How infected hosts behave critically affects both their fitness and the transmission, and thus fitness, of their parasites. Understanding host–parasite coevolution therefore requires understanding the factors affecting infected host behaviour. This can usefully be decomposed into the current behavioural expression level and any infection-induced changes in that expression: each of these may have different implications for host and parasite fitness. For example, from the perspective of socially transmitted parasites, the more contacts infected hosts make with susceptible conspecifics, the better, regardless of their pre-infection contact behaviour. For social hosts, however, minimizing infection-induced changes to fitness-enhancing behaviours (‘behavioural tolerance’) may reduce the risk of social exclusion and its associated costs. Here, we tested how the presence of conspecifics (‘social context’) affects the behaviour and behavioural tolerance of guppies, Poecilia reticulata, infected with the socially transmitted ectoparasite Gyrodactylus turnbulli. In the absence of a stimulus shoal of three females, males that lost the most mass over the course of infection (i.e. had lowest tissue-specific tolerance) were significantly less active, whereas in the presence of the shoal, male activity levels were high and not correlated with tissue-specific tolerance. Females were consistently active, regardless of their tissue-specific tolerance or social context. Behavioural tolerance, quantified as the per-parasite change in activity level, also differed between the sexes in the absence of the shoal: among females, behavioural tolerance was negatively correlated with tissue-specific tolerance, whereas in males the correlation was strikingly positive. In the presence of the shoal, however, tolerance components were not correlated in either sex. Overall, these sex differences in behaviour and behavioural tolerance indicate that females are highly competent at transmitting this parasite, and that males can conceal their disease when in the presence of females. Our results contribute to the growing literature on factors affecting variation between infected hosts, which fundamentally affects epidemiological predictions.