The alkaline phosphatase test is used as an indicator of adequate pasteurisation of milk and cream. A proprietary fluorimetric technique (Fluorophos) is a sensitive and quantitative method for the determination of alkaline phosphatase (ALP) activity in milk products. Currently, adequate pasteurisation of milk products is regarded as confirmed in samples that contain a residual bovine ALP activity of < or =500 mU/litre. This is equivalent to the statutory acceptable level of 4ug phenol/ml required by the EC analytical method. The purpose of the present study was to assess the effectiveness of pasteurisation of milk and cream produced by on-farm dairies. In a longitudinal study over a four-year period, 4,999 samples of milk and cream were collected from 130 on-farm dairies and from two large commercial dairies in NW England for comparison. Bovine ALP activity of >500 mU/litre was deemed as a failure and was found in 3.5% of whole milk, 2.4% semiskimmed milk, 5.0% of skimmed milk, and 39% of cream samples from on-farm dairies. Bovine ALP activity of >100 and <500 mU/litre was found in 18.4% of whole milk, 9.3% of semi-skimmed milk, 13.2% skimmed milk and 44.5% of cream samples from on-farm dairies. Results with skimmed milk samples showed significantly lower bovine ALP activity than whole milk. All 409 milk and cream samples from two large commercial dairies passed the fluorimetric test at less than 500 mU/litre of bovine ALP, and 99% of these milk and cream samples had bovine ALP activity of less than 100 mU/litre. The presence of residual bovine phosphatase indicates a failure and may be due to either inadequate pasteurisation or post pasteurisation contamination with raw milk. Residual bovine phosphatase was demonstrated in 108/114 (94.7%) of milk samples with a bovine ALP activity greater than 500 mU/litre, i.e. true failures. Of more concern is that residual bovine phosphatase was found in 395/401 (98.5%) of samples that gave bovine ALP activity greater than 100 mU/litre but equal to or less than 500 mU/litre. Residual bovine phosphatase was demonstrated in 37/108 (30.2%) of cream samples with bovine ALP activity greater than 500 mU/litre. Presence of reactivated bovine phosphatase is not an indication of a failure but can mask the presence of residual bovine phosphatase. Reactivated bovine phosphatase was found in 74/106 (69.8%) of cream samples. Our results confirm that the more sensitive fluorimetric method is suitable for testing pasteurised whole milk and semiskimmed milk, but for statutory purposes the acceptable level of residual bovine phosphatase should be <100 mU/litre. Our findings have highlighted a potential problem when testing skimmed milk and cream samples from on-farm dairies. To ensure public safety we need more stringent standards for the ALP test and new methods that will accurately confirm that pasteurisation of these products has been achieved.