The idea of payment for ecosystem or environmental services (PES) stems from societal awareness of the importance of these services for human well-being. The PES philosophy argues for the internalization of environmental externalities through the creation of ecosystem service markets, in which these services are commoditized and tradable and flow between sellers and providers. PES is lauded for its potential to both reverse ecosystem degradation and alleviate poverty. Despite this potential, PES faces dilemmas in both theory and practice; trade-off is a typical one. This paper analyzed four trade-offs confronted by PES programs. The first trade-off arises from the complex interdependencies among different ecosystem services. Current ecological knowledge is insufficient to characterize accurately the ecosystem services that underpin most PES schemes. In particular, incentives that encourage production of one service may have adverse effects on others. The second trade-off, between transaction costs and monitoring costs, is due to uncertainty about non-marginal changes in the provision of ecosystem services. Uncertainty can be reduced only by gathering further information, which in turn raises transaction costs and leads to an unworkable market mechanism. The third trade-off occurs between equity and efficiency. Equity means that income from PES is distributed evenly while efficiency demands that income be distributed in favor of those better able to provide ecosystem services, via analysis of benefits and costs. The latter is better adapted to market mechanisms, but may result in unfeasibility or unsustainability in scheme implementation. The last trade-off, between the provision of ecosystem services and poverty alleviation, occurs because allocation of more budgets on poverty alleviation reduces budget on ecosystem service, thus reduces the effectiveness and efficiency of PES programs. In the long term, poverty may remain a serious trap for ecological degradation, making it difficult for PES schemes to realize their objectives. Consequently, we conclude that trade-offs in PES are rooted in uncertainty about ecosystem service provision and their process mechanisms, the complex relationships in coupled socio-ecological systems, and heterogeneous implementation contexts. Our analysis suggests that PES should improve both in theory and in practice. More knowledge is needed about the relationships between specific variables and whole systems, and the design of PES schemes should be adapted to local conditions. Furthermore, a comprehensive evaluation system is indispensable for dynamically adjusting PES schemes during their design and implementation.