Friday, October 31, 2008

Rock-Paper-Scissors and mirror neurons: Executed and observed movements have different distributed representations in human aIPS

"Shane" left a comment on a previous post about a recently published paper by David Heeger's group.

I have heard about this paper, but haven't had a chance to read it yet. Here is the abstract for a quick summary:

How similar are the representations of executed and observed hand movements in the human brain? We used functional magnetic resonance imaging (fMRI) and multivariate pattern classification analysis to compare spatial distributions of cortical activity in response to several observed and executed movements. Subjects played the rock-paper-scissors game against a videotaped opponent, freely choosing their movement on each trial and observing the opponent's hand movement after a short delay. The identities of executed movements were correctly classified from fMRI responses in several areas of motor cortex, observed movements were classified from responses in visual cortex, and both observed and executed movements were classified from responses in either left or right anterior intraparietal sulcus (aIPS). We interpret above chance classification as evidence for reproducible, distributed patterns of cortical activity that were unique for execution and/or observation of each movement. Responses in aIPS enabled accurate classification of movement identity within each modality (visual or motor), but did not enable accurate classification across modalities (i.e., decoding observed movements from a classifier trained on executed movements and vice versa). These results support theories regarding the central role of aIPS in the perception and execution of movements. However, the spatial pattern of activity for a particular observed movement was distinctly different from that for the same movement when executed, suggesting that observed and executed movements are mostly represented by distinctly different subpopulations of neurons in aIPS.
(Italics added.)

So this is an anti-mirror neuron paper. While I'm fully on-board with the anti-mirror neuron conclusion, I'm not sure the data really support this view. Again, I haven't yet read the paper and am basing my argument on the abstract only, so somebody correct me if I'm missing something. The study found that aIPS activated both for action production and action viewing. No surprise there. The interesting and novel contribution of this paper is that within the activated region, they found different patterns of activation for observation and execution of movements. From this they conclude that the these two functions are supported by distinctly different subpopulations of neurons.

I like the methodology employed here, and I believe their findings do indicate that observation and execution involve non-identical populations of neurons, but I don't think this is strong evidence against a mirror neuron view. Here's why: Suppose there are three types of cells in aIPS:

1. sensory-only cells
2. motor-only cells
3. sensory-motor cells (mirror neurons)

There is evidence for this kind of distribution of cells in parietal sensory-motor areas. Suppose further that action understanding is achieved by cell type #3, the mirror neurons. If this were true, the ROI as a whole would activate for both action observation and action execution, as the study found, but sensory vs. motor events would nonetheless activate non-identical populations of cells within the ROI: observation would activate cell types 1 & 3, whereas execution would activate cell types 2 & 3. This difference may be enough to allow for above chance pattern classification that is based on non-mirror neurons within the ROI.

So if I've got the basics of the study correct (based on the abstract), this is not strong evidence against mirror neurons supporting action understanding. Neither is it evidence FOR mirror neurons, however.

I. Dinstein, J. L. Gardner, M. Jazayeri, D. J. Heeger (2008). Executed and Observed Movements Have Different Distributed Representations in Human aIPS Journal of Neuroscience, 28 (44), 11231-11239 DOI: 10.1523/JNEUROSCI.3585-08.2008


andreja said...

It's probably right to say that the mirror neurons would survive this data pattern (just like almost any other, for that matter). Not knowing much about the classifiers (and not having read the paper), i wonder whether you could still get some type of a gradient in classification correctness if you could design a similar experiment featuring additional control conditions which would activate only cells 2 and cells 1.
If not, what else could you do to further investigate the topic - could repetition suppression work?

Greg Hickok said...

I think it is going to be very difficult to test the "action understanding" hypothesis with this kind of study, and in fact I think it is a waste of time to try to do so.

There is already everyday evidence that you don't need mirror neurons to understand action: I understand the flight of a bird (I can't fly), bassoon playing (never touched one), and pretzel folding (tried but can't do it). And there is everyday evidence that motor codes are not sufficient to action code meaning: the motor code(s) for tipping a pitcher and allowing water to flow into a glass is(are) ambiguous with respect to the very different meanings of the actions of pouring, filling, or emptying, never mind tipping, rotating, tilting, spilling, dumping, splashing, etc., etc. Clearly much of the meaning of an action can't possibly be coded in terms of a motor representation.

I propose we start studying the "mirror system" as a sensory-motor integration system, not as the end-all solution to action semantics, mind reading, and autism.

andreja said...

Well, i principally agree.
But, what bugs me is the ultimate looseness of terms being used, which makes it quite inviting to try and twist the arguments around.
For one thing, it seems like the terms "motor" and "mirror" sometimes get interchanged. For the second, the term "mirror system" is vague, to say the least.
E.g., it is clear that you can understand action without being able to do the movement yourself, but some would still say that you may profit (or need) your motor system for this understanding - if you think in terms of using motor system for prediction of any dynamic event or if you take into account affordance accounts, it sounds plausible (you may not be able to fly like a ball, but you may be able to catch it which may be helpful (needed) for understanding its dynamics). This can go for both action understanding or pure perceiving.
On the other hand, motor codes may be ambiguous for action meaning, but as i understood mirror neuron people, MNS is about goals and intentions - you simulate the movement and reach the goal (regardless of who/how you actually get there) - related to this, i think at least in primates, MNS is restricted to goal-directed actions.
Ironically, I can't think it through right now, but I have a feeling my first and second example would eventually contradict each other, but this is exactly what happens when "everything goes" in terms of levels and scopes of explanation.
The bizzare thing about MNS is that it is a system defined by the presence of a certain type of cells - not many within these regions, i would guess. Properties of these cells are used for defining the functioning of the whole network ... which is really ill defined. (By the way, I didn't read this paper (yet?), but isn't MNS supposed to be more ventral in the IPL, not really in the sulcus? And, even if I am wrong the way MNS is localized and then conceptualized in many studies sometimes looks very circular, If you go for the more liberal accounts, it is not even a system, but systems - e.g., empathy would be insula, motor parietal-premotor...)

Greg Hickok said...

Andreja, Your point about one "profiting" from the motor system is important. I actually think that our comprehension of certain action may profit from (or via) motor systems, and maybe exactly in the ways you discuss: prediction/forward modeling, affordances, etc. But this is a very different thing than saying that the motor system is the "basis" for action understanding. I'm only arguing against this more extreme view.

Once we give up on "basis" and start talking about how the motor system may have a modulatory effect on understanding, then we have a viable game, I think.

Regarding ambiguity of motor codes for action. If it really is the GOAL that is coded -- e.g., "to pour" -- then you could achieve that goal with any number of motor actions including telling someone else to do it! Are we still in the motor system at that point?

Thanks for the thoughtful comments!

andreja said...

:) thanks for a motivating post and a cool discussion.
p.s. i agree; the integrative and modulatory effects are fascinating without the panexplanatory power transcending all possible phenomena out there. and, as for the last question, i don't think we are still in the motor system at that point.

Anonymous said...

Dear Greg,
Our paper is not "anti mirror neuron"! For some reason everyone is interpreting it that way and I guess we did not do a good enough job making that clear in the abstract. But had you read the paper you would have seen that we entirely agree. We suggested that there are three neural populations in aIPS and that the visual and motor populations are those dominating the fMRI responses, hence we get different patterns for observation and execution. This is actually what you would predict from the monkey literature where mirror neurons are a minority of those responding in this area. Nevertheless, this is an important characterization of the responses of this area and has strong implications for interpreting about 10 years of fMRI, EEG, and MEG experiments regarding the human mirror system. Responses of "mirror system areas" do not equal responses of mirror neurons as many previous papers seem to suggest.
All the best!