Psychiatric diagnoses currently rely on a patient’s presenting symptoms or signs, lacking much-needed theory-based biomarkers. Our neuropsychological theory of anxiety, recently supported by human imaging, is founded on a longstanding, reliable, rodent ‘theta’ brain rhythm model of human clinical anxiolytic drug action. We have now developed a human scalp EEG homolog—goal-conflict-specific rhythmicity (GCSR), i.e., EEG rhythmicity specific to a balanced conflict between goals (e.g., approach-avoidance). Critically, GCSR is consistently reduced by different classes of anxiolytic drug and correlates with clinically-relevant trait anxiety scores (STAI-T). Here we show elevated GCSR in student volunteers divided, after testing, on their STAI-T scores into low, medium, and high (typical of clinical anxiety) groups. We then tested anxiety disorder patients (meeting diagnostic criteria) and similar controls recruited separately from the community. The patient group had higher average GCSR than their controls—with a mixture of high and low GCSR that varied with, but cut across, conventional disorder diagnosis. Consequently, GCSR scores should provide the first theoretically-based biomarker that could help diagnose, and so redefine, a psychiatric disorder.