Medulloblastoma (MB) is the commonest malignant brain cancer in children that arises from a structure called the cerebellum. Four subtypes are recognised which have different biology and vary in their outlook. One subtype termed sonic hedgehog (SHH) medulloblastoma (SHH-MB) includes a lethal form that has a mutation in the gene, TP53. In these tumours, primitive so-called cancer stem cells are responsible for tumour initiation and relapse. To understand their biology and find therapeutic targets it is essential to grow tumour cells in the laboratory. In general however, these cells are hard to grow with a few exceptions including a cell line that harbours a mutation in TP53. Unfortunately, these laboratory-grown cells poorly recapitulate patient tumours. Moreover, medulloblastoma mouse models also differ in their properties from patient tumours. For these reasons, therapies that seem promising in the laboratory usually fail in the clinic. Accumulating evidence points to the lack of a human environment for laboratory-grown medulloblastoma cells as the reason for tumour model deficiencies. We have grown ‘mini’ human cerebellum organoids in vitro, from a type of human stem cell which can make any cell type in the body. By growing medulloblastoma cell lines in these organoids we were able to sample thousands of healthy and tumour cells to determine their properties in a technique termed single cell sequencing. Our data shows the primitive tumour cells regain the capacity to form specialised nerve cells called neurons, when grown in their native micro-environment, just as in patient tumours. Surprisingly, a subgroup of SHH-MB cells adopts a cancer stem cell identity that is not observed in cell lines grown in isolation. We identified a candidate signal that might trigger this change present in cerebellar organoids. In future, we hope to exploit our model to identify therapeutic vulnerabilities of these putative cancer stem cells.