RANDOM population sampling is rarely achieved in human research. Rosenthal and Rosnow1 have concluded that human behavioural science is largely the science of punctual college sophomore volunteers. Volunteers typically have more education, higher occupational status, earlier birth position, lower chronological age, higher need for approval and lower authoritarianism than nonvolunteers2–9. We report evidence here to suggest that the volunteering behaviour of college women varies with the stage of their menstrual cycle. This implies that many experiments which report perceptual, cognitive or physiological response differences between the sexes may be biassed in systematic fashion by sampling error.