In principle, alpha and beta can be obtained from single-dose survival curves using standard linear-quadratic fitting; however, alpha and beta being interdependent, it is difficult to evaluate them together with good precision. On the assumption that full recovery from a split-dose treatment gives a result that is the product of the single-dose surviving fraction, it has been suggested that the measurement of split-dose recovery should provide a method to measure beta alone using the formula: beta RR = lnRR/2d2. Most of the studies published to date have been carried out on cancer cell lines or transformed normal cells. We have systematically tested the above proposal on two normal human fibroblast cell lines (HF19 and 1BR3) in two different situations: growing cells, and plateau-phase cells. Two different protocols were used to assess both the potential influence of a priming dose on the surviving cells and the extent of the split-dose recovery. The survival curves generated after different priming doses did not show any significant change in comparison with those achieved without previous irradiation. In addition, the split-dose survival was not different from the square of the corresponding single-dose survival (model free). In these conditions, beta RR's obtained by a linear regression of the recovery ratio data were very similar to the beta's obtained by single doses. However, a curvilinear regression (with a very small negative term at high doses) appears to be more appropriate for cells in plateau phase. This has the result that, as the dose increases, the cell survival curves tend to become less bending than would be expected from the linear-quadratic model; however, the linear-quadratic fitting is still a reasonable characterization of the radiation response since the in vitro colony formation method does not allow measurement of survival < 10(-4).