Tuesday, August 16, 2011

Mirror Neuron Forum - Some additional discussion - Part I

Now that people have had a chance to digest the recently published "Mirror Neuron Forum" (Perspectives on Psychological Science 6(4) 369–407) I think it would be useful to revisit some of the claims and counter-claims. I will start working through some of the points in a series of posts. Of course, my focus will be on the parts of the forum that I participated in, but if you have some comments and thoughts on any part of it, feel free to email me and I'll post it as "guest post".

I would like to get the forum participants involved in this less formal discussion. I hope they will join in and of course, I will be happy to post anything from my co-forum participants.

Question 1: Do Mirror Neurons in Macaques or Humans Make an Important Contribution to Action Understanding?

I suggested in my answer to this question that if we have reason to question the role of macaque mirror neurons (MNs) in action understanding then we have reason to question the role of the mirror system in humans in action understanding and the various functions to which the system has been generalized.

Here is Gallese's response to my point:

VG. I will respond to some of the points raised by GH. First, the major premise of his initial argument is debatable. Accord- ing to GH, if one can question the relevance of MNs to action understanding in monkeys, then this would automatically jeopardize any conclusion about the role of MNs in human social cognition. Why? Do we assume that a given trait or neural mechanism found in different species must necessarily preserve identical characteristics? Evolutionary theory patently contradicts this assumption.

Gallese is right, of course. There is no reason to assume that a given system's function will be preserved evolutionarily. It is perfectly legitimate to assess the empirical merits of theories of MN function in macaque separately from the empirical merits of theories of the so-called mirror system function in humans; they may have very different functions, a point I made in my Eight Problems paper (it was problem #4). However, this is not how many people, including Gallese's group, have approached the two systems. Rather, theories of the function of the human mirror system has been built on the foundation of the MN theory of action understanding in macaques. Consider the abstract from one of Gallese's highly influential papers:

How do we understand other people's behavior? How can we assign goals, intentions, or beliefs to the inhabitants of our social world? A possible way to answer these challenging questions is to adopt an evolutionary frame of reference, both in phylogenetical and ontogenetical terms, envisaging these ‘mind-reading' capacities as rooted in antecedent, more ‘ancient' and simple mechanisms. This approach can capitalize on the results of different fields of investigation: neurophysiology can investigate the neural correlates of precursors of these mechanisms in lower species of social primates such as macaque monkeys. Developmental psychology can study how the capacity to attribute propositional attitudes to others develops.
In the present article we will propose that humans' mind-reading abilities rely on the capacity to adopt a simulation routine. This capacity might have evolved from an action execution/observation matching system whose neural correlate is represented by a class of neurons recently discovered in the macaque monkey premotor cortex: mirror neurons (MNs). Gallese and Goldman, 1998

So the logic is:
-MNs in macaques perform motor simulation/action recognition during observation
-humans posses a parallel motor simulation system (an assumption not mentioned in the abstract but one that gets its own section in the paper)
-therefore, mind-reading is evolutionarily built on this motor simulation function of MNs.

The point in my response to the MN Forum question was that if macaque MNs are not doing motor simulation/action recognition then the standard argument put forward by Gallese and others needs to be re-evaluated. Why? Because the logic of the argument in Gallese and Goldman, for example, would become this:

-MNs in macaques are NOT performing motor simulation/action understanding during observation
-humans posses a parallel system that may or may not be performing motor simulation/action understanding during observation
*therefore, mind-reading is evolutionarily built on this motor simulation function of macaque MNs -- Oops, doesn't follow! (neither did it follow logically above either, but here it is a larger stretch to even attempt the linkage).

Practical point: Gallese and others have used the claimed function of macaque mirror neurons to develop theories of the neural basis of everything from speech perception to mind-reading in humans. This is a reasonable approach. But if we learn that macaque mirror neurons are doing something else, then by the same approach -- i.e., to "adopt an evolutionary frame of reference, both in phylogenetical and ontogenetical terms" and to "capitalize on the results of different fields of investigation" -- we need to reconsidered our theories of the human mirror system.


Gallese, V., Gernsbacher, M.A., Heyes, C., Hickok, G., & Iacoboni, M. (2011). Mirror neuron forum. Perspectives on Psychological Science, 6, 369–407.

Gallese V, & Goldman A (1998). Mirror neurons and the simulation theory of mind-reading. Trends in cognitive sciences, 2 (12), 493-501 PMID: 21227300


Brian Barton said...

Actually, I thought the issue regarding evolution was probably the weakest point in the entire forum. Effectively, Gallese argues that, "Macaques have macaque mirror neurons (mMNs). Humans have something close to mirror neurons, but not quite (human mirror neurons; hMNs). If the features of hMNs differ from those of mMNs in a way that does not fit nicely into my framework (e.g., hMNs respond to mimicry, when mMNs do not), it is due to an evolutionary adaptation, and so is not really a problem for my framework."

This argument is simply lazy, because any problem that applies to a feature that is shared between the two species' MNs also applies to both, unless some third feature of one type of MN overrides this problem. Furthermore, if the hMNs differ from mMNs in any way, they can differ in a number of unanticipated ways, all of which would need to be evaluated before concluding which properties of mMNs can be used as supporting evidence for hMNs.

Anonymous said...


I'm currently reading this forum. It is very interesting to have a direct comparison between the different points of views.

I was struck by the role that was attributed to the motor system in speech perception. You talk about it as a modulation of the auditory signals by motor signals. I might think that it's more than modulation. Indeed, in the motor control theories, when we have two inputs about the same signal (let's say vision and proprioception about hand position), one of them does not modulate the other. Rather, an estimate of hand position is obtained through combination of the visual and proprioceptive inputs (in a Bayesian way). Indeed, your estimate will always be more accurate (smaller SD) if you have several sources than if you have only one of them.

So here is my question: Can't we have two speech perception signals that originate from different systems and that are combined together to obtain the best estimate?


PS: I was amazed by the accuracy of your description of motor control system (p.376).

Greg Hickok said...

JJ - You are right to suggest that if you have two inputs about the same signal they can potentially combine rather than one driving and the other modulating. But this depends on what the inputs are. Notice that in your example you used two sensory inputs, not a sensory and a motor one. In speech we see the same combining of two sensory inputs when auditory and visual speech combine to yield the McGurk effect.

I don't think this kind of combination of inputs occurs in the case of a sensory and a motor signal because it would defeat the whole purpose of forward prediction in motor control. Here's why: If my motor system generates a forward prediction about a sensory state and that prediction doesn't match my sensory feedback, combining these signals would not result in a veridical estimate of the error between predicted and actual states (which is what you need for motor control) but rather a blurring of the difference. In fact, given the goals of motor-based forward prediction, I'm coming to think that I'm wrong that these signals can be used productively to modulate perception of others' speech.

vkodytek said...

I am confused by your last sentence, Greg. If perception of others’ speech is not a strictly bottom-up process, what does it mean? That one’s motor system may not participate in analysis by synthesis at all?

Greg Hickok said...

Right. I don't think the motor system participates in analysis by synthesis at all. I believe in the concept of AxS but I don't think it is coming out of the motor system. Instead I think it is coming out of higher levels (e.g., lexical) of the ventral stream.

I currently think that existing observations regarding motor effects on speech perception via TMS, fatigue, etc. involve bending the system in way it was not intended to operate and indeed doesn't operate normally. By analogy, the field has found a screwdriver (the motor system) and has shown that you can make a little headway hammering nails (speech perception) with it.

vkodytek said...

Thank you. So in your MN Forum Fig.2 you've moved to the square A.

By the way, humans are unique in the ability to use tools in many different ways. :-)

Greg Hickok said...

For speech perception yes, I'm in box A at the moment until some bit of evidence pushes me out. For speech production I'm in box C though: speech motor control needs the auditory system!

vkodytek said...

I see.