Looking into Ramachandran’s broken mirror


I visited Vilayanur S. Ramachandran‘s lab at the University of California, San Diego recently, and interviewed him and several members of his lab about their work. Rama and I talked, among other things, about the controversial broken mirror hypothesis, which he and others independently proposed in the early 1990s as an explanation for autism. I’ve written a short article about it for the Simons Foundation Autism Research Initiative (SFARI), and the transcript of that part of the interview is below. I also wrote an article summarizing the latest findings about the molecular genetics of autism, which were presented in a symposium held at the Society for Neuroscience annual meeting last November.

MC: Autism is an umbrella term referring to numerous conditions. Can the broken mirror hypothesis account for all of them?

Ramachandran: Autism is characterized by a specific subset of symptoms. There may be three or four that are lumped together, but by and large it is one syndrome, as good a syndrome as any in neurology. It’s not like dyslexia, where there are half a dozen or a dozen types. With autism, people are debating whether high functioning and low functioning autistics should be lumped together or not. There’s a tendency to group them together rather than saying they’re distinct.

We have suggested that the mirror neuron system is deficient in autism, and there’s mixed evidence of that, but most groups support our view. [Marco] Iaconobi‘s group at UCLA did a brain imaging study showing that the mirror neuron system is deficient, but others claim that it’s normal. That may partly be based on the heterogeneity of autism. The mirror neuron system itself could be normal but its projections, or the regions it’s projecting to, could be abnormal. It’s still up in the air.

One of the things I say in my book [The Tell-Tale Brain] is that the mirror neuron system allows you to take an allocentric view of other peoples’ actions, to view the intentions to their actions. It may even be turning inwards and looking at one’s self from an allocentric perspective, so it may be partly contributing to self-awareness. In addition to an allocentric perceptual view, the same system then evolved into adopting an allocentric conceptual, or metaphorical, view – “I see your point of view”. This could have been an evolutionary step from perception to conception, but we don’t know exactly when than magic line was crossed.

MC: There’s another hypothesis, which states that theory of mind is defective in autism. What are the similarities and differences between this and the broken mirror hypothesis? It seems to me that the two aren’t necessarily mutually exclusive, because mirror neurons are likely to be involved in theory of mind.

Ramachandran: The broken mirror hypothesis and defective theory of mind are complementary. It’s like saying genetics excludes DNA, or something. They complement each other. The broken mirror theory is the one we proposed. We also proposed in the same paper that the salience landscape is defective, due to a derangement of connections between the amygdala and other limbic structures. Normally you assign zero salience to that [points to coffee] – well, it’s tasty, so I assign some salience to it – but not much salience to that [taps the table]. The brain is constructing a salience landscape, and in autism that gets messed up, for some reason. You get trivial things provoking a fight or flight response, so you get the autonomic storms that characterize autism.

MC: You did an early experiment showing mu wave suppression in one person with autism. Have you obtained any more evidence for the phenomenon since then?

Ramachandran: We repeated it on 25 or 30 autistic kids and got the same result, and others have found evidence supporting the idea. Mu wave suppression occurs when they perform an action, but not when they watch somebody else performing it. There are people, like Iacoboni, who used functional neuroimaging to show mu wave suppression. The salience landscape idea was proposed with Bill Hirstein, and we published that in Proceedings of the Royal Society about 10 years ago. It’s also described in our broken mirror article in Scientific American.

MC: One recent study failed to find any evidence for the existence of mirror neurons in the human brain, and some say there’s a lack of evidence that the human brain actually contains mirror neurons.

Ramachandran: That’s nonsense. First of all, there’s functional brain imaging data. Secondly, why would the monkeys and anthropoid apes have mirror neurons, only to suddenly disappear in humans? It doesn’t make intuitive sense and empirically it’s flawed.

On the other hand, what people have said is that there’s a lot of media hype about the magical properties that have been attributed to mirror neurons. But just because there’s hype, doesn’t mean it’s wrong. That’s a stupid argument.

The third argument is that some of the properties attributed to mirror neurons are acquired through learning. Every time a monkey reaches and grabs a peanut, a motor neuron fires when it sees its hand reach out, creating a Hebbian link, and the same neuron fires when it watches a peanut being grabbed. The counter-argument there is, why is it that only 30% of the neurons are mirror neurons? If it’s just Hebbian association, what happens to the other neurons? Why don’t they learn? So there are some hardware constraints. Secondly, even if it is learnt, what’s the big deal? If we’re interested in knowing what the mechanism is, the question of nature versus nurture is orthogonal. The existence of stereoscopic neurons and their circuitry is what you want to understand. How the behaviour is learned, or acquired or can be modified is a separate issue.

There’s also a strong anti-reductionist streak, with psychologists saying it’s the same as theory of mind. That’s an idiotic argument, it’s like saying that if you know Mendellian genetics you don’t need to understand DNA. Mirror neurons provide a mechanism for the theory of mind, not a substitute for the idea. They complement each other very nicely. So that criticism goes down the tube.


  1. #1 Jon Brock
    March 10, 2011

    “even if it is learnt, what’s the big deal?”

    If you’re trying to argue that dysfunctional mirror neurons *cause* autism (rather than vice versa) it’s pretty crucial, non?

  2. #2 Mo
    March 10, 2011

    Oui, but are you suggesting that defective mirror neurons might be an *effect* of autism?

  3. #3 Jon Brock
    March 10, 2011

    I have no idea to be honest, but it seems a reasonable hypothesis (assuming that MNs are dysfunctional in the first place).

    In general, I think people are way too quick to assume that brain abnormalities (usually identified in studies of adults or older kids) cause autism rather than the other way around.

  4. #4 Evelyn Wolke
    March 10, 2011

    I correspond with (and know personally) folks who are very functional and still bear the “autistic spectrum/Aspberger’s” diagnosis. They might argue that their mirror neurons are not “dysfunctional” but “different”, and I’ve even had one say that autistics are a human species. I think that’s going too far, but it’s tough to judge when someone articulate like that has such a different way of perceiving the world.

    Perhaps those folks are the ones Dr. Ramachandran views as “lumped in with others”, who are more globally affected (and I worked with many in my younger days, in MR/DD settings). The book is incredibly fascinating – I’m about 2/3 of the way through it now. Thanks for the great interview.

  5. #5 Chris Phoenix
    March 11, 2011

    Interesting that he compares autism and dyslexia.

    His statement that there are 6-12 types of dyslexia… I’d restate that to say there are 6-12 _major symptoms_ of dyslexia, depending on how you lump them.

    If you’re a researcher looking to invent a dyslexia cure, you find a sensory modality that dyslexia affects, and an exercise that reduces that effect, and you observe that the exercise improves dyslexics’ reading. So you think you’ve discovered the cause of dyslexia.

    If you try to understand how Geiger and Lettvin can correct dyslexia with visual exercises, and Merzenich and Shaywitz can correct dyslexia with auditory exercises, you might conclude that there are various types of dyslexia.

    But take another step back and notice a few more things.
    1) Multiple researchers, studying and manipulating various sensory modalities, claim improvement in well over 50% of dyslexics.
    2) Some dyslexics have “conditional dyslexia” – the severity of their disability varies, sometimes in just a few seconds.

    Point 2 suggests that whatever is going on in dyslexia, it is not just a simple dysfunction. I’d argue that there’s a psychological component – that it’s triggered by strong emotion, especially confusion.

    Point 1 suggests that whatever is going on, it’s at least sometimes multisensory… and that normalizing any of the senses can mitigate the underlying mechanism.

    I’ve done pilot studies that seemed to show that both vision (apparent motion in blinking lights) and hearing (classifying a range of synthesized speech sounds) could be affected by going in and out of “dyslexic mode.” But I lost access to the lab before I could ramp up the studies.

    The flip side of perceptual distortion is perceptual flexibility. It may be a stretch, but what if dyslexic-type sensory distortion were actually useful… except when trying to interpret symbols?

    (The confusion link, and the usefulness of perceptual flexibility, are from Ronald Davis’s work. Davis would probably argue that all dyslexics are conditional dyslexics.)

    Anyway, back to autism… after seeing how many different symptoms can arise from part-time sensory distortion, I could believe almost anything could arise from a relatively simple but more chronic processing difference in autistic brains. I know less about autism than about dyslexia, but I tend to lean toward attentional problems as one of the roots of autism: if you pay as much attention to random things as to people and people-salient stimuli, you’re likely to 1) abreact to small stimuli, 2) not develop your mirror-neuron circuitry very well, 3) not learn to generalize properly about people-relevant problems. From these, lots of other symptoms can emerge.


  6. #6 Wayne Williamson
    March 14, 2011

    nice article…I think that mirror neurons are needed to create a socially functional person…If you cannot imagine what others are thinking, how can you interact with them…

  7. #7 David Johnson
    March 23, 2011

    Thanks for the nice interview… I’m not sure I’m familiar with the latest research regarding this issue, but just wanted to point out that there is some evidence disassociating the neural circuitry underlying theory of mind from the mirror neuron system. There are, at least, brain regions unique to each system. Also, primates have mirror neurons but don’t seem to possess theory of mind…I’m not sure that its clear the degree to which these systems are related. Thanks again for all the blogging!

  8. #8 Birger Johansson
    March 24, 2011

    (OT) -A reminder that you do not need to have a diagnosis of some neurological condition to behave less then 100% logical!

    Cerebral prejudice favours the beautiful“.

    Another case of how subtle effects twist our thinking…”The important thing here is that, if we are aware that these stereotypes exist and affect our cognitive processes, we have a possibility to counteract them,”

  9. #9 Andreas
    March 24, 2011

    Interesting topic. Connecting to David Johnson’s comment, there is this meta-analysis by van Overwalle and Baetens (2009) showing little or no overlap between fMRI activations reported from studies of ToM (called “mentalizing”) and the MNS (Mirror Neuron System), respectively.

    This could pose an argument against the idea of a “ToM system” and MNS working together in social perception. However, I think that the real problem here is how ToM is conceptualised and measured. Most legacy ToM tasks is about conscious reasoning about the mental states of other people, and in fMRI studies participants are normally instructed to reason as such. This kind of ToM is hardly a candidate for use in most situations of social interaction, due to the speed limits of conscious reasoning.

    This said, there are some emerging evidence of automatic ToM-like processing, a topic which I think would be interesting to study in conjunction with the MNS. This is presumably more interesting also for autism, given that many high-functioning autistic individuals eventually learn to handle verbal ToM-tasks but still has problems with social interaction.

  10. #10 David Johnson
    March 24, 2011

    Andreas – Thanks for the link to the meta-analysis. That’s what I was thinking of but couldn’t recall the source.

    I worked in a lab for a couple of years that was focused on the neural correlates of mentalizing, and most of the studies were designed to capture automatic, nonconscious processes (e.g. subjects weren’t instructed to explicitly consider the mental states of others but the tasks were designed such that it would have been difficult to complete them without mentalizing). Our data showed that there is a distinct network of brain regions preferentially active when one is engaged in using one’s self, or one’s own mental state, as a proxy to understand another persons mental state, e.g. intentions, beliefs, preferences and such. The more similar that target is to the self, the more people seem to use themselves as proxies. My understanding about mirror neurons is that most of the evidence seems to be more related to imitation, particularly of motor actions, and not as much to mental states per se. That’s not to say there might not be some overlap, but I think the case is way overstated.

  11. #11 Andreas
    March 25, 2011

    David Johnson – That’s interesting. The data you mention, are they published? I don’t readily recognise having read about what you describe, but it sounds interesting to me. I have been thinking of doing EEG of mentalising as a part of my PhD, but I’m hesitating since I’ve had problems designing a task to capture implicit mentalising over repeated trials. The problem as I see it is to prevent the participants from seeing through the task structure and adopting some task-specific heuristics not related to mentalising. So I would be interested in how you did it (I guess fMRI was used)? (You can email me if you like: atbf dot spamcollect at, or post a link here)

    I think that I agree with you on mirror neurons. From my understanding the MNS is also sensitive to goal-directed action, but it’s not obvious that understanding such actions necessitates mental state attribution. The opposite is more consistent with van Overwalle and Baetens, showing ToM activation in conjunction with goal-directed actions only when the actions did not make sense to the participant. Put another way, understanding a goal-directed action is one thing and understanding it’s reason is something quite different.

  12. #12 russell
    April 4, 2011

    about the lack of mirror neurons:

    from a purely theoretical standpoint – mirror neurons are just the same neurons that fire when the monkey watches the other monkey eat a banana as when the watching monkey is actually eating the banana.

    close your eyes, and pretend you`re eating a banana. there, your `mirror neurons` just fired.

    actually, not really…mirror neurons do not exist. sensation is neurons firing, and when you imagine yourself doing something, its the same neurons firing…just less of them.

    for example. imagine how it would feel if you flicked yourself in the nose.

    now, flick yourself in the nose.

    now, imagine how it feels, again.

    which imagined flick was more vivid, more real? the second one obviously…the brain had better access to the sensation and could recreate it more accurately.

    my bet would be that ”mirror neurons” would NOT fire if the monkey had no experience of what it was like to eat a banana and watched another monkey eating one.

    do you get my meaning? that was a bit of a rant, i can be more clear if you want..

  13. #13 mmozart
    April 13, 2011

    Digression…science bloging is something like brainstorming hub. Most of science blogs are actually news blogs but there are some where writers publish their own ideas, work etc. Using blog, they get feedback, opinions, new interesting ideas and that makes their work event better. :) Even through comments on news you can share something interesting,your attitude, opinion what can make buterfuly effect. – Just to share my thoughts. :) Greetings

  14. #14 Mike Alexander
    May 19, 2011

    My 4-year-old son has a high-functioning autism diagnosis. One of his most conspicuous symptoms is an apparent inability to mimic the physical actions of his peers. He can’t seem to learn physical tasks by straightforward imitation; he needs to explore how to do these tasks for himself – learning the hard way. This seems more obviously and directly related to the mirror neuron hypothesis than the social imagination/theory of mind aspects of autism.

  15. #15 C. W.
    July 19, 2011

    Ramachandran talking about autism:
    “….by and large it is one syndrome, as good a syndrome as any in neurology.”

    One syndrome? Really? So all of those countless people who have been diagnosed with one from a huge range of rare genetic syndromes and are also diagnosed as autistic must therefore coincidentally have two different genetic syndromes, autism and another rare syndrome, and the two syndromes are completely unrelated. Sounds like bullshit to me. I’ve gone to the OMIM database and searched on the term “autism” and retrieved 162 different records. Go ask a medical geneticist if autism is one genetic syndrome. I’m sure you will induce laughter.