Distributed Processing vs. Dedicated Neurons in the Production of Simple Behavioral Acts

Abstract

When faced with complex devices like most nervous systems, we invent schemes to make them easier to understand. Hierarchies, modules, feature detectors, command cells — these are all pious simplifications that we create to try to make sense of how a nervous system might function. We admit that they are simplifications, and assure ourselves that we will keep track of the complications as they arise. However, when we face data that doesn’t fit into these simple schemata, we are remarkably resistant to abandoning these comfortable mental constructs. In my own case, I started to look for the interneurons in the medicinal leech that produced a very simple reflexive response, a localized bending response to a mechanical stimulus. I expected to find interneurons that were activated during dorsal bending, others during lateral bending, and still others during ventral bending. None of the interneurons that I found were so finely tuned, so I abandoned the search. When Shawn Lockery picked up the search several years later, he found the same kind of interneurons as I had, but came to a different conclusion: that the behavior was caused by a neuronal organization that is quite different from what we expected; i.e., it was produced by a distributed network. The implications of this kind of network are not easy to grasp, because the outcome is the sum, often a nonlinear sum, of many competing interneuronal influences.