Abstract
Data from a large field trial were used to test the propositions that the HoNOS is a unidimensional scale and that the original subscale structure fits the data and is valid. Low inter-item correlations suggest that the scale in not unidimensional. Low internal consistencies of the original subscales were also found. Confirmatory factor analysis was used to test the fit of the original subscales, which was found to be poor. An alternative subscale structure is described which fits the data much better, differentiates diagnostic groups better, and accounts for movements between in-patient and out-patient status better. The implications of the findings are discussed.