Despite its biological richness, the rodent and marsupial fauna of many parts of the Amazon Basin remains poorly known and the efficacy of different methods in assessing its diversity are poorly understood. We present results of small mammal trapping at a previously unsurveyed site in the Xingú Basin of the southeastern Amazon, Pará, Brazil; provide details on a new method for arboreal trapping; and compare species richness among traps at different heights and between different trap types. Mammals were livetrapped at 3 trap heights: ground, understory (1–3 m above ground), and canopy (mean height = 11.5 m ± 3.2 SD, range = 4.8–16.8 m, n = 76 trap stations). We recorded 1,769 captures of 1,178 individuals of 8 marsupial and 17 rodent species in 38,090 trap-nights (or 19,320 station-nights, where a station-night consisted of 1 Sherman and 1 Tomahawk live trap at 1 height for 1 night). Overall trap success was 6.1 individuals per 100 station-nights; success by trap position was 9.0% for ground (7,850 station-nights), 4.7% for understory (7,850 station-nights), and 2.9% for canopy (3,490 station-nights). Success by trap type was 4.7% for Shermans and 5.9% for Tomahawks, with Tomahawk traps showing a more rapid accumulation of species. Rarefied species accumulation curves showed little improvement with the inclusion of canopy trapping, which we attribute to high variability in the development of vertical structure at the site. We suggest that in areas with low and often-broken canopies, only ground and understory traps need be employed for long-term studies; however, some form of canopy trapping should be used during initial surveys so that the utility of arboreal trapping can be evaluated.