During a 4-year routine screening programme for cystic fibrosis (CF) 15 464 specimens were examined for raised meconium albumin levels by a test strip method and by electroimmunoassay. The incidence of false-positive results was about 5 per 1000 specimens in either test. This could be reduced by 90% by determining the ratio of albumin : alpha-1-trypsin inhibitor (a ratio below 2.0 being considered as a negative result), and it could be reduced to zero by determining the ratio in subsequent faecal specimens. Three of 12 meconium specimens from infants with proved CF gave false-negative results in all 3 tests. The other 9 specimens had greater than 100 mg albumin/g dry weight and albumin: alpha-1-trypsin inhibitor ratios of greater than 3.0; in subsequent faecal specimens the ratios were over 4.0. 176 meconium specimens from elsewhere in the UK were examined and these included 23 from infants who were subsequently proved to have CF. Six of these 23 CF specimens gave false-negative results, the other 17 being strongly positive. The origins of meconium serum protein suggest that infants with CF in whom meconium gives false-negative results have normal pancreatic functions at birth. The specificity of current meconium tests therefore cannot be improved as they depend on pancreatic dysfunction.