Multilevel-consistency of social behaviour in a cockroach

Group-living animals may show consistency in both individual behaviour and group-level behaviour. Further, while individuals show consistent differences in their responses to environmental change (plasticity), it is less clear if groups differ in plasticity. Here we exposed groups of cockroaches ( Blaptica dubia ) to three humidities and quantified their social networks. We repeated our design across multiple groups with individuals reassigned to different groups and used machine vision analysis of images to extract 152,800 records of associations across 16 different groups. 116 unique individuals contributed >4,500 individual social network measures and 192 measures of entire networks. Lower humidity led to individuals that had slightly better connectedness to the whole network, but otherwise did not impact individual or group-level phenotypes. Individuals showed some consistency in mean behaviours (12-39%) but scant consistency in plasticity (0-9%). In contrast, groups showed less consistency in mean behaviours than individuals (0-26%) but showed relatively higher among-group differences in plasticity compared to individuals (0-22%). Patterns of group-level variance in individual traits matched patterns of variance in characteristics of entire networks, suggesting these two approaches for quantifying group phenotypes are compatible. While consistency in individual mean behaviour is the norm across taxa, consistency in group-level plasticity is less well understood and may be an emergent phenomenon of collective dynamics that deserves further investigation. Using automated marker recognition techniques such as machine vision allows us to collect the large datasets necessary to simultaneously test hypotheses at the individual and group level and so can be more widely adopted.