During the TL-and ESR-Conference
in Clermont-Ferrand in 1990 we initiated a second "Intercomparison project on ESR dating" in which 14 ESR groups participated. Each group was provided with two samples of a coral powder, one of which (Sample A) was a fossil coral with a mass spectrometric determined U/Th age. Sample B was a recent coral, irradiated with a definite y-dose (checked by alanine dosimetry). In both cases the AD had to be determined (for sample A the U-content and age as well). Additionally, the chance of a calibration of y-sources by alanine dosimeters was offered (GSF). The results show that (i) the y-source calibration is better than f5 R, (ii) the mean value of the AD from sample A seems to agree with the expected AD but the mean AD value from sample B is overestimated, (iii) systematic errors occur due to the fitting procedure: the AD estimate depends on the maximum y-dose used for the irradiation curve, (iv) the AD determination including the smallest systematic error gives correct values for sample B but too low values for sample A which may be caused by fading of the signal g = 2.0006.