A survey was conducted on the categories considered relevant by teachers when judging first-year schoolchildren. Six rating scales were derived and applied to 180 English children in a multi-racial urban area, yielding data on perceived differences between ethnic groups and between boys and girls. Changes occurring during the school year were recorded, and the structure of the scales was analysed using `repertory grid' techniques. Results suggest that the rating technique developed gives reliable and sensitive measures, but that teachers' judgements are highly global; two-thirds of the total variance in them was accounted for by a single Principal Component. Sex differences diminished over the first year, but ethnic ones did not, while ethnic differences in a school for educationally subnormal children were the reverse of those found in normal schools.