Perceived group membership, perceptions about accompanying group norms and the degree to which a person identifies with a social group are predictive of a wide range of human behaviours. Behavioural clinical trials in general, and HIV-prevention intervention trials in particular, however, have not examined the degree to which individuals who join a large behavioural study (and hence, a group) may, in an unanticipated way, develop a sense of social identity related to the study, and how this identity or associated group norms may influence participants' behaviours and, potentially, study outcomes. Project EXPLORE was a large-scale behavioural intervention trial in six US cities to prevent HIV seroconversion among men who have sex with men (MSM) (EXPLORE Study Team, 2004). We previously found that participants (examined at one study site) were more likely to engage in high-risk sexual activities with other MSM who were EXPLORE participants than other partners. The present ancillary study (n=271) sought to examine the degree to which perceived group membership, group identity and group norms among EXPLORE study participants was associated with sexual behaviour with other EXPLORE participants, high-risk sexual behaviours with other EXPLORE participants and intentions to engage in high-risk sex with other EXPLORE participants. A principal components analysis of a 14-item scale assessing perceived group membership and norms regarding being part of EXPLORE yielded six principal components (PCs): PC1: perception that EXPLORE participants engage in safer sex; PC2: social comfort with EXPLORE participants; PC3: perceived group identity with EXPLORE; PC4: trust of other EXPLORE participants; PC5: perception that EXPLORE participants are cunning; and PC6: feeling detached from EXPLORE. Social comfort with other EXPLORE participants (OR = 1.24; p = 0.013) and trust of other EXPLORE participants (OR = 1.44; p=0.003) was significantly associated with a higher odds of having sex with another EXPLORE participant. Feeling detached from EXPLORE (OR = 0.56; p=0.020) was significantly associated with a lower odds of engaging in high-risk sexual behaviour with other EXPLORE participants. Regarding intentions to engage in high-risk sexual behaviour with other EXPLORE participants, social comfort with EXPLORE participants (OR = 1.39; p<0.001) and trust of other EXPLORE participants (OR = 1.30; p<0.027) were significantly associated with higher odds of this outcome and the perception that EXPLORE participants are cunning (OR = 0.66; p<0.004) and feeling detached from EXPLORE (OR = 0.68; p<0.007) were significantly associated with lower odds of this outcome. Final models controlled for potential confounders found to be statistically significant in the bivariate analyses. These findings suggest that large-scale studies such as EXPLORE may result in participant's perceptions about group membership, identity and norms, and that these perceptions can influence study outcomes.