The patch electrode technique was used to record single channel current pulses in tissue-cultured mouse spinal cord neurons. In agreement with earlier noise studies, channels activated by gamma-aminobutyric acid (GABA), muscimol, and (-)-pentobarbital were found to have equal unit conductances. The kinetics of channel closing were studied by analyzing the distributions of open state lifetimes. Channels activated by (-)- pentobarbital and muscimol had longer mean open times than channels activated by GABA. As a result, the kinetics of (-)-pentobarbital- and muscimol-activated channels could be studied in greater detail. Most observed open state lifetime distributions were not exponential but contained an excess of short duration events. A sum of two exponential functions gave a much better fit than a single exponential function to most observed open state lifetime distributions. A critical comparison of noise analysis with single channel recording shows that the fast process responsible for the rapid closures would be very difficult to detect in a noise experiment. The channel noise is dominated by the slower process, and as a result, the relaxation time of the slower kinetic component derived from single channel studies is close to the mean open state lifetime derived from noise measurements. The observation of a faster process points toward either an additional population of channels or a scheme for the channel closing transition which is not a simple first order process.