Fifty consecutive panic patients had M-mode echocardiographs read independently by two cardiologists with expertise in echocardiography. In this prospective study, there was poor interrater reliability (22 of 50; K = 0.11) for diagnosis of mitral valve prolapse (MVP). On repeat evaluation 10 months later there was also unacceptable intrarater reliability for each reader: 22 of 35 (K = 0.41) and 22 of 35 (K = 0.45). We conclude that M-mode echocardiography is clinically unreliable for establishing the diagnosis of mitral valve prolapse. These findings suggest that the variable reporting of M-mode-determined mitral valve prolapse in psychiatric populations may reflect differences among echocardiologists rather than differences in cardiac pathology. The clinical implications of these findings are discussed.