We have studied the metabolism of compounds containing 31P in normal breast using magnetic resonance spectroscopy (MRS). Spectra were acquired from non-lactating pre-menopausal breast (n = 14 women), lactating breast (n = 8) and post-menopausal breast (n = 8). The standard acquisition protocol used a 5.5 cm surface coil with the volunteer prone to minimize chest wall signal contamination. In pre-menopausal non-lactating women the phosphocreatine (PCr) peak area, expressed relative to the sum of all 31P peak areas, was negatively correlated with breast size (r = -0.56, p = 0.02) suggesting that much of the PCr signal originated from the chest wall. The phosphodiester (PDE) relative peak area was positively correlated with breast size (r = 0.71; p = 0.002). Spectra could be acquired at all phases of the menstrual cycle. In sequential examinations of five women not taking the oral contraceptive pill (OCP), phosphomonoester (PME) relative peak area was significantly lower on Week 2 than other weeks of the cycle (p = 0.03). Among pre-menopausal women no clear difference was apparent between the spectra from women taking the OCP and those not taking the OCP. Lactating breast had significantly higher PME relative peak area than non-lactating pre-menopausal breast (p = 0.02), probably reflecting the higher proportion of epithelial tissue in lactation; the lower PCr relative peak area in lactating breast (p = 0.05) is probably due to the greater size of the breast during lactation. Spectra were acquired from post-menopausal women but with a relatively low signal-to-noise ratio. The only significant difference between 31P relative peak areas of breast spectra acquired from pre- and post-menopausal women was that less PCr was detected in the post-menopausal volunteers (p = 0.03), probably as a result of differences in breast size.