on. So we need to make a distinction between people’sperformanceas
moral agents (which will most likely fall short of the ideal for a multitude
of reasons) and their underlyingcompetencein judging what is morally
right and wrong. In a similar way, Cohen suggests that people may make
performance errors on reasoning tasks, but this should not lead to a
diagnosis of irrationality at the level of competence because the normative
criteria of rationality which we (philosophers and psychologists included)
may propound are all developed in an attempt to systematise and extend
ordinary intuitions about reasoning.
Cohen therefore claims that psychological research on reasoning falls
into four categories: (1) it may reveal conditions under which subjects are
prone to genuine cognitive illusions, in which case special mechanisms
must be postulated to explain them, and (2) it may investigate circumstan-
ces in which subjects reason poorly because of mathematical or scientiWc
ignorance; but more commonly, fallacious or deWcient reasoning may be
wrongly imputed to subjects by experimenters because the latter (3) apply
the relevant normative criteria in an inappropriate way, or (4) apply
normative criteria which are not the appropriate ones.
However, there is an important metaphysical diVerence between reXec-
tive equilibrium in the case of ethics and in the case of rationality. In the
moral case, the rightness or wrongness of conduct is (many believe)
ultimately dependent upon what humans value. But in the case of rational-
ity, the truth or falsity of beliefs about the world depends upon the way the
world is, not upon how humans think. There is thus available a standard
which is independent of people’s intuitions of rationality (if not indepen-
dent of their total belief-set), which can be used in assessing inferential
norms. We can ask, ‘Are theyreliablenorms, being such as to reliably
generate truth from truth?’ When we raise this question, we may discover
that the norms ordinarily employed are highlyunreliable.
Moreover, the argument from reXective equilibrium fails in a further
respect. For it assumes an identity of purpose behind the standards gov-
erning ordinary competence and our normative – consciously developed –
criteria. In the case of morality this assumption is justiWed. The moral
theorist and the sincere, intelligent moral agent share the same concern to
develop and apply principles which make clear how one ought to act and
why. Indeed, whenever confronted by a moral dilemma a sincere and
intelligent moral agentbecomesa moral theorist. By contrast, we cannot
assume that the objectives served by normative criteria of rationality are
exactly the same as the constraints shaping the functioning of cognitive
systems of reasoning. Normative criteria governing inference may place
overriding importance on preserving truth and avoiding falsehood. Given
suYcient time, investment, co-operative eVort, education and study, we
118 Reasoning and irrationality