Social scientists (most recently John Weeks) tend to reply to arguments abo
ut epistemological difficulties for belief ascription with two responses. A
n epistemological response goes: Behavioral evidence for underlying beliefs
'is all we have and all we need'. A metaphysical response argues that beli
efs, being merely instrumental entities are not really subject to worries a
bout underdetermination. In this article, I argue that these responses are
inadequate, and that social science practice can be greatly improved by inc
orporating cognitive-science constraints into the practice of ascribing bel
iefs.