We report an investigation of the validity of teachers' ratings of children's progress in "phonics" as a screener for dyslexia. Seventy-three 6-year-olds from a whole school population were identified as "at risk" of dyslexia according to teacher judgements of slow progression through phonic phases. Six months later, children's attainments in literacy and phonological skills were compared with those of their typically developing peers matched on age and gender. Teacher assessments of risk were related to individual differences in performance on a standardised test of reading ability. Teacher assessments overestimated "risk of dyslexia", defined as below-average reading performance. However, teacher judgements, supplemented by tests of phoneme awareness and rapid naming, allowed a sensitive and specific identification of children who subsequently experienced reading difficulties. These findings show teachers can identify risk of dyslexia; the accuracy of this process can be improved by administering two tests of phonological skills.