Is it possible to experimentally determine the extension of cognition?

Abstract

Various analytical tools originally developed for theories of mechanistic explanation have recently been imported into the ongoing debate on the hypothesis of (extracerebrally) extended cognition (HEC). One such tool that appears particularly relevant to that debate is Craver’s mutual manipulability account of constitution (MM), most of all because it promises to settle the debate on experimental grounds. This paper investigates whether it is possible to deliver on that promise. We first find that, far from grounding an experimental evaluation of HEC, MM is conceptually incompatible with both internalist and externalist accounts of cognition. Next, we propose a suitable modification of MM, MM*, but it turns out that MM* presupposes rather than produces clarity on the extension of cognition. Moreover, subject to MM* the inference to constitution is radically empirically underdetermined. Finally, we argue that our results can be generalized and conclude that, for principled reasons, it is impossible to experimentally determine whether cognitive processes have extracerebral constituents. Determining the extension of cognition is an inherently pragmatic matter.

Keywords:

• Constitution
• extended cognition
• extended mind
• mechanistic explanation
• mutual manipulability

Acknowledgments

We thank Carlos Zednik for valuable discussions and two anonymous reviewers for their helpful comments on earlier drafts. Moreover, we are grateful to the audience at the GAP.9 conference, University of Osnabrück, September 2015.

Notes

1. Note that Kaplan (2012, p. 560) misreads MM as only providing a sufficient condition for constitutive relevance. Textual evidence, however, contradicts that assessment. For instance, Craver (2007, p. 159) presents mutual manipulability as a sufficient condition for constitutive relevance and the absence of mutual manipulability as a sufficient condition for constitutive irrelevance, which entails that mutual manipulability is sufficient and necessary for constitutive relevance. If MM were to merely provide a sufficient condition, it would be strongly biased in favor of HEC’s proponents. Based on a merely sufficient condition, supporters of HEC could license an inference to constitution whenever extracerebral processes and cognitive phenomena happen to be mutually manipulable. Yet, based on a merely sufficient condition, critics of HEC could never build a case against cognition extending beyond the brain, for failures of mutual manipulability could not be taken to falsify HEC.

2. While the weakened notion of an intervention discussed in this paragraph is Woodward’s, not ours, referring to such interventions as permissibly fat-handed is our terminology, not Woodward’s. Woodward (2015, p. 334), instead, speaks of IV*-interventions.

3. Note that the claim here is not that causation and constitution can never be distinguished on evidence-based grounds. There are many causal structures that can be empirically distinguished from constitutive ones. For instance, in a causal chain , the instances of A and C are entailed not to spatiotemporally overlap; A can therefore not be a constituent of C. Rather, the claim is that for every constitutive structure, there exists one particular type of causal structure, a common-cause structure, that is empirically indistinguishable from it. More specifically, whenever A can be modeled as constituents of C, A and C can also be modeled as parallel effects of common causes, for example, .

4. As is common in scientific modeling, we assume here that modeled variable sets do not contain variables that are logically or conceptually dependent on one another.

5. Correspondingly, in scientific practice, numerous top-down and bottom-up experiments, under many different circumstances, are typically conducted on mechanisms before constitutive relations are considered established – a case in point being the study by Ballard et al. (1995). In our view, this indicates that scientists do not think that MM provides a sufficient condition for constitution.

6. This finding echoes Sprevak (2010), who argues that neither externalist nor internalist accounts of cognition have a clear edge over their rivals with respect to criteria of explanatory power or coherence with scientific practice.