| Firstly, the Oscar acceptance speech: I would like to thank my research director Dr. Christian Huyck for the help and inspiration that he gave me in carrying out this work. |
Cell assemblies are a type of neural network which are just beginning to be investigated, and one in which I am particularly interested. This section of my web site outlines the research that I have carried out in this particular field.
The idea of cell assemblies was first put forward by Canadian psychologist Donald Oldings Hebb in his book Organization of Behavior in 1949. He proposed two radically new theories about how the brain worked.
His first idea was that the strengths of the connections (the synapses) between two neurons could alter with time. If the neuron providing the signal (the pre-synaptic neuron) and the neuron receiving the signal (the post-synaptic neuron) were both active, then the strength of the synapse would increase. This became known as Hebbian learning, and a good analogy is a muscle strengthening with exercise. The idea was expanded later to include the strength of the synapse weakening if one neuron was active and the other was not.
The second idea was that the connections between large groups of neurons would alter themselves thanks to Hebbian learning in such a way that loops would form, i.e. if neuron A fired and activated neuron B, which fired and activated neuron C, which in turn fired and activated neuron A again, then the synapses linking these neurons would all strengthen and increase the possibility of the whole thing happening again in the future. If neuron A is activated in the future, then it would be more likely to activate neuron B, and hence neuron C and back to A again. This is a form of positive feedback that leads to strengthening of the connections and a strong loop being formed.
In practice, these loops, which Hebb termed cell assemblies, would consist of many thousands of cells, but the connections would develop in such a way that activating any part of the loop would cause the rest of it to activate and the whole process would be self-sustaining. Even if the stimulus that caused the loop to activate in the first place were removed, then the assembly would remain active for quite a while longer. Hebb likened this to the ringing of a bell after the hammer that struck it had been removed.
The concept of cell assemblies follows logically from the principle of Hebbian learning, but in 1949 (and for quite a while longer) there was precious little evidence for it. Nevertheless, the theory has survived, and it is only since the discovery of neural analysis tools such as Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) that evidence is coming in for it. Nowadays, the idea of cell assemblies and Hebbian learning in the brain is rapidly becoming orthodoxy.
So where do I come in to this? Well, I have been investigating whether these assemblies can be simulated in a computer program, and if so, to what extent. I'm not the first to do it (as I mentioned earlier, Rochester et al. first did a simulation in 1956), but the field of cell assembly simulation is still a fairly new one, and still presents a wide open frontier.
If any of this sounds interesting to you, please click on one of these links:
If you have carried out any work in this field, or you can think of any academic papers that I have missed, please do let me know. You can E-mail me at bowles64@hotmail.com.
Back to main menu