Monday, March 22, 2010

Mirror neurons support action understanding -- "from the inside"?

I think we are getting closer to understanding what mirror neurons are doing. No longer is it claimed that mirror neurons are THE basis for "action understanding". Now, according to Rizzolatti & Sinigaglia's new review (2010), there are several non-mirror mechanisms that can accomplish this:

We conclude that, although there are several mechanisms through which one can understand the behaviour of other individuals...

Mirror neurons do something else (R&S would probably prefer the term, more, but else is better I think)

the parieto-frontal mechanism is the only one that allows an individual to understand the action of others 'from the inside'

Let me explain what "from the inside" means, or at least provide an alternative to whatever R&S mean by it. In my "Eight Problems" paper (Hickok, 2009) I noted that I can understand the action of saxophone playing even though I've never performed the actions associated with saxophone playing. R&S acknowledge that, indeed, I do understand the action of saxophone playing, and do so without the benefit of mirror neurons. But they suggest my understanding is lacking something, namely that extra bit of knowledge that comes from knowing how to play a saxophone. So recognizing an action that I know how to perform = basic action understanding from non-mirror systems + mirror neuron-driven knowledge that hey I know how to make MY motor system do that!

In other words, mirror neurons support the knowledge of how to perform an action that one is observing -- that is, mirror neurons are part of the same old "how" stream that vision neuroscientists, and more recently auditory neuroscientists have been working on for more than a decade (Milner & Goodale, 1995). The "how" stream, of course, supports sensory-motor integration, or in R&S's terms, action understanding "from the inside". This is why you see motor activation during the perception of actions that you can perform: it is sensory-motor association.

Importantly, notice that there is no magical semantic knowledge that suddenly falls from heaven when we know how to perform an action. I can teach you a new word, glemph, and you can learn to reproduce it with your vocal tract so that subsequent presentations of acoustic glemph will activate your motor system by association. I could do the same with a sign language gesture. It doesn't take on meaning until you link the sensory-motor ensemble to a conceptual structure. So for example, I could define glemph (or the sign language equivalent) as 'the act of publishing on the topic of mirror neurons'. Now, the sensory-motor ensemble has meaning and you understand it, but does that mean that the meaning and the understanding is now suddenly coded or even augmented by the sensory-motor ensemble itself? Put differently, do you now have a different understanding of the concept, 'the act of publishing on the topic of mirror neurons' just because you have a new sensory-motor associate of the concept? No, you just know how to articulate the word that is associated with the concept.

I suggest that this interpretation of "understanding from the inside" explains every mirror neuron-related observation, does so more parsimoniously than Rizzolatti's account, and has more empirical support from research on aphasia and apraxia.


Hickok G (2009). Eight problems for the mirror neuron theory of action understanding in monkeys and humans. Journal of cognitive neuroscience, 21 (7), 1229-43 PMID: 19199415

Milner, A.D., and Goodale, M.A. (1995). The visual brain in action, Oxford: Oxford University Press.

Rizzolatti, G., & Sinigaglia, C. (2010). The functional role of the parieto-frontal mirror circuit: interpretations and misinterpretations Nature Reviews Neuroscience, 11 (4), 264-274 DOI: 10.1038/nrn2805


shane said...

It looks like you have three main points here:

1. Mirror neurons are just neuronal ensembles that link perception and action.

2. These sensory-motor associates don't constitute meaning. It the activation of the conceptual structure that is important. ("It doesn't take on meaning until you link the sensory-motor ensemble to a conceptual structure.")

3. Knowledge of how to perform an action is not part of the conceptual system ("there is no magical semantic knowledge that suddenly falls from heaven when we know how to perform an action")

So I think claim 3 is false. If you have never seen the colour blue, you will have a different concept of blue from someone who has. If you have never played a chord on a piano, you will have a different concept of a chord to a piano player. If I have never glemphed, then I will have a different understanding of glemphing to someone that has.

But perhaps you are right about the claims that mirror neurons are just sensory-motor associates and have no meaning on their own. However, if they do have a functional role, then it is to link perception and action. This linking of perception and action could lead to the activation of the motor system (in some cases). The activation of the motor system could then influence action understanding. You might not have a different understanding of the concept 'x' just because you have a sensory-motor associate of the concept, but having and using the motor side of the association can lead to different knowledge associated with 'x'. This might be a weaker claim about mirror neurons than some make, but I think it is more defensible.

tom said...

Hi Greg,

I really like this idea!

I think it also needs to explain why actions with an assumed real-world goal (i.e. grasping an actual object) seem to cause MN firing, whereas arbitrary 'pantomime' gestures do not.

One possible explanation maybe that the MN activation upon observation is some function of the strength/efficiency of the motor programs present in the animal. This would assume that motor programs for real-world actions repeatedly performed throughout the animal's life will have stronger representations than those induced by relatively short periods of experiment-specific training, or no training at all - i.e. in your terms, you understand 'how' a lot better if you've had more practice.

Greg Hickok said...

The point about #3 is not that there isn't necessarily some "knowledge" encoded in the sensory-motor ensemble, but that this knowledge is restricted to information regarding HOW to perform an action not WHAT an action means.

Yes, having a sensory-motor associate of a concept does add some kind of knowledge, but having this knowledge of HOW to perform an action does not in an of itself tell you the goals and intentions of the actor, which is what Rizzolatti is claiming.

Anonymous said...

I like this idea too.

I assume you have seen Celia Heye's substantial body of studies suggesting that the properties of MNs reflect sensorimotor associations rather than anything innate and and special to social cognition. This includes things like the Catmur study that R&S try to dismiss. I'm not sure what Celia thinks about the functional role of MNs though.

Greg Hickok said...

I'm glad you guys like the idea. Funny though -- this is just an elaboration of what I claimed in my Eight Problems paper. It didn't get much attention there. Cecelia Hayes' recent paper is also a much elaborated version of the sensory-motor theory. I'll have to look more closely at her paper to understand exactly what she is claiming; I bet it is very similar to what I said here in this last post.

I think what Tom said in response to my previous post is right: No one is going to abandon the action understanding theory without a serious competitor theory. And while I did propose a competitor theory, I certainly failed to emphasize it, focusing instead on the why Rizzolatti's theory fails. Tactical error on my part I suppose.

Howard Book said...

Being a old guy, I'm still attempting to understand the ''mirror neuron " phenominon. My understanding is that when a monkey sees or hears another monkey snap open the shell of a peanut, the same neurons are activated in the watching/hearing monkey as in the monkey carrying out the snapping.

But, does this mean that the watching/hearing monkey necessarily demonstrates the same - or similar - neruomuscular activity in those muscles involved in doing the 'snapping'.

Now let me jump to humans and the implications this may have for our capacity for "understanding" - a term often used in discussions about mirror neurons. What do we mean by, and how do we define, the term "understanding"? Are we referring to the capacity to mentalize which Peter Fonagy and his collegues at Baylor have described as ' the capacity to infer that behavoirs of others and of oneself are a reflection of unobserved inferred mental states ( in the other and in oneself) and that other's mantal state may be different than one's own'? Without such understanding we are at risk for logical thinking to regress to teleological state.

Or are we referring to the capacity to empathize, which has been described as 'the capacity to non-judgementally experience the other persons conscious and pre-conscious perspective of him/her-self and of his/her world, regardless of the foreingness that experience may have for us? Kohut refers to this as 'vicarious introspection' learning of the other's experience by looking inward.

Thanks for your comments.

Greg Hickok said...

Understanding is not clearly defined by Rizzolatti which is one of the problems of the account.

Anonymous said...

When reading your Eight problems for the mirror neuron theory It would have been very useful to find your opinion on what action understanding is, that will help so much as to take advantange of the large amount of papers writen on the subject.
A.M Acosta