This paper examines the effect of learning curve spillovers on market structure and performance. We derive a simple characterization of the “true” marginal cost for a broad class of learning curves, and use calculated examples to show that spillovers substantially undercut the barriers to entry erected by proprietary learning. Unlike the case of cost-reducing research and development, spillovers also tend to improve market performance; the increased efficiency of the industrywide reduction process typically outweighs the decrease in firms' incentives to reduce costs by expanding output.