Since this article was published, I have discovered two additional articles that are highly relevant but were not cited. I did cite Barlow and Foldiak (1989) for their proposal that anti-Hebbian plasticity could allow neurons to learn to make predictions. However, I did not cite Foldiak (1990), who demonstrated some of the virtues of combining Hebbian and anti-Hebbian synaptic plasticity within single neurons. My theory also combines Hebbian and anti-Hebbian synaptic plasticity within single neurons, and additionally extends the use of anti-Hebbian plasticity to selecting amongst a neuron’s non-synaptic ion channels.

I also failed to cite Berg et al (2007). They performed intracellular recordings from turtle motor neurons during rhythmic scratch-like behavior. They found that synaptic excitation and inhibition occurred in-phase with one another. Thus motor neurons showed rhythmic periods of high conductance during which the neurons were depolarized by synaptic excitation despite a simultaneous increase in synaptic inhibition. The coincidence of inhibition with excitation could be explained by anti-Hebbian selection of inhibitory inputs, which would cause inhibition to predict and partially cancel excitation. I had cited several papers that described a similarly “balanced” inhibition and excitation in sensory neurons (17-20), but the work of Berg et al (2007) extends this phenomenon to motor neurons. This supports my theory’s proposal that the rules that govern the selection of inputs are common to both sensory and motor neurons.

Berg RW, Alaburda A, Hounsgaard J (2007) Balanced inhibition and excitation drive spike activity in spinal half-centers. Science 315: 390-393.

Foldiak P (1990) Forming sparse representations by local anti-Hebbian learning. Biological Cybernetics 64: 165-170.