An automated most-probable-number (MPN) system for the enumeration of total bacterial flora and Escherichia coli was compared with plate count agar and tryptone-bile-glucuronide (TBX) and ColiID (in-house method) agar methodology. The MPN partitioning of sample aliquots was done automatically on a disposable card containing 48 wells of 3 different volumes, i.e., 16 replicates per volume. Bacterial growth was detected by the formation of fluorescent 4-methylumbilliferone. After incubation, the number of fluorescent wells was read with a separate device, and the MPN was calculated automatically. A total of 180 naturally contaminated samples were tested (pig and cattle carcass surfaces, n = 63; frozen minced meat, n = 62; and refrigerated minced meat, n = 55). Plate count agar results and MPN were highly correlated (r = 0.99), with log MPN = -0.25 + 1.05 x log CFU (plate count agar) (n = 163; range, 2.2 to 7.5 log CFU/g or cm2). Only a few discrepancies were recorded. In two samples (1.1%), the differences were > or = 1.0 log; in three samples (1.7%), the differences were > or = 0.5 log. For E. coli, regression analysis was done for all three methods for 80 minced meat samples, which were above the limit of detection (1.0 log CFU/g): log MPN = 0.18 + 0.98 x log CFU (TBX), r = 0.96, and log MPN = -0.02 + 0.99 x log CFU (ColiID), r = 0.99 (range, 1.0 to 4.2 log CFU/g). Four discrepant results were recorded, with differences of > 0.5 but < 1.0 log unit. These results suggest that the automated MPN method described is a suitable and labor-saving alternative to colony count techniques for total bacterial flora and E. coli determination in minced meat or on carcass surfaces.