Oocyst prevalence and intensity have been recorded in 349 laboratory infections of Anopheles stephensi with Plasmodium berghei. Intensity and prevalence of infection are shown to be predictably related. The structure and heterogeneity in the infections has been analysed with the objective of describing the biological mechanisms by which the observed negative binomial oocyst distributions are generated. The analysis has revealed that the most likely processes lie within the population dynamic events of malaria within the mosquito, namely gametogenesis, fertilization and mortality. The distribution is similar in all Plasmodium – mosquito combinations examined so far, whether they are of laboratory (P. gallinaceum in Aedes aegypti) or field (P. vivax in An. albimanus and P. falciparum in An. gambiae s.l. and An. funestus) origin. Further we conclude that there is competition between parasites in the vector. Oocyst frequency distribution analysis shows that under natural conditions of transmission intensity, and even under the best laboratory conditions, significant numbers (> 10%) of fully susceptible mosquitoes will not be infected under conditions where the mean infection is as high as 250 oocysts. Failure to infect is not therefore an absolute indicator of refractoriness. In assessing transmission data it is shown that sample sizes should not be less than 50, and ideally 100 mosquitoes, if reliable data are to be obtained. In the field it is suggested that difficulties in determining the low natural intensity of oocyst infections indicate that prevalence estimates are a useful and accessible parameter to measure. In determining the impact of transmission blocking mechanisms we predict that under conditions where high oocyst intensities prevail, large reductions in intensity will be required before a reduction in prevalence can be expected i.e. here it will be necessary to measure intensity of infection. Conversely, under conditions where low oocyst intensities prevail, a rapid reduction in prevalence will occur with little concurrent reduction in intensity i.e. prevalence determination will be the more sensitive estimate.