Tag Archives: 4EA cognitive science

Tyler Burge begs the question against nonrepresentationalism

There is an interesting article by Tyler Burge in the NY Times philosophy blog called “A Real Science of Mind” that I happen to disgree with vehemently. He basically claims that representationalism is the only game in town when it comes to explaining visual perception. In fact, he doesn’t even hint at the fact that representationalism is but one theory, and one supported by philosophically ambiguous explanations of what it means to actually “represent” something. Indeed, he says:

Explanation in perceptual psychology is a sub-type of task-focused explanation.  What makes it distinctively psychological is that it uses notions like representational accuracy, a specific type of correlation…Why are explanations in terms of representational accuracy needed?  They explain perceptual constanciesVisual perception is getting the environment right — seeing it, representing it accurately…Perceptual psychology explains how perceptual states that represent environmental properties are formed.

Now, it seems to me that Burge has massively begged the question against nonrepresentational explanations of low-level visual perception.

In  making this claim, I put myself in a precarious position. One of the main points of Burge’s article is that vision science is a highly developed and “mathematically rigorous” science. Burge is insistent that vision science is on solid explanatory ground and I have no intention of challenging the mountain of empirical evidence gathered by orthodox representational visual science. No, the question is not about the facts, but rather, about the interpretation of the facts. It is my claim that representationalism is but one way of interpreting the empirical facts gathered by orthodox visual science.

My claim goes as follows: talk about the visual creature “accurately representing” the environment can be replaced, without losing any explanatory power, by talk of “discriminating information” in the environment. Some would say this is merely a matter of semantics, and in a way they would be right. But when it comes to philosophical explanations of visual perception, semantics are of the utmost importance. But why bother with this semantic triviality between “representation” and “discrimination”? Aren’t they the same thing? In a way, yes. But, as William Ramsey has argued in his important book Representation Reconsidered, this theoretical equivalency is actually the result of orthodox visual science moving away from classic forms of representationalism. For when a visual scientist claims that the organism “accurately” represents a feature of the environment in perception, all the explanatory work is being done by the neural workhorse that is the brain. And, naturally, this explanation is ultimately cashed out in physiological terms, against Burge’s claim that visual science is truly representational.

It is my contention that talk of “differentiation” or “discrimination” is just as psychological as talk of “representation”, but discrimination is more ontologically coherent. Take the example of a hungry primate perceiving a juicy red strawberry. Orthodox visual science would say that in successfully perceiving the strawberry the primate must have accurately represented the red strawberry as a red strawberry, (and not, say, as a purple poisonberry). This is the classic representationalist explanation. On my view, it would be more philosophically parsimonious to say that in successfully perceiving the strawberry, the primate discriminated the strawberry from out of the ambient array of energy surrounding the strawberry. Another way of putting it would be that in perceiving the red strawberry the primate attended to the information specific to the features of the strawberry that were relevant to its internal needs, namely, hunger.

On this account, the primate can be said to perceive the red strawberry as nutritious, not as a strawberry. Notice how this is starkly different from the representationalist interpretation. For the representationalist, the primate’s perception of the strawberry is cashed out in terms of how accurately the internal representation is in comparison to the objective features of the strawberry. If the primate represents the strawberry as being red, and the strawberry really is red, then the primate’s perception of the strawberry is said to be “accurate”, and thus successful. It is then said that the brain consults the representation when forming its intentions to act. Orthodox visual theory is thus committed to what some philosophers have called the sense-represent-planact model. The primate receives proximal sense-data, tries to form an accurate representation of distal stimulus, consults the representation to form a plan, and then executes a motor command to pluck the strawberry and bring it into its mouth.

On my interpretation, we can eliminate the “represent” and “plan” stages and replace it with a sensorimotor model. On this account, the task of the brain is to discriminate the meaningful information already in the environment by attending to it. Neurally speaking, the discrimination supervenes upon the neural patterns of activity. So how is this different from the representationalist story? Because unlike Burge, I think the behavioral nature of discriminatory perception is actually a plus, not a downside (and of course, behavioral explanations are a kind of psychological explanation unless we beg the question against behaviorism). So we shouldn’t expect visual neuroscience to engage in representational theorization when the proper explanatory level of description is behavioral, not representational. I have never seen a representational theory that avoided the homunculus problem without merely collapsing into descriptions of the behavior of  neurons.

And for good reason. Although Burge claims that representation is well understood by visual science, he is only half-right. Representation is well understood if by that we mean that we understand the neural underpinning and physiological correlations of the representation. But as William Ramsey has argued, this is precisely the point. Orthodox visual science has never actually successfully explained how a representation actually functions as a representation, as opposed to being a merely physiological mediator in a long chain of neural activity that ultimately leads to effective motor behavior.

So while Burge is perfectly right to say that “neuralbabble” is nonexplanatory on the psychological level, I believe he is mistaken when he claims that representationalism offers a philosophically rigorous interpretative framework that explains the phenomena at hand. Burge recognizes this when he talks about “generic representations” that apply so widely to any causal correlation as to no longer being explanatorily useful in cognitive science. To make representation explanatorily worthwhile, he introduces the notion of “accuracy”. But as I attempted to explain above, there is an alternative interpretation of accuracy available that focuses on the accurate perception of an affordance. But, crucially, the accurate perception of an affordance is entirely different than the accurate representation of an objective feature. This is because the affordance is more directly tied into the motivational circuits and can thus undercut the “represent” and “plan” stages of the sense-represent-plan-act model and jump right into the scientifically respectable arena of “sensation” and “action”. Hence, sensorimotor models of visual perception. The notion of accurately representing objective features of the environment is replaced by the accurate discrimination of information specific to invariant properties of objects which are themselves specific to affordances (opportunities for behavior). Perceiving the strawberry then becomes a matter of attending to those features of the strawberry which either past experience or innate knowledge has taught to be relevant to homeostatic needs.

Hence, we can account for the normative or “psychological” component of perception (its possible success or failure) in terms of how well the organism is capable of detecting information specific to properties that are themselves specific to affordances. And this offers us a path towards a “real science of mind”. Why? Because affordance perception is directly tied into those sensorimotor causal pathways that have been so successfully studied by orthodox visual science. And it does this without invoking a notion of one thing somehow “standing in for” something else.

Now, my representational critics will respond by saying that the discrimination of information specific to affordances is no more understood than is the notion of accurately representing the environment. Point well taken. But it is my contention that orthodox visual science has been talking about discrimination all along. So I really don’t see myself as being a “revolutionary”. I contend that we could go into almost every single visual science article and change “represents” with “discriminates” without losing any explanatory value. In fact, I think this semantical change would actually enhance the explanatory power of visual science precisely because “discrimination” is more ontologically tractable insofar as it doesn’t make a sharp distinction between the “merely mechanical” sensation of a bacterium and the “cognitive” perceptual capacities of “representing creatures”. One could say that my theory offers a “flat ontology” wherein all lifeforms are said to share in the capacity for discrimination of information and reactivity in direct response to that discrimination. Accordingly, my interpretation is immediately amenable with the advances being made in evolutionary biology.

Moreover, and most importantly for my purposes, the rejection of representationalism for an explanation of basic visual perception would leave room for those phenomena that truly deserve a representational explanation: human symbolic cognition. Indeed, in rejecting representationalism for the explanation of basic visual perception I do not reject all representational explanations like Anthony Chemero does. I thus think, following Clark and Toribio, that some phenomena are “representation hungry”, while others aren’t. Following Gibson, I do not think that basic visual perception as shared by most animals on this planet is representation hungry. What I do think absolutely requires a representational explanation is the symbolic and linguistic cognition of humans. For the referential system that is language absolutely requires an explanation of how one thing (a linguistic symbol) could “stand in for” something else. For example, the word “strawberry” cognitively stands in for a real strawberry. Now, I’m not claiming to have a complete theory worked out about symbolic cognition. But I think significant progress in the mind sciences would be made if we all recognized this demarcation between the nonrepresentational, sensorimotor cognition we share with nonhuman animals and the representational, symbolic cognition seemingly unique to humans.


Filed under Philosophy, Psychology

Sutton et al on Adams and Aizawa's critique of "revolutionary" cog sci

We are bewildered at the dialectic on which Adams and Aizawa rely. We are entirely happy to treat study of ‘the kinds of processes that take place in the brain’ as scientifically valid, and to accept intracranial cognition: we have never argued otherwise, and nor to our knowledge has Clark (nor Rowlands, nor Wilson). Cognition is not necessarily or always extended (Wilson and Clark 2009, p. 74; Sutton 2010, p. 191; Rowlands 2010). And even when it is extended, the brain remains a unique part of the extended system, performing operations which are distinct from (though complementary to) those of the external resources (Clark 2010a; Sutton 2010). Adams and Aizawa misunderstand the nature of the extended cognition thesis: the revolutionary flag which they belittle is not one we have ever saluted. In defending a complementarity-based case for extended cognition, neither we nor Clark ally ourselves with radical anti-cognitivism, whether of dynamicist, enactivist, phenomenological, or Wittgensteinian stripe. This is why our version of the thesis has real bite. We may adopt some of the constructive (rather than the critical) aspects of these movements, but ultimately we are playing the same game as Adams and Aizawa: we too maintain a version of the representational-computational theory of mind, even if ours is a somewhat revised and amended version (Clark 2001b).This is why complementarity-based theorists of distributed and extended cognition are in turn sometimes criticised by more extreme anti-cognitivists for “not proposing that the very idea of cognition is itself a mistake,” and because we do “not renounce cognitive science” (Button 2008, pp. 88–89; compare Malafouris 2004, Dreyfus 2007). While we respond vigorously to such critiques, and seek more precisely to differentiate our views from these truly radical alternatives, these critics do in these respects characterize our position more accurately than Adams and Aizawa, who wrongly think that the hypothesis of extended cognition requires wholesale rejection of intracranial cognitive processes and their neural and psychological study.

So Adams and Aizawa first treat extended cognition as a “revolutionary” thesis which denies intracranial cognition, and then suggest that complementarity fails to deliver on the revolutionary promise. They are thus seeking to trap the extended cognition theorist in a dilemma: either maintain the extreme “revolutionary”  position, or collapse back into individualism. But we reject the alleged dilemma. Along with Clark and the others, we inhabit a rich middle ground, one which this paper continues to develop, which is entirely distinct both from internalist forms of cognitivism and from externalist anti-cognitivism. Yet when Adams and Aizawa do accurately acknowledge that our views are not anti-cognitivist, they try to assimilate us to a more conservative internalism. They lump Sutton’s treatment of memory together with the work of Lakoff and Gallagher on embodied cognition as examples “of a non-revolutionary approach” (2008, p. 179). Their aim is to deny the existence of that middle ground, and to assimilate any view which is not radically anticognitivist to a much more orthodox individualism. Sutton’s project, they say, ‘can be undertaken while leaving much of the cognitive psychology of memory as the study of processes that take place, essentially without exception, within nervous systems’ (2008, p. 179). We disagree: this reversion to internalism is not an implication of Sutton’s view. As the cognitive psychological research on memory which we describe below demonstrates, the scientific study of memory is not and should not be restricted to the examination of processes occurring within the brain.

link to online version

link to academia.edu version


Filed under Philosophy, Psychology

Skeptical about Chemero's interpretation of Gibson

Chemero makes a curious statement about Gibson in a recent paper:

The claim that the cognitive system is not in the head at all, that cognition is to be explained entirely in terms of the interactions of whole animals and their environments, may seem like an automatic nonstarter and an idea so crazy that no one would have held it. That is not so. Skinnerian behaviorists still make claims like this (Hineline 2006), and the later work of Gibson (1979) can be interpreted as making claims like this. In both cases, the claim is that all of the explanatory work can be done by carefully studying the ways active animals interact with the environment. In the Skinnerian case, one focuses on the subtle ways that animal behavior is shaped by environmental outcomes, and claims that reinforcement learning can account for the whole gamut of behavior.In the Gibsonian case, one focuses on the breathtaking amount of information available to a perceiver, especially one that is moving, and claims that this information is sufficient for perception of the environment without the addition of information stored in the brain. Note that neither Gibson nor Skinnerians claim that the brain is not importantly involved in cognition; rather they claim that psychologists can do all their explanatory work without referring to the brain.

First claim: ecological information in the environment is sufficient for perception of the environment.

This sounds weirdly put to me. I think a better way of stating Gibson’s claim would this: the discrimination of ecological information from the environment is sufficient for the control of action in that environment. The point is not that “cognition ain’t going on in the head”, but rather, that the brain doesn’t need to build an internal model of the environment in order to successfully navigate it. If we define cognition as the “coordination of motor control”, then it seems metaphysically plausible that this is taking place inside the head, without violating the spirit of Gibsonian psychology.

Second claim: Gibson thought that psychology can be complete without referring to the brain.

I’d like to see some solid textual evidence supporting this interpretation. I do not recall Gibson ever saying such a ridiculous thing. If we want to give a complete account of how information is discriminated then we are going to have start talking about the brain. Gibson never talked about the brain only because so much more conceptual work needed to be done. It is only after we have laid down the basic foundation of affordance theory that we should investigate the brain, otherwise we might be prone to improperly determine the computational task. Gibson was against cognitivist theories only because he thought they were positing unnecessary steps in the computational process. For example, we don’t need to compute information about depth from a 2D picture because the information specific for depth is already available in the ambient optic array.

However, if we redefine cognitive computation in terms of controlling action-perception cycles, then there is a real sense in which the brain is computing information. Rather than doing the computation in terms of discrete mental representations and internal, lingua-form models (that get experienced as a rich picture of the external world),  the computation of information is done in terms of sensorimotor connectivity. On Edward Reed’s selectionist account and Robbins’ subtractionist account, Gibson’s affordance theory actually makes empirical predictions about how variable neural activity is coordinated so as to produce functionally specific responses to changing environmental demands. Gibsonian information theory actually predicts that the brain will seek out invariances in the environment so as to “select” the most functionally advantageous course of action from out of the “virtual phasespace” of the neural population dynamics. This approach turns out to endorse a theoretically deep model of decision-making and attention-control at the prereflective level. Moreover, such a selectionist account is self-consciously compatible with evolutionary biology and developmental systems theory insofar as the emphasis is on plasticity and adaptation.

So, far from being a psychological theory that completes itself without referencing the brain, Gibsonian theory actually provides a radical new approach to understanding small and large scale brain dynamics. Once we have done the hard theoretical work of determining what kind of affordance information is available in the environment, we must then look into the brain and determine how the discrimination of such information allows for the coordination of neural dynamics in such a way as to bring about functionally specific, adaptive behavior in response to changing environmental demands.


Filed under Psychology