Some thoughts on the coupling-constitution fallacy and the mark of the cognitive

In Adams & Aizawa’s (A&A) recent paper in Menary’s anthology The Extended Mind (2010), they put put forward a relatively straightforward (and by now, well-rehearsed) objection to the extended mind argument:

[The coupling-constitution fallacy] is the most common mistake that extended mind theorists make. The fallacious pattern is to draw attention to cases, real or imagined, in which some object or process is coupled in some fashion to some cognitive agent. From this, one slides to the conclusion that the object or process constitutes part of the agent’s cognitive apparatus or cognitive processing. If you are coupled to your pocket notebook in the sense of always having it readily available, use it a lot, trust it implicitly, and so forth, then Clark infers that the pocket notebook constitutes a part of your memory store. If you are coupled to a rock in the sense of always having it readily available, use it a lot, trust it implicitly, and so forth, Clark infers that the rock constitutes a part of your memory store. Yet coupling relations are distinct from constitutive relations, and the fact that object or process X is coupled to object or process Y does not entail that X is part of Y . The neurons leading into a neuromuscular junction are coupled to the muscles they innervate, but the neurons are not a part of the muscles they innervate. The release of neurotransmitters at the neuromuscular junction is coupled to the process of muscular contraction, but the process of releasing neurotransmitters at the neuromuscular junction is not part of the process of muscular contraction…

So, if the fact that an object or process X is coupled to a cognitive agent does not entail that X is a part of the cognitive agent’s cognitive apparatus, what does? The nature of X , of course. One needs a theory of what makes a process a cognitive process rather than a noncognitive process. One needs a theory of the “mark of the cognitive.” It won’t do simply to say that a cognitive process is one that is coupled to a cognitive agent, since this only pushes back the question. One still needs a theory of what makes something a cognitive agent. This is another weakness of extended mind theories. Yet, in all fairness to Clark and other extended mind theorists, it must be admitted that one of the shortcomings of contemporary cognitive psychology is that there is no well-established theory of just exactly what constitutes the cognitive.

A&A’s objection is clear enough, but does this spell disaster for Clark and friends? I don’t think so.

First, I think that the recent Extended Mind literature has, for some reason, forgotten about the debates about anti-representationalism in the 90s that focused, I think, on the “mark of the cognitive” in terms of Heideggerian Cognition (embodied doing) versus Classic Cognition (explicit symbol manipulation). In light of these debates, I think Clark has already given a modest sketch of the mark of the cognitive. However, it seems like Clark does not like to self-consciously engage in this particular exercise of concept construction since in most of his responses to A&A he seems skeptical of the very idea of coming up with a “mark of the cognitive”. This puzzles me, because I am of the opinion that Clark has ample philosophical resources at his disposal for coming up with a mark of the cognitive that would adequately deter A&A from making their above objections. I am thinking of one paper in particular: Clark and Toribio’s “Doing without representing?” Here is a nice passage that, I think, can be construed as laying out a sketch of the mark of the cognitive:

On our account, the notion of Representation is thus re-constructed not as a dichotomy but as a continuum. At the non-representational end of that continuum we find cases in which the required responses can be powered by a direct coupling of the system to some straightforwardly physically specifiable parameters available by sampling the ambient environment in some computationally inexpensive way (eg. a toy car with a ‘bump’ sensor). Moving along the continuum we start to find cases in which the system is forced to dilate and compress the input space : to treat as similar cases which (qua bare input patterns) are quite unalike and to treat as different cases which (qua bare input patterns) are pretty similar. At this point, the systems are trafficking in ‘modest representations’. In addition, where such dilation and compression is achieved by the creation of a systematically related body of intermediate representations (as in the connectionist learning of such representations at the hidden unit level (see eg., P.M.Churchland, 1989) we begin to witness the emergence of full-blooded representational ‘systems’, albeit ones which remain quite unlike the classical vision of such systems as loci of moveable symbols capable of literal combination into complex wholes. (It is this classical vision of moveable symbols prone to engage in text-like recombinative antics which corresponds most closely to the vision of explicit representation which is, we claim, the proper target of many of the ‘antirepresentationalist’ arguments). And at the far end of the continuum we find cases in which the system is able to invoke various kinds of intermediate representations even in the absence of ambient environmental stimuli (ie as a result of ‘top-down’ influences.). At this point, we find systems capable of reasoning about the spatiotemporally remote etc.

Here, Clark and Toribio are giving a layercake mental taxonomy that starts with simple sensorimotor cognition, ramps up with moderate representational cognition, and achieves maturity with full-fledged representations that allow for “mental time timetravel”. I am particularly interested in the first layer: the powering of responses. This seems to me compatible with my earlier thoughts about the mark of the cognitive, where I said “I think it is reasonable to first define cognition as a regulatory or coordinating process that serves to select effective neural pathways out of internal variability.”

What do “effective neural pathways” do? First and foremost, they subserve the control of motor responses. Why should we accept this as true? Michael Wheeler asks us to consider “a compelling evolutionarily inspired thought: biological brains are, first and foremost, systems that have been designed for controlling action” (2005, p. 12). A fine example of this principle is the life cycle of the famous sea squirt. During the larval stage of its development, the sea squirt’s cerebral ganglion (the equivalent of its brain) is used to coordinate motor control for the task of finding a suitable resting place on a rock. Once attached to the rock, the sea squirt does not need its brain for controlling action and promptly digests it for sustenance and crucial energy savings. We thus have an answer to the question, “If brains are so costly to develop and maintain, what is their evolutionary usefulness?” The answer is that brains allow for a higher degree of what Wheeler calls online intelligence. “A creature displays online intelligence just when it produces a suite of fluid and flexible real-time adaptive responses to incoming sensory stimuli” (ibid.).

Accordingly, we can interpret the first layer of “powering responses” in terms of online intelligence. I believe this represents an adequate sketch of the mark of the cognitive. Coming back to A&A’s objection to the Extended Mind thesis then, I think we can plausibly argue that some of the typical case examples in the Extended Mind literature can be seen as helping regulate the production of fluid and flexible real-time adaptive responses to sensory stimuli. Otto’s notebook, for example, helps regulate his behavioral responses in a fluid and flexible way. Pen and paper helps the mathematician regulate and control his production of mathematical work. My laptop helps me regulate and control my adaptive behaviors (as a student, I must do research and write papers in order to be adaptive).

A&A’s response to these examples is that they are mere crutches, but do not constitute the cognitive process. But this objection only works if our proposed mark of the cognitive excludes the cases. I am of the opinion that these cases of extended cognition are compatible with our fundamental mark of the cognitive: the powering of responses in a functionally adaptive manner. If anything helps regulate and control the powering of behavioral responses in a reliable and automatic way, then it should be considered part of the cognitive system. The “control” criterion excludes “mere causal couplings” such as our reliance on oxygen in the air around us, since this does not control the cognitive system, but rather, causally supports its existence. I believe this criterion adequately stops worries about “cognitive bloat”. This neobehaviorist-functionalist definition of cognition is compatible with the Heideggerian paradigm of cognition. However, given the 2nd and 3rd layers of the cake that invoke representational prowess, any mature science of mind must also talk about how representations qua representations also help to regulate and control the powering of responses. In modest cases, this might involve topographic representations modulating and regulating sensorimotor connections in semisophisticated perceptual cognition. In more extreme and human-specific cases, this would involve the integration of linguistic-cultural representations into the cognitive system to help modulate and control sociocognitive behaviors. This would be a case of narratological cognition (See Dan Hutto’s Narrative Practice Hypothesis).

It seems to me then that 4EA theorists shouldn’t shy away from trying to define the mark of the cognitive. I think there are significant philosophical resources at hand sufficient for such a conceptual enterprise. My recent paper in Phenomenology and the Cognitive Sciences defending Julian Jaynes could be seen as an attempt to lay out a mark of the cognitive, starting with what I call “the reactive mind” and ending with what I call “Jaynesian consciousness”, which Jaynes defines as ““[T]he development on the basis of linguistic metaphors of an operation of space in which an ‘I’ could narratize out alternative actions to their consequences”.