SUMMARYA total of 245 test cross progenies obtained by crossing cytoplasmic male sterile (CMS) lines and restorer (RHA) lines were evaluated in replicated trials in 1995 and 1997. Experiments were carried out in two locations (Osimo, East Central Italy and Budrio, Northern Italy) in 1995, whereas in 1997 trials were conducted only in Osimo. General combining ability (GCA) and specific combining ability (SCA) of the parents was evaluated to identify genotypes suitable to be used as testers for breeding programs and to verify the performance of hybrids. The results obtained in 1995 were used to make test crosses in 1996 to be evaluated in 1997. The GCA and SCA of RHA and CMS lines were nearly always significant for all the studied traits when lines were randomly taken in 1995. When selected testers were used on the basis of the results of the first year, the SCA of new RHA lines and GCA of new CMS lines were not significant for achene yield, but they were for the other traits.The best GCA estimates for RHA were often higher than those of CMS, indicating that selection for RHA could be more useful than for CMS. As expected, GCA was always lower than SCA. CMS and RHA, with high variance among their test crosses for yield and many other traits, could be used as testers. The lack of a tester with high variance in all traits requires more than one tester in evaluating lines. The GCA of a line can change in function of the germplasm with which it is combined.