1. The process by which stretch of the external muscle of the intestine leads to excitation of myenteric neurons was investigated by intracellular recording from neurons in isolated longitudinal muscle-myenteric plexus preparations from the guinea-pig. 2. Intestinal muscle that was stretched by 40 % beyond its resting size in either the longitudinal or circular direction contracted irregularly. Both multipolar, Dogiel type II, neurons and uniaxonal neurons generated action potentials in stretched tissue. Action potentials persisted when the membrane potential was hyperpolarized by passing current through the recording electrode for 10 of 14 Dogiel type II neurons and 1 of 18 uniaxonal neurons, indicating that the action potentials originated in the processes of these neurons. For the remaining four Dogiel type II and 17 uniaxonal neurons, the action potentials were abolished, suggesting that they were the result of synaptic activation of the cell bodies. 3. Neurons did not fire action potentials when the muscle was paralysed by nicardipine (3 microM), even when the preparations were simultaneously stretched by 50 % beyond resting length in longitudinal and circular directions. Spontaneous action potentials were not recorded in unstretched (slack) tissue, but when the L-type calcium channel agonist (-)-Bay K 8644 (1 microM) was added, the muscle contracted and action potentials were observed in Dogiel type II neurons and uniaxonal neurons. 4. The proteolytic enzyme dispase (1 mg ml-1) added to preparations that were stretched 40 % beyond slack width caused the myenteric plexus to lift away from the muscle, but did not prevent muscle contraction. In the presence of dispase, the neurons ceased firing action potentials spontaneously, although action potentials could still be evoked by intracellular current pulses. After the action of dispase, (-)-Bay K 8644 (1 microM) contracted the muscle but did not cause neurons to fire action potentials. 5. Gadolinium ions (1 microM), which block some stretch activated ion channels, stopped muscle contraction and prevented action potential firing in tissue stretched by 40 %. However, when (-)-Bay K 8644 (1 microM) was added in the presence of gadolinium, the muscle again contracted and action potentials were recorded from myenteric neurons. 6. Stretching the tissue 40 % beyond its slack width caused action potential firing in preparations that had been extrinsically denervated and in which time had been allowed for the cut axons to degenerate. 7. The present results lead to the following hypotheses. The neural response to stretching depends on the opening of stretch activated channels in the muscle, muscle contraction in response to this opening, and mechanical communication from the contracting muscle to myenteric neurons. Distortion of sensitive sites in the processes of the neurons opens channels to initiate action potentials that are propagated to the soma, where they are recorded. Neurons are also excited indirectly by slow synaptic transmission from neurons that respond directly to distortion.