The purpose of developing new crop technology is generally to increase yields per unit of land and to reduce costs per unit of product. A rather elaborate' methodology has been developed to measure the rates of return from such investments, principally involving ex post analysis. Improvements to this methodology are continually being suggested.1/ Yet much of this sophisticated methodology is built on a rather limited statistical base--particularly when it comes to the actual measurement of the effect of the technology on yields. Some of these yield measurement problems, such as the difficulties in isolating the yield effect of a specific technology when it is part of a package of interacting technologies, have long been recognized. The influence of weather and other factors must, of course, also be considered. Production functions have been an important tool in sorting out the roles of these factors. But relatively little methodological attention seems to have been given to the basic yield data themselves. These are of critical importance in index number analysis. Just how good are the available statistics? What are their advantages? What are their limitations? What alternative sources are available? Could existing data be utilized in a better way? What additional data might be gathered? As one who has given considerable attention over the past decade to the collection of area data on wheat and rice, 2/ I must confess to have largely neglected these questions. Yet to judge from the literature, I am not alone. Perhaps the time has come to balance improvements in methodology with concern for the basic data. The purpose of this paper is simply to highlight the issue and to try to stimulate others to explore the matter further.