Efficacy of different methods in screening for iron deficiency was re-examined in a randomly selected sample of 38-year-old women (n = 203) with known iron status based on absence/presence of stainable iron in bone-marrow smears. The study was made in 1968-69. Serum ferritin (SF) was determined in 1978 in frozen sera using the Ramco IRMA and, in 1992, samples were re-analysed using a RIA calibrated with the International Standard 80/602 for SF determination. The effect of storage on SF was calculated from a previously established relationship (courtesy of Dr Mark Worwood, Cardiff) between the results obtained with the Ramco assay and assays calibrated with IS 80/602. The distributions in iron replete and iron deficient women showed less overlap (diagnostic efficiency 91%) for SF than for other haematological parameters. The best discrimination was obtained at SF < 16 micrograms/l (specificity 98%; sensitivity 75%). Absence of iron stores was associated with signs of an iron deficient erythropoiesis, starting already at SF 25-40 micrograms/l. Use of multiple criteria to diagnose iron deficiency falsely reduces prevalence figures for iron deficiency.