In 1978 and 1979 a large number of US children were fed chloride-deficient soy-based infant formula. A representative sample of such children was identified in a southern county through a mailing to the homes of 3639 first- and second-grade children in the public schools. Of the 2329 (64%) who responded, 56 reported use of deficient formula and were invited to have developmental testing by one of four study psychologists at their school. Of the 310 users of other soy formulas, 112 were selected for testing as matched controls on the basis of their sex, feeding history, age, birth weight, and socioeconomic status (as indicated by school attended). After exclusions and refusals, 42 children who used deficient formula and 66 control children were tested using the McCarthy Scales of Children's Abilities. Examiners were unaware of the child's history of formula use. The mean General Cognitive Index was 102.8 in those using deficient formula and 105.4 in controls. After adjustment for demographic differences the children who used chloride-deficient formula were found to average 4.9 points less than the controls (P = .04, 1-tailed). The largest difference was in the Quantitative subscale (P = .005). These data show a statistically significant although small effect of chloride-deficient formula on the long-term developmental outcome of exposed children; however, further study of these results is needed for full confirmation.