Unusual repetitive behaviours in confined animals (including those sometimes called stereotypies) have long been used as welfare indicators because they are disturbing and fairly easy to assess. Because it is not yet certain which behaviours should be included (for example, is wheel-running normal or abnormal?), how much the heterogeneity of different forms matters, and which are most analogous with stereotypies versus OCDs versus other clinical abnormal behaviours, I will group all as “abnormal repetitive behaviour” (ARB). Despite these unknowns, enough is now understood to use ARBs in welfare assessment (where welfare means affective state: moods, and quality of life). I will review the aetiology of ARBs, and discuss their underlying mechanisms (including unintended reinforcement by owners), to help illustrate why it is that they emerge. To evaluate their validity as indicators of welfare, I will review whether they are increased by exposing animals to aversive stimuli and stimuli that are ancestrally bad for fitness. I will show that the prevalence and/or frequency of ARBs typically reflects suboptimal husbandry and uncomfortable health problems, and that they are quite specific to negative states (though perhaps as experienced over the lifetime, rather than just present state alone). Overall, ARBs are thus reliable signs of poor welfare. However, general activity can be a confound. Indeed, some negative states never promote ARBs; and in some species, strains and individuals show little ARBs, even in extremis, becoming inactive instead (such that all else being equal, we should not assume that high ARB individuals have worse welfare than low ARB individuals).