Evidentialism, like most philosophical theories, begins with a truism and ends in some confusion. ‘One is epistemically justified in accepting that p if and only if believing that p fits the evidence that one has for p’ (Lehrer, Evidentialism and its Discontents (hereafter ED), p. 55, after Conee and Feldman). But what is it to ‘believe’, and can we choose our beliefs? I may find myself simultaneously believing p and also believing that I have no objective right to that belief (no one would find it odd that I might desire that p though agreeing that I shouldn't). Will the first belief (if I am virtuous) slowly evaporate, or the second (if I am not)? Does believing require believing that some proposition is really true (so that non-human animals and infants have no beliefs)? Must enquirers seek to hold back from ‘believing’ till the case is proved, or may they have unprovable beliefs? Is my own immediate conviction or experience that p itself evidence for p? Or should I only grant as much weight to my own ‘evident experience’ as I would or should to anyone else's? If knowledge is – as many suppose – the same as a true belief caused, ‘in the right sort of way, by relevant facts’, and my belief that p counts as knowledge if the circumstance identified in p plays a proper part in causing my belief, may I not properly believe that p even if I have no evidence that p? My belief, perhaps, is engendered by the relevant fact, even though no perception of that fact proves the truth of the belief, and even if I don't know what has caused it (I can recognize a friend – from her walk, her silhouette, the way she turns her head? – before I can articulate what it is that cues my recognition, and long before those properties prove her identity).

These volumes deal chiefly with those problems of ‘evidentialism’ that turn on the nature of ‘evidence’ and the grounds or reasons for believing or disbelieving, rather than problems with the notion of ‘belief’ itself. The seventeen papers of ED focus on defences of evidentialism by Conee and Feldman, who provide detailed and helpful replies to their critics. The twelve papers of Evidence and Religious Belief (hereafter ERB) address Plantinga's ‘Reformed epistemology’ as it applies to ‘religion’, drawing on George Mavrodes's work (there is one hint that the book was intended specifically to honour him: Rowe, ERB, p. 185). Both collections are hard and rewarding work in the analytical tradition.

This does not make for easy reading. The second volume (ERB) is likely to be of more direct interest to this journal's readers – and is also (mostly) better written. There is much talk – especially in ED – of epistemic significance, epistemological disjunctivism, meta-evidence and metalevels, parity conditions, undercutters, defeaters, the Rational Uniqueness Thesis (‘hereafter RUT’: Axtell, ED, p. 76), and assorted other acronyms (BIV, GPO, RWH) – all the paraphernalia of analytic argument. One essay, by Williamson (ED, pp. 147–164), on whether we may know p without knowing or even justifiably believing that one does, must be opaque to any ordinary anglophone (as being written mostly in formal logic). There are more lucid observations – like Huemer's dictum that ‘we must not require an epistemic agent to verify the reliability of his own appearances before trusting them, because the agent would have no way of verifying anything without trusting his appearances’ (ED, p. 28; see also Zagzebski, ERB, pp. 22–36). Since those appearances include immediate intuitions and convictions (as ED, p. 10), it may be that ‘evidentialism’ is here being stretched to its limits: I may continue believing p without any ‘evidence’ of a kind that should influence anyone else. But the point of the initial rule was to demand that we correct our own convictions, or at least our habit of being convinced, in the light of objective reason, and not rely on what seems ‘evident’ to us! ‘Call the view that a belief cannot provide evidence for itself the independence constraint’, and its denial ‘the self-support condition’ (Lehrer, ED, p. 59). Once agreed that any beliefs are warranted because they are generated by some seemingly reliable mechanism there is less reason to require any further evidence (at least until there is evidence against that reliability: maybe my friend is known to be somewhere else).

Some authors (as Axtell, ED, pp. 78–81) look back to the disputes between William James and William Clifford (thus providing evidence that our Victorian predecessors were much better stylists). Locke's ruling that it ‘would become all men to maintain peace and the common offices of humanity and friendship in the diversity of opinions’ might be threatened by too firm an insistence – implied by RUT – that we should all unite in believing (or publicly avowing) all and only what we should. By this account, it is not only true that either theists or atheists are incorrect in fact, but that one party is culpably irrational: whereas the more civil understanding would be that each side, even if one is factually mistaken, is rationally entitled to its view and should not be harangued or bullied to silence or insincerity.

Does such tolerance verge on scepticism? It is axiomatic amongst epistemologists that ‘scepticism’ is ‘debilitating’ and therefore (?) must be wrong, and that they are able to prove it. Swinburne (ED, p. 196) is open in asserting that any theory which implies that this life is a dream and a delirium, as Marcus Aurelius – apparently – found it, must be unacceptable. But notoriously, the closure principle (knowing p and knowing that p implies q at least permits me to say that I know q) and modus tollens (if p then q; not-q and so not-p) together imply that since I cannot prove (and so do not know) that my experience is not ‘virtual’ (typically and obscurely, that I am not a ‘brain-in-a-vat’ or ‘BIV’) I also don't know any of my ordinary knowledge claims (for example, that I am sitting at my computer, typing with two hands). Steup (ED, pp. 105–122), after dismissing two other common replies (that either I simply know I'm not a BIV, or else the closure principle is mistaken), argues that we do have evidence against ‘the BIV hypothesis’ sufficient to show that it isn't true (but the evidence he adduces may, by hypothesis, be deceitful). Unfortunately, Steup, Fumerton (ED, pp. 179–191), Rysiew (ED, pp. 207–225), Swinburne, and Pritchard (ED, pp. 235–253) do not address the simulation argument advanced, for example, by Nick Bostrom (‘Are You Living in a Computer Simulation?’, in Philosophical Quarterly, 53 (2003), 243–255): that it is not merely possible that we're living a virtual reality, but actually very probable! Nor do they adequately address what lies behind the rejection of the closure principle: namely, that all knowledge claims are contextual, claiming no more than that – within a certain range of options – all but one can be dismissed for now. I am sitting at my computer rather than feeding the cats in the kitchen, but I cannot similarly bring evidence to decide whether I am a twenty-first-century anglophone, a 500th-century post-human enjoying a historical drama, or an immortal spirit imprisoned in Plato's Cave. All I can do is carry on, without knowing what I am doing! Is there anything wrong with withholding judgement?

Is there anything really wrong, on the other hand, with not withholding judgement? Swinburne's Principle of Credulity (ED, p. 202), that we are right to believe ‘everything that we do in fact believe as strongly as [we believe] it until it is rendered improbable by something else [we believe]’, may be a useful basis for thought or action in this world. We have to start from where we are, or at least where we seem to be – the principle of Phenomenal Conservatism defended by Tucker (ERB, pp. 52–73), with appropriate checks. But where do ‘we’ seem to be? If truth is what sensible people say, are we quite sure we're sensible? If we need a sound understanding before we can have good evidence (Rysiew, ED, pp. 221–222), how should we know if we do? Aristotle's dictum, that truth is what good people say, is only useful for those who think they're good (and perhaps they are deluded). DeRose (ED, pp. 137–146) and Baehr (ED, pp. 88–102) have interesting comments to make on epistemic virtues, but they too assume too easily that ‘we’ are virtuous!

The examples given in ED of well-attested beliefs, or even items of knowledge, are immediate, local, or personal items: that I am sitting at a computer, or can recognize a pink-spotted fly-catcher. My ‘evidence’ may often simply be that that is how things look to me just now (and that I remember being proved right often enough in the past). These are not what most people worry about: the more serious cases are to be heard in Parliament and courts of law, banks, laboratories, or even seminars. That is where we most need evidence, of one sort or another, not simply personal conviction: criminal and civil cases have somewhat different standards of ‘proof’, as do the experimental or mathematical sciences, investment policy, historical investigation, and philosophical enquiry. The most contentious area – ‘religion’ – is addressed in ERB. Must ‘religious’ belief be founded, like other beliefs, on ‘evidence’ (and if so, what), or may it be ‘properly basic’, held for good reasons (or not) but without ‘objective’ evidence? Conversely, are there ‘defeaters’ for the received beliefs: for example, whether God as traditionally conceived is free, or morally respectable (Rowe, ERB, pp. 175–185), or whether – if He exists – He ought to make His existence obvious (Coffman & Cervantes, ERB, pp. 95–113)? In ED the question, what ‘belief’ amounts to, is not fully addressed. In ERB, what ‘religion’ is does not receive much attention, nor even ‘God’, nor whether ‘belief’ is always an appropriate concept with which to describe ‘religion’.

Even when ‘belief’ is central to a ‘religious’ tradition it may not rest on any inference from secular ‘beliefs’, nor are religious doctrines meant to explain only what seems obvious to the irreligious. Credal religions offer their own context of argument and inference: we may find ourselves alive within them, without ever having been argued into accepting them. As Ross argues (ERB, pp. 13–21), we rely on our faculties and forms of life – like other animals – not because we have assessed ‘the evidence’ but because it pays to do so! Believing that there is a real world or a real past at all – some say – is not an inference from more solid evidence (since those beliefs define what evidence there can be). The same might apply to Christian faith: in Ross's description, ‘a form of life aimed at respectful love and care of mankind and at life unending with God’ (ERB, p. 21). Maybe one reason for adopting a mainstream theism (with the prospect of a real and eternal reconciliation of all our separate goods) is that it gives us better reason to be ‘morally good’ (as Clark & Samuel, ERB, pp. 157–174). Theists can also simply accept the consensus gentium (Zagzebski, ERB, pp. 22–36; cf. Kelly, ERB, pp. 135–156) without supposing that the consensus itself is evidence of God's existence. Believers may also ‘see’ the moral law, the starry heavens, or a flower as signs of God: perhaps those experiences count as ‘evidence’ (even if the non-believer doesn't think so), or perhaps the believers simply experience God at work in the law, the stars, the flower (see Evans, after Mavrodes, ERB, p. 44). Maybe, as Evans suggests (see also Goldman, ED, pp. 254–280), there is at least a ‘partial rapprochement between evidentialist advocates of natural theology and Reformed epistemologists’: simply a question of how the experience is described, whether it is the vision (the experience) that is evident to the visionary or the reality (the object of the vision) that she sees. ‘In one case the person simply “reads” the sign spontaneously and perceives God, while in the other case the individual perceives the sign and infers God's reality’ (ERB, p. 48). In yet other cases, of course, the individual neither reads the event as a sign, nor infers God's reality from whatever emotion she feels.

Robust evidentialists may insist that we ought to cultivate an epistemic character that trusts the public evidence more than the private. Seeing ‘heaven in a wild flower’ is not what every reasonable person does – but Evans plausibly retorts that ‘natural signs for God’ are very widely attested, even by philosophers who reject the formal arguments for God's existence (ERB, pp. 46–47). And Reformed epistemologists add that the fact that Calvin's sensus divinitatis (misquoted as ‘sensus divinitatus’ in the Introduction, but not elsewhere) is obscured in some cases does not compel us to reject its message. As Wainwright observes, in one of the best essays of ERB (pp. 77–94), ‘a properly cultivated emotional nature is essential to sound ethical reasoning’, and the same may be true in reasoning about God (ERB, p. 87). What epistemic character we cultivate depends, in circular fashion, on the conclusions we have already reached about what sort of place the world is.

Wainwright's essay – whose main theme is an answer to the suggestion that it is morally and epistemically better to be agnostic than a believer – also stands out because he draws on Muslim and Hindu argument as well as familiar western Christian tropes. Hasker (ERB, pp. 186–199), in calling John Hick to task for his apparent polytheism, would have found Hick's position easier to fathom if he had attended to Muslim or Orthodox metaphysics: for example, to Ibn Arabi's account of the imaginal world which mediates between us and the One (see Henry Corbin, Alone with the Alone (Princeton University Press, 1998) ).

Another of Plantinga's arguments is addressed by Crisp (ERB, pp. 114–131), and put to novel use. Atheistical naturalism, if true, would make it very unlikely that we have the rational powers we need to deal with ‘recondite philosophical [or scientific] issues’. Whatever sensory or intellectual gifts we have were selected, by Neo-Darwinian hypothesis, because they gave a reproductive edge. This makes it likely enough that we can discriminate between potential prey and predator; we may even have sufficient foresight and self-control sometimes to defer gratification, and sufficient empathic skill to cope with social relations. What reason have we to suppose that ‘Reason’ has any more powers than that, in science or in philosophy? As Crisp (ERB, pp. 116) puts it: ‘[the probability of] the proposition that our cognitive faculties are reliable with respect to recondite philosophical issues [in particular, ‘the problem of evil’] … is low or inscrutable’! This need not – or at least will not – deter philosophers too long. Even if we cannot know the truth, the pursuit (as the sceptics say) is worth the trouble.