284 ANALYZING DATA
beliefs by locating them in wider webs of beliefs in large part because they think, first, that beliefs
are in some way constitutive of actions and, second, that beliefs are necessarily holistic in nature.
Consider the constitutive nature of beliefs in relation to actions. When other human scientists
study voting behavior in terms of, say, surveys of the attitudes of voters or models of rational
action given certain beliefs and preferences, they thereby differentiate beliefs from actions so as
to seek a correlation or deductive link between the two. By contrast, interpretivists often suggest
that such surveys and models cannot tell us why, say, raising one’s hand should amount to voting,
or why there would be uproar if someone forced someone else to raise their hand against their
will, or why only certain people should be regarded as eligible to vote (C. Taylor 1971). To
explain these sorts of things, they continue, we must appeal to the intersubjective beliefs that
underpin the practice of concern to us. We need to know, for example, that voting is associated
with making a free choice and so with a particular concept of the self. We need to know what
counts as an infringement of free choice and who is regarded as being capable of making such a
choice. Practices and beliefs are constitutive of one another: Practices could not exist if people
did not have appropriate beliefs; and beliefs or meanings would not make sense in the absence of
the practices to which they refer.
Now consider the holistic nature of beliefs. Many interpretivists emphasize that people hold
beliefs for reasons of their own, so we can make sense of their beliefs only by locating them in the
context of the other beliefs that provide reasons for their holding them. Hence, even if human
scientists establish a correlation between, say, a positive attitude to social justice and voting for
the Democrats, they cannot properly explain people’s voting Democrat by reference to this atti-
tude; after all, people who have a positive attitude to social justice might nonetheless vote Repub-
lican if they believe still more strongly in conservative values or if they believe the Democrats
will not implement the policies they avow. To grasp why someone with a positive attitude toward
social justice votes Democrat, we have to unpack the other relevant beliefs and desires that relate
that attitude to that vote. To explain an action, we cannot just correlate it with a single isolated
attitude; we must interpret it in relation to a whole set of beliefs and desires. A wide range of
human scientists typically treat beliefs, meanings, ideas, norms, and the like as if, first, they could
be differentiated from actions and, second, they could be related individually to actions.
Interpretivists stand out in their insistence that beliefs or meanings form holistic webs that are
constitutive of actions and so of practices.
Interpretivists concentrate on unpacking the beliefs or meaning embodied in actions and prac-
tices. The concept of meaning does well here in that it has less cognitive a ring than does “belief,”
thereby reminding us to extend our concern from big ideas and movements to the subconscious
beliefs that inform habitual acts. Equally, the concept of belief does well in that it serves the
extraordinarily valuable role of challenging the spurious dichotomy between understanding and
explanation, for it echoes the micro-level commitment of rational choice theory and folk psychol-
ogy to explaining actions in terms of beliefs and desires. Interpretivists often imply that the hu-
man sciences explain actions and practices by pointing to beliefs and desires that cause them.
Models, typologies, and correlations can do explanatory work only if they are unpacked in terms
of such a narrative. This analysis of narrative challenges those political scientists who appear to
think of ideas as one kind of variable to which we might give more prominence within a constitu-
tive logic that compliments a more general causal logic (Wendt 1998). Our analysis of narrative
suggests that non-ideational variables can do explanatory work only if they are unpacked in terms
of ideas or beliefs. It implies that correlations between variables never do any explanatory work;
they merely point to a conjuncture that might inspire us to adopt a particular narrative to explain
something. It leads us to conclude that constitutive logics, far from being useful additions to more