In earlier studies using asynchronously growing Chinese hamster cells, we observed substructure in the survival response at low doses. The substructure appeared to result from subpopulations of cells having different, cell cycle phase-dependent radiosensitivity. We have now applied the same flow cytometry and cell sorting technique to accurately measure the responses of cells of eight different asynchronously growing human tumor cell lines, representing a wide range in radiosensitivity. When the data were fitted with a linear-quadratic (LQ) function, most of these lines showed substructure similar to that observed in Chinese hamster cells, with the result that values of alpha and beta were dependent on the dose range used for fitting. Values of alpha describing the low-dose response were typically smaller (by as much as 2.2 times) than the alpha describing the high-dose response, while values of beta were larger at low doses. Values of alpha/beta from our measurements are in reasonable agreement with other values published recently if we fit the data for the high-dose range (excluding, for example, 0-4 Gy), which corresponds to a conventional survival response measurement. However, the values of alpha/beta describing the low-dose range were, on average, 2.8-fold smaller. The results show that the usual laboratory measurement of cell survival over 2 or 3 logs of cell killing, if fitted with a single LQ function, will yield alpha and beta values which may give a rather poor description of cell inactivation at low dose in asynchronous cells, no matter how carefully those measurements are done, unless the low-dose range is fitted separately. The contribution of killing represented by the beta coefficient at low doses was found to be surprisingly large, accounting for 40-70% of cell inactivation at 2 Gy in these cell lines. A two-population LQ model provides excellent fits to the data for most of the cell lines though, as one might expect with a five-parameter model, the best-fitting value of the various parameters is far from unique, and the values are probably not reliable indicators of the size and radiosensitivity of the different cell subpopulations. At very low dose, below 0.5-1 Gy, another order of substructure is observed: the hypersensitive response; this is described in the accompanying paper (Wouters et al., Radiat. Res. 146, 399-413, 1996).