Chest ultrasonography (CU) has been advocated as an effective tool for diagnosis and localization of pleural fluid. Studies to date supporting the technique have been anecdotal and nonrandomized. To determine if CU was beneficial when thoracentesis was performed by clinicians or house staff, we evaluated prospectively 205 patients presenting with pleural effusion at 2 community teaching hospitals. Decubitus roentgenograms were obtained on all patients, but CU with targeting by skin marker was performed on a randomized basis. Results were evaluated as to (1) whether the quantity of fluid obtained was sufficient for the intent of the procedure, (2) the number of needle insertions required to obtain the fluid, and (3) the incidence of complications such as pneumothorax. One hundred three effusions were evaluated by CU and 102 by roentgenography alone. The effusions in each group were stratified as small (obliteration of less than half of the hemidiaphragm on roentgenogram) or large. Small effusions were further stratified as free flowing or loculated (no layering of fluid on decubitus roentgenograms). By chi-square test, CU was significantly superior to decubitus roentgenograms alone for obtaining adequate fluid samples in small effusions (p less than 0.01). This was true regardless of whether the effusion was loculated (p less than 0.02) or free flowing (p less than 0.05). The technique had no such advantage in large effusions. We did not find that CU significantly reduced the need for multiple attempts nor incidence of complications in any group.