Abstract: Foraging by screaming hairy armadillos, Chaetophractus vellerosus, and guinea pigs, Cavia porcellus, was studied in the laboratory. The main question was whether patch exploitation varies with overall capture rate as predicted by the marginal value theorem (MVT). Armadillos in experiment I and guinea pigs in experiment II experienced a single travel time between depleting patches of two kinds: good and poor. There were two treatments, which differed in the quality of poor patches. MVT predicts that within a treatment, more prey should be taken from good than from poor patches and between treatments, good patches should be exploited in inverse relation to the quality of poor patches and poor patches should be exploited in direct relation to their own quality. In experiment III, guinea pigs experienced three treatments which differed in the travel requirement, while the two patch types remained the same. MVT predicts that within a treatment more prey should be taken from good than from poor patches and that between treatments more prey should be taken from both patch types as travel requirement increases. The qualitative predictions were supported in the three experiments. The quantitative fit was good but there was a bias towards more severe patch exploitation than predicted. The results indicate that in these species patch exploitation depends on overall food availability as predicted by the MVT when overall food availability differs either because of patch type composition or because of differences in travel requirement between patches.