Five breeding lines and 4 cultivars were grown in replicated trials at 1 location in 1999-2000 and 2 locations in 2000-2001 cropping seasons in izmir in order to determine the usefulness of a safety-first selection index in making selection among high yielding wheat genotypes for test weight and thousand-kernel weight. Genotype x environment interactions were significant for these 2 quality traits. The regression coefficient (b(i)) of genotypes, variance (S(I)(2)) of a genotype across environments, and a safety-first index (I(I)) were calculated to evaluate the stability of genotypes for test weight and thousand-kernel weights. Wheat genotypes with lower b(i) values and small variance (S(2)) across environments (the most stable ones) tended also to have lower means for these 2 traits, However, those genotypes with the largest value of index (I), which is considered desirable, were found to have the highest mean values for both test weight and thousand-kernel weight. The rank-correlation coefficient between (I(I)) and ((Y) over bar (1)) values was significant for thousand-kernel weight. The ranking from I was significantly correlated with the rankings from the stability parameters b(I) and S(I)(2) for test weight. The rank correlation coefficients between b(I) and S(2) for the 2 physical quality traits were highly significant and of similar magnitude (r = 0.983). It was concluded that a safety-first selection index can be useful in plant breeding for the improvement of wheat cultivars with enhanced stability and consistency of quality characteristics when genotype x environment interaction is present.