Mirror Neurons, Animacy, and Gesture

So mirror neurons have been back in the news recently, as the result of a paper in the July 2007 issue of PLoS one titled, "Do you see what I mean? Corticospinal excitability during observation of culture-specific gestures"(1). Sounds interesting in a geeky sort of way, right? The paper starts with the observation that

At this stage we still know very little about how special classes of actions such as communicative hand gestures are understood. (p. 1)

So the authors, Molnar--Szakacs et al., conducted an experiment in an attempt to rectify this. Let me tell you about it...

But before I do, let's step into the Wayback Machine and travel back to June, 2007, when a paper by Wheatley, Milleville, and Martin(2), and was published the journal Psychological Science. That paper has been aptly summarized by fellow ScienceBlogger Dave Munger here, but in case you're too lazy to click over, I'll give you the gist.

Wheatley et al. set out to test two competing types of hypotheses about how we human folk understand the actions of animate agents (and other human folk in particular). The first type hypothesizes that a broad range of social cognitive processes (e.g., theory of mind, value judgments, moral judgments, all influenced by domain specific and domain general reasoning processes as well as affect) go into interpreting the agents' actions. These processes utilize what Wheatley et al. call the brain's "social network" (see the images below, from Wheatley et al.'s Figure 1, p. 470), which consists of the superior temporal sulcus (1), insula (2), medial prefrontal cortex (3), posterior cingulate (4), amygdala (5), and lateral fusiform gyrus (6), each of which is associated with specific social cognitive functions (e.g., the insula and amygdala are associated with "affective processing").

i-bdbef0d6d40731614ff5b4eb94c76ae2-WheatleyetalFig1.JPG

The second type of hypothesis, generally called simulation theory, is much simpler, and dare I say, vaguer (that is, it's not clear how it works). It argues that our understanding of the actions of others takes place in the motor system, and in the mirror system (the inferior parietal cortex, 7, and the ventral-premotor/inferior-frontal cortex, 8, in the images above). In case you've forgotten, mirror neurons are cells, first discovered in monkeys, that fire both when performing a movement and when observing a similar movement performed by another individual. There is now a fair amount of evidence that humans, like monkeys, have a mirror system, though it's not at all clear from that evidence what they have it for. Simulation theory is so-called because it argues that we understand the actions of others by activating the same mirror cells that we'd activate when performing similar actions, thereby simulating the actions in our head.

In order to test these competing types of hypotheses, Wheatley et al. conducted a pretty simple experiment. They constructed twelve moving objects that could be interpreted as animate or inanimate depending on the background. For example, one could be seen as an ice-skater or a spinning top, depending on whether it was presented as moving around on a frozen pond or a table with toys on it (see Dave's post for video of this example). Each participant was presented with each moving object in one of its two possible backgrounds. For each object they first saw the background alone ("look" condition), then the object moving against the background (the "watch") condition, and finally the background alone with instructions to image the object in it ("imagine" condition). Participants observed all of this while in an fMRI machine, so Wheatley et al. could take pretty pictures of their brains.

Focusing on the "social network" and the mirror network, including all the areas in the images above, Wheatley et al. looked made a few different comparisons. The most important question is, are the target areas more active when viewing and/or imaginging animate motion than inanimate motion? For the "social network," the answer is yes. Most of the areas associated with social cognition, including the left fusiform gyrus, the right superior temporal sulcus, the left amygdala, the left anterior insula, and the posterior cingulate, were more active when viewing animate motion, when imaginging it, or both, than when viewing the background or inanimate motion. For the mirror system, the answer is no. That is, both the left inferior parietal lobe and the left inferior frontal gyrus responded more to animate motion than to the background in both the watch and imagine conditions. However, it responded more to inanimate motion than the background in both the watch and imagine conditions as well. In fact, the left inferior parietal lobe seemed to respond more strongly to inanimate than animate motion in the image condition.

What does this mean? Well, it suggests that the social network is important for processing animate motion, both when viewed and when imagined, as the regions in this network seem to respond selectively to animate as opposed to inanimate motion. It also suggests that the mirror system doesn't really care whether motion is animate or inanimate; it's going to respond either way. That's a bit of a blow to the simulation theory of action understanding, I'd say. It's not a falsification of course, 'cause let's face it, fMRI studies can't tell us anything about what the mirror system is doing when it responds to animate and inanimate motion. Perhaps it's doing different things for each, and therefore could play a role in understanding animate actions. Who knows?

I bring up this study and the problems it raises for the simulation theory because in their introduction, Molnar-Szakacs et al., write,

It has been proposed that action perception involves an internal simulation or replication of the observed action. (p. 1)

Oops, they might be a month too late, but as I said, the Wheatley et al. data is equivocal, even if it's a bad omen, with regard to simulation theory (though it bodes pretty well for theory of mind theories, I'd say), so I suppose there's no reason to panic. On to the experiment, then.

The experiment used transcranial magnetic stimulation (TMS) to measure corticospinal excitability (CSE). Basically, you put magnets on people's heads and see how little stimulation is needed to cause neurons in the target area of the brain to fire. The less stimulation needed, the higher the excitability in that brain area. More excitability implies more activity, or something like that.

Molnar-Szakacs et al. hypothesized that if people share a "motor repertoire" (that is, they have knowledge of the same gestures), they'll be able to communicate more effectively. That would mean that if people are using motor simulations to understand gestures, then witnessing a known gesture from someone who also knows the gesture should increase CSE in the mirror system. They tested this in a study with a whole bunch of conditions. These conditions involved watching videos of either shared gestures (American gestures like "thumbtwiddle," "OK," "shamefingers," and "hang loose"), gestures from a different culture (Nicaraguan gestures like "play marbles," "get caught," "I swear," and "neat/well done"), unknown gestures from the same culture (signs from American Sign Language), or a person standing still, and the gesturer was either from the same culture (American) or a different culture (Nicaraguan). So participants (4 American males) either saw an American or Nicaraguan perform familiar American gestures, unfamiliar Nicaraguan gestures, unfamiliar ASL gestures, or a person standing still.

i-3b22ff1ea8f73ad46998738a7b5ae8e9-Molnar-SzakacsetalFig1.JPG

Molnar-Szakacs et al. Figure 1, p. 2.

Just to be clear, the prediction here is that when the performer of a gesture is from the same culture, and is performing a familiar gesture, CSE will be higher than when a gesture (familiar or unfamiliar) is performed by someone from a different culture. And here's what they found (from their Figure 3, p. 3):

i-b7034fe12f530d55178969bb61e3a566-Molnar-SzakacsetalFig3.JPG

That graph basically represents the percent that the MEP (that is, the measure of CSE) changed from the "overall mean" (that is, across conditions) in the different conditions. As you can see, when the person performing the gesture was American, CSE was well above the average for each type of gesture (control = ASL in the graph) except the Nicaraguan. On the other hand, CSE was well below average when the Nicaraguan actor performed American or ASL gestures, but not when he performed Nicaraguan gestures or just stood there.

How this data relates to the authors' hypothesis is, well, difficult to say, but I'll try. It looks as though the mirror system is responding to the American actor regardless of what he does, including if he just stands there, unless what he's doing looks weird (like a Nicaraguan gesture). On the other hand, the mirror system is much less responsive for the Nicaraguan actor if he's doing American things, while it doesn't really care about Nicaraguans doing Nicaraguan things or just standing there. Since the authors' hypothesis had to do with familiar gestures, the fact that response to ASL signs was so high is problematic, as is the response to the American just standing there. Molnar--Szakacs et al. suggest that this pattern may indicate that the mirror system is involved in determining whether someone's a member of an in-group (because, you know, the mirror system's involved in everything!), but it's hard to tell what, if any role it plays in gesture understanding. It just seems to like familiar-looking people doing stuff that doesn't look too unfamiliar (as if it came from a very different culture), even if the stuff they're doing doesn't involve moving at all, lights up the mirror system. Combined with the Wheatley et al. data, it seems that mirror neurons respond to animate actions, inanimate actions, and no action whatsoever, as long as the actors aren't doing anything too weird and look like they're from around here.

It seems, then, that we've learned absolutely nothing about gestures, leaving us right where we started, and haven't learned a whole hell of a lot about mirror neurons either. Man, I love mirror neuron research.


1Molnar-Szakacs, I., Wu, A.D., Rbles, F.J., & Iacobini, M. (2007). Do you see what I mean? Corticospinal excitability during observation of culture-specific gestures. PLoS One, 7.
2Wheatley, T., Milleville, S.C., Martin, A. (2007). Understanding animate agents: Distinct roles for the social network and mirror system. Psychological Science, 18(6), 469-474.

More like this

Hi Chris,
Nice take on the paper. I had similar sentiments. Really, wake me up when 'mirror neuron'/ 'mirror systems' approaches have evolved to a stage where they make refutable predictions. By golly, I'd have thought that the data in this paper would be consistent with quite a few types of null hypotheses (that mirror system approaches would want to provide alternatives (H1s) to). On that matter, it could be good if fMRI studies of this type would be more clear on what the null hypothesis is from the get-go rather than fishing for any pattern whatsoever. Glad I got that out of my system.

For another interesting "failure to mirrorize" result (in a study using action verbs) see
Rüschemeyer et al. Comprehending prehending: neural correlates of processing verbs with motor stems. J Cogn Neurosci, Vol. 19, No. 5. (May 2007), pp. 855-865.

By ThursdayMan (not verified) on 09 Aug 2007 #permalink

I wonder if those mirror neurons were tripping up the Queen from Snow White and the Seven Dwarfs.

By Bill Benzon (not verified) on 09 Aug 2007 #permalink

Excellent critique!

One novel contribution of the current study is our finding that the human mirror neuron system specifically, is differentially sensitive to ingroup versus outgroup members.

"Specifically" - do they mean as opposed to the monkey mirror neuron system? What the authors fail to mention is that monkey mirror neurons respond to actions performed by non-conspecifics such as the experimenter (Gallese et al., 1996).

In a study of actions performed by conspecifics and non-conspecifics, Buccino and colleagues showed that actions belonging to the motor repertoire of the observer were more successful in eliciting activity within the fronto-parietal circuit for action representation than foreign actions.

However, here they fail to mention that the observation of biting movements -- whether performed by humans, monkeys, or dogs -- all produce the same activations in that very same fMRI study (Buccino et al., 2004).

So how do those findings fit with the "racist" mirror system in the present study?

Haha... how do any findings fit with it? 'Cause I mean, looking at that graph, I can't figure out what the hell the data imply.

But I do know this: this is the first result implicating the mirror system in our current immigration debate.

Although the original critique is cleverly written to lead the reader to think that the argument makes sense, it uses evidence from a study of OBJECT perception with fMRI, to criticize the results of a study of ACTION (made by biological agents) perception using TMS. First of all, the PLoS study isn't about animacy versus non-animacy, all stimuli are not only animate, but they in fact exhibit biological motion, in a way that objects don't. Had it been a study of interacting with robots, I would buy more of the critique.

The Neurocritic evidently read the Buccino study very carefully, but not the one discussed here, because if one looks at the figure posted, it shows that in fact the mirror system is not racist at all, in fact there is higher activity for a Nicaraguan actor making the NIcaraguan gestures, than american gestures. If the MN system was racist, why would it try to simulate outgroup actions?

There are very subtle effects presented here, that are being completely overlooked by people who are trying to critcise ANY findings about the human MNS for whatever reason.. and run around stating that the MNS is 'racist'. Now we can see where the current administration gets its science advisors...

Toast, but the Wheatley et al. paper was on action. The whole point of the study was to use the same objects in a way that in one case, they looked like animate agents, and the other case they looked like moving inanimate objects, and to look at whether the mirror system (and social cognition system) responds selectively to to animate motion, or action. Since they didn't, that's a problem for the simulation theory: unless you think we're simulating the actions of inanimate agents too, which would be odd, because, like, they don't have hands and stuff.

The relevance, then, is that it would be odd for the mirror system to not selectively respond to animate agents acting, but to selectively respond to in-group vs. out-group actions. That would just be... well, I don't know what it would be, but it'd be a pain in the ass to come up with a theory to explain it. And hell, the data doesn't seem to imply in-group/out-group discrimination in a straightfoward way anyway.