action [Furley]. Rather, swerves explain the fact that people have characters capable of change and reaction that goes beyond mechanical response to stimuli. We act freely because we have characters that are flexible and spontaneous, and this is because we are composed of atoms which swerve occasionally. On this account, swerves do not have to be frequent, since they are not part of any mechanism of action; one swerve in your soul is enough for the kind of character flexibility that is required. Such an account avoids the problems attaching to any account that brings swerves into free action, but at the cost of not answering very closely to the evidence; the Lucretius passage certainly suggests that swerves are in some way relevant at the point of action.

Another kind of suggestion is that swerves are not the causes of free actions at all. Rather, they come into the process whereby free actions are brought about. Swerves are supposed to explain something about the nature of free agency and how it works, but they do not cause free actions (by cutting across causal chains, for example). This suggestion can be developed in several ways. The boldest version holds that swerves do not explain the existence of free volitions at all; [Sedley] rather Epicurus holds anyway that volitions are nonphysical, "emergent" entities.

Random swerves provide alternative possibilities for an adequately determined will to choose from

The role of swerves is to provide alternative possibilities for volitions to choose between, for there would be no point in having free will if there were no genuinely open possibilities between which to select. This suggestion depends on the strong thesis that Epicurus regards the mind as something nonphysical, which we have seen to be highly contentious; and also it likewise does not really answer to the evidence, in which it is not merely the possibility of swerves, but actual swerves, which play a role at the level of action. A second kind of account gives the swerve a role in enabling the mind to focus on one thing rather than another by way of the mind's selective "grasp" or epibole tes dianoias. A third sees it as parallel to Aristotle's use of the connate pneuma; that is, it creates a new kind of physical substance which explains, within a physicalist system, how human minds can be active, and in particular can initiate action.

It is undoubtedly more attractive to find a role for swerves in the mechanism of free action, rather than as mysterious events enabling free action to come about. However, all such accounts face the problem of evidence: Lucretius, the only source who gives us much detail about the swerve in human action, associates it with the formation of impulse (voluntas), not with any subsequent mechanism to carry it out. However embarrassing we may find the thesis that the swerve explains the formation of free impulses, and in some way explains how they are free, that remains the view best supported by the ancient evidence.

As we have seen, however, occasional random swerves cannot produce reliable free actions. The only way that the theory has a hope of working is on the assumption that swerves are extremely frequent, so as to produce a standing physical con dition. How, though, do we avoid the obvious objection that trees and stones would also contain frequent swerves, given that it is an important aspect of Epicureanism that human beings are parts of nature, atomic compounds like the others? We can meet this objection by the consideration that swerves are indeed everywhere frequent, but that they produce effects only in human souls, perhaps indeed only in the rational parts of human souls. This is because the human rational soul is a compound of the finest and most tenuous atoms, and only this kind of compound permits swerves to have effects. Thus we are free, and trees are not, because of a physical difference: in our minds atomic swerves produce effects, which somehow enable us to act freely. While the mechanism remains somewhat sketchy, we can see the general idea. Swerves do not operate one per action; rather, because we (and some animals) are the kinds of atomic compound that we are, we are able to act freely, in a way that genuinely chooses between alternatives.

But now we find a striking redundancy, for Epicurus has already postulated the nameless atoms in the soul to account for the complexity of sentient and intelligent behavior. Why do we need swerves as well to account for the same fact? Impressive as the fact may be, we hardly need two such physical differences to account for it. It might be objected that nameless atoms account only for agency, while we need swerves to account for free agency. But it is quite unclear from our evidence what this difference would be taken to consist in. This is especially so since Lucretius uses animal behavior as an example of free agency, ruling out the otherwise promising idea that freedom might be a matter of informed choice between alternatives, or something similar which is plausibly found only in humans.

It is very hard not to feel pressured here toward a developmental hypothesis, namely, that Epicurus had both these ideas, but not at the same time. It has been suspected on other grounds that the swerve was a late idea of Epicurus', one developed after he had written his major works, possibly in response to objections. It is also possible that Epicurus himself had no very definite theory of how the swerve underpins free agency, and that later Epicureans filled in the story, possibly in divergent ways, just as modern scholars do. It is hard to conclude, however, that the swerve was a good idea, and the disproportionate emphasis which it has received in discussion of Epicurus' ideas about the mind has been unfortunate.
(Julia Annas, The Hellenistic Philosophy of Mind, pp.184-88)

Purinton makes the all-too-common common error of translating Lucretius' libera as 'free volition'

(1) First and foremost, there is De Rerum Natura 2.251-93, where Lucretius presents what I shall call 'the libertarian argument' for the thesis that atoms swerve. Lucretius maintains that the swerve is that "whence" (unde) arises the "free volition" (libera voluntas) whereby "we likewise swerve our motions" (declinamus item motus) when and where we wish. And the natural way to read this is as claiming that swerves cause volitions from the bottom up. (I shall call this 'the argument from Lucretius.')

Purinton simply identifies swerves (one or more) with volitions

(2) Other than Lucretius, none of the authors who mention the swerve gives any account at all of the relation between swerves and volitions. That is not very surprising, if the relation between swerves and volitions is so simple that it can go without saying (as it is on my view, according to which volitions 'at bottom' just are swerves). But it would be very surprising, if Epicurus had a complicated view of the relation of swerves to volitions, such as that swerves are very rare events which are not directly linked to volitions, but function only to break the chain of causation every once in a long while (as Furley would have it) or that swerves cause us to focus on images of actions, which then cause volitions (as Fowler would have it) or that first volitions occur and then, after a short wait, swerves occur to trigger the desired bodily motion (as Englert would have it) or that swerves are caused by volitions, as emergent properties of the mind, from the top down (as Sedley would have it). For none of our sources say any such thing. (I shall call this 'the argument from silence.')

Purinton is wrong here. As Long and Sedley argued, our thoughts and alternative possibilities can be free, and our willed actions adequately determined

(3) Now set aside all textual evidence and simply ask what a would-be libertarian atomist is obliged to say. Since, to be a libertarian, one must say that volitions are fresh starts of motion, and since, to be an atomist, one must say that all mental events are caused from the bottom up by the motions of the mind's constituent atoms, a would-be libertarian atomist is obliged to say that volitions are caused from the bottom up by fresh starts of atomic motion. (I shall call this 'the a priori argument,' since it does not depend on any textual evidence.)

(4) At Ennead 3.1.1, Plotinus formulates the problem with Epicurus' position thus:

One must not admit the uncaused by positing vain swerves or a sudden motion of bodies which happens with no antecedent cause or a sudden volition (ὁρμή) of the soul with nothing moving it toward doing what it was not doing before. Or else, by this very thing, a greater necessity would hold the soul, that of not belonging to itself, but of being borne along with such motions as are undesired and uncaused.

Plotinus does not mention Epicurus here, but the allusions to "the uncaused" and "swerves" strongly suggest that it is Epicurus' view that Plotinus is here criticizing. And that is significant. For the objection that Plotinus makes - that we would not be in control of our own lives if we were borne along by random atomic motions - is basically Furley's objection to the view that swerves cause volitions from the bottom up. But, whereas Furley presents this as an objection to the thesis that Epicurus held such a view, Plotinus presents it as an argument against Epicurus' view. And that supports my thesis that, as a matter of historical fact, Epicurus did hold that swerves cause volitions from the bottom up. (I shall call this 'the argument from Plotinus.')

Notice, by the way, that Plotinus speaks here of "swerves" in the plural but of "volition" in the singular. This raises a question: what are we to say is Epicurus' view of volition in the singular? We can be sure that, according to Epicurus, just as a mind is at bottom a plurality of atoms, so a volition, as a motion of the mind, is at bottom the motions of a plurality of atoms. But, in a given volition, how many of these many atomic motions are swerves? All of them? Only one? Some, but neither all nor only one? My guess is that Epicurus did not believe that, in a typical volition, only one atom swerves. He rather believed that many do, and that more do the more strenuous the action; more mind-atoms swerve when one tries to turn one's body sharply, for instance, than when one tries to deviate just a little from one's path. But this is just guesswork, which I want to keep separate from my main thesis. So here is how I shall formulate in the singular my thesis that Epicurus held that volitions (in the plural) are caused by swerves from the bottom up: Epicurus held that an agent's volition (in the singular) is caused from the bottom up by that agent's mind's atoms' motions, at least one of which is a swerve.

That is my main thesis.
(Jeffrey Purinton, "Epicurus on 'Free Volition' and the Atomic Swerve,' Phronesis, 44, pp.256-59)

In her book and a 1998 article in Phronesis (Vol. 43, No. 2 (May, 1998), pp. 133-175), Bobzien identified several variations on the theme of human freedom that were important in antiquity. Three of them are indeterminist freedoms, by which she means the decision is partly or wholly a matter of chance, and does not involve the character and values of the agent:

1) freedom to do otherwise: I am free to do otherwise if, being the same agent, with the same desires and beliefs, and being in the same circumstances, it is possible for me to do or not to do something in the sense that it is not fully causally determined whether or not I do it.

2) freedom of decision: a subtype of freedom to do otherwise. I am free in my decision, if being the same agent, with the same desires and beliefs, and being in the same circumstances, it is possible for me to decide between altemative courses of action in the sense that it is not fully causally determined which way I decide. 1) differs from 2) in that it leaves it undecided in which way it is possible for the agent to do or not to do something.

3) freedom of the will: a subtype of freedom of decision. I act from free will, if I am in the possession of a will, i.e. a specific part or faculty of the soul by means of which I can decide between alternative courses of actions independently of my desires and beliefs, in the sense that it is not fully causally determined in which way I decide. 2) differs from 3) in that the latter postulates a specific causally independent faculty or part of the soul which functions as a "decision making faculty."
(Phronesis, p.133)

In 1967 Epicurus was credited with the discovery of the problem of free will and determinism. Among the contestants were Aristotle and the early Stoics.

Furley merely de-emphasized the direct involvement of the random swerve in volition, as had Bailey before him, to avoid the standard argument against free will

Epicurus emerged victorious, because — so the argument went — Aristotle did not yet have the problem, and the Stoics inherited it from Epicurus. In the same year David Furley published his essay 'Aristotle and Epicurus on Voluntary Action', in which he argued that Epicurus' problem was not the free will problem. In the thirty-odd years since then, a lot has been published about Epicurus on freedom and determinism.

But it has only rarely been questioned whether Epicurus, in one way or another, found himself face to face with some version of the free will problem. In this paper I intend to take up the case for those who have questioned the point, combining a fresh perspective on the debate with a selection of new arguments and a detailed textual analysis of the relevant passages. Let me begin with a brief sketch of the problem of freedom and determinism which Epicurus is widely taken to have been concerned with.

The determinism Epicurus defends himself against is usually understood as causal determinism: every event is fully determined in all its details by preceding causes. These causes are commonly pictured as forming an uninterrupted chain or network, reaching back infinitely into the past, and as governed by an all-embracing set of laws of nature, or as manifestations of such a set of laws of nature.

Freedom to do otherwise, freedom of decision, and extreme freedom of the will Bobzien now labels "two-sided" freedom

On the side of freedom, Epicurus is generally understood to have been concerned with freedom of decision (the freedom to decide whether or not to do some action) or freedom of choice (the freedom to choose between doing and not doing some action) or freedom of the will (where the freedom to will to do something entails the freedom to will not to do it, and vice versa; I call this two-sided freedom of the will). Epicurus is taken to have introduced an indeterminist conception of free decision or free choice or two-sided free will: agents are free in this sense only if they are causally undetermined (or not fully causally determined) in their decision whether or not to act or their choice between alternative courses of action; undetermined, that is, by external and internal causal factors alike. There is assumed to be a gap in the causal chain immediately before, or simultaneously with, the decision or choice, a gap which allows the coming into being of a spontaneous motion.

Epicurus did not make actions directly the result of random atomic swerves, but he and Aristotle did think volitions were "up to us."

In this way every human decision or choice is directly linked with causal indeterminism. The assumption of such indeterminist free decision, free choice, or two-sided free will does not presuppose that one specifies an independent mental faculty, like e.g. a will, and indeed it is not usually assumed that Epicurus' theory involved such a faculty.

The 'free will problem' that Epicurus is assumed to have faced is then roughly as follows: If determinism is true, every decision or choice of an agent between alternative courses of actions is fully determined by preceding causes, and forms part of an uninterrupted causal chain. On the other hand, if an agent has (two-sided) freedom of the will, it seems that the agent's decision or choice must not be fully determined by preceding causes. Hence, it appears, determinism and freedom of the will (freedom of decision, freedom of choice) are incompatible.

I do not believe that Epicurus ever considered a problem along the lines of the one just described. In particular, I am sceptical about the assumption that he shared in a conception of free decision or free choice akin to the one I have sketched. (I also have my doubts that he ever conceived of a determinism characterized by a comprehensive set of laws of nature; but this is a point I only mention in passing.) To avoid misunderstandings, I should stress that I do believe that Epicurus was an indeterminist of sorts — only that he did not advocate indeterminist free decision or indeterminist free choice.

But Bobzien is wrong to suggest that Epicurus did not see a problem between human freedom and the causal determinism of his fellow atomist Democritus, and that his atomic swerve was not his proposed solution to that "free will problem." She notes that

Whether Epicurus discussed free will depends on what one means by 'free will'. For example, if one intends 'free will' to render Lucretius "libera voluntas," and to mean whatever element of Epicurus' doctrine Lucretius meant to capture by this phrase, then Epicurus evidently was concerned with free will. My concern is only to show that he did not discuss a problem of free will that involves a conception of freedom of decision or choice as adumbrated in the main text. [namely, "extreme" libertarianism in which chance is the direct cause of action.]

My own thesis is that Epicurus' main concern is not with justified praise and blame, but with preserving the rationality and efficacy of deliberating about one's future actions, although he thinks that determinism is incompatible with both. The reason for this is that a necessary condition on effective deliberation is the openness and contingency of the future, and determinism makes the future necessary. Furthermore, even though Epicurus posits the swerve in order to render causal determinism false, the sort of deterministic argument that Epicurus is concerned to rebut is the fatalist argument given in de Int. 9 and by the Megarians, which moves from considerations of future truth, to the fixity of the future, to the pointlessness of deliberation. Epicurus thinks that, if the Principle of Bivalence (the principle that every statement either is true or is false) held universally, this would make the future fixed in a way such as to render us helpless. (And so we can call my view the 'bivalence' interpretation.) Epicurus thinks that both logical and causal determinism are incompatible with the contingency of the future, and the swerve renders both false, since logical and causal determinism are mutually entailing. The swerve plays no direct role in the production of action or the formation of character.

The main textual support for attributing this role for the swerve to Epicurus is Cicero's De fato. There is precedent for the sort of position Epicurus adopts in Aristotle's rejection of the Principle of Bivalence for similar reasons in de Int. 9. If I am right about this, to assimilate Epicurus' concerns to those of modern libertarians is highly misleading.

In order to establish the 'bivalence interpretation,' I need to go through die texts that bear on the Epicurean position regarding human freedom. Rut before doing so, let me first establish its initial plausibility by showing that none of the terminology Epicureans use when discussing human freedom preclude it, and that the sort of 'free will and determinism' problem that I take Epicurus to be concerned with is one he should be concerned with, given his ethics and psychology, whereas — even apart from considerations of how successfully the swerve addresses these problems the other sorts of free will and determinism problems should not even trouble Epicurus at all.

Epicurus is concerned to defend human freedom, but none of the terminology he uses (or that others use who report on the Epicurean position) show that he is worried about preserving the ability of an agent to do otherwise than he does, much less that he conceived of this two-way ability in libertarian terms. Libera voluntas is often translated `free will,' but depending on the context, it can mean something like 'unfettered impulse.' After all, Cicero is willing to describe even the compatibilist Chrysippus, who certainly did not have a two-sided libertarian conception of freedom of will, as wanting to free (libero) our minds from necessity of motion and to accommodate the views of those who think that the movements of our minds are voluntary (voluntarius)." (For this reason, I will usually translate voluntas as 'volition,' since it is not potentially misleading in the way 'will' is, and while I think that the Epicurean theory of libera voluntas actually ends up being something like a theory of 'unfettered impulse,' using that as a translation would be highly tendentious.)

Likewise, Epicurus says that how we act and develop "depends on us" (παρ’ ῆμᾶς) and that our actions arise through us ourselves or from us ourselves (δι’ ἡμῶν αὑτῶν or ἑξ ἡμῶν αὑτῶν). That our actions are παρ’ ῆμᾶς is compatible with them simply being caused by us (e.g., that I caused myself to walk, so that my walking "depended on me"), and need not imply that, however we act, it is "up to us" whether to act one way rather than another (e.g., that it was up to me whether or not to walk). In fact, Bobzien argues at length, and I think convincingly, that to say our actions are παρ’ ῆμᾶς is more naturally read as indicating that we are causally responsible for our actions (what she calls a 'one-sided causative' παρ’ ῆμᾶς and has no implications about free choice." To say that actions are δι’ ἡμῶν αὑτῶν or ἑξ ἡμῶν αὑτῶν also has no implications of free choice. In fact, Chrysippus defends the thesis that certain things originate "from us" (ἑξ ἡμῶν), and when the Stoics are concerned to describe precisely what type of agency we do have and are at pains to deny the thesis that freedom is a matter of having free choice between opposite actions, they say repeatedly that what is up to us is what happens through us (δι’ ἡμῶν)"

The passages which report the Epicurean views on determinism and freedom indicate that Epicurus is concerned about defending something like the view that we have moral responsibility. The much later Epicurean Diogenes of Oinoanda claims that all censure and admonition would be abolished if fate controlled what we did, and Cicero, in reporting the worry that motivates the various parties to the fate and free volition debate, says that if fate were operative, there would be no justice in either praise or blame. Epicurus, likewise, when arguing that some things 'depend on us,' says praise and blame properly attach to such things, and in his anti-fatalist argument in On Nature 25, he asserts that our practices of rebuking, opposing and reforming each other presuppose that the cause of actions is 'in ourselves.'

However, even more prominent in Epicurean thought is the theme that determinism would render us helpless. When Lucretius describes the libera voluntas that the swerve snatches from the fates, he says nothing about responsibility, praise, or blame. Instead, libera voluntas is what allows each animal to go where pleasure leads it and the mind to move itself (DRN 2 257-260). Although Lucretius' later discussions of voluntas do not mention the swerve, they do confirm that it is voluntas that allows us to act as we wish to act — to visualize what we wish to visualize, to move our limbs as we desire, etc. (DRN 4 777-780, 877-880) So, if determinism threatens this voluntas, the implication is that determinism would render us unable to act as we wish to act, to move our limbs as we will, etc.

A similar concern with fatalism lurks in Epicurus' discussion of the "fate of the natural philosophers" in Ep. Men. 133-134. Epicurus contrasts what is ἁνάγκη, which is ἁνυπεύθυνος; — 'unanswerable' or 'beyond human control' — with what is παρ’ ῆμᾶς, which is ἁδέσποτος; — 'without master' or 'autonomous.' He then goes on to say that it would better to believe in the meddling Olympian gods than to be a 'slave' to the fate of the natural philosophers, since at least one can try to placate the Olympian gods, whereas the necessity of the natural philosophers is inescapable. In On Nature 25, the target of Epicurus' argument is a fatalist: this person denies that our decisions make any difference; what we do is not a cause or explanation (aitia) of what happens.' Finally, Cicero's discussions of the Epicurean and Stoic positions in the De fato show that a major concern of theirs was whether, if what will occur in the future has always been true and always been causally determined, the future is necessary in a way that makes all deliberation and action pointless.

Now, before going through the texts, let's step back and ask: given Epicurus' ethics and his psychology, what sort of freedom should he be worried about? Epicurean ethics is egoistic and hedonistic. In every choice, one should strive to attain the 'goal of nature,' pleasure (KD 25).

This extreme form of libertarianism, in which chance is the direct cause of action, developed by Epicurus critics to attack him, is unlikely to have been what he had in mind

The problem with most of the traditional 'libertarian' interpretations of Epicurus is that they dissociate one's actions not only from external causation but also from being caused by the psychological states of the agent present at the moment of choice: his beliefs, desires, and character, since being causally determined by these states is incompatible with a robust ability to do otherwise than one does.

Let us leave aside, for the moment, the objection that a random atomic swerving in one's mind is an unpromising basis for the production of free and responsible actions, instead of random and blameless twitches. Why should Epicureans be concerned to try to defend this sort of freedom of choice in the first place? If one has correct beliefs about the workings of the world and the limits of what is required for happiness, and one knows what one needs to do in the present situation to attain a pleasurable life, then having one's actions determined by these psychological states would not be ethically problematic — in fact, it is exactly what one would want to happen. This is the state of the Epicurean Sage. It is hard to see how having a 'two-sided' libertarian freedom of choice would help in the pursuit of ataraxia, or tranquility, which Epicurus maintains is the chief constituent of the happy life. And if this freedom would not help in the pursuit of ataraxia, it is hard to see why any good Epicurean should care abhout defending its possibility against the threat of determinism.

Salles discusses several aspects of Stoicism that seem to demand complete determinism.

• The Principle of Bivalence: Statements about the future are already true or false. A true statement about the future necessitates future events, the causes for which must already exist.

  (T_c) For any event S that occurs at some future time t, the proposition S occurs at t cannot be true now, unless there is a cause now (that is, a causal chain stretching from the present to the future time t) for S's occurring at t.

  Chrysippus infers causation from prior truth through (T_c). This inference, in turn, explains why the first premiss of Chrysippus' argument: - If there were causeless events, propositions about future occurrents would lack a truth value' - holds true and, in particular, why propositions second premiss of the argument — 'propositions about future occurrents are already either true or false' — yields that no future event is causeless.

  We may generalize the thesis to all times, which is Chrysippus' intended conclusion ('motus ergo sine causa nullus est'). Since in the past what is now present was in the future and what is now past was present, then, for any present event, there was in the past a true proposition asserting its future occurrence; in consequence, there was at that time a cause of its (then) future occurrence; so every event in the present has a cause; mutatis mutandis, every event in the past has a cause; thus, every event (past, present and future) has a cause. (Salles, pp.8-9)

• Eternal Recurrence: After a given world-cycle, the planets are arranged as they were at the start, and everything that happened in one cycle now reoccurs. Nothing can be discernibly different (ἀπαράλλακτος).

  Stoic indiscernibility is metaphysical and conveys numerical identity. It is metaphysical, as opposed to merely epistemological, because it does not mean just that no observer can register the differences — it also means that there are no differences there to be registered between them; and it conveys numerical identity because if A and B are indiscernible, A and B are not really 'two' discrete things, but rather one and the same object numerically. It is because they are indiscernible that they are in fact numerically the same. As applied to everlasting recurrence, indiscernibility implies that the world of the present cycle is the same in all respects (and hence the same even in number)" as the world of any other cycle: in the qualities of its objects, in how each of them is intrinsically disposed, and in how they are related to each other." As Nemesius indicates further down in his report: 'Everything will be just the same and indiscernible down to the smallest details.' (Salles, p.23)

• Predetermination: The Stoic God foreknows everything about the world, being essentially synonymous with the laws of Nature.

  Since (a) divine providence presupposes predetermination and (b) predetermination, in turn, presupposes foreknowledge. Providence implies predetermination because, to ensure that the world as a whole (in extension and duration) turns out to be as good as possible, god has to predetermine right from the start of the cycle everything that will happen and exist in it. And predetermination implies foreknowledge because god could not be said to predetermine X to happen unless he thereby acquires foreknowledge that X will happen. Divine foreknowledge is not just a sufficient condition for predetermination, but also a necessary one (god predetermines X to happen if and only if god foreknows that X will happen). (Salles, p.27)

• The Future Is Already Fixed: This is a special form of Stoic fatalism.

  Fatalism, or the idea that propositions about future occurrents are already true or false and that the future is already fixed, is earlier than Stoic philosophy. It is a position that was already the target of Aristotelian criticism in the difficult chapter 9 of the de interpretatione. And it is to Aristotle that we owe the objection that fatalism encourages idleness: if it is already true that I will win the elections (or already false that I will lose), why should I do a campaign? Is not a campaign superfluous? And if it is, why should I not sit back and relax until the elections? Chrysippus attempted a defence of fatalism against this objection. He provided strong reasons for thinking that fatalism does not render our actions and efforts superfluous.

• Everything Has A Cause, and Causation is Necessitating: Given the same circumstances, the same effects must obtain. This idea goes back to the first Stoic, Zeno of Citium

  'It is impossible that the cause be present yet that of which it is the cause not obtain' (ἀδύνατον δ’ εἴναι τὸ μὲν αἴτιον παρεῖναι, οὖ δέ ἐστιν αἴτιον μὴ ὑπάρχειν).

• General determinism:

  General determinism holds that every counterfactual state or event is forever impossible and, correlatively, that every factual state or event is forever necessary. An example of a state that is subject to this kind of necessity — which I shall call `general' necessity — is that expressed by the factual proposition snow is cold. As a matter of fact, snow is always cold, and it cannot be hot so long as it remains snow. This proposition expresses a state that does seem to be subject to a general necessity, hut is every factual state and event subject to this kind of necessity? According to general determinism the answer should be in the affirmative. (Salles, p.xiv)

• Crude fatalism:

  Crude fatalism departs from general determinism in that it is compatible with the possibility of change. In particular, the crude fatalist plainly accepts that the future may differ from the present and the present from the past. However, according to crude fatalism, the future is already fixed in a way that what is due to happen, or be the case, will happen regardless of what states or events obtain in the present or the past. For example, if I am ill but am due to recover, then I will recover whether or not I call in a doctor and follow his prescriptions. More generally, the obtention of states and events at a particular time is not dependent upon the obtention of earlier states or events — a point that one may express by saying that the former would have obtained even if, per impossibile, the latter hadn't. This sheds light on an important aspect of crude fatalism: factual states and events at a particular time do not obtain because of the states or events that obtained earlier. There is no explanatory relation between past, present and future. To pursue the example, if I do call in the doctor and recover from illness, then, given that I would have recovered even if I had not called in a doctor, I did not recover because I called in the doctor. (Salles, p.xv)

• External determinism

  External determinism maintains that any event or state is contingent upon earlier events or states. The distinctive claim of external determinism is that the prior causes of what we do and of what we are may all be traced back to things that are external to us: our present environment, our teachers, our family and even the biological make-up of our ancestors. In consequence, everything we do and everything we are is, in fact, ultimately fully determined by external causes alone. Now, external determinism does not seem to be compatible with responsibility, either in a moral or in a legal sense of the term 'responsibility'. To take one example, I cannot be morally blamed for having missed my daughter's graduation ceremony if the reason why I missed it is that I was kidnapped by some ruffians. There is one exception to this principle: I may be responsible for something that happens to me, if it happens to me as a result of some earlier thing that I did and for which I am responsible. For example, I am morally blameworthy for being at the mercy of the ruffians who kidnapped me if the cause of my kidnap was, to some extent, my lack of care in circumstances that I knew were dangerous. However, this is precisely what external determinism denies is possible. According to external determinism, all the states or events that we supposedly bring about, or to the production of which we supposedly contribute, are ultimately fully determined by causes external to us. My lack of care, the external determinist would argue, is itself ultimately fully determined by external factors alone.

  Some philosophers have argued that any form of causal determinism is, willingly or not, a form of external determinism. In fact, by the time of Aristotle, 'force', or fully external determination (βία), was already one of the connotations of the term 'necessity' (ἀνάγκη) and its cognates. (Salles, p.xvii)

The philosophical question addressed is whether this `internality requirement', as I shall call it, can be met in a world governed by determinism. One major objection against compatibilism turns on this issue. The objection (henceforth the 'externalist objection') is that, if every state and event is determined by prior causes, then everything we do is in fact fully determined by external factors alone. In consequence, the internality requirement cannot be met and causal determinism would remove any possible ground for the justified ascription of responsibility. But is the externalist objection cogent? A central compatibilist argument developed by Chrysippus was designed to rebut it. On his view 'everything is determined by prior causes' does not have to imply that we are always at the mercy of purely external forces. The internality requirement, which is a necessary condition for responsibility (either legal or moral, as Aristotle claims), can in some relevant cases be perfectly met in a world governed by determinism. (Salles, p.33)

Frankfurt and Chrysippus explain moral responsibility by appealing to factors that are substantially the same. In Frankfurt's theory, the responsibility for the action derives from the agent's decision to perform it, but also from that decision's being based on a previous all-things-considered practical reflection. Similarly, the responsibility for the action in Chrysippus derives from the agent's exercise of an impulse for it (or his assenting to the impression where the action is presented as valuable), but also, and crucially, from the impulse's being fully rational, which involves a reflection concerning the all-things-considered desirability or appropriateness of the action. It is noteworthy that in some of his later works Frankfurt's account lays a certain emphasis on second-order desires: To be responsible for Φ-ing, it is sufficient to desire having the desire to Φ, provided that the former, second-order, desire is based on a previous practical reflection concerning the desirability of the latter: the agent is responsible because he came to have the desire to desire to Φ as a result of a reflection about whether the desire to Φ is worth having. Therefore, the question addressed in the reflection is mainly 'should I desire to Φ?' Its focus is a desire — whether or not one should have it. In the Chrysippean account, by contrast, the focus of the reflection is on action. As we have seen, the question it addresses is 'Is it appropriate for me to Φ (given the present circumstances)?' (Salles, p.66)

I Φ is capable of being true if I am fit, or strong, enough to Φ; and it is capable of being false if I have the physical strength to refrain from Φ-ing. As for the other condition — being or not being prevented by external factors (τὰ ἐκτός) from being true or false — it refers to the presence or absence of factors external to us that either prevent us from acting in a certain way or force an event or state to take place at us. In contrast with the former condition, the latter is an innovation of Chrysippus.

The non-necessity of a proposition in this modal system is compatible with there being necessitating causes for the event in question. Consider a situation where my action is to stand still. I now have the intrinsic fitness required for walking and nothing external prevents me from doing so. Therefore, the proposition I stand still now is non-necessary in the sense envisaged by this modal system. Yet, my standing still is causally necessitated, namely by the whole rational process by which I came to the conclusion that I should remain still and that caused me to act accordingly. In other words, the proposition I stand still now is non-necessary in Chrysippus' modal system, even though my action is, at the same time, necessary in a causal sense. In fact, as has been hypothesized in recent scholarship, there seems to be two kinds, or at least senses, of necessity in Chrysippean Stoicism. One sense is that required by the modal system just described, whose aim, as I shall argue in some detail later on in this section, is twofold: (i) to establish that some states and events that are counterfactual at all times are nevertheless possible; (ii) to preserve the interdefinability of the four central modal notions. It follows from (i) and (ii) that a factual action whose opposite is counterfactual at all times but possible is, thereby, non-necessary.

The other sense of necessity is that required by Stoic causation, according to which an effect is necessitated by its cause — an idea that Chrysippus never contradicted and that goes back, as we know, to Zeno: 'it is impossible that the cause be present yet that of which it is the cause not obtain' (ἀδύνατον δ’ εἴναι τὸ μὲν αἴτιον παρεῖναι, οὖ δέ ἐστιν αἴτιον μὴ ὑπάρχειν). Thus, a factual event necessitated by its cause may nevertheless be non-necessary from the point of view of Chrysippus' modal system. There is no contradiction as long as we bear in mind that these are two different kinds or senses of necessity that were not meant by Chrysippus to be equivalent to each other. (Salles, pp.82-84)