ABSTRACT The current EPA protocol for testing the effectiveness of dispersants in the laboratory—the Swirling Flask Test—is fraught with difficulties because it has been found to give widely varying results in the hands of different testing laboratories. The objective of this study is to develop an understanding of the factors most responsible for this lack of reproducibility. Once that objective has been accomplished, the protocol will be modified to achieve better reproducibility in the hands of multiple laboratories. To accomplish the objective, fractional factorial experiments were conducted to determine which factors have the greatest impact on the variance of the method. Such factors as oil type, mixing speed, settling time, dispersant type, flask type, and operator were evaluated systematically and in a statistically rigorous manner. In addition, since the calibration curve used for analyzing the oil concentration dispersed in the seawater solution may also be a major source of variability, a systematic evaluation of the way such measurements are made was conducted (such as comparing various types of instruments and evaluating the variance due to operator vs. that due to the instrument per se). Instruments evaluated were the spectrophotometer, diode array detector, and gas Chromatograph with a flame ionization detector. Results from these experiments to date have indicated that the factors most responsible for the lack of reproducibility of the method are mixing speed, settling time, and the type of flask in which the test is conducted. This has led to a redesign of the flask used in the test, which is characterized as being baffled and having a stopcock located at the bottom of the flask to prevent remixing of the dispersed oil mixture when transferring to the extraction vessel. The paper discusses the design of the new flask and the important experiments leading to the refinement of the protocol to reflect a more accurate simulation of mixing and dispersion in the open sea.