Taking the grounding problem seriously

Cognitivistic approaches to human thought and behavior have been very successful in tracing, identifying, and analyzing the processes and mechanisms underlying human cognition. However, they have created and strengthened the image of humans as couch potatoes that have incredibly interesting thoughts and mental simulations but hardly translate any of those into real action. Moreover, the strategy to construe human cognition as symbol manipulation, that is, as the processing of abstract units of information that are not related to, and grounded in real-world events in straightforward and theoretically well-understood ways, has led to a growing dissatisfaction with traditional cognitivistic approaches. A promising alternative is the embodied-cognition approach that construes cognition and cognitive representations as emerging from, and as being grounded in perceptual, affective, and action-related states and processes (see Pecher & Zwaan, 2005). Ideally, the meaning of a perceived or produced event can be reduced entirely to the sensorimotor (and affective) states and processes directly involved in its perception or production, so that cognitive representations lose their explanatory overhead and become mere summaries of, or pointers to well-understood sensorimotor component processes—as in the Theory of Event Coding (TEC; Hommel, 2009; Hommel, Müsseler, Aschersleben, & Prinz, 2001). Rueschemeyer, Lindemann, van Elk, and Bekkering (2009; henceforth RLvEB) make an attempt to apply an embodied-cognition approach to the interface between language and action, and they put forward two major claims: That the new concept of ‘‘semantic resonance’’ is needed to understand how language and action control interact and that a dedicated cognitive control mechanism is needed to regulate this interaction and tailor it to the situation at hand. I strongly sympathize with the general approach defended by RLvEB because the embodied-cognition approach is healthy in forcing us (more than traditional cognitivistic approaches) to think of how mind, brain, and body interact, and how our cognitions relate to our physical and social environment, and because relating nonverbal perception and action to verbal perception and action is likely to be very productive both theoretically and empirically. At the same time, however, I have doubts whether the concrete suggestions RLvEB make really advance our understanding of embodied cognition in general and of the relationship between language and action in particular. In fact, I believe that their approach actually represents a significant setback on the way to a comprehensive theory of embodied cognition. As I will explain in the following, this is because their approach increases, rather than decreases, the gap between cognition and the sensorimotor processes that according to be embodied-cognition perspective should represent its basis and substrate. It, thus, effectively disembodies cognition and, as I will also explain, it does so without any need, that is, in the face of obvious theoretical alternatives that perfectly fit with the notion of embodied cognition. For the sake of the argument, let us take an extreme alternative and assume that the semantics of human perception and action, and of any language-related cognitive process, can be entirely reduced to sensorimotor (and affective) states and processes, as implied by TEC. Hence, let us assume that there are no separate semantic representations or processes or resonance—the meaning of a perceived or produced event would thus consist of and be embodied by the total of activated sensorimotor states. If so, one would clearly expect that semantically related objects and actions interact, simply because