Play is commonly used to assess affective states in both humans and non-human animals. Play appears to be most common when animals are well-fed and not under any direct threats to fitness. Could play and playfulness therefore indicate pre-existing positive emotions, and thence optimal animal welfare? We examine this question by surveying the internal and external conditions that promote or suppress play in a variety of species, starting with humans. We find that negative affective states and poor welfare usually do suppress play (although there are notable exceptions where the opposite occurs). Furthermore, research in children suggests that beyond the frequency or total duration of play, poor welfare may additionally be reflected in qualitative aspects of this heterogeneous behaviour (e.g. display of solitary over social play; and the ‘fragmentation’ of play bouts) that are often overlooked in animals. There are surprisingly few studies of play in subjects with pre-existing optimal welfare or in unambiguously highly positive affective states, making it currently impossible to determine whether play can distinguish optimal or good welfare from merely neutral welfare. This therefore represents an important and exciting area for future research. © 2017 Elsevier B.V.