We present a novel argument demonstrating that when trade is characterized by uncertainty the comparative advantages doctrine is misleading and a positive level of diversification is growth enhancing. Applying a result developed in the mathematical biological literature, we show that, in Ricardian trade model in which capital available for investment depends on previous periods returns, incomplete specialization is optimal. We also demonstrate that, in this case, the decentralized solution is characterized by an inefficiently high level of specialization with respect to the social optimal one. Finally, we present a taxation scheme that, reconciling individual incentives and social optimum, is able to induce individual agents to adopt the optimal specialization strategy, i.e. the one that maximizes the country growth rate.