Cognitive ontology in flux: the possibility of protean brains

Abstract

This paper motivates taking seriously the possibility that brains are basically protean: that they make use of neural structures in inventive, on-the-fly improvisations to suit circumstance and context. Accordingly, we should not always expect cognition to divide into functionally stable neural parts and pieces. We begin by reviewing recent work in cognitive ontology that highlights the inadequacy of traditional neuroscientific approaches when it comes to divining the function and structure of cognition. Cathy J. Price and Karl J. Friston, and Colin Klein identify the limitations of relying on forward and reverse inferences to cast light on the relation between cognitive functions and neural structures. There is reason to prefer Klein’s approach to that of Price and Friston’s. But Klein’s approach is neurocentric - it assumes that we ought to look solely at neural contexts to fix cognitive ontology. Using recent work on mindreading as a case study, we motivate adopting a radically different approach to cognitive ontology. Promoting the Protean Brain Hypothesis, we posit the possibility that we may need to look beyond the brain when deciding which functions are being performed in acts of cognition and in understanding how the brain contributes to such acts by adapting to circumstance.

Keywords:

• enactivism
• cognitive ontology
• functionalism
• neurodynamics
• pluripotency

Acknowledgements

We would also like to thank three anonymous reviewers for their helpful comments on an earlier draft of this article.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Daniel D. Hutto is Professor of Philosophical Psychology at the University of Wollongong and member of the Australian Research Council College of Experts. His most recent books, include: Wittgenstein and the End of Philosophy (Palgrave, 2006), Folk Psychological Narratives (MIT, 2008). He is co-author of the award-winning Radicalizing Enactivism (MIT, 2013) and its forthcoming sequel, Evolving Enactivism (MIT, 2017). He is editor of Narrative and Understanding Persons (CUP, 2007) and Narrative and Folk Psychology (Imprint Academic, 2009). A special yearbook, Radical Enactivism, focusing on his philosophy of intentionality, phenomenology and narrative, was published in 2006.

Anco Peeters is a Ph.D. candidate at the University of Wollongong.

Miguel Segundo-Ortin is a Ph.D. candidate at the University of Wollongong.

ORCID

Daniel D. Hutto http://orcid.org/0000-0002-3585-4618

Anco Peeters http://orcid.org/0000-0002-2380-4154

Miguel Segundo-Ortin http://orcid.org/0000-0002-3787-0013

Notes

1. This project is one of specifying “the entities that exist in [the cognitive] domain and the relations between them” (Lenartowicz et al. 2010, 679). The systematic descriptions of structure-function relations of neural systems are “referred to as ontologies” (Price and Friston 2005, 263).

2. In line with realism, it is assumed that “this mapping can only be successful if the cognitive constructs being mapped to the brain are actually implemented in the brain as separate constructs” (Lenartowicz et al. 2010, 680, our emphasis). Thus the process of articulating the correct cognitive ontology requires us to “clarify, refine, and test theories of brain and cognitive function” (Poldrack and Yarkoni 2016, 587). They are meant to “facilitate the integration of cognitive and anatomical models and organize the cognitive components of diverse tasks into a single framework” (Price and Friston 2005, 262).

3. Our proposal about the protean character of cognitive ontology fits most naturally with embodied, enactive approaches at the extreme end of the spectrum – namely, those that endorse an extensive enactivist approach to cognition (see Hutto, Kirchhoff, and Myin 2014 and Hutto and Myin 2013, Chapter 7).

4. Enactivist authors, such as Thompson (2007) and Noë (2009), stress the importance of looking at life not computers when trying to understand and model the mind. Respectively, they tell us, “life and mind share a set of basic organizational properties, and the organizational properties distinctive of mind are an enriched version of those fundamental to life. Mind is life-like and life is mind-like” (Thompson 2007, 128); and that “What biology brings into focus is the living being, but where we discern life, we have everything we need to discern mind” (Noë 2009, 41).

5. Some may be reticent about giving up on a computational, componential approach to mind because of putative past successes that the classical cognitivist framework has allegedly delivered. Take the favourite parade case of Marr’s (1982) computational theory of vision as a case in point. It is frequently taken to have heralded “the arrival of visual psychology as a maturing science” (Burge 2010, 93). It is widely regarded as both “extremely impressive and influential” (Tye 2000, 83) and as one of cognitivism's most “stunning successes” (Shapiro 2014, 2). In short, it is the canonical cognitivist success story. Yet there are well-known limitations in modelling vision in Marrian terms. A major worry is that it fails to accommodate the interactive complexity of cognition because it thinks of vision in overly linear terms. Its classic style of sequential processing is deemed too slow and rigid to properly account for the dynamically up-dated, on-the-fly character of intelligent responses. Although there have been attempts to address these issues by trying to build feedback into cognitivist stories in various ways, arguably, such developments are at best epicyclical patches lacking the advantage of more radical overhauls to cognitivist thinking (see Clark 2016, 51–52 and Anderson 2014, 166–167). Thus despite having many staunch defenders – who frequently attempt to justify sticking with classical cognitivism by pointing to its seeming success – many in the field today now regard Marr's theory of vision as marred by serious explanatory limitations and empirical inadequacies.

6. In some of Anderson's earlier work a non-protean reading of reuse is present, but we are following Anderson's later work here which understands reuse more explicitly as protean. Poldrack and Yarkoni (2016) also favour these more revisionary approaches. Thanks go to an anonymous reviewer for pointing this out.

7. Of course, talk of the brain anticipating future events is likely to induce a representationalist view of anticipatory neurodynamics. The worry goes like this, or something close to it: if the brain is in the business of anticipating future events, then it does not have access to such events and must therefore represent them. But this conclusion need not follow. In the literature on dynamical systems, it is common to treat two separate pendulums as coupled, and therefore as a constituting a single nonlinear dynamical system. It is relatively straightforward to show that any dynamical system A – for example, an organism – coupled to a second dynamical system B – for example, an environment – can be understood as anticipating the dynamics of B “when it reliably covaries with the dynamics of B and it is robust to the noise inherent in the coupling” (Bruineberg and Rietveld 2014, 7). This is the notion of anticipation we have in mind when saying that the brain continuously tries to anticipate unfolding events in the immediate future. If it is not representations that maintain a connection between brain and world, what is it? Work on global network dynamics and dynamical systems theory suggests that it is synergies. A synergy is an assembly (typically short-lived) of processes enslaved to act as a single coherent and functional unit (Kelso 1995). Specifically, synergies “are defined as compensatory, low dimensional relations in the dynamic activities of neuromuscular components (Kelso 2009), not as static representational structures such as motor programs” (Riley, Shockley, and Orden 2012, 23).

8. In any case, talk of states when thinking about dynamical systems appears to be an idealization at best. As Spivey argues

[Claiming] that a system was in a particular “state,” X, at a particular point in time, really boils down to saying that the average of the system's states during that period of time was X. This kind of coarse averaging measurement is often a practical necessity in science, but should not be mistaken as genuine evidence for the system actually resting in a discrete stable state. (2008, 30)

9. This way of characterizing functionalism best suits the needs of classical computational theories of mind – those that deem computational mechanisms to be a special sub-class of functional mechanisms, distinguished by the fact that they “manipulate vehicles based solely on differences between portions of the vehicles in accordance with a rule that is defined over the vehicles and, possibly, certain internal states of the mechanism” (Piccinini 2015, 1). Thus, “computational states and processes are individuated functionally, i.e. formally and syntactically” (Piccinini 2015, 2).

10. This may come as no surprise. After all, as Putnam explained long ago, when first introducing functionalism, it was only ever a framework for theorizing and advancing empirical hypotheses; in itself functionalism is “the putting-forward, not of detailed scientifically ‘finished’ hypotheses, but of schemata for hypotheses” (Putnam 1967/1992, 54). Of course, even if enactivism is compatible with a minimal functionalism, this leaves open the question of whether a minimal functionalism adds any explanatory payoff or punch to enactivism.

Additional information

Funding

The authors acknowledge that Hutto’s contribution to this article was supported by the Australian Research Council Discovery Project “Minds in Skilled Performance” (DP170102987).