This study investigated the accuracy of sixteen models of commercial dental radiometers (DR) in measuring the output of thirty-eight LED light curing units (LCUs) compared with a 'gold standard' laboratory-grade spectrometer integrating-sphere (IS) assembly. Nineteen Type I (fiber-bundle light guide) and nineteen Type II (light source in head) LED LCUs were tested, some using different output modes and light guides, resulting in 61 test subsets per radiometer. Gold standard (GS) output measurements (n = 3) were taken using the IS and confirmed with two types of laboratory-grade power meter (PowerMax-Pro 150 HD and PM10-19C; Coherent). One DR (Bluephase Meter II, Ivoclar; BM II) allowed power (mW) as well as irradiance (mW/cm.sup.2) recordings. Irradiance readings (n = 3) for each DR/LCU were compared with the IS derived irradiance. Individual LCU irradiance values were normalized against IS data. The GS method yielded reproducible data with a 0.4% pooled coefficient of variation for the LCUs. Mean power values ranged from 0.19 W to 2.40 W. Overall power values for the laboratory-grade power meters were within 5% of GS values. Individual LCU/DR normalized irradiance values ranged from 7% to 535% of the GS; an order of magnitude greater than previous reports. BM II was the only radiometer to average within 20% of normalized pooled GS irradiance values, whereas other radiometers differed by up to 85%. Ten radiometers failed to provide any reading for 1 LCU. When tested with the PowerMax-Pro in high speed (20 kHz) mode, eight LCUs demonstrated pulsing outputs undetectable at the standard (10 Hz) data acquisition rate. Sufficient light exposure is critical for the successful curing of dental resin-based materials. Substantial discrepancies may occur between actual and estimated radiometric data using current DRs. More accurate DRs need to be developed. Manufacturers' accuracy claims for DRs should specify compatible LCUs and testing parameters.