Because of its vigorous growth, poplar may be a good candidate for sustainable biomass production to cover renewable energy needs. Due to the wide diversity available within the genus, a screening of the genotypes to be used is necessary. However, selection of suitable genotypes is a time-demanding process and it has to base on relevant traits. For this purpose, the relevance of the traits is mostly estimated through the strength and robustness of their relationships with the target properties, e.g. yield, resistance to biotic and abiotic constraints, efficiency to use water, wood quality, etc. For bioenergy purposes, a high biomass production remains the main selection objective. However, to avoid concurrence for land use with food agriculture, bioenergy tree plantations are likely to be principally installed on marginal areas, where water and nutrients may be scarce. Traits such as the efficiency with which the trees use resources may be of interest in selection programmes. Yield is dependent on the individual growth of the trees which is itself defined by numerous determinants: phenology (e.g. length of growing season), canopy and whole tree architecture (e.g. foliar index, ramification, rooting), leaf function (e.g. photosynthetic assimilation, stomatal behaviour, efficiency to use nutrients and water), or biochemistry (e.g. carbon and nitrogen storage and allocation). During the last decade, numerous studies have aimed at disentangling the relationships among these different categories of traits, at leaf, tree and canopy levels, in the genus under various growth conditions and for diverse plant materials in terms of age and parentage. If some traits have proved their relevance as selection criteria, some others showed erratic relationships with yield as well as with resource-use efficiencies. We propose here a synthesis of the results from 15 studies dealing with productivity determinants. The strength and robustness of the relationships among potential productivity determinants in response to (1) the pedoclimatic context, (2) the age of the plant material, (3) the water availability, (4) the hybrid formula, and (5) the time course, are summarized. The comparison of these studies shows that many traits can not be considered as relevant selection criteria because genetic variation within poplar clones is too narrow and/or because their relationships with wood production is highly dependent on the growth conditions. Basically, traits can be classified in three categories: (1) traits linked to wood production irrespective of growth conditions and plant material (e.g., leaf area, sylleptic branchiness, bud burst), (2) traits for which the relationships with wood production are dependent on environment and/or plant material (e.g., specific leaf area, nitrogen content, water-use efficiency), and (3) traits showing no link with wood production irrespective of growth conditions and plant material (e.g., chlorophyll content, carbon content).