We evaluated the initial assessments of the 380 diabetic patients with and without polyneuropathy in the Rochester Diabetic Neuropathy Study for (1) associations among neuropathy test results, (2) usefulness of different tests for diagnosing and staging polyneuropathy, (3) appropriateness of different minimal criteria for the diagnosis of polyneuropathy, and (4) significant differences in test results with increasing stage of polyneuropathy. Nerve conduction ([NC]; abnormality in two or more nerves) and quantitative autonomic examination ([QAE]; decreased heart-beat response to deep breathing [DB] or the Valsalva maneuver [VAL]) were the most sensitive and objective and were especially suitable for detection of subclinical neuropathy. We propose the following minimal criteria for the diagnosis of diabetic polyneuropathy: greater than or equal to 2 abnormal evaluations (from among neuropathic symptoms, neuropathic deficits, NC, quantitative sensory examination [QSE], and QAE) with one of the two being abnormality of NC or QAE (DB or VAL). Neuropathy Symptom Score, Neuropathy Disability Score, QSE (vibratory or cooling detection threshold), and summated compound muscle action potential of ulnar, peroneal, and tibial nerves were best for judging severity. Inability to walk on heels provided a discrete separation of diabetic patients into those with mild and those with more severe neuropathy--a separation helpful in staging.