There are grounds for suspecting that, to varying degrees, smoking, alcohol consumption, oral contraceptive use, vasectomy and induced abortion are markers for high steroid hormone levels. So in epidemiological studies, false inferences may be drawn that these markers (treated as risk factors) have causal or exacerbating effects on diseases which are truly partially caused by high levels of hormones (e.g. probably prostatic cancer and breast cancer). Analogously, such studies of conditions which are truly partially caused by low levels of hormones (e.g. bone fractures, poor sperm quality, and perhaps testicular cancer and rheumatoid arthritis) may yield spurious suggestions of an ameliorative effect. The results of epidemiological studies of the above five "risk factors" for the above six pathologies are-in many cases-in striking disarray. I suggest that this is, at least partially, because of this form of confounding. The point may be tested by contrasting the hormone levels of people who self-select for smoking, vasectomy, etc., at the time that self-selection is made with those of appropriately selected control subjects.