We conducted an experiment to test explicitly the accuracy of upper ocean sediment trap fluxes using the particle-reactive radionuclide 234Th (t1/2=24.1 days). Two independent VERTEX-style multitrap arrays were used for collection of sinking particles at 95 m and 97 m depths over a four-day period in May 1992 at the U.S. Joint Global Ocean Flux Study Bermuda Atlantic Time-series Study (BATS) site. Samples for total 234Th were collected every 8 m between the surface and 96 m and immediate combined for analysis to obtain the vertically integrated activity of 234Th. We collected a total of 27 samples over the four-day period. The 234Th samples were collected daily at each of the two traps and every other day on a 6¿6 km grid to characterize the entire source region for particles collected in the traps. In situ flow sensors at one trap array indicated low horizontal shear at the trap mouth (5--10 cm/s) compared to normal values at BATS. The predicted 234Th flux from the watercolumn profiles was not significantly different from zero (-30¿140 disintegrations per minute/m2/d). The measured trap 234Th flux at both arrays was significantly higher (290¿15 dpm/m2/d). We hypothesize that upper ocean traps at Bermuda may overcollect during low-flux periods and undercollect during high-flux periods, thus recording a biased signal of the true particle flux. ¿ American Geophysical Union 1994