We present results of a study of changes in residential natural-gas consumption in the St. Louis region. The Princeton Scorekeeping Method (PRISM) is used for measuring changes in weather-adjusted consumption and for isolating the components of conservation. PRISM also provides a basis for separating changes in the thermal integrity of the housing stock from changes in average indoor temperatures. We decompose savings further by considering demographic variables, fuel substitution, and the composition of internal heat gains. The results show that a substantial fraction of savings can be attributed to changes in weather-independent consumption (30–40% of total savings for St. Louis since 1971), but a considerable fraction of this may be caused by changes in household size. Savings associated with structural improvements to the housing stock, which generally have long-term effects on energy consumption, represent the principal component of savings for St. Louis (about 53%). Savings from behavioral changes that involve lower indoor temperatures and possibly decreased comfort may be readily implemented by households as a first reaction to fuel shortages or sudden price increases, but at least in some areas of the country, they have had little, if any, lasting effect. Over the long term, the principal cause of conservation may be increases in energy prices. For St. Louis, 85% of the change in natural-gas consumption since 1971 can be explained by variations in price alone. The remaining unexplained variation may be correlated with other factors such as changes in energy awareness. The sources of conservation which are identified in analyses of this type are important for developing effective conservation policies.