being assessed), andprocess rationality(where it is the person’s belief-
forming processes which are at issue). And as we shall see,state rationality
should probably be further bifurcated into type- and token-state
rationality.
To say of particular people that they are rational (in the domain of
belief) is, we think, to say that most of their belief-formingprocessesare
rational ones. For imagine someone whosebeliefs(or many of them) are
clearly irrational, but where those beliefs were induced by hypnotic sugges-
tion or brain-washing. Provided that the person ‘keeps their wits about
them’, as we say, and that the processes by which they now evaluate and
form beliefs are still rational ones, then we think their status, as a rational
creature, remains uncompromised.
What is it to say of a belief that it is rational? Here we think a further
distinction needs to be drawn. Imagine someone whose beliefthat plants
need waterhas not been formed in the normal sort of way, but has been
induced a-rationally by post-hypnotic suggestion. Here they have a belief
which is rational, in the sense that it is a belief of atypewhich can be
formed by rational processes. Buttheir holdingthat belief – or thattoken
belief of theirs – isirrational, since it wasnotformed by a rational process.
When we ask after the rationality of someone’s beliefs, then, we need to be
clear whether we are considering them as types, or as tokens.
It should be plain from the above that the most fundamental notion is
that of a rational belief-formingprocess, since it is in terms of this notion
that both creature rationality and state rationality are to be explained.
What is it, then, for a belief-forming process, or a method of inference, to
be rational? Plainly it would be a mistake to say, ‘When it is avalidone,
guaranteeing truth from truth’. For that would conWne to the scrap-heap
of irrationality all non-deductive modes of inference, including induction
and inference to the best explanation. We might try saying that a rational
process is, not justreliable(where ‘reliable’ means ‘not leading from truth
to falsehood’), but themostreliable available. Then where valid principles
can be employed, the use of an invalid one would be irrational. But in other
domains the use of an invalid principle of inference, such as induction, can
be rational provided that it is more reliable than the competing principles
on oVer. Although tempting, to pursue this line of explication would be a
mistake. For our goal as enquirers is not to formonlytrue beliefs, but to
formenoughtrue beliefs in a short enough time-span. And the most
reliable methods might be ones which hardly deliver any beliefs at all.
So standards of rationality for belief-forming processes should be
relativised to our needs as situated,Wnite, enquirers after truth. This is an
important point, to which we return at greater length in section 5. Others
have suggested that there may be a multiplicity of purposes in relation to
Introduction: the fragmentation of rationality 107