The Positive and Negative Syndrome Scale (PANSS) is widely used as a method for the assessment of symptoms of schizophrenia but the most complete model of how symptoms are structured has not been determined. Using the methods of confirmatory factor analysis with a large sample of 1,233 of schizophrenic subjects this study examined the goodness of fit of 20 previously proposed models. None of these proposed models met criteria for adequate fit to the empirical data. The sample was then stratified and half of the data was used to calibrate a new model. The model was validated in the second half of the data. The new pentagonal model uses 25 of the 30 items of the PANSS in 5 factors: positive, negative, dysphoric mood, activation, and and autistic preoccupation. Patients who varied widely in age, severity, and chronicity of illness did not differ in their overall symptom structure. The results of this study also implicated some problems in the validity of the PANSS as currently configured when used to assess symptoms of schizophrenia.