Assessing tardive dyskinesia (TD) has been complicated by the use of different research criteria and rating scales. We studied concordance between two commonly used scales, the Abnormal Involuntary Movement Scale (AIMS) and Extrapyramidal Symptom Rating Scale (ESRS), to study interscale concordance and criteria to define TD. Patients with schizophrenia or schizoaffective disorder (N = 374) were rated at baseline with both scales. Linear and logistic regression models explored relationships between scale ratings and mapped scores for corresponding items. TD was defined as at least mild in > or = 2 anatomical areas, or moderate or greater symptoms in > or = 1 area at baseline. Logistic regression was used to find simplified criteria for predicting AIMS-defined TD by ESRS scores. There was a strong association on corresponding item ratings. "Mild" was defined as AIMS score of 2 and ESRS 2 or 3, and "moderate or greater" as AIMS score > or = 3 and ESRS > or = 4. Using these criteria, there was 96.0% (359/374) agreement between AIMS- and ESRS-defined TD cases. The ESRS Clinical Global Impressions of severity of dyskinesia (CGI-SD) best predicted AIMS-defined TD. An ESRS CGI-SD > or = 4 (95% CI: 3.61, 4.76) was associated with > or = 95% probability of AIMS-defined TD. High concordance between the scales for dyskinesia scores was found. Findings suggest that the ESRS CGI-SD score can serve as a simplified criterion for identifying AIMS-defined TD, and may be a useful tool for future research-based TD analyses, when occurring in the context of a full movement disorder assessment.