During the evolution of multicellularity, cells undergo an evolutionary transition in individuality, such that groups become the subject of Darwinian evolution. Comparative work, supported by theory, suggests that a life cycle in which cells 'stay together' following cellular division (termed clonal development) should facilitate this transition. While central to our understanding of multicellular evolution, this hypothesis has never been directly tested in a single experimental system. We circumvent this limitation by creating an isogenic yeast system capable of either clonal or aggregative development. We evolved 20 populations of either clonally-reproducing 'snowflake' yeast or aggregative 'floc' yeast with daily selection for rapid growth followed by sedimentation, an environment where multicellularity is adaptive. While both genotypes adapted to this regime, growing faster and having higher survival during the group-selection phase, there was a stark difference in evolutionary dynamics. Competitions reveal that evolved floc obtained nearly all of their increased fitness from faster growth, not improved group survival, while snowflake yeast mainly benefited from higher group-dependent fitness. Through a combination of genome sequencing and mathematical modeling, we identify a trade-off: clonal development can allow selection to act more efficiently on group-beneficial traits, but dramatically increases the overall rate of genetic drift due to mutational bottlenecking. Our results demonstrate how simple differences in the mode of group formation can have profound impacts on the transition to multicellularity: clonal development, but not aggregation, precipitated a transition from cells to groups as the primary level of Darwinian individuality.