Epidemiological models are used to analyse 8 published data sets reporting age–prevalence curves for trypanosome infections of the tsetse fly Glossina pallidipes. A model assuming a fixed maturation period and a rate of infection which is independent of fly age is adequate for Trypanosoma vivax-type infections, explaining 98% of observed variance in prevalence by site and age, allowing that the rate of infection may be site dependent. This model is not adequate for T. congolense-type infections and the fit can be improved by allowing (i) the rates of infection to decline with age (although non-teneral flies remain susceptible), (ii) a fraction of resistant flies, which may vary between sites, (iii) increased mortality of infected flies and (iv) variation in the maturation period. Models with these features can explain up to 97% of observed variance. Parameter estimates from published experimental data suggest that all may contribute in practice but that (i) and/or (ii) are likely to be the most important.