Friday, November 9, 2012

What does "cognitive" mean to you?

Just curious... what counts as "cognitive" to you? I've been reading a bit of the embodied cognition literature and I find statements like this rather odd:  "the traditional conceptualization of cognition as a stage in the perception–cognition–action pipeline."  Is cognition just high-level stuff?  I don't see it that way.  Perception is cognition.  Action is cognition.  Language is cognition.  Categorization, memory, attention, are all cognition.  Is this "cognitive sandwich" notion just a straw man given modern conceptualization of cognition?


Jon Brock said...

For me, cognition is everything between brain and behaviour.

Anonymous said...

What is a reflex? Is it a cognitive process or just one subjected to“hard laws of ecophysics”. Is the jaw jerk reflex cognitive? Is the behaviour of, say, the earthworm cognitive? And what about the “nano-intentional” behaviour of cells? Is it possible to draw a clear-cut context-free definition of cognition at all?

Strange Loops said...

Replace "perception" with "sensation" and it makes some more sense to me (as a cognitive psychologist), at least if we simplify sensation to the mere activation/transduction of peripheral receptors. But I can see how separating the action component is also problematic, even if we want to define "action" as the peripheral motor output. Clearly the efferent signals aren't just one-way output, but involved in an impressively fast process involving lots of feedback (in addition to the modulation from expectations, efference copies, interaction with other senses, experience, etc.).

I think it makes more sense to have a fairly inclusive, broad label "cognition"/"cognitive", and then break it down into various [non-independent] components when useful. So we can talk about memory as one cognitive process, perception as another, motor control as another -- in fact, this is how most cognition texts work, right? And we can still recognize that they are interconnected, and that a more complete picture is a more complicated one, where all aspects of cognition dynamically affect each other in smaller or larger ways.

Pakl said...

I define cognition, or cognitive actions, as the activation of reliable neural patterns that do not immediately generate motor output.

(Examples: mental rotation, incrementing a number in working memory, covert visual attention shifting, and individuation of entities in a plural.)

Furthermore, cognition (cognitive actions) can also be delayed in time from a triggering perceptual input, or abstracted from it.

Jessica Feldman said...

Who are you reading? There is a philosophy of cog. camp and a basic neurophysiology camp. This is precisely the area of my work - electrophysiology studies, humans. High level perception is equiv to high level cognitive processes - consider imagery for example. Perception is Is also used as external inputs- better phrased as sensation because they are low level inputs while high level perceptual-cognitive processes involve direct precepts processed at higher levels and are also nonexperiential, not accessible to consciousness. High level perceptual representations are the stuff of images, which can lead to action, or back to perceptual processes that are considered to be high level cognitive, like visual imagey.
Embodied cognition from the philosophical side has way to many variants on the same theme.

Greg Hickok said...

I've been reading Barsalou and related theorists. So it seems we have a mix of definitions of "cognitive" with some restricting it to higher-level processes and others happy to have cognition involving lower level stuff.

A couple questions that come to my mind in light of some of the comments above.

Is there a meaningful distinction between perception and sensation? (I don't think so)

For those of you who believe cognition is restricted to higher level processes, how do you decide the boundary point between cognition and non-cognitive mental operations? I would challenge you to come up with a definition that is non-arbitrary.

My ultimate point of this discussion: if we get the definition of cognition right, "embodied cognition" is a meaningless concept.


Martim said...

That is exactly my problem with the mirror neurotic community. They say: "Actions can be understood by mirroring, without any cognitive processes". This is obviously ridiculous.

But this view is apparently shared by some commentators. It seems like there is a direct route from the very earliest visual level to the lowest motor output level. And actions are understood by activating this motor output level. Hmm, this doesn't make sense. Even if this logic is true I would expect some backflow from the lowest motor level to levels were actions are planned, let's say a goal level (which then enables action understanding). I cannot see why this is not cognition?

I agree that the concept of embodied cognition has a problem here, because you need a sharp boundary. Otherwise you have to state that brain IS your body, and therefore everything in the brain is embodied. I can live with that.

Henrik said...

Pakl said...
I define cognition, or cognitive actions, as the activation of reliable neural patterns that do not immediately generate motor output.

I'd strongly disagree. Let me phrase it the other way round: Is neural activity necessary for something to be cognitive? (What about machines? Is a Turing machine a cognitive device? Does a speaking and acting robot exhibit something like cognition?) This brings me to Greg's point:

Is there a meaningful distinction between perception and sensation? (I don't think so)

I am sure that a robot will not FEEL/SENSE something we call heat or coldness. However, the robot is able to PERCEIVE the temperature in that it may change its behaviour and avoid a dangerous heat source.

What I would like to put forward is the idea of mental representation and information processing. The term 'cognition' might therefore be defined as something like as an operation on mentally represented information processes.

Anonymous Martim said...
I agree that the concept of embodied cognition has a problem here, because you need a sharp boundary. Otherwise you have to state that brain IS your body, and therefore everything in the brain is embodied.

Of course, you're brain IS not your (whole) body. But it's an integral part of it. What people in the embodiment section are claiming is that cognition wouldn't work the way it works without the rest of your body since your mind (!) is embodied.

Greg Hickok said...

I think Henrik hit on the main issue and the ultimate point of my original question. Cognition is a bad term. It means different things to different people. Some view it as only high level mental function, some view it as everything the brain does.

What we need to be talking about instead is representations and computations. From this perspective the notion of the "cognitive sandwich" (perception --> cognition --> action) that so many embodied theorists are throwing around becomes vacuous as best and misleading at worst because there is nothing to sandwich; it's all just representation and computation. Get rid of the misleading terminology and we are left with "embodied computation" which basically means that sensory and motor systems perform complex computations. We knew that already.

Henrik said...

Greg, could you give us an example of how some theorists are throwing around the 'cognitive sandwich'? I'm not sure if I got your point. Solely because it seems to be all about computation and representation, this does not imply that perception, cognition and action are the same.

I just had a look into the literature and came across an interesting and even more confusing answer to what cognition is:

"Enaction: A history of structural coupling that brings forth a world." (Varela, Thompson, Rosch 1991: p. 206)

Whew! Concerning computation, there is even the view that "cognition is best studied by means of noncomputation and nonrepresentational ideas [...]" (Clark, 1997: 148).

Greg, If you say that

Perception is cognition. Action is cognition. Language is cognition. Categorization, memory, attention, are all cognition.

What do these have in common? Maybe that might give us an answer to what cognition might be.

Greg Hickok said...

"The first and foremost challenge is that cognition cannot be studied as a module independent from other modules (sensory and motor), as suggested by the “cognitive sandwich” metaphor. Instead, cognition is deeply interrelated with sensorimotor action and affect."

-Pezzulo, G., et al. (2011). The mechanics of embodiment: a dialog on embodiment and computational modeling. Front Psychol, 2, 5.

"The classical sandwich conception of the mind – widespread across philosophy and empirical sciences of the mind – regards perception as input from world to mind, action as output from mind to world, and cognition as sandwiched between."

-Hurley, S. (2008). The shared circuits model (SCM): how control, mirroring, and simulation can enable imitation, deliberation, and mindreading. Behav Brain Sci, 31, 1-22; discussion 22-58.

"I am skeptical of the Classical Sandwich myself, because I believe that thinking incorporates representations used for perception and motor control"

-Prinz (forthcoming) Is Consciousness Embodied? In P. Robbins and. M. Aydede (Eds.) Cambridge Handbook of Situated Cognition.

"traditional psycholinguistics also assumes the cognitive sandwich, with the thinking “meat” keeping apart the production and comprehension “bread.” But if action and perception are interwoven, then production and comprehension are interwoven as well, and so accounts of language processing should also reject the cognitive sandwich."

-Pickering and Garrod (forthcoming). An Integrated Theory of Language Production and Comprehension. Behavioral and Brain Sciences (target article).

What do perception, action, and "cognition" have in common? Computation.

Anonymous said...

If I am not mistaken, the (pejorative) metaphor of the classical cognitive sandwich model comes from the radical embodied cognitivists who reject both representations and computation. For them, cognition is a kind of Watts governor, perception and action in one, with nothing much noteworthy in between. A poor sandwich!

Anonymous said...

"robot' might be a useful way of approaching how we conceptualize 'cognition'. The reason being that a robot can only perceive or sense temperature if it has some sort of thermal sensor, in that case the distinction of hardware versus software becomes clear(of course for some embedded controllers the sw/hw distinctiion can become blurred). Similarly the robot only effects behavior through some sort of actuator or connection to a mechanical system, Thus cognition has to do with software, i.e., the program doing the processing, everything else is either a sensor or an actuator. For our human wetware the distinction is much more fuzzy. We have not too much problem with the boundary of sensor/actuator at hair cells/muscle. We might imagine reflexes as non-cognitive, i.e. a very short number of synapses between hot and move your hand. Anything that takes more synapses is then cognitive. -RC

Greg Hickok said...

Why can't a reflex be "cognitive"? Is a reflex a form of computation? Is the input from one pool of neurons causing another pool of neurons to fire in a reflex circuit? Is that fundamentally different from a pool of neurons in PPC causing another pool of neurons to fire in PM?

Anonymous said...

To go on with the stupid examples in the second comment below the post (which I neglected to sign): If a plant turns to the sun, is it cognition? A crude definition of cognition could be processes that can’t be reduced to physic or biology. Then perhaps Pavlov’s dog salivation could be cognitive while the jaw jerk reflex not. Vilem

Greg Hickok said...

Cognition is just a bad term. It confuses things in my view.

Fred said...

I don't know that "computation" leaves us in much better shape, really...see, e.g. point 5 at

(this is Brian Cantwell-Smith's [] new/ongoing take on what computation is)

Ryan Ogilvie said...

Traditionally, the term 'cognition' was used by philosophers to refer to thought and knowledge. Although figures like Descartes, Hume and Kant can be considered proto-cognitive scientists, they don't really use the term to refer to the physical underpinnings of thought. (This is particularly true in the case of Descartes, as he was a card-carrying dualist. But thinking about how the mind can be instantiated in something physical continues to be a difficult problem.) This came later, with the rise of computationalism.

The embodied cognition folks are worried about what was once a dominant way of conceptualizing thought in the mid 20th century, namely, as a central processor that received inputs from sensory apparatus and delivered outputs to motor systems. Sensory input was taken to be a complicated transduction story, while action output was just a complicated effector story--hence, the "sandwich" metaphor.

If one thinks that there is perceptual processing is essential in someway to thought (e.g., beliefs, judgments, desires, etc.), then you will want to reject the sandwich model of cognition. In other words, I don't think we're dealing with a strawman, here, for this is a substantive question.

However, you might reject the notion that perceptual processing (or motor output) is essential to thought, yet nonetheless think of sensory processing as cognition. In such as case, one just needs to be clear about how she or he is using the term.

So, yes, the use of cognition in contemporary and historical debates does introduce significant confusion. I guess we just have to carefully work our way through the mess.



Greg Hickok said...

My problem is that the embodied folks refer to the "classical sandwich model of the *mind*" when really the sandwich model only applies to certain domains, mostly notably the representation of conceptual knowledge. It is a straw man to present embodiment as an alternative model of the mind unless you want to restrict "mind" or "cognition" only to include conceptual representation or problem solving, etc.

Are state feedback control models in the motor control literature supposed to be outside of the "mind"? Are visual computations that "infer" structure from motion not part of the "mind"?

Here's a wild thought: the embodied cognition movement is nothing more than the realization by theorists in "high level cognition" that sensory and motor systems are complex information processing systems in their own right capable of categorizing, drawing inferences, making predictions and so on (which sensory and motor scientist have known for quite a while).