Embryoid bodies (EBs) have been popular in vitro differentiation models for pluripotent stem cells for more than five decades. Initially, defined as aggregates formed by embryonal carcinoma cells, EBs gained more prominence after the derivation of karyotypically normal embryonic stem cells from early mouse blastocysts. In many cases, formation of EBs constitutes an important initial step in directed differentiation protocols aimed at generated specific cell types from undifferentiated stem cells. Indeed state-of-the-art protocols for directed differentiation of cardiomyocytes still rely on this initial EB step. Analyses of spontaneous differentiation of embryonic stem cells in EBs have yielded important insights into the molecules that direct primitive endoderm differentiation and many of the lessons we have learned about the signals and transcription factors governing this differentiation event is owed to EB models, which later were extensively validated in studies of early mouse embryos. EBs show a degree of self-organization that mimics some aspects of early embryonic development, but with important exceptions. Recent studies that employ modern signaling reporters and tracers of lineage commitment have revealed both the strengths and the weaknesses of EBs as a model of embryonic axis formation. In this review, we discuss the history, application, and future potential of EBs as an experimental model. For further resources related to this article, please visit the WIREs website.