To determine the accuracy with which spirometric measurements of FVC and expiratory flow rates can diagnose the presence of a restrictive impairment. The pulmonary function tests of 1,831 consecutive white adult patients who had undergone both spirometry and lung volume measurements on the same visit over a 2-year period were analyzed. The probability of restrictive pulmonary impairment, defined as a reduced total lung capacity (TLC) below the lower limit of the 95% confidence interval, was determined for each of several categoric classifications of the spirometric data, and additionally for each of several interval levels of the FVC and the FEV1/FVC ratio. A large clinical laboratory in a university teaching hospital using quality-assured and standardized spirometry and lung volume measurement techniques according to American Thoracic Society standards. Two hundred twenty-five of 1,831 patients (12.3%) had a restrictive defect. The positive predictive value of spirometry for predicting restriction was relatively low; of 470 patients with a low FVC on spirometry, only 41% had restriction confirmed on lung volume measurements. When the analysis was confined to the 264 patients with a restrictive pattern on spirometry (ie, low FVC and normal or above normal FEV1/FVC ratio), the positive predictive value was 58%. Conversely, spirometry had a very favorable negative predictive value; only 2.4% of patients (32 of 1,361) with a normal vital capacity (VC) on spirometry had a restrictive defect by TLC measurement. The probability of a restrictive defect was directly and linearly related to the degree of reduction of FVC when the FVC was < 80% of predicted (p = 6.002). Combining the FVC and the FEV1/FVC ratio improved the predictive ability of spirometry; for all values of FVC < 80% of the predicted amount, the likelihood of restrictive disease increased as the FEV1/FVC ratio increased. Spirometry is very useful at excluding a restrictive defect. When the VC is within the normal range, the probability of a restrictive defect is < 3%, and unless restrictive lung disease is suspected a priori, measurement of lung volumes can be avoided. However, spirometry is not able to accurately predict lung restriction; < 60% of patients with a classical spirometric restrictive pattern had pulmonary restriction confirmed on lung volume measurements. For these patients, measurement of the TLC is needed to confirm a true restrictive defect.