We compare non-blind teacher assessments with blind national test scores in maths to examine teacher-test score disparities by children's height and weight. Relative to test scores, shorter and heavier children are rated less favourably by teachers. This teacher-test score discrepancy cannot be explained by the child's behaviours, motivation to learn or cognitive ability. Unobserved student fixed effects across subjects explain the teacher-test score discrepancy by height, but not weight. Our analysis points to biased teacher assessments as the most plausible explanation for the remaining teacher-test score gap by weight. We find harsher teacher assessments are associated with a reduction in both the child's future test performance and liking for maths 4 years later.