To examine the possibility that there is an early sensitive period for the effects of malnutrition on cognitive development, three groups of children (N = 197) were recruited from a birth cohort with known growth characteristics in south-west Ethiopia (N = 1563). All had initial weights [ges ]2500g. Early growth falterers dropped in weight below the third centile (z<−1·88) of the NCHS/WHO reference population in the first 4 months. Late growth falterers were children not in the first group whose weights were below the third centile at 10 and 12 months. Controls were a stratified random sample with weights above the third centile throughout the first year. All children were tested blind at 2 years using the Bayley Scales of Infant Development, adapted for use in Ethiopia. Mean (SD) scores on the psychomotor scale were 10·2 (3·7) in the controls, 6·6(4·2) in the early growth falterers, and 8·5(4·3) in the late growth falterers. For the mental scale they were 28·9(5·8), 22·6(6·2), and 26·6(6·1) respectively. Both overall differences were statistically significant at p<·001, and planned comparisons between the control and the combined growth faltering groups, and between the early and later growth faltering groups, showed that each difference was statistically significant for both scales. However, early weight faltering was associated with weight at the time of testing (r = .33), which was associated with scores both on the psychomotor (r = .53) and the mental scale (r = .49). After taking weight at the time of testing into account there was no additional effect attributable to the timing of growth faltering. In this population, therefore, early malnutrition does not have specific adverse effect beyond the contribution that it makes to enduring malnutrition over the first 2 years.