This paper shows that the presence of conditional staging in R&D (Research & Development) has a critical impact on portfolio risk, and changes diversification arguments when a portfolio is constructed. When R&D projects exhibit option-like characteristics, correlation between projects plays a more complicated role than traditional portfolio diversification would suggest. Real option theory argues that research projects with conditional phases have option-like risk and return properties, and are different from unconditional projects. We show that although the risk of a portfolio always depends on the correlation between projects, a portfolio of conditional R&D projects with real option characteristics has fundamentally different risk than a portfolio of unconditional projects. When conditional R&D projects are negatively correlated, portfolio risk is hardly reduced by diversification. When projects are positively correlated, however, diversification is more effective than these tools predict.