Overview

In 1970, R. J. O'Connell observed a ‘strangely Kantian’ tone in the lex aeterna of Augustine's De libero arbitrio I. Although nearly fifty years have passed, O'Connell's tantalizing observation remains undeveloped. My purpose in this article is to bring out more fully the role played by the concept of law in De libero arbitrio I, including aspects that indeed might be called Kantian. My method is reconstructive. First, I examine a proper fragment of Augustine's account of moral agency, expressly not including his concept of law, and I note some of the fragment's implications. In brief, knotty problems emerge, including, rather startlingly, a puzzle familiar from Stoic philosophy. I then restore Augustine's concept of law and observe the result – in brief, a neat resolution of all of the problems generated by removing it. The role of the concept of law in Augustine's account of moral agency makes it clear that, far from being superficially Kantian, as one might surmise from O'Connell's comment, Augustine's complete account of moral agency in De libero arbitrio I is in fact deeply Kantian. There are, however, important differences between the two approaches, two of which I conclude by indicating.

Background

We begin by setting the stage. Augustine reflected on the problem of postlapsarian human moral agency continuously over the course of his long career. Whether or not his commitments changed over that time, his emphasis certainly did: he began by insisting that postlapsarian humans are free and ended by insisting that humans are doomed to sin. This change is partly explained by the positions against which he was disputing. His first efforts were directed against the Manichaeans, who explained sin as the result of a power of darkness overcoming the soul. Against this, Augustine argued that postlapsarian humans are wholly responsible for the evil they do, a claim that presupposed, or so he took it, that humans are free. Augustine's later works were directed against the Pelagians, who held that humans can earn salvation by rightly exercising their free will. Against this, Augustine argued that postlapsarian humans are saved by divine grace alone, a claim that he took to presuppose an otherwise insuperable propensity to sin. The Pelagians were quick to note the apparent similarity of Augustine's later position to Manichaeism, and they enthusiastically cited Augustine's earlier work against him. Despite thus having ample opportunity to disavow his youthful commitments, Augustine never did. The absorption of those commitments into his mature philosophy is a subject of extraordinary and, one may still say, urgent interest, but I shall not examine it here.

The first book of De libero arbitrio, written in 387–389, soon after Augustine's baptism, belongs squarely to the early period. He is principally concerned in this book with finding agents responsible for the evil they do, but he makes it clear, as we will see, that postlapsarian human freedom extends also to the attainment of a narrow class of goods. The book takes the form of a dialogue between Augustine and his friend Evodius. As the book opens, Evodius is seeking to learn whether God is the ultimate cause of the evils committed by postlapsarian humans; Augustine directs the conversation as one who has already deeply considered the question. The two begin by specifying the class of evils concerned in their discussion. We take up their discussion at this point.

Puzzles

Just punishment is evil in the suffering, Augustine and Evodius agree, but not in the doing. This acquits God of doing evil when he punishes sinners, but it leaves unclear what, if not suffering, makes evil done blameworthy. ‘[A]ll evildoings are evil precisely because they come about from lust, that is, from a blameworthy desire’;Footnote ¹ in turn, blameworthy desire is ‘the love of things one can lose against one's will [earum rerum amorem, quas potest quisque inultus amittere]’.Footnote ² Augustine calls those things that one can lose against one's will temporal, while those that cannot be so lost he calls eternal.Footnote ³ These definitions are rendered without qualification.

When one thinks of temporal as against eternal desires, what leap to mind are such morally dubious desires as those for physical pleasure and money. But Augustine's definition is radically more encompassing than the usual, with troubling implications for his account of moral agency. Three implications stand out.

First, on Augustine's unsupplemented definitions there can be no distinction between evils as to their severity. Temporal aims are temporal without degree, and so likewise are the evils they constitute. But surely negligent homicide, while bad, is less bad than premeditated murder.

Second, Augustine's definitions afford no moral distinction between temporal desires that are, on their face, some of them good and some of them bad. For example, one can no more save the life of a drowning child on merely willing it than one can drown a child on merely willing it, yet one should like to distinguish ethically between those willings. On Augustine's definition, they are both temporal, hence blameworthy.

Third, even where the account correctly identifies an evil, it does so on peculiar grounds: willing an innocent's drowning counts as evil, as one should expect, but the evil consists in willing something that might not come to pass despite one's willing it. Similarly, any good is good on account solely of the agent's ability to secure it at will.

To my knowledge, no philosopher before Augustine commits himself to the second and third implications, but the first implication is in essentials a well-known ‘paradox’ attributed by Cicero to the Stoics. As Epictetus puts it, more than a century after Cicero, a person should so direct himself that he ‘may never fail to get what he desires, nor fall into what he avoids’.Footnote ⁴ The concept of eternality in lib. arb. has its root here: the will should be trained on those objects that it cannot fail to obtain. Now, the Stoics, as observed by Cicero, held the reward for so training the will to be maximal: ‘No one who is totally self-reliant, and contains within himself all that he owns, can fail to be completely happy.’Footnote ⁵ By the same token, every failure of will, no matter how small, counts as evil: ‘We must consider what custom forbids as a crime, but what law forbids as sacrilege. ‘Even in little things?’ Yes, since we may not of course be able to shape the way things are, but we can control our own minds.’Footnote ⁶ By avenues that Cicero leaves dark – Epictetus, too, for that matter – these considerations lead to the ‘Stoic paradox’ identified by Cicero: ‘all evils are equal, and similarly all right actions [Aequalia esse peccata et recte facta]’.Footnote ⁷ This is the first troubling implication mentioned above.Footnote ⁸

Be Augustine's historical antecedents what they may, the first implication of Augustine's definitions is indeed troubling. An account of moral agency should, where possible, avoid flattening evil in just that way. Something similar goes for the second and third implications: an account of moral agency does well to avoid them. We will find it useful, in the discussion that follows, to regard these implications as puzzles raised by Augustine's account of moral agency, puzzles that his account should, if possible, supply the resources to solve.

Solution

We can think of the puzzles identified in the previous section as laying down conditions on philosophically and exegetically satisfactory readings of lib. arb. So formulated, the conditions are these:

1. 1. A reading of lib. arb. should furnish a distinction between worse and less bad blameworthy desires (first puzzle);

2. 2. A reading of lib. arb. should furnish a distinction between, as it were, bad blameworthy desires (desire for a drowning) and good blameworthy desires (desire for a child's physical well-being) (second puzzle);

3. 3. A reading of lib. arb. should furnish plausible grounds for judging something evil or, likewise, good (third puzzle).

Together these conditions suggest another:

1. 4. A reading of lib. arb. should identify, wherever there are, as it were, good blameworthy desires, blameless desires that conduce to the same ends. So, for example, if it is evil to desire the physical well-being of an innocent, there should be some blameless desire that motivates agents to preserve the physical well-being of innocents. In short, a reading of lib. arb. should supply at least the outlines of a plausible theory of practical reason.

The fifth and final condition on satisfactory readings of lib. arb. is the general one of fealty to the text:

1. 5. A reading of lib. arb. must be not only consistent with Augustine's text but, to the degree possible, firmly rooted in his expressed views.

For a reading of lib. arb. satisfying these conditions, we turn to Augustine's conception of the law. First, however, it will pay to motivate that conception with an example.

Suppose that you are walking along a river when you see a drowning child. The child is near enough that you might save him, and you dive in. If we take as given the lib. arb. doctrines so far discussed, the physical well-being of the child is temporal and your desire for his well-being is evil. What possible justification could there be for this? The conclusion seems perverse, but a motivation is available. Suppose that, despite your best effort, you fall one foot short of saving the child's life. Clearly, there must have been something in the circumstance that it was right for you to do. But it could not have been actually saving the child's life – for, if it had been, you would have failed to do the right thing and, so, committed evil, your unsparing effort notwithstanding. Identifying the child's well-being with the good it was yours to achieve renders the rightness or wrongness of your action a matter of luck.

Note that a retrospective strategy fails against this scenario. For example, let it be laid down after the fact that your falling one foot short was the right thing to do all along. Then doing right depended on your not suffering a stroke at two feet short. But surely your doing right could depend on no such thing. Since at no point in one's progress towards saving a drowning child is one immune to bad luck, no single step along the way can constitute the right thing to do. Indeed, a retrospective strategy must fail even leaving aside the possibility of bad luck: right action should reach as far as the child's actual rescue, circumstances permitting.Footnote ⁹

What is the right thing to do in the case of the drowning child? Is an answer available that satisfies the five conditions? Here we take up Augustine's discussion of law. But Augustine prefigures that discussion in a short discourse on learning, which it will be useful to review first.

Evodius' initial thought, when embarking on the search for an ‘author’ of evil, is that postlapsarian humans learn to do evil, hence that evil can be traced to a teacher. Augustine rejects this, on the grounds that ‘if [someone] is evil he is not a teacher, and if he is a teacher he is not evil [Si enim malus est, doctor non est: si doctor est, malus non est]’.Footnote ¹⁰ The principal steps toward this conclusion are characterizations of disciplina (learning) and intelligentia (understanding) as success-terms: something counts as learning only if it improves understanding, and something counts as understanding only if it is good. Whatever goes on in a case that looks like evil instruction, its true description must be something else. This not only removes teachers as possible authors of evil, but supplies the outline of a negative criterion for evil: does coaching in the matter – to put it neutrally – count as genuine instruction? If so, it cannot be evil. This outline remains empty until grounds are supplied for telling instances of genuine instruction from counterfeits. Augustine's reflections on pedagogy – the principal text would be De magistro, contemporaneous with lib. arb. – are not by and large directed to this question, but the conception of learning he elaborates in them, on which learning includes not just knowing but also rational comprehension, supplies the raw materials for an answer.

We need enter no further on Augustine's theory of learning. The point is simply that Augustine's treatment of law has a parallel structure. ‘Law’ [lex] is a success-term: ‘a law that is not just does not seem to me to be a law [Nam mihi lex esse non videtur, quae iusta non fuerit].’Footnote ¹¹ Again this supplies the outline of a criterion of the good: something is good if it is mandated by a law. The question is what counts as a law. By a consideration of various cases, including some of apparent inconsistency – a people should govern itself, but a benevolent dictator should put order as needed – Augustine and Evodius arrive at an application to laws of the temporal/eternal distinction: ‘let us call a law temporal if, although it is just, it can justly be changed in the course of time’, while a law is eternal in so far as ‘it should always be obeyed; through it good people deserve a happy life and evil people an unhappy one; and finally through it temporal law is both rightly enacted and rightly changed’.Footnote ¹² All laws are eternal to some degree, otherwise they would not count as laws. For example, a ‘law’ remaining in effect only while agents comply with it does not count as a law.Footnote ¹³ That law is most eternal ‘according to which it is just for all things to be completely in order’.Footnote ¹⁴ Thus the question of the content of the legal criterion for goodness is reduced to the question ‘What counts as order?’. We will return to this question later; here, we simply note that, for Augustine, something is ordered to the degree that it exists.

Now to the conditions. We must, according to (1), distinguish among blameworthy desires as to their severity; according to (2), distinguish, as it were, bad blameworthy desires from good blameworthy desires; according to (3), furnish plausible grounds for judgments of good and evil; according to (4), identify, for any given good blameworthy desire, a blameless desire that conduces to the same practical end; and, according to (5), remain faithful to Augustine's text. The conditions can be met. Consider your blameworthy desire for the well-being of the drowning child and a law to the effect that, say, strangers in peril should be helped.Footnote ¹⁵ Willing the law's fulfilment conduces to the same practical end as blameworthy desire for the child's physical well-being. Thus we meet the condition in (4). In turn, compliance with the law distinguishes your desire to save the child's life from a morbid desire to, say, witness a drowning, satisfying the condition in (2). As to (3), we required plausible grounds for judgements of good and evil, and now love of the law, implying willing submission to its obligations, furnishes those grounds, provided, as Augustine does provide, that laws are just.

The condition in (1) requires more careful handling. Augustine expressly recognizes a ranking of evils:

The question is whether rational grounds are available for the ranking. For present purposes, it suffices to observe that the temporality of a law mirrors the severity of the evil it proscribes: where two laws conflict, the less eternal will give.

The beginnings of a solution to the puzzle in (1) are now visible. Suppose that we have a law L prohibiting a and a law L′ prohibiting a′, and suppose further that L is more eternal than L′. Nothing stands in the way of our counting desire for a as more blameworthy than desire for a′. Admittedly, not all laws are directly comparable, or if they are, it is not always clear that one is more eternal than the other. For example, is a law that prohibits larceny more eternal than a law that prohibits fraud? Perhaps we may lay it down that in such a case the laws are equally eternal. Another provision is required to compare laws formulated positively (mandates) and laws formulated negatively (prohibitions). Assuming such concerns can be met, however, it seems that we have at least begun to meet the condition in (1): in brief, a desire is more blameworthy the more eternal the law that we violate in acting on it.

We have met the conditions in (1)–(4), at least to a first approximation. Moreover, we have introduced nothing essential from outside of Augustine's text, and so we have met the condition in (5), too.

Eternal law and the categorical imperative

We have seen that omitting Augustine's notion of law from his account of moral agency yields several puzzles, and that restoring his notion of law solves them. The notion of law that turns this trick is the one that long ago struck R. J. O'Connell as Kantian, and its Kantian aspect is no accident.Footnote ¹⁷ In this section, we see how deep the resemblance between Augustine and Kant goes.

Augustine recognizes, on one side, our absolute ethical responsibility – bad luck is never an excuse – and, on the other side, the brute fact of our limited power. He steers a course between them by admitting the will's complete dominion over a restricted domain. A child might drown despite one's willing the contrary, but there can be no such thing as failing against one's will to will the fulfilment of the law.Footnote ¹⁸ So a good will is eternal and, indeed, the only good of which humans are capable (the case is different with God, for whom there is never a gap between willing and attaining what is willed). Precisely this conception of human will undergirds Kant's ethics, on which, famously, only a good will is unqualifiedly good. Again, this similarity is no coincidence: Kant's conception of free will rides between the same constraints as Augustine's, namely, moral responsibility, on one side, and impinging circumstance, on the other.

If the notion of law is to mediate between moral responsibility and impinging circumstance, the notion must be, so to say, flexible on the side of circumstances and rigid on the side of the will. By this I mean that it must be possible for an agent to count as law-loving under a wide variety of circumstances and in view of any number of possible actions taken by the agent. Augustine and Kant achieve this two-sidedness in part by conceiving laws hierarchically, according to the range of circumstances they comprehend. So, for example, a law-loving agent might be released from the obligation to prevent a theft provided there is a sufficient risk of injury in acting to prevent it. For Augustine, release would proceed in accordance with the lex aeterna, by which ‘it is just for all things to be completely in order [ordinatissima]’.Footnote ¹⁹ According to (one formulation of) Kant's categorical imperative, which governs the activation and deactivation of maxims in light of contingent circumstance, one should ‘act always in accordance with that maxim whose universality as law you can at the same time will’, for this is ‘the single condition under which a will can never be in conflict with itself’.Footnote ²⁰ In both cases, a condition on lawfulness, preserved under all circumstances, is raised to the status of a super-law regulating lower-level laws. Augustine's condition is order, Kant's is consistency.Footnote ²¹

We have seen that, for both Augustine and Kant, good faith failure is never grounds for blame, rather blameworthiness depends entirely on the will's disposition. We have seen, further, that the ultimate test of good will is not the same for Kant as for Augustine. Augustine asks whether one sincerely desires the maximization of order, while Kant asks whether what one desires is consistent with the idea of a universal law mandating it.

A.-H. Chroust once imputed to Augustine ‘a formalistic and abstract definition of the nature of right and justice’.Footnote ²² Consideration of two further differences between Augustine's and Kant's conceptions of the law will show, however, that while Chroust's imputation might be true of Kant, it is certainly false of Augustine.

First difference: ethical bivalence

In elaborating the example of the drowning child, we assumed without further ado several non-trivial ethical principles. The first such principle is that anything morally evaluable at all must have exactly one of two values, namely, good or evil. We took this principle for granted when we assumed that in failing to do good, you did evil. In semantics, the principle of bivalence holds that every proposition has exactly one of two truth-values, true or false; let us call the corresponding ethical principle ethical bivalence. Whatever might be said for or against ethical bivalence, it recurs throughout Augustine's corpus, if only implicitly, and to my knowledge he never evinces any interest in relaxing it.Footnote ²³

Now to understand the role played by ethical bivalence in Augustine's account of moral agency, we must see how ethical bivalence works together with a second principle unstated in the case of the drowning child. According to this principle, we always have it in our power to do the right thing. Only given this principle does it follow from the fact that your failure to save the child was in good faith that saving the child was not the right thing to do. This principle, call it P, is actually entailed by Augustine's basic definitions: after all, if the good were beyond one's reach then, absurdly, desire for it would be temporal and, so, evil. Note that the contrary of P, call it P′, according to which it is always in our power to avoid evil, also follows from Augustine's definitions: if some evil turned out to be unavoidable then, absurdly, its avoidance would be temporal and, so, evil. From P′ it follows that we are blameworthy only if we are responsible.Footnote ²⁴ By ethical bivalence, anything that is ethically evaluable is either good or evil; by P, the good is always within reach.

We must now consider the relation between ethical bivalence and principle P. Principle P involves an uncertain level of modal commitment: if we say that it is always in our power to do the right thing, it may be asked what we mean by ‘in our power’. Is it always logically possible? Physically possible? Possible given the particular circumstances? Or is it, perhaps, attainable on merely willing it? (The same holds, mutatis mutandis, for P′: we must ask in what sense it is always in our power to avoid evil.) The scope for right action in a given case will depend, in part, on the degree of modal commitment in P. Recall your good faith failure to save the drowning child. If we both interpret P liberally, so that saving the child might count as the good it was in your power to do, and maintain ethical bivalence, it is hard to see how we can avoid charging you with evil for failing to save the child. If, on the other hand, we relax ethical bivalence or interpret P more narrowly, the charge can be avoided. When Augustine trims the domain of the ethically evaluable so as to exclude such items as your failing to save the child, he is, more exactly stated, maintaining ethical bivalence while interpreting P narrowly.

But that is not quite the whole story. Augustine starts, as I have suggested, by admitting into the domain of the ethically evaluable all willings and assigning the value evil to every willing of something temporal. He thus undoubtedly advances what Mann has memorably called an ‘inner-life ethics’, that is, a view on which ‘moral appraisal should focus primarily – perhaps exclusively – on the agent's mental states themselves’.Footnote ²⁵ Having settled the core of the domain of the ethically evaluable, however, Augustine extends that domain to include the will to attain lawfully mandated temporal aims, then again to include those temporal aims themselves (the physical well-being of a child, for example), and finally to include the instruments necessary to attain those temporal aims – provided, always, that the aims and instruments are desired not for their own sake. Law leverages this domain-extension at every stage, by distinguishing the blameworthy willing of temporal aims as ends in their own right from the morally acceptable willing of temporal aims for the sake of a law's fulfilment.Footnote ²⁶ We may say, therefore, that for Augustine temporal aims are in fact morally evaluable, but only derivatively.

With Augustine's account of moral agency now clearly in view, we turn to Kant. Any statement of Kant's position on these matters must be, as so often with him, the tip of an iceberg. However, it can be said quite definitely that Kant accepts three categories of action, namely, prohibited actions, required actions, and indifferent actions (adiaphora), and hence that Kant rejects ethical bivalence.Footnote ²⁷ Once ethical bivalence is rejected, it becomes possible, of course, to acquit you of committing evil in failing to save the drowning child while maintaining that saving the child was the good it was in your power to do, for it becomes available to say that failing to save him was merely indifferent. Kant, however, declines to count saving the child as the good it was in your power to do, instead maintaining, with Augustine, that whatever good it was in your power to do, it was in your power to do by merely willing it.

In sum, both Augustine and Kant maintain P, but only Augustine affirms the intricately related principle of ethical bivalence.

Second difference: grounds of law

As mentioned earlier, Augustine holds that all genuine laws are just. But how are we to tell genuine laws from counterfeits, or, as we might be inclined to call them, ‘bad laws’? It cannot be according to whether the ends they mandate are good, given that we are making out goods as ends mandated by law.

To set out the problem clearly, suppose that a ‘law’ were passed mandating that blue-eyed children be drowned. We might like to argue that this ‘law’ is unjust, hence a counterfeit, on the grounds that it is evil; but we have specified that the grounds for judging something evil are proscription by law, in fact, that an evil is worse the more eternal the law that proscribes it. If there are no grounds for judging something good or evil independently of the law, there is little we can bring against this abject ‘law’ beyond brute insistence that it is a counterfeit.

Augustine does, however, offer independent grounds for ruling out this ‘law’, when he identifies goodness with being and being with order.Footnote ²⁸ This twofold identification makes available a conception of law as an existence-promoting norm. If evil is the privation of being, drowning blue-eyed children represents an obvious and great evil, for rather than promote existence it reduces it. Indeed, it reduces being of a particularly high order, for Augustine takes it, on scriptural grounds, that human beings are the image of God. This answers the further question, Why can we not freely will evil, or again, why is evil itself not eternal? Augustine's answer, detailed after lib. arb., most forcefully in Confessions, is that an evil will undoes itself, as it declines into habit and, eventually, degeneration so complete that only divine grace can restore it.

We have asked what test Augustine offers to distinguish genuine laws from counterfeits, and we have answered that among principles of action, laws and only laws – that is, just and only just principles of action – promote order, hence being, hence the good. With Kant, the question must be different. He already gives us a partial test for morally binding principles of action in the categorical imperative: if a principle of action is moral, it can be consistently universalized. Augustine would readily agree to this; after all, a principle of action that would obligate one to both follow it and also not follow it hardly merits recognition as a law. But how do we tell morally binding universalizable principles of action from universalizable principles of action that are not morally binding? Kant holds that no further test is needed: consistent universalizability in a principle of action is not only necessary for its morality but also sufficient.Footnote ²⁹ Whether Kant is right in holding that consistent universalizability is sufficient for morality is a deep question; for our purposes, it is enough to observe that there is no obvious reason why there could not be a consistently universalizable but immoral principle of action, for example, one for whose sake a person might passively observe a drowning child.

It is now clear why Chroust rightly could have attributed a ‘formalistic and abstract definition of the nature of right and justice’ to Kant but not to Augustine. The difference between them on this score is, in fact, deep, and not without consequences for their other commitments. I conclude this section by considering one such consequence.

Why should an agent want to be good? Augustine has a ready answer: to be good is simply to be, and surely being is preferable to non-being. This answer turns on the interpretation of evil as the privation of being, a non-obvious interpretation, to be sure, but one that I will suppose can be carried through. If the question is pressed why being should be loved, Augustine will appeal, as he does especially in his later writings, to the divine grace whereby humans come to know love at all.Footnote ³⁰ Kant, for his part, roots moral agency so firmly in pure reason that the question of agent interest emerges only as a devilishly slippery loose end. Here is Kant taking up the thread at the end of the Groundwork:

Despite its appearance as a loose end, Kant allows that the problem of ‘moral feeling’ brings us to ‘the supreme boundary of all moral inquiry’, a horizon at which ‘all knowledge has an end’ but on the other side of which pure reason must be arousing ardour by some means akin to causation.Footnote ³² Augustine, too, recognizes a sharp distinction between the hidden and the visible: ‘From a depth which you cannot see rises everything that you can see.’Footnote ³³ But what we do not see is God, to whom we must credit inter alia our affective motives. It is no part of Augustine's thought that the claims of morality derive ultimately from pure reason, much less that those claims are self-legislated. He thus leaves himself an appeal that Kant does not have.