The poverty of the disembodiment of embodied cognition
Wednesday, June 15, 2016
The #symbodiment special issue of the Psychonomic Bulletin and Review empaneling “Arguments about the nature of concepts” embodies much that is salutary about our field—vigorous, sometimes heated, criticism of recent creative attempts to ground cognition.
How symbols are grounded to their worldly referents has long been recognized as a central problem for psychology by scholars such as Harnad and Skinner. The problem is exacerbated for the more abstract concept of concept. What the last decade has offered, summarized in the #symbodiment issue, are research programs and clarifications that go well beyond academic speculations over brains in vats in Chinese Rooms.
But not all are convinced, including my colleague Stephen Goldinger and his students, who offer a “ruthless skeptical scrutiny of all ideas, old and new”, in their enthusiastic disembodiment of cognition in their contribution to the #symbodiment issue. Under the premise that “cognitive science examines mental life”, the authors open with a fable: A young woman on her daily rounds is “thinking about a Mother’s day gift, remembering to pay bills, and …mixing daily concern with novel routines”. The story is then limned in familiar cognitive terms—“categorizing… awakening of stored memories and associations…attention waxes and wanes … language … ‘inner dialog’ … private thought”. Alas, none “of this life of the mind…can be plausibly explained, or even meaningfully addressed, by the principles of embodied cognition”.
Each turn in the journey presents new hurdles that embodied cognition (EC) cannot compass, let alone surmount; but cognitive psychology has names for each of them, and has even studied them in the laboratory. That picturesque defeat of EC leads to a litany of shortcomings of the approach: The basic principles are “old”, “often co-opted from other sources, such as evolution”; are “vacuous with respect to virtually all cognitive phenomena”; and “vague so that model building is not feasible”; and so on.
I was bemused by Goldinger and colleagues’ account of the cognitive progress and vicissitude of a young woman’s mind—the examination of which is seldom a simple task—and agree that little of her Joycean walkabout could be predicted by embodied cognition. That is not what embodied cognition does. (Neither is it what most of cognitive science does; but that is another matter.) Their paper exemplifies a classic category error. All of their praise for cognitive science concerned the admirable progress it has made in linking theoretical accounts to behavioral measures. I share that admiration.
Embodied cognition does not offer itself as an alternative to those formal accounts. It is not about how to measure and model mentation: It is about where it occurs. To fault embodied cognition as the authors do for not explaining that the woman noticed that someone or something had cleared street litter, is like faulting cognitive neuroscience for not explaining it. These are different, often complementary, levels of discourse. When they do overlap, important value is added, as ably documented for example by Kemmerer.
What of that photo-ID problem showing the amazing difficulty of matching faces to driver’s licenses, that bill-boarded a page of their paper? Well, how does cognitive psychology solve this deep problem? Goldinger and associates talk about cognitive research that talks about face perception. The problem has, to a large extent, been solved—not by cognitive psychology, but by deep-learning nets and eigen-faces; but those are not so much cognitive models as gold plated black boxes. What does cognitive psychology bring to the table? Can EC add value? We know that blocking facial mimicry can selectively impair recognition of emotional expressions. Documented: Our own facial behavior affects our perceptions of others. A step forward. Why not take another step together?
What of the claims of co-option, vacuity, and vagueness? Co-opt means to “assimilate, take, or win-over”; the example given is evolutionary theory. EC would blush to think that anyone would think it would assimilate evolutionary theory; it would blush scarlet if it were inconsistent with it; and it is proud to incorporate insights from it. That is not co-opting; it is sensible theory construction. Vacuity? Counter-examples abound in this special issue. Too vague to model? Barsalou points to many models in his #symbodiment contribution. Too many versions? Always the case in a new venture; they will shake out.
Embodied cognition is a movement that recognizes that the whole body and parts of its environment often play crucial roles that interact with other cognitive processes to guide perception and action. This interaction can range from things as mundane as counting on one’s fingers or gesticulating to facilitate and clarify speech, to actively tuning the information received by the body. Dimitriou has recently shown that the stage of learning of motor tasks determines how much information is gated to the muscle spindles from afferent neurons. From turning our heads to twitching our muscles, information processing starts at—and even extends past—the skin. Just as we read the faces and postures of others, the set of our own face and posture guides our emotions and decisions. Goldinger and Hanson showed 10 years ago that a subliminal stimulus increased subjects’ frequency of “old” responses in a memory task, presumably because they “credited it to stimulus familiarity”. That these reports are not new does not disqualify them as evidences for the EC perspective, even as that evolves to a more defined form.
By starting their critique with an introspective report, perhaps Goldinger and associates wished to focus attention on conscious cognitive processes. But those are a proper subset of cognition, as he amongst a legion has shown. If we are not drawing the semantic line at consciousness, why draw it at the central nervous system? So much that gets there is actively conditioned by the peripheral nervous system; and that in turn affects and is affected by the musculature. All such interactions are parts of the computations that guide our behavior. Such whole-body involvement in thought and action is part of cognition, and deserving to be studied under that rubric, for it is not a disembodied mind that thinks, but a body, brain and all, that does so.