BACKGROUND More boys than girls receive provision to meet special educational needs. It has been suggested that teachers' evaluations of children's difficulties in school may be subject to gender bias. But the evidence is inconsistent, and the methodology of some work that reported bias has been criticised because the type of rating task that was used may have encouraged stereotypic thinking (Langfeldt, 1993). OBJECTIVE This study investigated whether gender of child would still have an influence on teacher's judgments if a fuller context was provided for the stimulus and there was a more realistic rating task. METHODS The sample comprised 523 teachers in 79 primary schools in London, the home counties and the North-West of England. METHODS Participants completed questionnaires on what action might be required in their school for children with difficulties who were described in short vignettes. The children's gender was varied systematically. The construct of 'seriousness', which had generally been left vague in earlier research, was given a concrete definition. RESULTS It was found that, when the experimental task was contextualised in this way, the gender-of-child effect disappeared. Teacher characteristics such as gender did not influence the results. CONCLUSIONS These findings should be treated with some caution as they relate to the small and restricted range of types of difficulty that were included in the study and to a sample of primary schools in one society. However, they give support to an emphasis on 'realism' in the method of investigation that is used for exploring teachers' judgments.