Salivary steroid immunoassays are widely used in psychoneuroendocrinological studies of menstrual cycle phase, puberty, and menopause. Though manufacturers advertise their assays as suitable, they have not been rigorously validated for these purposes. We collated data from eight menstrual cycle studies across >1,200 women and >9,500 time points. Seven studies collected saliva and one collected serum. All assayed estradiol and progesterone and had an independent measure of cycle phase (LH-surge, menstrual onset). In serum, cycle phase measures strongly predicted steroid concentrations. In saliva, cycle phase poorly predicted estradiol values, which showed an upward bias compared to expectations from serum. For salivary progesterone, predictability from cycle phase was mixed. Widely used enzyme-linked assays performed poorly, while LC-MS/MS performed better. Imputing the population-average serum steroid changes from cycle phase may yield more valid values of hormonal changes for an independent person than directly assessing their hormone levels using salivary immunoassays.