Thirty-two radiotherapy centers in the USA and Canada cooperated in a study of the variability of clinical thermoluminescent dosimetry (TLD) systems. The primary purpose of the survey was to ascertain the accuracy of TLD for the determination of in vivo dose measurements. Each participating institution provided two TLD packets for irradiation on a Clinac 4, at a prearranged time. Two batch irradiations were made. Thirty-two TLD packets, one from each institution, were uniformly irradiated to a dose of 22.35 cGy (known by us, but not by the participants). A second group of 32 packets were likewise irradiated to a dose of 179.0 cGy. Participants were told only that their TLD''s would be irradiated to doses between 10 and 50 cGy, and 100 to 200 cGy. TLD''s were then returned to the institutions of origin for readout, and the doses reported to use for analysis. Calibration factors, readout and annealing procedures, etc., were all established independently by each participant. Although these procedures varied widely between institutions, the mean values of the reported doses were within 5% and 3% of the expected values for the low and high doses, respectively. Standard deviations in the reported doses were 10% and 5%. Also of interest, however, is the finding that 22% (i.e., 14 out of 64) of the dose reportings were in error by more than 10%. The implications of these findings vis a vis radiotherapy are discussed.