From Molinism to Existentialism

A Treatise on the Metaphysics of Freedom, Agency and Volition

by John Davenport

Ph.D. Candidate

University of Notre Dame

I. Introduction: Molinism vs Existentialism

Alvin Plantinga's Curley example; intuitive warrant for the Molinist presupposition; the contrasting Existentialist position; the meaning of 'existence precedes essence;' two theses attributable to Heidegger; outline of the paper.

II. The Existentialist Critique of Molinism

(A) The Preliminary Objection

Tom Flint's definition of Molinism; the analogy between causal modality and Molinist feasibility; Plantingian transworld depravity; the 'ephemeral' nature of counterfactual power to do otherwise.

(B) The Haskar/Adams Anti-Molinist Strategy, and the Molinist Response

Comparison of the preliminary objection to Haskar's and Adams's critiques of Molinism. The difficulty with Haskar's 'Power Inference Principle;' Adams's "New Anti-Molinist Argument;" William Lane Craig's recent response to Adams; Craig's defense of the 'emphemeral' Molinist logical possibility of doing otherwise; the two stages of the Existentialist's alternative strategy against Molinism.

(C) A Theory of Synthetic Modalities and the Truths of Middle Knowledge

The error of misinterpreting Molinism as Essentialism; Flint's axiom: the truth of counterfactuals of freedom as logically contingent; the actuality of sub-maximal states of affairs 'prior' to the complete actuality of as the only possible explanation for the world-neutral status of truths of middle knowledge; transworld depravity reinterpreted; synthetic modalities as 'existential modalities' that arise from the 'prior' existence of sub-maximal states of affairs (with diagram); feasibility as a kind of synthetic modality; comparison to nomological modality again; the unsaturated sense of the metaphysical 'priority' (P) implied by Molinism.

(D) The Truths of Middle Knowledge as 'Hard Facts'

Response to Craig's first objection: proof that the truth of counterfactuals of freedom are univocally 'prior' to the actions they are about, if P is a transitive relation; why the Molinist cannot assume that P is intransitive; response to Craig's second objection: an argument to show that the Molinist must presuppose that P is transitive.

(E) Why Molinism Must Attribute Truths of Middle Knowledge to Our Agency

The Existentialist's moral responsibility test; the initial objection to the Molinist counterfactual 'power to do otherwise' renewed; William Alston's objection interpreted; the Molinist's only remaining option: atttribute even hard fact truths of middle knowledge to our agency.

(F) The Existentialist Refutation of Molinism

The common objection to Molinist metaphysics; the Existentialist's two conditions for moral responsibility; the Existentialist's moral responsibility objection to Molinism: the 'Antinomy of Authenticity'; David Lewis's analysis of Molinism criticized; the best Molinist response refuted; formal summary of the Existentialist's argument against Molinism; historical explanation of how the Molinist error came about.

III. Feasibility Without Molinism: Our Pretheoretical Modal Intuitions About Volition

(A) The Ground Underlying Molinist Intuitions

The ubiquity of intuitions for counterfactuals of freedom; the characterological ground of our modal intuitions of feasibility; the morally significant level of 'character'; volitional dispositions.

(B) The Challenge to Existentialism

Simplistic Existentialist libertarianism; the intuition of a synthetic modality of volition; Tom Morris's concept of moral capability; Harry Frankfurt's concept of volitional necessity; the danger of picturing agency as pure negative liberty.

(C) Molinism as an Atomistic Conception of Feasibility

Theoretical conceptions of pretheoretical modal intuitions as interpretations of the ontological grounds delimiting the relevant sphere; the simplest theoretical ground of feasibility model (S). Molinism reinterpreted as the theory of the supervenience of feasible worlds on the primitive feasibility of individual acts (theory P); the metaphysical primitive problem; 'atomistic' supervenience and the impossibility of any reason for patterns of feasibility; David Lewis's criticism of Plantinga's free will defense; Molinism (theory P) as the default 'minimal theory' of feasibility; the possibility of alternative conceptions.

IV. Existentialism vs Other Conceptions of Feasibility

(A) Original Sin

The first (O) conception of feasibility; inconsistency with Molina's distinction between sufficient and enabling grace; critique of an Adam-centered interpretation; impossibility of theodicy on this theory; critique of transworld depravity as a feasibility model for original sin.

(B) Strong Providence

The (SP) conception of feasibility; Hugh McCann's defense of (SP) and its Thomistic roots; critique of McCann's argument that we are responsible for our decisions brought about by God's creative fiat; Dominic Báez's version; Kathryn Tanner's version of the Thomist theory; the impossibility of theodicy on this theory.

(C) The Kantian Approach

Kant's notion of the ultimate 'disposition' in the Religion; link to his theory of spontaneity; single ultimate character as a restrictor of feasibility; comparison to Frankfurt's conception of care and volitional necessity; Philip Quinn's analysis of Kant's theory; theory (K) as a 'Noumenalist' version of Molinism; the possibility of revolution in ultimate moral character vs theory (K).

(D) Kant and the Eschatological Requirement of Openness for Ultimate Moral Character

The joint pedagogical significance of radical evil and the possibility of moral revolution; relation to the moral significance of eschatology; the theme of good works in moral religion; an argument for the 'eschatological requirement' of the possibility of revolution in ultimate moral character.

V. The Habit-Based Approach to Feasibility

(A) Intuitive Features of the Characterological Approach

Advantages of the habit-based approach; the problem of an atomistic habit-based conception;

(B) Robert Adams's Criticisms and the Necessity of Probabilistic Qualification

Adams's argument for irreducibly probabilistic (IP) qualification of the relation of habits to particular actions; the critique of theory (R), a non-probabilistic attempt to accomodate the appearance of probabilistic qualification in a habit-based approach; a moral argument for the IP axiom.

(C) The Incompleteness Problem for Atomistic Habit-Based Truths of Feasibility

The problem of the acquisition of volitional dispositions; Aquinas's notion of habit; argument for a level-interactive (LI) conception of the habit-action relation and a broader ontological basis for truths of feasibility; the problem of altering one's habits of will; Aristotle's habit-based model (A); critique of Aristotle's conception by the 'eschatological requirement'.

VI. The New Existentialist Characterological Theory of Feasibility

The IP and LI principles as distinctive of the Existentialist habit-based approach.

(A) The Concept of Volitional Difficulty

The interpretation of the character-grounded irreducibly probabilistic truths of feasibility for particular actions as difference in openness of logical alternatives; the phenomenon of volitional difficulty contrasted with other sorts of obstruction; the compatibility of Existentialism with the existence of volitional difficulties.

(B) The Universalistic Conception of Feasibility

The anti-atomistic implications of level-interactivity; laws of feasibility equally true for all possible persons as the ground of truths of feasibility for the acquisition of habits, the change of habits, and the relations of habits to volitional difficulty; the 'spiritual sphere' of worlds; Kant's objection to synthetic truths not based on particular natures but holding for all possible persons; response to Kant by way of the phenomenological model of morally significant personhood; summary of the moral personhood theory of feasibility (M); theory (M) as inherently compatible with Existentialism.

VII: Conclusion: Existentialism and Spiritual Laws

Summary of the Existentialist view of personal volition; the inalterability of spiritual laws of feasibility; the example of 'Faustus's Law;' Existentialism as compatible with a kind of moral teleology; conclusion on the warrant for accepting non-simplistic Existentialism as the best possible framework for conceiving feasibility.

In analytic philosophy of religion today, it is well-known that the particular conception of libertarian free will offered by Luis de Molina and Francis Suárez has been revived and plays a controversial role in the analysis of several problems. Although its classical purpose was to provide a way to reconcile libertarian freedom with providence and divine omniscience, the Molinist idea that there are contingent 'counterfactuals of freedom'⁽¹⁾--facts about what free creatures would do in any completely specified circumstance they could be in--is also critical to the best currently available formulation of the free will defense against the logical problem of evil. Although he does not emphasize its role in his versions of the free will defense, as presented in his God, Freedom, and Evil⁽²⁾, The Nature of Necessity⁽³⁾, and other articles, Alvin Plantinga's argument relies on the Molinist intuition that some logically possible worlds are not feasible or "weakly actualizable,"⁽⁴⁾ because they include at least one choice of a particular alternative by a free agent, which in fact that agent would not make if she were in the circumstances.

Plantinga's now-famous example is a certain Curley, who has been offered one bribe but would have accepted a lesser one.⁽⁵⁾ As Plantinga explains this case, there is a maximal segment of a (logically) possible world in which Curley is offered the lesser bribe: i.e. a segment complete in every possible respect excepting Curley's decision either to take or not take this bribe, which is left open, and excepting the truth-value of whatever propositions are logically dependent on that. This world-segment (S) exists, as Plantinga says, in two different worlds, W and W', such that W = S + Curley's accepting the bribe, and W' = S + Curley's refusing the bribe.⁽⁶⁾ Now comes the Molinist part. Professor Plantinga then suggests that we accept, hypothetically, that if this maximal world segment S obtained, Curley would have taken the bribe. Plantinga admits, however, that the viability of this premise rests on the assumption that there is some truth about what Curley would have done in the counterfactual situation, at least when it is sufficiently specified: "a state of affairs that includes Curley's having been offered $20,000, all relevant conditions--Curley's financial situation, his general acquisitive tendencies, his venality--being the same as in fact, in the actual world."⁽⁷⁾ Yet Plantinga makes a persuasive plea that this assumption accords with some of the deepest modal intuitions involved in moral experience:

There is something Curley would have done, had that state of affairs [as just specified] obtained. But I do not know how to produce a conclusive argument for this supposition, in case you are inclined to dispute it. I do think it is the natural view, the one we take in reflecting on our own moral failures and triumphs...⁽⁸⁾

On the contrary, Existentialists (myself included) are more than inclined to believe that Plantinga's Molinist premise must misrepresent our modal intuitions in some fundamental way. In fact, Existentialism is the conception of negative liberty which takes the rejection of this Molinist premise as its first principle.

Virtually all major Existentialist authors⁽⁹⁾ hold not only that moral freedom requires some range of liberty to choose among different logically possible alternatives, but also that a person can only be morally responsible for choices that are not completely determined by factors other than broadly logical necessity. Limitations imposed by physical laws, physiology, and perhaps even innate predilections must not be deterministic, but must allow at least some alternative "possibilities" (in the physical sense, physiological sense, etc.) if the decision is a responsible one. Existentialist authors have thus generally followed Kant's Anselmian interpretation of the ought-implies-can principle. For example, because I can be held morally responsible for my career choice, even though I have been socialized and nurtured to enter into one particular field, it follows that I am free to do otherwise, even if choosing a different career may be very difficult for me. In the same way, an Existentialist denies that there can be any absolutely determined particular facts about what a given individual would choose in a given situation, prior to that individual's being actualized and making the choice.

Thus it turns out that we can gain a more perspicuous understanding of Existentialism itself by thinking of it as the antithesis of Molinism. The most an Existentialist could accept is that, if there are counterfactuals of freedom for particular choices such as Curley's, then these counterfactuals themselves must be determined by the individual's other free choices in time. This is the rigorous meaning behind the enigmatic existentialist motto that a free being's existence precedes their essence. By "essence," the Existentialist does not mean logical essence in Plantinga's sense, but rather the ultimate character of the person over a complete life. The Molinist's idea of the total set of a person's true counterfactuals of freedom not only expresses the person's ultimate character in this sense, but also suggests that this ultimate character is metaphysically prior to the person's being actualized and living their life. Existence precedes essence reverses the Molinist priority, since it means that ultimate character is posterior to existing and leading a life. The Molinist and Existentialist are thus diametric opponents within the general ambit of metaphysical libertarianism.

In support of this interpretation, consider Martin Heidegger's claim that "Dasein's 'Essence' is grounded in its existence."⁽¹⁰⁾ By Dasein, Heidegger means the kind of being constitutive of personhood. Heidegger holds that what makes an entity a person capable of personal identity is not any sort of soul-"substance" or "Being-what-it-is" (Aristotelian essentia), but rather a certain kind of agency with a volitional structure that makes the entity able to relate itself volitionally to itself, to others, and to its world: "These entities, in their Being, comport themselves towards their Being."⁽¹¹⁾ This is shown partly in the fact that only for Dasein can questions of Being be an issue--or, to translate, only persons are so constituted (ontically) that they have the sort of ontological relations to themselves, to their world, and to others that involve modal intuitions with ontological import. As he explains in more detail:

And because we cannot define Dasein's essence by citing a "what" of the kind that pertains to a subject matter, and because its essence lies rather in the fact that in each case it has its Being to be, and has it as its own, we have chosen to designate this entity as "Dasein."⁽¹²⁾

The term Dasein is explicitly chosen to contrast with Sosein. The type of agency necessary for personhood is not a "Sosein" or multiply instantiable essence, precisely because it requires first-personal individuality, self-possession, in order to "comport" itself towards itself--or, in the language of analytic philosophy, to establish a reflexive volitional relation.⁽¹³⁾ Dasein is individual being as distinct from a kind-essence that makes an entity 'what it is.'

Dasein's individual existence, however, should not be misunderstood simply as the medieval concept of existentia that contrasts with essentia. Heidegger certainly does not mean that Dasein is a bare entity without color or quality or character. Rather, the existence which is proper to or distinctive of Dasein is free being, with others persons, in a world of natural environments and artifacts. Heidegger explains the 'existence' which characterizes Dasein as the ontological condition in which free will expresses itself:

Dasein always understands itself in terms of its existence--in terms of a possibility of itself: to be itself or not itself. Dasein has either chosen these possibilities itself, or got itself into them, or grown up in them already. Only particular Dasein decides its existence, whether it does so by taking hold or neglecting. The question of existence never gets straightened out except through existing itself.⁽¹⁴⁾

For Heidegger, Dasein is faced with possibilities that are not yet closed or determined (or "actualized" to speak Plantingian). In existence, an individual's social identity and ultimate personal character are still open, and do not become finalized except through his or her own choice.⁽¹⁵⁾ Moreover, even the range of future possibilities for one's being are determined existentially, or in time, both by what the individual contingently inherits from the past and by their own response to these conditions.

In his later Letter on Humanism, Heidegger makes clear that his focus on the existence of Dasein is not intended simply as a metaphysical reversal of Plato's doctrine that "essentia precedes existentia."⁽¹⁶⁾ In a look back at the crucial passage in Being and Time, Heidegger rephrases his earlier statement by saying that "What man is..the 'essence' of man--lies in his ek-sistence".⁽¹⁷⁾ "Ek-sistence" here "does not coincide with existentia in either form or content," but means "standing out into the truth of being;" it designates an irreducible capacity for awareness of ontology, an access to the significance of ontological questions, that is constitutive of personhood. Ek-sistence in this sense means having access to ontological meanings (i.e. meanings with modal content), and for this reason, ek-sistence itself "can also never be thought of as a specific kind of living creature among others".⁽¹⁸⁾ For being aware of the ontological, standing in the light of being, cannot itself be construed as just a kind of being. For a kind is something an entity exemplifies simply by being it, but Dasein questions what its own being is, and is able to grasp the significance of its having being--therefore Dasein never just has its being, or just is it. Ek-sistence is not a kind-essence, but nor is it bare existentia: rather it is the ontological state of being ontological, that is, open to modal truth, which includes non-actual possibilities that are not merely represented, but given by Being itself.⁽¹⁹⁾ Its awareness of its own possibilities, which arises from Dasein's freedom, is the origin of its peculiar 'standing-out' from kind-existence. By contrast, other creatures "are as they are without standing outside their being as such.."; they are "lodged in their respective environments but are never placed freely in the lighting of being.."⁽²⁰⁾ This clarifies the true relation of freedom and existence for Heidegger: man's "ek-sistence" is constituted by an openness in time to personal possibilities that are conditioned but not determined or closed by history--personal possibilities which open Dasein's eyes to the ontological generally.

We have, then, two Heideggerian theses that help define the Existentialist position: (1) Persons with free agency do not act outside actual existence in time, and (2) even in actual existence, ek-sisting or being a person means that one's ultimate character is not closed.⁽²¹⁾ As we will see, these two theses are the basis for the Existentialist's two strongest objections to Molinism and other conceptions of feasibility.

My primary aim in §II is to show how the first Heideggerian thesis addresses the deepest problem with Molinism. After discussing several other recent critiques of Molinism and responses to them, I will argue that the Molinist conception of 'feasibility' must be understood as a synthetic modality analogous to nomological modality. When this is understood, it becomes clearer that Molinists cannot block the inference that the truths of middle knowledge are 'hard facts' which are metaphysically prior to the actions and choices that they are about. Nevertheless, the Molinist must still attribute these truths to our agency in order to create a disanalogy to causal necessitation. These preliminary conclusions (or lemmas, if you like) provide the basis for the main Existentialist refutation of Molinism. This refutation focuses on the conception of agency implicit in Molinism to argue that authentic moral responsibility for one's actions in a Molinist framework leads to an antinomy.

Taken together, these arguments should be fatal to the classical Molinist position. However, as we see in §III, this analysis will hardly be complete or decisive for the Existentialist. There remains a certain deficiency in the 'simplest' Existentialist conception of human freedom, a deficiency which is related to its apparent failure to accommodate precisely the modal intuitions which Plantinga claims should support his Molinist supposition. I will argue that there is something to this claim, but that Plantinga and other Molinists have not fully analyzed the underlying grounds from which the intuitive warrant for the Molinist suppositions actually derives.

This concludes the negative side of the paper's argument. In Part III, however, I argue that there are ubiquitous modal intuitions about motivation and behavior which Molinism answers to. Plantinga claims that these intuitions should support his Molinist theory, but, as I will argue, Molinists have not fully analyzed the underlying ground from which these modal intuitions arise. In §III (A), I will argue that the modal intuitions which account for the appeal of Molinism arise from our sense that there is modality for volition which limits freedom in its purely libertarian sense. This allows us to separate the notion of "feasibility" and the intuitions which support it from the Molinist theory of feasibility. Feasibility really stands for the pretheoretical intuitions of this synthetic modality, which is closely bound up with our notion of volitional character and 'habits of choice'. Since our common beliefs about feasibility are actually based on intuitions about character or dispositions of the will, it becomes clear that traditional Molinism (including its modern analytic versions) is only one specific conception of this synthetic modality called 'feasibility,' which is associated with volition and character.

When we understand that the grounds of feasibility as a synthetic modality are essentially characterological, the potential for an alternative (non-Molinist) account of the same modal intuitions becomes clearer. Yet, for an Existentialist theory which rejects Molinism, offering such an alternative is not just a potential: it is also a requirement. Any positive Existentialist account of human agency will have to explain the same modal intuitions in an alternative (non-Molinist) fashion. I suggest that the failure of the 'simplest' Existentialist conception of human freedom to meet this need is the root of the most common criticism brought against Existentialism: namely, the danger of reducing free choice to complete arbitrariness among logically possible alternatives. I will discuss this critique using some examples from Harry Frankfurt which will help us to see the need for a 'thicker' conception of volitional character than that shared by utilitarianism and 'simplistic Existentialism.' It is only by developing a characterological account of differentials in the 'feasibility' of our different options for action that Existentialism will ultimately be able to respond to this critique.

In §III-(C), I argue that there should be alternative possible theoretical conceptions of feasibility as a synthetic modality, which would attempt to capture the same modal intuitions as traditional Molinism, but diverge from it in implications. Different theoretical conceptions acknowledge that they are describing the same synthetic modality, but they have different accounts of the 'world-neutral' which ground the modality. Molinism, for example, has an atomistic conception of the truths which ground feasibility: the Molinist sees the ultimate limits on the sphere of feasible worlds as resulting from the interaction of indenumerable, primitive facts about individuals. Seen in this light, however, Molinism is a highly unattractive way of explaining our pretheoretical modal intuitions. There is every reason to belief an alternative conception could do better.

In §IV, I will consider three alternative conceptions which the Existentialist can reject on good grounds. In critiquing the conception of feasibility initially offered by Kant, I will return to the second Heideggerian thesis mentioned above. It forms the basis for the Existentialist's conception of the relation between ultimate moral character and eschatological possibility, which justifies the rejection not only of Molinism but also of other conceptions of feasibility.

In §V, I will explore the habit-based approach to feasibility as a synthetic modality of free choice, beginning a characterological model Robert Adams considers and critiques, as well as another model which must be rejected in light of our findings in previous sections. Finally, in §VI, I will show how to form a new character-based conception of 'feasibility' as a synthetic modality of free choice, which not only avoids the problems of the other models, but is even consistent with Existentialism. This moral conception of volitional modality (or feasibility), which is based on a phenomenological model of personhood, shows how it is possible for Existentialism to be committed to a more substantive account of free agency than it critics realize.

In short, by the end of the paper I will have shown, (A) that the modal intuitions at stake in Molinism are deeper than the particular conception offered by Molinists indicates; (B) that traditional Molinism itself, like several closely related models, captures the content of these intuitions in ways that are both ontologically inadequate and inconsistent with moral responsibility, and (C) that by contrast, Existentialism itself can adequately capture the same intuitions in a new model that clarifies the true relation of character to transcendental freedom, with radical implications. Where Molinism and other alternative conceptions must fail, Existentialism can succeed.

II. The Existentialist Critique of Molinism

Modern analytic forms of Molinism naturally follow Luis de Molina's conception of God's "middle knowledge," which Professor Thomas Flint has conveniently summarized as follows:

...for any person who does or might have existed, God, (being omniscient) would know, prior to any creative act on his part, what that person would freely do in any situation in which that person might be created and left free. Given his commitment to libertarianism, though, Molina also argued that the truths God would thus know--truths of the form "If person P were to be placed in circumstances C, P would freely do A"--would not be truths over which God had any control, despite the fact that such conditionals would be only contingently true.⁽²²⁾

For the Molinist, true counterfactuals of freedom are regarded as particular facts about individuals, which God does not cause but uses as the basis for his providential activity. Both the contingency of these truths and their independence from divine activity is crucial, because the Molinist wants to maintain that these truths are completely consistent with persons having the power to choose among alternatives in a robust sense. But in order to complete God's foreknowledge and guide the creation of the best feasible world, these truths about what a person would do in any situation must metaphysically precede their actual existence.

(A): The Preliminary Objection

The Existentialist is convinced that this picture of our freedom is fundamentally wrong, but it is far from easy to say precisely where the problem lies. My plan is to introduce an initial, largely intuitive objection, which will force us to clarify several of the key features of Molinist metaphysics. Only working through this process will reveal the element of Molinism to which the Existentialist's objection applies.

The initial objection is that intuitively, there is a certain loose analogy between the Molinist's notion of unfeasibility and nomological or causal necessity.⁽²³⁾ A causally necessary truth is logically contingent, but its alternatives are not 'physically feasible.' The counterfactuals which give this truth its modal force are all logically contingent, but presumably true of our physical universe. In a similar fashion, true counterfactuals of freedom are logically contingent, but a complete set of them determines which logically possible (l.p.) worlds are personally feasible (or feasible given the free persons included in these worlds). Just as the sphere of l.p. worlds that are physically possible (relative to our universe) expresses certain natural laws of our universe,⁽²⁴⁾ in the same way, the sphere of l.p. worlds that are 'feasible' in the Molinist's sense might be thought to express certain 'laws' pertaining to the choices of each possible person. We could even say that the logically contingent truth about all a person's counterfactuals of freedom establishes another sort of modality (analogous to causal)--let us call it 'personal necessity' or feasibility--in which we could say that it is personally necessary that Curley would accept the $20,000 bribe if offered it. It is unfeasible for Curley to refuse.

The idea behind this comparison is that feasibility in the Molinist sense, like causal possibility, refers to a kind of synthetic modality distinct from 'broadly logical modality.' Quine, of course, has made a famous attack on the Kantian distinction between 'analytic' and 'synthetic' which requires me to explain what I mean by a synthetic modality very clearly. While Kant's notion of 'analytic' may be inexact,⁽²⁵⁾ suppose we identify 'analyticity' with Plantinga's notion of broadly logical necessity. Then a new definition of metaphysical syntheticity, analogous to Kant's old one, may be given: a proposition is synthetic if it is neither logically necessary (necessary_l) nor necessarily_l false (impossible). Given this reconstruction, Kant's characterization of 'metaphysical' truths as synthetic and necessary propositions can only mean that there are kind(s) of 'necessity' (or modal relevance) which are narrower in scope than broadly logical modality.⁽²⁶⁾ Just as the meaning of causal or physical possibility can be understood in terms of a 'sphere' of l.p. worlds (which excludes some l.p. worlds as physically impossible),⁽²⁷⁾ feasibility can be interpreted as referring to a proper subset of l.p. worlds.

It is perhaps easier to see the relevance of this modal interpretation of counterfactuals of freedom if we reflect on possible patterns of true counterfactuals of freedom that might show up across many or even all possible persons. An example of this would be what Plantinga calls "transworld depravity." If we admit the Molinist assertion that there is always a truth of the matter about what a person would do in each possible fully specified circumstance they could ever be in, it is a short step for Plantinga to assert that it is logically possible that for any world in which Curley is "significantly free" (or possesses negative liberty in the logical sense of having alternative possibilities), there are some choices with respect to which he would always go wrong, if placed in situations requiring him to make them.⁽²⁸⁾ This means that Curley is in what Plantinga calls transworld depravity: the worlds in which he never sins (or "goes wrong" in a morally significant free choice) are all unfeasible in the Molinist sense.⁽²⁹⁾ If we accept that it is possible that "there be persons who suffer from transworld depravity" in this sense--a possibility which Plantinga considers "obvious"--then in addition "it is possible that everybody suffers from it,"⁽³⁰⁾ and even that the essences of all possible free agents could be in transworld depravity, in the sense that in the only l.p. worlds containing them that are in fact creatable by God, they would sin.⁽³¹⁾ In sum, it is logically possible that no feasible world is without sin.⁽³²⁾ If this possibility were true, it would surely look and function very much like an empirical law for persons. Although it would be logically contingent, if such a 'general law of transworld depravity' were true, it would tell us something specific about the limits which determine the sphere of feasible worlds: namely that this sphere excludes logically possible worlds in which free persons are sinless. Within this sphere, we therefore have a synthetic necessity: it is unfeasible (or false in all feasible worlds) for any possible person to live a sinless life. Since this claim extends beyond the range of all actual persons, it seems to have the sort of modal force which is usually thought distinguishes nomological necessities from mere empirical regularities.

If the analogy to causal laws still seems unclear, consider the cosmological function transworld depravity is meant to play. G.W. Leibniz once thought that God could find the best possible world by looking at the overall order (including natural laws) and variety in each world, and identifying the one which had the most 'efficient' combination of these factors. For Plantinga, analogously, among feasible l.p. worlds, there would be an ordering ranging from the world(s) with the least depraved compossible⁽³³⁾ set of persons, to those with more and more depraved compossible sets of persons.⁽³⁴⁾ Then Molinism can provide a "defense" of the existence of the particular world we have, on the grounds that it is possible that this is the best of all feasible worlds--best in mixture of moral good and moral evil, at least.⁽³⁵⁾ If this possibility were true, then every other set of persons God could have actualized would be more depraved in their actions, because their counterfactuals of freedom ensure it, than the set of persons existing in this the actual world.⁽³⁶⁾ It is then not logically impossible but rather unfeasible for God to have done better.

The Molinist would surely argue against the analogy between contingently true causal laws and contingently true patterns of counterfactuals of freedom by insisting that these counterfactuals do not constrain the alternatives to which our free will is open, as causal laws do. God's knowledge of our counterfactuals of freedom does not constrain us because, as Tom Flint says, Molinism allows that we have "counterfactual power" over God's middle knowledge:

"..if I'm free, then I do have the power to do things such that, were I to do them, both God's middle knowledge and his simple foreknowledge would be different."⁽³⁷⁾

But this power to bring about logically possible alternatives is somewhat deceiving, because according to Molinism, it is bound (in a nearly 'nomic' sense) to remain unused. There is no real possibility--possibility in a synthetic sense distinct from logical possibility--of my now exercising my so-called "power" to do things such that, were I to do them, what God foreknows would be different. The reason is simple: God's middle knowledge (as part of his foreknowledge) is based on a set of facts which is metaphysically prior to my being actualized as a real person in an actual, temporal world. Whatever I do here is posterior to the facts that determined God's middle knowledge. Thus, by the same token, what I do as an actual person is also metaphysically posterior to God's middle knowledge. In this respect, then, Professor Flint's so-called "power" is like my power to do things such that, were I to do them, the laws of nature would be other than they are: in other words, it is rather ephemeral at best. Flint admits as much when he notes that for the Molinist, God's middle knowledge is not causally dependent on any of our actions: rather, "our actions are causally dependent on God's decision to create us in the circumstances in which those actions are performed."⁽³⁸⁾ The sense in which I lack the power to do otherwise, then, is something analogous to nomological impossibility. This comparison of Molinist 'feasibility' to causal modality depends, of course, on idea that there can be real (not merely conceptual) kinds of modality more restricted in scope that logical possibility.

(B) The Haskar/Adams Anti-Molinist Strategy and the Molinist Response

This preliminary Existentialist objection to Molinism, however, is quite similar to other critiques which Molinists have insisted they can meet. In his God, Time, and Knowledge, William Haskar published an argument against Molinism⁽³⁹⁾ which can be summarized as follows: (1) the truth of an individual person's counterfactuals of freedom are prior, in some important sense, to their powers of agency; hence (2) because the truth of a counterfactual of freedom necessitates the action it predicts if its antecedent conditions are actualized, an individual in a choice circumstance is not free to act contrary to his or her counterfactual of freedom for that circumstance. In his own "Anti-Molinist Argument," Robert Adams praises Haskar's line of reasoning and portrays his own argument as a "recasting of Haskar's."⁽⁴⁰⁾ Like Adams, I am convinced that Haskar's intuition is on the right track and locates the essential reason why Existentialists, like many others, cannot accept Molinism.

But there have been qualms about the strength of Haskar's proof. In a recent article, William Lane Craig reiterates his opinion that Haskar's argument rests on the "clearly false" premise "that on the Molinist view, counterfactuals of freedom are more fundamental features of the world than categorical facts."⁽⁴¹⁾ Craig can make this objection because the kind of "priority" enjoyed by counterfactuals freedom, which allegedly makes individual agents unfree with respect to them, is quite unclear. As Adams suggests, the first stage of the argument is to show that "created, supposedly free agents do not bring about the truth of counterfactuals of freedom about them, if Molinism is true."⁽⁴²⁾ As an alternative to Haskar's own argument for this, Adams proposes to use "the idea of explanatory priority" to make clearer the sense in which the truth of counterfactuals of freedom is prior to the agent's actions.⁽⁴³⁾ The second stage is to argue "that it follows that such agents do not have the power to act otherwise than they in fact do."⁽⁴⁴⁾ But even supposing we have completed the first stage, Adams identifies several premises in Haskar's subsequent inference that agents lack the power to do otherwise than their counterfactuals of freedom dictate. One premise is that "for Molinism, it is a necessary truth that every counterfactual of freedom whose consequent is true is true."⁽⁴⁵⁾ As Adams says, it would be hard for the Molinist to do without this presupposition, but a more controversial premise is:

...Haskar's assumption that if Molinism implies, as he says, that we do not bring about the truth of counterfactuals of freedom about us, it also implies that we do not have the power to bring about their truth. The assumption is plausible, but I am not sure it has been proved.⁽⁴⁶⁾

Let me dub this the 'Power Inference Principle' (since Adams does not name it). In my own view, the problem with this assumption is that the Molinist can always insist that we have this power in the weak sense that alternative actions are logically possible for us in a given choice-circumstance. This response to the Power Inference Principle (PIP) is damning because Haskar's overall argument is an attempt to show that the metaphysical priority of the truth of counterfactuals of freedom conflicts with our logical possibility of doing otherwise. Thus the inference from our not bringing about the truth of our counterfactuals of freedom to our lacking the power to do so tacitly assumes the crucial point.

What this shows, I think, is not that Haskar's proof strategy simply needs emending, but rather than the goal of his anti-Molinist critique needs to be reconceived. What we need to show is not that Molinism implies that we lack the logical possibility of doing otherwise, but rather that it implies a different sort of powerlessness. In my development of the Existentialist's 'preliminary objection,' I already suggested that it must be in something analogous to a nomological or causal sense that true counterfactuals of our freedom prevent us from doing otherwise. Developing this intuition will both help us clarify the type of 'priority' which the truth of counterfactuals of freedom have, and open up what I think is a better way of pressing Haskar's anti-Molinist intuition.

Adams, however, believes he can construct an alternative anti-Molinist argument without Haskar's conclusion that "it follows from Molinism...that we do not bring about the truth of an counterfactuals of freedom about us."⁽⁴⁷⁾ Adams's argument begins as follows:

(1) According to Molinism, the truth of all true counterfactuals of freedom about us is explanatorily prior to God's decision to create us.

(2) God's decision to create us is explanatorily prior to our existence.

(3) Our existence is explanatorily prior to all of our choices and actions.

(4) The relation of explanatory priority is transitive.

(5) Therefore, it follows from Molinism (by 1-4) that the truth of all counterfactuals of freedom about us is explanatorily prior to our choices and actions.⁽⁴⁸⁾

Adams then skips his steps (6)-(9), which constitute the argument for Haskar's conclusion, and continues with six more steps of his own, beginning with

(10) It also follows from Molinism that if I freely do action A in circumstances C, then there is a true counterfactual of freedom F* which says that if I were in C, then I would (freely) do A.⁽⁴⁹⁾

Since from (5) is follows that (11) "the truth of F* is explanatorily prior to my choosing and acting as I do in C," and since (13), "the truth of F* ..is strictly inconsistent with my refraining from doing A in C," I am compelled by F* to do A in C in a way that is incompatible with my libertarian freedom: (15) "if Molinism is true, then I do not freely do A in C," which is a reductio of Molinism.⁽⁵⁰⁾ To reach this conclusion, Adams only needs the conception of freedom in the following premise:

(12) If I freely do A in C, no truth that is strictly inconsistent with my refraining from A in C is explanatorily prior to my choosing and acting as I do in C.

As Adams explains, this premise is supposed to capture freedom in the "incompatibilist sense," and its implication is that "the totality of my action in situation C" must be the first thing in the order of explanation which logically excludes my acting otherwise.⁽⁵¹⁾

The basic problem with Adams's approach is that the notion of "explanatory priority" is too vague a concept to capture precisely the metaphysical priority that the truth of our counterfactuals of freedom seems to have on the Molinist account. As a result, William Lane Craig has been able to challenge Adams's argument as using "equivocal" senses of "explanatory priority."⁽⁵²⁾ Craig argues that explanatory priority in Adams's (2) and (3) is "metaphysical" necessity, but the truth of our counterfactuals of freedom is required in the same sense by God's decision to create us:

The root of the difficulty seems to be a conflation of reasons and causes on Adam's part. The priority in (2) and (3) is a sort of causal or ontic priority, but the priority in (1) is not causal or ontic, since the truth of all counterfactuals of creaturely freedom is neither a necessary nor a sufficient condition of God's decision to create us.⁽⁵³⁾

Craig also charges, somewhat obscurely, that the sense of explanatory priority in Adams's inference (5) is different still.⁽⁵⁴⁾ Finally, Craig also charges that Adams's premise (12) is false, because the explanatory priority of my counterfactual of freedom does not entail my powerlessness to alter it:

Though F* is (ex concessionis) in fact explanatorily prior to my freely doing A in C, it is within my power to refrain from doing A in C; only if I were to do so, F* would not then be explanatorily prior to my action nor a part of God's middle knowledge.⁽⁵⁵⁾

Although Craig does not say in what sense it remains "within my power" in C to alter the counterfactual of my freedom which truly describes what I would do in C, presumably he means that it is still logically possible for me to do so. In this second objection, Craig is trying to show that the inference Adams's wants to draw from his premise (12) implicitly requires Haskar's conclusion that the person's choice mentioned in the consequent of the counterfactual of freedom cannot be what actually brings about the truth of the counterfactual. Only if it is not within my power to bring about the truth of my counterfactuals of freedom, could we be assured that their (alleged) explanatory priority makes it logically impossible for me to do otherwise. Thus Craig's second objection forces Adams's case against Molinism to depend once again on Haskar's conclusion that we do not cause the truth of our counterfactuals, and his 'Power Inference Principle.' Without these, Adams will not be able to show that Molinism contradicts our logical possibility of doing otherwise. Craig is satisfied to hang on this Molinist libertarian power--ephemeral though it actually is.

Ultimately, I think Adams's new anti-Molinist argument leaves space for these countermoves because it does not give a clear enough account of the 'special priority' which counterfactuals of freedom must have to be of any use to divine providence: taking this status as an explanatory priority leads Adams to try to show that the truth of my counterfactual of freedom makes it logically impossible to do otherwise. The only way to put a permanently stop to the Molinist's dance back and forth between allowing priority to counterfactuals of freedom (when talking about providence) and then refusing them priority (when talking about freedom) is to develop Adams's and Haskar's insight a different way.

The Existentialist's alternative strategy also has two stages: to analyze Molinist feasibility as a type of synthetic modality, and then to make clear that the synthetic necessity of counterfactuals of freedom is already enough to cancel moral responsibility for one's actions, no matter what ephemeral powers of doing otherwise might remain. Specifically, the type of "explanatory priority" which the truth of counterfactuals of freedom have for God, as in Adams's premise (1), must be interpreted as a kind of necessity which is independent of God's will. Moreover, although these truths are not causally prior to God's decision to create us, in themselves they are already analogous to causal necessities which restrict God's options. God's decision to create us may not be necessitated by these truths, but in analogous fashion, the laws of nature may also be neither necessary nor sufficient conditions for my dropping this penny: nevertheless, these laws are still metaphysically prior to my actions, since they necessitate certain consequences of whatever I do. Analogously, it is deceptive for Craig to suggest that the truth of counterfactuals of freedom serve only as reasons for God's decisions: Molinism very clearly implies that these truths also determine what the consequences of creating free creatures in different situations will be. In fact, it is precisely these real determinations which serve the Divine intellect as reasons. The Existentialist's objective will be to show that the synthetic necessity established by true counterfactuals of freedom is incompatible with moral responsibility, even if it is compatible with the logical possibility of doing otherwise. The first step for the Existentialist is therefore to explain what it means for the truths of middle knowledge to constitute a kind of synthetic modality.

(C): A Theory of Synthetic Modalities and the Truths of Middle Knowledge

To say that the Molinist power to do otherwise is "ephemeral" is not to accuse the Molinist of contradicting the logical conditions of contingency required for libertarian free will. The Existentialist's objections to Molinism do not rest on the misunderstanding that the truth of an individual's counterfactuals of freedom are part of their logical essence, which would leave them only nominal power to bring about different alternatives. But the fact that the Molinist position is so often misunderstood this way is not an accident, and I think it reveals something interesting about Molinism. For example, Plantinga's notion of transworld depravity as a possible pattern of true counterfactuals of freedom lends itself to this common misreading of Molinism, because Plantinga makes it sound as if it is a person's essence by itself which accounts for their transworld depravity. Because he describes individual essences themselves as depraved, it would be natural to interpret Plantinga as meaning that the truth of certain counterfactuals of freedom (those ensuring transworld depravity) is essential to the person--i.e. true in every possible world in which the person exists.⁽⁵⁶⁾

But in fact, Molinism does not hold this: rather, to remain officially consistent with libertarian freedom, it allows that in world W', in which Curley does not take the $20,000 bribe, what Curley would have done (his counterfactuals of freedom) is also different.⁽⁵⁷⁾ A person's counterfactuals of freedom themselves must all be (logically) contingent properties of that person, according to Molinism.⁽⁵⁸⁾ This point, which I will call Flint's axiom, is a corrective to the misunderstanding that an individual's counterfactuals of freedom are logically essential to him in the de re sense. The truths of a person's counterfactuals of freedom are thus not a function of their logical identity alone, since the same person exists with the same logical identity in worlds in which he has different counterfactuals of freedom.

Flint's axiom, however, naturally engenders a new misunderstanding of its own. The problem is that contingency normally means simply being true at some logically possible worlds, and false at others. We may think of contingent truth as truth with respect to a set of logically possible worlds, or truth that is world-relative.⁽⁵⁹⁾ Although an individual's counterfactuals of freedom certainly are contingent in this sense, the truth of those counterfactuals of freedom that become part of God's middle knowledge, cannot be said to be contingent in precisely this simple sense. Flint's axiom is deceptive to the extent that it makes it sound as if the truth of an individual's counterfactuals of freedom is a world-relative truth for the Molinist. We can say that (') Curley would refuse the bribe in C is true with respect to W' (and would be true if W' were actualized), and () Curley would accept the bribe in C is true with respect to W (and would be true if W were actualized), but the Molinist claims that there is an additional fact which breaks this nice symmetry: namely, that Curley in fact would take the bribe. () is true simpliciter, and not just with respect to W. That is why actualizing W' is unfeasible: the counterfactual of freedom ' which would be true if that world were actualized, is in fact false. But is this just because W has been actualized instead? Certainly not, for this truth has metaphysical priority in the following precise sense: is true prior to the actualization of any single, maximally complete, logically possible world.

This makes more explicit how the explanatory priority enjoyed by the truths of middle knowledge must function according to Molinism. The Molinist claims that counterfactuals of an individual's freedom can be true in a world-neutral sense,⁽⁶⁰⁾ where 'world-neutral' is the complementary property to 'world-relative' as used above. Counterfactual is true of Curley, not merely at W or relative to W, but simpliciter. And it is this "fact" that determines that only l.p. worlds (including W) for which is true in the world-relative sense can be feasible.⁽⁶¹⁾ By 'world-neutral,' I do not mean that counterfactuals of freedom such as are true at no l.p. world (which would mean that they are logically impossible), nor that they are true at every world (which would make them necessary). Rather, I mean that the truth of , for example, is independent of the actuality of any particular complete l.p. world out of the l.p. worlds at which is true in the world-relative sense. World neutral truths are prior in actuality to a maximal state of affairs, or to the truth of a maximal proposition (or whatever we take a logically possible world to be).⁽⁶²⁾

And this is what makes Molinism so liable to be misunderstood. In S5 modal semantics, we normally think of a truth about an entity as either logically essential to that entity, or logically contingent for it, i.e. true only at certain worlds in which it exists. But the Molinist implies, without fully clarifying it, that there is an intermediate ontological status truths about entities can have. There are some truths which are contingent in the de re sense, because they do not hold for the entity in every world in which it exists, but which are nevertheless true in the 'world-neutral' sense. These counterfactuals of freedom are true simpliciter, but their world-neutrality does not mean that they are logically essential to the individual: rather, they are neutral among a sphere of l.p. worlds, which must exclude some l.p. worlds in which the individual exists⁽⁶³⁾ but has different counterfactuals of freedom, and would act differently in the same circumstances. For example, although the truth of is neutral among the worlds at which Curley exists and takes the bribe in C, but it unactualizable the (non-empty) set of logically possible worlds in which Curley exists and refuses the bribe in C;⁽⁶⁴⁾ so is world-neutrally true for Curley, although it is not logically essential to him.

This interpretation implies that transworld depravity is also more complex notion than it might appear to be. Speaking rigorously, we should really say that a person's essence contributes to such depravity only by telling us all the sets of choice-circumstances a person would face in all the possible worlds in which they exist. These sets will differ in each world, despite overlaps.⁽⁶⁵⁾ But if we index to each l.p. world the set of choices an individual faces in that world, then the set of these world-indexed sets is essential to the person. This is all the person's essence by itself gets us--it does not by itself tell us that the person is "depraved." For in fact, in the world-relative sense, it is not true that he behaves depravedly in every l.p. world. Rather, it is the 'intermediate' contingent but world-neutral truths of his counterfactuals of the freedom that determine that in every logically compossible set of world-choices of the agent, there are some choices (at least one) in which he would go wrong. It is not really his essence, but his unique set of world-neutral truths about what he would do, which is 'depraved.'

The Molinist is thus an actualist about an individual's counterfactuals of freedom (henceforth, c.o.fs). A given individual has exactly one consistent set of these c.o.fs that are all true for him, in the world-neutral sense, in any world. In other words, while ' is true for Curley at W', this possible counterfactual of freedom ' is nevertheless false in quite another world-neutral sense, which is precisely what makes W' unactualizable. Only when one grasps this does the full strangeness of the Molinist's position become apparent. While the Molinist says that an agent has different c.o.fs in worlds W and W', he cannot only mean that one set of c.o.fs would be true if W' were actual, while another set of c.o.fs would be true if W were actual. While it may be true that if W' were actual, a different set of c.o.fs would be true for Curley, there are facts (unusual, world-neutral ones) that privilege one set of Curley's c.o.fs as the true set for Curley. Which set of c.o.fs this is does not depend merely on which complete logically possible world is actualized. Rather, Molinism insists that it is the other way around: which logically possible worlds can be actualized depends on what these world-neutral truths are.

As a result, the Molinist cannot mean that the true set of c.o.fs about Curley is simply the set that happens to be true in (Plantinga's rigid designator for whichever complete l.p. world turns out to be instantiated as the actual world).⁽⁶⁶⁾ Rather, which set of Curley's c.o.fs is true in the privileged, world-neutral sense determines which logically possible worlds even could become . A set of Curley's c.o.fs is true, prior to the complete actualization of any maximal logically possible world.

If we like, instead of a truth about Curley, we can interpret a true counterfactual of his freedom as a subjunctive state of affairs 'S' (something like 'Curley's would-doing A were C actual'), which obtains, or is contingently actual, 'prior' to the complete actualization of any l.p. world. This 'priority' itself is still primitive or undefined, but at the moment, we are only concerned with how it functions in Molinism. In practice, the 'priority' enjoyed by a subjunctive state of affairs such as 'S' means that it does not entail the actuality of a single possible world, but it does entail that l.p. worlds in which Curley exists and does not do A in C cannot be . For with 'S', the state of affairs corresponding to , we already have a piece of --a piece that is actual 'prior' to the rest of . Thus cannot turn out to be a world that does not include 'S.' The real implication of Molinism is that there is a set of c.o.fs for every possible person like Curley, which is determined from all time to be a segment of . Yet the actuality of all the subjunctive states of affairs corresponding to all the true counterfactuals of freedom put together does not pick out a complete logically possible world: thus taken together, the truths of counterfactuals of freedom is still neutral among many logically possible world.

Let us call this set of truths CF. We could think of those l.p. worlds which contain CF (or which contain nothing whose contrary is entailed by CF) as '-relevant' in the following sense: those worlds remain candidates for becoming . This is what feasibility in the Molinist sense means: it is a relevance notion, since it refers to a particular kind of relevance which some logically possible worlds have. We can think of this relevance as a property which 'colors' a sphere of logically possible worlds. This 'sphere of relevance' is established by the world-neutral truths of CF, or the actuality of the subjunctive states of affairs corresponding to them. The metaphysical priority of CF, which I said consisted in the notion that the truth of counterfactuals of creaturely freedom are given prior to the truth of a complete l.p. world, can thus be translated as follows: the -relevance of a certain sphere of l.p. worlds metaphysically precedes the actuality of a single l.p. worlds as . The determination (by God and natural chance, or whatever) of which l.p. world is instantiated as is thus conditioned by the -relevance of our sphere of worlds. The results can be represented in the following diagram:

This model helps us understand why Molinism ought to be interpreted as implying the existence of a kind of synthetic modality. If there is any kind of real synthetic modality, i.e. a modality narrower in scope than logical possibility that refers to distinctions that are not simply a result of our language, our ideals of economy in theoretical systems, and so on, such a synthetic modality could only arise on the sort of ontological ground we are considering here. For some real synthetic form of modality 'K' to exist, there has to be a logically sub-maximal set of states of affairs (in the case of Molinism, those corresponding to CF), whose actuality has some sort of 'metaphysical priority' (either primitive or to be interpreted in some way) relative to the actuality of --the complete logically possible world that is uniquely actual on 'actualist' interpretations of the metaphysics of modality. The prior actuality of this set of states of affairs grounds the relevance of the 'sphere of worlds' which gives meaning to the K-kind of synthetic modality. We can then say that propositions that are true throughout this 'sphere' are K-necessary, those that are true in some worlds in this sphere are K-possible, and those that are false in every world in the 'sphere' are K-impossible.

A synthetic modality with this type of metaphysical structure is not merely 'conceptual:' it would be better described as existential, since the relevance of the sphere of worlds that give the synthetic modality its sense arises from the privileged existence of a certain set of states of affairs. The notion that any such kinds of real synthetic modality exist may seem wild or even scandalous to some readers,⁽⁶⁷⁾ but it must be admitted that the idea makes sense, if we can credit that the truth of some propositions (or the actuality of corresponding states of affairs) may be 'prior' in some atemporal, non-causal sense, to the actuality of an entire possible world. The crucial point is that Molinism itself implies such an account of the truths of middle knowledge when it lets these truths restrict the states of affairs it is possible for God to actualize. All my model does is show why the special priority, or world-neutrality, which Molinism thus gives to the truth of some counterfactuals of freedom inevitably establishes a form of synthetic modality: a real form of synthetic modality just is a sense of modal quantification made relevant by the -relevance of a sphere of l.p. worlds. Whether he likes it or not, the Molinist cannot avoid the implication that 'feasibility' in his sense is a synthetic modality in my sense.

This theory of synthetic modalities is bound to seem unfamiliar for a straightforward historical reason. The type of sub-maximal actuality which ground a kind of -relevance of a sphere of worlds and thus provides the semantic basis for a synthetic modality therefore ruptures both of the standard conceptions of actuality available until now in the metaphysics of modality. On the one hand, David Lewis's 'possibilist' conception implies that every (concrete) l.p. world is equally 'actual.' On the other hand, Alvin Plantinga's theory holds that a single (ersatz) l.p. world is uniquely actualized.⁽⁶⁸⁾ If these extremes represent an exclusive alternative, then the paradigm case of actuality is either (1) an indexical property shared throughout the sphere of l.p. worlds, or (2) the complete actuality of a single possible world . But it is the exclusiveness of this alternative which my theory of synthetic modalities challenges. If synthetic modalities exist, then the paradigm case of actuality is the existence of sub-maximal states of affairs, which determine the -relevance of a limited sphere of l.p. worlds. This is the deepest intuition at work in the theory--a conception of actuality which succumbs neither to Lewis's nor Plantinga's extreme alternatives.⁽⁶⁹⁾ And more importantly, whether they like it or not, their conception of feasibility already implicitly commits Molinists to the coherence of such this 'third' conception of primordial actuality.

Once we grasp this, it is easy to see why the comparison of feasibility to causal or nomological modality is so intuitive. If Molinism is true, then there is some sense of metaphysical priority which allows for 'world-neutral' truths, and hence by parity of reason it will also be possible to interpret nomological modality as a 'kind' of synthetic modality in my sense. Imagine that instead of CF, we have a set of indeterministic natural laws L which are true in a world-neutral sense, like truths of middle knowledge, due to the metaphysical 'priority' of their truth to the full actuality of . Since they are indeterministic, the physical universe these laws constitute will exist in a plurality of logically possible worlds. The 'sphere' of worlds in which L is satisfied thus becomes the 'modal range' for nomological possibility. Events such as exceeding the speed of light, for example, will only exist in worlds outside the -relevant sphere L establishes, and so such an event will be physically impossible for this universe. Although I think this way of conceiving causal modality has much to recommend it on its own terms, my only point here is that such an interpretation of causal possibility has the same sort of metaphysical basis as Molinist feasibility.

Someone is bound to object that I still have not explained this "metaphysical priority" that some segments of enjoy over the complete actuality of itself, when they constitute the modal range for a synthetic modality. But the Existentialist is willing to let the Molinist's account of God's providence and middle knowledge decide exactly what this metaphysical precedence means. It is sufficient that the Molinist account implies that there is some type of priority P, such that the -relevance of a sphere of l.p. worlds is 'P-prior' to God's decision to actualize a single possible world as . As Craig indicates, 'P' cannot be a causal priority, but beyond that it does not matter to the Existentialist what P is, as long as the Molinist considers it sufficient to make it impossible for God to actualize a world that is not -relevant. The Existentialist is only interested in the formal features which result from P-priority being attributed to the truth of a logically sub-maximal set propositions. For whatever P is, the P-priority of such a set of propositions (counterfactual or otherwise) will establish a kind of synthetic modality, as explained. And this synthetic modality acts as a real constraint on God's creative decision, even if it does not cause any particular divine decision. It makes certain creative actions (such as actualizing W' in which Curley refuses the bribe) synthetically impossible for God. And for this to be the case, the synthetic modality must exist P-prior to the existence of particular free creatures. This is as much as the Existentialist will need to demonstrate the untenability of the Molinist position.⁽⁷⁰⁾

(D) The Truths of Middle Knowledge as 'Hard Facts'

As a result of our analysis of Molinist 'feasibility' as a kind of synthetic modality, we can locate the sense in which the truths of middle knowledge are 'hard facts' relative to the choices they entail when a person is actualized in a given circumstance. In his objection to premise (12) of Adams's "New Anti-Molinist Argument," Craig urges that Adams has not shown that "the content of God's middle knowledge is a 'hard fact.'"⁽⁷¹⁾ Yet divine middle knowledge about me includes only world-neutrally true counterfactual of freedom such as , which says that 'JD would do A in C.' It is the world-neutrality of this truth which makes it unfeasible for me to refrain from doing A in C. Thus:

(1) is a truth of middle knowledge.

(2) It is unfeasible for God to actualize an l.p. world in which JD refrains from doing A in C,

i.e. it is unfeasible for God to make false (follows from 1 according to Molinism).

(3) The truth of is P-prior⁽⁷²⁾ to the complete actualization of (interpretation of 2).

(4) The proposition E, which says "JD does A in C," is true (supposition).

(5) It is feasible for God to refrain from creating JD (according to Molinism).

(6) It is unfeasible for JD not to exist and for JD to do A in C

(logical impossibility entails unfeasibility, although not the reverse).

(7) Hence it is feasible for God to make E false (from 5 & 6).

(8) Hence E is true only relative to , not P-prior to the complete actualization of

(interpretation of 7).⁽⁷³⁾

Notice that (8) interprets (7) by the same criterion as (3) interprets (2). The truth of E cannot precede the complete actualization of , since Molinism maintains (5), whether (4) is true or not. Hence, in whatever sense the truth of is 'prior' to the complete actualization of , the truth of E is not 'prior' to the complete actualization of . E is true only relative to various l.p. worlds, including the one that turns out to be . Then if we grant that

(9) P is a transitive priority relation,

(10) the truth of is P-prior to the truth of E

follows from (3) and (8). Thus we have shown that if Molinism is true, there must be a univocal sense of priority at work in results such as (3) and (8), and that irrespective of what it is, if it is transitive, the truths of middle knowledge are prior in that sense to the actions and choices they are about. Thus we can say that the truth of is a hard fact relative to the truth of E.

This overcomes both Craig's objection about the equivocal use of "explanatory priority" and his objection that Adams's requirement for incompatibilist freedom tacitly depends on a demonstration that truths of middle knowledge are 'hard facts.' One doubt remains, however. In anticipation such a reconstruction, Craig says that if a univocal meaning for the needed priority relation is found, "I suspect that any such notion will be so generic and weak" that it will lead to absurd conclusions unless we "deny its transitivity."⁽⁷⁴⁾ On my approach, however, the possibility of such a reductio of P's transitivity does not arise, because I have shown that Molinism implies the existence of a priority relation which can be used univocally throughout the argument, without needing to specify it. The onus is thus on the Molinist, not his or her critic, to show that there is some non-transitive relation of priority which is adequate to make truths of counterfactuals 'world-neutral' in the way necessary for them to restrict the worlds God can actualize. And of course, unless this result is preserved by the interpretation of P, Molinism will not be of any use for the purposes for which its defenders want it, such as answering the problem of evil and explaining the consistency of providence with our moral responsibility for our actions.

It is the Molinist, then, who has to come up with a sense of P strong enough to establish the synthetic modality of feasibility, and yet still weak enough to remain intransitive. And since the metaphysical priority which makes contingent truths 'world-neutral' in the sense necessary to ground any real synthetic modality is almost bound to remain a metaphysical primitive, the Molinist has little chance of vindicating his or her position.

Can the Molinist urge that although that P is a metaphysical primitive, our common intuition is that it is intransitive? This would be very unconvincing, because to most philosophers of religion, even many of those sympathetic to Molinism (let alone laypersons), it comes as a complete surprise that its defenders would even consider suggesting that the truth of a counterfactual of freedom in middle knowledge is a 'soft fact' in the same way as truths in the past about what I will do in the future are considered 'soft' by Ockhamist theories of divine foreknowledge. Rather, it is usually assumed that a Molinist theory of middle knowledge as the basis for divine foreknowledge is an alternative to an Ockhamist interpretation of divine foreknowledge.⁽⁷⁵⁾ For the Molinist who suggests that P is intuitively intransitive claims that (a) there are truths about what I would do 'prior' to my existence; (b) these truths are 'prior' to the divine decision to create me; and yet (c) these truths are posterior to my actual actions and choices! If intuition is our measure, then it must be frankly confessed that the circle which such intransitivity in 'priority' implies seems to border on insanity: when caught in such a circle, the very notion of our existence becomes so pliable that it is virtually reduced to meaninglessness. Clearly, the 'soft fact' Molinist's claim is about as anti-intuitive as a philosophical claim could possibly be.⁽⁷⁶⁾ The natural view is that, if Molinism makes sense at all, the truths of its counterfactuals of freedom constitute facts that are "hard" not only at an earlier time (as in Ockhamism), but even metaphysically prior to all time.

But I think we can say something stronger. There is good reason to think that on the Molinist account, the 'metaphysical priority' attributed to the truths of counterfactuals which establish feasibility must be understood as transitive, even if nothing else can be said about it. The reason is that otherwise, the division of logically possible states of affairs into feasible and unfeasible will not even be 'prior' to God's creative decisions. One can see this as follows:

(1) Suppose that Molinism is true.

(2) Suppose that God makes some free decision D, such as the decision to actualize some

possible world W (consistent with 1).

(3) There is then a contingently true proposition D* which says 'God decides D.'

(4) D* must be true only relative to specific l.p. worlds, i.e. D* is not true P-prior to the

complete actualization of .

Proof (By reductio)

(4i) Suppose that D* is 'world-neutrally' true P-prior to the actualization of a complete l.p. world.

(4ii) Then it would be unfeasible for God not to decide D (from 2, 4i and the interpretation we have given for 'unfeasibility').

(4iii) Then W is the only feasible l.p. world (4ii & our interpretation of feasibility).

[Explanation: since D is the decision to actualize W, if refraining from deciding D is unfeasible, the actualization of any other l.p. world is unfeasible].

(4iv) But in that case, the truths of counterfactuals of creaturely freedom are irrelevant for divine decision-making.

(4v) By supposition (1), (4iv) is false.

This shows us why a Molinist cannot hold that God's decision to actualize a particular l.p. world is P-prior to the complete actuality of . If it was, then this decision would have to be neutral between a plurality of worlds, but it would reduce the sphere of the feasible to a single l.p. world. In that case, the counterfactuals of creaturely freedom would be irrelevant for every divine decision they are supposed to inform--since these decisions themselves would already be the only feasible option for God. The actuality of decision like D is therefore a part of , but they are not actual P-prior to the actuality of as a whole. Then we continue:

(5) The truth of a counterfactual of creaturely freedom such as ('Curley would take the bribe

in C') is P-prior to the complete actualization of (as established by our analysis of the

meaning of feasibility).

(6) Suppose (for reductio) that P not transitive

(7) Then the truth of is not P-prior to the truth of D*.

(8) But the truth of is P-prior to the truth of D* (according to Molinism--premise 1).

(9) Thus P is transitive (from reductio of 6 by 7 & 8).

What this argument shows is that if P-priority is not a transitive relation, the truths of counterfactuals of creaturely freedom that makes up Divine middle knowledge cannot play the role Molinism requires them to play. We know that God's decision is not P-prior to the complete actuality of (as per 4), and that the truth of counterfactuals such as is P-prior to the complete actuality of (5), but from these premises, we can only infer that these truths of middle knowledge are prior to the divine decisions whose consequences that are supposed to restrict, if P is transitive [(4 & 5) 8 only if P is transitive]. So the Molinist always implicitly treats the priority he implicitly grants to truths of middle knowledge as a transitive priority relation. The possibility of making a case for a 'soft fact' interpretation of truths of middle knowledge thus looks very bleak indeed.

(E) Why Molinism Must Attribute Truths of Middle Knowledge to Our Agency

Truths of middle knowledge, then, can only be coherently interpreted as putative hard facts. But by itself, this concession will not cause the Molinist to capitulate completely. As Craig urges at the end of his response to Adams, "even if they [c.o.fs] were in some peculiar sense explanatorily prior to our actions because they are true and known by God logically prior to categorical contingent propositions, that would not be incompatible with the freedom of our actions."⁽⁷⁷⁾ To refute this, the best strategy for the Existentialist is to show that if Molinism is true and truths of middle knowledge are hard facts, whatever remaining sense our actions are "free" is insufficient for moral responsibility for our choices and actions.

It is a primary concern of Molinism to maintain its consistency with the conditions of agency necessary for moral responsibility. This is why, as Adams acknowledges, Haskar suggests that "Molinists may wish to hold that in the case of a true counterfactual of freedom,⁽⁷⁸⁾ it is the agent of the free action described in the consequent who brings it about that the conditional is true."⁽⁷⁹⁾ I have vindicated Haskar's view that the Molinist cannot coherently attribute these truths to the agent in the sense of making them soft facts dependent on the agent's actual choices. Nevertheless, if even if the world-neutral truths of counterfactuals of freedom admitted to be are 'hard facts,' I think the Molinist has no option but to hold that the agent is in some sense⁽⁸⁰⁾ still the author of truths of middle knowledge that refer to her possible choices and actions. The reason why, as we will see, is brought out by an 'initial' Existentialist argument that unfeasibility of refraining from my actions voids my moral responsibility for them.

The Existentialist finds the Molinist's counterfactual 'power to do otherwise' ephemeral not because Molinism says that I am committed by my essence to do what I will do in C, but rather because Molinism implies that the truth of the c.o.f. (or fulfilled 'subjunctive conditional,' to be accurate) which entails my actual choice is metaphysically prior to the choice itself. The truth of this c.o.f. makes it unfeasible for me to do otherwise than it describes. Thus if I am not somehow the author of the truth of my c.o.f. itself, the fact that it is logically possible for me to choose otherwise in C is irrelevant: I am no more responsible for it than if I had been caused to do it. The Existentialist insists that moral responsibility requires more than the logically possibility of doing otherwise: it also requires the synthetic possibility of doing otherwise, in every valid sense of synthetic modality.

Molinism allows us only 'ephemeral' power to do otherwise because feasibility is a form of synthetic modality, like nomological possibility. I am in a position relative to the truth of my counterfactuals of freedom analogous to the position I am in relative to causal laws, because the truth of both, although contingent, is metaphysically prior to my processes of deliberating, deciding, willing and acting as an actualized person.

William Alston has helped to make clear the force of this objection by showing that any conception of divine foreknowledge which acts like causal necessitation is incompatible with libertarian freedom of the will. In a 1985 paper, he argues that a condition Plantinga proposes in God, Freedom, and Evil is not sufficient for "significant" libertarian freedom in the full sense:

P: "It was within Jones's power at t₂ to do something such that if he had done it, then God would not have held a belief that in fact he did hold."⁽⁸¹⁾

Alston points out that this power P only guarantees the truth of the compatibilist causal counterfactual that "the actual situation in which Jones found himself is such that a contrary decision, inserted into that situation, would give rise to a contrary action."⁽⁸²⁾ In other words, this power P only says that no antecedent conditions and causal laws causally necessitate Jones's doing at t₂ what God foreknows he will do. In this sense, Jones has a "real" and not just logical possibility of doing otherwise. But Alston argues that Jones will not have the logical possibility of doing otherwise even with P:

For clearly God believes that p at t₁ entails Jones does not do something at t₂ such that if he had done it God would not have believed that p at t₁. And so if divine beliefs are "antecedent conditions" in the relevant sense, i.e. hard facts about the time at which a given such belief is held, then Plantinga's condition for something being within a person's power [i.e. significant freedom] is not met by Jones and power P.⁽⁸³⁾

Alston acknowledges that in God, Freedom, and Evil, Plantinga does not draw the distinction between "hard" and "soft" facts, which becomes crucial in his later Ockhamist view of divine foreknowledge. But even if divine beliefs about an action at a time earlier than that action can be treated as 'soft facts,' the analogous beliefs based on middle knowledge cannot be, as we have seen. Thus we should be able to construct an analogous 'Alstonian' argument that works against Molinism, even if the Ockhamist can escape the original one directed at him.

The key is to realize that "real" possibility means possibility in a synthetic sense, but this is not limited to causal modality: Molinist 'feasibility' is also synthetic in the relevant sense, as we have seen in detail. The Molinist is able to make power P look sufficient to guarantee "real possibility" of doing otherwise only by suppressing the fact that a 'hard' truth of middle knowledge constitutes a different sort of real (or quasi-nomic) impossibility of doing otherwise. Consider the following example.

There is a true Molinist counterfactual saying that if Jones were to exist in a given situation C at t₂, he would do X; yet Jones also has power P at t₂.

Suppose that at t₁ God beliefs that Jones does X in C at t₂. But this is foreknowledge of a 'soft fact' that Jones has the causal power to change by making a decision at t₂. Moreover, let us even extend power P to say that Jones has a real causal possibility of not doing X at t₂: i.e. if he decided to do something else, no causal factors would block the fulfillment of that alternative decision. But even with P in this extended sense, the truth of denies him the real feasibility of deciding differently in the first place: thus his causal power to implement a different choice, were he to make it, is not much use to him. Compatibilist freedom secured by P in this case hides a synthetic limitation which functions as a "real" impossibility in a way analogous to causal necessity.

Thus the Molinist cannot reject this comparison of causal laws and patterns among true counterfactuals of freedom by saying that in life, we are free for our c.o.fs in a sense in which we are not for natural laws. Nor can the Molinist retort that the comparison simply misses the logically contingent nature of Molinist true c.o.fs, since we have shown that is not the issue. The Molinist only has one remaining way of asserting a disanalogy between nomological impossibility and the unfeasibility of acting contrary to the world-neutral truth of one's c.o.f. He can point out that causal laws, especially if interpreted in a strong modal fashion (as I have done), embody limitations on action which confront us as absolutely alien to our volition. The reason is that our agency has nothing to do with creating these limitations: we identify with pure physical necessities even less than with desires and impulses that enter our conscious life without our consent. By contrast, the Molinist must say, the so-called 'restrictions' imposed by the prior truth of our counterfactuals of freedom arise from, or in some sense just are, our agency. We are identified⁽⁸⁴⁾ not only minimally (as with the conscious impulses) but wholly, with the truth of our c.o.fs. Far from being alien to us, our true c.o.fs are what is most authentically and absolutely our own. If truths of middle knowledge are admitted to be 'hard facts,' then asserting that our agency is sufficiently associated with these facts to allow moral responsibility for our actions is the Molinist's only way of responding to the 'initial' Existentialist critique.

(D): The Existentialist Refutation of Molinism

We have now forced the Molinist's hand. The association he posits between our agency and the contingent but world-neutral truth of our counterfactuals of freedom is the only way he can resist the initial objection that the priority of these truths to our actual existence restricts our freedom in a sense analogous to causal necessitation. It is only at this point that the Existentialist's main critique of Molinism can begin.

It is assumed by most analytic philosophers familiar with the issue that the worst problem for the Molinist is that he can give no account of what makes true the set of an individual's c.o.fs that make up middle knowledge.⁽⁸⁵⁾ But the limited force of this objection has only served to obscure the real problem. For this standard objection fails to recognize that the Molinist does claim that we are identified with our true c.o.fs in the practical sense. Even the 'soft fact' interpretation of Molinism which we have critiqued is an extreme attempt to preserve this essential element in the Molinist's program. The Existentialist's objection is more damning than the standard one, because it aims at this real heart of Molinism.

The Existentialist's main argument begins with two conditions I have already alluded to, which both Molinists and Existentialists typically recognize as necessary conditions for moral responsibility for an action:

(1) The individual must have multiple logically possible alternatives for any action.

(2) To be an action, the description (including the purpose or end) under which the individual acted (i.e. the maxim) must be attributable to the individual's agency. In other words, she must be identified with that maxim, intending her action under that teleological description.⁽⁸⁶⁾

The Molinist accepts condition (1). To accommodate (2), he must somehow explain how it is possible for us be identified with the true c.o.fs which determine the maxims we would adopt in each possible circumstance. The difficulty arises, however, from a third premise required by Molinism but rejected by Existentialism:

(3) Truths about what a person would do in different circumstances are made true, in some manner, metaphysically prior to the person's actual existence.

From (1), (2), and (3) taken together, there arises an antinomy which I will call the Antinomy of Authenticity. The general problem this antinomy poses for any sort of Molinism can best be understood in terms of a trilemma.

The first horn of the trilemma is Suárezian Molinism, which maintains that the truth of a person's counterfactuals of freedom is a primitive and inexplicable 'fact.'⁽⁸⁷⁾ This is obviously the position against which the 'standard' analytic objection to Molinism is directed. This Suárezian position saves conditions (1) and (3) through the contingent but world-neutral ontological status of the truth of c.o.fs. But it is unable to satisfy condition (2). Saying that the truth-maker of the individual's counterfactuals of freedom is unanalyzable divorces the truth of his c.o.fs from an individual's agency: in that case, the individual cannot be authentically identified with his true c.o.fs.⁽⁸⁸⁾

The second horn of the trilemma is what I will call sophisticated Molinism, which I think represents the best alternative the Molinist has.⁽⁸⁹⁾ The sophisticated Molinist answers the objection to the Suárezian position as follows: on pain of destroying an individual's moral responsibility for all his actions, she says that the individual's own agency is somehow the truth-maker of his c.o.fs. Now she appears to have a real explanation for why the individual is identified with the truth of his counterfactuals of freedom, and hence why all his actions in accordance with these c.o.fs are morally imputable to him. The sophisticated Molinist thus appears to satisfy both conditions (1) and (2). But her way of preserving (2) contradicts condition (3), to which any Molinist is committed. 'Soft fact' Molinism is precisely an attempt to deny this conflict, but as we have seen, the 'soft fact' interpretation of the truths of counterfactuals of freedom fails when their world-neutrality is recognized. The contradiction between (2) and (3) is then inevitable: for in what conceivable sense could person's agency, purely as possible and metaphysically prior to its actual existence, be the authentic the source of anything? We might accurately characterize Existentialism by saying that its fundamental intuition is that a logically possible person, prior to their existence, is not the source of anything that identifies them with any element of volitional character. To deny this by denying the transitivity of the relevant sense of 'priority' is to obscure the very meaning of existence--as well as to strip truths of middle knowledge of their power to restrict the results of divine choices (as we saw).

Thus the problem with 'hard fact' Molinism (the only coherent kind) is not that it leaves the individual with no logical possibility of doing otherwise. Rather, the problem is that it destroys the agent's moral responsibility for her actions by imposing on her a synthetic necessity to act and choose in a certain way, when this synthetic necessity itself cannot be attributed to her agency. The requirement of authentic identification which is necessary for moral responsibility cannot be satisfied by satisfying conditions (3) and (1) at the expense of (2), as 'hard fact' Molinism does.

Finally, the trilemma allows for one other conceivable resolution, which we might call Essentialism. The Essentialist says that it is simply part of an individual's essence that his agency would choose certain actions, or would will its acts for the sake of certain ends, rather than others. Essentialism thus conceived satisfies (3) and (2), but only at the price of violating condition (1) by giving the individual an agency not logically free to do otherwise in action and motivation. On this account, the individual's true counterfactuals of freedom are logically essential to them. Such an Essentialist position escapes the 'antinomy of authenticity' only if we accept that condition (1) is not necessary for authentic identification with (and thus moral responsibility for) our actions.⁽⁹⁰⁾

But as we have seen, this is not what the Molinist claims. For the Molinist, the individual's c.o.fs are free in the sense of being logically contingent for him, thus satisfying (1). Moreover, this is seen as a necessary condition for moral responsibility. But (3) entails that there can be no intelligible sense in which his agency can possibly have identified him with these contingencies--and this violates condition (2). The notion of a possible but non-actual agency which nevertheless contingently makes some set of choices attributable to itself in se, is nonsense--like the round square cupola.

Thus agency, as explicitly conceived by sophisticated Molinism--and implicitly conceived in all Molinism--makes no sense at all. If this conclusion surprises you as extreme, remember that the Molinist does not merely claim that some set of c.o.fs is possibly adopted by the individual's agency, i.e. would be attributable to him if a given logically possible world were actual. That view might make sense. But no--we are told that, out of all the sets of c.o.fs an individual has in different l.p. worlds, one set is supposed to be actually (but contingently) attributable to him, prior to his existence and in fact prior to the complete actualization of , thus restricting the set of l.p. worlds in which it is feasible for him to be actualized. Indeed, because of these counterfactual truths, which moral responsibility for our actions requires us to attribute to the possible person's agency, God may even chose not to actualize the possible person to whom they are attributed. The Existentialist refutes Molinism by pointing out that this stronger, world-neutral truth of one's c.o.fs which Molinism requires, cannot be attributable to the agent. The essence of any agency is to act, and that is something it cannot actually do prior to being actual itself. This is the fixed point on which the Existentialist stands.

The problem here is highlighted by David Lewis's remarks in his recent effort to evaluate the results of basing a free will defense on Molinism. Lewis argues that the Molinist can only mean that the truths of unfulfilled counterfactuals of freedom are modal primitives, because they can supervene neither on the will of God nor on personal agency. Lewis first asserts that fulfilled counterfactuals of freedom, such as the truth of the fact that Judas would betray Christ, must be made true by the agent's free action. He continues:

Unfulfilled counterfactuals of freedom are very different. They're not rendered true by the free choice of the agent, since they concern choices that never actually take place. Some of them even concern agents who never actually exist.⁽⁹¹⁾

But this actually involves a double misrepresentation. First, the Molinist really treats fulfilled and unfulfilled counterfactuals of freedom in the same way: both are true prior to any choices the fully actualized agent makes in time.⁽⁹²⁾ Second, as I have argued contra Suárez, the Molinist must also maintain that in another sense, all of the individual's counterfactuals of freedom, fulfilled and unfulfilled, are "rendered true" by the individual's "free agency." Otherwise, the actions arising from the truth of these c.o.fs in actualized circumstances would not be morally imputable to the agent.

Lewis simply assumes that the Molinist cannot mean that a person's free agency is the truth-maker for their counterfactuals of freedom, because he sees so clearly why such a claim would make no sense. But Lewis fails to see the 'antinomy of authenticity' in which the Molinist is caught: unfortunately, the Molinist must (to preserve moral responsibility) make precisely the incoherent claim about individual free agencies which Lewis is too generous to attribute to him.

Nevertheless, it is hard to see immediately that there is no rational way the Molinist can claim that my agency is identified in the necessary sense with my actions. For surely my true counterfactuals of freedom describe what I would do, what choices my agency itself would generate. Doesn't this fact--that the counterfactual is about my agency--establish a sufficient connection for identification? Accordingly, the sophisticated Molinist might try to respond to the Antinomy of Authenticity as follows: "A true counterfactual of Curley's freedom tells us what Curley would do in situation S. But in relating this to Curley's agency, what I mean is that Curley's agency would bring about the action predicted by the true counterfactual in S. The fact that this truth holds metaphysically prior to Curley's being actualized in S is no object to its being attributable to his agency, because it describes what his agency would bring about."⁽⁹³⁾

This is no doubt what the sophisticated Molinist must think. But this response misses the point of the Existentialist's objection. The Existentialist's point is that what a person's agency would do must itself be attributable to his agency, if the person is to be morally responsible for what he would do. If we accept the Molinist hypothesis that there is such a truth for every possible circumstance Curley could be in, including the ones he actually passes through in the real world, then his actions in every actual circumstance reflect the truths about what his agency would do in those circumstances. If his agency is not responsible for these truths about what it would do, then Curley is not responsible for his actions. But there is no way his agency can be responsible for the truth of these fulfilled counterfactuals about his agency, or any other counterfactuals true for it.

A subjunctive relation of the agency to its actions within the counterfactual is thus not good enough for moral responsibility. The reason why is not hard to see in light of our previous analysis of truths of feasibility as synthetic necessities for the person analogous to causal necessities. The Existentialist's argument against Molinism can be summarized as follows:

i. To be morally responsible for an action X, an agent A must be volitionally identified with that action [principle (2) above].

ii. An agent A is alien to (non-identified with) any synthetic necessity which he has not brought about by his own agency [this is clear in the case of causal necessity].

iii. A true synthetic counterfactual C of A's freedom telling us that he would do X in situation S is a synthetic necessity for A in the following sense: its actuality precedes the actualization of any particular complete logically possible world, and hence it holds for A in every feasible world [implied by Molinism].

iv. An agent A is alien to (non-identified with) any state of affairs X that is logically or synthetically necessitated by the actuality of another state of affairs Y with which he is not identified [corrolary to principle (2)].

In support of (iv), the Principle of Transitivity of Alienation, I should point out that this principle cannot be in dispute between the Existentialist and Molinist. For it is the reason why non-compatibilists accept that an agent cannot be morally responsible for an action which is causally necessitated by some prior state of affairs, if he did not freely bring about that causally prior state of affairs. The key is to see that the transitivity of non-identification (or alienation) holds in general for any relevant form of synthetic necessitation, and not just for causal necessitation. If X synthetically necessitates Y, and individual I is not identified with X, then individual I is not identified with Y. We continue:

..how an atemporal God could act at different times, in different ways, while never himself changing. The story is simple: There is one eternal divine act outside of time that has a great number of different effects in time, at different times.⁽⁹⁶⁾

This picture of the divine agency leads to Molinism in two ways. First, in order for God so conceived to know in atemporal eternity all that is necessary for the multiple effects of his single divine act to include responses to creaturely initiatives and choice, one has to "defend something like the Molinist conception of divine knowledge.."⁽⁹⁷⁾ But in addition, this divine agency which atemporally and all at once performs in a single act everything it will ever do from our perspective intuitively inspires the sense that there might be an analogous notion of human agency. Against this background, the notion that the truth of a person's counterfactuals of freedom must somehow be attributable to their agency, as if they had made them true in act prior to temporal existence, must not have seemed strange.

In this, the early Molinists were not only led astray by a wholly inadequate conception of divine agency.⁽⁹⁸⁾ In addition, they failed to see that even if such a conception of agency could work for God, it could not properly be extended to other persons. For on this account, God is at least actual in all possible worlds, and His 'single ultimate act' could therefore be thought of as world-neutrally actual⁽⁹⁹⁾, whereas the 'single act' of an individual human person, by analogy, would implicitly attributed to his agency prior to its very actuality. This implicit but unrecognized equivocation was the root of the fatal flaw in Molinism.

III. Feasibility Without Molinism: Our Pretheoretical Modal Intuitions About Volition

(A) The Ground Underlying Molinist Intuitions

If all this is accepted, however, the debate is far from won for the Existentialist. Plantinga is still right that intuitively, we think we know what it would mean for something like counterfactuals of creaturely freedom to be true. Indeed, these intuitions are a crucial element in a vast preponderance of our rational deliberations about actions that will create circumstances in which others are expected to react. We think literally all the time about things such as what A would think of me if I gave her X for her birthday, or what B would do if he were offered a position at the college, and so on. Thousands of everyday interactions which seem to involve something like 'counterfactuals of freedom' range from the narrowest egoistic anticipations of others' responses in bartering, to developing friendships and making much longer range plans that depend on possible future reactions of others to situations we either consider bringing about, or anticipate coming about without our help. Every kind of rational 'contingency planning' apparently depends on the intuition that there are truths about what persons would do in different institutional arrangements, faced with different set of options, and so on. Existentialists must also accept this: in their terms, we could say that reasoning about social existence in all its forms, and indeed our very sense of being in a lifeworld inhabited by other agents with whom interaction is possible, seems to arise from the intuition that there are true counterfactuals of agent-freedom--the intuition which Plantinga has claimed is 'natural.'

This at least explains why Molinism has a certain intuitive appeal. But the ground of these modal intuitions requires critical analysis before we can tell if it is really a ground for Molinism, or if some other theory might not better capture our intuitions. What underlies them? Why, at bottom, do we believe that we can meaningfully think about what people would do in all sorts of more or less completely specified hypothetical circumstances? When we reflect on this phenomenon--which we so often take for granted--it seems obvious that the reasons we have such beliefs always have to do with some aspect of what in general we call character. Our belief that there is a truth about what individuals would do arises from our sense that they have predispositions, habits, and tendencies which give substance to their volitions. Robert Adams provides a convincing example of this:

There does not normally seem to be any uncertainty at all about what a butcher, for example, would have done if I had asked him to sell me a pound of ground beef, although we suppose he would have had free will in the matter. We say he would certainly have sold me the meat, if he had it to sell. What makes us regard it as certain? Chiefly his character, habits, desires, intentions, and the absence of countervailing dispositions. (He would have had no motive to refuse me).⁽¹⁰⁰⁾

This example should prompt us to note, however, that there are different 'levels' of character, and different types of 'facticity' involved in more or less determining what a person would do, at each of these levels. At the 'lowest' level, we can agree with Plantinga that "If Robbins had slipped and fallen at Thanksgiving Ledge, he would have been killed," because of the law of gravity.⁽¹⁰¹⁾ Here the facticity, which is determining rather than merely predisposing,⁽¹⁰²⁾ is purely part of 'physical character,' and thus not part of 'character' in the sense related to morally significant actions at all. Then there is physiological character: if I were to hold a lit match to his palm, he would instinctively try to withdraw his hand. We still have not reached character in the moral sense. Next, there are broad tendencies which make up 'economic laws' based on the 'character' of people purely as buyers and sellers in different types of markets: Adams's butcher obeys these laws. Similarly, if I were to raise the price of gasoline nationally, Jones would still buy almost as much of it, because his demand for it is inelastic. We might think these laws do not hold for all actual people, since they are to some extent culture-relative and dependent on a value system which says it is okay to order one's decision preferences purely as a rational utility maximizer when visiting the supermarket, if not in other contexts. In other cultures--and certainly in other possible worlds--where these values are not operative, persons would make different decisions in the same or similar circumstances.

But even at this stage, we have only reached the level of impulses, desires, and preferences which people typically allow to dictate their economic behavior (at least to a large extent), without necessarily taking an attitude towards these preferences, even when acting on them. Most if not all of what we commonly think of as moral character relates to higher-order attitudes one takes towards one's first order impulses, desires, and preferences.⁽¹⁰³⁾ Being courageous, for example, does not mean that one would not have a fearful impulse if one found oneself in a situation of grave danger. Nor does it simply mean having the second-order desire or volition not to act on one's fearful impulse (for an unwilling coward--one who recently decided he no longer wants to be a coward--would have this volition). Rather, it means a settled and effective higher-order disposition to resist acting on one's fearful impulses, a lasting tendency to the right sorts of higher order volitions, which has brought one's first-order will under its control.⁽¹⁰⁴⁾

I believe the intuition that there can be morally significant facts about what a person would do--facts for which they are morally responsible even if they were never in the circumstance to do these actions--arises from our sense that, on this higher order level of character, quite independently of physiological factors and other basic needs and impulses that cause desires, volitional dispositions⁽¹⁰⁵⁾ can make it true (or at least probable) that I would or would not do something in a hypothetical situation. When the modal intuitions to which the Molinist appeals are actually examined, their ground turns out to be entirely characterological. This basis for Molinist intuitions is apparent even in the language used by the late medieval authors. Robert Adams notes that Suárez believes a person has one of two properties, "either the property of being a possible agent who would in s [a situation] freely do a, or the property of being a possible agent who would in s freely refrain from doing a." It is no accident that Suárez called such a property a "habitudo."⁽¹⁰⁶⁾ The fact that habits of will or volitional dispositions are the basis for our belief in apparently Molinist counterfactuals also explains, I think, why intuitively we do not always believe there is some truth of the matter about what someone would do (i.e. why we believe that the principle of Conditional Excluded Middle fails). If I had been born on a futuristic Martian colony and had the chance to perform some heroic feat there, would I have taken it? Probably we have no intuitive sense that there is a truth of the matter for a case thus specified. But if we added in my complete circumstances there, including every fact about my character, would not our intuition become stronger that there is a truth about what I would do? The reason is that in moving to the maximally specified circumstance, we have captured the factors that really matter, and much more. What matters is not actually the complete circumstance, but rather my character and those aspects of the situation to which aspects of my character relate in significant ways.

(B) The Challenge to Existentialism.

The same intuitions about the significance of character which remained unanalyzed but implicit in traditional Molinism are the root of dissatisfaction with certain 'simplistic' versions of Existentialist libertarianism. Because of the power of character, we have a commonly shared modal intuition that freedom as pure negative liberty, i.e. the mere logical possibility of our bringing about different alternative actions, by itself provides an inadequate picture of human agency. Some philosophers accommodate this intuition by arguing against a libertarian conception of freedom altogether.⁽¹⁰⁷⁾ Many others have felt that even if there is always an element of negative liberty in our actions, we never find ourselves equally at liberty to choose any alternative (at least not in choices that matter). For this reason, people rebel against any description of free will which sounds as if it is emphasizing its spontaneous, arbitrary side: for example, Jean-Paul Sartre's picture of human transcendence as a level of radical freedom, in which we are forced to make choices between alternatives that are ultimately arbitrary because no sufficient reason can ground a stable preference for one alternative over another.⁽¹⁰⁸⁾ When contrasted with such 'simplistic' Existentialism, Molinism seems attractive because it attempts to accommodate libertarian negative freedom without implying sheer arbitrariness in our choices.

The problem with simplistic Existentialism, then, is that there are limitations on our freedom of choice which are not logical or physical impossibilities, nor mere results of socialization and psychology, but something else. This 'something else' is a synthetic modality of volition which explains certain patterns of free choice not from physiological or economic grounds, but from characterological grounds. I call this a form of modality because it is supposed to refer to a type of possibility and impossibility. But it is synthetic because the sense of possibility and impossibility it involves is distinct from the broadly logical. As we already saw in §II, to say that a logically possible world is feasible can usefully be thought of as characterizing it as "possible" in this synthetic sense.

In comparing Molinist contingently true counterfactuals of freedom to contingently true causal laws, I used the locution personally possible for the specific version of synthetic modality operative in traditional Molinism: thus for example, it is 'personally impossible' (although logically possible) for Curley to reject the $20,000 bribe if offered it: in all feasible worlds in which Curley exists, he takes the bribe. This explains how the type of fact involved in a counterfactual of Curley's freedom being true becomes a "facticity" in the Existentialist's sense--namely, something that restricts one's broadly logical freedom in substantive ways.⁽¹⁰⁹⁾

What is different about this type of facticity, however, is that it is not supposed to be alien to one's will, but rather an integral part of it. Our intuition that broadly logical freedom is restricted by facticities of this special kind reflects our conception of volitional character. This analysis of the ground of our intuitions explains why the Molinist wants to say that the truth of our counterfactuals of freedom goes back to our agency⁽¹¹⁰⁾--for in fact, such counterfactuals are meaningful only to the extent that they express a level of deep character we expect to find in a person.

The Molinist, of course, does not think of 'feasibility' as a type of synthetic possibility in the first place--but as we have seen, this interpretation of feasibility is inevitable given the role Molinism itself attributes the truths of middle knowledge. Moreover, the intuitions which underlie the Molinist's own version of 'feasibility,' pretheoretically do express a notion of synthetic modality, in particular, a modality of volition. It is crucial to see that 'feasibility' in this intuitive sense corresponds to some unanalyzed notion of a synthetic modality--quite independent of any particular account of what grounds this modality, such as the particularly Molinist conception that world-neutral truths about individual counterfactuals of freedom provide its basis. The very idea of a synthetic modality of volition must be separated from particular theories about its metaphysical grounds.

If you have any doubts about whether we really have a pretheoretical notion that is properly described as a synthetic modality of volition, consider the following examples from two philosophers working quite independently of Molinism. In discussing the notion of power, Tom Morris develops what he calls the concept of "moral capability," which he distinguishes from all sorts of enabling means, such as cognitive and mechanical abilities, skill, practical knowledge, opportunity, and so on. Suppose, Morris says, we think of an example in which Jones has all these requisite means to kill someone obnoxious to him, and yet his neighbor says, "Jones could not possibly do anything like that. He's not capable of such behavior."⁽¹¹¹⁾ Notice that capability in this sense refers to a certain pretheoretical modality. To be incapable of murdering someone is for it not to be "possible" in a certain sense, which is clearly distinct from the broadly logical, causal, as well as physiological and economic senses:

Often, when we say that a person is not "capable" of a morally dubious or improper line of action, what we mean to indicate is that doing such a thing would be contrary to a firmly entrenched character that person has, that the desire or inclination to perform the action is not within the range of his possible desires or inclinations...⁽¹¹²⁾

If we interpret this not as meaning that the person could not have the desire to kill, but could not possibly have the higher-order volition to act on that desire (i.e. to make that desire his 'will,' to identify with it), then Morris's concept of capability will closely parallel Harry Frankfurt's concept of volitional necessity.

In introducing it for the first time, Frankfurt describes volitional necessity as a "familiar but nonetheless somewhat obscure kind of necessity, in which of which [a person's] caring is not altogether under his control."⁽¹¹³⁾ The example Frankfurt gives is Luther's declaration, "Here I stand; I can do no other." In explaining this, Frankfurt notes that the impossibility "was a matter neither of logical nor of causal necessity...What he was unable to muster was not the power to forbear, but the will." A volitional necessity in this sense makes it apparent to Luther than anything but standing by his Theses is "unthinkable," in the same way as actually committing murder is "unthinkable" for Morris's Jones.⁽¹¹⁴⁾

In a later essay, "Rationality and the Unthinkable," Frankfurt uses this notion of unthinkability to explain what is wrong with the notion of personal identity usually thought to emerge from 'simplistic' Existentialism. In a passage that parallels Morris, Frankfurt notes that "Sometimes..a person may be incapable of performing a certain action even though, so far as considerations of these kinds go, he is in a perfectly good position to perform it."⁽¹¹⁵⁾ Frankfurt argues that it is only because the possibilities open to our will are limited by aspects of our character which make many actions "unthinkable" in this sense that we have personal identity and autonomy at all. If the personal world of volitional possibilities which are relevant or salient for our decisions were not thus limited, we would become a "bare person" for whom there are no limits to what is practically possible.⁽¹¹⁶⁾ In other words, if we have no volitional necessities, if we can alter our "specific volitional character" purely at will, there are no set cares and priorities to rule out alternatives and make it possible to decide what to will:

A person like that is so vacant of identifiable tendencies and constraints that he will be unable to deliberate or to make conscientious decisions. He may possibly remain capable of some hollow semblance of choice. If he does, however, it will only be by virtue of a vestigial susceptibility to inchoate volitional spasms...wholly devoid of authentic personal significance.⁽¹¹⁷⁾

It is incumbent on the Existentialist to show that her ontology of human freedom does not lead to such a result. But how is this possible, if we have decided that the Molinist way of conceptualizing volitional or personal necessity is void?

What the Existentialist requires is an new theory--a theory fully compatible with Existentialism--that will make sense of our pretheoretical modal intuition that there is a kind of synthetic modality which limits the arbitrariness of logically possible alternatives. The Existentialist has to show that it is possible to develop on his own terms a full theoretical conception of the pretheoretical modality which we have alternatively called "feasibility" or "capability."

(C) Molinism as an Atomistic Conception of Feasibility

As our analysis of synthetic modalities in general has shown us, the semantics for any such modalities requires that a particular 'sphere' of logically possible worlds be delimited as -relevant by the world-neutral actuality of some sort of sub-maximal states of affairs. If we want to claim that feasibility is a real (as opposed to merely conceptual, mind-dependent) synthetic modality, we must explain what grounds make one possible world feasible and another unfeasible, thus delimiting the sphere of l.p. worlds to which the feasibility as a synthetic form of 'possibility' corresponds. In other words, we need to explain what unifies the sphere of logically possible worlds relevant for the synthetic modality at issue.⁽¹¹⁸⁾

Although this need has been clearly perceived for some time by philosophers attempting to justify a nomological account of causal laws, the analogous task for metaphysical feasibility has not been clearly perceived, because it has not even been recognized that feasibility is a synthetic modality. Explaining the criterion or ground which determines the limits of the sphere for the synthetic modality is a more systematic way of conceiving the question, "what is it that makes counterfactuals of freedom true?" The problem with this traditional version of the question is that it already assumes that we are thinking of pretheoretical feasibility in a Molinist way: i.e. assuming that the sphere of 'the feasible' will be determined by the world-neutral truth of particular counterfactuals. But since we explicitly deny that Molinism is the only possible account of the grounds of feasibility as a synthetic modality, the neutral way to state the question is: what grounds truths of feasibility in general?, i.e. what ultimate metaphysical grounds give rise to the -relevance of a certain sphere of l.p. worlds, distinguishing them in a salient way that corresponds to the pretheoretical notion (evident in our modal intuitions) of capability, or volitional modality. When we ask this neutral systematic question about how a theory can account for our modal intuitions, we can see that Molinism is simply one particular answer to this more general problem. As a result, Molinism is only one particular theoretical conception of pretheoretical feasibility, among several alternative conceptions.

The first possible conception, which I will call the 'simplest' conception of feasibility (S), would be to suggest that there is a 'spiritual' world-sphere including all possible choices of possible free agents. By this designation, I mean 'spiritual' in the Husserlian rather than Pentecostal sense: Kant's "Kingdom of Ends" or society of all possible free and rational beings would be an analogous notion to this spiritual sphere. However, whereas we think it is logically possible that many different physical universes exist wholly incommensurable to our own, each of which would have its own 'sphere' of relevance for defining its natural laws (if the analysis of physical modality proposed above is accepted), there can for the Molinist (as for the Kantian) only be one 'spiritual world'⁽¹¹⁹⁾ (only one sphere of logically possible worlds relevant for 'feasibility'). The central problem for the Molinist is then to explain by what metaphysical criteria this sphere of logically possible worlds is unified. The simplest conception says that the relevant sphere includes exactly all those logically possible worlds (l.p. worlds) in which there is any action taken freely by a rational agent. Drawn this widely, however, the Molinist would certainly find it impossible to maintain that the spiritual world-sphere even could be bounded as he is inclined to argue it might be. For example, according to the simplest conception (S), the spiritual world-sphere would include all logically possible combinations of free acts, including even combinations in which all free agents involved freely choose what is right. On this model, it is clear that transworld depravity could not hold as a 'law' throughout the entire sphere of feasible worlds.

The Molinist answers the question not by picking out some unifying quality which all feasible worlds share, but rather by referring to a more primitive set of alleged facts: some free choices which are logically possible for agents are nevertheless not feasible in the relevant sense. It will then have to be by feasible combinations of feasible choices that we determine which logically possible combinations of free choices are also feasible. This approach amounts to saying that the more primitive set of facts about feasible choices--the true counterfactuals of individual freedom--determines which l.p. worlds are feasible. The feasibility of worlds supervenes on the primitive feasibility of particular actions of particular individuals in their circumstances in each world: I will call this Molinist conception (P) for "primitive atomism". On this conception, there is an infinity of independent, primitive facts about what all possible agents would do in every circumstance they could be in, in every possible world. The sphere of feasible l.p. worlds is built up atomistically by all such facts.

Once we understand the Molinist model in these terms, it is easy to see why Molinists would not think of feasibility as analogous to causal necessity. We think of the law of gravity as holding not because each physical object just happens, quite independently, to obey this law. Rather, we think of this object as co-existing with other objects in a context which is constituted by principles such as the law of gravity. But on the Molinist picture, patterns such as transworld depravity could never constitute the sphere of feasible worlds. They hold throughout this sphere, if they do, quite accidentally as a result of the counterfactuals of freedom that happen to be true for individual people. However, it is useful to describe this Molinist model as an atomistic conception of the grounds that unify the feasibility-sphere, because it allows us to gain a more general and objective perspective. It reveals the sense in which Molinism is a particular variant among alternative answers--which need not be atomistic--to the same general problem.

Seen in this light, however, the primitive atomistic conception of Molinism, or theory (P), does not look like a particularly attractive theoretical conception of feasibility, since it amounts to reducing feasibility to a primitive modal status of particular actions. This criticism of Molinism is quite independent of the problems with its implicit conception of agency (as described in §II), but is related instead to the 'intermediate' ontological status of the truth of counterfactuals of freedom. The truth of an unrealized counterfactual of freedom cannot be understood merely as a logical possibility, of course, because the whole point of Molinism is to say that not everything (or every combination) that is logically possible for free agents to choose is such that it is feasible for them. But if the counterfactual which expresses this synthetic modality is defined in terms of what is feasibly possible_F--rather than logically possible--then it would appear to be circular to say that the sphere of relevance is established by counterfactuals of freedom defined this way. The Molinist must therefore say that the feasibility of particular actions and choices of this given individual is are unanalyzable facts, although they must be referred to the individual's agency in some sense. The effect of this, as David Lewis says, is that feasibility is left as a metaphysical primitive, and the feasibility of l.p. worlds is determined by logical combination of agent-choices which are feasible in the primitive sense.

Metaphysical primitivism is one problem for (P), but there are quite distinct problems which arise from its atomism.⁽¹²⁰⁾ If we granted for the sake of argument that each agent's own agency is the primitive truth-maker for certain counterfactuals of their freedom, would there then be any reason to think that what agents would choose would fall into certain patterns, such as transworld depravity, rather than others? Is transworld depravity, or any other conceivable pattern of feasibility, just a wild chance that entails no logical contradiction, for which there could in principle be no explanation from the grounds of feasibility itself, if it turned out to be true? The atomism of (P) does seem to entail this unhappy result.

This makes (P) unsuited to at least some of the theological purposes for which Christian philosophers want it. In a recent paper, David Lewis has pointed out the weakness of theological "defense" in Plantinga's sense, which means "just any hypothesis" that serves to "rebut the contention that there is no possible way" that an omniscient, omnipotent, and omnibenevolent God could permit evil.⁽¹²¹⁾ I think Lewis is right that "defense" in this sense is too weak to provide much warrant for theism: the job, as Lewis says, is to "devise hypotheses that are at least somewhat plausible" and provide grounds for an answer to the problem of evil, even if they fall short of full theodicy.⁽¹²²⁾ As part of this general criticism, Lewis suggests that when Plantinga includes the bare possibility that Molinism is true as part of his defense, the plausibility of his argument for theism is further weakened, because Plantinga does not stop to prove that Molinism is plausible.

However, I think this last criticism is misplaced. In some of his writings, as we have seen, Plantinga does provide us with intuitive grounds for thinking that something like Molinism is plausible--although he does not analyze the grounds of these intuitions. Lewis's allegation about the weakness of Plantinga's "defense" against the problem of evil bites in a different way. Plantinga's defense openly depends upon the possibility of a particular pattern of feasibility holding--namely, transworld depravity--but even if in general there are Molinist true counterfactuals freedom, Plantinga's Molinist conception of feasibility (P) makes it impossible in principle that there could ever be any argument or explanation to warrant the claim that transworld depravity is not only possible, but plausible. Because of the atomistic ground Molinists postulate for them, patterns of feasibility can never be anything but wild chances, and therefore answers to the problem of evil employing them can never be anything more than 'mere defense' in the objectionable sense.⁽¹²³⁾ It is surely unattractive that the maximum a Molinist can ever contend is that opponents cannot absolutely prove that some wild chance pattern of feasibility is not a possible result of unanalyzable metaphysical primitives. If we think can do better, we ought to try.⁽¹²⁴⁾

There is also a more general sense in which the Molinist's reliance on a primitive feasibility makes his position unattractive. The problem is that the Molinist seems to arrive at his conception (P) simply because he has never entertained other possible explanations for how the sphere of feasible worlds is delimited: (P) is simply the only option he has for avoiding (S). If this is right, then conception (P), is just the minimal theory that naturally falls out when (S) is rejected but there is no account for why the sphere of feasible worlds is limited in some significant way or other. From this it follows that, if there are other alternative conceptions to (S) which provide more substantive accounts, including defensible grounds for holding that the sphere of feasible worlds is limited in some morally or religiously significant way, then any of these other conceptions would be rationally preferable to (P). If the postulation of unanalyzable facts of freedom given by (P) is warranted at all, it could only be because there is no other more substantive theory which better satisfies our basic modal intuitions. This explains the relevance of considering alternative conceptions which can satisfy the intuition that there are synthetic limitations on free choice, but construe their grounds differently than conception (P).

The possibility of alternative conceptions of feasibility raises some interesting issues. Certain patterns of feasibility, such as transworld depravity, depend for their very possibility on particular conceptions of feasibility, such as (P), and are unable to survive in other conceptions which might fit our basic modal intuitions more adequately. As a result, if any alternative, 'non-Molinist' approaches to feasibility as a synthetic modality work out, then it may be possible to deny the possibility of patterns such as transworld depravity, but without having to give up entirely on the notion of a type of synthetic modality which limits free choice in substantive ways. It is only the lack of alternative conceptions to (P) which has allowed traditional Molinists to maintain a monopoly on the idea of feasibility.

What alternative conceptions are available which might be used to explain feasibility as a synthetic modality, as well as particular patterns of feasible choices that might hold across the entire sphere of feasible l.p. worlds? To my knowledge, only four sorts of metaphysical principles have ever been considered as candidates for this role. They are, respectively, original sin, strong 'metaphysical' providence, Immanuel Kant's notion of an 'ultimate maxim' for each individual, and finally, a related 'family' of theories which would ground feasibility on volitional dispositions or 'habits' of the will.

IV. Existentialism vs Three Other Inadequate Theories of Feasibility

Let us begin by asking if either original sin or strong 'metaphysical' providence could provide grounds for believing that a pattern such as transworld depravity is likely in the counterfactuals of freedom. Either of these grounds of course would have the disadvantage of making belief in such a pattern of modal facts impossible without theism, but at least they might show that belief in the likelihood of certain patterns of feasibility is more than just barely possible within a Christian framework.

However, I will argue that upon closer examination, neither original sin nor providence, can provide warrant for transworld depravity even within a religious framework that could accommodate them. They both fail to achieve the task for which the theistic defender desires them: namely, to reconcile negative liberty with the cosmological thesis that God actualizes the best world open to him to create.

(A) Original Sin

An interpretation of original sin strong enough to provide a metaphysical ground for feasibility would hold that all free agents are born so constituted such that some good choices, or some combination of good choices, are such that they would not make them, while some evil choices are such that they would make them, even though both alternatives are logically possible for them in every case. Let us call this the (O) conception of feasibility.

But Christian theology has usually given a different interpretation of original sin: it refers to the belief that there are some kinds of good actions and habits that cannot be acquired without Divine grace. What Molina called "sufficient" or enabling grace⁽¹²⁵⁾ would, on this interpretation, be a necessary condition for salutary actions and all virtues (not only the theological ones). Molina and his followers distinguished such "sufficient" grace from "efficacious" or "congruent" grace, which would freely be accepted by the person in their circumstance.⁽¹²⁶⁾ Suppose we accepted that these concepts are at least provisionally distinct. Then, if the condition of original sin itself were the only factor determining which free acts of consent to grace are feasible, and granted that merely "sufficient" grace is enough to alter this condition of sin, it follows that sufficient grace can make its own acceptance feasible, i.e. it can make itself efficacious. The distinction between sufficient and efficacious grace thereby collapses altogether.

On this conception, then, the truth of counterfactuals of freedom is simply determined by God's free dispensation of sufficient grace, which thereby reduces feasibility to authoritarian divine choice, with no logos beyond the divine will. Clearly, if this were right, true counterfactuals of freedom could not in any way limit what omnipotence can actualize: facticities acquired by original sin could limit the feasible worlds only in a way intended by God, and therefore this limitation could not play the role required of it in Plantinga's or any other defense against the problem of evil. Original sin as an account of feasibility would imply that only God makes logically possible worlds unfeasible by withholding the "sufficient" grace which (given original sin) is both necessary and sufficient to make them feasible.

Another version of the (O) conception of feasibility is possible. One could hold that 'In Adam's fall sinned we all' means that certain counterfactuals of freedom for every person were made true not by their agency in re, but by Adam and Eve's actual choices in time. Interestingly, this version of (O) is closer to Existentialism than is traditional Molinism, because it grants that counterfactuals of freedom could only be made true by an agent if the agent is actualized. But it clearly fails to satisfy the criteria for moral responsibility, since on this account, my counterfactuals of freedom are literally imputable to Adam, the actual, existing person, not to me.

If we ignored this problem, however, would the traditional conception of original sin and grace require us to admit that Adam's choice was not by itself the sufficient condition for the truth of everyone's counterfactuals? Our non-acceptance of divine grace that could change these truths would seem to be an additional necessary condition for the truth of our retrograde c.o.fs. It might appear so at first glace, because in this variant of (O), there remains a distinction between sufficient and efficacious grace. But unfortunately, in this model it could only be Adam's actual choice which, independently of God's will, makes the acceptance of grace unfeasible for persons. If so, then it is Adam who brings it about that sufficient grace offered to me would not become efficacious. In this version of (O), then, Adam plays the role that God plays in the former version: his action determines for whom sufficient grace becomes efficacious. But surely this is far too much: Adam is now responsible not only for original sin, but for the reprobate status of souls that must end up damned.

Theory (O), at least in the two versions I have examined, is untenable. This result is interesting because Plantinga suggests that his conception of "transworld depravity" as a possible pattern of feasible worlds captures the meaning of the Christian doctrine of original sin. At least, Plantinga hints in this direction: for example, he says, "I leave as homework the problem of comparing transworld depravity with what Calvinists call 'Total Depravity.'"⁽¹²⁷⁾ But ironically, if we were to make a pattern of counterfactuals of freedom such as transworld depravity depend on a prevailing state of original sin, as understood in the orthodox sense, these counterfactuals could not support Plantinga's overall free will defense against the problem of evil. The reason is obvious: on an original sin model of feasibility, a completely sinless world would be feasible if only God would give sufficient grace to all agents therein. Unless we postulate counterfactuals of God's freedom to prevent it being feasible for him to give such universal grace, although it is logically possible for him to do so, the refutation of logical problem of evil will not go through.

One cautionary note: a rejection of theory (O) as a conception of feasibility in no way implies a rejection of the doctrine of original sin altogether. How one ultimately understands this theological concept is obviously bound up with how one makes sense of the synthetic modality of feasibility. What my critique of theory (O) shows is that, if for some reason one is committed to the doctrine of original sin, one would have to make sense of it in terms of a tenable theory of feasibility, rather than the other way around. Nor should it be presupposed that no conception of original sin is consistent with Existentialism. Only when we have arrived at a full theory of feasibility consistent with Existentialism, could we decide if any acceptable conception of original sin can be generated within the parameters of that model.⁽¹²⁸⁾

(B) Strong Providence

A strong metaphysical version of providence might account for the truth of counterfactuals of freedom in a different way. It would hold that different individuals are simply so constituted that they cannot in fact freely choose certain combinations of action (such as living a completely sinless life), because God's sovereign control of creation makes it so. Let us call this theory (SP), for strong providence. On this conception, providence (as a stronger concept than omniscience) has simply determined that persons have (and continue to have) the natures they do, which also determines the patterns of their counterfactuals of freedom for different possible circumstances. This account would explain why such a set of possible individuals, once they are so constituted, could be in a pattern such as transworld depravity.

The motives for this type of approach are well summarized by Hugh McCann in a new paper on "Divine Sovereignty and the Freedom of the Will."⁽¹²⁹⁾ First, it suggests a theistic solution to the problem of showing that there can be a sufficient reason for my making one choice rather than another in a given situation: "If, as creator, God is responsible for my choices, then even though they are not subject to deterministic law they still have a full explanation."⁽¹³⁰⁾ As McCann notes, this problem of showing that there can in principle be a sufficient reason for my voluntary acts is made particularly acute by all liberatarian theories of free will.⁽¹³¹⁾ The other independent motive for such a theory is our understanding of what it means for God as creator to preserve in existence the actual world and all its inhabitants as it unfolds. McCann suggests that

It is hard to see...how God can be responsible for my existence at every moment without being responsible also for my characteristics. As Malebranche observed, God cannot create a physical object that is in no particular place, that is neither at rest nor in motion, etc. All of these things need to be settled in the production of the object. And in the same way, it would seem, God cannot create me with an indeterminate will--that is, a will that is neither deciding to A nor not deciding to A.⁽¹³²⁾

This is a view that dates back to Thomas Aquinas's conviction that God is directly involved through his creative agent-causation in all world-internal processes of causation among secondary causes (including 'efficient' or 'agent'-causes). McCann is simply extending this analysis of what takes place in Divine creative/preservative acts to our voluntary decisions: "short of God's action as creator, there is no me to do any deciding, and the same act of His will that is responsible for my being puts in place my decisions as well."⁽¹³³⁾

The Existentialist must thank McCann for clarifying the analysis of creative activity that underlies the strong providence theory, however, because in the process its defects become more apparent. Aside from the fact that this analysis seems to presuppose some (highly suspect) kind of kalam conception of 'time-atomism,' its worst problem is that it assumes that creating the existence of an entity necessarily involves actualizing it with a kind-essence complete enough to account for all of its accidents. This is indeed the reason that St. Thomas posits his notoriously problematic conception of a completely individualized substance. The most radical objection would be to argue that existence in the sense deriving from the Judeo-Christian notion of creation ex nihilo is an 'absolute' which is entirely distinct from ousai in the Greek sense, i.e. the entity's being 'what it is.' This is Duns Scotus's objection: the esse is not the Sosein, q.e.d. The weaker version of this objection would be to argue that God can constitute an entity as a person with a voluntary will, without 'strongly actualizing' their particular choices: at this point, there is a real metaphysical joint at which to cut. 'Once' the person exists, they can do the rest. McCann is wholly unwarranted in asserting that God cannot create me with an indeterminate will (note that indeterministic does not have to mean entirely arbitrary, either⁽¹³⁴⁾). It is true that my will cannot be preserved in existence at the 'moment' (if there is such a precise point) when the decision is made, without its being in a state of having made a determinate choice, but it does not follow that God directly actualizes that choice itself. The Existentialist insists that in metaphysical (but not temporal) order of priority, God can create the person with a will that must terminate in some choice (possibly, of a given type), but the person thus constituted then creates the particular choice itself. McCann cannot rule this out, although we are then attributing to humanity a power which is literally a limited form of God's ability to create ex nihilo.

The Existentialist thus has strong objections to the most plausible rationale for the strong providence theory. But there are even stronger objections to be raised against the implications of this (SP) model. This theory was held by Dominicans such as Dominic Báez who explicitly opposed Molina. According to Copleston, Báez held that the Thomistic creation theory was necessary to secure principle that God is the source of all goodness in existence. Thus, "God is the cause of all salutary acts:"

According to Báez...God knows the future free acts of men, even conditional future free acts, in virtue of His predetermining decrees, by which he decides to give the 'physical premotion' which is necessary for any human act.⁽¹³⁵⁾

On this conception, God's efficient causality directly brings about actual choices in time, in a way known prior to the actual existence of the persons.

The implications of this (SP) conception may perhaps be better understood if we consider another version of it recently defended by Kathryn Tanner. Professor Tanner presents a theory developed over a number of years to defend the Thomistic thesis that creatures depend on God's agency not only for their existence, but also for the every modality of their actions;⁽¹³⁶⁾--as we have seen, this is the principle McCann needs to justify. As a recent reviewer explained Tanner's theory,

God's immediate creative agency brings about not only my ongoing existence as an agent with the capacity to make choices, it also brings about my choices. Although God does not act alongside or among secondary agencies to cause me to act as I do, God's primary activity constitutes me as the agent who performs these acts. God's creative will includes each of my choices, and God's will infallibly effects what it intends.⁽¹³⁷⁾

This version of (SP) has one advantage over Molinism, in that the truths of an agent's c.o.fs are made before the agent is actualized, but they are so made by an actual God, not by unactualized persons. But in other respects, Tanner and Báez's (SP)-conception of feasibility is at least as bad off as Molinism. Tanner holds a compatibilist view of human freedom like that described by William Alston⁽¹³⁸⁾: negative liberty obtains for Tanner only between an agent's power for deliberative choice and other secondary causes: "an agent's free actions are not determined by the prior history of the world and the laws of nature."⁽¹³⁹⁾ This is what Báez also means by his claim that God moves "free agents, when they act as free agents, to act freely."⁽¹⁴⁰⁾ But if moral responsibility demanded negative liberty on the 'horizontal' level of such secondary causes, it would require some type of negative liberty in the 'vertical' relation between creature and Creator as well.⁽¹⁴¹⁾ If it turns out that I am just so constituted that, although I have many alternative logical possibilities to hating my brother, I cannot in fact avoid doing so (I will always hate him, given the chance), then how can I be justly blamed for it? This problem is especially acute if the individual's c.o.fs cannot in any way be attributed to the individual themselves, since God could have made them otherwise but just chose not to.

(SP) thus turns out to be a version of the Essentialist horn of the 'antinomy of authenticity.' Recall that this position rejects condition (1), the requirement of negative liberty. But it is particularly clear that this (SP) version of Essentialism also fails to satisfy condition (2), the authentic identification condition for moral responsibility. Since it is God and not I who makes my essence, constituting me as the person who would do X in circumstance C, clearly the truth of this c.o.f. is not imputable to me. This makes not only the good but also the evil I do attributable to God, and not to me. For his part, Báez seems to accept the principle of attribution from which I draw this conclusion: for otherwise, we could not attribute the good of my virtuous decision to God. Báez presumably imagined that in evil actions, I and not God am the source--but whether such an asymmetry could ever be justified by any means, Thomistic or other, remains very doubtful at best.

McCann disagrees. He argues that my decision (and thus presumably its goodness as well as its evil) are solely attributable to me, and not to God:

..even though God creates me [as] the person who decides to burn down the church, He does not make my decision. I do. My decision, and any evil that lies with it, are predicated of me. What belongs to God is his decision that I shall decide to do what I do, which is an altogether different matter.⁽¹⁴²⁾

This is a surprising conclusion to draw--perhaps shocking would be a better description. McCann seems to think that just because I can make a mere conceptual distinction between my decision and God's decision that I shall so decide, it is obvious that the former is only attributable to me (and the latter is certainly only attributable to God). What is shocking is that this move completely fails to take into acocunt recognized conditions for authentic attribution of states of affairs to agencies. In particular, McCann ignores the requirement that I must be volitionally identified with my act--condition (2) in §II-E.

Once the relevance of this requirement is accepted, it is easy to see that the same argument we ran against 'sophisticated Molinism' applies to McCann's version of (SP). McCann admits that I am not identified with God's decision to sustain me in such a way that I shall decide to X. We need only one more premise, which no self-respecting Thomist could dispute: I shall decide to X follows with broadly logical necessity from God decides that I shall decide to X. Since I am not identified with the latter, I also do not identify with the former: it is no part of my authentic self, but alien to me, even if, from a third-person point of view, I perform the act X describes. Hence my decision to X is not morally imputable to me.

Thus on (SP), God is responsible for what appears to be my moral evil. McCann seems to think this is the only serious objection based on the problem of evil the defender of (SP) has to answer. But not so. On the strong providence account he and Tanner share, it is clear that God, and not my agency, has determined the truth of my counterfactuals of freedom, which limit or even uniquely determine the choices which are feasible for me in each possible circumstance. But if so, there are no synthetic constraints independent of God which limit compossible arrangements in a world: In such a case, David Lewis would be right to assert that God's omnipotence implies that "He never has to allow evil so that good may come of it."⁽¹⁴³⁾

Yet McCann and other defenders of (SP)--Leibniz, for example--do assume that something more than logical consistency will limit compossible co-existence in the possible worlds. For example, McCann says that God's decision to have me make a decision X may involve making me do something evil in itself, but such "harmful consequences...need hardly be the point of His plan. They may only be obliquely intended, as necessary to some superior good God wishes to accomplish."⁽¹⁴⁴⁾ The Existentialist's response must be to ask what "necessity" means in this context. If the (SP) defender replies "broadly logical necessity," then he will not have limitations sufficient to explain the evil in the world: there are surely logically possible worlds in which some evils (say, a person's venial thought right before their death, foreover unknown to any other person) are ommitted and the same goods remain. But the (SP) defender is not entitled to bring in any sort of synthetic necessity to help explain the limits on God's options, since he has stripped these of all their modal force (relative to God) by reducing them to regularities supervening on a kind of divine 'impressionist painting,' whose 'dots' are each thing's being recreated at each atom of time with just the individual nature and accidents it has at that moment.

Thus for (SP), feasibility can render no assistance in answering the problem of evil. This is the same difficulty as we encountered in the (O) conception of feasibility: without any limit on what is feasible independent of God's agency, the limits on feasibility can obviously be of no value at all in explaining the existence of evil in God's world. In both (O) and (SP), synthetic modalities of feasibility are conceived as grounded in a way that undermines virtually any plausible theistic response to the problem of evil.

(C) The Kantian Approach

In his Religion within the Limits of Reason Alone, Immanuel Kant works out a theory of ultimate moral character which is another theoretical conception to make sense of the prereflective intuition of 'feasibility.' According to Kant, we must postulate the existence, in each person, of a single deepest "disposition" of their will, which determines the character which underlies all their actions as either "good" or "evil." This disposition is thought of as a maxim which is itself freely adopted by the will, but conditions the adoption of all lower-order maxims:

..the good or evil in man (as the ultimate subjective ground of the adoption of this or that maxim with reference to the moral law) is termed innate only in this sense, that it is posited as the ground antecedent to every use of freedom in experience (in earliest youth as far back as birth) and is thus conceived of as present in man at birth...⁽¹⁴⁵⁾

The ultimate "disposition" in this sense is a concept analogous to "virtue," which Kant defines as "the firmly grounded disposition strictly to fulfil our duty.."⁽¹⁴⁶⁾ Thus, although an individual's ultimate disposition is spontaneously incorporate as the ultimate maxim of his will, it has continuity and stability over time, like "cares" in Frankfurt's sense. Kant, however, makes much stronger claims for his ultimate disposition. Not only must a person's ultimate maxim form the single underlying ground of all their incorporations of lower-order maxims at a given time; it must ground all their exercises of freedom in experience over time, and hence in a sense defines their single ultimate character over a 'complete life:'

To have a good or an evil disposition as an inborn natural constitution does not here mean that it has not been acquired by the man who harbors it, that he is not the author of it, but rather that it has not been acquired in time...The disposition, i.e. the ultimate subjective ground of the adoption of maxims, can be only one and applies universally to the whole use of freedom.⁽¹⁴⁷⁾

Kant's arguments for this position are fairly simple, and are directly related to the conception of willkür he presented in the Third Antinomy of his Critique of Pure Reason: "reason creates for itself the idea of a spontaneity which can begin to act of itself, without requiring to be determined to action by an antecedent cause in accordance with the law of causality."⁽¹⁴⁸⁾ Because the adoption or "incorporation" of all maxims must trace to spontaneity of this kind,⁽¹⁴⁹⁾ an individual's "subjective ground" for the adoption of particular good or evil maxims must be a higher-order maxim: "If this ground itself were not ultimately a maxim, but a mere natural impulse, it would be possible to trace the use of our freedom wholly to determination by natural causes."⁽¹⁵⁰⁾ For Kant, it follows from the nature of moral responsibility that, in the order of grounding, any maxim must trace either to a higher-order maxim, or finally to a pure spontaneous act of adoption for which no sufficient reason can be given. But free will also cannot consist in an infinite regression of maxims behind maxims: spontaneity is the idea of an absolute beginning for a series. Hence the use of freedom 'in experience' (acting on particular maxims) must trace a last higher-order grounding maxim which itself is spontaneously adopted.

This regression-argument does not by itself assure us of the unity of a person's highest-order maxim(s) which are themselves grounded only in spontaneous adoption. This further step involves Kant's transcendental idealism, which requires him to conceive the adoption of a first maxim in the grounding-order as a purely noumenal action. Kant holds that when the ground-consequent relation or category of causation is schematized, or applied to a temporally ordered manifold, we find that in principle, we can always extend the regression of causes to prior causes, without end: this is the reason why the Third Antinomy arises.⁽¹⁵¹⁾ As a result, the highest-order or first maxim(s) whose spontaneous adoption brings the grounding series to an end must be incorporated atemporally for Kant. Although the ultimate "disposition" must itself be freely adopted to be imputable,

..the subjective ground of cause of this adoption cannot further be known...since otherwise still another maxim would have to be adduced in which this disposition must have been incorporated, a maxim which itself in turn must have its ground. Since, therefore, we are unable to derive this disposition, or rather its ultimate ground, from any original act of the will in time, we call it a property of the will which belongs to it by nature (although actually this disposition is grounded in freedom).⁽¹⁵²⁾

Although Kant is not entirely perspicuous about it, his argument is clearly from the impossibility of an infinite regression to the conclusion that a first maxim must be acquired atemporally or noumenally. This explains why he could urge earlier that a person can have only one highest maxim which must underlie the whole of their lower-order use of freedom: noumenal spontaneity cannot change from one ultimate disposition to another, or it would be possible to order it in a time-like series of states, i.e. to schematize it.⁽¹⁵³⁾

With this theory of an ultimate maxim, Kant is moving towards a notion of character capable of determining the feasibility of particular actions and choices. As Philip Quinn says in an article on Kant's theory in the Religion, the disposition to moral evil is

..present in the agent independent of and therefore antecedent to all exercises of freedom in time and [is] inextirpable in the sense that it cannot be eradicated by means of exercises of freedom in time."⁽¹⁵⁴⁾

Like the truth of one's counterfactuals of freedom in Molinism, one's ultimate disposition cannot be altered by actions in time, but rather, it constrains the range of particular choices the person could make in a given circumstance. This constraint is not absolute, because an evil ultimate disposition, for example, doesn't determine any particular action-maxims, but only makes it more probable that one will adopt certain action-maxims rather than others in a given circumstance.⁽¹⁵⁵⁾ This ultimate maxim exercises this qualified constrainst on particular maxims because the pattern in particular action-maxims which reflect the underlying ultimate maxim are the result of that ultimate disposition: as John Silber says in his commentary, "The specific acts of Willkür...do not establish the motive of action but are largely the products of the motivational force of the dispositional act."⁽¹⁵⁶⁾ Therefore the pattern of feasibility which his ultimate disposition imposes for a person is not an 'atomistic' result of the primitive feasibility of particular actions, as was the case with Molinism. Rather, the pattern comes about because of an underlying reality which in large part brings about that pattern.

At the same time, unlike Molinist counterfactuals of freedom, Kant's ultimate disposition does not completely specify the pattern of first-order action-maxims which one would adopt in every circumstance one could ever be in. As Quinn puts it, the dispositional maxim "does not causally necessitate particular actions at particular times," but rather acts as a propensity for particular uses of freedom.⁽¹⁵⁷⁾ In this respect, John Silber rightly emphasizes that is because of this flexibility that Kant can maintain his rigorist thesis that a person's choices are always subordinated either to an evil or good ultimate maxim, even though the patterns of particular choices thus bivalently subordinated range through a continuum of virtuous quality. For example, the will can decline in quality as even while "the Willkür remains good in its dispositional aspect.."⁽¹⁵⁸⁾ In each of the stages of "radical evil," however, the character of the will still constrains what is feasible for the person at the practical level. For example, it is not feasible for a will whose character is nearly that of full "autonomous volition," which suffers only from occasional weakness, to commit an act of heinous evil and feel no guilt for it.⁽¹⁵⁹⁾

Kant's ultimate disposition thus functions in several ways like Frankfurt's volitional necessity, which (as we saw) is a disposition to higher-order volitions (or "care") which is so strong that it makes certain options for one's will volitionally impossible or "unthinkable," even though one might be able to consider them abstractly speaking. As Frankfurt says, maintenance of such a necessity is not voluntary in every decision which falls under it, because it is not alterable "merely at will."⁽¹⁶⁰⁾ The noumenal, atemporal status of the ultimate maxim secures the same result for Kant. And yet, Frankfurt also insists that at another level, volitional necessity is "to a certain extent self-imposed." The reason is that caring₁ about something only becomes a "volitional necessity" for a person when he cares₂ about his own caring₁:

[Volitional necessity] is generated when someone requires himself to avoid being guided in what he does by any forces other than those by which he most deeply wants to be guided. In order to prevent himself from caring about anything as much as he wants to care about them, he suppresses or disassociates himself from whatever motives or desires he regards as inconsistent with the stability and effectiveness of his commitment.⁽¹⁶¹⁾

In other words, the stability which constitutes his care₁ is something he actively shields from alternative options whose consideration would break down this stability. The aversion he has to letting his care be changed is thus an aversion that "has his endorsement."⁽¹⁶²⁾ This is the reason why an individual does not find volitional necessity alien to his own will, like animal impulses which enslave unwilling addicts, but rather finds that the forces which maintain the necessity "are in the most authentic sense his own forces."⁽¹⁶³⁾ Kant's "innate" ultimate disposition is similar, precisely because he insists that its innateness (or incorrigibility) "does not..mean that it has not been acquired by the man who harbors it, that he is not the author of it.."⁽¹⁶⁴⁾ The "disposition" itself, the restriction it brings about on the feasibility of particular action-maxims, and the pattern of particular free acts which issues from it are thus freely chosen, although not apparently alterable in time.

At this point, we can see that Kant's theory of feasibility as grounded in a person's ultimate disposition, which I will call conception (K), occupies an unusual intermediate position between Molinism and Existentialism. It would not be unfair to say that in Religion within the Limits of Reason Alone, Kant is trying to be a kind of Molinist, or at least a 'Noumenalist' analog of a Molinist. Kant avoids the fundamental antinomy in Molinism by insisting that the free agency that adopts the ultimate disposition, which in turn rigorously determines which more particular maxims are feasible for the agent, is itself an actually existing agency. Yet, like the Molinist and unlike his Existentialist followers, Kant believes that this agency and its single 'ultimate act' must lie outside time, or lack temporal schematization, and therefore apparently be unalterable. This intermediate position held by (K) is possible because of the unusual intermediate status of the noumenal realm or aspect itself: it is actual and yet supposedly atemporal. The theory of feasibility which results from this could well be called 'Existential Molinism,' if that is not too paradoxical.

It might seem, then, that for a philosopher sympathetic with Molinism, theory (K) would be a good position to retreat to. By retreating to a version of (K), she could avoid the 'antinomy of authenticity' objection (section §II): she can maintain that a person's counterfactuals of freedom are determined spontaneously--thus satisfying condition (1)--by their actual agency, thus satisfying condition (2), and that the determination occurs only noumenally outside of time, thus satisfying (3). Kant's position is the only resolution which preserves all three of these conditions, and avoids the antinomy of authenticity. To accept a noumenal basis for the truth of counterfactuals of freedom, the Molinist would have to accept one amendment of fundamental importance. For Kant, the ultimate disposition never absolutely determines an individual's particular action in a given circumstance: it is only revealed phenomenally in trends over a whole life. So the 'Kantian Molinist' would have to accept that counterfactuals of freedom are probabilistically qualified.⁽¹⁶⁵⁾ But if they accepted this, could a 'Kantian Molinist' still hold that there is just one set of noumenal truths about what a person (existing noumenally) would probably do in any given phenomenal situation?

Fortunately for the Existentialist, however, Kant himself reveals why his own intermediate theory (K) is untenable. Kant goes on in the Religion to affirm the Christian view that it is possible to change one's ultimate disposition from evil to good. Although it cannot be done through a gradual reform of one's action-maxims to conform to the law (and thus not through a temporal progression in the phenomenal world), a "change of heart" in which he adopts motivation by pure practical reason as his highest maxim is possible through "a revolution in the man's disposition."⁽¹⁶⁶⁾ Although Kant apparently thought this was consistent with the notion that a new ultimate disposition cannot be "acquired in time."⁽¹⁶⁷⁾

Professor Quinn has argued convincingly that allowing such a reversal in the ultimate maxim commits Kant to an irretrievably inconsistent position. Given the "thesis of rigorism," which entails that a person's cannot be both good and evil at once, "the thesis of radical evil" is inconsistent with the "thesis of moral revolution" for the following reason:

..the supposition that the human had at one time adopted a morally evil supreme maxim and at some later time given it up contradicts the part of the thesis of radical evil which asserts that amorally evil supreme maxim, because it is the product of an act independent of all temporal conditions, is inextirpable once adopted.⁽¹⁶⁸⁾

Quinn is certainly right: Kant would have to modify his conception of radical evil to hold his thesis of the possibility of revolution in ultimate moral character. Even more radically, the possibility of such revolution seems to imply that there is some time-like succession in the noumenal realm itself, some history for the Willkür. If this is accepted, the transcendental will has lost its unschematized character which made intermediate theory (K) possible. Instead, we have a conception that finally vindicates the Existentialist's claim that the feasibility of choices can only be determined within some actual temporal order. Kant's theory of moral revolution is thus a decisive step in our progression from Molinism to Existentialism.

(D) Kant & the Eschatological Requirement of Openness for Ultimate Moral Character

Because Quinn is mainly interested in the prospect that Kant's theory might provide a coherent conception of original sin which preserves both its blameworthiness and inextirpability,⁽¹⁶⁹⁾ he finds it unfortunate that Kant went on to include his account of moral revolution.⁽¹⁷⁰⁾ But from my perspective, it seems clear that it is Kant's own determination to construct a theory that accurately captures our intuitions about feasibility which leads him to take this last, crucial step towards an Existentialist position. Kant argues that the transformation of our ultimate maxim remains feasible for us because "the original predisposition to good in us" still remains in us, no matter how evil we have become: "It must be presupposed throughout that a seed of goodness still remains in its entire purity, incapable of being extirpated and corrupted," even though we have refused to adopt this remaining incentive to respect the law into our ultimate maxim. Yet Kant admits that if this remaining "seed" or vestige of the good predisposition were utterly lost, "we could never get it again," and the transformation of one's will would then be impossible.⁽¹⁷¹⁾ Thus for Kant, the feasibility of this moral revolution or transformation of one's ultimate maxim corresponds to the belief that a human will cannot become "reprobate" in this life: its ultimate character cannot be so utterly fixed that salvation would already be unfeasible for it. Notice that denying the feasibility of a moral revolution in ultimate character would entail that our ultimate character is fixed, in one single atemporal act--theory (K) implies this result as much as the traditional Molinist theory (P). I believe Kant finds this result unacceptable, in the last analysis, because it is inconsistent with the very meaning of eschatological religion. The nature of eschatology requires that a person's final moral character is cannot be fixed before their death.

Since this point is crucial for understanding the Existentialist's approach to a theory of feasibility, it is worth explaining why Kant thought eschatological concerns would require the revolution thesis. Scholars usually account for the moral revolution thesis without any reference to its eschatological significance. Typically, they find sufficient explanation in Kant's stated conviction that the possibility of revolution in the ultimate maxim follows from the ought-implies-can principle: because the moral law includes "the injunction that we ought to become better men" Kant says "this must be within our power."⁽¹⁷²⁾ But the reasons why Kant requires the moral revolution thesis extend beyond this stated explanation. To understand Kant's deepest reasons, it is more important to discern what lies behind his ought-implies-can argument.

At first, it seems that Kant's main motive for the moral revolution thesis is that it is necessary for proper moral education. Kant is extremely worried about moral habituation that only tries to teach the virtue of "empirical character," or external conformance to the law, which suggests to students that "a change of heart is not necessary, but only a change of practices."⁽¹⁷³⁾ Real moral virtue, by contrast, can be taught only through the Pietist theme that a fundamental turn, an umkehr, must underlie the "continual progress from bad to better" in empirical character: the latter must be founded on the former.⁽¹⁷⁴⁾

..it follows that man's moral growth of necessity begins not in the improvement of his practices but rather in the transforming of his cast of mind and in the grounding of a character; though customarily man goes about the matter otherwise and fights against vices one by one, leaving undisturbed their common root.⁽¹⁷⁵⁾

Without the possibility of a revolution in character, the reformative potential in moral education could only be understood in a way that would conflict with the unity of virtue.

This apparent pedagogical motivation for the revolution thesis also agrees with Kant's belief that the thesis of radical evil itself is only of pedagogical value: it leaves "moral dogmatics" (the strict derivation of duties) untouched, but in "moral discipline" it warns us to "begin with the assumption of a wickedness of the will," and thus to conceive moral pedagogy not as merely reformative but revolutionary. For Kant, it is quite clear that the thesis of radical evil cannot have its proper pedagogical (rather than metaphysical) significance if moral revolution is impossible. Moreover, the pedagogical significance of the possibility of revolution in ultimate character would be in danger of reduction to mere admonition for reforming outward practices, if radical evil were not the diagnosis of our condition. Far from being opposed, in their pedagogical implications, the thesis of radical evil and the revolution thesis require one another.⁽¹⁷⁶⁾

But even this does not get to the real root of the matter. Why do we need to assume that any potential for moral progress (and educative processes to encourage it) are necessary at all? Kant's answer is that without this assumption, the promise of salvation can only be understood as divine favor-seeking, rather than moral religion, the religion in accord with pure practical reason. In religions "which are endeavors to win favor,"

...man flatters himself by believing either that God can make him eternally happy (through remission of sins) without his having to become a better man, or else, if this seems impossible, that God can certainly make him a better man without his having to do anything more than to ask for it.⁽¹⁷⁷⁾

The thesis of moral revolution, then, is ultimately required to ensure the right interpretation of, and the right practical relation to⁽¹⁷⁸⁾, the eschatological promise that constitutes religion. The importance of this theme cannot be overstated, for it becomes the central principle of Kierkegaard's philosophy following Kant.⁽¹⁷⁹⁾ Without the principle that it is feasible for us to bring about the change in our own ultimate maxim, or at least to take the initiative without which salvation is impossible, good works are not necessary at all. The thesis of moral revolution, then, embodies Kant's surprising Pietist opposition to Luther's doctrine of the sola fide. Our ultimate character cannot be fixed from all time, because if it is, then good works and moral duty have no religious significance at all.⁽¹⁸⁰⁾

For Existentialists, the requirement that "good work" on one's ultimate moral character remain possible has a deeper metaphysical meaning, which can be expressed by an extension to the eschatological category of their general view that counterfactuals of freedom could only be made true by existing persons in actual time. The Existentialist not only holds that which logically possible world turns out to be 'the' actual l.p. world depends on what happens in actual time, as it becomes actual; she also insists that which eschatologically possible hereafter becomes actual is not yet determined, but likewise depends at least partially on choices agents make in time. But any conception that gives each person just one ultimate character, fixed prior to any developments in the unfolding time of their life up to their death, limits the range of feasible distributions of persons into final states of damnation or salvation, to a single atemporally feasible distribution, a single atemporally feasible hereafter. In Kierkegaard's terms, such a conception has misappropriated the finality of the 'eternity to come,' and given it to an atemporal universal 'aeternitas,' the realm of unchanging metaphysical truth. It fatally confuses an eschatological finality with a Platonic metaphysical ground.

This eschatological requirement of existential openness for one's ultimate moral character before death, which I will call the eschatological requirement for short, forms a strong objection to Molinism quite distinct from the 'antinomy of authenticity' proved in §II. Moreover, while Kant's theory of a single ultimate maxim (theory K) escapes the antinomy of authenticity, it does not escape the objection that it fails to satisfy the eschatological requirement. Of course, since Kant affirms the possibility of moral revolution, his full theory in the Religion satisfies the eschatological requirement, but in doing so he contradicts his own theory (K), which does not satisfy this requirement.

As an objection to both Molinism (theory P) and Kant's narrower theory in the Religion (theory K), failure to satisfy the eschatological requirement is not as strong an objection as the 'antinomy of authenticity,' since it cannot by itself show that theory (K) or (P) is metaphysically inconsistent. But their failure to meet this requirement does show that the conception of ultimate moral character which these theories entail is contrary to the most essential requirement of what Kant calls "moral religion," i.e. a religion which insists on the significance of good works and the religious relevance of reformative progress in ultimate character. It seems quite plain that any theory which implies that counterfactuals of one's eschatological status are complete and closed not only prior in time to one's death, but from all time, completely eliminates the existential significance of the eschatological promise, the meaning which is essential to eschatological divinity as an idea throughout world-history.

I realize this is a strong charge, but it seems inevitable. I have no absolute demonstration for the 'eschatological requirement.' But more can be said than simply that it is a fundamental part of the Existentialist perspective. I maintain that any theory of human volition and character which rejects this requirement makes it impossible in principle that eschatology could have any relevance for life and moral progress. But this is surely too extreme a violation of our modal intuitions, even if theism is false. It is one thing to say that thousands of years of human history were wrong, because there is no God; it is another to say that what they believed, namely the divine eschatological promise, necessarily could not ever have turned out to be meaningful, even if had turned out to be true. Rejecting the Existentialist's 'eschatological requirement,' as far as I can tell, is tantamount to claiming that the eschatological promise is not false, but meaningless. That extreme claim, however, is as intuitively false as a claim could be. If we were not so constituted that eschatological possibility could be meaningful to us, it could never have become the center of religious world-views for several millennia.

Anthropology is on the Existentialist's side. So is Christian theology. I urge that anyone starting from a Christian framework should consider themselves already committed to the 'eschatological requirement' in their commitment to the orthodox doctrine that salvation is never impossible before death. For a vivid expression of this belief, we have C.S. Lewis's powerful portrayal of a wicked man's dying moment:

Escape for the soul, if not for the body, was offered him. He became able to know (and simultaneously refused the knowledge) that he had been wrong from the beginning, that souls and personal responsibility existed. He half-saw. He wholly hated...With one supreme effort he flung himself back into the illusion. In that attitude, eternity overtook him as sunrise in old tales overtakes trolls and turns them into unchangeable stone.⁽¹⁸¹⁾

Lewis is attempting to capture orthodox doctrine in this imagined scene. But if we read him as meaning that salvation is always just logically possible until the moment of death, but perhaps unfeasible, then there would be no point for the final offering Lewis imagines God making to the soul near damnation. In addition, the turning point of eschatological finality, which, as Lewis sees, is symbolized in endings of fairy tales, would occur with the acquisition of the mortal vice, not with death. To give the theological doctrine its proper meaning, we must interpret it to read, salvation is never impossible_F before death.

V. The Habit-Based Approach to Feasibility

(A) Intuitive Features of the Characterological Approach

We have now ruled out several alternative theoretical accounts of feasibility as a synthetic modality, including (S), (P), (O), (SP), and (K). The last model, however, has introduced the idea that volitional dispositions could be the basis for the feasible or unfeasible status of certain actions. Because the Kantian approach considered an ultimate disposition which did not supervene on particular 'atomistic' and absolutely discrete 'facts' about what a person would do in each possible choice situation, it already takes us towards a habit-based approach to analyzing feasibility. The hallmark of this approach is that by distinguishing deep volitional character from particular actions, it aims to ground truths of feasibility in the former without necessarily completely determining every particular action a person would do.

There are three good reasons for exploring the possibilities of such a 'habit-based' or 'characterological' approach, as I will call it. First and foremost, it accords better than Molinism with the real characterological ground of the pretheoretical intuitions that lead to the notion of feasibility as a synthetic modality of freedom. Second, we already have in Frankfurt's model of cares as habits of higher-order willing a satisfactory non-noumenal account which allows us to distinguish the relevant morally significant 'level' of character which feasibility will concern on a (non-Kantian) habit-based approach. Finally, although traditional virtue ethics has long regarded habits of free choice as synthetic limitations on libertarian freedom of logical possibilities, the habit-based or 'characterological' approach does not necessarily presuppose a theistic framework or any theistic premises, such as divine middle knowledge, providence, or original sin.

Developing a habit-based conception of feasibility proves difficult, however, precisely because of features that are distinctive of habitual dispositions of the will. Although in the relation of habit to actions, the former does not simply supervene on the sum of the latter (taken atomistically), since habits refer to the characteristic behavior and motivation of an individual when he or she is in certain circumstances, such habits would still seem to provide a criterion of what is feasible only for the individual. In this sense, it seems at first that a habit-based theory of feasibility would be atomistic in another sense than Molinism: what is feasible for persons in general--and therefore which patterns of feasibility obtain--would supervene on what is feasible more primordially for each particular individual with their character (although not on what particular action is more primitively feasible for them in each circumstance). For Molinism, there is an atomistic 'truths' of feasibility for each intersection of two variables: the identity of the person and the circumstance which determines the action-decision to be made. It seems at least initially that a habit-based approach would also sanction an atomistic truth of feasibility for each intersection of two variables: the individual's morally relevant character⁽¹⁸²⁾ and their choice circumstance.

If volitional character or habit of will is the primordial level for truths of feasibility, however, there is a problem about moving from this level to more particular truths about the feasibility of different actions a person might take in various circumstances. A disposition or habit would not, it seems, be sufficient grounds for asserting that if an individual was placed in the appropriate situation, he would not deviate from his habitual response. It is not a fact that he would not deviate, especially if the "appropriate situation" were described generally enough to apply to an indefinite range of occasions, rather than to maximally determinate circumstances. Rather, the habit, insofar as it relates to particular actions, determines only an outward trend of that individual, a statistical phenomenon in her courses of action. Being in the habit of accepting bribes, for example, does not determine that a certain mayor will accept a given possible bribe on the possible occasion of being offered it. If there is a discrete fact about his decision in that hypothetical situation, as some philosophers have thought, it seems that it could not be determined by the habit alone.

(B) Robert Adams's Criticisms and the Necessity of Probabilistic Qualification

Robert Adams is to my knowledge the only philosopher who has previously given serious attention to the possibility of a habit-based or characterological ground for truths of feasibility. Adams, however, rejects this approach as inadequate precisely because of the difficultly just outlined--namely, the problem of relating volitional dispositions to truths for counterfactuals about determinate actions. Adams develops his argument with reference to the biblical event where David is warned by the Lord that if he stays in the city of Keilah, Saul will attack it and the men of the city will surrender him--"a famous proof text for the Jesuit theologians" who founded Molinism, which yields two cases of middle knowledge:

(1) If David stayed in Keilah, Saul would besiege the city.

(2) If David stayed in Keilah and Saul besieged the city, the men of Keilah would

surrender David to Saul.⁽¹⁸³⁾

After noting that these sentences cannot be made true by logically or causally necessary connections between their antecedents and consequents, as long as we assume that they refer to free actions, Adams admits that "we might seek non-necessitating grounds for the truth of (1) and (2) in the actual intentions, desires, and character of Saul and the Keilahites."⁽¹⁸⁴⁾ But he argues that such a basis for (1) and (2) is "inadequate precisely because it is non-necessitating:" it could only justify a probabilistic modification of (1) and (2):

(1') If David stayed in Keilah, Saul would probably besiege the city.

(2') If David stayed in Keilah and Saul besieged the city, the men of Keilah

would probably surrender David to Saul.⁽¹⁸⁵⁾

Adams also argues against any attempt to say that (1) and (2) are true because the possible world in which David stays in Keilah and Saul attacks is more similar to the actual world in the relevant respects than the possible world in which David stays in Keilah and Saul does not attack: any such judgement of the similarities would violate our "well-entrenched belief that under many counterfactuals conditionals many a person might have acted out of character, although he probably would not have."⁽¹⁸⁶⁾

I believe Adams is right that a characterological approach can never justify anything more than (1'), for example, when it comes to particular actions in fully specified circumstances. But Adams is wrong to draw the conclusion that a habit-based theory of feasibility in general--i.e. a theory that would make sense of our modal intuitions about such cases--is therefore impossible. This error comes from failing to distinguish feasibility in general from the Molinist account of it. Adams's argument is that if a ground G for the truth of a given statement about feasibility SF is posited, then G must necessitate SF. If SF is a counterfactual of freedom in the Molinist's sense--i.e. an alleged discrete fact about the single action a person would take in a given circumstance--then indeed, G would be inadequate to necessitate SF. But once we see the possibility of alternative non-Molinist theories of feasibility, there is no reason to retain the prejudice that the metaphysically basic level of truths of feasibility must be counterfactuals of freedom in the Molinist's sense. Suppose we accept it as a metatheoretical condition that for any theory of feasibility, the grounds which establish the truths of statements of feasibility (SFs) must necessitate whatever sorts of SFs they are associated with. Then the correct conclusion to draw is that starting with habit-based account of the grounds of feasibility entails that at least for particular actions, probabilistic conditions such as (1') and (2') just are the only appropriate type of SFs, the only ones which can have determinate truth-values. I will refer to this result as the IP Axiom: the irreducibly probabilistic nature of primordial truths of feasibility.

I claim that this axiom follows from the adoption of characterological grounds in general. To prove that the probabilistic qualification required by this axioms holds for truths of feasibility about particular actions on any habit-based conception, I will give two arguments to show that no theory can hold both the following at once:

(a) All truths of feasibility are grounded on characterological states;

(b) These truths grounded on habits of higher-order willing necessitate particular actions in given situations.

The converse of (b) is obviously the IP axiom.

The first argument shows that holding (b) (or denying the IP axiom) entails rejecting (a). Let us consider a theory (R) which tries to hold both (a) and (b) at once by assimilating the characterological approach to a more limited version of traditional Molinism. It might seem possible to preserve the Molinist intuition expressed in (b) along with (a) if we claim that there will be an absolutely discrete fact about what Saul would do if David remains in Keilah, if the circumstance is sufficiently specified and includes Saul's morally relevant character and habits. Where C includes David's staying in Keilah as well as Saul's avarice and paranoid habit of trying to eliminate potential enemies by the swiftest available method,

(3) If King Saul were in C, Saul would attack the city.

is true. However, if a less maximally specified situation C- includes David's remaining in Keilah but excludes the relevant elements of Saul's character, then there is no truth-value for either of the following counterfactuals:

(4) If King Saul were in C-, Saul would attack the city.

(5) If King Saul were in C-, Saul would not attack the city.

This revised version of traditional Molinism (R), in other words, would agree that truths of feasibility have a characterological base (a), but still maintain that truths of feasibility grounded by habits of will are non-probabilistic (b). Features of moral character are relevant on this model because their inclusion in the circumstances is the decisive factor in determining whether or not there is a discrete truth-value for a traditional non-probabilistic counterfactual of freedom. In (R), the difficulty of probabilistically qualified truths of feasibility seems to be avoided, and the habit-based account is assimilated to the traditional Molinist approach with the concession that without including the individual's habits of choice, there is no Molinist 'fact' about what they would do.

At first glance, this might look like a promising response to Adams's objections and a way around the IP axiom. But theory (R) faces several difficulties. The Existentialist will insist that theory (R) must somehow be able to account for our intuition that habits of character do not necessitate that one act in accordance with them. We cannot possibly believe that any person in the same position as Saul is in C (including his bad habits of character), will always attack Keilah, or we would be saying that it is unfeasible in general to act contrary to the habits or dispositions Saul possesses. This is plainly untrue, or we would not even call Saul's vices habits or dispositions. Thus we have

(c) Habits of the will do not in general (for every person possessing them, at all times) necessitate actions in accordance with the habits.

The only way theory (R) could explain this intuition (c) is by allowing that there are contrary but still quite determinate counterfactuals of freedom for different persons with the same character-traits as Saul has in similar situations. For example, imagine that in C, Darth Vader has all of Saul's vices, and otherwise occupies his place in history (being King of the Hebrews at the appropriate date, etc). Then let us further imagine that

(6) If Darth Vader were in C, and David stayed in Keilah, Darth Vader would

not attack the city.⁽¹⁸⁷⁾

is true. Thus, according to (R), it only appears that in general it is feasible to act against the habits of character included in C, because it just happens that most people (such as Saul) absolutely would not, and a few people (such as Darth Vader) absolutely would--and we have no way of telling which kind true counterfactual of freedom is made true by the person's habit in any particular case.

Theory (R) is a clever way of assimilating the intuitions which support the IP axiom into a broadly Molinist framework. But the only way (c) can be explained to remain consistent with (b) involves implicitly denying (a). In trying to explain away the appearance that habits are non-necessitating, (R) turns out not to be a habit-based account after all. This is already apparent from the fact that Darth Vader's bad habits themselves can hardly be the reason why he would not act in accordance with them, making (6) true. Rather, theory (R) requires that the truths of counterfactuals of freedom such as (3) and (6) simply happen to fall into the trans-individual patterns that make us call certain features of their circumstances habits or dispositions. It turns out that on (R), the very existence of habits of volition in the intuitive sense supervenes on the primitive truth of counterfactuals such as (3) and (6)--it is not the habit which makes them true but rather the reverse. In denying (a), (R) becomes simply another person-circumstance-choice atomism which in this case offers a minimalist explanation for the appearance that there are such things as virtues, vices, and other aspects of character, and the appearance that they are relevant for determining what their owners would do in different circumstances. (R) is then heir to all the failings of Molinism--the antinomy of authenticity, the lack of any possible explanation for its patterns of feasibility, and its violation of the eschatological requirement--which we hoped to avoid by turning to a non-noumenalist habit-based conception.

This analysis of (R) suggests that there is no way consistently to hold (a) and (b) at once, at least not while satisfying the burden of explaining the intuition expressed in (c). Furthermore, there is a second argument to show that any theory which tries to hold that truths of feasibility for actions are non-probabilistic truths (a), while basing these truths on habits of choice (b), necessarily contradicts certain well-entrenched moral intuitions. If his vicious habits make it completely unfeasible for Saul to act against his vice, then he is moral responsible only for freely acquiring the vice, not also for refusing to make the sacrifices necessary to overcome it. If the habit ensures that Saul would never choose against it, then never overcoming it cannot be separated from simply possessing the habit, at least for Saul, and thus it cannot be the additional sin we always think it is.⁽¹⁸⁸⁾

Taken together, the above considerations show that there is no tenable way of denying the irreducibly probabilistic relation between habits of will and particular actions, while still holding that volitional dispositions are the grounds for truths of feasibility for particular actions. Premise (a) entails (IP): therefore, we are justified in accepting IP as an axiom for any conception committed to basing truths of feasibility on volitional dispositions and features of character.

(C) The Incompleteness Problem for Atomistic Habit-Based Truths of Feasibility

We have seen in the preceding section that the basic premise of the habit-based approach to feasibility entails the IP axiom. In this section, I will argue that if (a) is accepted, there is another set of difficulties which, taken together, show that the habit-based approach to feasibility must be broadened in a fundamental way. As we noted earlier, the habit-based approach naturally lends itself to a different sort of atomistic conception of feasibility than Molinism involves: it suggests that the truths of feasibility which are metaphysically basic are probabilistically qualified truths for each relevant combination of volitional dispositions and habits of the individual, and the options for action they have in their circumstances. The problem is that any characterological theory must have some explanation for the acquisition of morally relevant habits and dispositions (virtues and vices). If we accept principle (a) as it stands, then there are no truths of feasibility for an individual without a fully determinate moral character. In that case, what about the following type of case:

(d) If A does X in C, then A will acquire a habit of higher-order will H.

(e) But then C cannot include A's having H.

(f) Assume that H is the only habit of higher-order volition which could be relevant to determining the feasibility of doing X in C.

(g) Then by (a) and (e), there cannot even be a probabilistic truth about the feasibility of X's doing A in C.

Given that we have no non-question-begging grounds for insisting that (f) is false for every X and H, then we would be forced to say that it is possible that feasibility-relevant volitional dispositions and habits are themselves often acquired through actions and motivations in which, eo ipso, the individual has no such habits, and therefore lakcs even probabilistic truths of feasibility with respect to these free habit-acquiring actions and their motives. I will call this the Incompleteness Problem of the habit-based approach.

The Incompleteness Problem for the acquisition of dispositions is unavoidable for the habit-based approach with the IP principle. Scholastic philosophers such as Aquinas who forged our modern notion of morally significant habits (including virtues and vices of character) thought of habits as arising in the person because of chosen actions--actions with whose deliberate intention the person identified, for the sake of an apparent good leading to happiness, certainly, but nevertheless with the consent of negative liberty.⁽¹⁸⁹⁾ But if this is right, then it was always feasible for the person not to acquire the habits they did acquire through their own free actions.⁽¹⁹⁰⁾ In Aquinas's system, for any habit of the will, it is feasible for me not to acquire it (but to acquire some other volitional disposition instead). Clearly, then, no truths of feasibility which are grounded by habits of the will can have the kind of 'middle' ontological status granted to counterfactuals of freedom in Molinism. For example, it is feasible for me not to acquire the habit of taking bribes. Thus, if it was because of the existence of the habit alone that refusing a bribe would not be feasible (or would be very improbable) for me, were I offered it, then indeed refusing the bribe would be feasible for me, because it is feasible for me to bring about that this habit does not exist. We could say that "world-segment" which forms the context of the choice is specified to include the earlier choices of mine which led to the habit of bribe-taking. But then there is a real sense in which this specification is giving us a false picture: the circumstance includes facts about my character which it was feasible for it not to include. In an analysis of volitional freedom, we must focus on feasible circumstances. And in this model, there are feasible circumstances in which there is no truth of feasibility, even probabilisitcally qualified, for my choice as to whether to accept the bribe. In other words, on this atomistic habit-based account of feasibility, it will always be feasible for the agent to be in some choice-circumstances such that their choice would be entirely indifferent or arbitrary. This is another way of phrasing the Incompleteness Problem.

The Existentialist also wants to maintain that it cannot be true, metaphysically prior to all ytime, that refusing the bribe in a given situation is unfeasible for me, simpliciter; it could at most be true that refusing the bribe is unfeasible (or difficult) for me given that I have made certain earlier choices in which both alternatives were feasible for me. But ideally, the Existentialist would like to maintain this and still avoid the Incompleteness Problem. The problem of 'Simplistic Existentialism' would be most completely overcome by showing that there is a consistent Existentialist theory from which it follows that feasibility is complete--that there is no feasible choice situation in which the action would be entirely indifferent.

As far as I can see, for the habit-based approach there is only one way to avoid the possibility that there are no synthetic limitations on the feasibility of actions and motivations through which volitional dispositions are acquired. This is to allow that habits and dispositions of the higher-order will not only have a probabilistic 'top-down' effect on the feasibility of actions, but also that specific types of action are synthetically connected (in a 'bottom-up' fashion) with the acquisition of certain kinds of habits and dispositions. In such a conception, the metaphysical grounds for truths of feasibility will not be the mere possession of habits of will themselves, but rather the reciprocal relations that govern the feasible interactions over time of elements in a person's deep character and her particular actions in different sorts of circumstance. I will refer to this crucial feature of a broadened characterological theory of feasibility as level-interactivity (LI), since it involves interactive relations between the different 'levels' of personality (from particular actions to ultimate maxims) which we distinguished intuitively in §III-A.

The Incompleteness Problem for the initial acquisition of habits is also closely related to the problem of moral revolution, or bringing about changes in the deep and morally relevant aspects of one's volitional character. What if the feasibility of actions through which I change my volitional habit H itself depends on H? We run into a problem similar to the previous one, but more complex. If we do not broaden the base for truths of feasibility beyond that indicated in (a), then somewhere in the system we will get either or both of the following: (i) some aspects of our character which are supposed to ground limitations on the feasibility of particular options for action in different circumstances, which we are nevertheless completely free to change at will; (ii): some aspects of character that are completely unchangeable once acquired.

Option (i) allows for an incompleteness of feasibility for changing some habits of will, i.e. those at the 'top' of the system for determining truths of feasibility based on habits. For these habits H_t, the feasibility of the actions which would allow an individual to change H_t habits depends on no other habits of will other than H_t. If there is such a set of habits, then some actions will be arbitrary. Suppose, for example, that Darth Vader has two H_t habits of will: an overwhelming care for the power to dominate (H_t1), and an overwhelming cowardly volition not to will any actions that would put him in conflict with the Emperor (H_t2). Let us suppose that the only way for Darth Vader to change H_t1 is to save Luke by killing the Emperor: losing his desire for power without such an action is unfeasible. Yet H_t2 makes saving Luke by killing the Emperor unfeasible in the sense of highly improbable for Darth. But it is feasible to overcome H_t2 by changing H_t1. Is Darth Vader in a catch-22?

According to option (i), he is in the reverse of a catch-22: defying the Emperor and saving Luke is arbitrarily open to him, for the following reason. If we specify Darth's choice-circumstance C to include both his having H_t2 and H_t1, then it seems that the former makes saving Luke/killing the Emperor highly improbable. But in fact, in C, it is feasible for Darth to change C to C', in which he lacks H_t2. Why?--because it is feasible first for him to change C to C*, in which he lacks H_t1. How could he change C to C* and then to C'? By exercing his remaining negative liberty--which by the IP axiom H_t2 never ruled out--to choose to save Luke and kill the Emperor. Since making this choice eo ipso involves overcoming H_t1 and hence overcoming H_t2, the negative liberty Darth has to make this improbable choice against his vices, is itself already the liberty to remove the habits which made this habit-changing action unfeasible. The very limited negative liberty Darth has for this action thus expands to remove its own limitations. Hence we can only say that it is open to Darth at will to kill the Emperor and save Luke, since it is open to him at will to remove any limitations to what is open to him at will. Yet something seems wrong about this analysis of the catch-22: it seems to lead to the false conclusion of 'Simplistic Existentialism.'⁽¹⁹¹⁾

Option (ii) is the opposite way out of the catch-22. We can understand this option as a rough approximation of the conception of the dynamics of human volitional character which Aristotle offers us in his Nicomachean Ethics.

..if without being ignorant a man does the things which will make him unjust, he will be unjust voluntarily. Yet is does not follow that if he wishes he will cease to be unjust and will be just. For neither does the man who is ill become well on those terms. We may suppose a case in which he is ill voluntarily, through living incontinently and disobeying his doctors. In that case it was then open to him not to be ill, but not now, when he has thrown away his chance, just as when you have let go a stone it is too late to recover it; but yet it was in your power to throw it, since the moving principle was in you. So, too, to the unjust and to the self-indulgent man it was open at the beginning not to become men of this kind, and so they are unjust and self-indulgent voluntarily; but now that they have become so it is not possible for them not to be so [1114a 13-23]:⁽¹⁹²⁾

In this statement, Aristotle clearly does not mean that it is broadly logically impossible for the unjust man to bring about that he becomes just: Aristotle means that in our synthetic sense, it is unfeasible for him to bring this fundamental reform about, at least by his own initiative. Clearly it is a conception of the synthetic modality of freedom which is at stake in the intriguing picture Aristotle provides with his metaphor of moral character as a 'thrown stone.' For convenience, let us refer to this Aristotelian conception as theory (A).

As a conception of feasibility, (A) has many features we have seen to be advantageous. It maintains that truths of feasibility are grounded on habits of volition that are acquired by actual choices of the agent in time, rather than noumenally. It also admits that habits do not necessitate actions, and thus truths of feasibility are only probabilistic with respect to particular actions. It is also a better option than (i)--i.e. 'Simplistic Existentialism'--because it at least preserves the meaning of some truths of feasibility, namely those grounded on unchangeable habits or dispositions of will. But it still conceives feasibility as atomistically grounded on the habits and dispositions of particular individuals and their circumstances, and consequently it can only solve the problem of changing these truths-of-feasibility-grounding aspects of moral character by ruling out such changes. For Aristotle, it was feasible for the man to avoid acquiring his bad habits, but once they are acquired, it is not just too difficult in practice, but even unfeasible in principle for him to overcome them.

Aristotle's is a sort of no-way-back model of free choice and character. Theory (A) thus avoids the Incompleteness Problem for changing habits of will, but at quite a cost. Medieval theologians such as Aquinas rejected the second half of this Aristotelian model (A), thus making possible a new conception of virtue in general, as well as new higher-order virtues specifically concerned with self-reform, such as 'Purity,' 'Patience' and 'Humility.' I believe there are two reasons why they rejected Aristotle's solution to the question of how the feasibility of bringing about changes in ones own habits of will is governed. First, it is not true to experience. In Return of the Jedi, Darth does kill the Emperor, save Luke, and change his moral character in the process: 'revolutions' such as this are possible in everyday life as well. But secondly, and perhaps more importantly for medieval theologians, the extreme implications of theory (A) violate the 'eschatological requirement.' Being in a state where it is absolutely unfeasible to overcome a deadly vice or lose a combination of virtues is already to have realized an eschatological state of reprobate vice or perfect virtue. The realization of such a state is not possible in this life: a system of feasibility implying otherwise would be guilty of misappropriating eschatological finality to states of will in time, in characterizing vices and virtues of particular mortal individuals.

Here the Existentialist is in harmony with medieval reformers of Aristotelian virtue theory: it is part of the existential meaning of habits of disposition and moral character that, for any possible mortal person, until their death it will always be not just logically possible but also feasible for them to act contrary to and even to overcome habits for whose acquisition and continuation they are morally responsible. This blocks not only system (R)'s attempt to remove the irreducibly probabilistic aspect of habits while using them to account for the feasibility of choices, but also the less unsatisfactory theory (A).⁽¹⁹³⁾

In sum, the Incompleteness Problem and Eschatological Requirement, taken together, show that the basic premise of the habit-based approach entails the thesis of level-interactivity (LI): the acquisition and alteration of dispositional states of higher-order will is never feasibly arbitrary or feasibly impossible, but is governed, like logically possible particular actions, by probabilistically qualified truths of feasibility arising not from the bare existence of habits of the will, but rather from inter-level connections between such habitual states and the first-order acts of will which make particular actions `deliberate.' These 'connections' govern the feasibility for acquiring and changing volitional dispositions, as well as the feasibility of different types of actions conditional upon one's character states once acquired and until overcome. By focusing on these connections, we provide a basis for the system of truths of feasibility which is wide enough to let that system include all actions of persons, including those that involve self-reform and the acquisition of new characterological features.

VI. The New Existentialist Characterological Theory of Feasibility

In this last section, I will attempt to outline a new theory of feasibility that takes into account all our previous findings, and indicate how it might be grounded. The result will be, I hope, the most adequate conception of feasibility to date. Moreover, once its implications are understood, we will see that this new model is really is a form of Existentialism, preserving both authenticity, moral responsibility, and the eschatological requirement of openness for ultimate moral character.

The new model's two most distinctive features are its non-atomistic conception of the relation between truths of feasibility and individual character-states, and its acceptance of the irreducibly probabilistic nature of truths of feasibility relating to particular actions and dispositional states of individuals. Our analysis has brought out the important systematic relation between this non-atomistic aspect of the theory, which allows it to be level-interactive (LI) in the sense defined above, and its acceptance of the IP axiom. First, each of these features is entailed by premise (a) of the habit-based approach, as we saw in §V-b and §V-c respectively. Furthermore, it is possible to argue that LI directly entails IP. Once we accept that the metaphysical ground for the synthetic modality of freedom must be broadened in order to cover both the feasible limitations on the acquisition of habits of will through actions and the feasibility of altering one's volitional dispositions, it follows that the IP axiom must be accepted.

Consider a woman who has acquired the habit of taking bribes. Since by LI it is feasible for her to overcome her habit, if she is in circumstance C faced with a bribe, it is feasible for her to bring about that she is the same characteristic situation C', a "maximal world segment" identical to C except that in C' she lacks the habit of taking bribes. If she would refuse the bribe in this situation C', then, since it was feasible for her to be in C', it was feasible for her to refuse the bribe. But if she would accept the bribe in C', in which she lacks the habit, then obviously the habit cannot be the sole reason why. Either way, possessing the habit by itself cannot be sufficient for making one alternative absolutely unfeasible for the individual. Because the interactive relation between particular actions and the acquisition of habits of will does not cease once dispositions are acquired (contra Aristotle), it ensures that particular actions can never be made more than probable by dispositions of the higher-order will. Hence IP follows from LI.⁽¹⁹⁴⁾

(A) The Concept of Volitional Difficulty

The new Existentialist theory of feasibility accommodates the IP axiom by refusing to speak of discrete 'counterfactuals of freedom' in the traditional Molinist sense. Counterfactuals of freedom, understood as pure facts about the maximally specific result of an individual's being faced with a particular choice, cannot be entailed by habits, which are by definition non-deterministic.

The solution is not to give up on a habit-based approach, but to give up the prejudice which says that the modal intuition that there are synthetic limitations on free moral choice can only be cashed out in a theory that is deterministic with respect to particular choices by particular agents. On the contrary, Existentialism holds that the connections between choices persons can make and truths of feasibility for them are not deterministic. A viable alternative is to maintain that what follows from certain kinds of choices made by agent A is not an absolute fact about the feasibility or unfeasibility of this particular choice for A, but rather a difference in the openness of options that in themselves are logically possible alternatives for A.

This notion of a "difference in openness" amounts to a synthetic modal interpretation of the medieval concept of difficulty. Whether it is expressed as a propensity to avail oneself of one option more than others, as a probability, or in other ways, this difference is something which cannot be captured simply by the logical possibility of the alternatives (say, of refusing or accepting the bribe). For logical possibility is always symmetric: neither alternative can be "more" logically possible than the other. But by definition, a difference in the openness of logically possible alternatives is not symmetrical. It is thus synthetic, i.e. something over and above the 'broadly logical' modalities of their circumstance. And this is what makes such differences in openness plausible candidates for interpreting our intuitions about feasibility. Thus in Robert Adams's case, for example,

(7) If David stayed in Keilah, it would be difficult for Saul not to attack the

city.

which, by the definition of 'difficulty,' means:

(8) If David stayed in Keilah, it would be more volitionally open to Saul to

attack the city than not to attack it.

The qualifier 'volitionally' is inserted here to indicate that Saul's 'difficulty' does not mean any sort of trouble arising from causal factors tending to bring him into conflict with Keilah, or from problems in rational deliberation which prevent him from considering other options, or from purely instinctive impulses to attack something that threatens him. Rather, the difficulty comes from a disposition of the higher-order volition, which in Saul's case is the vice of arrogant care for power.

It is a point of fundamental philosophical importance to see that Existentialism not only compatible, but even consonant with the existence of volitional difficulties of the higher-order will in the modal sense here envisioned. Although Existentialism defends negative liberty of logically possible alternatives, and insists that further synthetic limitations on free will must be acquired in actual time, neither of these claims entails that the openness of different options to a free agent's negative liberty must be entirely unaffected by the past history of their choices. Accordingly, it is consistent both to hold that there is negative liberty in every choice, and yet to maintain that feasibility is complete: in no choice is there a synthetic arbitrariness or complete balance in the openness of options.

(B) The Universalistic Conception of Feasibility

The only way to avoid positing an atomistic conception of the metaphysical relation between truths of feasibility and individual character-states, which, as we have seen, makes level-interactivity impossible, is to accept that the metaphysically primary level of truths of feasibility must be the same for all possible persons. But how is this possible? This idea is bound to seem unfamiliar, since it requires precisely the opposite approach to that traditionally taken by Molinism. It is possible, however, by accepting the deepest implication of our intuitive experience of level-interactivity in the synthetic limitations on freedom. This is the implication that the ground of primary truths of feasibility is not simply the temporal acquisition of higher-order volitional dispositions, but rather a set of laws which determine the feasible interactive relations between actions and habits of will. The Existentialist hypothesis is that there are certain connections between earlier actions, the acquisition of habits of will, and the subsequent differences in the openness of options, which are not broadly logically necessary, but are nevertheless the same for all possible persons. And it is these general metaphysical connections which are the ultimate ground of particular truths of feasibility for specific persons in given circumstances, which fall under the general connections as instances.

This proposal is connected in an interesting way with the idea of explaining synthetic necessities in terms of the relevant sphere of logically possible worlds, which we introduced in §III-C. The Existentialist theory implies that the range of l.p. worlds which includes all possible persons is the modal range which defines the sense of necessities of the relevant synthetic kind, i.e. feasibly necessary truths. For example, it could be necessaryF⁽¹⁹⁵⁾ for a given individual that if he chooses actions X, Y, and Z, then later, in certain characteristic circumstances where he must choose between doing A or B, it will be more difficult for him to elect A than B. This connection between a certain kind of action and the acquisition of certain habitual propensities of choice will not just be true for him, but feasibly necessary_F because it holds for all possible persons. I will speak of such a synthetic necessary connection as a law of feasibility, holding for connections of higher-order volition to actions for all possible moral agents.

But do we have any basis besides the difficulty with other models of feasibility for believing that there actually are laws of feasibility in this sense? If observable connections between choices and habitual 'differences of openness' for logical alternatives is merely local, holding only for specific individuals or only in explicitly designated circumstances, it can hardly have the generality looked for in a conception of synthetic necessity. This was Kant's fear when in the Groundwork of the Metaphysic of Morals he argued against trying to base a conception of moral rightness on "the activities and conditions of human willing as such, which are drawn for the most part from psychology."⁽¹⁹⁶⁾ The traditional Molinist is content to accept a model in which the feasibility of choices is relativized not just to beings of a particular psychological make-up, but to individuals per se. In system (P), what is feasible for one agent may well be unfeasible for another in the very same fully specified circumstances (including the same habits of higher-order will): in such a case, only the difference in their identity makes a difference to what is feasible for them. This is because in (P), as we saw, differences in feasibility supervene on differences in primitive facts particular to each agent. But level-interactivity not only compels us to abandons this extreme particularism of the default conception (P); it also identifies the synthetic necessary connections of feasibility with connections between actions and volitional habits/propensities of choice that are the same for all possible persons. Level-interactivity implies a universalistic conception of truths of feasibility.

These universal connections, if they exist, by definition can have no empirically adequate basis. Whether there is an a priori basis for postulating the existence of action-habit connections that are feasibly necessary_F in this sense, can only be decided by asking what would be required to ground laws of feasibility that apply to the range of all possible morally significant agents. The connections envisioned by such laws would have to be true only in virtue of features essential to all such beings, rather than because of accidents of this or that physiology, neurology, or individual hacceity. Only the connections between certain kinds of actions, volitional habits and propensities in which they result, and volitional difficulties which derive from these acquired characterological traits, all specified without quidditative reference to the individual identity of the person, or to groups of persons distinguishes by 'purely local' qualities or 'natural kinds,' could be asserted as necessary connections in the relevant synthetic sense ('necessary_F'). These connections would have to be pure of merely 'contingent_F' basis relative to the sphere of possible morally significant agents, and based solely in the nature which all entities in that sphere of relevance share. Understood this way, the laws of feasibility would, pace Kant, pick out universal features of "willing as such" that are highly relevant for moral theory.

What this analysis shows is that if there are any connections between global kinds of choice and habitual propensities to prefer certain kinds of alternative options to others, which hold not merely because of human psychology but for all free and morally significant beings, then there would be synthetic laws of feasibility determining feasible histories of choices.

There is a reason why such a theory has never been formulated before. Standing in its way is a broad presupposition, shared by authors ranging from Aquinas to Kant, that all connections between choices and habit in our experience are either determined by specific features of human nature, corrigible facts of human psychology, or even less general grounds such as cultural differences.⁽¹⁹⁷⁾ This perspective will seem natural if we assume, with Kant, that the only thing all possible morally significant agents have in common is their bare negative liberty itself.⁽¹⁹⁸⁾ If we operate with this presupposition, then we will assume that the sphere of all possible morally significant agents will be equivalent to the sphere of all possible beings with negative liberty--the sphere which defined the simplest conception of feasibility (S) in §III-C. The Kantian view assumes that the sphere of l.p. worlds in which there exist beings to whom moral laws could apply is equivalent with the sphere of feasible worlds as construed according to the simplest conception (S).

Suppose, however, that we abandon this unwarranted presupposition in favor of an account of the morally significant agent in general--one to whom moral laws could relevantly apply--which is inspired by a phenomenological analysis of personhood.⁽¹⁹⁹⁾ According to this account, in addition to bare negative liberty, moral agents would necessarily have such fundamental features as the ability to be aware of, and volitionally related to, three kinds of beings: other persons, their natural worlds, and divinity. If we restrict our attention to the first of these 'fundamental orientations,' as I will call them, the capcity for a fundamental volitional orientation to others depends not only on the co-existence of persons in a temporal order of some sort, but also on their ability to use their negative liberty to identify themselves, in acts of higher-order willing, with certain enduring propensities of choice referring to relations with others. In other words, on this account personhood would require a capacity for free identification of the self with habitual dispositions, enduring through a time-like ordered series, which establish fundamental orientations towards other persons (as well as towards nature, and towards divinity). These fundamental orientations would consist in synthetic restrictions on the range of options which the individual agent identifies as feasible for his responses to other beings. Beings capable of free and even rational decision outside of any time-like order, or without any capacity for identification and selfhood, would be logically possible, but could not constitute persons with agency or choice in the substantial sense necessary to fall under the scope of moral law.

At this juncture, further details of this conception of personhood are beside the main point. If personhood or moral agency can convincingly be conceived along such lines, then the sphere of morally significant free agents can be viewed as already limited in crucial ways: it will leave out some of the logically possible free and rational beings included within the sphere of feasibility according to conceptions (P) and (S). On this new phenomenological conception, all logically possible moral agents would be capable of identifying themselves with habits of higher-order will concerning their relations with other persons (and nature and divinity). If this is the proper sphere of relevance for conceiving feasibility, then the sphere of feasible action-habit connections (or spiritual laws) is already limited to those possible for free beings constituted as moral agents or persons are constituted.

I will call this the moral conception (M) of the spiritual world-sphere. In this conception, the sphere of feasible l.p. worlds is determined by the inclusion of free agents who also possess whatever set of features is necessary for moral agency and selfhood. The central idea in theory (M), then, is that it is not individual agency, but the proto-social nature of moral freedom and personhood itself which make necessary the laws of feasibility that ground all particular truths of feasibility. According to (M), the synthetic sphere of feasibility is determined at bottom by the sphere of logically possible full moral agents. The spiritual laws of action-habit connection which establish truths of feasibility are just those which are logically essential to full moral agents as conceived in the phenomenological conception of personhood.

On this moral conception of feasible worlds, if features such as those sketched in the phenomenological account above are common to all moral agents, then there is less reason to make the Kantian assumption that any laws governing connections between actions, dispositions, and volitional difficulties could at most be general to human psychology. Although many such connections are culturally contingent, or based on facts of specifically human nature, are we certain that they all are? It is surely not certain that there are no connections of this sort left over that are universal for all possible free beings capable of other-regarding habits of choice? Given the phenomenological account of personhood sketched above, one cannot rule out a priori that there are some connections between certain global kinds of choices and global kinds of other-regarding habits of will and volitional difficulties⁽²⁰⁰⁾ which attend moral personhood in all its possible forms. The phenomenological account of personhood gives an a priori plausibility to the hypothesis that there are 'spiritual laws,' laws of feasibility true for all possible persons.

VII. Conclusion: Existentialism and Spiritual Laws

This moral conception of truths of feasibility as grounded in spiritual laws is as fully consistent with Existentialism as the existence of volitional difficulty. Existentialism only requires that in their application to particular existing persons, laws of feasibility cannot have any initial conditions which are not adopted voluntarily by the moral agent herself. The individual must exist, not just in potentia, but in actuality, in order to determine for herself, through her own negative liberty, the initial conditions of her character which influence her habitual propensities of higher-order will and the probabilistic truths of feasibility which depend on them. Nor can there be any law that directly makes it feasibly necessary for any individual to do or refrain from any act.

Understood this way, Existentialism hardly looks as irrational or incomprehensible as many critics of continental philosophy presume it to be. The central claim of Existentialism is equivalent to saying that for all individual persons, their logical essence can be actualized without thereby determining any particular truths of feasibility or volitional difficulties they may in time acquire. Despite this, however, the negative liberty of moral persons, who are so constituted that they cannot avoid standing in some volitional relation to other persons in their world, has what we might call a spiritual temporality to its exercise, a temporal order in which through prior choices the agent will necessarily acquire volitional dispositions and habits which synthetically limit her future options. The laws of feasibility, on this account, are the laws that shape feasible courses through this temporality of the will or 'spirit.' It is completely compatible with libertarian freedom for there to exist laws which govern the possible histories of choices of free agents, because negative liberty does not entail that real freedom of choice is as absolutely arbitrary with respect to the history of one's spirit (or free will).

To understand the notion of spiritual laws of feasibility, it is crucial to realize that these laws refer to synthetic connections which cannot be altered by any actions or choices of the moral agents to whom they apply. Our choices determine the specific results which develop in accordance with these laws, in the same way as the laws of physics tell me what will happen if I throw a penny off the Eiffel Tower. But my bringing about a specific result which the laws of feasibility do not by themselves necessitate_F no more changes these laws than deciding to throw the penny changes the law of gravity which determines the result. If spiritual laws exist, they are a rigorous system which determines, willy-nilly, the types of volitional difficulties that will inevitably follow from different sorts of volitional dispositions, and which of these habits of will inevitably follow from particular acts of willing.

This implies that there can be something like a system of moral teleology for the Existentialist. To accord with negative liberty, this system must allow for more than one single possible highest telos of volition for all possible persons, since it is feasible for the moral character of any person to go in at least two directions by their own fully informed decisions. But the direction of increasing probabilities for further decisions and increasing difficulty in turning back which follow from alternative choices may be quite inexorable, "set fast" like the Wierd of old Nordic mythology. Indeed, what we have in the (M) conception of spiritual laws for persons is a certain almost paradoxical combination of fate and freedom. Persons so conceived are free from the individual destinies of particular choices Molina and Leibniz prepared for them, but they are not free from the different feasible lines of fate into which their own choices may lead them. Fate on this conception is not singular, not determinate; it is like an uncollapsed wavepacket. Yet the few different progressions according to which it may unfold are utterly compelling for our volitional character. This can be clarified if we consider a single possible example of a putative spiritual law of feasibility:

Faustus's Law: It is unfeasible to pursue power for its own sake with arrogance and yet remain nobly committed only to the loftiest and grandest purposes. Pursuing this course exponentially increases the probability that one will acquire all sorts of little meanness and dispositions to petty villainy until these, rather than one's grand schemes, come to be all one cares about. It is absolutely necessary_F that this descent into puerile spite will occur unless one turns back entirely from the whole course that led to it (which it is always feasible but ever more difficult to do).

Of course, we have no empirical evidence adequate to prove that such a law holds for all possible persons. We only have introspection, the records of history, and the combined testimony of human experience inscribed in art and literature to suggest the likelihood that Faustus's Law is an unavoidable reality. Indeed, once we realize the possibility that spiritual laws for possible histories of volition exist, the philosophical relevance of classics in literature which bear directly on this theme greatly increases.⁽²⁰¹⁾

Of course, I also acknowledge that anyone, including an Existentialist, is free to challenge our specific characterization of any alleged spiritual law, which can never be more than imperfect approximations. My concern in this paper has not been to begin the large and vital task of describing particular spiritual laws themselves--this paper is only the propadeutic and prolegomena to the real task which lies ahead. In preparation for a full theory of spiritual laws, my goals in this paper have been (1) to explain the very notion of spiritual laws; (2) to demonstrate its philosophically crucial role in providing an account of our pre-theoretical modal intuitions about character and volition; (3) to show how a particular 'phenomenological' approach to personhood grounds the a priori postulation of such laws; (4) and finally to argue that such laws of feasible histories of willing are fully compatible with Existentialism. An Existentialist philosophy such as Soren Kierkegaard's which acknowledges the existence of spiritual laws of feasibility is thus not only consistent: it is also saved from the objection of arbitrariness which is so often levelled against Existentialist conceptions of human volition and freedom. Because of this, Existentialism can succeed in accommodating an adequate theory of feasibility which makes sense of our modal intuitions, while all the other possible conceptions of feasibility tried in the history of philosophy have failed. It is hard to conceive how there could be any stronger warrant for accepting Existentialism.

1. As Robert M. Adams has pointed out, the truths of "middle knowledge" in de Molina's sense are, strictly speaking, "subjunctive conditional propositions, many of which are strictly counterfactual" (Adams, "An Anti-Molinist Argument," in Philosophical Perspectives 5: Philosophy of Religion, 1991, ed. James E. Tomberlin [Ridgeview Publishing, 1991]). According to Molina, there are subjunctive truths of middle knowledge even for choice-circumstances which actually occur, so they are not all strictly speaking counterfactual.

2. Alvin Plantinga, God, Freedom, and Evil (Grand Rapids, MI: Wm. B. Eerdmans, 1974). All references are to the 1977 paperback edition (reprinted 1987).

3. Alvin Plantinga, The Nature of Necessity (Oxford University Press, 1974). All references are to the 1992 paperback reprint.

4. This is the phrase Plantinga prefers in The Nature of Necessity, ch.IX (see p.173).

5. Plantinga, The Nature of Necessity, pp.173-174, p.180-181. As Plantinga puts it, "suppose it is true as a matter of fact that if God had actualized T [the initial segment leading up to the choice], Curley would have accepted the bribe: then God could not have actualized W* [the world in which Curley rejects the bribe]" (p.181). See also God, Freedom, and Evil, pp.45-48.

6. This is my interpretation of Plantinga, God, Freedom, and Evil, p.46.

7. Plantinga, The Nature of Necessity, p.180.

8. ibid. Compare God, Freedom, and Evil, p.41, where, after giving two alternative counterfactual conditionals similar to those in Curley's case, Plantinga says "It seems clear that at least one of these conditionals is true, but naturally they can't both be....either way, there are possible worlds God could not have actualized." In The Nature of Necessity, however, Plantinga goes on to suggest that after we have understood "Leibniz's Lapse," "this assumption, harmless as it no doubt is, can be dispensed with" (p.180). But this further claim is quite false: Plantinga's entire free will defense absolutely depends on his Molinist supposition. Without it, there would be no way to make sense of the possibility that Curley and other possible persons suffer from "transworld depravity" (pp.185-6). More on this later.

9. I include Kierkegaard, Heidegger, Sartre, and Jaspers, although not Nietzsche, who is not an "existentialist" in any sense that makes him comparable to these other figures. It might be doubted that Heidegger could be described as a "libertarian," since it is well-known that he disputed Sartre's portrayal of radical freedom. Nevertheless, a strong Augustinian element of negative liberty is central to Heidegger's conception of freedom for unactualized possibilities. Part of the purpose of this paper, in fact, is to show how it is possible to hold the position Heidegger does hold, namely that negative liberty combines in a complex way with the substantive, 'historical' character of the self.

10. Martin Heidegger, Being and Time, tr. John Macquarrie and Edward Robinson (Harper and Row, 1962); ¶25, p.152 (H117).

11. ibid, ¶9, p.67 (H42). What Heidegger means by this, I think, is that a person is able to identify with or alienate their volitions (and even persisting aspects of their volitional character) in much the same sense described by Harry Frankfurt in his well-known essay on "The Freedom of the Will and the Concept of a Person" (Journal of Philosophy, LXVII No.1, January 1971), and later essays. Heidegger's point, like Frankfurt's, is that personhood requires a certain kind of self-relatable agency with a differentiated structure. There is much more to Heidegger's analysis of Dasein, but this feature is crucial to it.

12. Heidegger, H12, p.32-33: my italics.

13. ibid, p.67. As Heidegger says, "That Being which is an issue for this entity in its very Being, is in each case mine" (meaning first-personal).

14. Heidegger, H12, p.33: my italics.

15. Even if this is only a tacit choice to be "neglecting," or to be wanton in Harry Frankfurt's sense.

16. Letter on Humanism, tr. Frank A. Capuzzi & J. Glenn Gray, in Martin Heidegger, Basic Writings, ed. David F. Krell (Harper & Row, 1977): p.208.

17. ibid, p.205.

18. ibid, p.204.

19. ibid, p.196.

20. ibid, p.206; my italics.

21. In the Letter on Humanism, Heidegger is suspicious of "the personal" as a correct interpretation of Dasein or ek-sistence (see p.207). But this is because he assuming that "person" will have a reductive, 'mentalist' meaning. If "subject, person, or spirit" is treated as an extension of the rational soul as the "specific difference" definitive of the human as animal (p.203), then taking ek-sistence as personhood would be a way of illegitimately construing ek-sistence as a kind-essence. By person in my sense, I mean an individual being with an existential will and an irreducibly first-personal perspective, not a natural kind concept.

22. Thomas Flint, "Praying for Things to Have Happened," delivered at the University of Notre Dame, Spring 1994; mss p.5-6.

23. An intuitive sense of this analogy, I think, must be what leads David Lewis to think in a complete Molinist conception, God would have "middle knowledge about chance systems other than free creatures, for instance, radium atoms.."(See David Lewis, "Evil For Freedom's Sake?," Philosophical Papers, Vol. XXII (1993), No. 3.; p.159).

24. For example, all the l.p. worlds in which some information-bearing process travels faster than light are 'physically unfeasible,' relative to our universe.

25. Actually, there are good grounds for asserting that what Kant meant by analyticity already was a modal conception of necessity. In paper on Frege's concepts of logical consistency and independence, Professor Patty Blanchete has shown that for Frege, the consistency of a set of sentences was not a model-theoretic notion, but an implicitly modal notion (Notre Dame, Presentation Nov.11, 19914). It is very likely that Frege's richer notion of 'analyticity' was precisely the one Kant had held.

26. Quine's positivistic objection that concepts of synthetic modality are unclear is overcome by construing synthetic forms of modality this way. If one doubts that there are synthetic modalities in this new sense, at this point I simply urge that physical modality and Molinist feasibility are candidates, if we can provide an account of them. And since this is what I am examining, I ask that there be no positivistic foreclosure on the result in advance.

27. This idea of a Kripkean semantics for causal or 'nomological' modality has, of course, had many critics, such as Alexander Rosenberg and Norman Swartz. To answer their objections requires explaining how the limitation on the range of 'physically' possible worlds which is needed for such a semantics is grounded. Although I think this is far from impossible (and am working on such an account), for now it is sufficient if the very notion of a synthetic modality is clear enough, whether or not any such modality turns out to be well-grounded.

28. Plantinga, The Nature of Necessity, p.47.

29. Note that this does not mean that, in each logically possible world in which Curley exists, there is at least one situation requiring Curley to make a morally significant choice in which he goes wrong. Rather, if Curley exists in any logically possible world W1 in which he is sinless (which he surely does, for he is free with respect to each sin he commits), then there is at least one maximal world-segment representing a choice situation in W1, which is also a maximal segment in world W2 (with the difference that Curley does go wrong in W2), and if that maximal world segment became actual, he would do what he does in W2, not in W1.

30. Plantinga, p.48.

31. Plantinga, p.53. Creatable in this sense is a synonym for 'weakly actualizable' or 'feasible.'

32. This much suffices to answer J.L. Mackie's argument against the Free Will defense, but to show that Molinism provides a logically possibility for explaining the amount of evil in the world, more is required (as the discussion of cosmology explains).

33. We should remember that who else coexists with a given individual in a world will affect how depraved the individuals in that world are taken together. Other people's free choices (and their true counterfactuals of freedom) have much to do with determining which choice situations our individual finds herself in. Imagine two possible persons, L and G (for Launcelot and Guinevere). Imagine (for Molinism makes possible things of this kind) that the following counterfactual of G's freedom is true: in no situation in which G would be in a position to tempt someone to infidelity, would G do so, except for the one situation where G is in a situation to tempt L. And, likewise, in no situation in which L is tempted to choose infidelity, would L do so, except for the one case where it is G doing the tempting. Now, neither L and G are very depraved in their counterfactuals of freedom, looked at as a whole, and individually, either of them could be members of many other worlds without adding much to the total depravity of these worlds. But when circumstance brings them together, they add quite a bit more depravity. This situation creates even further opportunities for complex good for A (Arthur), and so on. In short, the depravity of a compossible set of persons is not simply the sum of their individual essential 'depravities.'

34. Whether there is a most depraved feasible world, of course, is irrelevant.

35. Plantinga, p.55.

36. Plantinga, p.56-57.

37. Flint, "Praying for Things to Have Happened," mss. p.8.

38. Flint, mss. p.7. In fairness to Flint, later in the paper he openly recognizes the limits of the this counterfactual power by distinguishing between weak counterfactual power over the past (which is ephemeral in the way indicated) and strong counterfactual power over the past, which is realized in counterfactuals with incomplete conditionals excluding, at least, the relevant fact about the past. Flint admits that only such strong counterfactual power could be relevant for the practice of praying. The problem, however, is that no matter how much is left out of the conditional, the Molinist only has two alternatives: either in the circumstance as incompletely specified, there is no fact about what agent A would do, or there is a set fact about what that agent would do--which again is not a fact they can change. Thus whether it is possible to maintain the existence of any strong counterfactual powers that do not ultimately reduce to weak or 'ephemeral' powers in a traditional Molinist system remains controversial at best.

39. William Haskar, God, Time, and Knowledge (Ithaca: Cornell University Press), pp. 29-52.

40. Robert M. Adams, "An Anti-Molinist Argument" in Philosophical Perspectives, 5: Philosophy of Religion, 1991, ed. James E. Tomberlin (Atascadero, CA: Ridgeview, 1991); p. 346.

41. William Lane Craig, "Robert Adams's New Anti-Molinist Argument," Philosophy and Phenomenological Research, Vol. LIV, No. 4 (December, 1994), p.857.

42. Adams, "An Anti-Molinist Argument," p.348.

43. Adams, "An Anti-Molinist Argument," p.346.

44. Adams, "An Anti-Molinist Argument," p.348.

45. Adams, "An Anti-Molinist Argument," p.348.

46. Adams, "An Anti-Molinist Argument," p.349.

47. Adams, "An Anti-Molinist Argument," p.347. This is conclusion "(9)" of Adams's "first stage of the argument." In this argument, Adams attempts to show how he can justify Haskar's conclusion in nine steps, using his own idea of explanatory priority. But then, to avoid depending on the premise that "(6) The relation of explanatory priority is asymmetrical," Adams develops a different argument against Molinism, starting from the same steps (1)-(5), but not going through the Haskarian conclusion (9).

48. Adams, "An Anti-Molinist Argument," p.346. Adams reiterates these steps on page 349 as the beginning of his new argument.

49. Adams, "An Anti-Molinist Argument," p.349. However, Adams should really have said that if I do A in a maximally specified circumstance C, then Molinism implies a counterfactual such as F*.

50. Adams, "An Anti-Molinist Argument," p.350.

51. Adams, "An Anti-Molinist Argument," p.350.

52. Craig, "Robert Adams's New Anti-Molinist Argument," p.858.

53. Craig, p.859.

54. Craig develops a parallel line of reasoning, starting from the premise that "A*. If children were born to us, they would come to love God." If such a counterfactual is true, and is the basis for the his decision to have children, then it ends up that "The truth of (A*) is explanatorily prior to our children's coming to love God" (Craig, p.860). Craig complains that he cannot even understand the sense of "explanatory priority" in this sentence. But is this the anti-Molinist's fault, as Craig seems to assume? The Existentialist has a ready explanation for why an incomprehensible explanatory priority is entailed by A*: it is because the truth of a counterfactual such as A*, prior to the existence and free choices of Craig's children, itself makes no sense! If the senselessness of the premise is carried through to the conclusion, this is the Molinist's fault.

55. Craig, p.860.

56. For Plantinga, essence is defined in terms of truth at every world in which the entity exists, i.e. in terms of a sum of what I am now calling world-relative truths (Plantinga, The Nature of Necessity: p.70 and p.77).

57. In W', he would take the bribe, because he does in W'. We could even say that at W', Curley has a whole set of counterfactuals of freedom that are true relative to W'.

58. In other words, the individual's counterfactuals of freedom are contingent in the de re sense, not only in the de dicto sense. I am indebted to Tom Flint for clarification of this point.

59. Of course, logically necessary truths can strictly be considered true relative to worlds--but in this case, the relativity is dispensable, since they are true at every l.p. world. At whichever one becomes the actual world (assuming actuality is not simply indexical), the necessary truth will be true.

60. In earlier drafts, I had used the term "world-independent" instead, but in discussion, Alvin Plantinga objected that this implied that there could be propositions which are true, but at no logically possible world--which of course is nonsense. By world-neutral truths, I clearly mean propositions whose truth precedes the actualization of any of the particular logically possible worlds at which they are true in the world-relative sense, not propositions that are true despite being impossible.

61. Whether they are feasible, of course, depends on all the true counterfactuals of freedom. To find the feasible worlds, we take each true counterfactual of freedom C*, find the worlds at which C* is true in the world-relative sense, and then derive the intersection set of these sets of worlds. The feasible worlds are those in which every counterfactual of freedom which is true simpliciter, is also true in the world-relative sense.

62. This theory is neutral between most 'ersatz' metaphysical interpretations of the logically possible worlds needed for Kripke-style modal semantics, but it does presuppose some version of "actualism" about possible worlds, whatever they are. If David Lewis is right that all the l.p. worlds are equally actual, or each actual-at-themselves in the same way as every other, then eo ipso there can be no world-neutral truths of the kind required for Molinist unfeasibility, or any kind of real synthetic modality, to exist.

63. Because Molinism insists on the individuals's logical possibility of doing otherwise.

64. is also of course compatible with worlds in which Curley exists but is never in C, and with worlds in which Curley does not exist at all.

65. Note that even though W' and W share the maximally specified choice circumstance S, in W Curley will face some choices as a result of taking the bribe which he cannot face in W'.

66. here is understood as a logically possible world in the ersatz sense, as distinct from a physical universe. is just the logically possible world that happens to be instantiated or actualized.

67. To any reader with Quinean scruples, broadly logical modality, logical essences and so on, are already bad enough. To add to this real synthetic modalities may seem like a final step to far. But my aim at this point is not to insist than any real synthetic modalities exist--it is only to claim that the very idea makes sense. In fact, Molinist 'feasibility' is useful to the extent that it has given us an example of the general structure which real synthetic modalities must have.

68. See, for example, Alvin Plantinga, "Actualism and Possible Worlds," originally published in Theoria 42 (1976); reprinted in The Possible and the Actual, ed. Michael J. Loux (Cornell University Press, 1979), pp. 253-273. Plantinga explicitly defines "actualism" as "the view that there neither are nor could have been nonexistent objects" (p.257). He also affirms that actuality is a more basic notion that "inclusion in the actual world" (p.258), which would seem to open the door to the sort of actuality of sub-maximal states of affairs which I am proposing. But in fact, 'the actual' is implicitly conceived to be maximal, so that it always constitutes the "domain of " (p.268).

69. Ultimately, I believe it is this notion of actuality which is integral to Kant's conception of metaphysics, which is inherited in large part by the existentialists. Therefore it should not be surprising that existentialists generally take the existence of submaximal states of affairs, such as those constituting time, to be more primordial than the complete actuality of a logical construct, such as a possible world. To the existentialist, actuality is always a fragment. Plantinga's conception of actuality gives priority to a deceptive abstraction, while Lewis's levels off the entire significance of 'actuality' as a mode distinct from 'possibility.'

70. The Existentialist, incidentally, cannot object to the very notion of real synthetic modalities in the sense I have described. The basis of her objection to Molinism cannot be simply that the Molinist must hold that there is some metaphysical 'priority' which makes world-neutral truths conceivable. For the Existentialist herself actually requires some such 'priority' with a continuum of ordinal positions. The reason is that Existentialism maintains that at any point in real time in which free beings are actualized, only a fragment of is actualized; which single and complete logically possible world turns out to be 'the' actual world is determined only internally within actual time itself, both by physical chances and choices ultimately rooted in spontaneity. The Existentialist is committed to holding that as a matter of fact, there is no maximal, logically possible 'world-proposition' whose truth is metaphysically prior to the actual unfolding of time and collapsing of its wavepackets and 'choicepackets' (if you'll pardon the phrase) in any partially-actualized physical universe that exists.

In my model, what is actual at each point in time, as conceived by the Existentialist, is a set of world-neutral truths that 'color' a certain plurality of l.p. worlds as -relevant. At each subsequent point, the 'sphere' of the -relevant worlds has grown smaller. Time is thus a continuum of synthetic ranges of possibility, in which the 'past' states of affairs ever more narrowly restrict the range of the 'temporally possible' in the future. The 'truth' of propositions is thus literally time-indexed for the Existentialist, which explains why for her, there can as yet be no truth or falsehood for contingent propositions about the future. All states of affairs whose 'actuality' is metaphysically prior to the present actualities exist in whatever l.p. world turns out to be actual, but they underdetermine it. The eventual actuality of that maximal, l.p. world, including the actual future of our universe, is posterior to and dependent on the unfolding actuality of sub-maximal parts of it we already have. This radically metaphysical interpretation of time thus depends on the same idea of metaphysical priority and world-neutral truth as does the Molinist's account of the truths of divine middle knowledge.

71. Craig, p.860.

72. Again, in our placeholder 'P' sense, letting the Molinist fill in the value of P.

73. This will be so as long as we have picked a C that did not depend on unfeasible actions of other people.

74. Craig, p.860.

75. For example, see Tom Morris's discussion of these approaches as strict alternatives in Our Idea of God (University of Notre Dame Press, 1991), ch.5, pp. 94-96. Morris's natural impression that truths of middle knowledge must be understood as hard facts is conveyed by the following passage: "Unlike Ockhamism, it [Molinism] offers an interesting account of how God is able to know the future. And it carries no hint of an implication that, in order to be free, I must be able to act in such a way as to change or determine what God's beliefs about me have already been" (p.96).

76. For this reason I find Tom Flint's suggestion that "the everyday believer's view of providence is simply rough-hewn Molinism" dubious at best (Flint, "Middle Knowledge and the Doctrine of Infallibility," in Philosophical Perspectives 5: Philosophy of Religion, ed. James E. Tomberlin). If by this Flint means that Christians intuitively hold beliefs that can be expressed in a soft fact interpretation of Molinism, the claim is absolutely unconvincing. Soft fact Molinism is the sort of doctrine that holds together a position which is obviously contradictory on its face by splitting up concepts in such a way as to reduce all the terms in the claim to virtual meaninglessness--the kind of mystification which gave scholasticist thought such a bad name. Soft fact Molinism, when unmasked, is a thousand times more inscrutable than Duns Scotus's modest suggestion that there could be epistemologically objective "formal distinctions" between entities that are nevertheless metaphysically inseparable. This claim at least has an intelligible meaning, but the 'soft fact' Molinist's circle of an actual act of the agent prior to his actuality can literally be said to be without meaning.

77. Craig, p.861.

78. This will have to mean 'true' in world-neutral sense, as we have seen, since not all c.o.fs about a person that are true relative to some l.p. world are also true in the privileged sense required for a truth of middle knowledge. For example, it is not part of middle knowledge that at W', Curley would refuse the bribe in C. What middle knowledge contains is the world-neutral truth about what Curley would do in C--namely, take the bribe.

79. Adams, "An Anti-Molinist Argument," p.345.

80. However, this certainly cannot be a straightforward causal sense. See Adams's "An Anti-Molinist Argument," endnote 7, p.352.

81. William P. Alston, "Divine Foreknowledge and Alternative Conceptions of Human Freedom," first published in the International Journal for the Philosophy of Religion, 18 (1985): 19-32; reprinted in God, Foreknowledge, and Freedom, ed. John Martin Fischer (Stanford University Press, 1989): p.270. The original quotation comes from Plantinga, God, Freedom, and Evil, p.71.

82. Alston, p.266.

83. Alston, p.271.

84. I am using "identification" here in the sense in which it is often used by Harry Frankfurt, namely as the central problematic of personhood.

85. See, for example, David Lewis, "Evil for Freedom's Sake?," Philosophical Papers, Vol. XXII (1193), No.3. Lewis asks, "What does make unfulfilled counterfactuals of freedom true?" (p.161). See also Robert Adams's first critique of Molinism in his "Middle Knowledge and the Problem of Evil," first published in American Philosophical Quarterly, 14 (1977), pp. 109-17; reprinted in The Problem of Evil, ed. Marilyn McCord Adams and Robert Merrihew Adams (Oxford University Press, 1990), p. 113f [All further citations of this paper refer to the pagination in The Problem of Evil]. In this paper, Adams argues that the Molinist has no clear conception of how anything could make counterfactuals of creaturely freedom true.

86. It was Anscombe who first analyzed agency along these lines, but this standard for agency is also found in Donagan's The Theory of Morality and in many other discussions of action theory.

87. See Robert Adams, "Middle Knowledge and the Problem of Evil," p.115.

88. Adams is therefore wrong to say that Suarez's position is "the least clearly unsatisfactory type of explanation for the alleged possibility of middle knowledge" (Adams, p.115). Although Suarez can simply bite the bullet in the face of the standard analytic objection, his position reduces human responsibility for our actions to an absurdity.

89. This is a view I do not find represented in the literature. Obviously, it is not introduced here as a straw man version of Molinism, for my purpose is to show that every possible version of Molinism, including this one, is untenable because of the trilemma. If you don't agree with me that sophisticated Molinism is the Molinist's best option, choose some other way of trying to accommodate (1), (2), and (3): all will fail.

90. Notice that Harry Frankfurt's understanding of human freedom comes close to 'Essentialism' of this kind. Frankfurt tries to escape the 'antinomy of authenticity' by arguing that the logical possibility of acting otherwise is not a necessary condition for authentic identification (condition 2). This position also fails, I believe, because, whatever we say about actions, alternative logical possibilities for higher order willing and caring themselves are required if these are to be the ground of our identification with actions. Kant's Incorporation Thesis will still apply to higher-order maxims, whatever we say about action-maxims. But critiquing Frankfurt's position is not my business in this paper.

91. David Lewis, "Evil for Freedom's Sake?", pp.160-161.

92. This point is clarified by a distinction Robert Adams makes between truths of middle knowledge as "deliberative conditionals" versus "counterfactuals:" as he says, a deliberative conditional "ought not, in strictness, to be called counterfactual. For in asserting one of them, one does not commit oneself to the falsity of the antecedent" (Adams, "Middle Knowledge and the Problem of Evil, p.118). The conditionals involved in Molinism are neutral as to the actuality of their antecedent.

93. I am indebted to Alvin Plantinga for this counterargument. However, I have rephrased it in my own words, and I am solely responsible for any infelicities it might involve.

94. This, of course, is the rigorous meaning of the term "world-independent," which I used earlier in the analysis to explain the middle ontological status of the truth of c.o.fs.

95. Tom Morris, Our Idea of God (University of Notre Dame Press, 1991), p.120

96. Morris, p.131.

97. Morris, p.132.

98. For in fact, a even a single "act" outside of time would not be an action if there is not a "time" or ordered series for God in which the state of having acted is metaphysically prior to the "action." Action intrinsically involves a time-like ordering in actuality, and hence the notion of an eternally present and unchanging state without antecedent, which is nevertheless called an "action," is once again incoherent.

99. In fact, God in the Augustinian-Anselmian view has a kind of noumenal actuality with no possible change of character. This is very similar to the conception of agency used to ground feasibility on the 'Kantian approach:' see §IV (C).

100. Robert Adams, "Middle Knowledge and the Problem of Evil," p.122.

101. Plantinga, The Nature of Necessity, p.177.

102. Unless one insists on remembering that there is an extremely small quantum mechanical probability of Robins not falling but reappearing on the ledge.

103. This was the basic insight Harry Frankfurt used in developing his theory that second-order volitions are a necessary condition for personhood in the moral sense. See his paper, "Freedom of the Will and the Concept of a Person," reprinted in Frankfurt, The Importance of What We Care About (Cambridge University Press, 1988).

104. Virtues and vices are volitional dispositions of this kind. Frankfurt refers to such lasting dispositions as volitional states of caring. He also insists that a person's character is more determined by cares of this type than even by higher-order decisions to take certain attitudes towards one's impulses and desires--for the latter, despite being higher-order, may not last and become effective in the way a care does by definition: "If we consider that a person's will is that by which he moves himself, then what he cares about is far more germane to that character of his will than the decisions or choices he makes. The latter may pertain to what he intends to be his will, but not necessarily to what his will truly is" ["The Importance of What We Care About," reprinted in The Importance of What We Care About, p.84].

105. I will always use this term to mean second-order dispositions, or dispositions to acquire and maintain certain lasting second-order volitions, which in time become effective in controlling the first-order will on which one acts in different types of circumstance.

106. Robert Adams, "Middle Knowledge and the Problem of Evil," p.115.

107. For example, Harry Frankfurt and David Lewis.

108. This is one possible way of construing Sartre's theory in any case. The usual simplistic reading of him as celebrating arbitrary choice without any recognition of the importance of character and its levels, however, is grossly misrepresentative. Still, I have picked Sartre's theory as popularly understood for an example of 'simplistic' existentialism in order to make the point clear.

109. Facticity in general simply means synthetic necessity.

110. Although, as we saw, it turns out that on his own metaphysics, the Molinist is not entitled to this claim.

111. Morris, Our Idea of God, p.71.

112. ibid; my italics.

113. Frankfurt, "The Importance of What We Care About," in The Importance of What We Care About, p.86.

114. Frankfurt, ibid, p.86. Frankfurt says here that the volitional necessity of a course of action is the experience of finding "that every apparent alternative to that course is unthinkable," and he goes on to analyze the sense in which this extreme form of care is must be involuntary (not simply voidable at will), although self-imposed at some level. He also uses the phrase "unthinkability" to refer to the same sense of volitional necessity in later essays, but in some sense it is an unfortunate choice of terms, because Frankfurt obviously accepts that Luther think about retracting his Theses: he just cares about them in such a way that he could not will to retract them.

115. Frankfurt, "Rationality and the Unthinkable," new in The Importance of What We Care About, p.181.

116. Harry G. Frankfurt, "Rationality and the Unthinkable," p.179. Frankfurt borrows the phrase from Rawls, who criticizes utilitarianism for defining persons as "what we may call 'bare persons'" who lack any determinate conceptions of the good which commit them to substantive loyalties--see Rawls, "Social Unity and the Primary Goods."

117. ibid, p.178.

118. In the analysis of physical modality proposed above, obviously what unifies the relevant sphere of logically possible worlds is that the same physical universe exists in all of them. This is an idea I have developed more extensively in a paper on 'Real Physical Modality.'

119. Note that the 'spiritual world' in this sense is a sphere of -relevant logically possible worlds. Thus neither the spiritual nor physical 'world' are worlds in the ersatz sense, such as logically possible worlds. The term spirit is only used in the Husserlian sense to indicate volition or will or agency.

120. A very different theory might hold that the feasibility of entire l.p. worlds is a modal primitive, without making this supervene on the primitive feasibility of individual choices. This theory would rely on a modal primitive, but would not be 'atomistic' in the relevant sense.

121. Lewis, "Evil for Freedom's Sake?," p.151.

122. ibid, p.152.

123. This strategy also makes the free will defense particularly vulnerable if we have any independent grounds for thinking that transworld depravity is implausible. Robert Adams, for example, argues that it is not plausible, and that Molina and Suarez would deny that it is true of free creatures ["Middle Knowledge and the Problem of Evil," p.123]. I should also make clear that this reduction to mere defense is not the only problem which Molinism makes for answering the problem of evil: there are others I cannot discuss here.

124. By this, I am as far as possible from wishing to imply that Plantinga's free will defense was not an invaluable contribution to Christianity and the philosophy of religion in general. Plantinga's refutation of atheistic claims that theism in the Judeo-Christian sense is logically inconsistent achieved a groundclearing without which no further progress in the philosophy of religion was impossible. My admiration and debt to Plantinga's groundbreaking argument cannot be overstated. Doing 'better' would ultimately require showing that the Existentialist can produce a more satisfying defense against the problem of evil, and thus produce an advance on the defense Plantinga has already made.

125. Copleston, p.342.

126. ibid, p.342-3. This notion of efficacious grace, of course, was the motivation for the postulation of Molina's scientia media, for it means sufficient grace that would be freely accepted--and thereby includes the intuitive idea of the synthetic modality of feasibility.

127. Plantinga, p.48.

128. I believe Soren Kierkegaard's conception of original sin as presented in The Concept of Anxiety and later works will probably meet this requirement.

129. Hugh MaCann, "Divine Sovereignty and the Freedom of the Will," delivered at the Conference on the Philosophy of Mind, University of Notre Dame (November 3-5, 1994); forthcoming in Faith and Philosophy. My references are to the manuscript pages of Professor McCann's address (cited with permission of the author).

130. McCann, mss. p.1

131. In the fourth section of his paper, McCann wants to argue that liberatarianism can be reconciled with the view that "God originates my decisions and actions, simply creating me the person I am" (mss. p.13-14). I will not comment on this part of the paper, although I am pessimistic about its prospects.

132. McCann, mss. p.8-9.

133. McCann, mss. p.14.

134. When Aquinas makes room for negative liberty in his account of the will, as he certainly does in Summa Theologiae I Qu.87 Art. 3-4, I believe his own account requires that our will is created or held in existence in a kind of 'superposition' between a range of options (which are not arbitrary, since they are determined by my collected apprehensions). It is I, however, and not God, who extend God's creative action to a particular decision through my act of comparative judgement (for it is at this stage that negative liberty enters in for St. Thomas). Notice that this is not quite the Existentialist view that the person literally actualizes their decision ex nihilo, but it is coming close to this.

135. Copleston, p.343; my italics.

136. Kathryn Tanner, God and Creation in Christian Theology: Tyranny or Empowerment? (Basil Blackwell Ltd., 1988).

137. Thomas F. Tracy, Review of Kathryn Tanner's God and Creation in Christian Theology, in Faith and Philosophy, Vol.9 No.1 (January, 1992); p.123.

138. See §II (B).

139. ibid.

140. Copleston, p.343.

141. Tanner forgets that the puppet also has a 'vertical' relation to the puppeteer, while nevertheless not being determined in its movements by other objects on the stage.

142. McCann, mss. p.12.

143. Lewis, "Evil for Freedom's Sake?", p.149. Even an approach such as John Hick's "soul-making theodicy" in fact presupposes some connections between free actions and valuable states of character which are feasibly necessary--or 'necessary' in a synthetic sense independent of God's will. Only this explains why complex goods would require a "vale of soul-making."

144. McCann, p.12-13.

145. Immanuel Kant, Religion within the Limits of Reason Alone, tr. Theodore Greene and Hoyt Hudson, with introductory essays by Theodore Greene and John Silber (Harper & Row, 1960); p.17.

146. ibid, p.19, footnote on Schiller.

147. ibid, p.20.

148. Immanuel Kant, Critique of Pure Reason, Unabridged Edition, tr. Norman Kemp Smith (St. Martin's Press, 1965); A533/B561, p.465.

149. Kant, Religion, p.19: this is the famous principle which Henry Allison calls Kant's Incorporation Thesis.

150. Kant, Religion, p.17.

151. Kant, Critique of Pure Reason, A532/B500, p.464.

152. Kant, Religion, p.20-21; my italics.

153. Note that it is necessary to distinguish the disposition or ultimate character of a person's will from the natural predispositions to good (e.g. the predisposition to "personality") which are implied in the very structure of human nature. Human tendency to evil, however, just is a disposition in the acquired sense: this disposition, which serves Kant as an analog to original sin, is itself blameworthy (see Philip Quinn's article, pp.191-193). The natural rather than acquired predisposition which makes a disposition to evil possible, however, comes about from the inevitable conflict of the predispositions to personality, humanity, and animality--i.e. from the difference in good predispositions between which Willkür must chose its disposition (see Quinn, p.196-7).

154. Philip L. Quinn, "Original Sin, Radical Evil, and Moral Identity," Faith and Philosophy, Vol.1 No.2 (April, 1984); p.193.

155. I am indebted to Phil Quinn for clarifying this point in discussion.

156. John Silber, "The Ethical Significance of Kant's Religion," in the Introduction to Religion within the Limits of Reason Alone; p.cxvi.

157. Quinn, p.193-4.

158. Silber, p.cxix. I think Silber is careless, however, when he adds that "Since Willkür is free, its dispositional aspect is alterable from one moment to the next" (p.cxvii). Quinn's analysis makes clear why this interpretation is erroneous.

159. As Silber says, "The goodness of the weak will is revealed by the continuing prominence and even dominance (when one thinks of the great moral suffering in comparison with the slight satisfaction of the illegal desire) of the Wille" (p.Cxxi).

160. Frankfurt, "The Importance of What We Care About," p.88.

161. Frankfurt, ibid, p.87-8. This description seems to me to fit particularly well with Kant's view of the wicked man, who wills "to ignore the moral law and to oppose its demands when it interferes with his non-moral incentives" (Silber, Kant's Religion, p.cxxiv). The wicked man in a sense cares about preserving his disposition to adopt heteronomous action-maxims: thus he works to reduce the remaining capacity in his Wille to undermine the steadiness of this disposition.

162. Frankfurt, "Rationality and the Unthinkable," p.182.

163. ibid, p.184.

164. Kant, p.20. It might seem paradoxical that I am comparing this aspect of Kant's theory to Frankfurt's notion of the identification involved in volitional necessities, since Frankfurt famously rejects the negative liberty which Kant's Willkür embodies. This is unavoidable, however, for it is precisely at this point that Frankfurt is paradoxical: although he rejects the principle of alternative possibilities for actions, he has not realized that principle must apply to higher-order volitions themselves if these are to be the bearers of identification or "incorporation" in the Kantian sense.

165. This is of course another critical step towards the habit-based account of feasibility which fits with Existentialism. Sections V-B and VI-A will discuss the question of probabilistic qualification in greater detail.

166. Kant, Religion, p.42-3.

167. Kant, Religion, p.20.

168. Quinn, p.199.

169. I should note in passing that Kant's theory (K) probably provides a better model for original sin than does transworld depravity. As Quinn notes, "it seems very improbable that a propensity to moral evil should be both a product of freedom and universal among mankind" (Quinn, p.194). Yet, as we saw, on a Molinist foundation any pattern of feasibility, however vastly improbable, is in principle without explanation. The same would hold for so improbable a pattern as one that constitutes original sin. But Kant, by contrast, "sees quite clearly that the universality he attributes to the propensity to moral evil needs an explanation of some sort" (Quinn, p.194). For Kant, it is the differentiated structure of the Willkür itself which is supposed to provide this explanation.

170. Quinn, p.197.

171. Kant, p.41-42.

172. Kant, p.40. Kant gives versions of this argument for the possibility of moral revolution several times in the succeeding pages: pp.41-43. Here and at the end of Book I, however, he allows it is possible that some inexplicable assistance by divine grace is also a necessary condition for the revolution. But if so, he does insist that it is our own free choice by which we "render ourselves susceptible of...inscrutable assistance" (p.40-1). In other words, sufficient grace is never by itself effective grace.

173. Kant, p.42.

174. For an example of this theme in another Pietist author, see Martin Buber, The Way of Man According to the Teaching of Hasidism. Without the essential "turning" in which one obeys the command to "turn wholly away from evil" (p.32-33), a man has not united his soul, and any outward progress he makes will be "patchwork" (p.22).

175. Kant, p.43-44.

176. Kant, p.46. And thus we have the following argument:

(1) Assume the thesis of radical evil.

(2) Radical evil entails that moral education is reformative.

(3) Assume the unity of virtue.

(4) If moral education is reformative and the unity of the virtues holds, then

moral education aims at a fundamental revolution in one's ultimate maxim.

(5) Thus, if moral education is possible, it follows from (1) & (3) that a fundamental revolution in

one's ultimate maxim is possible.

177. Kant, p.47.

178. This is the theme which of course is emphasized in Kierkegaard's argument that religious truth consists in subjective appropriation, an orientation of the whole self in faith.

179. It is well-known that Kierkegaard was influenced by Kant's Religion, although as Silber rightly points out, Kierkegaard opposed Kant's belief that strong 'demonic' evil is impossible for human beings (p.cxxix). What is less well-understood is that for Kierkegaard just like Kant, the absolute indispensability of good works for salvation through faith was a central principle. Not only is this clear in his later religious writings [see Walter Lowrie's Introduction to Attack Upon Christendom (Princeton University Press, 1968): p.xvii]. It is also suggested in passages early in Fear and Trembling that are clearly intended as barbs against Danish adherents to sola fide. For example, Kierkegaard tells us that justice holds in the "world of spirit" and "Here it holds true that only the one who works gets bread.." [Fear and Trembling, tr. Howard V. Hong & Edna H. Hong (Princeton University Press, 1984): p.27; my italics]. The indispensability of good works is expressed for Kierkegaard by the fact that the religious stage is only accessible through the ethical stage, and strong moral virtue remains a necessary condition within the stage 'beyond' it. Faith that does not dialectically contain rational morality within it--i.e. cumulatively retain the demand for good works--is aesthetic childishness p.47). Moreover, the movement of faith is also essentially a volitional movement of the self rather than an assertion of any proposition to be believed sola fide. Thus, the sense in which the religious stage of faith is 'beyond reason,' then, is the same as the sense in which it is 'beyond the ethical:' it contains reason dialectically within it, just as it contains the ethical. Kierkegaard's 'beyonds,' unlike Nietszche's, are cumulative, and for this reason, Kierkegaard is not in any sense a fideist. On perhaps no other point has he been more misunderstood.

180. It is a measure of the strength of Kant's insistence on good works and moral progress for eschatological salvation that John Silber even felt that Kant's theory is inimical to any adequate notion of divine forgiveness or mercy (Silber, p. cxxxi). I am not sure if this is a fair criticism of Kant, but it certainly indicates how strong his emphasis on morality as an essential element of religiousness was.

181. C.S. Lewis, That Hideous Strength, p.358.

182. One could of course consider each distinct 'value' of this variable to be a constellation or combination of volitional dispositions and relevant character-traits.

183. Robert Adams, "Middle Knowledge and the Problem of Evil," p.112.

184. Adams, p.113.

185. Adams, p.114; in Adams, these are numbered (5) and (6). I have renumbered them for convenience in this essay.

186. Adams, p.117.

187. This example is prompted, obviously, by the surprising reversal in this character's behavior in the end of the film, Return of the Jedi, in which Vader overcomes his vices and makes a moral choice.

188. Particular theological difficulties would flow from this--for example, it would remove a large measure of the accepted theological meaning of the sin of pride if pride were a state that is unfeasible for any person to overcome. In addition, it would collapse the distinction between simple vice and reprobate vice for those individuals who happened, by their hacceity alone, to be certain never to overcome the habit.

189. St. Thomas Aquinas, Summa Theologiae, Part Ia Qu.83 Art.4.

190. That Aquinas believed this may be inferred from the fact that he considered the acquisition of malicious habits, even from quite a young age, to be a sin.

191. See §III (B) for a full discussion of the problems with 'Simplistic Existentialism.'

192. The Basic Works of Aristotle, ed. W. D. Ross (Random House, 1941); Nicomachean Ethics (tr. W. D. Ross), Book III, ch.5: p.973.

193. We might say that within the habit-based approach, theory (R) is to theory (A) much as (P) was to (K). (R) fails for the same reasons as traditional Molinism, and (A) is inadequate for the same reasons as (K).

194. This surprising theorem helps bring out the fully systematic nature of the Existentialist conception of feasibility. It helps us to see why other conceptions, such as (A), have only partial or incomplete pictures. Since it is also possible to argue that IP makes no sense on any account but a habit-based account, IP will also ential premise (a). If this is granted, since (a) entails LI, we could then say that IP LI.

195. By necessary in this sense, of course, I do not mean broadly logically necessary. Rather, the connection envisaged by this paradigm form is a synthetic one, having something to do with the nature of the choices, the circumstances specified, and the person.

196. Immanuel Kant, Groundwork of the Metaphysic of Morals, tr. H.J. Paton (Harper Torchbooks, 1964); Academy pages 390-1; Paton page 58.

197. This is why in Kant, for example, there can be no schematism for morality. Schemata are determinate connections between phenomena which Kant assumes must be due to the nature of specifically human faculties--e.g. the pure intuitions of space and time.

198. Note that it is precisely because they accept this presupposition that Thomists argue that any substantive, non-formalist account of the Good must be based on a single telos of human nature. Although Kant draws the opposite conclusion because of his emphasis on transcendental spontaneity as a condition for moral responsibility, his explicit arguments that all determinations of the Good in terms of that which leads to happiness or well-being will be particular to human psychology actually make use of the same presuppositions as Aquinas employs in his arguments concerning the nature of the last end for human beings. On my account, we can find something more than bare negative liberty--something substantive, if not as `thick' as a Thomistic rational soul--at the level of persons in general.

199. The defense of the features added to negative liberty here as actually found in persons and as definitive for personhood in the forensic sense is the topic for another examination, which I have carried out in depth in other papers. Here I simply intend to show how the results of these other arguments can guide our attempt to reformulate Molinism.

200. Or, more technically, global connections relating choices to differences in openness of options for relations with other persons in the social world.

201. One might cite C.S. Lewis's Screwtape Letters, for example. This work is really a study in spiritual laws in the sense that has now been rigorously defined.