Electrothermal graphite-furnace atomic-absorption spectroscopy with pyrocoated graphite tubes, integrated platform and matrix modification was used to determine submicromolar concentrations of trace lithium in human red blood cells. Matrix-matched samples were used to establish calibration curves for concentrations up to 0.58 microM (addition-calibration method) with satisfactory linearity (r2 > 0.99) and intra- and inter-day variability (CV < 11.4%). The median concentration of trace lithium in the cells of 40 healthy Caucasian volunteers devoid of medical or psychiatric history was 0.23 microM (inter-quartile range 0.20-0.30). The levels of trace lithium in the red blood cells correlated (r2 = 0.83) with plasma concentrations (median 0.13 microM, inter-quartile range 0.11-0.19) measured in the same blood sample. Dietary factors (e.g. consumption of lithium-containing mineral water) affected both levels. The red blood cell/plasma lithium ratio had a median value of 1.57 (inter-quartile range 1.16-2.07), implying that trace lithium is accumulated in erythrocytes. This contrasts with most reports of red blood cell/plasma ratio, measured during therapeutic treatment with lithium, for which the average value is 0.5-0.8, albeit for much higher concentrations of lithium (approx. 500-800 microM). The proposed analytical method has the required sensitivity and accuracy for determination of trace lithium in red blood cells and makes it possible to perform epidemiological studies to assess human exposure to environmental lithium in diet and beverages, and inter-individual variations in trans-membrane and renal lithium kinetics at the submicromolar level.