Knowledge turned upside down

Do not think that the goal of belief is truth and truth only; the goal of belief is knowledge. Do not think that belief is conceptually prior to knowledge; if it is just part of the nature of belief that belief aims at knowledge then knowledge is conceptually prior to belief. Do not think that knowledge is a composite of belief, truth and other more elusive features; knowledge has no composition at all. These are just some of this book's many striking claims, backed by one original argument after another.

Timothy Williamson is well known to the philosophical community and beyond. His previous book, Vagueness , was universally admired and proved to be something of a philosophical bestseller. This new book is broader in scope, far more ambitious and much less forgiving to the inattentive reader (although one or two of the more technical chapters can be skipped without spoiling the overall effect). Although it is largely based on a string of previously published articles, it has more unity of theme than might be expected.

There is also plenty of scope to navigate through it as you please (at least if you read the introduction carefully). Try the penultimate chapter on assertion first - it is crucial to the book as a whole. It is here that we are told why only knowledge warrants assertion.

Williamson's basic idea is simple: it would be a kind of pathology to assert "John is in town but I don't know that he is" because such a sentence may well be true, even though one trips up over the rules of language in asserting it. In asserting, one implicitly represents oneself as having the authority to make that assertion. That is why it is pathological to add "but I don't know that he is".

Williamson takes this to reflect a simple "constitutive rule" governing assertion, a rule that forbids asserting a sentence when one does not know that the sentence is true. This is why Williamson takes the goal of belief to be knowledge and not mere truth.

Many readers may find this incredible. The worry is that Williamson's assertion rule is too strong; it forbids an assertion when one lacks knowledge but is nonetheless in possession of strong evidence. The simple and original reply is that knowledge is evidence.

If one thinks otherwise, one is likely to be working from within some internalist conception of knowledge, whereby one is always in a position to know what one's evidence is. Not so for Williamson - one can satisfy the assertion rule without knowing that one has done so.

Moreover, if one thinks that something weaker than knowledge warrants belief or assertion, one is then required to explain just why it is pathological to assert that "John is in town but I don't know that he is". What better explanation is there than the equation of knowledge with evidence?

Williamson's view is less incredible when one appreciates the theory behind it. Knowledge is easier to come by than many philosophers have imagined. Williamson recommends that we rethink how knowledge requires reliability.

In chapter seven (previously unpublished), Williamson debunks any conception of knowledge in which to know that something is the case, one's belief must be sensitive to the nearest possible situation in which it is not the case.

Knowledge requires safety not sensitivity - a theme that forms the backbone of the middle chapters. It is here that Williamson develops and applies a novel version of reliabilism. The idea is that our beliefs are reliable when and only when they are safe from error in relevantly similar cases. One attractive feature of safety-reliabilism is that, unlike sensitivity-reliabilism, very few concessions are made to the sceptic - the theme of chapter eight.

If knowledge requires safety, then to know one must leave a margin for error, shielding one from forming false beliefs. Here Williamson attacks the Cartesian conception of the mental: that if one is in a certain mental state, then one is always in a position to know that one is in that state. He also attacks the internalist conception of knowledge that if one knows, then one is always in a position to know that one knows. These principles have been attacked before, but Williamson's assault on them is novel.

Nonetheless, Williamson's arguments reveal certain tensions. To get from his safety requirement on knowledge to the idea that knowledge must leave a margin for error, he relies on the principle that there is limited discrimination in the methods via which we form beliefs.

One worry is that this principle is neither necessary nor knowable a priori , making Williamson's margin for error principles likewise neither necessary nor knowable a priori. This has consequences for whether these principles can be used to support the case for the existence of unknowable truths - the theme of the last chapter.

Another worry concerns Williamson's entitlement to use a principle concerning belief to substantiate insights concerning knowledge. Employing this principle conflicts with his key idea that knowledge is both conceptually and explanatorily prior to belief.

Who should read this book? Decision theorists, game theorists, cognitive scientists, psychologists, linguists, philosophers - anyone whose discipline has traditionally subordinated the place of knowledge and elevated that of belief or justification. This is one of the most important and refreshing books on epistemology written in the past 20 years. Williamson is to be commended for turning the theory of knowledge upside down.

Patrick Greenough is lecturer in philosophy, University of Bristol.