1. The response to bitter-tasting substances was recorded in outside-out membrane patches excised from the taste receptor cell of the bullfrog fungiform papilla. 2. Application of a bitter-tasting substance, quinine or denatonium, induced channel openings under conditions in which none of the second messenger candidates or their precursors (e.g. cyclic nucleotide, inositol 1,4,5-trisphosphate, Ca2+, ATP and GTP) were present on either side of the membrane. The response could be recorded > 10 min after excision of the patch membrane. These data suggest that the channel was directly gated by the bitter-tasting substances. 3. No change in response was detected upon addition to the cytoplasmic side of either GDPbetaS (1 mM) or GTPgammaS (1 mM), suggesting that the G protein cascade has no direct relation to response generation. 4. The quinine-induced current was dose dependent. The lowest effective concentration was approximately 0.1 mM, and the saturating concentration was near 1 mM. The dose-response curve was fitted by the Hill equation with a K of 0.52 mM and a Hill coefficient of 3.8. 5. The single channel conductance measured in 120 mM NaCl solution was 10 pS. The channel was cation selective, and the ratio of the permeabilities for Na+, K+ and Cs+ (PNa : PK : PCs) was 1 : 0.48 : 0.39. The unitary conductance was dependent on the extracellular Ca2+ concentration ([Ca2+]o); 9.2 pS in a nominally Ca2+-free solution, and 4.5 pS in 1. 8 mM [Ca2+]o. 6. The dose dependence, the ion selectivity and the dependence of the unitary conductance on [Ca2+]o were almost identical to those of the quinine-induced whole-cell current reported previously, indicating that the channel activity observed in the excised membrane is the basis of the whole-cell current. 7. The present observations suggest the new possibility that the cationic channel directly gated by bitter substances is involved in the bitter taste transduction mechanism.