We have previously reported significant strain-host incompatibility between the microsymbiont Rhizobium legumillosarum biovar trifolii and Trifolium spp. related to geographic and phenological barriers. Additionally, we have shown that an effective symbiosis between strains of R. l. trifolii and clover was established despite the soil harbouring ineffective R. L trifolii capable of nodulating the host. We termed this "selection" for effective symbiosis. This paper reports glasshouse-based experiments to validate and further explore this in situ selection phenomenon. The effect of cell density and strain ratio at the time of inoculation, as well as soil pH, were investigated on two hosts (Trifolium purpureum and Trifolium polymorphum) that were each exposed to one microsymbiont capable and one incapable of N-2 fixation. In co-inoculation experiments at a cell density of 10(4) cells/mL, each host nodulated solely with its effective strain, even when this strain was outnumbered 100-fold by the ineffective strain. However, the selection process ceased when the effective strain was outnumbered 1000-fold. At higher basal cell concentrations of 10(5)-10(8) cells/mL, selection for WSM 1325 to form effective nodulation on T purpureum was evident, but was significantly reduced as the ratio of ineffective cells in the inoculum increased above 4-fold. The results indicate that the selection mechanism is highly dependent on the basal rhizobial cell density. Soil pH did not significantly alter the process, and both strains grew at similar rates and formed nodules at similar rates. A preliminary investigation into the genetic backgrounds of WSM1325 and WSM2304 revealed that although their 16S rRNA sequences were identical, they have considerable differences in their symbiotic and chromosomal replicons through examination of atpD, GSII and nodD sequences. (c) 2007 Elsevier Ltd. All rights reserved.