Protoplast fusion commonly has three problems that prevent it from being applied to practical breeding of industrial yeasts : the difficult separation of fusants, introduction of undesired phenotypes, and potential instability in yeasts. In this study, we investigated solutions to these problems by model breeding of yeasts. A nonflocculent Brewers’ yeast was converted to flocculent by electrofusion. The fusants contained undesired phenotypes for braving. Refusion between the Brewers’ yeast and the fusants was carried out and the refusants were separated by flow cytometry. Sixteen strains of the refusants were characterized by small-scale brewing (2 l); they were improved in comparison with the fusants from fermentation and flocculation activities. The brewing was repeated seven times for the refusants (RF22, RF195, RF212) and a practical Brewers’ yeast. Both RF195 and the Brewers’ yeast had almost the same fermentation curves and their ploidies were constant. However, both RF22 and RF212 had unstable curves and their ploidies became lower as the brewings were repeated. These results suggest that flow cytometry can separate prototroph yeast fusants. The refusion was effective for exclusion of undesired phenotypes from yeasts. Small-scale brewing was also effective for screening of stable yeasts.