The concept that medulloblastomas represent cerebellar neuroblastic tumours was championed by del Rio Hortega in the 1930s and was critically reappraised in the 1960s by Moises Polak. Whereas the aetiology and molecular pathogenesis of medulloblastomas remain unresolved, there is now compelling evidence in support of a fundamentally neuronal tumour phenotype. Tumour cells express in a differentiation-dependent manner a repertoire of neuronal cytoskeletal, synaptic, and other lineage-associated proteins. Neuronal differentiation is more pronounced in the so-called nodular/desmoplastic medulloblastomas, which are typified by areas of neoplastic neuritogenesis ("pale islands") marked by the co-expression of neuronal marker proteins and neurotrophin receptors TrkA and TrkC, low proliferative indices, and apoptosis. The pale islands contain meshworks of reactive astrocytes as part of mutually inductive tumour-stromal cell interactions. However, overt glial differentiation or gliomatous transformation are uncommon. There is growing evidence to support the hypothesis that distinct subtypes of medulloblastomas may implicate transformed neuroblasts from two separate neuroepithelial sources: (a) the velum medullare for a subset of classic medulloblastomas, and (b) the external granule layer for the nodular/desmoplastic medulloblastomas as well as certain classic medulloblastomas. The nosological position of medulloblastomas is discussed in the context of the so-called embryonal central nervous system tumours with emphasis on the cerebral and cerebellar neuroblastomas. We give credence to the view that the medulloblastoma belongs to a group of central neuronal/neuroblastic tumours and call for a critical re-evaluation of its present taxonomic placement.