Abstract
Some proponents of structural representations (henceforth, structuralists) claim that no other theory of representation can legitimatize the explanatory appeals that cognitive science makes to mental content. Because other naturalistic approaches to representation purportedly posit an arbitrary relation between representing vehicles and representational content, these approaches must appeal to the role played by a representation, i.e., how it is used by the system in which it is embedded, to ground its content. This is in supposed contrast to structural representations, in which the relation of resemblance results in a non-arbitrary relationship between vehicles and content. Structuralists argue that, as a result, approaches that posit structural representations can, and alternative approaches cannot, explain how representational content can be causally relevant in the production of behavior. In this paper, I will argue that structural representations are susceptible to the very same critiques that proponents level against what they sometimes refer to as “use” theories. This, I contend, is not surprising given that a theory of structural representations is, in fact, just as much a use-theory as alternative approaches.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Availability of Data and Material
Not applicable.
Code Availability
Not applicable.
Notes
In fact, as some philosophers have pointed out (Ramsey, 2016), what would otherwise be viewed as competing accounts of representation need not be seen as such, if we were to take them as distinct, and complementary, answers to these two questions. See Piccinini (2020; 2022) for a defense of just such a hybrid account, focusing on structural resemblance’s capacity to answer the representational role question but not the representational content question.
The terminology of “use” theories may be somewhat misleading in that the same terminology is sometimes employed to refer to consumer based teleosemantic theories (e.g., Millikan (1984)). In the current context, “use” is intended to refer to a much broader range of theories, specifically, all naturalistic approaches to representation in which content is not grounded in structural resemblance. However, it is worth noting that this shared terminology is not coincidental. Robert Cummins initially employed the term in the context of an argument to the effect that all non-structuralist naturalistic approaches, when examined more carefully, must appeal to consumer systems to fix representational content (Cummins, 1996), and so, in that sense, are all use-based theories.
This is sometimes referred to as the job description challenge (See Ramsey (2007)).
I call these structuralists moderate because there are structuralists who aim only to develop an account that satisfies NATURALISM. That is, a weak structuralist sees their goal as offering a naturalistically legitimate way to talk about resemblance in the context of mental representations (e.g., Isaac (2013)).
Gładziejewski & Miłkowski’s focus on establishing a theoretical distinction between s-reps and u-reps may make it seem as if they are more accurately described as moderate structuralists. However, I place them in the strong camp given their explicit arguments for the need for representationalists to move away from indicator or detector-based theories and towards structural representations. According to Gładziejewski & Miłkowski, the former theories cannot help themselves to the kinds of resources that allow structural representations to satisfy EXPLANATION.
See Clark and Toribio’s (1994) discussion of representation hungry situations for one prominent strategy.
Gładziejewski & Miłkowski (2017, p. 342) also appeal to interventionalist considerations. For instance, had we intervened on the resemblance between the map and the city, Annie would fail to navigate the city successfully.
Taking the thermostat’s bi-metallic strip to be an s-rep is not uncontroversial. Some structuralists explicitly argue that bi-metallic strips are not s-reps (e.g., Gładziejewski & Miłkowski (2017)). Others explicitly defend the bi-metallic strip as an s-rep (e.g., O’Brien (2016), Williams and Colling (2018, p. 1947) seem to endorse O’Brien’s argument). The use of a bi-metallic strip as an example s-rep is not crucial to my argument in Sect. 3 (that is, the form of the argument should apply to any paradigm s-rep), but I keep it nonetheless. In part, this is to maintain consistency for a discussion in Sect. 4 that appeals to a pre-existing dialectic regarding bi-metallic strips.
Calling this the content grounding relation may be misleading, as resemblance is only a partial ground of content. The reader should keep in mind that what this relation is intended to refer to is whatever relationship obtains between a representational vehicle and its content, whether it be resemblance, causation, indication, etc.
And had we, for instance, intervened on the tokening of the state and its being caused by, or carrying information about, an object, the frog would similarly fail to engage in successful tongue flicking behavior (see Nirshberg and Shapiro (2021).
According to Gładziejewski & Miłkowski, explanations that invoke structural representations are causal explanations because a system that exploits structural similarity uses a strategy whose success is causally dependent on the obtaining of a similarity relation between the representing vehicle and what is represented (Gładziejewski & Miłkowski, 2017, p. 340).
This is perhaps most obvious on a view such as Dretske’s (1988), where at some point in the past, a state is a candidate for representational status in virtue of indicating flies. But only after the relevant state is recruited to modify the behavior of the frog to produce tongue flicking, does the state have the content FLY.
One might wonder whether it’s true that causal theories must posit u-reps. For instance, on an account such as Fodor’s (1990), that P represents Q seemingly depends only on whether Ps are caused by Qs, and that Ps being caused by Qs does not asymmetrically depend on Ps being caused by Rs. However, without invoking something downstream from P, causal accounts are unable to explain the successful behavior of the system towards Qs. See Cummins (1996, ch. 5) for a more in-depth argument for the claim that causal theories must posit u-reps.
Though Cummins’ critique of use theories does not invoke causality, it can be fruitfully interpreted as a version of the circularity critique. Cummins’ explicit argument is that use theories cannot explain misrepresentation or learning, because to explain misrepresentation and learning requires making a distinction between the content of a representation and how that representation is used. However, Opie & O'Brien themselves are explicit that they see their critique of use theories as equivalent to (or at least entailed by) Cummins’ critique (Opie & O'Brien, 2004, p. 3). That is, both are concerned with the role of the behavior grounding relation in constituting the content grounding relation. And though Cummins’ focus is on misrepresentation, Opie & O'Brien point out that cognitive scientists invoke misrepresentation (or representational error) to account for misbehavior.
Some philosophers have defended the ability of u-reps to satisfy the causal constraint (see, for example, Rupert (2018) and Artiga (2022)). However, two points are worth mentioning regarding such defenses. First, I am not aware of any who have defended the ability of u-reps to satisfy the causal constraint specifically in response to the causal impotence or circularity critiques. Second, and more importantly, because my main claim is that s-reps also succumb to both critiques, I will forgo any explicit discussion regarding the success of such defenses of u-reps.
Cummins (1996) being a notable exception. Cummins defends an isomorphism condition and argues it is both necessary and sufficient for representation. However, Cummins, too, invokes use in his account of misrepresentation, and though it is beyond the scope of this paper, I believe his view is susceptible to basically the same critique I develop in the rest of this section.
See, also, Lee (2019) for a discussion of s-reps and the action guidance condition.
Elsewhere, while addressing the problem of content indeterminacy, O’Brien also explicitly appeals to the content limiting role of interpretation (2016, p. 11).
There are theories of content that do claim (or could be reasonably interpreted as claiming that) use fully determines content. Examples include functional role semantics (Block, 1999), action guidance theory (Rosenberg & Anderson, 2008), interactivism (Bickhard, 2009), success semantics (Nanay, 2013; Whyte, 1990) and an early version of Millikan’s consumer system approach (1984). Perhaps the circularity critique does apply to such theories in a way that it does not to u-reps. However, as long as there are some u-rep approaches in which content is only partially determined by use, then the move to the content limiting, as opposed to determining, role of use will not help establish that all u-rep approaches succumb to the circularity critique. Further, recall that my main aim is not to defend u-reps from the circularity critique, but to show that it applies equally to s-reps. As I continue this section, the reader should keep in mind that my claim is that the circularity critique applies (or does not apply) equally to u-rep and s-rep approaches in which use plays a content limiting role.
The same is of course true for the use theorist.
Strictly speaking, this too strong. However, it’s quite improbable that Annie successfully navigates Chicago without this antecedent knowledge. For instance, suppose Annie had no idea that the map was a map of Chicago. In such a case, its resemblance to Chicago would not cause Annie to navigate successfully through Chicago, as she would not think to use it for that purpose. That is, Annie would not constitute a suitably arranged system with respect to the map, thereby not allowing the map to play a causal role in producing appropriate behavior.
Or because the relation obtained in token frogs during the evolutionary history of the species.
Though one way of interpreting Isaac (2013, p. 700) is that second-order resemblance is created rather than revealed by the processes that operate over the vehicles of content.
They think this is true in virtue of the way that the successful performance of s-reps depends on structural resemblance while with u-reps this is not the case.
See, also, Facchin (2021). Facchin takes this to show that structural representations thereby fail to satisfy the job description challenge (2021, p. 5495). According to Facchin, since indicators are merely causal mediators, and indicators fail to satisfy the job description challenge, then if indicators are s-reps, then s-reps fail the job description challenge, too. This is a similar conclusion to the one I reach in this paper, though by different means.
References
Artiga, M. (2022). Strong liberal representationalism. Phenomenology and the Cognitive Sciences, 21(3), 645–667.
Beer, R. (2003). The dynamics of active categorical perception in an evolved model agent. Adaptive Behavior, 11(4), 209–243.
Bickhard, M. H. (2009). Interactivism: A manifesto. New Ideas in Psychology, 27, 85–89.
Block, N. (1999). Functional role semantics. In R. Wilson & F. Keil (Eds.), The mit encyclopedia of the cognitive sciences (pp. 331–332). MIT Press.
Brooks, R. (1991). Intelligence without representation. Artificial Intelligence, 47(1–3), 139–159.
Chemero, A. (2001). Dynamical explanation and mental representations. Trends in Cognitive Sciences, 5(4), 141–142.
Clark, A., & Toribio, J. (1994). Doing without representing? Synthese, 101(3), 401–431.
Cummins, R. (1996). Representations, targets, and attitudes. MIT Press.
Dretske, F. (1988). Explaining behavior. MIT Press.
Facchin, M. (2021). Structural representations do not meet the job description challenge. Synthese, 199(3–4), 5479–5508.
Fodor, J. (1990). A theory of content and other essays. MIT Press.
Gładziejewski, P. (2015). Explaining cognitive phenomena with internal representations: a mechanistic perspective. Studies in Logic, Grammar and Rhetoric, 40(1), 63–90.
Gładziejewski, P. (2016). Predictive coding and representationalism. Synthese, 193(2), 559–582.
Gładziejewski, P., & Miłkowski, M. (2017). Structural representations: Causally relevant and different from detectors. Biology and Philosophy, 32(3), 337–355.
Isaac, A. M. (2013). Objective similarity and mental representation. Australasian Journal of Philosophy, 91(4), 683–704.
Lee, J. (2019). Structural representation and the two problems of content. Mind and Language, 34(5), 606–626.
Millikan, R. (1984). Language. MIT Press.
Morgan, A. (2014). Representations gone mental. Synthese, 191(2), 213–244.
Nanay, B. (2013). Success semantics: The sequel. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 165(1), 151–165.
Nirshberg, G., & Shapiro, L. (2021). Structural and indicator representations: A difference in degree, not kind. Synthese, 198, 7647–7664.
O’Brien, G. (2016). How does mind matter? Solving the content causation problem. In T. Metzinger (Ed.), Open MIND philosophy and the mind sciences in the 21st century (pp. 1137–1150). MIT Press.
Opie, J., & O’Brien, G. (2004). Notes toward a structuralist theory of mental representation. In H. Clapin, P. Staines, & P. Slezak (Eds.), Representation in mind: New approaches to mental representation (pp. 1–20). Elsevier.
Piccinini, G. (2020). Nonnatural mental representation. In J. Smortchkova, K. Dolega, & T. Schilcht (Eds.), What are mental representations? (pp. 254–287). Oxford University Press.
Piccinini, G. (2022). Situated neural representations: Solving the problems of content. Frontiers in Neurorobotics. https://doi.org/10.3389/fnbot.2022.846979
Ramsey, W. (2007). Representation reconsidered. Cambridge University Press.
Ramsey, W. (2016). Untangling two questions about mental representation. New Ideas in Psychology, 40(A), 3–12.
Rosenberg, G. H., & Anderson, M. L. (2008). Content and action: The guidance theory of representation. Journal of Mind and Behavior, 29(1–2), 55–86.
Rupert, R. D. (2018). Representation and mental representation. Philosophical Explorations, 21(2), 204–225.
Segal, G., & Sober, E. (1991). The causal efficacy of content. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 63(1), 1–30.
Shea, N. (2018). Representation in cognitive science. Oxford University Press.
Van Gelder, T. (1995). What might cognition be, if not computation? Journal of Philosophy, 92(7), 345–381.
Whyte, J. T. (1990). Success semantics. Analysis, 50(3), 149–157.
Williams, D., & Colling, L. (2018). From symbols to icons: The return of resemblance in the cognitive neuroscience revolution. Synthese, 195(5), 1941–1967.
Acknowledgements
I am very grateful to the following for their comments and invaluable feedback on earlier drafts of this manuscript: Christopher Blake-Turner, Hayley Clatterbuck, Hubert Marciniec, Farid Masrour, Michael Schon, Lawrence Shapiro and four anonymous referees. Thank you also to the members of the MadMindLab at the University of Wisconsin-Madison for discussion of an early draft of the paper.
Funding
The author did not receive support from any organization for their submitted work.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author has no conflicts of interest to declare that are relevant to the content of this article.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Nirshberg, G. Structural Resemblance and the Causal Role of Content. Erkenn (2023). https://doi.org/10.1007/s10670-023-00699-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10670-023-00699-y