A mark-recapture experiment was carried out in northern Tanzania to determine whether Anopheles arabiensis exhibits memory, by investigating if bloodfed individuals would return to either the location or the host where or on which they had obtained a previous bloodmeal, behaviours termed site-fidelity and host-fidelity respectively. Over 4300 mosquitoes were collected from 2 houses, marked with different fluorescent colours according to whether they were caught in cattle sheds, 'cattle-fed', or within human bednets, 'human-fed', at either location, then released from a third location. Over 17,000 mosquitoes were collected and examined over the next 8 days. In total, 1% of released mosquitoes were recaptured. Of these, 68% had returned to the house where they were first caught, demonstrating site-fidelity (P = 0.007). However, 86% of recaptured mosquitoes were caught on cattle regardless of where they were initially caught (P = 0.185). Bloodmeal identification showed that a high proportion of mosquitoes classed as human-fed contained bovine blood, thereby confounding the investigation into host-fidelity. Notably, the proportion of mosquitoes with mixed bloodmeals depended on the proximity of cattle and humans, with significantly higher proportions of mixed bloodmeals occurring when cattle and humans slept in close proximity. The effects of the observed behaviours on malaria epidemiology are discussed.