In this paper we provide a brief recount of alternative approaches to what we argue is a fundamental issue for our understanding of sociality – the micro-macro problem or, as we refer to it here, the problem of social emergence. We then discuss recent attempts to identify how the range and type of emergent phenomena changes as a result of changes in the fundamental characteristics of micro-agents. We conclude that there appear to be a number of critical thresholds, notably that which arises when agents become constitutively autonomous and subsequently also develop behavioural (sensori-motor) autonomy. It is the combination of these two levels of autonomy which accounts for what we typically call ‘cognition’ in biological agents. Current artificial intelligence models attempt to replicate the ability without autonomy. While this approach is being seen as increasingly problematic in robotics it appears yet to have influenced approaches to social simulation. We propose achieving behavioural autonomy as a goal and focal point for future simulation research. We argue that this is the minimum threshold needed to achieve social emergence. We illustrate this by discussing the concept of social ‘norm’ as an ‘attractor’ in a phenomenal domain of structurally coupled behaviour.