Previous results are contradictory regarding the concentration of thyroxin in human milk. Using a sensitive radioimmunoassay, we have found a lack of parallelism between the standard curve for thyroxin and the curve for serial dilutions of whole human milk, skimmed milk, or ethanol extracts of milk. Nonspecific binding also indicated the presence of analytical artifacts. Thus we have separated thyroxin from other milk components by means of a strongly basic Bio-Rad anion-exchange resin with quaternary ammonium exchange groups attached to a styrene divinyl benzene copolymer lattice, radioimmunoassaying the fractions eluted with an equivolume mixture of acetic acid and water. Parallelism with the standard curve was good, and results were the same whether or not the resin eluate was further purified by paper chromatography. The range of thyroxin concentration in 21 samples of human milk was 0.29-2.00 micrograms/L (mean 0.71, SD 0.40, microgram/L). Such concentrations are unlikely to afford protection to the developing brain of a breast-fed athyreotic baby, as previously claimed.