1. The contributions of monosynaptic and polysynaptic circuitry to the tail-withdrawal reflex in the marine mollusk Aplysia californica were assessed by the use of physiologically based neural network models. Effects of monosynaptic circuitry were examined by the use of a two-layer network model with four sensory neurons in the input layer and one motor neuron in the output layer. Results of these simulations indicated that the monosynaptic circuit could not account fully for long-duration responses of tail motor neurons elicited by tail stimulation. 2. A three-layer network model was constructed by interposing a layer of two excitatory interneurons between the input and output layers of the two-layer network model. These interneurons had properties mimicking those of the recently described interneuron LP117, receiving excitatory input from pleural sensory neurons and evoking a biphasic excitatory postsynaptic potential (EPSP) in pedal motor neurons (Cleary and Byrne 1993). The three-layer model could account for long-duration responses in motor neurons. 3. Sensory neurons are a known site of plasticity in Aplysia. Synaptic plasticity was incorporated into the three-layer model by altering the magnitudes of conductance changes evoked in motor neurons and interneurons by presynaptic sensory neurons. In these simulations the excitatory interneurons converted an amplitude-coded input into an amplitude- and duration-coded output, allowing the three-layer network to support a large range of output amplitudes and durations. 4. Synaptic plasticity at more than one locus modified dramatically the input-output relationship of the three-layer network model. This feature gave the model redundancy in its plastic properties and points to the possibility of distributed memory in the circuitry mediating withdrawal reflexes in Aplysia. Multiple sites of control over the response of the network would likely allow a more diverse repertoire of responses.