BACKGROUND The baseline findings from a controlled study of the effect of a public education campaign on community attitudes to mental illness are presented. METHOD A census of attitudes to mental illness was conducted in two areas, prior to the opening of supported houses for the mentally ill. Factor analysis of the Community Attitudes toward the Mentally Ill (CAMI) inventory revealed three components: Fear and Exclusion, Social Control and Goodwill. RESULTS The only determinant of Fear and Exclusion was having children. The main determinants of Social Control were social class, ethnic origin, age, having suffered mental illness and having children. The main determinant of Goodwill was educational level. The attitude factors were predictive of respondents' behavioural intentions toward the mentally ill. Respondents with children and non-Caucasians were more likely to object to the mentally ill living in their neighbourhood. CONCLUSIONS Any intervention aimed at changing attitudes to mentally ill people in the community should be targeted at people with children and non-Caucasians, as these groups are more likely to object.