Twenty women, who had been randomly selected from women with subclinical hypothyroidism identified in a population study were treated with L-thyroxine and placebo in a double-blind cross-over design during 2 x 6 months. Three women did not complete the study, one because she moved to another part of the country, and two because of nervousness and sense of tachycardia. None of these 'drop-outs' had any objective signs of overtreatment; they had normal pulse rate and a serum T3 concentration within the reference interval. During L-thyroxine treatment serum procollagen-III-peptide concentration increased in 13 women out of the 17 women completing the study and at the end of treatment the mean concentration was significantly raised (P less than 0.001). Serum concentrations of procollagen-III-peptide then correlated with those of free thyroxine (P less than 0.01), total thyroxine (P less than 0.05), and reverse triiodothyronine (P less than 0.05). The same comparison revealed little or no effect on the concentrations of serum creatine kinase activity, transcortin or sex-hormone binding globulin. Heart rate-corrected preejection period and symptom score decreased (P less than 0.05). Four women starting with L-thyroxine showed a marked and prolonged (4-6 months) rise in thyrotrophin concentration during the subsequent placebo period, but remained clinically euthyroid. Four women (of 17) improved during therapy as judged by psychometric testing and their own rating. We could not by pretreatment observations identify these four women apart from serum free and total 3,5,3'-triiodothyronine concentrations in the lower part of the health-associated reference interval. Subclinical hypothyroidism is common among middle-aged and old women, and our findings indicate that approximately one woman in four with this 'subclinical' condition will benefit from L-thyroxine treatment.