Our group has shown that many of the progeny of cells which survive irradiation, as determined using a clonogenic assay, die out after 8-30 further cell divisions. Correction of conventional survival curves for this extra component of defective cells (termed lethally mutated cells) reduces or eliminates the 'shoulder' which is characteristic of the radiation response of many mammalian cell lines. Since the size of the shoulder is theoretically and experimentally linked with the extent of split-dose recovery, this paper examines the occurrence of lethal mutations following fractionated irradiation using a variety of experimental conditions. The results show that, when curves are corrected for lethal mutations to give residual survival, the size of the single-dose survival-curve shoulder is indeed reduced but the extent of recovery after split-dose irradiation remains the same or, under some conditions, is actually increased. However, the rate of increase in survival with time between doses is reduced over the first 2 h if the data are corrected for lethal mutations, suggesting that early postirradiation repair may be error prone. When a metabolic inhibitor which depletes cellular ATP was used, the single-dose and split-dose curves corrected for lethal mutations were coincident with each other and with the corrected single-dose control curve, all being exponential and with an extrapolation number of one. It is concluded that the mechanisms leading to the production of the primary survival-curve shoulder are different from those leading to split-dose recovery. The results strongly suggest that a mechanism involving induction of repair/resistance by an initial dose of radiation is involved in split-dose recovery.