Friday, July 22, 2005

The case of the missing gap junctions

Christof Koch and Stuart Hameroff have an ongoing debate about the role of gap junctions in gamma synchrony (think brainbeat of consciousness, ~40 Hz). Hameroff is sure it's based in the dendritic feltwork of outer layers of neocortex, in so-called hyper-neuronal gap junctions. Koch isn't so sure.

In a recent response to Hameroff on SCR Koch cited some highly germane research findings. I didn't mention these in my article, but they strongly support my hypothesis. Koch:

Gap junctions and consciousness
flurry of studies over the last handful years have shown that electrical gap junctions play a much larger role in cortex than previously realized (for a review, see Bennett and Zuki, 2004). The focus has been on gap junctions mediated by the Connexin family of proteins (which has at least 10 members expressed in the mammalian central nervous system); they provide a low-resistance, electrical pathway between two neurons –– hence their common alias, electrical synapses. The most important connexin of the adult brain is Cx36. These proteins link groups of local, inhibitory interneurons into large networks (forming a “hyperneuron”), restricted to specific layers in neocortex and hippocampus. Different from their conventional chemical synapses that would serve to inhibit the firing activity of their postsynaptic targets, gap junctions can cause the membrane depolarization in one interneuron to spread to others. This has given rise to the hypothesis that Cx36 gap junctions synchronize the firing of these interneurons, enabling the entire population to fire in lock step in the 30-70 Hz (gamma) range. If such synchronized and rhythmic firing is important for certain aspects of attention, perception and consciousness, then mice that lack gap junctions should show major deficits.

It has been possible to breed mice that lack the gene for the Cx35 protein. In these knockout mice, interneuron coupling is greatly reduced; this goes hand-in-hand with a loss of spike synchrony among them (however, gamma range oscillation persist, albeit at a reduced amplitude). The animals have decreased retinal function at low light levels (as expected from a loss of gap junctions in the retina) and reduced reproductive rates, but otherwise no major behavioral deficiencies; they display no obvious loss of motor coordination, and can stay on a rotating cylinder (rotorod test) as well as normal, wildtype mice.

The idea that the firing activity of groups of interneurons, possibly extending over several cortical columns, is tightly synchronized and may underlie the coalitions of neurons that are sufficient for any one conscious percept is a fascinating one to me. However, no positive evidence links neuron-to-neuron coupling in cortex via gap junctions to perception, let alone consciousness. Thus, the cautionary statement in my book, “ Not enough is known about this phenomenon to implicate it in conscious perception.” remains true today. This may always change in the future, of course.

I believe the time for change has come, in the form of Dale Antanitus' Theory of Neocortical Neuron-Astrocyte Interaction. Antanitus hypothesizes that astrocytic syncytia, groups of astrocytes connected via gap junctions, are responsible for
synchronous neuronal firing. This would account for why, when the mediator of hyper-neuronal gap junctions was eliminated, gamma synchrony persisted, albeit at diminished amplitude. It suggests both neuronal and astrocytic syncytia are involved in gamma synchrony. Such redundancy is of course ubiquitous in nature and should come as no surprise; critical systems are often backed up.

There's been a lot of research into astrocytes in the past five years (e.g., 87 titles from The Journal of Neuroscience alone), but no one has yet produced a grand integrative theory as to their relationship to their neuron brothers, to my knowledge, except Antanitus. I hypothesize that his ideas are the neurophysiological basis for not only gamma synchrony but Hawkins' regions of cortex, the fundamental units of the Memory Prediction Framework, as well.

In time, after The Memory Prediction Theater is published at SCR, I hope to discuss the implications of Antanitus' theory, in the context of recent research findings, with Koch, Hameroff and others in the consciousness community.

Thursday, July 21, 2005


Just learned that my article introducing The Memory Prediction Theater is going to be published by Science and Consciousness Review. I feel very grateful. I'm not a scientist. I'm a software developer with an interest in AI. I've always been drawn to the subject of consciousness, but never seriously approached it until I tried to imagine how to create a piece of software that could pass the Turing Test. I needed a model of human cognition. What was the best model science had produced? That became my driving question.

The question soon led me to Bernard Baars and Global Workspace. I read Baars for a couple weeks, becoming increasingly consumed, especially by the theater model he used to elaborate Global Workspace. I was just starting to really dig into the work of Stan Franklin, Baars' computer science colleague at the University of Memphis, who's developed a "conscious" software agent based on Global Workspace, when an epiphany occurred.

I happened to see something on Steve Jurvetson's blog that caught my eye. He said he'd been reading Jeff Hawkins' new book. If Steve was reading it, I thought I'd better check it out. I googled, one-clicked and was immersed within two days. I devoured the book. Easy to read. Absolutely brilliant. Totally original. The basic problem of consciousness, what Hawkins calls intelligence, solved.

Stunned, I immediately began to wonder how this Memory Prediction Framework of Hawkins' might play out in Baars' Theater of Consciousness. That's how the MPT came into being. That was November of last year. I realized today, on hearing that SCR was going to publish the hypothesis, I might be getting some questions on it, so decided to launch this blog.