Knowing the serotype of a salmonella does not help in the treatment or prognosis of a case, nor is infection with different serotypes distinguishable clinically: all serotypes can cause serious disease, including septicaemia.1 Why then is it important for almost all 30 000 or so salmonellas reported annually to Britain's Public Health Laboratory Service (PHLS) to undergo detailed identification, and for its Laboratory of Enteric Pathogens routinely to subdivide common serotypes into phage types that are irrelevant to patient management? One reason is to allow epidemiologists to use surveillance to identify linked cases. They can then trace sources and vehicles of infection, and curtail outbreaks. How has the case control method, as employed by Killelea et al and Shohat et al in this week's BMJ,2 3 helped achieve this? Their papers (pp 1105, 1107) describe the latest in a series of successful investigations of national and international outbreaks of salmonellosis4 5 6 7 8 which have two striking similarities: in all cases, the salmonella and the food …