Agronomy is a scientific discipline whose main object of study is the cultivated field. It usually includes the biophysical functioning of the field seen as an agrosystem and the reasoning of actions applied on this agrosystem by the farmer. This second aspect leads to an up-scaling process towards wider levels of organization such as farm and landscape, through farming system research and land-use management. In the humid tropics, agroforestry systems (AFS) are the result of farmers¿ actions and can be considered as cultivated fields characterized both by the combination of biophysical interactions and farmer practices. To study these complex systems, concepts and tools of agronomy should be applicable, but the majority of them were constructed from and applied to temperate agriculture. Moreover temperate agriculture is based on the rotation of a reduced number of annual monocrops. As a result, most of the concepts and tools of agronomy aimed at maximizing the productivity of simplified systems, nowadays in a sustainable way. But AFS have opposite characteristics, that is, i) the presence of perennial and ligneous species, ii) a high number of cultivated and/or harvested crops, iii) a high specific botanical richness, iv) a high structural heterogeneity, and iv) a significant renewal of species on a long time scale. These characteristics generate a more complex functioning, based on emergent properties, and various ecological and social services. In this context, are concepts and tools of agronomy applicable to AFS? What are the limits, interests, and necessary adaptations to use them in order to understand and optimize AFS? This paper analyses such limits and interests and offers directions for improvement. The concept of cropping systems, the basis of agronomy, is analysed for that purpose, from case studies on cocoa and coffee agroforests in Africa and Central America. (Texte intégral)