Abstract

Realism and non-realism are usually construed as competing views regarding the proper interpretation of completed scientific theories; only rarely are they construed as endorsing alternative research methodologies. In my dissertation I argue that the realist/non-realist debate is as much a debate about methodology as it is a debate about the proper interpretation of completed theories: realism and non-realism endorse different methodologies that can be expected to lead to empirically different theories. Realism, I further argue, is preferable on methodological grounds because it is more likely to lead to theories that are predictively successful. ;Whatever the fortunes of realist methodology in the physical and biological sciences, one might question whether it can be expected to prove successful in cognitive psychology, since it seems to conflict with the prevailing non-reductionist tenets of that field. This view, I argue, is mistaken. Materialism may be true; psychology may not be reducible to physics: yet psychology may nevertheless be best understood realistically. The issue has to do with whether theoretical terms in cognitive psychology refer to natural kinds, even if those kinds are not physical kinds. Mental images, I argue, are psychological kinds, though not physical kinds, even though particular mental images are presumably physical states or processes. ;Finally, it is argued that there is empirical evidence to the effect that the theories of mental imagery phenomena resulting from realist methodology are in fact superior to those resulting from non-realist methodology. The former are superior not only in the light of realist criteria, but also in the light of non-realist criteria: 'mental imagery' theories have had more predictive success than 'propositional' theories