New forms of catalytic epidemic models were developed to estimate the incidence of primary toxoplasmosis infection from age- and time-specific seroprevalence data collected from persons aged 0-100 years in South Yorkshire, England, 1969-1990. Piecewise constant and exponential polynomial functions were used to assess the way in which incidence depended on age and time, and to guide the choice of parametric models suitable for prediction. Incidence estimates were biased unless both age- and time-dependence were allowed for. New findings on the epidemiology of this infection emerged. Incidence appears to have fallen sixfold between 1915 and 1970, but has remained stable for the last 20 years. There is a marked peak in incidence in childhood. The incidence throughout the childbearing period is currently estimated to be 0.07 or less per 100 susceptible persons per year. However, these predictions were highly sensitive to assumptions about incidence in childhood, and the 95% confidence limits for a range of models were between 0.003 and 0.32% per year. Age- and time-specific seroprevalence data can be collected inexpensively on a mass population basis, and, with appropriate incidence modeling, may prove to be a powerful method for the study of infectious disease and for incidence prediction.