This article deals with the topic of social psychological research methods in practice, by examining how informed consent is gained from research participants. In most research, the consent-gaining process is hidden from analytic scrutiny and is dealt with before data collection has begun. In contrast, conversation analytic research, which records interactional encounters from beginning to end, enables examination of this methodological 'black box'. We explored how 'requests' to consent in research played out across different institutional settings. We found that participants had to 'opt-out' of a research process that was already underway. Consent-gaining sequences constrained opting out in two ways: (1) because research activity was already underway, it must be stopped affirmatively by participants; (2) consent-gaining turns were tilted in favour of continued participation, making opting out a dispreferred response. We also found a mismatch between what ethics guidelines specify about consent-gaining 'in theory' and what actually happens 'in practice'. Finally, we make suggestions about interventions in and recommendations for existing practice to best achieve informed consent.