Fifty-two children aged 18 months who had fallen across two weight centile lines were compared to 52 controls. Five years after the initial study, 89% of these children were traced. At follow-up, four of the cases but none of the controls had been placed in care and cases were significantly shorter and lighter than controls. Reanalysis of the early weight data revealed that the screening criterion had been over-inclusive, identifying a majority with only borderline failure to thrive (FTT). Although the mean IQ was lower in the cases than in controls, the difference was not statistically significant. However, a significant association was found within the cases between severity of FTT and IQ. There was no difference in the number of behaviour problems reported. The use of a simple centile shift-based definition of FTT, while proving to be over-sensitive, has identified a subgroup of children who suffered adverse long-term cognitive outcomes.