Published studies of rhizobial populations, communities and other strain collections were analysed to identify trends in strain richness, strain dominance and genetic diversity within and between locations. For individual populations and communities, strain richness indices were calculated by dividing the number of strain types identified by the number of isolates recovered. Where possible, strain dominance (the proportion of rhizobial isolates represented by each strain type) was also calculated. Analysis of the genetic diversity of populations, communities and rhizobial strain collections originating from diverse legume hosts and locations, was confined to studies using multilocus enzyme electrophoresis (MLEE) so that diversity could be compared on the basis of published H values. Strain richness indices were highly variable (0.02-0.94) and were influenced by both the discriminatory power of the strain typing method and the type and number of legume species used to recover rhizobia from soil. The strain richness of populations recovered either directly from soil, or from the nodules of trap hosts inoculated with the same soil, was similar. Because the arithmetic relationship between the number of strain types and number of isolates varies between different populations and communities, strain richness curves are proposed as the most appropriate method for reporting rhizobial structural diversity. Comparison of over 50 rhizobial populations and communities from published studies showed that strain dominance patterns in nodules were often similar. Typically, a single strain type occupies more than 30% of nodules with the majority of strain types being recovered at low frequency (similar to 75% of published populations and communities). Rhizobial populations and communities characterised by MLEE varied in their genetic diversity, with H values ranging from 0.06 to 0.78. In several studies, most of the genetic diversity within a site could be recovered from the nodules of a single plant. When hierarchical analyses were performed on populations within and between sites, the genetic diversity within sites was similar to the genetic diversity between sites. Similarly, the genetic diversity of strain collections originating from multiple hosts and locations was no more diverse than some individual populations and communities. Strain richness and genetic diversity measures were not always correlated for rhizobial populations. Populations with low strain richness were sometimes genetically diverse, and the relationship between the diversity measures varied for different legume species at the same location. We suggest that both strain richness and genetic diversity measures are required to fully describe rhizobial population and community diversity. (C) 2004 Elsevier Ltd. All rights reserved.