The reliability of the Spanish version of the Social Behaviour Schedule (SBS) was tested in a vocational setting on a sample of 64 subjects with learning disabilities. Test-retest assessment showed a good percentage of agreement (80%) and adequate kappa values for most SBS items. The overall percentage of agreement of inter-rater reliability was 85% and kappa values were moderate to nearly perfect for 52% of items. Inter-informant analyses produced poorer results, with an average agreement of 43% and inadequate kappa values on 42% of items. The intraclass correlation coefficient (ICC) was 0.64 for test-retest, 0.76 for inter-rater assessment and 0.94 for inter-informant assessment. The Spearman correlation coefficient was adequate on the test-retest and inter-rater analyses, but not on inter-informant analysis. This low inter-informant agreement could be attributed to environmental factors which alter the reliability of reports from different informants in community settings with high levels of normalization. In such environments, an interview with a key informant may not suffice, and both a careful review of the clinical record and a direct interview with subjects may enhance the reliability of the information attained.