The vast majority of multicellular organisms develop clonally via 'staying together' after mitotic reproduction. Evolutionary theory predicts that cells staying together provides several key advantages over multicellular construction via cells 'coming together', but little empirical work has directly compared these developmental modes. In our previous work evolving multicellularity de novo in the yeast Saccharomyces cerevisiae, cells evolved to form clonal clusters exclusively through postdivision adhesion of mitotically-produced cells, a result that reflects the strong bias towards clonal development in extant multicellular taxa. An equally parsimonious explanation, however, is that cluster development through incomplete cell separation is simply easier to evolve than the production of the adhesive compounds required for aggregation. To disentangle these hypotheses we repeated the experiment of Ratcliff et al (2012), selecting for rapid settling through liquid medium. Instead of using a unicellular ancestor, however, we started our experiment with five wild strains of yeast capable of aggregating into clusters via flocculation. Clonally-developing 'snowflake' yeast evolved and invaded 36/40 experimental populations within 155 transfers, and competition experiments revealed that invading snowflake yeast were substantially more fit than their floc contemporaries. These results support the hypothesis that clonal development is evolutionarily superior to aggregation, and demonstrate that 'snowflake' yeast can readily evolve in diverse, wild-collected yeast strains.