The distribution and ecological roles of ground-dwelling invertebrates colonizing dry beds of temporary rivers are poorly known, due in part to the lack of tested sampling methods. Sampling efficiency and completeness and detection biases associated with pitfall-trap and quadrat samples were compared at six sites in the dry beds of two New Zealand rivers. On average, pitfall-trap samples contained 3.5 times more taxa than quadrat samples, and pitfall traps required less time to collect as many taxa as quadrats. Among all taxa collected, 80% were exclusive to pitfall traps and 4% to quadrat samples. Among-sites differences in invertebrate composition were greater in pitfall-trap samples than in quadrat samples. Rarefaction curves indicated that eight to eleven pitfall traps/100 m2 were necessary to collect most of the common taxa in the study sites. Differences in the performance of pitfall traps and quadrats in dry riverbeds may be due to the absence of vegetation, to the large range of taxa considered, and to diel movements of some taxa. The large number of samples required at our study sites to reach a high level of sampling completeness suggests that the diversity of ground-dwelling invertebrates in dry riverbeds has probably been underestimated in previous studies.