Consequences on mutation induction of a defective excision repair (exr-) system have been studied in Drosophila for a series of 30 carcinogens, representing 17 mono-, six bi- or trifunctional agents, five cyclic alkylating agents and two polycyclic aromatic compounds. Repair-inactive spermatozoa or (late) spermatids were mutagenized and then transferred to excision-defective (mei-9L1) or appropriate excision proficient (exr+) oocytes. Hypermutability responses in exr- in relation to the exr+ genotype were determined by calculating frep-/frep- (ratio = y) indices at x = 1% X-linked recessive lethal mutations (SLRL); frep- denotes induced SLRL frequency in mei-9L1, frep+ that induced in repair-proficient condition, and x = 1 means 1% SLRL in exr+, A linear positive relationship between frep-/frep+ estimates and nucleophilic selectivity (Swain and Scott's constants s) was established for 18 carcinogens, representing either mono- and cyclic alkylating agents: frep-/frep+ = 12.4s - 1.9; r = 0.79, r2 = 0.62, P less than 0.01 Noticeable exceptions to this linear correlation indicated that, although nucleophilicity is a principle determinator for hypermutability response in exr- mutants, other cellular factors play a significant role as well. These are (i) the faster rate of removal of methyl adducts compared to ethyl derivatives, (ii) the complex metabolism of some of the carcinogens (PC, DTIC, DMPT, Cl3DMPT) and (iii) the length of the time period between DNA modification and onset of replication after fertilization. By contrast, CEO and polyfunctional agents (FA, HMPA, MC, MCT, Thio-TEPA and BCNU) did not follow the linear relationship. They provided lower frep-/frep+ indices than would be anticipated on the basis of their nucleophilic selectivity (MC, BCNU, Thio-TEPA) or were even inactive (FA, MCT, HMPA) in the Drosophila repair assay. Both in terms of consistency (mono- and cyclic alkylating agents) and exceptional behavior (CEO, the six crosslinking agents), there is an intriguing positive correlation between relative efficiency ranking of carcinogens and with respect to their position on the potency scale for hypermutability. Thus, in genetic terms, as most potent carcinogens in rodents appear to be those agents giving no effect or a low activity in the exr- genotype in Drosophila: carcinogens with high potential for direct miscoding [ENU, ENNG, DEN, iPMS, CEO and presumably DMBA (?)], and those capable of forming crosslinks (MC, MCT, HMPA, Thio-TEPA and BCNU).