Are mirror neurons evidence for simulation theory?

Hi folks. I’m one of the new kids on the virtual block. Thanks to Gualtiero for giving me the opportunity to contribute here.

I’m interested in the relation between mirror neurons and social cognition. Mirror neurons are neurons in our brains that fire for boththe first-person experience and the third-person observation of actions, emotions and sensations. For example, mirror neurons are activated endogenously when I  grab a cup, and these same neurons are activated exogenously when I observe you grab a cup. That’s the basic gist of mirror neurons. There have been many bombastic claims made onbehalf of mirror neurons (e.g., they’re allegedly responsible for our abilities to understand others’ actions, to grasp others’ intentions, to learn to use tools, to imitate, to empathize, to understand language, our moral development, and cultural evolution, among other things). Of particular interest to me is the claim that mirror neurons are evidence for the Simulation Theory of mindreading (ST). From what Ican tell, there is a pretty strong consensus that this claim is true, but I’m not sure why. The argument that seems most influential is offered by Gallese and Goldman (1998).

ST holds that we understand others by replicating their mental states. As Gallese and Goldman put it, to understand a target’s current behavior,

“The attributor starts with the question, ‘What goal did the target have that led him to perform action m?’  He conjectures that it was goal g,and tries out this conjecture by pretending to have g as well ascertain beliefs about the effectiveness or ineffectiveness of the action m vis-à-vis goal g. This simulation leads him to form a (pretend) decision to do m. He therefore uses this result to conclude that the target did indeed have goal g. In this fashion, the attributor ultimately makes a ‘backward’ inference from the observed action to a hypothesized goal state” (Gallese and Goldman, 1998, 497).

When mirror neurons are activated endogenously in execution mode, the mirror neuron activity constitutes a motor plan.  When mirror neurons are activated exogenously in observation mode, the mirror neuron activity still constitutes a motor plan, but one that is tagged as belonging to the target.  The exogenous activation of mirror neurons constitutes the retrodictive simulation that ST postulates.  Gallese and Goldman conclude, “Thus MN activity seems to be nature’s way of getting the observer into the same ‘mental shoes’ as the target –exactly what the conjectured simulation heuristic aims to do” (Galleseand Goldman, 1998, pp. 497-98).

Further, Gallese and Goldman argue, the Theory Theory (TT) does not predict anything like the mirror neuron system. “A proponent of TT might say that TT also has ways of accounting for retrodictive attributions of mental states. Is it clear that anything similar to  simulation occurs in externally generated MN activity? The point isthat MN activity is not mere theoretical inference. It creates in the observer a state that matches that of the target. This is how it resembles the simulation heuristic. Nothing about TT leads us to expect this kind of matching” (Gallese and Goldman, 1998, 498).

Do others find this argument persuasive evidence for ST and against TT? Why exactly? Are there other (good) arguments connecting mirror neurons with theories of mindreading? I don’t find the Gallese and Goldman argument particularly persuasive, but I’ll keep that for the comments section if there’s interest.

Gallese, V. and Goldman, A. 1998: Mirror neurons and the simulationtheory of mindreading. Trends in Cognitive Sciences, 2, 493-501.

[NB: This entry was originally posted by Shannon Spaulding]

13 Comments

  1. Bill

    I think that what the fMRI studies show is that we activate the same areas of nonverbal cortex when observing another’s behaviour as we use when producing similar nonverbal behaviours. To the extent that Theory Theory has the common philosophical flaw of reducing nonverbal thinking to verbal thinking, I think the fMRI studies show it is incorrect. However this does not mean that we don’t use such theories in predicting others. It’s likely both TT and simulation theories have some truth.

  2. James

    Cecilia Heyes had a counter at the Susan Hurley memorial conference earlier this year to the MN evidence for ST thesis.

    I can’t find a reference at the moment. But this is from something I wrote at the time:

    “Heyes has a different—rather deflationary—story to tell. She notes that associative learning mechanisms, whereby brain regions become associated with others, are not unusual in the brain. All that is needed is some sort of initial synchronicity in the firing of the two brain regions. Given frequent repetition of synchronous activation over time brain regions display activation akin to ‘mirroring’, about which however we may tell a story about the individuals past experience, not about evolutionary benefits.

    Heyes presents evidence that mirroring does not actually appear to apparent from birth, and that it is subject to the development of expertise. Her suggestion is that mirroring develops because of the frequency with which we both perform and action and observe the same action at the same time. These occasions include self-observation, self-observation in mirrors and participation is activities such as training, dancing and sports in which one has to perform the same actions as others around you.

    One snippet of information in support of this is the fact that Male and Female ballet dancers show higher activation when observing gender-specific dance movements of their own gender. Such moves have reinforced mirroring due to the fact that ballet dancers always practise in front of mirrors, and in early training, alongside dancers performing the same actions.

    Thus, she suggests, mirror neurons are simply another instance of associative learning, a phenomenon which has many benefits and most likely evolved due the the advantage of much more basic functions rather than because they facilitated the higher human social functions suggested which most likely emerged much later in the evolution of humans.”

  3. Daniel Weiskopf

    Hi Shannon. This is somewhat orthogonal to your main question, but I (and others) am pretty sceptical about mirror neurons and their standard interpretation. There was a great series of posts at the Talking Brains blog about this. Some choice bits to consult are:

    http://talkingbrains.blogspot.com/2009/09/what-mirror-neurons-are-really-doing.html

    http://talkingbrains.blogspot.com/2009/09/mirrors-in-brain-comments-on-rizzolatti.html

    http://talkingbrains.blogspot.com/2009/09/final-post-on-mirror-neurons.html

  4. shannon spaulding

    gualtiero: just as i suspected. this internet business is all voodoo magic!

    James: independently of the mindreading issue, i argue for a deflationary view of MNs as well, but maybe not so deflationary as Heyes. why not explain the ballet dancers’ MN behavior as some sort of top-down influence? there’s evidence that people’s MN systems are more active when observing a notable political figure of their own party, e.g., my MN system is more active when I observe Obama than when I observe Palin. (Iacoboni talks about this in his 2008 book). seems hard to explain this feature in terms of a mirror…

    Dan: thanks for the references!

  5. shannon spaulding

    hi eric. somehow my comment on your question got deleted earlier. there’s an interesting literature on autistics, who have dysfunctional mindreading capacities and (it seems) dysfunctional mirror neuron system. some argue that the damaged MN system causes the dysfunctional mindreading, but there are other plausible hypotheses, the most plausible of which is that there’s a common cause for the dysfunctional mindreading and MNS. Oberman and Ramachandran (2007) and Southgate and Hamilton (2008) are good starting places.

  6. shannon spaulding

    brendan’s comment got deleted earlier before i had a chance to respond. here’s what he wrote:

    Hi Shannon,

    For my own part, I am skeptical on many fronts. 1. I am highly suspicious of mirror neurons, let alone the existence of a mirror neuron “system”. But perhaps this is a reflection of my own ignorance on the topic. 2. Even supposing such mirror neurons exist, I find the above conjectures (What is the argument?) highly dubious, and 10 years after conjectured, have little to support them. 3. Nichols & Stich (2003) and Goldman (2006) both incorporate aspects of simulation theory and theory-theory, and I also think any plausible view is likely to be a hybrid. So, in that respect, MNs don’t seem to provide evidence of ST against TT in general, but perhaps just certain claims, or “pure” versions of TT. In fact, when I read Goldman (2006), it struck me that his arguments for anything other than low-level (emotional) simulation are incredibly implausible, and the MN stuff does little to help (see Saxe and Carruthers’ responses to Goldman in the phil studies book commentary) 4. The second argument is dubious: if you think MN have nothing to do with mindreading, which I think is the right position, then there is of course no problem that TT predicts “nothing like” MN. And even if MN does have something to do with mindreading, this is only a criticism of an overly pure TT. 5. Consensus? Really? Wow, maybe I am swimming in small circles (near DC), but I have met far more people who find the mirror neuron-to-mindreading connection to be dubious, than people who support it.

    Thanks for bringing up these issues!!!

    brendan: thanks for your comments!
    i agree with a lot of what you say, but i’ll note that i do think there is good reason to believe humans have MN systems and that MNs have *something* to do with mindreading– just what that something is is a very contentious matter. and at least as far as what’s published on mindreading and MNs, there does seem to be a consensus that MNs support ST.

  7. For those interested in what happens if the mirror neuron systems are damaged (by which I mean ablated or pharmacologically inactivated) there isn’t much data. The one article I found was cited in the review article Shannon mentioned:
    L.J. Buxbaum et al., On beyond mirror neurons: internal representations subserving imitation and recognition of skilled object-related actions in humans, Brain Res. Cogn. Brain Res. 25 (2005), pp. 226–239

    It shows that deficits in recognition and deficits in performance of the same act are correlated.

    Are there other primary research articles that address the effects of ablation/inactivation on MN regions?

    More generally, I’m still not convinced there is a substantive difference between the ‘theory’ and ‘simulation’ views. Simulations are perfectly good ways to implement a theory.

  8. Shannon Spaulding

    hi eric,

    there is some stuff on anger mirror neuron systems that may interest you.  Lawrence, et al (2002) showed that when subjects undergo a drug-induced of the dopamine system, their ability to recognize a face as angry is temporarily selectively impaired. when the drugs wear off, the subjects are again able to recognize faces as angry.  article here

  9. Mitch Herschbach

    I’ve been working on a paper questioning the interpretation of all mirroring as simulation. It developed out of a related (and still in progress) project analyzing the connection between the sense of “simulation” in the simulation theory of mindreading (focusing on Goldman’s account) and accounts of motor control pitched in terms of control theory (which Goldman addressed briefly in his 2006 book, and is prominent in Susan Hurley’s late work on the “shared circuits model”). There I argue that the motor control mechanisms called “forward models”, when re-used for goal/intention attribution, are better described as performing theorizing than simulation. An early version of this paper is available here.

    This analysis extends naturally to the issue of mirroring because some researchers have explicitly identified F5 mirror neurons as implementing a sensory forward model. I argue that this example is a case of neural mirroring found during mindreading that should not be interpreted as simulation, and try to develop what further conditions are required for simulation beyond interpersonal resemblance (the core of Goldman’s definition of simulation). This paper is still in its early stages, but I’ll be presenting a version of it at the 2010 APA Pacific meeting in San Francisco. Comments on it would definitely be appreciated.

Comments are closed.