Major events that led to acceptance of 30 to 300 as the most suitable number of colonies on plates for counting were reviewed. Three new sets of data were collected, involving triplicate plates of fifteen 1: 1.4 serial dilutions of 65 samples of raw milk. Statistical methods were developed to analyze bias (variability introduced primarily by crowding and analyst counting errors) and variance (sampling and dilution errors). Bias and variance were combined as mean-squared error, which was expresed as a function of the number of colonies per plate, The counting range that minimized the mean squared error could then be determined for selected dilution series. For two-fold, five-fold and ten-fold dilution series, respectively, the most suitable limits on plates for counting were 70 to 140, 40 to 200 and 25 to 250 colonies/plate. A range of 25 to 250 colonies/plate was suggested for the analysis of dairy products. Limitations in application of the data to other systems are discussed.