Abstract Background Persistent negative symptoms are an important area of clinical research. However, there is limited consensus on how to define positive and negative symptom severity thresholds for inclusion in clinical trials. From the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) dataset we evaluated the performance of varying baseline negative and positive symptom thresholds on baseline symptom severity, change from baseline, and negative symptom variance attributable to positive symptom change. Methods In CATIE, we first identified the group of subjects lacking an exacerbation prior to study enrollment. Six subsets with varying baseline positive and negative symptom thresholds were then identified. Baseline characteristics were summarized; negative and positive symptom change from baseline through month 3 was calculated both as crude change and change adjusted for corresponding baseline scores. Sensitivity analyses and correlations were calculated to assess the extent to which negative symptom variance was attributable to positive symptom change. Results More restrictive baseline symptom severity thresholds yielded a considerably smaller sample size and higher negative and lower positive symptoms at baseline. Unadjusted negative symptom change was greater with more restrictive criteria; when adjusted for baseline severity the magnitude of change was comparable across subsets. The amount of variance in negative symptom change attributed to positive symptom change was also comparable across subsets. Conclusions The use of restrictive positive and negative symptom thresholds yields a marked decrease in eligible sample size with no clear improvement in adjusted negative symptom change or amount of variance associated with positive symptom change.