Reading: Nozick Philosophical Explanations Chap 3, sections I and II
Martin 'Tracking Nozick's sceptic: A Better Method' Analysis 43 (1983)
Garrett 'Nozick on Knowledge' Analysis 43 (1983)
Gordon 'Knowledge, Reliable Methods and Nozick' Analysis 44 (1984)
Wright 'Keeping Track of Nozick' Analysis 43 (1983)
I will argue that Nozick's account of knowledge does not adequately answer the sceptic for two reasons. Firstly Nozick's account of knowledge can be criticised because it does not allow for the fact that someone may arrive at a true belief by way of an unreliable method, and secondly because, even if we grant that Nozick's account of knowledge is satisfactory, his attack on scepticism relies on an analysis of counterfactuals similar to that of Lewis, which can be shown to be faulty. I will firstly summarise Nozick's account of knowledge, then look at some examples which seem to undermine it (namely, examples in which we would not want to say we have knowledge, but that satisfy Nozick's four conditions for knowledge). I will then look at how Nozick's response to scepticism relies on an incorrect analysis of counterfactuals and show that, even if his four conditions for knowledge are correct, he has not adequately answered the sceptic.
In 'Philosophical Explanations' Nozick develops an account of knowledge designed to overcome the problems faced by accounts of knowledge as justified true belief. His analysis of knowledge relies on counterfactuals, and therefore overcomes Gettier counter-examples to the theory of knowledge as justified trues belief. According to Nozick, there are four conditions which have to be met in order for person S to have knowledge that P:
(1) P is true
(2) S believes that P
(3) ¬P -> ¬ (S believes that P)
(4) P -> S believes that P L ¬ (S believes that ¬P)
The first two of these on the own would show that knowledge is true belief, but the third condition if brought in to provide a means of overcoming Gettier examples. For instance, say a man, A, believed that someone in the office owned a Ford on the basis that person B drove to work every day in a Ford. However, he doe not know that person B has sold his Ford to person C, who happens to work in the same office. We would not want to say that A knows that someone in the office owns a Ford, and yet he has a justified true belief that this is the case. Nozick's third condition requires that if it were not true that someone in the office owned a Ford, person A would not believe that someone in the office owned a Ford. But this is not the case here, as person A would still hold his belief even if C sold his Ford to someone who did not work in the office. So it seems as though Nozick's refinement adequately deals with Gettier cases. The fourth condition is needed to solve problem cases like Nozick's own example of the man in a tank who is being brought to believe things by scientists stimulating his brain. We need to say not only that if he was not in the tank he would not believe that he was, but that if he was in the tank, he would believe that he was in the tank. Only then would he have knowledge about being in the tank.
So Nozick's account of knowledge is this:
"A person knows that P when he not only truly believes it, but also would truly believe it and would not falsely believe it. He not only actually has a true belief, he subjunctively has one" (P.178) Nozick calls this 'tracking' the truth.
So is Nozick's account of knowledge correct? I will argue that it is not, because it is possible to provide counter-examples to it, which show that Nozick's description has left something out, namely that it does not account for true beliefs arrived at by way of an unreliable method. Martin argues that Nozick's view is unacceptable because it does not require a belief to be justified in order for someone to have knowledge:
"A problem for it is that it seems that someone could have a belief that P which tracked the fact that P even though he was not justified in believing P" (Martin, P30)
He gives the example of person S who makes a bet that will gain him $10 if either Gumshoe wins the first race, or Tagalong wins the second race or they both win their respective races. S infers that Gumshoe won the first horse race from the fact that he receives $10 from the cashier, but we would not want to say that S knows that Gumshoe won, even though this case satisfies all Nozick's conditions for knowledge. Garrett argues that this example is faulty because Martin cannot show that the possible worlds in which Gumshoe does not win the first race and Tagalong wins the second and far enough removed from the actual world to be irrelevant. But, as Gordon points out, it seems we can make an addition to Martin's example which would enable it to satisfy Nozick's condition (3). We can suppose that Tagalong is such a bad horse that there is no hope of him winning the second race, so in this case we do seem to have a counter-example to Nozick. Gordon argues that this does not tell against Nozick's account, however, because we could just say that S does know that Gumshoe won the first race. But I am inclined to say that S does not have knowledge in this case and therefore Martin's objection to Nozick does follow through. It is a condition of the success of Martin's and Garrett's counter-examples to Nozick that one must not only have a reliable method, but must also believe one's method to be reliable in order to have knowledge.
This is the reason why I would argue that S does not have knowledge in Martin's example. S does not know that Tagalong has no chance of winning and so he does not know that his method will always produce the right answer. One way to test whether someone has knowledge or not may be to see how well S could persuade someone else that he knows that P. In Martin's case, S would have trouble persuading someone else that he knows that Gumshoe won, because S is not entitled to infer (given the limited information he possesses) from being given $10 by the cashier, that Gumshoe won. For all S knows, it is equally possible that Tagalong won and not Gumshoe. So there seems to be necessary not only to have a reliable method, but also to be aware that your method is reliable in order to have knowledge. S does not know that Gumshoe won because he cannot arrive at that as the only possibility, given the information he had.
So I hope we have seen that Nozick's theory of knowledge is inadequate and therefore he cannot provide an answer to the challenge of the sceptic. However, it remains to be said that even if we were convinced by Nozick's theory of knowledge and did not find the supposed counter-examples to it persuasive, the way in which Nozick replies to the sceptical challenge is problematic because it relies on the notion that counterfactuals are not transitive, which can be shown to be erroneous. I will firstly briefly examine Nozick's own arguments against scepticism and then turn to a criticism of Nozick's response. The sceptical argument basically holds that, since we cannot be sure that we are not being tricked by an evil demon, or that we are not just a brain in a vat, we cannot have knowledge such as 'I know I am now sitting at my desk' or 'I know that I have a hand'. Nozick's criticism of scepticism is that the sceptical argument presupposes logical implication, and Nozick argues that it is wrong to assume that knowledge has to be transitive across logical consequence. So an assessment of Nozick's treatment of the sceptical argument comes down to whether he is right in arguing that the subjunctive conditional is not transitive. Wright, in his paper 'Keeping Track of Nozick', seeks to show that Nozick is incorrect in his analysis of counterfactual conditionals and I would argue that Wright succeeds in this aim. It seems as though Nozick is entitled to presuppose that 'the range of counterfactual circumstances embraced by 'Had it not been the case that I have a hand...' must fall short of SK [sceptical possibilities] and its kin' because, as Wright's seat belt example shows, every assertion of a counterfactual rests on a tacit understanding that the range of counterfactual circumstances has to be limited. But we need to know the criteria for determining the 'extent of the range of relevant counterfactual circumstances'. Nozick follows Lewis' account under which transitivity is unacceptable. But Wright's objection to Lewis, and thus to Nozick, is that: "How 'far out' it is appropriate to search for relevant counter-examples to an asserted counterfactual looks to be determined not simply by its content but by the context in which it is asserted" (Wright, P138) As is shown in the skiing example, it seems that when a number of counterfactuals are used in the same context, then we must also be dealing with a single range of relevant worlds, and it is this point that Nozick fails to pick up on.
To conclude, then, I would argue that Nozick's answer to scepticism falls down on two accounts. Firstly, it can be shown that Nozick's account of knowledge is not sensitive to examples where true belief is arrived at by way of an unreliable method. Secondly, Nozick's answer to scepticism only works if counterfactuals are not transitive, and it can be shown that they do "like all conditionals, sustain transitivity in context" (Wright, P140)
© Anne Witton 1996. No part of this article may be copied without my permission.
