This volume contains three independent but related works, one by each of the authors. The result is a feast of contemporary epistemology approached through the question of how to account for the value of knowledge. Each contribution is highly readable, each quickly carries readers towards the cutting edge of discussions in recent epistemological literature, and each adds important, original arguments to those discussions.

Although this is in a sense an anthology (a ‘tryptich’, as Ernest Sosa describes it on the book jacket), it hangs together better as a whole than anthologies usually do. Despite differences of style, focus, and conclusions, the contributors have been collaborators in a project to see how far they can advance the study of knowledge through considering its value over and above the value of mere true belief, and their collaboration is evident in the organization and inter-responsiveness of the contributions.

Part I. Knowledge and Understanding

In the first piece, Duncan Pritchard provides a masterful account of the general problems that face any account of why knowledge is more valuable than mere true belief, and he goes on to argue that it is not knowledge but, instead, ‘understanding why’, (which often accompanies knowledge) that is necessarily more valuable than mere true belief.

He begins by distinguishing what he calls the value problem and the swamping problem. The value problem is the general problem of how to account for the greater value of knowledge. A solution to the value problem must explain why knowledge is distinctively valuable, and thus greater in kind, and not just degree, than mere true belief. The swamping problem is problem for theories, like reliabilisms, which conceive knowledge in terms of true belief plus some means of arriving at true belief. Since such theories conceive the added feature merely as a means to true belief, the value they add to knowledge, beyond the value of true belief, is swamped by the value of true belief. A theory of knowledge that will overcome both problems will imply that knowledge is intrinsically, or finally, valuable.

Pritchard thinks that the existing theory of knowledge with the best shot at overcoming both problems is ‘robust virtue epistemology’: the theory that knowledge is ‘…the product of one's reliable cognitive abilities’ (25–26). According to the best version of the theory, a la John Greco, knowledge is a cognitive achievement primarily creditable to the knower. Since achievements are plausibly distinctively valuable, in a way that mere true belief is not, this theory has potential for overcoming the value problem. Since the value of cognitive achievements is distinct from the value of mere truth-conducivity, this theory is not susceptible to the swamping problem.

Nevertheless, Pritchard argues that robust virtue theory is implausible. We can have knowledge without cognitive achievement and cognitive achievement without knowledge. So even if cognitive achievement is more valuable in kind, and not just quantity, than mere true belief, this theory fails to show that knowledge is finally valuable.

A problem with robust virtue theory is that it answers to only one of two fundamental intuitions we have about knowledge. It answers to our intuition that knowledge is the successful exercise of a reliable ability; but it fails to answer to is our intuition that knowledge cannot be lucky, in the sense that one's true belief could have easily been false. On this basis, Pritchard proposes a new theory, ‘anti-luck virtue epistemology’, which is roughly a grafting together of robust virtue epistemology and an anti-luck condition, which implies that the relevant kind of cognitive achievement is ‘…to a significant degree creditable to one's character.’ (51) Nevertheless, even this theory fails to connect knowledge with achievement in a way that is sufficient to show that knowledge is finally valuable. He concludes that knowledge does not have final value that is distinct from the value of mere true belief.

So why do theorists usually suppose that knowledge is more intrinsically valuable than true belief? He argues that it is because a related, nearly coextensive kind of cognitive achievement, understanding why such-and-such is the case, is more intrinsically valuable than true belief. This nicely sets the stage for Millar's and Haddock's contributions, since both of them think that understanding why in this sense is necessary for knowledge.

I question some of Pritchard's intuitions, especially the ones on which he bases his argument that knowledge is not always an achievement. They seem too sportsmanlike, as if he is conflating the relevant sense of the ‘distinctive value’ of knowledge, which requires that it be different in kind and not just quantity from true belief – with a sporting one which requires that it emanates from more than common skill or overcomes extraordinary obstacles.

Part II. Knowledge and Recognition

Traditional epistemology uses conceptual analysis to try to reduce the concept of knowledge to its constituent parts: true belief plus something(s) which conceptual analysis has so far failed to reveal. Alan Millar is skeptical about the prospects for success in this enterprise. He proposes instead to investigate knowledge on the basis of the role it plays in our thought and lives, in which we value knowledge because it ‘puts us in a position take [a] matter to be settled, to terminate enquiry and to vouch for [its] truth.’ (95)

Since Millar is skeptical that conceptual analysis can reveal necessary and sufficient conditions for knowledge that do not ‘implicate the concept’, he begins with a mainly descriptive investigation of the means by which we accomplish knowledge, or knowledge's purposes. However, he wants to do more than describe knowledge. He wants to provide normatively significant accounts of perceptual and testimonial knowledge.

He argues that perceptual knowledge is in part the result of the successful, responsible application of our perceptual-recognitional (P-R) abilities, conceived as broad competences. Our applications of these abilities are usually successful, in large part because of what he means by a broad competence. A broad competence implies a favorable environment – one in which the things one recognizes have distinct appearances, unlike in the famous problem about a person who is good at recognizing barns in his home territory, but who loses this ability as he enters a territory where most apparent barns are fake. A narrow competence, on the other hand, is not environment-sensitive in the same way.

In addition to perceptual belief being produced in the right way for knowledge, by the successful operation of relevant P-R abilities conceived as broad competences, we need higher-order knowledge in order to responsibly take a perceptual matter to be settled. He argues that we manage this through the operation of higher-order recognitional abilities . He is not worried that this generates a vicious regress of dependence on higher-order knowledges because he thinks that the requisite higher-order abilities operate upon the same data as our first-order ones, and he thinks the understandings of the wider world that inform our higher-order abilities do not need to be independently supported in order for us to responsibly rely on them.

Although basic perceptual recognition is psychologically immediate, its content is rich and takes us well beyond the ‘the superficial properties of things that go to make up their appearance.’ (92) This suggests the falsehood of the ‘covering generalization’ model of perceptual evidence, according to which we apply covering generalizations – like everything red and shaped thusly is a fire engine – to superficial appearances provided by perception. Instead, Millar argues, basic perceptual recognition is like direct recognition of the significance of indicators, such as the significance of a fuel-gauge reading, to one skilled at assessing the significance of fuel-gauge readings. Obtaining knowledge turns out to be much more like a practical skill than conceptual reductionists suppose.

He gives a similar account of straightforward cases of testimonial knowledge: cases of knowing that p because someone tells us that p. In transition between his accounts of perceptual and testimonial knowledge, he interestingly considers ‘free standing’ knowledge – knowledge for which we don't remember our sources, such as my knowledge that Salem is the capital of Oregon.

Part III. Knowledge and Action

Epistemologists usually respond to the Gettier problem by rejecting that justified true belief is sufficient for knowledge. But Adrian Haddock claims that this is too drastic. A better response – he calls it the ‘modest route’ – is to hold that justification of the sort involved in knowledge is truth-guaranteeing and, furthermore, knowledge-affording. He demonstrates this modest route with accounts of visual knowledge and self-knowledge of action.

In the case of visual knowledge, it is the fact that I see that things are a certain way that justifies me in believing that things are that way, and it is because I know that I see that things are that way that I know that things are that way. Thus, the kind of justification involved in visual knowledge implies second-order knowledge. The second-order knowledge is self-knowledge – for instance knowledge that I see that there is a lemon over there. The first-order knowledge is visual knowledge – for instance knowledge that there is a lemon over there.

The second-order knowledge is of a transparent fact: I know that I see the lemon over there because I see it over there and my capacity for visual knowledge is ‘suitably actualized’, so that this is intrinsically a case of my seeing it rather than, for instance, a Gettier-style case in which I see it but I do not thereby know that I see it.

At this point a regress seems to loom. Something I know, besides the fact that I see the lemon, must justify my second-order knowledge that I see it. This something – presumably higher-order knowledge – must in turn be justified by something I know, and so on, ad infinitum. Haddock argues that we can avoid this problem by conceiving the transparency of visual knowledge as entitling, rather than justifying us in holding the second-order knowledgeable beliefs which justify our visual knowledge. Thus, although justified true belief is sufficient for knowledge, it is not necessary for knowledge, since relevantly entitled true belief will do for higher-order knowledge.

The second-order self-knowledge that justifies visual knowledge is non-sensory – it is not the result of an external object affecting the mind (as is visual knowledge) but rather it is the result of the mind producing the object of the belief. Thus we have an account of fallible self-knowledge which Haddock sees as a natural ally of an Anscombian account of self-knowledge of action as a species of ‘knowledge without observation’.

He develops an Anscombian account of first-person knowledge of actions along these lines, which will be interesting to action theorists as well as epistemologists.

The distinctive value of knowledge on Haddock's account is Socratic: knowledge enables us to accomplish our goals in a way that is superior to that of mere true belief because knowledge is rationally immune to doubt in a way that mere true belief is not. To defend this he argues against Pritchard's view that in order for knowledge to be finally valuable it must be different in kind, rather than quantity, from mere true belief.