To: Kristian D.
From: Geoffrey Klempner
Subject: Can you know that you are not dreaming?
Date: 1st December 2011

Dear Kristian,

Thank you for your email of 21 November, with your essay for the University of London BA Epistemology module, in response to the question, 'Can you know that you are not dreaming?

You have used the question as an opportunity to offer an exposition of Nozick's 'truth tracking' definition of knowledge and consequent denial of the claim that knowledge is closed under entailment. This takes up about two thirds of the essay. Then, as an added bonus, you suggest alternative ways of tackling the dream hypothesis: inductive/ probabilistic, coherentist and finally a return to the view which you explored in your last essay, according to which we can effectively bracket the dreaming question and just not worry about it, assessing knowledge claims in the normal way. This time, you have consider 'realist' and 'anti-realist' versions of the bracketing theory, both of which you find unsatisfactory. More on that later.

Although it's often said, I don't agree with what you say in paragraph two that Descartes' scepticism is the same as the 'brain in a vat' variety; at least, if we take it in the full-blown version, which Descartes leads up to in stages. If I am being given a perfectly coherent dream by an evil demon, then there is no such thing as space or matter; whereas if I am in a pod in the Matrix, all I am wrong about is how things are in the physical universe.

This difference is important, because in the 'Refutation of Idealism' in the 2nd edition of Kant's 'Critique of Pure Reason', Kant believed that he had offered an effective response to Descartes. If I know that I exist, then I must know facts about objects in an external world. (Later, of course, we realize that this is qualified by the doctrine of transcendental idealism -- space and matter are 'empirically real but transcendentally ideal'.) When Kant comes to consider the possibility that I might still be dreaming, i.e. just asleep, he dismisses this with a remark to the effect that this kind of question is one which we deal with in the same way as we deal with any other question about the best explanation of the course of our experience.

That reply effectively makes the same point as your argument from induction/ probability. Indeed, if we look at the matter this way, then it could be argued that there is nothing whatsoever special about the question, 'Can you know that you are not dreaming?'. It's just one of many varieties of inductive scepticism, along with, 'Can you know that a terrorist missile is not going to blow you up before you have finished writing your email?', 'Can you know that the person you thought was a human being is not an alien from a planet in the vicinity of Betelgeuse?'

-- I won't repeat what I said last time about the contextual issues raised for knowledge here.

I don't have any comment on your exposition of Nozick. You are probably aware that Nozick's theory has come under attack, not least for its failure to deal adequately with Gettier counter-examples, although various fixes can be attempted. I think that the idea of 'truth tracking' is important and worth saving. Indeed, it has become part of contemporary philosophical vocabulary, even among philosophers who don't buy Nozick's definition of knowledge.

Truth tracking does also seem to be the core intuition in the bracketing theory (sorry, I couldn't think of a better name -- the allusion is to Husserl, not altogether apposite). Even if my life is one long dream, the question of whether or not my beliefs track truth (or, rather, dream-truth) is a valid one to ask, and enough to keep us busy. But then one has to deal with the objections I raised last time about 'what it all means'.

However, I don't agree with you about the anti-realist construal, although I can see how you get there. This happens to be a topic which I spent much time on when I was writing my doctoral thesis. Michael Dummett was lecturing on mathematical intuitionism. John McDowell (my supervisor) had written a seminal paper criticizing Dummett's argument for an anti-realist theory of meaning, and linking him to Quine. My counter-argument was that the anti-realist has far more subtle resources, doesn't need to offer a revisionary theory of meaning, can keep the law of excluded middle, in fact doesn't need to 'say' anything other than to reject dialectically every attempt the realist makes to 'state' what realism 'means'.

Forget about the question whether the question whether I am dreaming 'has' an answer (realism) or 'doesn't have' an answer (anti-realism). One thing that both parties can agree on is the possibility that I will wake up (or seemingly wake up). That's the cash value of the hypothesis. The anti-realist isn't required to offer an 'anti-realist definition of truth'. An anti-realist (pace Dummett) can perfectly well accept that 'truth transcends verification' in the sense that you can never offer an adequate substitute for 'is true' in any terms that involve verifying, testing etc. (even the notion of the 'limit of inquiry', which will always be relative to the capacities of the inquirers). I might wake up, like Neo. Or I might be blown up in the next minute by a terrorist missile. (As it happens there's a prominent mosque opposite my office.) Or you might be an alien from Betelgeuse.

All the best,

Geoffrey