This report presents the results of noncancer dose-response modeling for inhalation and oral exposures to nickel compounds using the NOAEL/LOAEL and benchmark dose (BMD) approaches. Several key issues associated with the implementation of the BMD approach were examined. Primary among them are difficulties associated with use of data for which the dose-response shape is poorly defined: nonuniqueness of maximum likelihood estimates and lower bounds equal to zero. In addition, several generalizable properties of the "hybrid approach" for modeling continuous endpoints were identified. A hybrid modeling approach allows one to consider "biological significance" on an individual (rather than group) basis; differences between individual- and group-based biological significance in the definition of benchmark response (BMR) levels are elucidated. In particular, it is shown that BMDs defined using group-based BMRs may be more like LOAELs than NOAELs. Application of cross-chemical and cross-endpoint comparisons suggest that, for chronic inhalation exposure, nickel sulfate appears to be as toxic or more toxic than nickel subsulfide and nickel oxide, although the high response rates for the latter two compounds at the lowest chronically administered concentration make such conclusions problematic. A nickel reference concentration could be derived based on the most sensitive benchmark concentration for chronic inhalation exposure to nickel sulfate, 1.7 x 10(-3) mg Ni/m3 for lung fibrosis in male rats. Analyses of oral studies of nickel sulfate and nickel chloride suggest that an appropriate basis for the nickel oral reference dose would be a BMD of 4-5 mg Ni/kg/day, based on increased prenatal mortality. (Uncertainty factors were not determined and neither an RfD nor an RfC was derived in this paper.) The BMD approach provides appropriate quantitative support for toxicological judgment; this paper addresses specific issues associated with the role of the BMD approach in noncancer risk assessment. Resolution of these and other issues may require the accumulation of a number of case studies such as the one presented here.