The Logic of Limax Learning

We wish to understand the neuronal computations performed on sensory inputs that result in the categorization of those inputs, their storage as memory states, and their associative combination. Dur experimental and theoretical work is focused on the neuronal computations performed on odor and taste inputs in the CNS of Limax maximus, a terrestrial mollusk convenient for behavioral, neuro­ physiologieal, and neurochemical experiments (Gelperin, 1983). The questions posed in this specific system are designed to illuminate issues of learning and mem­ ory storage with panphyletic generality. Several aspects of the learning behavior of Limax have encouraged us to attempt a unique synthesis of the approaches of behavioral biology, neurobiology, and neural modeling. While Limax displays many of the learning phenomena of higher organisms, including primates, it accomplishes these learning tasks using onlyabout 10,000 cells in its CNS. We are using behavioral experiments to define the major types of learning exhibited by Limax and for each type of learning to establish the critical interevent timing relations that allow learning to occur. Neu­ rophysiological and neurochemical experiments delimit the areas of CNS necessary for learning and describe the types of cellular elements and synaptic interactions available for modification during learning. The neural modeling asks how a collec­ tion of relatively simple neurons might interact synaptically to collectively accom­ plish the learning tasks. The model avoids assumptions about precise anatomical details, instead asking questions about the computational consequences of a simple set of rules governing synaptic interactions.