Model for olfactory discrimination and learning in Limax procerebrum incorporating oscillatory dynamics and wave propagation.

We extend our model of the procerebral (PC) lobe of Limax, which is comprised of a layer of coupled oscillators and a layer of memory neurons, each layer 4 rows by 20 columns, corresponding to the cell body layer (burster cells) and neuropil layer (nonburster cells) of the PC lobe. A gradient of connections in the layer of model burster cells induces periodic wave propagation, as measured in the PC lobe. We study odor representations in the biological PC lobe using the technique of Kimura and coworkers. Lucifer yellow injection into intact Limax after appetitive or aversive odor learning results in a band or patch of labeled cells in the PC lobe with the band long axis normal to the axis of wave propagation. Learning two odors yields two parallel bands of labeled PC cells. We introduce olfactory input to our model PC lobe such that each odor maximally activates a unique row of four cells which produces a short-term memory trace of odor stimulation. A winner-take-all synaptic competition enabled by collapse of the phase gradient during odor presentation produces a single short-term memory band for each odor. The short-term memory is converted to long-term memory if odor stimulation is followed by activation of an input pathway for the unconditioned stimulus (US) which presumably results in release of one or more neuromodulatory amines or peptides in the PC lobe.