1. To determine reliability of stimulus intensity coding we measured the response variability of 30 hydroxy-L-proline sensitive chemoreceptor cells from antennules of Homarus americanus. First, cells were tested with 5 repeated presentations of 10-5M and l0-4M hydroxy-L-proline. Then a stimulus response function was determined in each cell from l0-6to 5 × l0-4M in half log steps. 2. Population mean responses to repetitive stimulation showed little variability and mean stimulus response functions increased monotonically. In contrast, most individual cells showed a high response variability and no single cell stimulus-response function increased monotonically. The greatest response variability occurred between different cells. 3. Mean response magnitude of a 30 cell population allowed reliable discrimination between 10-5M and l0-4M hydroxy-L-proline for each of the 5 repetitions. It also allowed discrimination of all concentrations of hydroxy-L-proline tested. Across-fiber response patterns also differed significantly with concentration while they remained relatively stable with repetition of the same stimulus. 4. We applied different analytical methods to test the reliability of response magnitude and across-fiber pattern as possible codes for intensity discrimination. We demonstrated that a random population of rather unreliable receptor cells could provide remarkably reliable intensity information to the CNS based on response magnitude or across-fiber pattern.