Background Large community-based epidemiological surveys have consistently identified high co-morbidity between major depressive episode (MDE) and generalized anxiety disorder (GAD). Some have suggested that this co-morbidity may be artificial and the product of the current diagnostic system. Because of the added direct and indirect costs associated with co-morbidity, it is important to investigate whether methods of diagnostic classification are artificially increasing the level of observed co-morbidity.Method The item response theory (IRT) log-likelihood ratio procedure was used to test for differential item functioning (DIF) of MDE symptoms between respondents with and without a diagnosis of GAD in the 2001-2002 National Epidemiological Survey on Alcohol and Related Conditions (NESARC).Results The presence of GAD significantly increased the chances of reporting any symptom of MDE, with odds ratios ranging from 2.54 to 5.36. However, there was no indication of significant DIF of MDE symptoms in respondents with and without GAD.Conclusions The lack of any significant DIF indicates that cases with GAD do not present with a distinct MDE symptom profile, one that is consistent with the endorsement of symptoms that are conceptually similar in nature between the two disorders, compared to cases without GAD. This does not support the hypothesis that co-morbidity between MDE and GAD is artificially inflated because of the similar symptom criteria required by the current diagnostic system. Instead, MDE and GAD may be thought of as two distinct diagnostic entities that frequently co-occur because of a shared underlying trait.