The accuracy of fluid delivery via gravity-flow i.v. infusion systems in hospitalized patients was evaluated. All adult patients on the medical-surgical wards of a university hospital who were receiving i.v. fluids via gravity-flow infusion sets were studied during a four-day period. Data collected approximately every two hours over a 15-hour period daily included the prescribed i.v. flow rate, type of i.v. set (microdrop or macrodrop), drop rate, and the approximate volume of fluid remaining in the i.v. container. Drop rates were measured with a photocell device placed around the drip chamber of the i.v. set. A total of 509 observations involving 86 patients were recorded during the study; drop rates were evaluated at flow rates for which there were 20 or more observations. For the majority of flow rates and set types, less than 15% of observations were within +/- 10% of desired drop rates, while only 21% of observations fell within +/- 20% of desired drop rates. Mean versus desired volume of fluid delivered between observations differed substantially but not as much as anticipated based on drop rate variability, reflecting nurses' attempt to adjust fluid therapy based on volume of fluid delivered. Intravenous fluid delivery via gravity-flow i.v. infusion systems is highly inaccurate. To ensure appropriate fluid delivery, better monitoring or improvement of i.v. fluid administration systems or the use of electronic infusion control devices is recommended.