OBJECTIVE This study aimed to assess the influence of the level of initial injury on the long-term recovery kinetics of radiation damage in the central nervous system using a rat spinal cord model. METHODS The adult rat spinal cord (C2-T2) was initially given two or three daily fractions of 9 Gy, or three daily fractions of 10.25 Gy. At day 4 or weeks 6, 8, 12, 20, 28, 40, or 52, animals were reirradiated with graded single doses of X rays. The end point was forelimb paralysis caused by white-matter necrosis. RESULTS Latent times to paralysis as measured from the date of the initial treatment increased with increasing time interval between initial treatment and reirradiation but decreased with increasing size of initial injury. Retreatment ED50s were 14.1, 14.8, 15.4, 16.3, and 16.2 Gy for animals reirradiated at day 4 and weeks 8, 12, 20, and 28, respectively, after an initial dose of 9 Gy x 2. After 9 Gy x 3, the retreatment ED50s at day 4 and weeks 6, 8, 12, 20, 28, 40, and 52 were 10.0, 9.9, 9.8, 12.0, 13.9, 14.6, 14.7, and 15.5 Gy, respectively. For an initial dose of 10.25 Gy x 3, the retreatment ED50s at day 4 and weeks 8, 12, 20, 28, and 40 were 5.8, 6.1, 8.4, 10.6, 12.2, and 13.3 Gy, respectively. Using the linear-quadratic (LQ) model, alpha/beta of 3.0 Gy, to quantitate the biological effect of the different retreatment schedules, the initial doses of 9 Gy x 2 or 3, or 10.25 Gy x 3 were found to represent 47, 71, and 89% of the extrapolated response dose (ERD), respectively, and no significant increase in tolerance was observed for retreatment given within 8 weeks of initial treatment. Significant long-term recovery was observed thereafter and increased with increasing time interval to retreatment. The retreatment tolerance and radiation damage recovered at different intervals were influenced by the initial dose. Using direct analysis, the recovery kinetics could be best described by introducing a time function consisting of a linear and quadratic time component dependent on initial dose to the LQ model. CONCLUSIONS These results are consistent with the presence of significant long-term recovery of radiation damage in rat spinal cord, and suggest that the size of the initial damage influences the recovery kinetics, and hence the retreatment tolerance.