BACKGROUND: Two previous U.S. studies found that although generalized anxiety disorder (GAD) without self-perceived excessive worry was milder than GAD with excessive worry, its persistence, impairment, and risk for subsequent onset of other mental disorders were still substantial. This study examined the implications of relaxing the "excessiveness" criterion on the prevalence and socio-demographic profile of GAD in a Chinese population sample by considering both self and others' perception of excessive worry. METHOD: 2,005 respondents aged 15-65 years participated in a structured telephone interview that covered socio-demographic profile, 12-month DSM-IV diagnosis of GAD, core depressive symptoms, longest duration of worry episode, number of domains of worry, impairment measured by the Sheehan Disability Scale, and treatment-seeking. Excessive worry was assessed from the perception of both respondents and others as reported by respondents. RESULT: The 12-month prevalence of GAD increased from 3.4 to 4% when the excessiveness requirement was relaxed. Excessive GAD and nonexcessive GAD had similar socio-demographic, symptom, chronicity, impairment, depressive symptom, and treatment-seeking profiles. CONCLUSION: GAD without excessive worry was less common than GAD with excessive worry but was likely to be a valid nosological entity. Future iterations of the DSM-IV should clarify whether excessive worry should be retained and, if so, how individuals who only reported excessive worries perceived by others should be optimally assessed.