Clinical proton beam therapy has been based on the use of a generic relative biological effectiveness (RBE) of 1.0 or 1.1, since the available evidence has been interpreted as indicating that the magnitude of RBE variation with treatment parameters is small relative to our abilities to determine RBEs. As substantial clinical experience and additional experimental determinations of RBE have accumulated and the number of proton radiation therapy centers is projected to increase, it is appropriate to reassess the rationale for the continued use of a generic RBE and for that RBE to be 1.0-1.1. Results of experimental determinations of RBE of in vitro and in vivo systems are examined, and then several of the considerations critical to a decision to move from a generic to tissue-, dose/fraction-, and LET-specific RBE values are assessed. The impact of an error in the value assigned to RBE on normal tissue complication probability (NTCP) is discussed. The incidence of major morbidity in proton-treated patients at Massachusetts General Hospital (MGH) for malignant tumors of the skull base and of the prostate is reviewed. This is followed by an analysis of the magnitude of the experimental effort to exclude an error in RBE of >or=10% using in vivo systems. The published RBE values, using colony formation as the measure of cell survival, from in vitro studies indicate a substantial spread between the diverse cell lines. The average value at mid SOBP (Spread Out Bragg Peak) over all dose levels is approximately 1.2, ranging from 0.9 to 2.1. The average RBE value at mid SOBP in vivo is approximately 1.1, ranging from 0.7 to 1.6. Overall, both in vitro and in vivo data indicate a statistically significant increase in RBE for lower doses per fraction, which is much smaller for in vivo systems. There is agreement that there is a measurable increase in RBE over the terminal few millimeters of the SOBP, which results in an extension of the bioeffective range of the beam in the range of 1-2 mm. There is no published report to indicate that the RBE of 1.1 is low. However, a substantial proportion of patients treated at approximately 2 cobalt Gray equivalent (CGE)/fraction 5 or more years ago were treated by a combination of both proton and photon beams. Were the RBE to be erroneously underestimated by approximately 10%, the increase in complication frequency would be quite serious were the complication incidence for the reference treatment >or=3% and the slope of the dose response curves steep, e.g., a gamma(50) approximately 4. To exclude >or=1.2 as the correct RBE for a specific condition or tissue at the 95% confidence limit would require relatively large and multiple assays. At present, there is too much uncertainty in the RBE value for any human tissue to propose RBE values specific for tissue, dose/fraction, proton energy, etc. The experimental in vivo and clinical data indicate that continued employment of a generic RBE value and for that value to be 1.1 is reasonable. However, there is a local "hot region" over the terminal few millimeters of the SOBP and an extension of the biologically effective range. This needs to be considered in treatment planning, particularly for single field plans or for an end of range in or close to a critical structure. There is a clear need for prospective assessments of normal tissue reactions in proton irradiated patients and determinations of RBE values for several late responding tissues in laboratory animal systems, especially as a function of dose/fraction in the range of 1-4 Gy.