Factors were examined which may influence bacterial psychrotroph counts obtained by the European Community's directed preincubation test for pasteurized milk. Studies to determine the best protocol for sampling milk indicated that the least counting error was obtained by taking at least five consecutive samples of milk. Increasing the number of replicate plates or making subsamples did not decrease counting error. When milk samples from five commercial dairies were examined over an 11 month period there was a significant difference (p < 0·001) in counts obtained over the period, but this did not seem to be a seasonal effect. The counts between dairies were not significantly different nor was the position in the pasteurization run from which the samples were taken significant. The proportion of samples which failed the test standard was 9.7% over the period. The bacterial flora isolated from plate count agar were identified; Gramnegative rods were the most common isolates. The spiral plater system was compared with the pour plate method in the preincubation test and results were found to be in close agreement, suggesting that the spiral plater could be used to improve the current official methods. Varying the time and temperature of sample preincubation was found to give significantly different results in plate counts. This suggested that preincubation conditons were critical and that variation in conditions even within the specified ranges of the test could significantly alter the bacterial psychrotroph counts obtained. Because of these results, the validity of the statutory test was questioned.