Abstract Uranium dosimeter glasses SRM 612, CN 1 and CN 2 have been calibrated against Co monitors in 79 irradiations in the Herald reactor over seven years. Mean values of the calibration factors B for each glass are B 612 = (5.736 ± 0.055)·10 9 (74 irradiations), B CN 1 = (1.883 ± 0.026)·10 9 (21 irradiations) and B CN 2 = (2.014 ± 0.0 26)·10 9 (21 irradiations). Comparison of relative responses of the four dosimeters reveals that unaccounted errors exist in the response of the Co monitors and, to a lesser extent, in the response of SRM 612. The errors associated with the response of the natural uranium glasses CN 1 and CN 2 are represented by conventional (Poisson) “counting statistics”. These results show that attempts to calibrate a uranium glass against an activation monitor in only a small number of irradiations may produce results radically discrepant from the true value. The importance of systematic errors in neutron dosimetry for fission-track dating is also discussed. An alternative zeta (ζ) calibration approach is described, which circumvents absolute φ and λ f evaluation: each dosimeter glass is calibrated repeatedly against zircon age standards from the Fish Canyon and Bishop tuffs, the Tardree rhyolite and Southern African kimberlites, to obtain empirical calibration factors ζ. The weighted mean ζ-values are 339 ± 5 for SRM 612, 113.0 ± 2.6 for CN 1 and 121.0 ± 3.6 for CN 2. Independent K/Ar, 40Ar/ 39Ar and Rb/Sr calibrating ages for the standards are discussed. For two of the three glasses, the presented ζ-values derived from each of the zircon standards are consistent within error. Compatibility of the kimberlite data with that of the other samples is discussed. Age calculation by direct comparison of track density ratios in sample and standard is rejected as grossly imprecise. Examination of the reproducibility of results from repeated measurements indicates the conventional calculation of error to be reasonable, but shows the approach of Johnson et al. (1979) to give a serious over-estimate of precision.