The Employment of Intrinsically Defined Representations and Functions

Abstract

Nearly all of the ways philosophers currently attempt to define the terms "representation" and "function" undermine the scientific application of those terms by rendering the scientific explanations in which they occur vacuous. Since this is unacceptable, we must develop analyses of these terms that avoid this vacuity.Robert Cummins argues in this fashion in Representations, Targets, and Attitudes. He accuses "use theories" of representational content of generating vacuous explanations, claims that nearly all current theories of representational content are use theories, and offers a non-use theory of representational content which avoids explanatory vacuity. According to this theory, representations are physically instantiated structures, and represent whatever other structures are isomorphic to them, regardless of how or whether these structures are used by some cognitive system. Unfortunately, since isomorphism is a rather weak constraint, Cummins' theory underdetermines representational content so severely that it too undermines explanatory appeals to representation. One task I undertake is to develop an alternative non-use theory which avoids this difficulty.My second task is to adapt Cummins' argument to criticize most current analyses of "function," which undermine scientific explanation in an analogous way. Though Cummins does not explicitly argue in this manner, his own analysis of "function," by avoiding any appeal to use, avoids the explanatory vacuity to which they succumb. Consequently, I endorse Cummins' notion of function, both as it appears in cognitive science, and elsewhere. However, although use theories fail as analyses of the terms "representation" and "function," I argue that they can still make significant contributions to the sciences employing these terms. For, while philosophers seeking to define "representation" and "function" must avoid incorporating representational and functional uses into their definitions, scientists must still find a way to determine which representations and functions are being used. Suitably re-construed use theories of representation and function may in many cases assist them in this task by providing principles for theory choice in the face of empirical underdetermination of facts about representational and functional use.