1.AbstractUnderstanding variation in social organization that does not have a strong phylogenetic signal represents a key focus of research in behavioural and evolutionary ecology. In light of this, we established a sociality framework that identifies four categories of variation in social organisation that range from large-scale to fine-scale and can each be related to various ecological factors: (1) forms of sociality, (2) degree of sociality, (3) social plasticity, and (4) within-group plasticity. We modelled this framework by quantifying the four categories of variation over time, space and disturbance regime using multiple species of coral-dwelling gobies from the genusGobiodon. Gobies are a particularly interesting model system as they vary in social structure, have within-group cooperation and form mutualistic relationships with their coral hosts which are vulnerable to climatic disturbances. We found that gobies varied in forms of sociality – from being solitary, to paired or group-living depending on location and disturbance regime. Only low or moderate degrees of sociality were observed in gobies, and this was influenced by location or disturbance regime depending on species. Gobies were more often solitary or pair-forming than group-forming (which became extremely rare) in a high disturbance regime whereas they were more often found in groups in a moderate disturbance regime. The size of coral hosts affected the social plasticity of gobies, and corals were smaller due to climatic disturbances. Gobies did not exhibit within-group social plasticity, as there were no changes to the structure of size-based hierarchies or sex allocation patterns with location or disturbance regime. Lastly, by combining the four categories of variation, we find that there is a high loss of sociality in coral-dwelling gobies due environmental disturbances, which likely affects overall goby survival as living in groups can improve survival and fitness. By using our structured framework, we identified which categories of social variation were influenced by ecological factors like location and disturbance. This framework therefore provides an excellent tool for predicting future responses of animal societies to environmental stressors.