What are mirror neurons? Mirror neurons are multimodal association neurons that increase their activity during the execution of certain actions and while hearing or seeing corresponding actions being performed by others. Neurons responding to the sound or sight of some actions, but only to the execution of different actions, are not mirror neurons. Where are mirror neurons found? Three research groups have reported the existence of mirror neurons in three regions of the macaque cortex (Figure 1). Pending systematic explorations, we do not know whether mirror neurons exist elsewhere in the macaque brain. Recently, mirror neurons have also been reported in the song-bird. Do humans have mirror neurons? This issue has been highly contentious, with no individual piece of evidence generally accepted as definitive, but quite a lot of indirect evidence for human mirror neurons has been reported. First, if a subject moves, the power of the mu-rhythm in the electro-encephalogram (EEG) recorded from his or her brain decreases. Similarly, the EEG rhythm desynchronizes when the subject observes somebody else move. Second, behavioral experiments indicate that the execution of an action is facilitated by viewing someone else execute a similar action, but hindered by viewing an incompatible action. Moreover, transcranial magnetic stimulation (TMS) studies evidence that watching performance of an action facilitates the motor cortical representation of the muscles involved in doing the same action. This shows that some neurons involved in performing an action are indeed selectively activated by seeing a similar action — in other words, mirror neurons do exist somewhere in the human brain. If humans do have mirror neurons, where are they? Disrupting activity in the ventral premotor cortex using repetitive TMS reduces the effect of viewing other people perform an action. Functional magnetic resonance imaging (fMRI) has shown that a great number of locations in the brain that are active during the execution of an action are also active when the same action is seen or heard; these include the premotor and parietal regions in which mirror neurons are found in monkeys (Figure 1). Unless one abandons the concept of evolutionary continuity, the conclusion that mirror neurons akin to those found in monkeys (and birds) exist at least in these regions of the human brain is most parsimonious. The results of fMRI experiments, however, suggest that a number of additional brain regions have mirror properties in the human brain, including the dorsal premotor, supplementary motor, the primary and secondary somatosensory cortex, the posterior middle temporal gyrus and parts of the cerebellum (Figure 1). Whether these shared activations indeed reflect the activity of mirror neurons in each of these regions remains to be established: some could also contain distinct populations of neurons active during the perception and the execution of actions that just happen to share the same voxel. Pending further single cell recordings, it would seem best to refer to these human brain regions as parts of a putative mirror neuron system. Is there evidence against the existence of mirror neurons in humans? Not really: for each experiment that fails to find evidence for mirror neurons in humans there is at least one that succeeds. For example, recent experiments used fMRI to test the prediction that, for a brain region that is part of the mirror neuron system, repeated viewing of an action would lead to reduced activation during subsequent execution of the same action (and vice versa): three of the four experiments that tried found such an effect. Given statistics that limit false positives to <5%, this ratio of 3:4 is strong evidence for the existence of human mirror neurons. But if there are mirror neurons, shouldn't all experiments find evidence for them? A basic power analysis falsifies this intuition: using the typical voxelwise thresholds of p < 0.001 in fMRI, one would expect even large effect sizes to remain undetected over 50% of the time. Do mirror neurons help us perceive the actions of others? Experiments in which brain areas thought to contain mirror neurons were disrupted, either using repetitive TMS or as a result of localized brain damage, have demonstrated a drop in the accuracy with which participants can report what action another individual has performed — particularly for those actions they can no longer perform accurately themselves. Often, however, the accuracy remains above chance. These results indicate that brain areas thought to contain mirror neurons are indeed contributing to our perception of the actions of others, but that they are unlikely to be the only player. Can we develop new mirror neurons? Hebbian learning suggests that performing an action while seeing and hearing oneself perform it should be enough for neurons involved in performance to start responding to the sight and sound of the same action. The fact that five hours of piano lessons suffice for the premotor cortex to start responding to piano music supports this view. Is the mirror system broken in patients with social deficits? Given that the mirror neuron system is implicated in understanding the actions of others, it has been widely suggested that defects in the system may play a part in neurological disorders characterised by social deficits, in particular autism. While some studies do show that, in individuals with autism, the putative mirror neuron system is activated less than usual while they view the emotions or actions of others, other studies do not. We now need to find out when autistic individuals may activate this system less, and examine whether this can help us understand this disorder. Are mirror neurons the neural basis of mind reading, empathy and language? Activations in brain regions involved in executing actions have been measured while people try to read the minds of others, empathize with them or listen to spoken language. Examining how much of that activity really stems from mirror neurons, and in particular to what extent there is a causal link between this activity and these mental functions is a key challenge for future research and will require TMS and lesion studies. Do we have mirror neurons for emotions or sensations? A handful of experiments suggest that brain regions involved in the experience of emotions and sensations become reactivated while we view the emotions and sensations of others. These regions might therefore contain mirror-like neurons for emotions and sensations. Pending single cell recordings, this conclusion remains tentative, though it is an influential idea. Why is the mirror neuron system so controversial? About a dozen papers have reported direct evidence for mirror neuron activity in monkeys and birds. Approximately 100 times as many papers refer to mirror neurons without directly recording their activity, often implying a link between mirror neurons and higher cognitive functions. While the existence of mirror neurons in animals is beyond doubt, the causal relationship between these neurons and phenomena such as empathy, mind reading, language, autism, esthetics, morals and politics is so poorly established, that the frequency with which the term mirror neuron is encountered in the literature should trigger some unease. So what is next? After the initial enthusiasm for the discovery of mirror neurons, and the stimulating wave of speculation that followed, we now need to concentrate on developing methods that can: localize these neurons in the human brain; examine what information they convey about the actions of others; test for a causal relationship between putative mirror neurons in various nodes of the system and higher brain functions in humans; and try to understand the evolution of this system. Mirror neurons give us a fascinating glimpse into the neural basis of social cognition — let us use careful experimentation instead of wild speculations and controversies to transform this glimpse into solid scientific understanding.