We investigated the reproducibility of the cytochalasin B micronucleus (MN) assay in irradiated human lymphocytes to assess its suitability in predicting cancer predisposition and response to radiotherapy by virtue of defects in the processing of clastogenic lesions. G0 lymphocytes were exposed to 3.0 Gy 60Co gamma-rays at high (HDR) or low dose-rate (LDR). Six healthy donors were assayed three times each in nine experiments and compared with six ataxia-telangiectasia (A-T) heterozygotes. In controls, significant interexperiment variability in MN yields was observed at HDR and LDR, also in dose-rate sparing (i.e. reduction in MN yield at LDR compared with HDR). Significant inter-individual variability was seen at HDR, but not at LDR or for sparing. Average sparing was 66.4 +/- 4.8%. In spite of the experimental variability, a significant difference between controls and A-T heterozygotes was detected at LDR, and 5/6 heterozygotes had sparing values below the control range. This gives encouragement for the use of this assay in predictive testing if sources of experimental variability can be identified so as to improve discrimination between individuals. HDR and to a lesser extent LDR irradiation induced significant mitotic inhibition, seen as a reduction in binucleate cells after cytocholasin treatment. A positive correlation between mitotic inhibition and MN frequency suggests that similar lesions may be involved in these effects.