Paco Calvo & Toni Gomila, Handbook of Cognitive Science. An Embodied Approach, Amsterdam, Elsevier, 2008

“Is cognitive activity more similar to a game of chess than to a game of pool?” This is the opening question of this volume and every social scientist concerned with the explanation of our decisions should carefully consider the answer. At least, they should if they use standard intentional explanations, where decisions result from a particular combination of beliefs and desires that purportedly captures our folk understanding of action. If we are not uncomfortable with such foundation is mostly thanks to the progress of cognitive science that shows how our beliefs and desires can be processed, beyond folk psychology, as “a computational manipulation of representational inner states”. If you are already wondering if there is anything else to a decision, you probably consider cognitive ability akin to a game of chess. The authors in this volume would rather see it as a game of pool, that is, a non-formal game in which you need to take into account real-time physical interactions. In the case of decisions, our sensorimotor interaction with a given environment plus our social interaction with other agents. All this conceived as a continuous process that should be modeled (and explained) as such: i.e., describing the range of changes that the agent-cum-environment system experiences over real time. In principle, there is no need to invoke standard mental representations or a global plan of action.

This seems to be the explanatory approach emerging in the interdisciplinary field of embodied cognitive science, at least according to the editors of this Handbook (p.13). Calvo and Gomila are well aware that not every author in their volume would accept such an approach to explanation. The aim of this compilation is precisely to bring the different agendas in this new field to converge on a joint research program (p.15). Among these agendas, the editors cite: ecological psychology, behavior-based AI, embodied cognition, distributed cognition, perceptual symbol systems, some forms of connectionism, interactivism and dynamical systems theory. Their common thread, according to Calvo and Gomila, is to conceive of cognition and behavior “in terms of the dynamical interaction (coupling) of an embodied system that is embedded in the surrounding environment” (p. 7). The reader is properly warned that many of these terms are still awaiting a more precise definition ―including here “embodied” (p.12)―, but Calvo and Gomila believe that the success obtained by this approach in certain particular domains justifies a generalization that would first redefine the research agenda of cognitive science. And then eventually expand into every other field in the social sciences where cognition plays an explanatory role.

The structure of the volume somehow reflects the current disunity of this project: it goes through the fields listed above, including several surveys, a number of success stories and a few conceptual discussions of the pros and cons of this emerging approach as opposed to mainstream cognitive science. The main division, for the purposes of this review, is between the analysis of, so to speak, lower and higher cognitive processes. The former are covered in sections 2-4, namely: “Robotics and Autonomous Agents”, “Perceiving and Acting” and “A Dynamic Brain”. These three sections exemplify several tenets that the editors present as distinctive in the embodied approach. For instance, the claim about perception being active and action perceptually guided is explored in chapters dealing with a control system for human avatars (ch. 8), an analysis of the use of inconsistent visual information for the control of our actions (ch. 11), experimental evidence on visual processes guiding sorting tasks (ch. 10) and, finally, a dynamical system model of the interaction of the neural network, the body and the environment of an evolutionary agent featuring visually guided object discrimination (ch. 6).

The evidence presented in these three sections is fascinating, at least for readers like me without any competence in the topics addressed therein. However, it is not presented in the systematic fashion you would expect from a Handbook. It is more a collection of papers representing the diversity of perspectives announced in the Introduction, but they rarely engage with the claims made by each other. The editors have a point when they call for an empirical comparison of the different post-cognitive hypotheses in order to ponder their merit within the joint agenda (p. 15). But such comparison rarely features in the Handbook, which is perhaps an accurate portrait of the state of the art in this field.

Nonetheless, we should grant that the evidence accumulated at these lower levels of cognitive activity is compelling enough to reconsider several traditional tenets about them. E.g., whereas in the traditional approach (both in philosophy and in cognitive science) vision was most often understood as yielding “internal representations for general-purpose use”, the brick-sorting experiment presented in chapter 10 compellingly suggests that eye movements are task-oriented instead. The evidence for this hypothesis is provided by an experimental setup in which subjects operate in a virtual environment wearing a head mounted display tracking their eye movements and manipulating a mechanical arm with their hands. Variations in the visual cues of the bricks during the sorting task revealed, for instance, that the subjects retrieved the relevant information either from the scene or from their working memory. An implicit cost function regulating visual attention seems to be at work here, even if we still do not know much about the mechanism implementing it. It probably evaluates such aspects as metabolic cost, cognitive load, temporal urgency, etc. The subjects themselves are certainly unaware of it being at work. According to Calvo and Gomila (p. 12), in this experiment perceptions seems to be more than building visual representations: it seems active and guides action in quite a straightforward manner.

However, as the authors of chapter 10 (Droll and Hayhoe) point out the evidence presented is not contradictory with “formal models of executive control in which high-level decision processes [about the relevant visual parameters] affect lower level sensory selection” (p. 202). In other words, these experiments can also be interpreted as speaking for a certain continuity/compatibility between embodied and traditional approaches to cognition. The former may well help us in reconsidering certain low level cognitive activities, but maybe at a higher scale the latter may still play a role. This is the problem that the editors dub “scaling-up”: can we explain high level cognition in an embodied fashion? This is the topic of sections 5-7, which cover “Embodied Meaning”, “Emotion and Social Interaction” and a general discussion of the transition from lower to higher levels of cognition.

In chapter 15, Lotte Meteyard and Gabriela Vigliocco present a wonderful review of the embodied theories of semantic representation. As they recall, in this approach, we apprehend linguistic meaning simulating the sensory-motor information produced by the referent of a word or a proposition. They distinguish between stronger and weaker versions of this approach according to the degree to which semantic content depends on this sensory-motor information (what the authors call their engagement hypothesis), reviewing the available evidence (namely behavioral and neurological) for or against each version. The authors conclude that there is a tie between them, but the evidence speaks against those who deny the engagement hypothesis and claim absolute independence between semantic and sensory-motor information. Chapter 16 presents one particular approach to embodied meaning, stemming from the Neural Theory of Language project, taking concept learning as case in point. The two remaining chapters in this section on embodied meaning deal with mathematics: in the former, Rafael Núñez applies a metaphorical approach to mathematics he developed with Lakoff to the analysis of axiomatic systems; in the latter, Arthur Glenberg draws out the practical implications of this approach for the teaching of mathematics. This Handbook is mainly aimed at practitioners of the cognitive sciences, but, all in all, this is probably the section that impinges most on the main tenets of mainstream analytic philosophy and I miss a straightforward discussion of the philosophical “paradigm shift” implicit in its claims.

There is quite a contrast between sections 6 and 7. In the former, on “Scaling up” , two of the three papers compiled seem quite deflationary, at least if we measure by the standards of the editors. In chapter 19, Margaret Wilson explores the possible mechanisms by which abstract de-contextualized thought may have emerged from sensorimotor abilities applied to immediate situations. However, she argues explicitly against reducing human cognition to situated cognition, which, in principle, leaves some room for traditional approaches to the former. In a similar vein, Michael Anderson (ch. 21) analyses brain imaging results showing cognitive overlaps between different areas of the brain and discusses to what extent these images speak unambiguously for embodied cognition. E.g., there is evidence that perceiving objects an object names activate brain regions associated with grasping. But this may be explained as a result of the redeployment of neural circuits across different domains in the evolution of our brain. Some sort of functional inheritance would often ensue as a result, without any further implication about their “embodied” connection.

The tone in the two papers compiled in section 7 is inflationary, by contrast. For instance, Shaun Gallagher (ch. 22) argues for an embodied alternative to standard theories of mind, in which we would not need belief or desire attribution to understand each other’s actions. This understanding would often be primary, originating in body expressions that we would apprehend directly through perception without mental representations. Gallaguer’s paper puts forward a different worldview than the sort of empirically informed hypotheses that abound in this volume. However, it is worth reading, even if just to have a flavor of what a fully embodied approach would entail ―even more so for M. Sheets-Johnstone final chapter.

The volume ends abruptly or, at least, I miss a final overview taking stock of all the evidence compiled and assessing the viability of the research program outlined by the editors in the introduction. The only general discussion can be found at the beginning, in the first two chapters ―therefore written without any explicit reference to the volume. M. Bickhard presents a conceptual argument against standard models of representation in cognitive science: they cannot account, he claims, for the possibility that the organism detects and corrects its own errors. Following Bickhard, if we ground representations on embodied interaction instead, it is possible to account for errors. Interactions involve a circular causal flow between the system and the environment, according to a range of indicated possibilities. Errors will be detected by the system when this range is violated in the interaction. Again, we may wonder how this error-detection model applies to higher level representations, but the volume is not very rich on suggestions about this particular point.

Hence, the only really general discussion of the project in the volume is Andy Clark’s paper on “Embodiment in explanation” (ch.2). Clark defends a somewhat conservative position: mainstream cognitive science should take into account the many findings of the embodied approach, without abandoning its current paradigm. Clark’s argument is based on a review of a significant sample of current research. Had he used the evidence compiled in this Handbook, it would have made an excellent conclusion. His conservatism originates in his skepticism regarding the possibility of a total identification between an agent’s experience and the underlying sensorimotor exercise, as it is often assumed in the most radical versions of the embodied approach ―for instance, the connection between bodily experience and our basic conceptual repertoire, as it is sometimes presented by Lakoff and Johnson.

This Handbook certainly feeds Clark’s skepticism. Despite the effort of the editors, I cannot discern in the papers compiled the possibility of building a general paradigm for cognitive science, impinging on the very foundations of our many theories of social interaction. But I may be just short-sighted. Nonetheless, it is a good invitation to rethink many deeply rooted assumptions across the social sciences.

{February 2011}
{Philosophy of the social sciences }