We study mechanisms for selecting up to m out of n projects. Project managers' private information on quality is elicited through transfers. Under limited liability, the optimal mechanism selects projects that maximize some function of the project's observable and reported characteristics. When all reported qualities exceed their own project-specific thresholds, the selected set only depends on observable characteristics, not reported qualities. Each threshold is related to (i) the outside option level at which the cost and benefit of eliciting information on the project cancel out and (ii) the optimal value of selecting one among infinitely many ex ante identical projects. (JEL D21, D82, O32) M any individuals, firms, and government agencies face situations in which they need to choose between a number of projects. Often, when making this choice the decision maker (from now on "the firm") is not fully informed and needs to rely on better informed agents whose interests are not aligned with the firm's interests. Our paper studies the firm's project selection problem under asymmetric information. For example, a firm, or a government agency, deciding which R&D projects to pursue, from projects proposed by managers working for the firm; or a corporate board deciding which capital investment project to finance.Projects can be risky, the likelihood of success (or the net expected return) might not be known to the decision maker, who is not sufficiently familiar with the technical details. Project managers hold private information about the quality of their projects, and typically prefer that their own project be selected. Hence, their interests are not perfectly aligned with those of the decision maker, and an agency problem arises. As Paul Sharpe (vice president at SmithKline Beecham) and Tom Keelin described:Major resource-allocation decisions are never easy. For a pharmaceuticals company like SB [SmithKline Beecham], the problem is this: How do you make good decisions in a high-risk, technically complex business when the information you need to make those decisions comes largely ). We are indebted to Dirk Bergemann for his guidance and invaluable advice. We are grateful to