Locomotory state in C. elegans (forward vs. reverse) is thought to be controlled by a network of forward and reverse command neurons in two reciprocally connected pools, but how this network is regulated by sensory input is poorly understood. According to one hypothesis, known as the Stochastic Switch Model, each pool acts as a probabilistic binary unit (OFF = 0, ON = 1). The model assumes that the rate constant for transitions to the ON state (a₀₁) at any point in time t is a sigmoidal function S(x) of net synaptic input I such that a₀₁(t) = a_max S(I(t)) where a_max is a scale factor. According to the Stochastic Switch Model, sensory input to each unit acts to bias the probability of the ON state. In one configuration of the model, modulation of bias is reciprocal. For example, a sensory stimulus that promotes forward locomotion would excite the forward unit and inhibit the reverse unit; a stimulus that promotes reverse locomotion would produce the opposite pattern of synaptic effects. Alternatively, the modulation may be non-reciprocal such that sensory input affects the bias of one unit but not the other. To distinguish between these two configurations in the context of chemotaxis, we presented worms with stepwise changes in the concentration of the chemoattractant NaCl and measured the rate constants for forward-to-reverse and reverse-to-forward transitions. Rate constants were analyzed in terms of a reduced version of the Stochastic Switch Model that enabled us to compute I(t) for each unit using the above equation. We found that in response to upward steps in NaCl, a forward-promoting stimulus, the forward unit was excited whereas the reverse unit was inhibited. This result indicates that upsteps promote forward locomotion by a reciprocal control mechanism. In contrast, we found that in response to downward steps in NaCl, a reverse-promoting stimulus, the forward unit was unaffected whereas the reverse unit was excited. This result indicates that downsteps promote reverse locomotion by a non-reciprocal mechanism. Taken together, these results suggest the hypothesis that locomotory bias in response to chemosensory inputs is established by different control mechanisms depending on stimulus polarity. We plan to test this hypothesis by recording synaptic potentials from command neurons in response to direct activation of chemosensory neurons. Support: NIH MH051383 and NSF IOB-0543643.