In 1977 extensive data were collected to calibrate six Helley-Smith bed load samplers with four sediment particle sizes in a flume at the St. Anthony Falls Hydraulic Laboratory at the University of Minnesota. Because sampler data cannot be collected at the same time and place as ''true'' trap measurements, the ''probability-matching method'' was used to derive surrogate pairs for calibration analysis. The method is invalid since it implicitly assumes that sampler and trap data have no sampling or measurement errors, it gives biased and highly variable results, and it does not contain information enabling model specification. A new calibration model was developed that regresses transformed individual sampler measurements on daily means of transformed trap data and incorporates within-day variation in trap rates to explain part of the sampler variation. Three small-nozzle samplers performed more uniformly than three large-nozzle samplers did. There is evidence that samplers with higher nozzle ratios collect more bed load in most particle size classes tested. However, between the two small-nozzle samplers with ratios of 3.22 and 1.40, significant differences could be detected for only one particle size. The standard sampler with a 76¿76 nozzle trapped sediment less efficiently than a similar sampler with a 152¿152 mm nozzle in three of four particle sizes tested. Limitations in the data restricted more definitive statements about the samplers, but the results of this study can be used to design a more rigorous calibration experiment. ¿ American Geophysical Union 1993