Pieces of Mind

When reading claims such as ‘The root cap is an example of plant consciousness or awareness’ (Trewavas [2016]) or ‘Representation of confidence associated with a decision by neurons in the parietal cortex’ (Kiani and Shadlen [2009]) in peer-reviewed scientific journals, most of us would resist a literal interpretation of psychological terms in these contexts. Plants and neurons lack the complexity of human psychological agents and the use of psychological concepts to describe their behaviour might appear to be a category mistake.

In her thought-provoking book, Figdor challenges our belief that humans are the privileged bearers of psychological capacities and outlines a theory of the semantics of psychological terms, called ‘literalism’, which accounts for scientists’ uses of these terms to describe non-humans. According to literalism, psychological terms (such as believing, preferring, choosing, and so on) are being ‘used literally with the same reference across human and nonhuman domains’ (p. 184).

In addition to literalism, which is a semantic thesis, Figdor also sketches an ‘anti-exceptionalist’ metaphysics (p. 5), according to which psychological capacities can be extended to non-humans. However, her main project is one of conceptual revision (p. 1), which takes the scientific use of concepts as having priority over the common-sense use of concepts.

To make her case, Figdor uses examples of applications of psychological terms to plants, bacteria, fruit flies, and neurons. As this list shows, she is interested both in cases of horizontal extension of psychological predicates (from humans to non-humans) and in cases of vertical extension (from humans to parts of humans). She suggests that such extension is usually done via analogy: via qualitative analogy (Chapter 2), based on similarities between behaviour and structures; and via quantitative analogy (Chapter 3), based on the extension of a mathematical model. In her view, quantitative analogy is superior to qualitative analogy, because the latter does not challenge the ‘anthropocentric semantic standards of predicates’ (pp. 29–30), while quantitative analogy does. The success of quantitative analogy in extending psychological predicates to non-human domains is explained by the success of mathematical modelling in science. Figdor focuses on two case studies: the extension of a decision-making model (the drift-diffusion model) from humans to fruit flies, and the extension of the temporal difference model of reinforcement learning (which includes an appeal to prediction, anticipation, expectation, and surprise) from humans and animals to neurons. Figdor argues that, in each case, the extension of the model is justified when the model is successful in the new domain and a terminological revision would be costly.

The core of her view is then developed in Chapter 4, where Figdor argues in favour of the sameness of reference claim: Psychological predicates ‘are being used with literal intent to pick out the same scientifically discovered structures across the relevant human and nonhuman domains’ (p. 61). This does not imply the truth of such predications, since her view is about the proper interpretation of psychological predicates and not about whether their application is correct. The chief argument for literalism is based on the existence of scientific evidence that successfully explains natural phenomena in non-human domains by an appeal to psychological capacities. If one agrees (as naturalist philosophers do) that empirical data have priority over intuitions, then the existence of this evidence shifts the burden of proof to those who think entities in non-human domains don’t have the capacities in question. Figdor also offers an inference to the best explanation based on two premises. The first premise is the existence of successful quantitative cognitive models that can be applied both to human and non-human behaviours. The second premise is an appeal to explanatory parsimony: the better hypothesis is the one that assumes that the models’ components are the same for both human and non-human cases rather than that they are different. The conclusion is the application of mental predicates to the non-human entities picked out by the models.

In Chapters 5–7, Figdor replies to three main set of opponents. The first are the defenders of the nonsense view, and in particular the neuroscientist Maxwell Bennett and the philosopher Peter Hacker, who criticize the use of psychological predicates for non-human domains (Chapter 5). They accuse some neuroscientists of committing the mereological fallacy of applying predicates that belong to whole persons to their parts (at the level of brain physiology). They also endorse human behavioural criteria for the application of psychological concepts. Figdor argues that their critique misses some crucial developments in science and especially the success of mathematical models in explaining cognitive capacities, as well as the fact that human behaviour is not the primary criterion for the ascription of psychological predicates. According to her, the successful application of mathematical models at different levels of description makes it the case that the mereological fallacy does not arise. The second set of opponents take the extension of psychological predicates to non-human domains to be based on a metaphor (Chapter 6), and the third set take the extension to signal a change of reference (a technical use) either as reference to behaviours or as reference to a quasi-capacity (Chapter 7).

Chapter 8 explores the consequences of literalism for psychological explanation and challenges central assumptions of the homuncular functionalist approach. In homuncular functionalism, the method of decomposition of a complex capacity into simpler components goes hand in hand with the gradual elimination of mentalistic vocabulary in favour of mechanistic vocabulary. The aim is to avoid the homuncular fallacy—the explanation of intelligent behaviour by an appeal to further intelligent agents—which lacks explanatory force and leads to infinite regress. For Figdor, the homuncular fallacy is not a fallacy. She draws on examples from outside psychology to show that the reappearance of the same term at different levels does not undermine explanatory power. For example, when using the same law (or model) to explain the rotation of planets and of atoms, there is no need to discharge ‘rotation’ in terms of simpler capacities when changing levels. Rotation is ‘discharged’ by finding laws (or models) that apply at all levels. Psychology is not exceptional in this respect, and psychological capacities can also be found at all levels. Thus Figdor dissociates the thesis of decomposition from the thesis of metaphysical simplicity: we can use decomposition as our primary explanatory approach in cognitive science, but decomposition should not be thought of as ‘mapping onto a mereological hierarchy of objects and their relations of basicness and simplicity’ (p. 162). For the literalist, ‘activity individuation and composition are out of sync with object individuation and composition’ (p. 163).

The case against conflating decomposition with metaphysical simplicity is compelling, but in my opinion the discussion leaves open the question of what the best explanatory strategy is when we want to obtain a fine-grained understanding of a mental capacity (for example, if we want to understand what decision-making or believing or surprise are). Compare the example of rotation with colour in physics: Colour is a property found in nature and (arguably) in need of an explanation that does not appeal to the same property (colour is explained in terms of how objects reflect light). It could be the case that, depending on our explanatory purposes and the sorts of entities we are investigating, decomposition and metaphysical simplicity sometimes converge and sometimes come apart. In other words, it seems that we should decide case by case whether certain psychological capacities do call for a more fundamental treatment.

Chapter 9 of the book is dedicated to the consequences of literalism for the moral status of both humans and non-humans. Figdor offers a conciliatory view, where literalism does not entail a shift in social and moral boundaries in the short term (p. 175), even if it might in the long term (p. 178). I found this conciliatory response a bit underwhelming. Even if a change in semantic reference does not have an immediate impact on the social and moral realms, when the semantic shift is accompanied by an anti-exceptionalist metaphysics, then the case for an extension of moral status arguably becomes stronger, even in the short term.

In general, I think that Figdor tends to underplay the revisionary consequences of her view. Literalism is, in a certain sense, the polar opposite of eliminativism (Churchland [1981]): not only are psychological terms not eliminated from mature cognitive science, their domain is expanded from the realm of psychology to the realm of neuroscience (p. 5). It would be incorrect, however, to think of literalism as vindicating folk psychology, because it conflicts with some of its core assumptions. This brings me to a point of criticism: In asking us to allow for an extension of the referential domain of psychological terms via quantitative analogies, Figdor might have left folk psychology behind more than she is willing to admit. Indeed, if our (adult human) psychological capacities are to be understood via quantitative models, rather than in terms of common-sense beliefs about the psychological capacity in question, it seems that even the starting point is already the result of a conceptual change. The quantitative models that were first applied to humans were already highly abstract versions of the psychological capacities in situ, so the psychological term that is extended is a scientific psychological concept that may itself be considered non-literal. As a result, the defender of folk psychology might view the extension from humans to non-humans as an extension from one non-literal use of a psychological concept to another.

Another point of criticism concerns the application of literalism to other domains of cognitive science. Figdor focuses mostly on plants and neurons, since they are the most counterintuitive. It would have been interesting to discuss the possible contribution of literalism to debates about the application of psychological terms to complex non-human animals and pre-verbal infants. In these domains, there is an on-going dispute about the extension of the domain of some psychological concepts, such as episodic memory to non-human animals (Clayton and Russell [2009]) or mindreading to pre-verbal infants (Heyes [2014]). The existence of these disagreements casts doubt on the extent of the consensus among scientists with respect to a literal interpretation of psychological terms when applied to non-humans (or even to humans in their early developmental stages). While the lack of consensus does not undermine the literalist thesis, I think that it calls for an in-depth discussion of the roots of such disagreements. I suspect that Figdor would appeal to scientists’ anthropocentric bias as the main cognitive obstacle to conceptual extension, but I want to suggest that the disagreement is also methodological, and not just a sign of resistance to change. This connects to a related, but more general, worry: While quantitative modelling has been successful in many areas of cognitive science, it is unclear how widespread (or useful) it is in all areas. How can we decide on the proper extension of the psychological domain where quantitative analogies are not (or cannot be) applied? Are qualitative analogies the only other option?

Pieces of Mind is an extremely stimulating read, which will interest not only philosophers of mind and cognitive science, but also philosophically minded scientists. Figdor displays an impressive knowledge of the literature in neuroscience, psychology, philosophy of mind, philosophy of language, and linguistics. She offers a unique approach to the semantics of psychological terms, which is consistent with (many) current scientific practices, and which should have a major impact on our understanding of explanation in cognitive science.