This paper relates to the use of the Multi-Program Valuation Method in recreation demand modelling. This method, based on Lancaster's multi-attribute approach, is devoted to the study of substitution/complementarity phenomena in demand for environmental goods. We propose its application to stated-owned French South-West coastal forests, where three adjoining wilderness areas (the ocean, the sand and the forest) form one sole site. Results show that residents' mean Willingness-to-Pay is significantly lower than tourists' one and that they do not valuate the three assets with the same rank. Residents and tourists are willing to pay respectively, without taking into account time cost, 2.56 and 10.66¤/visit/household for forest, 7.81 and 14¤ for the sand and 7.51 and 21.28¤ for the ocean. When they specifically attach importance to one asset, visitors are willing to pay an important premium in addition to this mean value (up to 75.58¤/visit/household for forest for tourists). Furthermore, tourists consider the three assets as complementary in utility whereas residents treat them as substitutes. Evaluating recreation demand while distinguishing both sub-populations is in consequence pertinent. Finally, all these results support the idea that forests should be regarded like significant value-added assets, rather than of first importance in destinations' choices.