Infectious pancreatic necrosis is a disease that is causing increasing loses to Scottish Atlantic-salmon farms. I asked the question: is infection of individual salmon farms persistent or transient? Using Fisheries Research Services' Fish Health Inspectors' data, conditional probabilities that a farm would (time 0) be infected were estimated for farms that had been infected [P(I(0)|I(-T))] or free [P(I(0)|F(-T))] from infectious pancreatic necrosis virus (IPNV) when a sample was taken at some earlier time (-T). A logistic-regression model was used to estimate these conditional probabilities with T; this model was multilevel to account for regional and inter-annual level differences in prevalence of IPNV. In freshwater, conditional probabilities remained substantially different for periods of at least 4 years, so, although many farms did change infection status, IPNV either persisted or recurred at specific freshwater farms. Marine farms showed similar conditional probabilities after about 2 years following a positive or a negative sample, indicating that infection was transient. Management of larger areas and exchanges between farms might be more effective than farm-level management at controlling marine IPNV.