Parfit's account of personal survival is a causal theory. He gives an umbrella
characterisation of causal theories of personal survival: they all say that x
survives as y iff x is psychologically connected and/or
continuous with y, and this connectedness or continuity involves the
right kinds of causal connections (Parfit 1984, pp. 283-7).
Causal theories may differ in their account of what constitutes "the right kinds" of causal connections. In particular, Parfit distinguishes three kinds of causal theories: those which say that survival requires "normal" causal connections; those which require "reliable" causal connections; and those which say that "anything goes", that any causal connections will do. Interestingly, Parfit adopts the third alternative, placing no restrictions on the type of cause that suffices: according to Parfit, the "right" cause is any cause
What makes Parfit's "anything goes" stance particularly noteworthy is that there are many causal theories around, for many things other than personal survival, and these other causal theories almost never say that "anything goes". Causal theories of memory, action, reference, knowledge, and other things have all been tried, and it is seldom plausible to say that any kind of causal connection does equally well: an unrestricted causal theory in these other domains would immediately elicit a surfeit of counterexamples. We ask, what makes Parfit so confident that causal theories of personal survival stand alone among causal theories, in that in this case alone we should allow that "anything goes"?
Consider Davidson's causal theory of intentional action. Putting Davidson's view rather crudely, a person's intentional actions are those actions that are appropriately explained by that person's reasons for so acting. These reasons are particular pairs of beliefs and desires, and they explain the actions of the person in question because they cause those actions - Davidson's is a causal theory. But in elaborating the theory, Davidson put constraints on the kinds of cause that count, and he was well advised to do so. If he hadn't, then his theory would be an easy target. For instance:
A climber might want to rid himself of the weight and danger of holding another man on a rope, and he might know that by loosening his hold on the rope he could rid himself of the weight and danger. This belief and want might so unnerve him as to cause him to lose his hold, and yet it might be the case that he never chose to loosen his hold, nor did he do it intentionally (Davidson 1980, p. 79).
We shall argue (§3) that there are counterexamples to Parfit's account of survival which are no less damaging. Then, Parfit's reasons for holding an unfettered causal theory will be examined (§4). These reasons are embodied in an argument from analogy. We will contend that the "analogous" case appealed to by Parfit does nothing to support his claim: if anything, it supports a reliabilist causal theory of survival. But first, we will set the scene by giving a brief account of Parfit's views on survival.
§2. The Theory.
Personal survival does not, Parfit thinks, require personal identity. That is to say, for you to survive you do not have to go on existing - there does not need to be a single person who exists at both the earlier and the later times. Parfit says we do not and should not care about whether there is just one such a person. Rather, what is required for your survival is that at least one person exists at a future time, and that this person should have states which stand in the "right kinds of causal relations" to your current states. Parfit thinks it is rational for you to care about whether there will be such a person, but it is not rational to worry about whether that person will be you.
In articulating his theory, Parfit refers to something called Relation R. This is the name he gives to whatever relation it is, between our present selves and some future person, which constitutes the survival to that future time of our present selves. He then goes on to explain his more specific theory about what this Relation R is like. Relation R holds between x and y provided that x is psychologically connected and/or continuous with y, and provided that the connectedness or continuity involve the right kind of causal connections (Parfit 1984, p. 262). Here is a quick explanation of the terms present in this characterization of Relation R: x is psychologically connected with y when there are direct psychological connections between x and y (ibid., 206). Direct psychological connections may be of various kinds. For example, a direct psychological connection holds between x and y if x has an apparent memory (properly caused) of a particular experience that y had, or if y performs an action (properly caused) that x intended to perform (Parfit 1971, pp. 15-7 and 20). If there are enough direct connections between x and y, then Parfit says that they are strongly connected. x is psychologically continuous with y when there are overlapping chains of strong connectedness between x and y (Parfit 1984, p. 206). In passing, we may note that since psychological continuity can be a branching relation, so may Relation R. Personal Identity is non- branching, and so, Parfit maintains that although it generally accompanies Relation R, personal identity is not what matters to us. What matters is survival not identity, and the holding of Relation R is both necessary and sufficient for survival.
§3. The Counterexample.
One of our friends, Nick, claims to remember having particular experiences that Aristotle had. For instance, Nick says he remembers how he felt upon learning of Plato's death, he remembers writing his early piece "On Ideas", he remembers teaching Alexander of Macedon, and so on. For various reasons, we come to think that Nick is not lying - he does sincerely report on various memory-like experiences which he has. But we wonder if he is just having imagination attacks rather than memories.
So we express our skepticism to Nick and ask him if he is sure that he is having genuine memories. We observe that memory presupposes personal survival, but Nick is unmoved. He says that he is Aristotle. We are bemused by Nick's claim, and give him the incredulous stare. However, Nick is determined to soften us up a little.
"Well", he says, "In the past, you have told me that you agree with Derek Parfit's account of personal survival. Is this still true?"
We nod our heads.
"In that case", he continues, "At least let me show you why, by your own reckoning, it might be the case that I am Aristotle."
He then invites us to attend to the following considerations:
It is possible that there are direct connections between Aristotle and Nick. For example, it does seem that Aristotle experienced some things that Nick seems to remember, and had certain beliefs and desires that also characterize Nick. In virtue of these matchings between Aristotle's and Nick's psychological states, let's suppose at least that there are apparent direct connections between Nick and Aristotle. Let's further suppose that there are enough of these ostensible connections so that, if they are genuine connections, then Relation R holds between Aristotle and Nick. This may seem an outlandish possibility, that Nick should have so many memory-experiences (or quasi-memory-experiences), so many beliefs and desires like Aristotle's, that he should perform so many actions which correspond to Aristotle's intentions, and so on. But though it's hard to believe, it's easy to imagine.
What would it take for these apparent connections to be genuine? For Parfit, every genuine psychological connection must be causally underwritten. Thus, to count as genuine connections, each of Aristotle's experiences that Nick seems to remember must be a cause of Nick's corresponding apparent memory. Furthermore, Aristotle's having a certain belief/desire that Nick also has must be a cause of Nick's also having that belief/desire. And so on. Furthermore, according to Parfit, what constitutes personal survival is merely the existence of appropriate matching between Nick and Aristotle, together with the existence of causal connections connecting Aristotle's psychological states with Nick's. There is no need for these connections to be "normal" or "reliable" - anything goes - and according to the story that Nick asks us to imagine, the connections between himself and Aristotle are mediated largely by the weather! According to the story, here is how his psychological states arose.
An important deliverance of twentieth century science is that there are many physical systems which are remarkably sensitive to initial conditions. Seemingly inconsequential differences in initial conditions between certain systems are grossly magnified over time. One such class of systems is the class of atmospheric systems. Consider a world, W, with the same natural laws as our own. Suppose that W at t is qualitatively identical with our world at t, except that it contains an extra moth which flaps its wings. The consequences of this event snowball, so that there is little resemblance between the weather patterns of W and the actual world at, for instance, t+20 years. Consider a tsunami that engulfs Japan at t+20 years in W. It is very tempting, by Mill's method of differences, to regard the moth's wing-flapping as a cause of the tsunami since the flapping plays an integral, albeit miniscule, role in bringing about the tsunami. Moreover, the moth's flapping links in law-like ways with later differences such as the tsunami. So, it is plausible to cite the flapping as being among the many events which are causes of the tsunami.
Suppose that Aristotle experienced a wild horse bearing down upon him. This experience was a cause of his leaping out of the horse's path. However, this sudden movement on Aristotle's part caused a small atmospheric disturbance. The causal ripples of this disturbance extended across the globe until, eventually, every atmospheric event was partially caused by Aristotle's experience. Let's put this date at 200 years after Aristotle's experience and call it D (this is probably a rather conservative estimate). Thereafter, every event having an atmospheric event as a cause has Aristotle's experience as a cause.
It is easy to believe that every human action and experience of this century has as a cause some aspect (indeed, probably all aspects) of the Earth's atmospheric state at D. It is easy to believe this because the state of the atmosphere has a myriad of direct and indirect effects on human activities, and those human activities themselves have effects on further human activities, and so on. For instance, the course of entire wars have been greatly shaped by the weather (think of Germany's Russian incursion of World War Two). And on an ostensibly more trivial level, people have most likely become married and had children together because of a sunny day which made them feel like a walk in the park which resulted in their first meeting. Now, if it is true that every human action and experience of this century has as a cause some aspect of the Earth's atmospheric state at D, then it is plausible that each of these experiences and actions has Aristotle's experience of seeing a horse bearing down on him as a cause. This is because Aristotle's experience is a cause of every atmospheric state at D.
Now, imagine that Nick has a seeming memory of leaping out of a horse's path that just happens to match the content of Aristotle's experience. If we further suppose that that atmospheric event occurred after every atmospheric event had become causally dependent on Aristotle's original experience, then we have a causal connection between Aristotle's experience and Nick's apparent memory.
According to Nick's story, all of the direct connections between Aristotle and himself are mediated in a fashion like this. Such a story is certainly possible. In this case, Parfit should agree that these would be "genuine" or "proper" direct connections of the sort required for Relation R, given that he places no constraints on what sort of causal chains are admissible. Since we assumed that if these connections were direct then there were enough of them for the holding of Relation R, Parfit should also say that Aristotle survives as Nick.
Does Aristotle really survive as Nick? Given that there are sufficiently many psychological matches between Aristotle and Nick, a Parfitian seems committed to saying so. However, far from convincing us that he might be Aristotle, Nick has only succeeded in convincing us that Parfit must be wrong: in the circumstance just outlined, Aristotle doesn't survive as Nick, and therefore, some restrictions need to be placed on the sorts of causes that can underpin Relation R.
What is it about Nick's story that has changed our minds? Perhaps the most immediately striking feature is the vast expanse of time between Aristotle and Nick. But it is not the two thousand year gap in itself that is responsible for the feeling that Aristotle does not survive as Nick. To see this we need only note that we have no conceptual problems with a person's surviving for thousands of years. And such a person might remember doing something thousands of years ago, in which case, a direct psychological connection would hold between mental events thousands of years apart. So there are no conceptual problems with direct psychological connections holding over long lapses of time.
You might also be struck by the fact that the causal connections between Aristotle and Nick are not "normal" ones. But we agree with Parfit that there are good theoretical reasons (and strong intuitions) militating against the requirement that survival need always involve only "normal" causal connections.
What the long time lapse, and the intriciate and chaotic weather connections do entail, however, is that the causal chains connecting Aristotle's and Nick's mental states are extremely "unreliable" in a clearly specifiable sense: replace Aristotle's experience of seeing a horse bearing down on him with Experience A and there is nothing approaching a guarantee that Nick would have an apparent memory of A. - in fact, there is nothing approaching a guarantee that Nick would even exist.
So, perhaps the "proper" sort of cause is a "reliable" one. Perhaps reliability of connections is needed for Aristotle's surviving as Nick. But perhaps not. Parfit has reasons for shunning the reliabilist species of causal theory. We do not intend to make a definite commitment to reliabilism, only to undermine Parfit's "anything goes" view. It is not clear that adding "reliability" would logically guarantee that Aristotle has survived as Nick. But it is clear, we urge, that the mere existence of wildly unreliable connections is not logically sufficient.
§4. Parfit's Reasons.
Our discussion would be incomplete if we neglected to look at Parfit's reasons for holding that the type of causation underwriting Relation R should be unrestricted. Let's quickly recapitulate the three views he considers as to what the appropriate restrictions should be (Parfit 1984, p. 207). One view is that only the normal cause (involving the continued existence of the brain) can underwrite Relation R. Another view is the view we have already touched on, viz., that any reliable cause will do. And, of course, there is Parfit's preferred view, namely, that any cause whatsoever is good enough.
Parfit favours the view that any cause will do because he thinks "what matters" is that Relation R obtains, and not the means by which it obtains. He attempts to persuade us that he is right by appealing to analogous cases where we accept that what is brought about is what matters, not any feature of the causal mechanism which brings it about. For example, here is a case that Parfit raises against the reliable cause view:
Suppose that there is an unreliable treatment for some disease. In most cases the treatment achieves nothing. But in a few cases it completely cures this disease. In these few cases, only the effect matters. The effect is just as good, even though its cause was unreliable. We should claim the same about Relation R (ibid., 287).
It would be wrong, however, to think that Parfit can draw comfort from such examples. He supposes that it does not matter whether the cure was reliable or not. In one respect, this may be true, because what we care about is just that the person who was sick is now well. However, in this respect, it is not clear that we should even care whether the person has been cured by the treatment. In this respect, the effect qua effect does not matter to the patient - the treatment might be ineffective, but the person is cured in a freak uncaused occurrence. The cure, whether unreliably caused, or not caused at all, is just as good for the patient as it would have been if it had proceeded from some reliable causal mechanism. But in the case of Relation R, the effect qua effect does matter. It does matter, for Parfit, whether the psychological matches between Aristotle and Nick are causally connected.
This shows why the example in question doesn't help Parfit. For the example to work in his favour, it must matter to the patient that his recovery was caused by the treatment, but not that the treatment is reliable in effecting a cure. Yet, it is hard to imagine a reason why the patient should care about whether there is a causal connection at all, which is not also a reason to care about the nature and reliability of the causal connection. If all we care about is that we are now healthy then there is no reason why we should care whether one or another of the prior events called "treatments" really caused the return to health. But, if we do care about whether our health was caused by a prior event, then we may be very interested in the characteristics of this event, notably its reliability as a cure.
Here, we shall outline a respect in which, even after having received a successful treatment, we do care very much about the reliablilty of that treatment. And this respect shows us that there are in fact closer analogies between Parfit's example and a reliablist account of survival, than there are between the example and his "any cause will do" story.
Suppose that we have been receiving treatment for inoperable cancer from a charlatan at great expense. By a sheer quantum mechanical accident it happens to be a single-case cause of health, although, as we subsequently discover, it has never cured anyone in the past. Clearly, its causing health in this case is not all that matters to us. This example helps to show why, if we care about whether a prior event caused our health, we are going to care about whether it was a reliable cause: our moral evaluation of the person who treated us depends much more heavily on the reliability of the cure than on the bald fact that it caused our health. And here, there is an analogy with personal survival which suggests that Parfit's own example, if it supports anyone, supports the reliabilist. Remember, it is Parfit who has argued that we should begin by looking for what matters to us, and that we should put what matters into Relation R. If we are cured of an illness, the fact that we are cured matters most, but if the cure is caused then the reliability of its cause matters too. And the reliability of this causal connection matters more than its sheer existence.
Parfit supposes that we are interested in more than whether we have a person with beliefs, desires, apparent memory- experiences and so on, which closely match the beliefs, desires, apparently remembered experiences and so on, of some earlier person. Furthermore, he supposes it matters deeply to us, not just that we have a person like the one we would have if the person had survived in the normal way. It matters deeply to us that this person's states are caused by states of the earlier person. And yet, given that the later states were caused by the earlier ones, Parfit supposes that it is not rational to care at all about the reliability of these causal connections: he thinks that the theoretically and emotionally charged question of survival should be decided according to whether there are such causal connections, and not according to their reliability. Thus, he assumes, for instance, that no greater moral significance attaches to the presence of reliable causal connections than accompanies the mere presence of causal connections. We do not find this assumption compelling.
If we follow Parfit, we should not say that Aristotle has survived as Nick if Nick's psychological states are uncaused - we should not specially honour him, praise him, pay him royalties for his work, and so forth. The shift in moral treatment which follows a judgement of survival ought to be triggered by the presence of causal connections, however unreliable. So, for Parfit, it is of equal moral and theoretical significance if, on the one hand, there is a one-off collection of quantum mechanical causal links between matching pairs of Aristotle's and Nick's psychological states, or if, on the other hand, these caual links are reliable. We think there is insufficient motivation for this view. It is difficult to see why mere causal connection should be "what matters" for survival, when in so many ways it is reliability that seems far more important.
It sounds plausible when Parfit says that all that matters is that the person survived - it does not matter how reliable the mechanism was by which they survived. But beware! This sounds plausible, we think, only because it begs the question of personal survival in a way that is impermissible for a reductionist like Parfit. Remember what Parfit has to say. We care about whether Nick's psychological states appropriately "match" Aristotle's psychological states. But, in one important sense according to Parfit's views, we are not like the person who cares only that they are now healthy again. We are like the person who wants to know which of the various prior events actually "cured" them. Parfit holds that it is not enough for survival that Nick and Aristotle have a sufficient number of matching psychological states. In addition, Enough of Nick's psychological states also have to be brought about by Aristotle's corresponding psychological states. For instance, suppose that Aristotle's experience of speaking with Plato had been a cause of Nick's apparent memory of seeing a wild horse bolting towards him and not a cause of Nick's seeming to remember speaking with Plato. If most of the ostensible direct connections between Aristotle and Nick were thus mismatched, then Parfit would say that Nick is not a survivor of Aristotle. Thus, according to Parfit, we care deeply about the distinction between, on the one hand, a Nick whose psychological matches with Aristotle are purely accidental, and on the other hand, a Nick whose psychological matches with Aristotle were mediated by chaotic weather-connections. But we do not care at all about the distinction between a Nick weather-related to Aristotle, and a Nick reliably-related to Aristotle. How plausible is this? Not plausible at all. And, certainly, Parfit's "anything goes" doctrine gets no support at all from the analogy with the case of the "unreliable cures" of a disease.
And yet, suppose we were to grant that Parfit had undermined a reliabilist theory about what kinds of causal connections matter for survival. It is far from obvious that this justifies his conclusion that "anything goes". Nick is not Aristotle, and even if there were chaotic weather-links between his states and Aristotle's, that would not be enough to make it the case that Aristotle has survived.
Derek Parfit, "Personal Identity", Philosophical
Review 80 (January 1971).
Derek Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984).
Donald Davidson, Essays on Actions and Events (Oxford: Clarendon Press, 1980).
Thanks to Sam Gates, Cathy Legg and Steve Matthews for helpful comments and discussions.