The mirror-neuron hypothesis also lends insight into autistic language difficulties. Mirror neurons are almost certainly involved when an infant first repeats a sound or word that she hears. It may require internal translation: the mapping of sound patterns onto corresponding motor patterns and vice versa. There are two ways such a system could be set up. First, as soon as the word is heard, a memory trace of the phonemes (speech sounds) is set up in the auditory cortex. The baby then tries various random utterances and, using error feedback from the memory trace, progressively refines the output to match memory. (We all do this when we internally hum a recently heard tune and then sing it out loud, progressively refining the output to match the internal humming.) Second, the networks for translating heard sounds into spoken words may have been innately specified through natural selection. In either case the net result would be a system of neurons with properties of the kind we ascribe to mirror neurons. If the child could, without delay and opportunity for feedback from rehearsal, repeat a phoneme cluster it has just heard for the first time, that would argue for a hardwired translational mechanism. Thus there is a variety of ways this unique mechanism could be set up. But whatever the mechanism, our results suggest that a flaw in its initial setup might cause the fundamental deficit in autism. Our empirical results with mu-wave suppression support this and also allow us to provide a unitary explanation for an array of seemingly unrelated symptoms.
Finally, although the mirror-neuron system evolved initially to create an internal model of other people’s actions and intentions, in humans it may have evolved further—turning inward to represent (or re-rep-resent) one’s own mind to itself. A theory of mind is not only useful for intuiting what is happening in the minds of friends, strangers, and enemies; but in the unique case of
This may be a good place to add three qualifying remarks. First, small groups of cells with mirror-neuron-like properties are found in many parts of the brain, and should really be thought of as parts of a large, interconnected circuit—a “mirror network,” if you will. Second, as I noted earlier, we must be careful not to attribute all puzzling aspects about the brain to mirror neurons. They don’t do everything! Nonetheless, they seem to have been key players in our transcendence of apehood, and they keep turning up in study after study of various mental functions that go far beyond our original “monkey see, monkey do” conception of them. Third, ascribing certain cognitive capacities to certain neurons (in this case, mirror neurons) or brain regions is only a beginning; we still need to understand how the neurons carry out their computations. However, understanding the anatomy can substantially guide the way and help reduce the complexity of the problem. In particular anatomical data can constrain our theoretical speculations and help eliminate many initially promising hypotheses. On the other hand, saying that “mental capacities emerge in a homogeneous network” gets you nowhere and flies in the face of empirical evidence of the exquisite anatomical specialization in the brain. Diffuse networks capable of learning exist in pigs and apes as well, but only humans are capable of language and self-reflection.