Simulation during observation of human actions--theories, empirical studies, applications.

Historically, data from brain imaging and brain stimulation studies have supported the idea that the processing of observed actions recruits - among other areas - a distinct sub-set of brain sites in the sensory and motor cortices. These empirical findings have initially been linked with the thesis of direct matching as a mechanism of action understanding, i.e., the idea of motor resonance implemented by mirror neurons. In more recent approaches, it has been proposed that the mirror neuron system plays a role in minimizing prediction error when inferring the most likely cause of an observed action. According to these theories, motor resonance is thought to function as predictive coding. Other theoretical accounts suggest that action understanding might result from a hypothesis testing mechanism in which potential goals are continually fed into the system until the correct one is identified. In this review, we will explore the relationship of these theories to specific empirical findings. Finally, we will discuss the implications of these theoretical structures on action observation-based approaches to the optimization of skilled performance in athletes and patients.