To determine what role the technique plays in complications associated with thoracentesis performed by physicians in training, we undertook a prospective study of thoracentesis in the medical service at our institution in which the sampling method was randomized among needle, needle with catheter, and needle with direct sonographic guidance. Fifty-two spontaneously breathing, cooperative patients with free-flowing effusions obliterating more than half of the hemidiaphragm on an upright, posteroanterior chest roentgenogram were randomized. When we analyzed those complications that were potentially life-threatening (eg, pneumothorax) and/or placed patients at increased risk for further morbidity (eg, pneumothorax, dry tap, inadequate tap), the sonography-guided method was associated with significantly fewer serious complications (0 of 19) than the needle-catheter (9 of 18) or needle-only methods (5 of 15). The sonography-guided method was associated with fewer pneumothoraces (0 of 19) than the needle-catheter (7 of 18) or needle-only methods (3 of 15). The difference between needle-catheter and needle-only methods was not significant. From our results, we conclude that the method by which thoracentesis was performed significantly influenced the spectrum and frequency of complications, and the sonography-guided method was the safest.