The objective appearance of negative symptoms in schizophrenia and other psychotic disorders may be a direct reflection of a primary neural abnormality or may be secondary to a variety of factors such as neuroleptic side effects, depression, positive symptoms, or environmental understimulation. Although there is a consensus that it is important to be able to disentangle "primary" versus "secondary" negative symptoms, optimal strategies for doing so remain unclear. Concerns have been raised about making this distinction based on clinical judgment because of potential low reliability in the absence of extensive training and/or highly specialized rating scales. This is particularly important in terms of the application of DSM-IV criteria for schizophrenia, in which negative symptoms play a prominent role. In the context of the DSM-IV schizophrenia field trial project, we examined the reliability of making the primary versus secondary distinction in a multicenter sample of 462 subjects with nonorganic psychotic disorders. Each subject was assessed by two raters, half in an interrater design (i.e., conjoint interviews) and half in a test-retest design (i.e., independent interviews by two raters conducted 1 day apart). All raters used the same semistructured interview instrument, which included an abbreviated version of the Scale for the Assessment of Negative Symptoms (SANS). In addition to the usual SANS ratings, raters were asked to indicate their judgment as to whether the symptom was primary, secondary, or unknown (inadequate information to assess). No formal training was provided. Reliability, as quantified by kapp, indicated only a fair degree of agreement ranging from 0.48 to 0.68 for interrater reliability (median, 0.50) and 0.34 to 0.66 for test-retest reliability (median, 0.38). Negative symptoms were rated as primary approximately twice as often as secondary, and raters believed they had adequate information to make this distinction based only on cross-sectional evaluation in all but 10% of the cases. These data suggest that the primary versus secondary distinction should not be incorporated into the application of operationalized diagnostic criteria. Implications are discussed in terms of balancing reliability and validity in the assessment of negative symptoms.