The measurement and analysis of common recurrent conditions such as diarrhoea, respiratory infections or fever pose methodological challenges with regard to case definition, disease surveillance and statistical analysis. In this paper we describe a flexible and robust model that can generate simulated longitudinal datasets for a range of recurrent infections, reflecting the stochastic processes that underpin the data collected in the field. It can be used to evaluate and compare alternative disease definitions, surveillance strategies and statistical methods under 'controlled conditions'. Parameters in the model include: characterizing the distributions of the individual disease incidence and the duration of disease episodes; allowing the average disease duration to depend on an individual's number of episodes (simulating a correlation between incidence and duration); making the individual risk of disease depend on the occurrence of previous episodes (simulating autocorrelation of successive episodes); finally, incorporating seasonal variation of disease.