Aim: To determine the sensitivity of methods for detection of injured and uninjured Escherichia coli O157:H7 (E. coli O157) in raw and pasteurized milk. Methods and Results: Raw milk, pasteurized milk with 1.5% fat content and pasteurized milk with 3.5% fat content were spiked with E. coli O157 at low levels. The samples were enriched in modified tryptone soya broth with novobiocin (mTSBn) at 37degreesC. Aliquots of the enriched culture were analysed either by manual immunomagnetic separation (MIMS) and culturing on sorbitol MacConkey agar with or without cefixime and potassium tellurite (SMACct or SMAC), or by automated immunomagnetic separation and integrated ELISA (EiaFoss(TM)). Uninjured E. coli O157 organisms were detected in milk by both methods at 1 cfu 10 ml(-1) sample). Injured organisms were detected at levels of about 4 cfu 10 ml(-1) sample. Direct enrichment in mTSBn (22 h incubation) showed better sensitivity for injured cells than enrichment in buffered peptone water (BPW, 22 h incubation), or in a two-step enrichment consisting of BPW (6 h, 37degreesC) and mTSBn (16 h, 37degreesC), successively. Conclusions: The methods showed equal sensitivity in that they were both able to detect 1 cfu 10 ml(-1) milk sample. Injured organisms can be detected and isolated at a level almost as low as this. A resuscitation step is not recommended for the detection and isolation of injured and non-injured E. coli O157 from milk. Significance and Impact of the Study: Due to the dilution of contamination in the bulk tank, analysis of milk for the presence of E. coli O157 requires a very sensitive method. Both methods described here are useful for such analysis.