A critical review is presented of the many and varied procedures for sampling adult mosquitoes. All sampling methods are subject to bias, but human-bait collections are probably the most reliable single method for detecting and monitoring populations of anthropophilic species. The estimation of realistic man-biting rates, however, remains difficult. The recent advances in computer technology have generated considerable interest in population modelling, with the aim of getting a better insight into the complexities of population dynamics, mosquito control and disease transmission. Although such approaches may be informative, it is stressed that models should be based on meaningful values of the critical parameters and that these are best obtained from field populations. The difficulties of interpreting mosquito collections in both attractant traps, such as animal-baited traps, carbon-dioxide-baited traps and light-traps, and in non-attractant ones, such as Malaise traps, sticky traps, ramp traps, rotary traps, suction traps and vehicle-mounted traps, are described. Suction traps probably provide the least biased catches of aerial populations of mosquitoes, but, if populations as a whole are to be considered, then resting adults also must be adequately sampled, and this may prove difficult. It is emphasised that different trapping techniques usually sample different components of a population, and that the choice of sampling methods depends much on the type and quality of information required. The limitations of using mark-recapture techniques to study adult dispersal and obtain absolute population estimates are discussed, together with the advantages and disadvantages of applying simple and more complex mathematical procedures for analysing recapture data.