This experimental study examined the effects of accounts, ‘offender’ and ‘victim/witness’ sex, and sex-role orientation on rating scales and ‘behaviouroid’ and behavioural measures. Participants who spent up to 45 minutes completing questionnaires that proved unusable due to a confederate's error were exposed to one of four systematically varied accounts or explanations. Account type, participant and confederate sex, and participant sex-role orientation served as predictor variables. In general, concessions and excuses yielded more positive impressions, greater commitment on the part of participant victims to help the offender, and greater help than did refusals, but the effectiveness of accounts also depended on by whom they were proffered. Results demonstrated partial support for politeness theory predictions.