Igor Farkaš, Michal Malý, Kristína Rebrová:  Mirror neurons – theoretical and computational issues

Abstract:  The discovery of mirror neurons, first in macaques and most recently in humans, has provided an interesting research agenda on the possible roles of mirror neurons in various cognitive functions. Compared to that of monkeys, the functionality of the human mirror system seems more general and more abstract. However, the most recent views on mirror neurons point to disagreements regarding their development and functions in human cognition. In this review, we start with some of the theoretical points of disagreement and propose their reconciliation. Namely, we argue for a position between the "adaptation" hypothesis (emphasis on inborn biases) and the "association" hypothesis (emphasis on learning) of the origin of mirror neurons. Then we propose a graded action understanding hypothesis, to be placed between the motor theory and the perceptual theory of action understanding. We explain justification of our hypothesis in the extended context of actions whose "deep understanding" requires special training. In addition, we argue that the role of STS, subserving action recognition, has not been fully appreciated in the mirror neuron theory. In the second part, we critically review conceptual and computational models of the mirror neuron system that range from the simplest, Hebbian accounts, to more complex accounts, according to the interaction between sensory and motor representations. We propose that a plausible computational model of action understanding should account for acquisition of a mirroring property, and include mechanisms for yielding high-level invariant perceptual representations needed for a neural account of action recognition.