Environmental policies are characterized by a growing emphasis on participation, devolution and negotiated decision making. Increasingly, centralized top down decision making systems are being replaced by new forms of local governance. In their strongest versions, these involve delegation of formal authority to local stakeholders who are expected to decide collectively upon the management rules of local common-pool resources. Devolution is particularly important in relation to the allocation and management of scarce water resources. Indeed the French water law of 1992 institutionalised the notion of devolution by requiring that water management rules be negotiated at the river basin level between all stakeholders. Although stakeholder negotiation is becoming increasingly important, relatively little is understood about the influence of the structure of negotiations on their outcomes. There is therefore a rising demand for applied simulation models which can be used as 'negotiation-support tools' to address such questions as: what criteria should determine participation rules ? What kind of biases result from power asymmetries ? In this paper, we apply the Rausser-Simon (RS) framework to a specific negotiation process, involving water use and storage capacity. The RS framework is a computable non cooperative bargaining model designed to study complex real world multi-agent, multi-issue negotiation problems. The elements of the modelled negotiation structure include: the list of participants at the bargaining table; the set of issues being negotiated; the decision rule; political weights ("access"); and the nature of the outcome if agreement cannot be reached (the fall-back option). The application we consider focuses on the upper part of the Adour Basin in south-western France. Increased irrigation use in recent years has led to recurrent water shortages, resulting in increased tensions among water users and potential damages to aquatic life. Public authorities have offered to finance new water supply facilities provided that stakeholders can agree on management issues such as quota allocations, water prices and dam configuration. A number of issues arise concerning the relationship between the structure of the negotiating process and the effectiveness with which participating stakeholders can pursue their individual interests. The case study is modelled with seven aggregate players (three aggregate farmers representing three sub-basins - upstream, midstream and downstream - two environmental lobbies, the taxpayer and the water manager), and with up to nine negotiated variables (water quotas and water prices in each sub- basin, and the capacities of three dams). Farmers' utility functions are estimated using hydraulic data and calibrated linear programming farm models. The richness of the data and institutional information available to us provides a realistic environment in which to examine the effect of negotiation structure on participant power. We focus in particular on the three farmer stakeholder group, whose interests are aligned but distinct, and so form a natural negotiating coalition. We construct experiments that enable us to evaluate the effects of negotiation structure on the effectiveness of this coalition. Our comparative statics experiments highlight a number of aspects of the relationship between negotiation structure and bargaining power. In addition to the standard indices of bargaining power - the distribution of access and "default strength" (i.e., stakeholders' relative tolerance for the outcome if negotiations break down) - our analysis identities a number of other less obvious sources of power. A coalition member can advance his own interests by ceding his own political access to another coalition member who has a "better" strategic location. It is also shown that while the interests of the coalition as a whole will be advanced if its members cede access to a "spokesman" representing their common interests, some coalition members may be adversely affected. Finally, we consider the effect on the coalition of restricting the set of proposals that may be placed on the bargaining table in order to mitigate "beggar thy neighbour" behaviour by coalition members. This improves all farmers' utilities, except in some specific cases in which external constraints initially oblige players to act as if they were playing cooperatively. In conclusion, the value of this model is to be used by the negotiation stakeholders to explore the implications of alternative specifications of 2 the bargaining environment. In particular, the simulations reveal that the ways in which the negotiation structure interact with stakeholder bargaining power are complex and lead to non intuitive outcomes.