Background. An ascertained possibility of culling segregating populations on the basis on their yield or grain weight per spike as a character closely correlated with yield can increase the efficiency of breeding practice with spring bread wheat. Materials and methods. Sixteen F2 –F4 segregating populations of spring bread wheat were assessed for their grain weight per spike, average grain weight per spike among the top 10%, coefficient of variation for grain weight per spike, and yield. The pedigree method was applied to select 2 to 3 F7 lines from each population. The obtained lines were tested for their yield in F8 and F9 . The data of grain weight per main spike in the parents that had been crossed to produce the studied segregating populations were retrospectively taken into account while the results were processed. Results Segregating populations persistently differed in their average grain weight per spike and the coefficient of its variation in different years of research. None of the 36 late-generation lines significantly surpassed the reference cultivar in yield. All the considered indicators of segregating populations had no significant correlations with the yield of late-generation lines. Average values of grain weight per main spike in the parent genotypes and the best parent as well as the differences between the parents in the average grain weight per main spike tended to have a negative correlation with the yield of late-generation lines (r up to –0.58). Conclusion. The culling of segregating populations based solely on the yield or grain weight per spike is associated with a risk of losing high-yielding genotypes. Late-generation lines with high yields occur in combinations where the grain weight per main spike in one parent or on average among the parents is not expressed to its highest level or when both parents differ only slightly from each other in the expression of this character.