The measurement of agricultural productivity is important in understanding growth in agriculture and in assessing competitiveness. In this paper, some difficulties related to the empirical measurement of productivity are analyzed using a Canadian case study. The paper focuses in particular on the choice of index number procedures, comparing traditional fixed base weight indexes with flexible or superlative indexes such as the Divisia and Fisher. Indexes of aggregate agricultural output, total farm input use, and total factor productivity are estimated for Canada and for the prairie region of western Canada from 1948 to 1991. Alternative productivity growth rates are reported and compared. The productivity results based on the Tornqvist-Theil approximation to the Divisia index and the chained Fisher index are very similar. Both these flexible weight index procedures are to be preferred over the Laspeyres, the most commonly used approach in Canada.