Polymerase chain reaction (PCR)-based genotyping of oocysts dissected from mosquito midguts has previously been used to investigate overall levels of inbreeding within malaria parasite populations. We present a re-analysis of the population structure of Plasmodium falciparum malaria using diploid genotypes at three antigen-encoding loci in 118 oocysts dissected from 34 mosquitoes. We use these data to ask whether mating is occurring at random within the mosquito midgut, as is generally assumed. We observe a highly significant deficit of heterozygous oocysts within mosquitoes at all three loci, suggesting that fusion of gametes occurs non-randomly in the mosquito gut. A variety of biological explanations, such as interrupted feeding of mosquitoes, positive assortative mating and outcrossing depression, could account for this observation. However, an alternative artefactual explanation--the presence of non-amplifying or null alleles--can account for the observed data equally well, without the need to invoke non-random mating. To evaluate this explanation further, we estimate the frequencies of null alleles within the oocyst population using maximum likelihood, by making the assumption that non-amplifying oocysts at any of the three loci are homozygous for null alleles. Observed levels of visible heterozygotes fit closely with those expected under random mating when non-amplifying oocysts are accounted for. Other lines of evidence also support the artefactual explanation. Overall inbreeding coefficients have been recalculated in the light of this analysis, and may be considerably lower than those estimated previously. In conclusion, we suggest that the deficit of heterozygotes observed is unlikely to indicate non-random mating within the mosquito gut and is better explained by misscoring of heterozygotes as homozygotes.