THE FEEDBACK

By Caitlin Mace
Department of History and Philosophy of Science
University of Pittsburgh

Find Caitlin’s recently published paper in Philosophical Psychology here.

Abstract

Whether neurons represent or play a mere causal role is a foundational issue in philosophy of neuroscience. Evidence that neurons perform a representational role is weakened by the possibility of explaining experimental results by appeal to brute causal processes alone. Despite this, neuroscientists ascribe representational content to patterns of neural activity to explain experimental results. An important problem with this practice is determining which content to ascribe to the neural representation. One view is that researchers are only warranted in ascribing the content determined by particular experimental results. An alternative view is that researchers are warranted in appealing to the broader research domain to determine the content of a putative neural representation. In this paper, I argue that both are warranted; either alone is insufficient. Using optogenetics research on memory engrams as a case study, I show how researchers ascribe content to neural representations and justify their approach. Whether a particular content ascription is warranted depends on particular experimental results, the broader research domain that is appealed to, and how results from various animal models, probes, and experimental paradigms are generalized.

Commentary by Caitlin Mace

Mental representations are posited to explain how we can have thoughts, memories, and imaginings about the world and possible worlds. In theorizing about mental representations, philosophers helpfully distinguish between the content of the representations—what the representation is about—and the vehicle—the physical or physically realized particular that bears that content. It has traditionally been taken for granted by many philosophers that vehicles are somewhere in the brain waiting to be discovered by neuroscientists. Debates persisted instead about representational contents and whether a theory of representations can render contents determinate (see Fodor 1987). Without determinate contents that mental representations intuitively possess, the representational theory is unsuccessful. My paper shows how this familiar problem reappears in neuroscience.

Theories of content that raised the issue in philosophy appeal to causal or probabilistic information carried by the vehicle (Stampe 1977; Dretske 1981). The problem is that, for any particular case, many content ascriptions are fitting. To resolve indeterminacy issues, theories have appealed to what users or consumers of the representation take the content to be (Millikan 1984) or to what the information carried by the vehicle is supposed to be about (Dretske 1986; Neander 2017; Shea, Godfrey-Smith, & Cao 2018). In appealing to permissive evolutionary and developmental constraints, some have instead accepted that indeterminacy is the nature of representational content (Rosenberg 2013; Piccinini 2022). Current content indeterminacy debates have largely stalled out in appeals to scientific study of representations.

Finding and individuating representations is a major project of neuroscience. How do neuroscientists determine the contents that they ascribe to representational vehicles? Answering this question was the motivation for my paper. Measurements and interventions do not provide a straightforward readout by which neuroscientists can determine what the activity represents. Philosophers concerned with the scientific ascriptions of content to vehicles appeal to explanatory goals the researchers have as arbiters of which contents to ascribe (Egan 2020, 2025; Cao 2022). But appeals to explanatory goals raise a new issue: researchers have many explanatory goals to which they may appeal. Explanatory goals local to an experimental context will suggest that the experimental paradigm determines representational content, but broader explanatory goals will suggest that appealing to presumed functions of the targets of intervention will determine content. Both experimental paradigms and theories of brain function suggest multiple fitting content ascriptions, and often many content ascriptions are fitting and compatible with experimental results. This is the new content indeterminacy problem in neuroscience. 

I argue in this paper that neuroscientists already rely on results from adjacent research programs to constrain content ascriptions and that this practice is methodologically appropriate. I offer this argument as a critique of Cao’s (2022) recent view of representations, but one to which I think she would be amenable. On her view, representations are local to an experimental context. Her point is that appeals to other research should be made very carefully. I certainly agree and take my account of how content is made determinate in neuroscientific research to provide a basis for theorizing about content indeterminacy and generalization in neuroscience.