Ice streams are river-like corridors of fast flowing ice that account for the vast majority of ice discharge to the ocean in continental ice sheets. Their most outstanding feature is that they can appear spontaneously within a slowly moving ice sheet, self-organize in evenly spaced patterns, and switch on and off over time. Yet, a full explanation of ice stream formation and evolution is one of the longest standing open problems in glaciology. This knowledge gap has precluded fundamental investigations on the role of ice streams in driving ice sheet change, and also casts doubt on the ability of state-of-the-art ice sheet simulation codes to project future sea levels. In this talk, I strip away much of the sophistication involved in “operational” ice sheet models to look at the ingredients necessary to capture ice stream dynamics in minimal continuum models. I first identify fundamental flaws in the established theory of sliding initiation, to show instead how a region of subtemperate sliding (i.e., where sliding occurs below the melting point and sliding speeds are limited by the need to maintain the basal energy budget in balance) is essential to enable the onset of fast, sliding-dominated, ice stream flow. Next, I move on to outlining how ice streams may emerge spontaneously out of an otherwise uniform ice sheet as a result of a newly identified instability of subtemperate regions. Last, I discuss how the pattern of englacial deformation in the onset region of the Institute Ice Stream (West Antarctica) is consistent with an extended region of sliding initiation, thus lending support to the theoretical findings discussed above.