This paper relates to the use of a multi-attribute extension of dichotomous choice Contingent Valuation called the Multi-Program Contingent Valuation Method. Such a method is devoted to the study of substitution/complementarity relations in the demand for environmental goods and we propose to implement it when recreational uses take place in a multi-dimensional context. A new way to design policies via criteria and pictures is needed therefore. The field study concerns the French South-West coastal forests where various natural assets cohabit. Our results prove that residents' mean Willingness-to-Pay for management policies ranges from 2.50 to 6.65 euros per person per visit depending on where recreation quality is supposed to be maintained. A bivariate probit proved to be relevant, showing that choices are non independent one from another. In a special case, the three natural assets appear to be valuated independently. But, it is still difficult to conclude on a more general basis.