Mirror neurons have been in the news a lot lately, performing a dizzying array of functions, from the fight against pornography to explaining why World Cup Football is a good thing, to explaining the banking crisis. It seems like people love to use mirror neurons to explain just about anything they want to, it doesn’t really matter what. And they don’t seem to know a whole lot about mirror neurons, other than vague associations with “empathy” and “imitation”. These are common enough things in human affairs that mirror neurons can intrude anywhere, make any opinion seem more scientific.

This profligacy is not limited to lay people and journalists. Scientists have implicated them in things as varied as understanding intentions, empathizing with others, the development of language, a possible role in autism, and a host of other things. Hundreds of papers on these subjects have been published.  Can mirror neurons indeed have such near-miraculous explanatory powers?

What are Mirror Neurons?

What are mirror neurons, and why all this fuss about them? In brief, mirror neurons are certain neurons in the brain that fire not only when the individual performs an activity, but also when the individual observes someone else perform that activity. They were discovered in the 90’s by an Italian scientist named Giacomo Rizzolati, in the macaque monkey. He observed that in the F5 region of the monkey brain (part of the premotor cortex, a region concerned with planning actions); there was a certain population of neurons which had some peculiar properties:

  • They fire both when the monkey performs an action, and when the monkey sees another monkey (or human) perform that same action. Typical actions in which such activity is observed are things like reaching for an object, or grasping an object.
  • They fire only when the actions are goal-oriented, meaning when there is a perceptible goal to the action, some object towards which the action is directed, for example, reaching for food. They do not fire when the monkey observes another monkey performing meaningless actions, such as waving its arms randomly.
  • Further, they fire only when the monkey knows what the goal is. Performing a pantomime (reaching for food when food is not there) will not produce any activity in these neurons. However, if the monkey knows the object is there even though he cannot see it, then they will fire. For example, if the observing monkey sees the food, but then a screen is placed in front of the food, and he sees the acting monkey reach for the food behind the screen, these neurons will fire, even though the object (food) is itself hidden at the time. This is because the monkey saw the food previously, and knows it’s there.
  • They don’t fire when tools are used to perform the actions. For example, if food is delivered simply by pressing a button, there is no activity in the mirror neurons when this is observed. Only the observation of direct, goal-oriented action will provoke the activity of the mirror neurons.
  • They don’t respond to partial activity. They won’t fire if you show the monkey the food, or show the monkey another monkey looking at the food. Only when the other monkey reaches for the food will they fire in the observing monkey. Further, the response is quite specific for the type of activity. For example, consider two actions – reaching for an object and placing it in one’s mouth (as with food), and a different action, picking up an object and placing it in a container. Both are goal-directed activities, and both can generate mirror neuron responses in the observer, but different populations of neurons are involved in each response. The mirror neurons which fire when the monkey observes another monkey reaching for food and placing it in its mouth are different from the mirror neurons which fire when the monkey observes another monkey picking up some object and placing it in a box.
  • The neurons will fire as early as the initiation of activity (as soon as the monkey observes another monkey initiating the activity), even before the action is complete. In other words, they fire as soon as the monkey observes another monkey reaching for the food, before it has actually reached the food, grasped it, brought it back to its mouth, and deposited it there. Now it’s possible that some actions could be ambiguous, for example, a monkey may reach for an object either to place it in its mouth, or to place it in a box. However, in most tested examples where the activities are clearer cut, the appropriate population of mirror neurons for that activity begins to fire at the initiation of that activity, suggesting that the observing monkey is predicting the subsequent course of the action.
  • They don’t fire during imitative activities. Adult macaque monkeys do not imitate, but juvenile ones do. Imitation is not a “goal oriented” action in the previous sense. A juvenile monkey will imitate any meaningless action, for instance, if you stick your tongue out at it, it will stick its tongue out back at you. In juvenile macaques, mirror neurons do not fire during imitative activities – not during observation of the activity, and not during subsequent performance of that activity.
  • They don’t fire in preparation for the activity, only when the activity occurs. This is important because there are areas nearby in the monkey brain (for example, area 6) which have “set neurons” that fire in anticipation or preparation for an action, before the action is initiated. This is not true for mirror neurons.
  • They don’t seem to represent covert activity, that is, activity which happens in the brain but is not implemented by the body. More specifically, covert activity would be the brain going through the motions of reaching for an object, and yet the hands don’t move, the monkey doesn’t actually reach for the object. Covert activity is ruled out by the observation that during the mirror neuron firing, there is no corresponding activity in the primary motor cortex. For covert activity, one would expect some activity in the primary motor cortex, which does not get translated into movement of arm muscles because it’s sub-threshold, or because it’s blocked in some way. Such covert activity is not recorded.
  • Finally, it’s important to remember that mirror neurons are only a subpopulation of the neurons in the F5 area. Different estimates put them at somewhere around 20%-40% of population. The rest of the neurons in F5 which do not show mirror activity are called “canonical” neurons. Of the subpopulation of mirror neurons, some might respond only to visual stimuli, some only to auditory, and some only to somatic sensory information. Further, of those that respond to say visual stimuli, some might respond only to certain kinds of actions and others to a different kind of action, as mentioned earlier.

These observations on macaque monkeys were the basis of our understanding of mirror neurons. Any theory about mirror neurons must explain these observations. So, what are these mirror neurons, what are they doing, and why?

Rizzolati and others initially explained mirror neurons as a form of “action understanding”.  In Rizzolati’s words:

“Each time an individual sees an action done by another individual, neurons that represent that action are activated in the observer’s premotor cortex. This automatically induced, motor representation of the observed action corresponds to that which is spontaneously generated during active action and whose outcome is known to the acting individual. Thus, the mirror system transforms visual information into knowledge.”

Nielsen put it this way:

“A mere visual representation [of an action], without involvement of the motor system, provides a description of the visible aspects of the movement of the agent, but does not give information critical for understanding action semantics, i.e., what the action is about, what its goal is, and how it is related to other actions.”

What does this mean? They are basically saying that in a pre-language sense, our brain has a certain innate “vocabulary” which it uses in planning. Part of this vocabulary has to do with motor actions, and that is assisted by mirror neurons. So for example, our brain has a “word” or a “concept” or a “symbol” or some discrete “thing” (to be as non-committal as possible) which represents the action of grasping, and that mirror neurons are an automatic mechanism that identifies this action when we observe others do it. By activating the same neurons that we would activate if we ourselves performed this act, we can recognize or identify when others are doing it. This adds to our knowledge in some sense – by observing the other monkey stretch its arm out and grasp the food, we gain the knowledge that the monkey is going to eat the food, because our mirror neurons are firing in the pattern in which they would if we were to reach for the food and grasp it.

This is an interesting point to consider, especially if we generalize it beyond monkeys to humans. But before we do that, let’s consider a few problems with the monkey model. Biology is very messy, and for every case where someone says “X”, another person says “not-X” and often the evidence for both is inconclusive.  So what are some of the objections against this “action understanding” theory?

First, there is some evidence that disruption of the F5 area (the mirror neuron area postulated to be involved with “action understanding”) does not disrupt “action understanding”. That is, if you destroy the area supposedly responsible for this function, the function does not go away. This is sort of fuzzy, because some experiments seem to show that lesions of F5 do indeed disrupt some forms of action understanding, while others don’t support the idea. So I’ll leave this alone for now until we have more data.

Second, we know that action understanding can happen without mirror neurons. The visual system has the job of identifying objects we see, understanding relationships between them, and using that information as the basis for action. This can occur quite independently of F5. For example, the Superior Temporal Sulcus (STS) has cells which are much more sophisticated than the F5 mirror neurons in distinguishing between different types of observed actions, and they respond strongly when monkeys observe other people do various actions. They don’t, however, have a motor component. They do not fire when the monkey himself does the action. So it’s possible to have action understanding without the corresponding motor part.

Third, it now seems that there are mirror neurons in the monkey’s M1 region, or primary motor cortex. This is a problem because we assumed that mirror neurons aren’t simply responding to a covert action, they are actually helping us process information by helping us identify the actions of others. One thing that supported this view was that there was no activity in the primary motor cortex when one monkey observes another monkey act. But now that we know that there are mirror neurons in M1, this foundation is somewhat shaken, and it’s harder to rule out that what we are seeing is simply an associational covert action. Humans do this too, if you observe a sports fan watching his favorite game on TV. He will contort his body into actions mimic the player he’s observing, almost “willing” the player to respond as he does. This does not need to be explained via “action understanding”, but simply by the fact that the sports fan is reacting as he thinks the player ought to act. He may in fact act differently from the player on the screen, because in his judgment a different action was called for.

Mirror Neurons in Humans

This brings us into the problem of generalizing from macaques to humans. While there is some evidence that humans also have mirror neurons, there are a number of key differences, including:

  • Humans show “mirror neuron” activity even for imitation, while monkeys do not. Monkeys only show activity when the action is goal directed, not when it’s random.
  • Humans react to pantomime displays, while monkeys do not. When the object isn’t really there, but the actor does a good enough job of pretending that it is, humans will show mirror neuron activity as if the missing object were really there. Monkeys don’t do that. They need to know that the object is there, even though it may be hidden during the experiment.

Some researchers explain these discrepancies by saying that the human system is more “evolved” and responds to a wider range of stimuli. This may be so. However, it raises two problems. First, it shows that there are significant differences between humans and monkeys in this regard, and therefore research that was done on monkeys should not be uncritically generalized to humans, as it often is. Second, it weakens the original argument for monkeys, which was that mirror neurons are responsible for “action understanding”. Imitative or random behavior is obviously not goal directed in this sense, and therefore there is nothing corresponding to understand. Why then, are human mirror neurons activated in such cases?

The data for humans and monkeys isn’t easy to compare. There are a lot of studies at the single cell level with implanted electrodes that are available for monkeys. Parallel data for humans is non-existent or very scarce, because of the obvious problems with implanting electrodes into human research subjects. On the other hand, there is a wealth of human fMRI data, which is scarce for monkeys, because monkeys can’t easily be trained to perform tasks inside an MRI scanner. This imbalance of the data types makes direct comparisons difficult.

Some tests which can only be performed on humans show other differences as well. For example, Catmur et al performed a TMS/MEP study on human subjects (TMS is trans cranial magnetic stimulation, in which a part of the brain is stimulated with magnets placed on the outside of the skull. MEP is motor evoked potentials, which are small electrical signals recorded from peripheral nerves) which show dissociation between mirror neurons and “action understanding”. The experimental setup was to stimulate specific motor areas of the brain (through TMS) to produce a twitching of the abductor muscles of the hand and fingers (which was recorded by MEP). Under the standard test condition, the subjects watched a video of someone abducting the index finger of hand, followed by abducting the little finger. When the subjects were watching the video of the index finger being abducted, the MEPs recorded from their own index fingers were stronger than those recorded from their little fingers. When they observed videos of someone abducting his little finger, the MEPs recorded from their little fingers were stronger than those recorded from their index fingers. This was the standard response. However, in the test condition, this was reversed. The subjects were trained to trained to move their index finger when they saw the video of someone moving his little finger, and vice versa. After training, the MEPs recorded were reversed. In other words, the “mirror effect” dissociated from the “action understanding”. Given that this is a crude way to compare what was done will skull electrodes in monkeys, but it is still significant.

This sort of training to overcome default congruent behavior is very common among humans. The fact is that we don’t WANT to act exactly the way we observe someone else acting. If you see the same action, say someone tossing a ball at you, your action will be very different depending upon whether you have a bat in your hand, or if you are the catcher. Our expectation of what they are trying to do depends upon our own role and circumstances. How do monkeys deal with this? We don’t know.

The large numbers of fMRI studies have opened up many interesting areas for study in humans. By its nature, fMRI scans large parts of the brain, compared to the single cell recordings done in monkeys. When you are scanning large areas of the brain, other interesting things turn up. Humans apparently have mirror type neurons in many locations. Among the more interesting regions are the anterior cingulate cortex, the anterior insula, and the inferior frontal cortex. These areas are supposed to have roles in mediating emotions, which leads to the speculation about empathy.

Mirror Neurons and Empathy

We know that observing a certain emotion in others can produce the same emotion in ourselves. This has been amply demonstrated in several studies for emotions such as disgust and pain. Further, when we observe someone else feeling these emotions, there is increased activity in areas of the brain (the above mentioned areas) which are activated when we ourselves experience those emotions. This is strong evidence for mirror neurons in these areas.

There are supportive studies as well. People who self-report to having a high empathy in questionnaires show a greater mirror neuron activation of emotion centers in the brain in fMRI when they are exposed to images of other people feeling those emotions, compared to people who self-report to having lower empathy. Further, not only do the self-reporting empathic people show greater activation in areas to do with emotions, they also show greater mirror neuron activation in tests which have nothing to do with emotions, such as the reaching/grasping tasks described earlier. Does this mean that some people just have a stronger mirror neuron system, which they typically experience as “having more empathy”?

There is some evidence in support of this from the opposite end of the spectrum too. Some autistic people (who have very low empathy) have been found to have markedly smaller/thinner cortical regions that are associated with mirror neurons. Autistic people also generally show lower mirror neuron activity in fMRI and EEG experiments. However, the implications of these findings are disputed, and the matter is not settled.

Mirror Neuron Fever in the Media

While this is extremely fascinating stuff, the fact is that our theories about the role and function of mirror neurons are very hazy at this point. Many of the theories were built on monkey models, which have not been proven to apply to humans. In fact, there are a number of ways in which human mirror neurons are different from monkey mirror neurons, so we should not expect the monkey data to simply generalize to humans without modification. These details are often ignored by media and even by scientists, who have a poor understanding of mirror neurons.

For example, consider this article in the Washington Times about a group lobbying Congress to ban porn because of the effect it might have on children. They found a doctor, a “forensic pediatric physician” to bolster their case.

Quoting from the article:

Pornography normalizes sexual harm, Dr. Cooper said. It shows children a lack of any kind of emotional commitment or relationship between two consensual partners, shows unprotected sexual contact and visual examples often of violent rape.

When a child sees this image of adult pornography, the mirror neurons that are in their brain will convince them that they are actually experiencing what they are seeing,” she said.

Children are very vulnerable as compared to adults because of the presence of mirror neurons in the brain, Dr. Cooper said. Mirror neurons are part of the brain that convince us that when we see something we are actually experiencing it.

This is complete nonsense. It is in fact the opposite of what the theories about mirror neurons say. We don’t experience stuff because we see it, we experience stuff when we do it, and then this knowledge can later be used to understand the actions of another. You don’t experience playing a piano just because you see someone else playing a piano, if you have never played a piano yourself. This is the whole point of mirror neurons – that our brains can and do obtain information and make sense of it in many different ways. On an intellectual and analytical level, we can observe a person playing a piano, understand what he’s doing, get some sort of “action understanding” from the observation. If, however, we ourselves have played the piano before, then to this analytical understanding we can add an “experiential” understanding, and perhaps understand better what it feels like to play the piano, and to understand the person we are observing from an experiential level in addition to an analytical level. Even this is just a theory, and is by no means widely accepted. But to go from this to saying that mirror neurons can convince us that we are experiencing what we see someone do is supremely ridiculous. And then to add that “children are very vulnerable as compared to adults because of the presence of mirror neurons in the brain” is a gratuitously stupid statement. Why would children be more vulnerable compared to adults because of mirror neurons? Because they have mirror neurons while adults don’t? No, adults have them too. Because they have more mirror neurons than adults? No, they don’t. All of which assumes that mirror neurons enable these kids to experience what they see on the screen, which is a wrong conclusion based on a poor understanding of mirror neurons.

This is an obvious example of political activism misusing science to push an agenda, but not all such incidents are so obvious. There are plenty of scientists who have caught on to the mirror neuron idea as well, and use it to push their pet theories. Among the worst offenders are “evolutionary biologists”, which is a category that is fast becoming known for having miles of speculation hanging on to an inch or two of fact.

Mirror Neurons and Morality

One common theory these days is that mirror neurons explain how humans evolved a moral code. The idea is that since we can empathize and feel another’s pain, we know when he’s feeling bad. And for various reasons ranging from social cohesiveness to “seeing him feel bad makes me feel bad”, we came up with the idea that we shouldn’t make others feel bad. This, supposedly, became the basis for some sort of golden rule – don’t do stuff to others that would make you feel bad.

I know this sounds very simplistic, and certainly there are many people who express the same idea much more eloquently, but this is in essence what it amounts to. I have some serious problems with this idea.

First, I am not denying that mirror neurons are real. Of course they are. Second, I am not denying that they may have a role in empathy, in understanding how others feel because we can experientially know their emotion since we feel it ourselves. There are plenty of studies which support that too. However, as a basis for morality, this explanation is very lacking.

Mirror neurons are only one of the ways in which we know things about other people. As mentioned earlier, we have much more sophisticated systems (such as the STS) which allow much finer discrimination when judging and interpreting the actions of others. These systems are not based on mirror neurons at all. Secondly, there are many ways of learning that if you hit someone, he will probably hit you back. While it may help to know “oh, that must have hurt him, I know because I’ve felt pain myself” this knowledge is general enough that we don’t need mirror neurons to remind us of it constantly. If you’ve bumped your toe against a rock, you know it hurts. This does not require mirror neurons. It is not a stretch from that to understand that if you hit someone else with a rock, it will hurt him too. Mirror neurons can add to that, for example by seeing the grimace on his face once he gets hit, which may help you understand the extent and severity of his pain better after you have hit him. If you are empathic, you may feel some pain yourself, from watching his reaction. But nowhere near the pain he felt.  And the fact is, while it may have added something to your knowledge, you certainly knew beforehand, before hitting him, before watching him grimace, that it would hurt. That’s probably why you hit him in the first place, to make him hurt.

Now if empathy was strong enough that our own distress while watching someone else in pain would prevent us from hurting anyone, then we might have an argument. But it isn’t. Even the kindest, most empathic mother may occasionally slap her child for misbehavior. We know from experimental studies that empathy (and the corresponding mirror neuron activation) increases towards people we love. So if it’s still not strong enough to prevent us from hurting people we love, how would it prevent us from hurting those towards whom we are indifferent?

On the other hand, there are several ways in which we can derive the same morality without referring to mirror neurons, which make much more sense. For example, we know that if you hit someone, he might hit back. This dissuades us from hitting people very frequently, because the consequences to ourselves would be unpleasant. We also know that if we live in a family where two people are constantly fighting each other, it can get unpleasant for other family members as well. You don’t want to deal with angry individuals, even though they might not be angry specifically at you. This is why when humans live in social groups, as our ancestors did, it benefits everyone to see that peace is maintained, even those who are not involved in the fighting. These are common sense things that we all know. And they are powerful inducements towards “morality”, if morality is reduced to “don’t hurt other people” in mirror-neuron fashion. And this kind of analysis is available to everyone without any mirror neuron imperatives.

The second and more important question to me is whether “don’t hurt people” is a sufficient basis for morality. Or even a particularly good basis. I don’t think it is. Most of us require a concept of justification when it comes to morality. Someone might perform an action and feel bad about it, and through empathy, we may observe this person and feel bad ourselves because he feels bad. But we may still think that he deserves to feel bad, because of the nature of his actions. This is the essence of justice, which is really the foundation of our morality. A killer may feel bad because society locks him up; we may look upon his face and see fear and misery on it, and perhaps that will evoke fear and misery in our own minds. But our idea of morality might be that he deserves his fate, because he took a life. You can call that balancing empathies, if you will – empathy towards the victim versus empathy towards the killer. But if it is empathy, it’s not based on mirror neurons. We don’t have to see the killer’s face, we don’t have to observe the act of the murder, and we don’t have to see his victim’s face. None of the things that mirror neurons might have a role in are required.  All you need are analytical concerns, such as “is this justice” or “he can’t kill anyone else if he’s locked up”.

You can ask the same question in another way. If you cheat on your spouse, and you know that your spouse will never find out, is it okay? If your sole concern was empathy and “don’t hurt people”, then you might consider it okay, since your spouse will never find out and never be hurt by your action. However, many people might disagree based on other grounds. Such as, you promised to be faithful. Why should you keep your promise? If you do, I think it’s because you value integrity. That is the crux, not empathy.

So I don’t consider empathy to be a sufficient basis for morality. Considerations like justice and integrity, concern for the viability of the society in which you live by having rules that promote harmony – these are all essential for a system of morality. And as mentioned earlier, empathy isn’t the greatest or most powerful reason for the golden rule either. As the rule itself says “do unto others as you would have them do unto you” – meaning, the stress is on the consequences, on the “as you would have them do unto you”. What are the implications? That if you want to be treated right, you better treat others right. If you expect consequence Y, do action X. Why? Because that’s justice. Empathy is part of it, but by no means the only part. Nor is empathy even necessary to understand that people might feel bad if you treat them badly.

Mirror Neurons and Philosophy

Finally, I’d like to touch on some philosophical implications of mirror neurons, which I think are pretty exciting. One well known fact (and problem) in philosophy is that each of us only ever has access to his own mind. We never really know what another person experiences, what they feel. We can draw inferences from their behavior, but we cannot directly experience what they experience. This leads to philosophical views that can verge on the absurdly solipsistic – “I am the only person, everyone else is just an entity that responds in certain ways to certain things I do or observe. Perhaps they are not even real; perhaps they are just creations of my mind”.

There is no cure for true solipsism. If everything is a figment of your imagination, then you have no possible way to establish the reality of anything but your own mind. This is technically true, but uninteresting, since it leads nowhere. At best, you have to admit that the figments of your imagination behave in somewhat predictable ways, just as they might if they were real. Whether you continue to call them figments of your imagination or real make little difference to how you behave. If you believe that truck you are imagining can kill you, then you will move out of its way whether it’s real or not. It becomes an exercise in semantics.

However, there is a kernel of a real problem here. Even if we admit that other people exist, we don’t know what goes on inside their minds. There are many approaches people take to overcome this – by making certain assumptions (he’s a human like I am, with the same sort of machinery to think, reason, feel, that I do), and we can interpret their behavior in light of our own experiences. However mirror neurons provide an extra link in this chain. If mirror neurons allow us to experience what the other person experiences more directly, then this is an automatic connection between his brain and ours, in that we have within our own brains a parallel to what is happening inside his brain. By becoming aware of these parallels in our own minds, we have some sort of understanding that is more experiential than simply observing his behavior.