The results of a laboratory assessment of the performance characteristics of a fish monitor, designed for the detection of toxic pollution events at drinking water intakes, are presented. The equipment assessed was developed by WRc in the U.K. and monitors significant changes in fish ventilation frequency as an indicator of the occurrence of a toxic pollutant. Rainbow trout ( Salmo gairdneri) were used in these trials being most commonly used in operation. Data from a trial period of 1 year were analysed to establish sensitivity to ten test substances and the estimated rate of false responses for the sensitivities quoted. Comparison with trout acute toxicity data (96 h LC 50) show responses at between 10 and 250% of the LC 50 with a response within 40 min. Comparison is also with 24 h SNARLs (suggested no adverse response levels) for humans where this is available. The monitor is capable of responding at concentrations below or close to these levels where the test substance could be expected to produce acute toxic symptoms, but the concentration of response may be considerably higher than the SNARL where toxic symptoms are of a chronic, long-term nature. The estimated rate of false responses for the monitor was between 1 in 3 and 1 in 5 months.