The identification of memory and learning in medically important mosquito species has been of epidemiological interest mainly because of the implications of learning on the pattern of contact between vectors and hosts. Empirical results either showing or suggesting the existence of cognitive abilities in mosquitoes have been reported in a number of experimental studies, mainly based on the observation of individual fidelity towards subsets of specific resources, such as hosts, resting sites or breeding sites. A closer inspection of the design of these experiments shows that, with the exception of recent studies providing stronger evidence of learning in the genus Culex (Diptera: Culicidae), methodological shortcomings still hinder the possibility of eliminating alternative interpretations for these findings, in some cases because the experiments were not specifically designed to identify the phenomenon, but mostly because of a lack of appropriate controls or replication. By highlighting these limitations, while acknowledging the practical difficulties that are inherent to the field, we aim to help expel from future research the 'ghosts' that still preclude the achievement of more definite conclusions about the prevalence of memory and learning in this group of insects.