Making a definite diagnosis of Cushing's syndrome is a challenging problem. Unsuspected Cushing's syndrome occurs in 2-3% of patients with poorly controlled diabetes, 0.5-1% with hypertension, 6-9% with incidental adrenal masses, and 11% with unexplained osteoporosis and vertebral fractures. The increasing recognition of this syndrome highlights the need for a simple, sensitive, and specific diagnostic test. Patients with Cushing's syndrome consistently do not reach a normal nadir of cortisol secretion at night. The measurement of late-night salivary cortisol levels might, therefore, provide a new diagnostic approach for this disorder. Salivary cortisol concentrations reflect those of active free cortisol in plasma and saliva samples can easily be obtained in a nonstressful environment (e.g. at home). Late-night salivary cortisol measurement yields excellent overall diagnostic accuracy for Cushing's syndrome, with a sensitivity of 92-100% and a specificity of 93-100%. Several factors can, however, make interpretation of results difficult; these factors include disturbed sleep-wake cycles, contamination of samples (particularly by topical corticosteroids), and illnesses known to cause physiologic activation of the pituitary-adrenal axis. In this Review, we discuss the methods and value of measuring salivary cortisol for the diagnosis of Cushing's syndrome, and put forward some recommendations to maximize accuracy of results.