In a recent clinical trial, 248 triple-lumen catheters were removed from patients in an intensive care unit, and their tip and subcutaneous segments were cultured by both the sonication and roll plate methods; for 191 of these catheters, flush cultures of all three catheter lumens were also performed. Previously published quantitative endpoints were used to define significant catheter colonization. By using a composite index as a definition of colonization (any of the seven types of cultures meeting quantitative criteria), sonication of the subcutaneous segment was the most sensitive at detecting colonization (58%), followed by sonication of the catheter tip (53%). Sonication of both the subcutaneous and tip segments was 20% more sensitive than sonication of an adjacent catheter segment by the roll plate method (P < 0.05). The greater sensitivity of the sonication method could be attributed to its greater ability than the roll plate method to detect catheter lumen colonization (82 versus 57%, respectively; P = 0.01). A greater number of positive catheter segment cultures were found for colonized catheters from patients with associated bacteremia than for colonized catheters from patients without bacteremia (57 versus 37%; P = 0.004), making any culture method more likely to identify them. For catheters with significant colonization of only one site, the localization was as follows: 36.7% subcutaneous segment, 36.7% catheter lumen, and 26.6% tip segment. These findings suggest that the current practice of culturing a single segment of a central vascular catheter is inadequate and needs to be reexamined. They further suggest that initial colonization of the catheter lumen and tip segments may be more important than previously thought and may require a change in thinking of strategies designed to prevent catheter infection.