This paper presents the first results of a comprehensive application of the sensitivity theory developed for the FORSS code system to the analysis of fast reactor integral experiments. A variety of assemblies and performance parameters were studied to determine the nuclear data sensitivity as a function of nuclide, reaction type, and energy. Comprehensive libraries of energy-dependent sensitivity coefficients were developed in a computer retrievable format for several critical assemblies. Uncertainties induced by nuclear data were quantified using preliminary energy-dependent relative covariance matrices evaluated with ENDF/B-IV cross sections and processed for 238U(n,f), 238U(n,γ), 239Pu(n,f), 239Pu(n,γ), and . Calculational results, cross-section covariances, and integral results and their covariances were used in a consistent fashion to improve uncertainty estimates of fast reactor core performance. A first attempt was made to quantify specifications for new cross-section measurements required to satisfy specific design goals at minimum experimental cost. An analysis of several critical experiments indicated that design accuracy goals of 0.5% in k and 2% in the central 238U capture: 239Pu fission ratio (28c/49f) ratio in mixed oxide liquid-metal fast breeder reactor cores are unlikely to be attained in the near future. This assumes that the nuclear data are based only on microscopic measurements, and the current cross-section measurement program is not changed dramatically. Current estimates are 2.3% in k and 7.3% in central reaction ratio using only differential covariance information. Using the measurements in ZPR-6/7 for k and central 28c/49f in a cross-section adjustment scheme with assigned uncorrected standard deviations of 1 and 2%, respectively, standard deviations of the same parameters were computed to be 0.7 and 1.8%. Results of integral experiments, therefore, are needed to improve uncertainty estimates of reactor performance for current fast reactor design work.