Note: other papers in this cycle can be found in my web page Time and Supervenience in the Theories of Value and Free-Will

by Robert L. Pendleton

R. M. Hare is credited with adapting the term "supervenience" to describe the presumed metaphysical relationship between moral facts and natural facts. Since Hare's introduction of the term, supervenience has achieved a much wider use, chiefly to characterize the various notions of mind-body dependence. But the basic idea of supervenience can still be helpfully illustrated through Hare's original concept of moral supervenience. Moral facts, while never identical to any natural properties, still bear a relationship of logical dependence on those natural facts in the following way: if we judge a collection of natural facts to be "good", it seems that we contradict ourselves if later we judge those unaltered natural facts to be changed as to their "goodness". Moral facts thus supervene on natural facts in this sense.

I have argued elsewhere that this useful notion of supervenience might fruitfully be employed in resolving some of the puzzles associated with the metaphysics of time (Pendleton,1999a). I have advanced the view that time, though in one sense unreal, nevertheless supervenes on change, which is real. The resulting metaphysics of "time as change" echoes a view of time espoused by figures as diverse as Arthur Prior, Roderick Chisholm, Franz Brentano and St. Augustine. This paper pursues one of the key advantages of this view of supervenient time. I hope to show that, if time supervenes on change, then the lawlike natural world can be shown to be compatible with a meaningful view of free-will. In this paper I provide a more detailed account of this argument for free-will based on the supervenience of time on change.

First a brief sketch of the argument. It is a commonplace within philosophy to hypothesize "deterministic" constraints on human freedom resulting from our position as creatures within a lawlike, natural world. However, in the face of these presumed constraints our experience presents us with a puzzling anomaly. Apart from the familiar limitations posed by our coexisting physical environment, we generally think of ourselves as "free." Of course, philosophers have been quick to challenge this kind of naive empiricism, proposing various arguments to show that deterministic constraints, though not experienced, might still be real. The metaphysics of time is involved in this debate because many familiar arguments proposed to justify deterministic constraints appear to assume a certain picture of time. For example, deterministic constraints are imagined to occur because they are the "necessary temporal antecedents" of our lawlike behavior. It is reasoned that, even if not experienced, these necessary temporal antecedents might be theoretically justified as constraints on the basis of their analogy to the more familiar constraints we experience in dealing with our presently existing physical environment. In another kind of argument, the existence of past facts plus operative natural laws is thought to necessitate, in a logically binding sense, some or all of the facts that are presently the case, indeed, facts that will be the case as well. Here again, it appears that a view of time comes into play in this "logical entailment" of the present and the future by the past.

The idea of an analogy between temporally antecedent constraints and presently existing constraints has already been shown to be a deeply flawed notion by Smart (1967, 126) and Newman (1988, 545). \ 9/ But such a view of time continues to tempt many philosophers who view free-will as incompatible with "determinism." For this reason it is an important advantage of the view that time supervenes on change that this seductive, but false, picture of time is ruled out from the start. Arthur Prior put the ontological matter in stark relief when he stated,

If the past is not real, there can be no physical relation between that past and the present. Nor can there be a physical relation between the present and the future that is yet to be. The "lawlikeness" that seems to link the past, present, and future must therefore be conceived as a monadic property of changing things rather than a subtle relationship between the present and its imagined past or future. As a consequence, the past (or the future) cannot be thought to present a threat to freedom in any way analogous to those physical constraints familiar to us in dealing with the present. Actual constraints achieve their constraining character by their continuing physical relationships to us in the present, that is, as coexisting objects.

Furthermore, if we take stock of those coexisting physical properties we actually come to see as constraints on our freedom it is not clear that the lawlikeness of those constraining properties has any special role to play. While it is no doubt true that many constraints are due to lawlike phenomena, constraints can just as easily result from random, or anomalous, phenomena. When pondering the idea of lawlike natural constraints, Wittgenstein is said to have pictured railroad tracks that suddenly "change their shape" (Smythies, 1989, 85). It appears that an indeterminate railroad, though less predictable, might still constrain. On the other hand, we cannot even say with confidence that the elimination of our co-existing lawlike environment would result in freedom. Real, palpable, constraints on freedom can arise from the absence of a co-existing lawlike environment that would normally enable free action (think of the hapless astronaut on a space walk).

Supervenient time also renders powerless the idea that constraints on human freedom might follow as logically necessary consequences of past facts and the existence of natural laws. If time supervenes on change, it is difficult to express coherently the idea of a logically necessary consequence spanning time. In Prior’s tense grammar, the idea that past conditions plus natural laws can logically entail a future consequence comes out something like: P (p & (p -> Fq ) ) -> P (Fq). The meaning of the modal consequence "P(Fq)" is open to question in a way that a simple, non-modal consequence "q" would not be. As a result, a whole range of incompatibilist arguments, of which perhaps Peter van Inwagen's "Consequence Argument" is the most notorious, come into question. (1983, 16)

The preceding sketch suggests, but does not make explicit, a fairly complex argument involving premises ranging from epistemology to human psychology. In this section I flesh out the argument and its major assumptions.

First, I assume that freedoms are meaningful only in relation to real or imagined constraints. Our experience of freedoms is always conceived in relation to these presupposed mechanisms of constraint. I propose a distinction between three possible natural mechanisms that might be thought to have a constraining effect for the human self or will in a wholly or partially determined world:

These three mechanisms cover constraints from external factors (conceived as either positive or negative factors) in space and time, and also putative constraints from internal factors, i.e., from a subject’s own lawlike (or non-lawlike) properties. To my mind, this exhausts the possibilities. I use this three-fold analysis of the possible mechanisms of constraint to support a compatibilist argument based in part on our view of supervenient time:

1) A natural theory of freedom, to be relevant as a theory of freedom, must posit theoretical phenomena or objects corresponding in some sense to a domain of experienced freedoms. And by our assumptions, therefore, a natural theory of freedom must start with freedoms defined in relation to a domain of experienced mechanisms of constraint.

2) Arguably, only type 1 (environmental) constraints are involved in clear-cut, experienced constraints on freedom. I argue that constraints of types 2 and 3 are in fact not experienced. For this reason, a natural theory of freedom must start with theoretical freedoms defined relative to experienced type 1 mechanisms of constraint.

3) Even though mechanisms of types 2 and 3 are not experienced, examples from the natural sciences suggest that we might justify the inclusion of type 2 and 3 constraints in our natural theory of freedom through arguments demonstrating either a logical or analogical connection between such mechanisms and the experienced domain of type 1 constraints.

4) The view that Type 1 constraints are logically tied to of type-2 or type-3 constraints is not supported by our experience of type 1 constraints. Examples of type 1 constraints do not seem to suggest the kind of uniform natural correlation to lawlike phenomena that would lead us to posit a logical relationship between type 1 constraints and supposed constraints of types 2 or 3.

5) The proposed analogy of type 2 constraints (necessary temporal antecedents) to type 1 constraints is shown to be doubtful with the help of a recent argument by Newman (1988, p. 545), and an earlier argument from Smart (1967, Vol. 8, p. 126). It is also inconsistent with the metaphysics of time as supervenient on change.

6) The proposed analogy of type 3 constraints (subsumption under natural laws) to type 1 constraints violates the verifiability conditions that characterize experienced Type 1 constraints.

7) We are therefore left with type 1 mechanisms of constraint as the sole potential source of the deterministic constraints posited by incompatibilists. However, type 1 constraints, based on our experience of them, do not appear to be of a nature to cause the unremitting frustration of free choice that we associate with the problem of free-will and determinism. As previously remarked, lawlikeness has no special role to play in determining which environmental (co-present) factors we come to see as constraints on our freedom. Nor do constraints based on non-lawlike, or "absent" (negative) physical factors (as in the case of the hapless astronaut) appear to play such a "deterministic" role. We must therefore conclude that, aside from possible type-1 constraints of which we may be ignorant, we are free in roughly the manner and extent to which we normally think ourselves to be free.

My proposal to base the issue of free-will upon empirical considerations runs counter to many recent incompatibilist studies of the free-will problem, which are characterized by rigorous, logical argument. Van Inwagen’s ‘consequence argument’ is typical of this trend (1983, 16):

Such logical arguments demonstrate—it seems fair to say—that our current concepts of freedom are logically incompatible with our current concepts of determinism. Incompatibilists conclude from this that the respective facts of freedom and determinism are similarly incompatible. The underlying assumption appears to be that any acceptable theory of freedom must be incompatible with any acceptable theory of determinism. This latter thesis, in turn, cannot stand unless we assume that there is an analytic relationship between our current concepts of freedom and any acceptable theoretical concepts of freedom.

It is of interest that this logicism with regard to theories of freedom seems very much like the doctrine proposed by G. E. Moore for theories in ethics. Moore, it will be recalled, rejected all but analytic definitions of the presumed attribute Good. Violating this rule with respect to the definition of good Moore termed the naturalistic fallacy. However, this doctrine appears a bit strained when Moore attempts to extend it beyond ethics to the theory of ‘yellow’. In that case Moore flatly rejects the definition of ‘yellow’ provided by the electromagnetic theory of light, a theory that is universally accepted by scientists in our day, as it was in Moore’s own time. This seems a remarkable position for a ‘common sense’ philosopher to take, but Moore is quite explicit in his rejection (Moore 1903, 10):

If we are seeking a natural theory of something like ‘yellow’, there is an obvious problem in Moore’s method. Although we may be persuaded that good is not to be found among the features of the natural world, I suspect that as few philosophers in our day as in Moore’s time would question the electromagnetic theory of light as providing the basis for an acceptable theory of ‘yellow’. Still fewer would find merit in Moore’s stated reason for his rejection: that such scientific definitions are not "what we mean by yellow."

This aspect of Moore's philosophy, which is often evident in Principia Ethica, appears to involve a confusion of Sense and Reference. Gilbert Ryle, in discussing Moore's earlier philosophy, puts the matter quite bluntly:

In the construction of theories of real things in the natural world, it seems clear that a method which employs "scientists with telescopes" is to be preferred to a method which employs "logicians without telescopes."4 If natural science is our goal, Moore’s famous criterion of analyticity for theoretical definitions must be replaced with a criterion more suited to the kind of empirical investigations involved in establishing the use of terms and in discovering the possible referents of such use. This new criterion, I propose, should be the requirement of rough, or prima facie, co-extensionality. Let us say that a necessary condition of a valid theoretical definition in a natural science is that the proposed theoretical referent must stand to the explained domain in a relation of rough co-extensionality. That is to say, propositions experienced to be true in the domain of explananda must be roughly paralleled by verifiable propositions in the domain of the explanans. Theoretical definitions of this type have been noted by Sklar (1993, pp. 114-16) in his distinction between "scientific identificatory reduction" and "philosophical reduction." They can also be seen in Quine’s "proxy functions." Proxy functions reinterpret our observation sentences but retain the latter’s experiential support. Their truth depends on the verifiable one-to-one correspondence of the related functions, not on the relation of the "values of the variables" or "objects" claimed by each interpretation (Quine 1992, p. 8). Such non-analytic definitions are also consistent with the results of Kuhn with regard to the "incommensurability" of scientific paradigms. Statements bridging successive paradigms cannot be expected to exhibit analytic relationships in the normal, intratheoretic sense (Kuhn 1970, p. 120).

While differing from Moore’s criterion of analyticity, the proposed criterion of co-extensionality serves the same intent and function: it satisfies the theoretical requirement of explanatory relevance. The propositions of natural theory of yellow color must, at a minimum, exhibit a rough correspondence with some domain of experienced yellow objects. Only after that experiential criterion is satisfied can we begin to think of the proposed theoretical domain as constituting a relevant theory of yellow. Similarly, the propositions of natural theory of freedom must, at a minimum, exhibit a rough correspondence with some domain of experienced freedoms. Only after that experiential criterion is satisfied can we begin to think of the proposed theoretical domain as constituting a relevant theory of freedom.

A natural theory of freedom need not be limited to naively experienced freedoms. A scientific theory often greatly expands the original domain which it explains, revealing natural connections among phenomena never anticipated prior to the creation of the theory. For example, the electromagnetic theory of light allows us to imagine domains of ‘color’ not visible to the normal eye. Because of the likeness, the analogy between visible light waves and these invisible domains, we can speak with justification of "ultra-violet" and "infra-red" as "colors" of which we can have no visual experience. But this extension of the electromagnetic theory of colors beyond the visible spectrum can be countenanced only because the prior criterion has been satisfied linking the electromagnetic theory of light to our experience of colors. In a straightforward way we may apply this principle to a natural theory of freedom. We may indeed consider extensions of a theory of freedoms/contraints—through logical or analogical reasoning—to domains of freedoms and constraints that go beyond our experience. But we can do this only if we have already established a theory of freedom that corresponds to our experience of freedom.

First a brief sketch of what I take to be type-1 "environmental" constraints. Of the three types of constraints of this category are arguably the most real to us, the easiest to understand. Indeed, one form of the worry about natural determinism seems to be based on the unknown influences of this commonplace type of constraint. That is, the co-existing, natural environment in which human beings are created, nurtured, and developed is often thought to cast doubt on the freedom, or autonomy, of the human self or ‘will’ that results. This doubt seems to be increased if the environment, and human psychology itself, is conceived to be largely or wholly ‘determined’—merely a behavioral response to environmental stimuli. In so thinking of environmental constraints we focus on contemporaneous natural forces comprising the interaction of one physical entity with another. This type of relationship is clearly in evidence in a very familiar example of causation:

This type of physical interaction looms large in any analysis of human freedom, for it seems to be involved in many cases where we would unequivocally ascribe a constraint on human freedom: for example, a prisoner in a jail cell. Rousseau’s classic "[the human being] is born free; and everywhere [She] is in chains," attests to the virtually criterial character of this kind of type-1 mechanism of constraint (being unfree is like being "in chains"). We assume that we are not always "in chains." However, by extending this clear-cut type of environmental constraint, could it not be argued that, with all the lawlike, environmental factors subtly shaping and forming our choices, our free-will might not be as free as it first appears?

Quite unexpectedly, this worry about the "criterial" character of impending natural forces in creating constraints on freedom turns out to be groundless. Although type-1 constraints of this type form an undeniable domain of experienced constraints, they cannot by themselves amount to the kind of ubiquitous deterministic constraint imagined by incompatibilists. Just a cursory reflection on the types of environmental relationships that form our routine experience will disclose that, even when such relationships are involved, we cannot always ascribe a constraint on freedom.

Take just two examples. First, our human perception is made possible by the nearly continuous environmental relationships between our sense organs and the natural world around us. But we would not normally think of our perceptual processes, so affected by, so dependent on, our environment, as constraining, or limiting, our freedom. Quite the contrary: we would be more likely to look upon the absence of these environmental relationships as the true constraint on our freedom. Imagine a group of people at a party. Suddenly, everyone is stumbling about because a prankster has mischievously turned off the lights. These partygoers are constrained in a very straightforward, environmental sense. But they are not constrained by a positive physical condition that inhibits their actions. Rather they are constrained by a negative condition, the absence of the light, which would enable them to act in a normal manner. This kind of negative constraint is far from unusual. Consider again the helplessness of an astronaut attempting a zero-gravity spacewalk. The constant gravitational interaction with the earth under our feet is unremitting, assumed in all we do, and, indeed, limiting to our freedom at times. But freedom is not to be gained by simply removing this environmental relationship. It is not the existence of the gravitational relationship per se that results in a constraint on freedom.

These examples suggest that it is often a "negative" environmental factor that results in a true environmental constraint on freedom.5 The prevalence of this kind of negative environmental constraint means that we cannot conclude from the mere existence of positive environmental relationships affecting the human will that such relationships constitute a constraint on freedom. We cannot, in other words, define constraints in terms of the more familiar positive environmental constraints, however "criterial" these positive constraints may seem at times.

A second reason to doubt that type-1 constraints might constitute deterministic constraints is given by our example of Wittgenstein's reported "indeterministic" railroad. In considering the notion that natural laws might constrain in the way that railroad tracks are conceived to do, Wittgenstein is said to have remarked that railroad tracks do not as a rule "suddenly change their shape". Wittgenstein deftly undermines the tie that we normally make between familiar type-1 constraints and the presumed lawlike nature of those constraints. An indeterministic railroad might in fact be the type-1 constraint. And a lawlike railroad might enable us to travel to Philadelphia. If we restrict our attention to the domain of type-1 constraints, an indeterministic world is not necessarily more free for the human will. Nor is a lawlike world necessarily more constraining for that human will. We cannot assume that constraints on human freedom must in all cases result from the existence of the human self, or will, within a ‘determined’, lawlike environment.

I would be the first to admit that the Type-1 Constraint is something of a "straw model" from the point of view of the actual arguments put forward on this subject by incompatibilists. Our ‘criterial’ prisoner in the jail cell illustrates the point: if we maintain that the prisoner is less free when the cell door is locked, then it is natural to reason that the prisoner is more free when the door to the cell is left unlocked. Such verifiable freedom conditions are part of what it is to be able to claim that the jail cell is a verifiable constraint on freedom in the first place. If, upon unlocking the cell door, the prisoner were found on closer inspection to be a statue, we would be less likely to conceive of the jail cell as a constraint!

But for incompatibilists the freedom of natural human wills in a determined world is always illusory, regardless of the co-existing environmental relationships that may or may not obtain at any particular moment. Citing modern physical theories, they remind us, for example, that our criterial prisoner, though freed from the jail cell, is still not free if subject to natural laws, as is assumed by the thesis of determinism.

Compatibilists have instinctively resisted this extension of the concept of a constraint on freedom. For them type-1 concepts of freedom are primary, even ‘criterial’. But on what does the primacy of the compatibilist’s ‘criterial’ concept of freedom rest? To no one’s surprise, primacy arguments based solely on the compatibilists’ own characteristic intuitions of freedom have not been persuasive for incompatibilists, who are quite at home with the long philosophical tradition tying the concept of freedom to the broader context of natural law and causation.

However, from the naturalist perspective that favors "scientists with telescopes" over "logicians without telescopes", compatibilists have more than intuition on their side. The preference for type-1 concepts of freedom can be seen as a reflection of the naturalist requirement, at least prima facie, for a domain of experienced freedoms. Whatever their limitations, Type 1 constraints and their correlated freedoms do meet that experiential requirement. What the compatibilist must question is the experiential status of the other presumed types of constraints thought by incompatibilists to follow as consequences of determinism.

My contention that ‘determining’ natural laws or temporal antecedents are not experienced as constraints on freedom may seem debatable to some. Incompatibilists have pointed to problematic cases of human psychology as evidence of possible constraints by ‘determined’ psychological forces. For example, obsessions like kleptomania and drug addiction have been cited as cases warranting mitigation of responsibility, and thus as cases of diminished freedom of the agents so afflicted (Watson 1982, p. 97). Indeed, the psychological problem of ‘weak will’ has been a puzzle to philosophers since the beginning of the discipline. However, in view of the known complexities and divisions possible within the human mind, cases of ‘psychological constraints’ on freedom might with equal cogency be explained on a ‘psycho-environmental’ model. If we are to countenance, with Proust, or Freud, the fragmentation of the self, we must be open to the view that not all the ‘willings’ that emerge from our biological envelope are as connected with that core ‘self’ with which we would identify, as are others. There seems no compelling reason to regard psychological constraints on freedom as other than a subtle form of Type 1 environmental constraint. 6

Furthermore, it is puzzling that incompatibilists must resort to unusual cases of psychological pathology to provide evidence of experienced constraints resulting from ‘deterministic’ sources. One would think, with all the lawlike behavior of the natural components of the human will and psyche, experience of deterministic constraints ought to be an everyday phenomenon for even the healthiest of human psyches. We are, after all, entirely composed of molecules, cells, muscles, and neurons—materials which are all assumed to conform to natural laws of some sort or other. Why can we not readily perceive the Type 2 and 3 constraints resulting from this determined natural behavior? Are we just too accustomed to our ‘bondage’ to notice?

We cannot resolve this puzzle—as is often proposed—on the ground that the subatomic world is ‘indeterministic’. Sir Arthur Stanley Eddington, early in this century, was among the proponents of the theory of ‘freedom based on indeterminism’ (Nerlich 1967, p. 458). Yet Eddington unwittingly provides an effective rebuttal to this position in his own celebrated quip that the likelihood that the randomly ordered gas molecules in a cubic foot of air might end up at any instant in just one half of the space is far more remote than the probability that an ‘army of monkeys’ typing away randomly might someday complete all the books in the British Museum! (Eddington 1930, 71-72) If lawlike certainties are truly constraints, why would not a human self or will behaving according to this kind of indeterministic but overwhelming probability experience a constant, unremitting constraint on freedom?

In the philosophical tradition surrounding this problem, the metaphysical faculty of ‘free-will’ has often been introduced, perhaps in part to account for our puzzling lack of experience of these ‘deterministic’ constraints. Kant, for example, relying on his metaphysics of the transcendent Ding an sich, leaves open the possibility that, although we must appear to ourselves as heteronomous, and thus unfree, we are in a higher, transcendent sense autonomous, and thus free. But there are several problems with this. First, if it is true that we must always appear to ourselves as unfree, whence comes our understanding of our transcendental freedom? A second problem is that the metaphysical faculty of free-will does not appear to have a broad enough application to eliminate the puzzle entirely. For, in the traditional theory ‘lower’ animals are generally denied free-will. But if such ‘lower’ animals lack free-will, should we not expect to observe a constrained behavior on their part? (A caged tiger does, after all, exhibit a marked behavior.) Lower animals display no indication of constraints beyond the common type-1 constraints familiar to human beings. Are we to believe that these animals are deterministically constrained but have no experience of their constraints? Or must we say that the lower animals possess a metaphysical "free-will" too? Kant's transcendental free-will appears to serve no purpose but to obscure the empirical weaknesses of the incompatibilists’ concept of deterministic constraints. The removal of the Kantian faculty of free-will is thus not so much a threat to freedom, but rather the elimination of an unnecessary, superfluous element in a natural theory of freedom.

One way of requiring the inclusion of deterministic constraints in a natural theory of constraints is to claim that the three suggested types of constraint are in fact not logically distinct. If that were so, it would seem that deterministic natural laws might make constraints out of environmental conditions that otherwise would not be so. First, a word of caution: we are not primarily concerned here with the real or claimed logical relationships between our current concepts of experienced constraints and our current concepts of deterministic phenomena. A method that relies on "scientists with telescopes" must be concerned to evaluate the facts relating to experienced constraints to see if their existence, in any likely theory of freedom, might turn out to entail, or be entailed by, deterministic or lawlike facts.

We first consider the view that experienced constraints might be found in every case to entail a deterministic substrate. One argument is suggested by the existence of what we have heretofore termed "negative" type-1 constraints. In the case of such negative constraints we are prevented from achieving some willed objective by the absence of a key environmental factor. We are "blind" because the lights have failed, or "trapped" in space by the absence of rocket fuel. But, it cannot be maintained that negative constraints are tied to their ‘deterministic’ character (there being literally nothing to be determined in such cases)?

A second argument is derived from a point by Wittgenstein. Considering the idea of natural laws as "...rails, along which things had to move,..." Wittgenstein cautions:

Wittgenstein provides the image of what we might call an indeterminate railroad to counter our usual assumptions about natural laws and constraining physical objects. Constraining physical objects differ from natural laws in that they can conceivably change. Nevertheless, a train can be conceived to be physically constrained by type-1 constraints in the form of railway tracks, even if those tracks fail to obey natural laws. It is a mistake to assume that this kind of example might be limited to cartoonish flights of fancy. Since the advent of quantum physics, the accepted view of the actual behavior of physical objects like railway tracks is much more like the ‘indeterminate railway’ than we may imagine. The suddenly moving rails of Wittgenstein’s example are not so much counter to the theory of statistical mechanics as they are highly improbable according to that theory. In a world of quantum mechanics an ‘indeterminate railway’ can constrain as effectively as a ‘classic railway’ obeying the laws of Newton. It is thus unlikely that deterministic factors might come to be conceived as logically necessary conditions for type-1 constraints on freedom in a natural theory of freedom.

The second type of logical relationship that we must confront in this regard is the claim that type-1 constraints are entailed by constraints of types 2 and/or 3. It is quickly seen that from the existence of a type 1 constraint and the fact that such constraints are entailed by type 2 and/or type 3 constraints, we cannot infer the existence of types 2 and/or 3 constraints. Nevertheless, it seems that we could infer negatively, by modus tollens, that if freedom from type 1 constraints exists, then types 2 and/or 3 mechanisms do not exist. This indeed seems a significant "armchair" discovery until we remember that our scientific method requires that the logical entailment relationship be established not through linguistic scrutiny so much as through empirical investigation. If we subject this presumed entailment relationship to such empirical scrutiny, it is far less clear that the kind of unvarying natural correlation exists to support the hypothesis of such a "logical" relationship. Indeed, some of the examples we have given of negative environmental constraints seem to indicate that some relationships with our lawlike environment are invariably connected with experienced freedoms. It is their absence which is correlated with experienced constraints. Moreover, our previous arguments that types 2 and 3 constraints are not experienced, can be seen in one sense as a counter-argument to the view that our lawlike nature entails constraints on our freedom. The knowledge that we are, in all likelihood, lawlike beings, and yet do not experience our lawlikeness as a constraint serves as an prima facie rebuttal to the notion that, in a natural theory of constraints, there might come to be thought a relation of logical entailment between type 2 and 3 constraints and type 1 constraints.

We know that the idea of the type-2 necessary temporal antecedent (NTA) as a constraint has weaknesses not only with respect to direct experiential support, but also in terms of its supposed logical connection to the more familiar type-1 constraints. The only remaining task with regard to NTA’s is to evaluate the notion of an analogy between such determining temporal antecedents and type-1 constraints. However, this idea of an NTA analogy falls apart under detailed scrutiny. Three specific arguments can be raised against it.

First, whereas environmental conditions are by definition contemporaneous with the bodies they affect, temporal antecedents are by definition prior in time to the bodies they affect. This leaves us with the anomaly that, in the case of NTA, the lawlike human self or will is constrained by conditions which no longer exist. Our theory of supervenient time can be seen in this argument. Necessary temporal antecedents can only be imagined to be analogous to present physical relationships in those theories of time which grant some sort of existence to past or future events. The "presentist" view of time does not, however, eliminate lawlike attributes. But such lawlike attributes must be re-conceived as monadic properties that do not depend on the notion of a physical relationship between the existing present and the unreal past or future. Thus recast, lawlike properties re-emerge as potential type-3 constraints which I shall address in the following section.

Although the case against NTA's as constraints is helped by a "presentist" view of time, not all the arguments against NTA's rely on this admittedly controversial time theory. Another problem with the NTA analogy arises from the possibility of negative type-1 constraints. Because type-1 constraints can be negative, we cannot argue that positive NTA’s must be, by analogy, constraints on freedom. For all we know, such NTA's may be among the enabling conditions for freedom. Nor can we assume that a human self or will without positive NTA’s would for that reason be free in the required analogical sense.

Attempting to view NTA's as analogous to type-1 constraints has always involved problems of coherence. Two recent arguments are worth mentioning in this regard. J. J. C. Smart (1967, p. 126) focuses on the problematic notion of ‘movement in time’. Any analogy between our ‘movement’ through time and our ‘movement’ through space, Smart argues, runs up against the difficulty that travel through space presupposes temporality. A truly analogous ‘Movement’ through time would thus require an additional ‘temporal’ dimension to be truly analogous. Movement through that new ‘time’ would in turn require still another ‘temporal’ dimension, and so on ad infinitum. This argument works just as well if we substitute ‘interaction’ for ‘movement’. Interaction in space presupposes temporality. To suppose that the necessary time relationships assumed by the NTA position are analogous to such spatial interactive relationships involves us in the same infinite regress.

A more recent argument is provided by Andrew Newman in the course of his discussion of the notion of the causal relationship as a temporal relationship (Newman 1988, p. 545):

For the purposes of my argument, I do not need to claim, as Newman appears to, that the causal relationship cannot be a temporal relationship between temporally successive causes and effects. I have only to prove that this kind of temporal causation, if it exists, is not analogous to the kinds of interactions which are associated with type-1 constraints. It seems quite clear that Newman's argument provides ample proof that this supposed analogy is incoherent. But it is just such an analogy that we require of the NTA, according to our naturalist method, to provide a justification for the inclusion of this kind of unexperienced mechanism of constraint in our natural theory of freedom.

From these arguments we must conclude that the argument for NTA constraints from analogy cannot stand. This will no doubt be looked upon as a paradoxical result, perhaps the most counterintuitive position advanced in this paper. Nevertheless, I believe it to be a sound result. It is counterintuitive, I believe, for the reason that the presumed ‘relationship’ between the successive times of natural objects is inadequately understood. This has the consequence that lawlike properties, and thus determinism itself, are the objects of similar misunderstandings. I must, however, defer consideration of this deeper problem—which is at bottom a misunderstanding of time—to a separate paper (Pendleton, 1999a)

As with necessary temporal antecedents, we are left with the possibility that a demonstrated analogy might justify the inclusion of type-3 constraints within a natural theory of freedom. As we have remarked, to avoid the fate of the NTA, this type-3 analogy must be based on a conception of lawlike behavior other than that of a constraining NTA relationship (an analogy which we have now discredited). Let us consider what alternatives there might be to this flawed NTA picture. We must admit that lawlike behavior is tied to the existence of certain characteristic properties over time. Could these characteristic lawlike properties comprise an alternate concept of a subsumption constraint? Does the possession of such ‘lawlike’ properties by the natural human self or will constitute a constraint on the freedom of the self or will in any sense analogous to our experienced base of type-1 environmental constraints?

Such an analogy between type-3 and type-1 constraints may appear to be the case due to the apparent likeness of the picture of a "constraint by one’s lawlike willing properties" to an important class of type-1 constraints wherein properties seem indeed to constrain. For example, if I imagine myself as palpably obese, it's easy imagine myself as constrained in my actions by my bodily properties in the same way as I might be by my physical environment. My imagined properties of weight and bulk might ‘weigh me down’ much as if I were wearing a heavy coat. I am forced to admit such ‘alienable’ properties as obesity as potential environmental constraints because I cannot distinguish in a non-arbitrary way between such ‘alienable’ properties and the constraining properties of my contemporaneous environment.

On the other hand, there are some properties—and we may be tempted to call them essential properties—that do not seem to fit as easily into this pattern. We do indeed want to say that as an obese man I am constrained when my choices, my will, are hindered by my imagined immense weight. But could we say the same of my willing itself? Can we say that I am constrained by my very possession of acting or willing properties? My intention, of course, is to argue that we cannot—a clear case of ‘language going on a holiday’. But if some properties, like willing, are not capable of constraining, how are we to give an account of this special, essential, character of willing and other non-constraining properties? We doubtless cannot mean, when we use the term ‘essential’ to characterize these significant properties, that there is a metaphysical distinction to be made in the case of properties which seem incapable of constraint. All of our properties, it would seem, must be on the same metaphysical footing.

The key to this puzzle is to be found, once again, in the naturalistic requirement of verifiability for experienced, environmental constraints. Property constraints are indeed quite common to our experience in the case of certain ‘alienable’ properties, as arguably a species of type-1 constraints. However, the verifiability of such alienable property constraints presupposes a core of constitutive properties in relation to which the alienable properties can be verified as constraints. If there were no such constitutive properties, there would be no way of confirming the constraining effect of the claimed constraining property (remember the case of the ‘prisoner’ who turns out to be a statue). This may be difficult for us to fully grasp because we have come to think of the self as a ‘punctual’ entity (to use Charles Taylor’s term for the modern, Lockean view of the self) in relation to which any natural property might be conceived as a constraint (Taylor 1989, pp. 159-176). Nevertheless, our naturalistic requirement of verifiability has the consequence that the properties constitutive of that self or will which we conceive to be subject to constraint, cannot themselves be construed as potential constraints of that same self or will, without equivocation on the definition of the self or will in each case. Verifiable property constraints therefore exhibit a relative structure that appears to prevent the possibility of ‘self-constraint’ in the case of certain properties of the constrained object. Experienced freedoms, and their correlative environmental constraints, are always relative to an object defined to be free (or constrained, as the case may be). Thus, by any environmental definition of freedom there will be a collection of presupposed ‘essential’ properties defining the object that is conceived to be free. These essential properties cannot be potential constraints for the object so defined as free, unless we change the definition of the object. Thus, although it may make sense to say that in my imagined obesity my acting and willing are impeded because I am overweight, it does not make sense to say of my acting and willing, that they are impeded by my acting and willing properties as such. This is not to suggest that the relative structure of freedom, or of constraints on freedom, is essentially tied up with the concept of a human conative faculty or will. Other perfectly valid instances of ‘freedom’ exhibit this same structure, but do not involve the notion of an animate will. We may say, for example, that the water behind the dam is constrained in relation to its tendency to flow down to the sea. But we cannot say of this same constrained water that it is also constrained by its own tendency to flow down to the sea. This ‘double bind’ is possible only by an equivocation on the definition of the water in each case.

Even if, by arbitrarily re-defining the human self, we were to countenance constraints by one’s own will, this ‘alienable’ willing property, as a species of type-1 environmental constraint, would share the same limitations. For example, negative constraints would appear to be just as possible for an ‘alienable’ will as in the case of more familiar environmental constraints. Consider Hamlet with an ‘alienable’ will: is he unable to act against evil because he is constrained by his indecision, or because he lacks the required will? Secondly, constraints by an ‘alienable’ will might have nothing whatever to do with the lawlike character of that will. Just as Wittgenstein’s ‘indeterminate railway’ can be imagined to constrain, so might an ‘indeterminate’ alienable will. If wills are things that might constrain, Hamlet might as well be thought to be constrained by an ‘indeterminate’ will as by a ‘determined’ will.

We can see from this that supposed constraints resulting from subsumption under natural laws—if based on the idea of constraints by one’s ‘alienable’ properties—do not comprise a mechanism of constraint that can distinguish the freedom, or lack of freedom, of a lawlike human agent from that of other presumed types of agent, including the incompatibilist’s ‘free agent’. Lawlike ‘alienable’ properties may indeed constrain, but so may ‘alienable’ indeterminate properties. And, in order to meet the verifiability requirements of the naturalistic method, both types of constraints, conceived as constraints on an agent, must presuppose a collection of ‘constitutive’ properties comprising a definition of that constrained agent. These defining properties, lawlike or not, cannot themselves be said to constrain the agent without equivocation.

Without recourse to the seductive but incoherent analogy of constraining temporal antecedents we can see that the issue of whether lawlike properties might constrain the human self or will has little to do with the traditional worry that determinism might constrain. The real problem of free-will is not the traditional problem of the physical place of the self or will in its natural, presumably lawlike, context, but rather the problem of the logical place of human willing properties within that complex, natural object which we label the self.

I shall conclude with some remarks on the method and the likely results of an investigation as to the logical place of such active, willing properties in the human self. First, in order that the issue not degenerate into a futile controversy over the meanings of words, we would do well to remind ourselves of the guidelines that I have heretofore labeled ‘naturalistic’ principles. We must, for example, replace appeals to linguistic intuition like "What is the Self?" and "What is the Will?" with empirical questions like "How do we use statements that attribute a ‘self’ and a ‘will’ to someone?" 7 Moreover, having thus sharpened our awareness of the meanings of our concepts of the ‘self’ and the ‘will’, we must not then succumb to the fallacy of assuming that a theory of the self and the will must exhibit logical relationships to those concepts.

As to the likely results of such an investigation into the self and the will, it will be very difficult, I believe, to sustain a theory of the self in which there is a possible division between the knowing and willing faculties of the self. The idea that such faculties might be distinct is at least as old as Aristotle’s practical syllogism. However, when the accessible facts of selves and wills are examined, it becomes clearer that the will, to be my will, must have practically unrestricted, transparent, access to my knowing faculty (the will must have its own ‘mind’, as it were). And my knowing faculty, to be truly identifiable as my mind, must be replete with active, willing characteristics (the knowing faculty must, it appears, have its own ‘will’).

Granted this thoroughgoing dualism of knowing and willing attributes in the knowing and willing "faculties," it becomes puzzling to conceive of a self comprised of both faculties as separate physical entities. Consider the odd picture that results: the will, or "passion" has at times been likened to a horse that one controls to carry out one’s acts. But a ‘will’ that does no more than extend my ‘reach’ is no more my will than is my right arm. The mind in turn has sometimes been pictured as a mere perceptual and logical machine for the convenience of the will ("reason is passion’s slave," as Hume quipped). But a mind that does no more than extend my ‘view’ is no more my mind than is my eyeball. We end up questioning, therefore, the viability of the dual faculty theory of the self, and with it the notion that my will can be alienated from me. If a will must have a mind in order to be a will, then an alienated will is more correctly viewed as an alienated self. From our common experience with ‘alienated wills’, this should come as no surprise. The passions, the obsessions, that we experience as alien to our true selves are quite naturally conceived by us as alienated ‘selves’, selves with which even the healthiest of psyches must deal on an daily basis. Furthermore, this picture of ‘alienated selves’ in no way vitiates the sense that we have of the unalienable will that is an integral part of our concept of the self. It in fact depends upon such a concept. An alienated ‘self’ could not be viewed has a constraint on the freedom of the true self if that true self did not itself possess a will that might be in opposition to the alienated ‘self’. From a naturalistic point of view, then, the real problem of free-will is to come to a correct understanding of the human self and its integral will. The question of whether and to what degree that self-and-integral-will is free is a complicated empirical matter. The philosophical contribution we can make to this empirical investigation is to note, as we have, that the potential verifiable constraints on that self-and-integral-will arise not from its lawlike properties, nor from its temporal antecedents. Verifiable constraints on the human self and its integral will are to be found without exception among Type-1 factors, whether lawlike or not.