Glioblastoma multiforme (GBM) is the most common and lethal of human primary central nervous system (CNS) tumors, with a median survival of 14-16 months despite optimal surgery, radiation and chemotherapy. A reason for this dismal prognosis is insufficient understanding of the ontogeny of GBMs, which are highly heterogeneous at a pathological level. This pathological diversity, between and within GBMs as well as varying grades of gliomas, has not been fully explained solely on the grounds of oncogenic stimulus. Interaction with the tumor microenvironment is likely a source of this pathological heterogeneity, as well as the inherent characteristics of the tumor cell of origin. Currently, controversy exists on whether the initial transformed cell is a differentiated astrocyte, progenitor or neural stem cell. Putative cancer stem cells (CSCs), which have features of normal stem cell plus the ability to recapitulate the tumor phenotype in vivo in small numbers, have been identified from a variety of solid human cancers, including GBMs. Evidence suggesting that regions harboring normal stem cells in the adult CNS, such as the subventricular zone and the dentate gyrus, are more prone to viral and chemical oncogenesis, is supportive of the hypothesis that brain tumors arise from stem cells. However, it is still to be determined whether the appearance of brain tumor stem cells (BTSC) is the cause or consequence of tumor initiation and progression. This review discusses emerging evidence highlighting the relevance of the state of differentiation and regional heterogeneity in the ontogeny of GBM. This is an area of high interest in cancer in general, with potential significant therapeutic and prognostic implications.