OBJECTIVE: To improve the characterisation of chest pain by comparing symptoms in patients with normal and abnormal coronary angiograms. STUDY DESIGN: Prospective case-control study. SETTING: Single tertiary cardiac referral centre. PATIENTS: 65 consecutive patients with chest pain and completely normal coronary angiograms recruited over a period of one year, and 65 sex matched patients with significant stenoses at angiography. MAIN OUTCOME MEASURES: Standardised chest pain questionnaires. RESULTS: 61 of 65 patients (94%) and every control reported chest pain on exertion. There were no important differences in the site, quality, and radiation of pain but three symptoms had discriminatory value expressed in binary fashion ("typical" v "atypical"): the consistency with which pain was reproduced by exercise (typical, score index 10/10), the duration of pain episodes (typical, five minutes), and the frequency of pain at rest (typical, 10% all pain episodes). All three symptoms were atypical in 21 (32%) patients with normal coronary angiograms, but only one patient with an abnormal coronary angiogram. Patients with no typical features had a 2% chance of an abnormal coronary angiogram if aged under 55 years or 12% if aged 55 years or more. The additional impact of exercise stress testing was low. CONCLUSIONS: Chest pain characteristics which separate patients with normal coronary angiograms from patients with obstructive coronary heart disease can be defined objectively. This may allow improvements in referral patterns for specialist opinion or angiography, and in characterisation of patients in research studies.