Many aspects of the natural history of tuberculosis are poorly understood. Though it is recognized that clinical tuberculosis may follow shortly after initial infection ('primary' disease), or many years thereafter through either endogenous reactivation or after reinfection, the relative importance of these mechanisms is often disputed. The issue is complicated by the fact that the risks of developing disease are age-dependent, and reflect infection risks which may change over time. This paper estimates the age-dependent risks of developing tuberculosis using an age-structured deterministic model of the dynamics of tuberculous infection and disease in England and Wales since 1900. The work extends the classical studies of Sutherland and colleagues. The best estimates of the risks of developing 'primary' disease (within 5 years of initial infection) were approximately 4%, 9% and 14% for individuals infected at ages 0-10, 15 years and over 20 years respectively, and a previous infection appeared to impart little protection against (further) reinfection, but 16-41% protection against disease subsequent to reinfection for adolescents and adults. We also provide evidence that reinfection made an important contribution to tuberculous morbidity in the past, as (i) exclusion of exogenous disease from the model considerably worsened the fit to observed notification rates, and (ii) the dramatic decline in the risk of tuberculous infection from 1950 in England and Wales accelerated the decline in morbidity among all individuals, even among the older age groups with a high prevalence of tuberculous infection. We conclude that the risk of infection is the single most important factor affecting the magnitude of the tuberculous morbidity in a population, as it determines both the age pattern of initial infection (and hence the risk of developing disease) and the risk of reinfection.