Evolutionary transitions require the organization of genetic variation at two (or more) levels of selection so that fitness heritability may emerge at the new higher level. For example, in case of the transition from single cells to multicellular organisms, single cells must, as it were, relinquish their claim to flourish and multiply in favor of the multicellular group. In this paper we consider the consequences on fitness variation and heritability of two main modes of reproduction used in multicellular organisms: vegetative reproduction, where the offspring originates from a group of cells of the adult (a propagule), and single-cell reproduction, where development starts from only one cell. Most modern organisms pass through a single-cell stage during their life-cycle, a possible explanation being that the single-cell stage increases the effectiveness of organism selection relative to cell selection, by increasing the kinship among cells within the organism. To study this hypothesis we consider simple cell colonies reproducing by fragments or propagules of differing size, with mutations occurring during colony growth. Mutations are deleterious at the colony level, but can be advantageous or deleterious at the cell level (termed iselfishi or iuniformly deleteriousi mutants, respectively). In our model fragment size affects fitness in two ways, through a direct effect on group size (which in turn affects fitness) and by affecting the within and between group variances and opportunity for selection on mutations at the two levels. We show that the evolution of fragment size is determined primarily by its direct effects on group size, except when mutations are selfish. When mutations are selfish, smaller propagule size may be selected, including single-cell reproduction, even though smaller propagule size has a direct fitness cost by virtue of producing smaller groups. Using continuous distributions of mutational effects, we show that selfish mutants have an important effect on mutational load and selection on propagule size, even when selfish mutations are relatively infrequent. We then consider the role of deleterious mutation in the evolution of the germ line. Two possible ways to mediate conflict in the germ line are considered: reduction in development time (of the germ line relative to the soma) and lowered mutation rate in the germ line. The evolution of shorter development time in the germ line depends critically on whether and how the number of gametes influences fitness. If there is a direct effect of the number of gametes on fitness, it will be difficult for shorter development times in the germ line to evolve. We conclude that a lowered mutation rate in the germ line relative to the soma provides the most robust rationale for the origin of the germ line.