An estimated three billion people worldwide, most of them in developing countries, rely on the burning of biomass in order to cook their food. Biomass combustion is associated with a number of negative health and environmental effects, and much effort has been put into designing cookstoves that use less fuel and have fewer toxic emissions in order to reduce the severity of these effects. Laboratory tests are often used in the design process to determine whether a given alteration has improved the design, as a tool to compare different stoves' suitability to a chosen application, and as a means of certifying whether a stove meets a certain benchmark or regulation. In this thesis, I show that a limited set of variables that affects the field performance of a stove and its eventual acceptance by potential users can be tested in the laboratory, and only a subset of these variables is captured by the dominant laboratory protocol used by organizations in the United States. A number of questionable assumptions underlying this protocol are examined, and it is shown that there exists the potential for serious error in its reported values under certain conditions. The key assumptions examined are that the stove being tested in a continuous-feed stove, the fuel supply is homogeneous, all water in the test vaporizes at the local boiling point, the ash content of the fuel is so small as to be unimportant, the procedure of sorting char from unburned fuel is unbiased, and the calorific value of char is one and one-half times that of raw fuel. In the case of a high ash content fuel, such as dung or agricultural residues, the failure to properly account for ash may lead to a relative error in reported values of emissions, fuel consumption, and energy efficiency of greater than ten percent. The assumption of all water vaporizing at the local boiling point, although physically incorrect, is shown to lead to relative errors in test results on the order of one percent, and is thus judged to be an acceptable assumption under ordinary circumstances. Test modifications are shown that effectively address ash content and broaden the test's applicability to batch-feed stoves. The sort-and-mass procedure used for energy accounting in the test is shown to consistently bias results in the direction of overestimating efficiencies and underestimating fuel use.