1. The guinea-pig trachea was isolated with its extrinsic innervation intact and pinned to the bottom of a water-jacketed dissecting dish filled with warmed, oxygenated Krebs solution. The trachea was not separated from the oesophagus. Isometric tension was measured in a segment of the rostral portion of the trachea. 2. Stimulation of the vagus nerves caudal to the nodose ganglia elicited contractions of the trachealis that were blocked by the muscarinic receptor antagonist atropine. Following addition of atropine and contraction of the trachealis with prostaglandin F2 alpha (PGF2 alpha), vagus nerve stimulation elicited non-adrenergic, non-cholinergic relaxations. Both responses elicited by stimulation of the vagi were abolished by cutting the recurrent laryngeal nerves and were considered parasympathetic in nature as they were sensitive to the autonomic ganglion blockers trimetaphan and hexamethonium. 3. Experiments were designed in which ganglionic blockers were added to the buffer bathing the entire preparation or, alternatively, added only to the buffer perfusing the tracheal lumen. When given equal access to the trachea and oesophagus, hexamethonium was 56-fold more potent an inhibitor of vagally mediated relaxations of the trachealis than vagally mediated contractions. Selective administration of hexamethonium to the buffer perfusing the tracheal lumen did not decrease the potency of the ganglionic blocker versus vagally mediated contractions. By contrast, even at a concentration of 1 mM, intratracheally administered hexamethonium failed to inhibit vagally mediated relaxations by 50%. Comparable results were obtained using trimetaphan. 4. Consistent with previous observations, removing the portion of the oesophagus contiguous with the region of the trachea at which isometric tension was measured abolished parasympathetic relaxations of the trachealis. Oesophagus removal was without effect on parasympathetic nerve-induced contractions. Removing the dorsal half of the oesophagus or the mucosa and submucosa of the oesophagus did not affect the parasympathetic relaxant innervation. 5. The compound action potential of guinea-pig recurrent laryngeal nerves evoked by vagus nerve stimulation consisted of three distinct peaks representing populations of axons with fast, intermediate and slow conduction velocities. The voltage-response characteristics of vagally mediated contractions were identical to those of the compound action potential peak representing fibres with intermediate (10 m/s) conduction velocities. By contrast, the voltage-response characteristics of the vagally mediated relaxations were best correlated with the compound action potential peak representing fibres with slow (0.4-3 m/s) conduction velocities.(ABSTRACT TRUNCATED AT 400 WORDS)