Technological learning, i.e., cost reductions as technology manufacturers accumulate experience, is increasingly being incorporated in models to assess long-term energy strategies and related greenhouse gas emissions. Most of these applications use learning rates based on studies of non-energy technologies, or sparse results from a few energy studies. This report is a step towards a larger empirical basis for choosing learning rates (or learning rate distributions) of energy conversion technologies for energy models. We assemble data on experience accumulation and cost reductions for a number of energy technologies, estimate learning rates for the resulting 26 data sets, analyze their variability, and evaluate their usefulness for applications in long-term energy models.