Moral Responsibility and Determinism: An Analysis of Frankfurt-style Cases

 

Introduction

 

In this essay, I intend to examine whether the concept of moral responsibility is compatible with the truth of determinism. In particular, I will look at the so-called principle of alternative possibilities (which I will refer to as ‘PAP’) and argue that, despite recent attacks, the principle is indeed necessary for moral responsibility, as it is commonly understood, to apply. I will further argue that if determinism is true, then alternative possibilities are ruled out, and that therefore moral responsibility does not apply. I will then finally suggest that moral responsibility is an intuition that is psychologically too deeply ingrained in us simply to ditch as a result of any philosophical argument and suggest that if this is so, a re-examination of what it is to be morally responsible is called for.

 

The structure of the essay is as follows. I will briefly set out some non-exhaustive criteria identifying what might be required for an agent to be morally responsible, a definition of determinism, and say why the two concepts seem to be incompatible. I will then look at how certain of the elements required for moral responsibility have been undermined by what have become known as Frankfurt-style cases. I will then look at various responses to Frankfurt and suggest that the structure of these cases is such that they do not, as they appear to at first blush, show that PAP is wrong. In particular, I will suggest that they rely on constructing a dichotomy between ‘actual sequence control’ and ‘regulative control’. I will suggest that in fact actual sequence control reduces to regulative control. Since the latter relies on alternative possibilities it is ruled out by determinism and therefore so is moral responsibility. Finally, I will attempt to take a step back and suggest that conflicting intuitions are at the root of the difficulties in this area and suggest that there are ways of resolving these conflicts.

 

 

 

Moral responsibility and causal determinism

 

There are various different accounts of what it is to be morally responsible and what requirements are necessary for moral responsibility to apply. One account suggests that we should only be held morally responsible for our actions insofar as it is fair to hold us responsible for them (Wallace 1994). Another suggests that for us to be morally responsible, our actions must truly belong to us; they must truly be ours.[1] Another approach is to use the Kantian thesis ‘ought’ implies ‘can’, and its negation, ‘ought not’ implies ‘is able to refrain from’ (Kant 1997: 30). These various accounts appear to rely on the idea that in order to hold ourselves morally responsible there are two necessary elements: that it is we who freely choose our actions and that in doing so we are able to select from various alternative courses of action.

 

It seems then that one requirement for an agent to be morally responsible for an action is that he is able to do other than what he actually does. This is the principle of alternative possibilities or ‘PAP’ to which I have referred. I would suggest that PAP seems intuitively to be true because we regularly employ it as an excuse to avoid responsibility. For example, we might say, ‘I’m sorry I was unable to make the meeting, but my hand was trapped under a rock which delayed me for five days until I was forced to saw it off’. The agent’s inability to attend the meeting would appear to be a good excuse for his actually failing to attend it. We may want to suggest that he be held responsible for his decision to put himself in a situation in which he might be delayed. However, once in that situation, we would not want to hold him responsible for his inability to attend.

 

Having stated what PAP is and why it seems, at least at first glance, to be a prerequisite for moral responsibility, we now need to look briefly at determinism to understand why a problem might arise. The notion of causal determinism can be stated as follows: every action or state of affairs in the universe is causally necessitated by the laws of nature and an earlier state of the universe (and if the universe has a beginning, by the initial condition of the universe).[2] If determinism is true then, the notion of moral responsibility obtaining by virtue of PAP would appear to be undermined. As Fischer (1999: 100) puts it:

 

‘Consider some choice of mine, C, which I actually make at time T2. If causal determinism is true, then the total state of the world at T1, together with the laws of nature, entail that I make C at T2. Thus, in order for me to have made a different choice at T2, it must have been the case that the state of the world at T1 was different from the way it actually was. And I cannot at any time so act (or choose) that some actual law of nature would not have obtained. That is, the past and the natural laws are ‘fixed’ in the relevant sense. Thus it seems that I could not have chosen otherwise at T2.’

 

It would appear then that causal determinism undermines our concept of moral responsibility as it is commonly understood. One response to this is to argue that since we all know that we are obviously morally responsible beings, and since we do require the ability to do other than we actually do for moral responsibility to obtain, then causal determinism must be false. This, it seems to me, is putting the cart before the horse and reasoning from conclusion back to premises. It is also undermined by the fact that we do use the inability to do otherwise as an excuse in cases of what Dennett ( 1984: 554) calls local fatalism, such as the example I have used above concerning the rock climber with the trapped hand. There seems to be no distinction between this type of local fatalism, the claim that I was unable to do otherwise because my hand was trapped, and the global claim, that I was unable to do otherwise because the laws of nature make it so, since any case of local fatalism will be necessitated by the global fatalism of determinism.

 

However, whether you favour this approach or not may well depend on which intuition you take as more basic: the belief that we are indeed morally responsible beings or the belief that determinism is at least a possibility. If we accept that determinism is a possibility then it seems we have to accept the possibility of ruling out moral responsibility.

 

Frankfurt-style cases

 

Daniel Dennet (1984: 553) suggests that, ‘wherever progress is stalled on a philosophical problem, a tactic worth trying is to find some shared (and hence largely unexamined) assumption and deny it.’ This is precisely what Harry Frankfurt did in 1969, by proposing the following thought experiment, which, he suggests, undermines the notion that alternative possibilities are required for moral responsibility to obtain:

 

‘Suppose someone – Black, let us say – wants Jones to perform a certain action. Black is prepared to go to considerable lengths to get his way, but he prefers to avoid showing his hand unnecessarily. So he waits until Jones is about to make up his mind what to do, and he does nothing unless it is clear to him (Black is an excellent judge of such things) that Jones is going to decide to do something other than what he wants him to do. If it does become clear that Jones is going to decide to do something else, Black takes effective steps to ensure that Jones decides to do, and that he does do, what he wants him to do. Whatever Jones’s initial preferences, Black will have his way.’(1969: 835) Frankfurt suggests that this mechanism could be anything from a threat, hypnosis, or a manipulation of the processes within Jones’ brain.

 

The crucial point in the argument is that ‘Black never has to show his hand because Jones, for reasons of his own, decides to perform and does perform the very action that Black wants him to perform.’(1969: 835)  As Frankfurt puts it, ‘it seems clear, Jones will bear precisely the same moral responsibility for what he does as he would have borne if Black had not been ready to take steps to ensure that he do it.’ In other words, if you remove Black from the scene entirely, Jones appears to bear exactly the same moral responsibility as if Black is present but fails to act. Yet it would appear that in the instant case, Jones is not able to do otherwise. Whatever happens, Black will get his way and so, the notion that the ability to do otherwise as a prerequisite for moral responsibility is undermined.

 

The literature on moral responsibility contains many Frankfurt-type examples, usually with the Black character determined that Jones will perform some heinous act such as assassinating a mayor, choosing to refrain from saving a drowning boy or even, voting for George W Bush in preference to Al Gore. For the sake of clarity, I will refer to the actor, Jones, as the ‘agent’, and Black, the eminence grise, as the ‘counterfactual intervener’. These examples all adopt the following general form, ‘a person is in the process of deciding which of n alternative acts A1, . . . . ,Ak, . . . . , An to perform. He believes correctly that he cannot avoid performing some one of the acts, although there is no one act of which he believes that he can’t avoid performing it. He decides to perform and, acting on this decision, does perform Ak. But, unknown to him, there were various factors that would have prevented from performing (and also from deciding to perform) and of A1, . . . . , An, except Ak, these factors would have come into play if he had shown any tendency towards performing (or deciding to perform) any of the alternatives except Ak.’ (Van Ingwagen 1978: 202).

 

Frankfurt wasn’t the first philosopher to suggest such an example. John Locke proposed what might be called a blockage case: that of a man who wakes up in a locked room, not knowing it to be locked, but who voluntarily chooses to stay in it anyway. This example shows that the counterfactual intervener need not be a human actor. It can be state of affairs. A more modern alternative might be to suggest that certain pathways within the agent’s brain are blocked ruling out various alternative choices, yet the agent freely chooses to follow the only remaining path. However, despite the apparent differences, these examples still conform to the general structure outlined above.

 

On Frankfurt’s own admission, cases of this sort do not of themselves render incompatible the notion of determinism and moral responsibility. He suggests that the agent would not be responsible if he acts only because he is unable to do otherwise, that is his inability to do otherwise causes him to act in the actual sequence, as opposed to the alternative sequence. If, however, he acts for other reasons, acting on his wants or desires, then he is morally responsible.

 

Frankfurt’s approach has inspired a number of copycat examples, some of which I will outline during the remainder of this essay as necessary, but which I fear blur the picture somewhat by seeking to make ever finer distinctions.

 

The important points concerning Frankfurt-type examples are that they appear, intuitively at least, to show that the agent is morally responsible for his actions, despite his inability to do otherwise; further, that the agent apparently freely chooses his action and, finally, that the removal of the counterfactual intervener makes no difference to what actually happens. This last point is very important because, with the counterfactual intervener removed, the agent clearly has the possibility of doing otherwise and therefore would be morally responsible for his actions. Since in the Frankfurt-style cases the counterfactual intervener does not act, his removal makes no difference to the causal chain, and therefore the agent appears to bear the same moral responsibility.

 

One problem with Frankfurt-style cases

 

However, this is unsurprisingly not the end of the story. A crucial question, which enables us to distinguish between those cases where the counterfactual intervener is present and does not act, and where he is completely absent, arises in Frankfurt-style cases, and it is this: how does the counterfactual intervener know under what circumstances he is to intervene in the actions of the agent? In the case of Jones choosing whether to vote for Bush or Gore, how does Black know when he should activate his brain-controlling device? Clearly, if he is going to use it at all, he must activate it before the act of voting takes place, even if the interval between activating it and the act of voting is infinitesimally small. But in order for him to do so, he must have some sort of reliable indicator that Jones is going to vote for Gore rather than Bush: a “prior sign”.[3] Let us assume that Black has a very sophisticated computer that is able accurately to model Jones’ brain and which indicates when Jones is going to vote for Gore due to the firing of a particular sequence of neurons within the brain. (In fact, this indication need not be anything like as sophisticated. It could equally well be that Jones gives off an involuntary facial tick when he is going to vote for Gore). In any case, whenever this particular arrangement of neurons (or facial tick) occurs, Black forms the belief that Jones is going to vote for Gore, and activates his brain-controlling device accordingly, forcing Jones to vote for Bush.

 

This requirement of a prior sign raises a problem for the defender of Frankfurt-style cases, which appears to me to be insurmountable.[4] The problem is this: what is the status of the path between the prior sign and the act of voting? Are the two events linked deterministically or indeterministically? A similar problem, incidentally, arises with respect to the reliability of the brain-controlling device: is it completely reliable, or is it capable of on occasion failing to control Jones when activated? These two situations are sufficiently similar, that I will examine just the first: the link between the prior sign and the act of voting.

 

Let us suppose firstly, that the prior sign is indeterministically linked to the act of voting for Gore. What this means is that sometimes Jones will give off the prior sign indicating that he is going to vote for Gore and yet choose to vote for Bush. But, because he has given off the sign, Black will activate his brain-controlling device, forcing him to vote for Bush anyway. If so, it would appear to be the machine, and not Jones, that acts. In this sequence Jones is not morally responsible for voting for Bush because he is not the actor. Alternatively, Jones could choose to vote for Gore, but fail to give off the prior sign that he is going to do so. In this case, he will successfully vote for Gore due to the absence of the trigger that would have caused Black to activate his device. Black will fail to get what he wants and Jones has genuine alternative possibilities. Either way, Jones is not morally responsible for voting for Bush. He is either forced to do so, or he manages to avoid doing so.

 

One move then taken by the supporter of Frankfurt-style cases, is to propose examples that narrow down the agent’s freedom to do otherwise. In the example we have used, the prior sign was a particular firing of neurons or a facial tick. We can go back in time ever further narrowing the difference between what happens in the actual sequence and the alternative sequence. We might, for example, examine what it is that causes the difference between the occurrence of the facial tick and its failure to occur and find a differing arrangement of neurons or electrical impulses in the brain. We might then ask what causes these to be different. If we carry on going back in time, we should eventually find ourselves at what we might call a basic distinction – a point at which we cannot go back in time any further and find a distinction between what happens in the actual sequence and what happens in the alternative sequence. It might something utterly insignificant, for example, the orbit of one electron in one atom in the brain being ever so slightly different. Now we can run time forward again, having found the absolutely smallest sign that we can, and use this difference as the sign that triggers the brain-controlling device. Remembering that in this example, we have concluded that Jones is morally responsible (due to the indeterminacy of the link between the sign and the act of voting), the defender of Frankfurt-style cases suggests that PAP is not credible in this case, because although admittedly Jones has the freedom to do otherwise, what he actually possesses is a mere ‘flicker of freedom’.[5] That is, his freedom is based on something that is too small, too unimportant, to ground a concept as rich and important as moral responsibility. In this case, the argument goes, it is not credible to base moral responsibility on the fact that one electron is in a slightly different position from where it might have been.

 

However, what exactly is being advocated here? An obvious sign, such as Jones’s facial tick, does appear to be sufficient to ground moral responsibility. All we have done is to identify the first cause of that facial tick. The Frankfurt defender has gone further back along that path until he has found the very first point of divergence between the actual and alternative sequence and used the fact that at this point of divergence the difference between what happens in the two difference sequences is infinitesimally small as a basis denying that this difference is sufficient to ground moral responsibility. I would suggest, however, that it is necessarily the case that if one traces the difference between the two alternative sequences back to its starting point it will be extremely (possibly infinitesimally) small. This does not mean that it is incapable of being a ground for greater things. Since we are talking about divergent paths, we might use an actual pathway along the ground as an analogy. Suppose that a pathway leaves Durham and forks into two, one fork going to London, the other to Edinburgh. The actual point of divergence can be arbitrarily small, but this point, however small, is absolutely necessary for the two paths to come into being from the one and to reach their different destinations. In other words, size doesn’t matter! If so, then the indeterminacy of the link between the prior sign and the relevant act successfully undermines the Frankfurt cases.

 

What, then, if we suppose that the prior sign is deterministically linked to Jones’ act of voting for Gore? That is, it is necessarily the case that every time Jones chooses to vote for Gore, the requisite sequence of neurons does indeed fire (or he does give off the requisite facial tick) and Black intervenes. If Jones chooses to vote for Bush, he gives off no sign and Black fails to intervene. We now have a case where determinism applies throughout the example, which eliminates the problems that we have identified previously, but what we don’t have is an argument that Jones is morally responsible for his act of voting for Gore. Since the question we are ultimately addressing is whether Jones is responsible for his actions if they are determined, merely to assert that he is indeed morally responsible is to beg the question at issue. In order to assert that Jones is morally responsible in this case, the Frankfurt-defender needs to give an account of why Jones is morally responsible in the actual sequence.

 

Actual sequence and regulative control

 

Fischer attempts to do precisely this by examining the notion of control (1991: 266-270). He suggests that there are two types of control that we might have over an event: ‘actual sequence’ control and ‘alternative sequence’ or ‘regulative’ control. In the examples we have looked at so far, it is suggested that Jones has actual sequence control, while Black has alternative sequence control.

 

To illustrate this distinction, we are asked to picture a car occupied by Al, a learner driver and Bart, his instructor. The car is one of those cars that possess dual controls so that the instructor can take over in an emergency. Assuming that nothing goes wrong, Al steers the car, changes gear, pushes the brake and accelerator pedals and so on. If he turns the steering wheel to the left, the car will turn to the left. Al, then, has control in the ‘actual sequence’. Bart, on the other hand is able to override Al’s control if necessary. If it looks like Al is going to turn to the right rather than to the left, Bart is able to take control. Bart then has regulative control over the car. It would appear then that in this scenario, it is possible for there to be two types of control and that they can reside in different people: Al that has actual sequence control while Bart has regulative control.

 

However, I would suggest that Fischer is actually mistaken in this case and that he is guilty merely of restating the basic Frankfurt dilemma. This is because Al’s control doesn’t actually amount to very much. If we consider simply the manipulation of the car’s direction rather than the other factors that a driver has control over, and further suppose that Al decides he wants to drive to the local bar, while Bart is determined that they will go to the Kwik-e-Mart, only one of them can be assured of achieving his desired result and that is Bart. In other words, Bart is fully able to ensure the success or failure of reaching any particular destination. Any control that Al has, he has merely by virtue of the fact that Bart will allow him leeway to turn the steering wheel in one of a number of different directions consistent with reaching the Kwik-e-Mart. I would suggest therefore that any putative example of actual sequence control in fact reduces to alternative possibilities control. Alternative possibilities control is able to guarantee outcome P or not-P where P is any desired outcome. If that is the case, what role remains for actual sequence control?

 

This suggestion may seem to run counter to our intuitions. Surely as learner drivers we do have some sort of control over the car’s direction? I would reiterate that any control we do have is by virtue of alternative sequence control. This dichotomy is well illustrated by the following passage,

 

‘when I was four years old and learning to ride a bicycle, I reached a point where my father decided I no longer needed training wheels. But he still worried that I might fall. So on my first attempt “without a net,” he ran alongside as I pedalled. His arms encircled without touching me, his hands resting lightly upon me, but not holding me upright. I rode straight ahead. My rather did not push or guide me, but if I had faltered or veered suddenly to the side, he would have tightened his grip, keeping me vertical and on track. After finally braking to a stop, I was jubilant but somewhat hesitant over whether I should be. I wondered, had I really ridden my bike on my own?… was the triumph of riding straight down the street mine or not?’ (Wyma 1997: 57)

 

In this case, I would suggest that the more leeway the father allows the son, the more we can say that the son is truly responsible for riding the bike. By contrast, if the father encircles his arms ever more closely around the son, the son’s responsibility is reduced and tends towards zero the closer the father gets. If so, then it is apparent that the success or failure of the son to ride the bike rests on his ability to do otherwise. This supports the defender of PAP, and its incompatibility with determinism, as under determinism the alternatives also tend towards (or simply are) zero.

 

Implications for freedom and moral responsibility

 

If I am correct then, the lack of alternative possibilities does appear to pose severe problems because it threatens our very ability to be the author of our own actions. Our intuitive notions of freedom, perhaps even our sense of self, seem to require control, choice, or the ability to make decisions. All of these concepts I would suggest are inextricably linked with the notion of alternative possibility. A choice with only one option is not a choice, as we understand it. Similarly, the notion of a decision requires the ability to choose one path of behaviour over another. It would, I suggest, be meaningless to talk of a taking a decision without alternatives being available.

 

If this is the case then, assuming the truth of determinism, we have two intuitions that are in conflict. First, that we are in some sense free morally responsible beings who are the authors of our own actions, and second that we require alternative possibility control for this freedom, authorship, and hence moral responsibility to exist. If we rule out alternative possibilities then it seems we have to discard our notions of freedom and moral responsibility.

 

As I have stated earlier, certain philosophers, have used this as an argument that determinism is false, instead taking it as basic that we are the authors of our own actions and thereby morally responsible. However, I would suggest that this intuition is open to challenge. In a recent study a group of subjects was asked to perform a simple action – that of raising their finger three times at a random time of their own choosing (Ramachandran: 2003). They were wired up to a computer, which measured electrical potential in their brains. The interesting thing is that about a second before they willed the action of raising their finger, the electrical potentials in their brain changed and the computer lit up, suggesting that it is not our wills that are the authors of our actions, but some other process occurring within our brains which is prior both to the action and the act of willing the action.

 

Apparently, another experiment is now being conducted in which the test subjects are able to see the computer. In other words, they will be able to see it lighting up before they will the action that causes it to do so (I use the words causes here very loosely indeed). Now suppose that the computer could measure changes within the brain with such extreme accuracy that it was able to predict not just the action of raising our fingers, but all of our bodily actions before we consciously chose to perform them. It would tell us what when were about to speak, what we were about to say, when we were about to blink, whatever. Would we still retain the notion that our actions were free? Would we still retain this intuition in the face of the evidence?

 

I can only suggest an answer, but my suggestion is that we probably would. We would still cling, uneasily perhaps, to the notion that despite whatever the computer tells us, there is a sense in which we could have done otherwise and we would still retain our notion that we are morally responsible beings.

 

Strawson (1962) suggests that talk of moral responsibility is really talk about our reactive attitudes to one another; that is, how we as social beings react to others around us. If so, then there does not appear to be any compelling reason for moral responsibility to be discarded. These reactive attitudes do not appear to require either freedom or alternative possibilities and so are not threatened by determinism. However, it does appear to threaten the notion that moral responsibility is connected to fairness. I cannot see how we could retain the notion that it is fair to hold agent morally responsible for their actions in a deterministic universe, as the concept of fairness, I suggest, does require genuine control - the ability to avoid a certain course of action or outcome. However, our actual moral practices are equivocal when it comes to the importance on fairness in this sense. For instance, we could arguably say that all drivers who fall asleep at the wheel are equally at fault, but we in fact only punish those who cause accidents as a result of falling asleep. Those who fail to cause accidents are simply lucky, but we do not punish them.

 

Conclusion

In this essay, I have argued against Frankfurt that our intuitive notions of moral responsibility do, in fact, require alternative possibilities control, and further that determinism rules out such control. I have further suggested that even if it were shown that determinism were true, we would still retain our notions of moral responsibility and freedom. The conclusion that I have drawn from this is that our intuition that moral responsibility requires freedom in the alternative possibilities sense may have to be discarded. Doing so would still allow us to retain the concept of moral responsibility in a Strawsonian sense. The fact that we may well retain our feeling that we are free agents and originators of our actions, in spite of evidence to the contrary, need not be problematic. There are many cases where we hold on to certain basic beliefs despite a lack of evidence that such beliefs are true, such as beliefs, concerning the existence of an external world, the continuing existence of the world and the existence of other minds.[6] Beliefs such as these may be necessary for us to function as human beings irrespective of their truth and the notions that we are free, morally responsible beings may be another belief that falls into this category.

 

Gary Lorrison

Department of Philosophy

University of Durham

garylorrison@aol.com

 

References

Dennett, Daniel C. 1984. I Could not have Done Otherwise - - So What? The Journal of Philosophy, Volume 81, Issue 10: 553 – 565.

Frankfurt, Harry G. 1969. Alternate Possibilities and Moral Responsibility. The Journal of Philosophy, Volume 66, Issue 23: 829-839

Fischer, John and Ravizza, Mark. 1991. Responsibility and Inevitability. Ethics, Volume 101, Issue 2: 258-278.

Fischer, John.1999. Recent Work on Moral Responsibility. Ethics, Vol. 110, Issue 1: 93 – 139

Ginet, Carl. 1996. In Defence of the Principle of Alternative Possibilities: Why I don’t Find Frankfurt’s Argument Convincing. Philosophical Perspectives 10: 403-417

Kane, Robert. 1996. The Significance of Free Will. New York: Oxford University Press.

Kant, Immanuel. 1997. A Critique of Practical Reasoning (tr. and ed. by Mary Gregor). Cambridge University Press.

Locke, John. 1975. An Essay Concerning Human Understanding (1690). Oxford Clarendon Press.

Pereboom, Derek. 1999. Living Without Free Will. Burlington, Vt.: University of Vermont, Philosophy Department

Plantinga, Alvin. 1981. Is Belief in God Properly Basic? Nous. 1981: 41-51.

Ramachandran, Vilayanur. 2003. The Emerging Mind. Reith lectures. www.bbc.co.uk/radio4/reith2003

Strawson, P.F. 1962. Freedom and Resentment. Proceedings of the British Academy, volume 48: 187 – 211. London: Oxford University Press

Van Ingwagen, Peter 1978. Ability and Responsibility. Philosophical Review 87: 201-224

Wallace, R. Jay. 1994. Responsibility and the Moral Sentiments. Cambridge, Mass.: Harvard University Press.

Watson, G. 1996. Two Faces of Responsibility. Philosophical Topics 24: 227 – 248.

Widerker, David. 1995. Libertarianism and Frankfurt’s Attack on the Principle of Alternate Possibilities. Philosophical Review  104: 247-261

Wyma, Keith. 1997. Moral Responsibility and Leeway for Action. American Philosophical Quarterly 34: 57-70

 

---

[1] For example see Watson (1996) and Pereboom (1999)

[2] A number of commentators have suggested that our knowledge of the quantum world suggests that determinism is false. I do not intend to debate the truth or falsity of determinism in this essay. Such a project is beset with difficulty. However, I am going to suggest that the truth or falsity of determinism is still an open question. Using the language of possible worlds, determinism could be true if a possible world existed in which deterministic rules, however odd, produced the same results as are obtained when quantum mechanical measurements are made in the actual world. Determinism would be true if the putative possible world were in fact the actual world.

[3] I rely on primarily on Fischer (1999: 109-111) for this exegesis of ‘prior-sign’ type cases.

[4] These arguments were developed from, inter alia, Kane (1996), Widerker(1995), Ginet (1996) and Wyma (1997)

[5] Here I adopt Fischer’s (1999) terminology.

[6] See, for example, Plantinga 1981