I can recall an article I read a few months ago, where a couple of researchers took data from an instant messaging service and found that there was, on average, a ~6.6 degree separation between parties.
Love this. Determining how these "hidden metric spaces" arise in neural networks during the course of evolution could have great impact on the understanding of the brain. Dan Dennett talks of consciousness as a "meme machine" made up of lots of little unconscious "robots" or neurons. Perhaps this is not the entire story though. Perhaps the neurons have evolved to maximize efficiency in communication by means of some hidden metric space. I wonder if associations of "synchronicity" could come into play here. Say neuron A wants to get a signal to Location L following some Input I. Neuron A doesn't know Location L but knows that Neuron B always fires when something happens in L. So it sends the signal over to Neuron B, which, sure enough, passes it along to L, therefore, Neuron A is able to get the signal following a certain input over to L without knowing exactly the topological connections to L. Of course words like "know" and "want" are used metaphorically, obviously this arises dynamically via blind trial and error. So Dennett's robots take on a new layer of interest: perhaps by being able to form associations between temporally or spatially related events, they can construct signal patterns via a "hidden metric space" for better efficiency. Figuring out how this additional efficiency comes about by figuring out how these hidden connections evolve could make artificial neural nets more efficient. Right now, I think neural nets are largely "brute force" - add more neurons for better results. But perhaps hidden metric spaces could help us get better results from the same neuron number. Anyway. Way cool.