Previous data suggested that the linear-quadratic (LQ) model underestimated the sparing effect of small doses per fraction on the radiation tolerance of rat cervical spinal cord when 24 h was allowed between fractions for repair. In these experiments, animals had been given initial top-up doses consisting of 3 daily fractions of 9 Gy to represent 75% tolerance, followed by small fractionated doses in 1, 10, 20, 30 and 40 fractions given once daily. The end-point was forelimb paralysis secondary to white matter necrosis. To assess the possible perturbation of the initial top-up doses on the biological system, an experiment was performed with the small fractionated doses given initially, followed by the same top-up (final top-up) doses. The ED50s for 1, 10, 20, 30 and 40 fractions followed by final top-up doses were 9.5 +/- 0.3, 22.6 +/- 0.6, 32.4 +/- 0.6, 37.7 +/- 0.8 and 41.7 +/- 0.9 Gy, respectively; the corresponding ED50s obtained from the initial top-up experiment were 10.0 +/- 0.4, 20.7 +/- 0.5, 30.0 +/- 0.8, 37.0 +/- 0.8 and 39.9 +/- 0.7 Gy for 1, 10, 20, 30 and 40 fractions, respectively. Using the direct method of analysis and assuming complete repair between fractions, data from both experiments were not adequately described by the LQ model, which gave small alpha/beta values of 0.97 Gy for the initial top-up experiment and 1.23 Gy for the final top-up experiment, in contrast to an alpha/beta value of 2.41 Gy for the experiment with full course fractionation, fraction sizes down to 2 Gy.(ABSTRACT TRUNCATED AT 250 WORDS)