The paper describes a maximum-likelihood method for the estimation of age-related changes in the per capita rate of infection, from case notification records or serological data. The methods are applied to records of measles incidence in the UK and USA, for which the estimated rates of infection tend to rise to a maximum value at around 10 years of age and then to decline in the older age-classes. Longer-term and seasonal trends are analysed by reference to changes in the estimated average age at infection; a statistic derived from a knowledge of the age-specific rates of infection. Future data needs in the epidemiological study of directly transmitted viral and bacterial diseases are discussed with reference to the detection and interpretation of age-dependent rates of disease transmission.