A series of glasshouse, nursery and field experiments were undertaken in Kenya, Senegal, Costa Rica and Scotland in order to screen and select effective rhizobia and arbuscular mycorrhiza (AM) to be used as inoculants for provenances of C. calothyrsus Meisn. in a variety of tropical field environments. The experiment encompassed (i) enumeration of indigenous rhizobia nodulating C. calothyrsus in two Kenyon soils, (ii) screening rhizobia and AM in axenic glasshouse conditions in Kenya, Scotland and Senegal, (iii) screening effective rhizobia and AM in unsterile substrates or soils in glasshouse and nursery conditions in Kenya, Senegal and Costa Rica, (iv) verification of N2-fixation potential in a field condition in Senegal. Estimates of indigenous rhizobial populations using the most probable number (MPN) plant infection technique differed significantly between the two soils and ranged from 32 to 4,375 cells g-1 soil. The most effective strains of rhizobia (CCK13, KCC6, KCC17, KCC39 and KWN35 from Kenya; CCCR1 and CCCR15 from Costa Rica; CCN26 from New Caledonia, and CCC22 from Cameroon) and AM isolates (Gigaspora albida [types GA1b and GA2] and Scutellospora verrucosa [SV2c] from Kenya, S. calospora [SC2] from Guatemala, and Glomus etunicatum [GE1] from Honduras) identified and selected during screening experiments under axenic conditions were further tested either as single or mixed, and as dual (rhizobia and AM) inoculants in unsterile potted substrates or soils in glasshouse and nursery conditions. Inoculation responses were more pronounced in Senegal than in Kenya, and no response was recorded in Costa Rica. In addition, responses varied with regard to parameter, provenance, inoculant and substrate or soil type. The proportion of N2 fixed by C. calothyrsus, as estimated by the 15N natural abundance method, varied from 0-38.5% with significant differences for both inoculation and provenance factors occurring at 19 months after planting. On the basis of these res