Spermatogenesis in mice and other mammalians is supported by a robust stem cell system. Stem cells maintain themselves and continue to produce progeny that will differentiate into sperm over a long period. The pioneering studies conducted from the 1950s to the 1970s, which were based largely on extensive morphological analyses, have established the fundamentals of mammalian spermatogenesis and its stem cells. The prevailing so-called Asingle (As) model, which was originally established in 1971, proposes that singly isolated As spermatogonia are in fact the stem cells. In 1994, the first functional stem cell assay was established based on the formation of repopulating colonies after transplantation in germ cell-depleted host testes, which substantially accelerated the understanding of spermatogenic stem cells. However, because testicular tissues are dissociated into single-cell suspension before transplantation, it was impossible to evaluate the As and other classical models solely by this technique. From 2007 onwards, functional assessment of stem cells without destroying the tissue architecture has become feasible by means of pulse-labeling and live-imaging strategies. Results obtained from these experiments have been challenging the classical thought of stem cells, in which stem cells are a limited number of specialized cells undergoing asymmetric division to produce one self-renewing and one differentiating daughter cells. In contrast, the emerging data suggest that an extended and heterogeneous population of cells exhibiting different degrees of self-renewing and differentiating probabilities forms a reversible, flexible, and stochastic stem cell system as a population. These features may lead to establishment of a more universal principle on stem cells that is shared by other systems.