is preconWgured by an intuitive metaphysics which setsWrm divisions
between such categories aspeople,animals,plants, andartefacts(Sperberet
al., 1995b). Such an intuitive metaphysics will at least associate with
kind-concepts the sorts of inferential links which place them within an
appropriate hierarchical category. So, for example, I will not be able to
have such a belief as ‘Algernon is a tortoise’ unless I am ready to infer
‘Algernon is an animal’ and ‘Algernon is alive’. If I think that it is really an
open possibility that Algernon may turn out to be some sort of clockwork
toy, then although I maysay‘Algernon is a tortoise’ I will not be sharing
the belief which people usually have when they say that.
Admittedly, it does seem possible that someone – Sophie, say – should
believeoftortoises that they are not animals, but rather some sort of
mechanical device. But whether we can report her states of mind by using
the concepttortoisedepends upon the speciWc, purpose-relative interests
which apply in a particular context. For example, we can certainly coun-
tenance the report, ‘Sophie has noticed the tortoise in the garden’, since
this merely communicates the presence of a tortoise and Sophie’s aware-
ness – however conceptualised – of it. We can, perhaps, also report,
‘Sophie has noticed that there is a tortoise in the garden’, thereby at-
tributing to Sophie a classiWcation of the thing noticed in the garden as ‘a
tortoise’. But does Sophiethinkthat there is a tortoise in the garden? The
answer to this question hovers over the contexts in which content is
attributed. For predictive purposes it is just barely acceptable (because she
maysay, ‘A tortoise’, if you ask her what it is), but also seriously mis-
leading – because you cannot apply the strategy ofinferential enrichment
(see chapter 4, section 3), since Sophie will not share most of the beliefs
which people normally do about tortoises. (We will be returning to the
distinction betweencommunicativeandpredictive/explanatorycontexts of
content-attribution in chapter 6, section 5.1.)
Where there are content-constitutive inferential links, there will be
corresponding limitations on possible degrees of irrationality. To be sure,
nobody can be so irrational as to think that Algernon is a tortoise and
seriously doubt that he is an animal. (NobodyincludingSophie – who is
not irrational, but rather just does not understand what a tortoise is.)
However, because only some inferential links are constitutive of content,
the argument from content imposes no more than an ultimate backstop
against possible degrees of unreason, leaving wide tracts of irrationality
open to the ordinary thinker. At least, there is nothing in this line of
argument which would lead us to question the interpretation of psycho-
logical experiments on reasoning and inference.
116 Reasoning and irrationality