In recent years, giant African pouched rats (Cricetomys gambianus) have been used successfully for detecting landmines in Mozambique. The rats are trained to detect mines through the use of operant conditioning techniques, in which the indicator response (stopping and digging) is rewarded (reinforced) with food when it occurs within one meter of a landmine and not rewarded when it occurs under other circumstances. Because handlers do not know the location of landmines when the rats are used operationally, and therefore cannot ascertain if an indication response is correct (near a mine) or incorrect, no rewards are delivered in actual demining. That is, in operational use the rats work under extinction conditions. The present study, conducted on a training minefield in Tanzania, evaluated the effects of repeated extinction on the performance of nine rats trained to detect landmines. Five rats worked in boxes (100 m² areas) with one mine and four rats worked in boxes with four mines. Accuracy (percent mines detected) was good when rewards were provided but declined fairly quickly under extinction conditions and recovered slowly when rewards were reinstated. Arranging parallel training sessions in which rewards were provided lessened the deleterious effects of extinction. These findings suggest that it is unwise to use Cricetomys as mine detection animals for long periods under extinction conditions and that arranging parallel reward training sessions, as is done in actual demining, is an effective solution to lessening extinction effects.