1. Membrane currents of hair cells in acutely excised or cultured mouse utricles were recorded with the whole cell voltage-clamp method at temperatures between 23 and 36 degrees C. 2. Type I and II hair cells both had delayed rectifier conductances that activated positive to -55 mV. 3. Type I, but not type II, hair cells had an additional delayed rectifier conductance (gK,L) with an activation range that was unusually negative and variable. At 23-25 degrees C, V(1/2) values ranged from -88 to -62 mV in 57 cells. 4. gK,L was very large. At 23-25 degrees C, the average maximum chord conductance was 75 +/- 65 nS (mean +/- SD, n = 57; measured at -54 mV), or approximately 21 nS/pF of cell capacitance. 5. gK,L was highly selective for K+ over Na+ (permeability ratio PNa+/PK+:0.006), but unlike other delayed rectifiers, gK,L was significantly permeable to Cs+ (PCs+/PK+:0.31). gK,L was independent of extracellular Ca2+. 6. At -64 mV, Ba2+ and 4-aminopyridine blocked gK,L with apparent dissociation constants of 2.0 mM and 43 microM, respectively. Extracellular Cs+ (5 mM) blocked gK,L by 50% at -124 mV. Apamin (100 nM) and dendrotoxin (10 nM) has no effect. 7. The kinetic data of gK,L are consistent with a sequential gating model with at least two closed states and one open state. The slow activation kinetics (principal time constants at 23-25 degrees C:600-200 ms) had a thermal Q10 of 2.1. Inactivation (Q10:2.7) was partial at all temperatures. Deactivation followed a double-exponential time course and had a Q10 of 2.0. 8. At 23-25 degrees C, gK,L was appreciably activated at the mean resting potential of type I hair cells (-77 +/- 3.1 mV, n = 62), so that input conductances were often more than an order of magnitude larger than those of type II cells. If these conditions hold in vivo, type I cells would produce unusually small receptor potentials. Warming the cells to 36 degrees C produced parallel shifts in gK,L's activation range (0.8 +/- 0.3 mV/degrees C, n = 8), and in the resting potential (0.6 +/- 0.3 mV/degrees C, n = 4). Thus the high input conductances were not an artifact of unphysiological temperatures but remained high near body temperature. It remains possible that in vivo gK,L's activation range is less negative and input conductances are lower; the large variance in the voltage range of activation suggests that it may be subject to modulation.