Improved rubber agroforestry systems (RAS) have been introduced and tested since 1994 though a collaborative research program with ICRAF and CIRAD. After completing an on-farm trial network with 100 fields in 3 areas in Sumatra and Kalimantan, farmers have been monitored for 15 years in order to assess impact of technological change and context evolution. The 1994 2007 period has been characterized by completely different contexts: good rubber prices (1994 1996, severe economic and political crises combined with low rubber world prices (1997 2001) and recovering with high rubber prices (2002 2007). Meanwhile, farmers organizations have been allowed, oil palm took has been added to the landscape, tree tenure has been modified allowing farmers to take advantage of tree biodiversity. When rubber could have been condemned at the end of the 1990s compared to oil palm, rubber and rubber agroforestry systems not only survived but took a far greater place in local farming systems. The hypothesis of complementarity between oil palm and rubber systems has been confirmed; meanwhile ecological services as well as economic advantages of agroforestry practices have been taken into account by institutions and integrated by producers in their farming and livelihood strategies. Positive externalities such as soil fertility maintenance, biodiversity conservation, 'useful biodiversity' valorization (fruits and trees), long-term income stability and risk aversion, can now be taken into account in farming systems' strategies and income evaluation. Fifteen years of research results on RAS have led to significant knowledge on farmers' strategies and innovation processes that can provide useful recommendations for further development. (Texte intégral)