An audit of South Thames infectious disease surveillance systems in 1995 revealed large inconsistencies between the numbers of food poisoning records held on local databases and the numbers of food poisoning notifications reported to the Office for National Statistics (ONS), then called the Office of Population Censuses and Surveys. In March 1996 a questionnaire sent to each local authority in South Thames asked what action was usually taken when a laboratory report of cryptosporidium, campylobacter, salmonella, or giardia was received unsupported by a notification of food poisoning. All 51 local authorities responded to the questionnaire. Forty-eight reported salmonella to ONS, 38 reported cryptosporidium, 38 campylobacter, and 37 giardia. Some local authorities considered whether a food or water source was evident or suspected before reporting. Patterns of notification varied between geographical areas. Differences between local authorities' interpretations of the requirement to report to ONS the isolations of food poisoning organisms from patients make it difficult to analyse food poisoning statistics. We would recommend the adoption of a simple approach, in which laboratory reports and notifications are reconciled locally. A case should be reported to ONS only if the doctor who notified or arranged for an examination of stool suspected food or waterborne transmission initially.