The population biology of Streptococcus pneumoniae is poorly understood. Most of the important issues could be addressed by the molecular characterization of large, well sampled populations from carriage and from the different manifestations of pneumococcal disease. The authors have therefore developed a pneumococcal multilocus sequence typing scheme and database by sequencing approximately 450 bp fragments of seven housekeeping loci from 295 isolates. The combination of alleles at the seven loci provided an allelic profile, or sequence type (ST), and the relatedness between isolates was obtained by constructing a dendrogram from the matrix of pairwise differences between STs. The typing scheme was validated using pneumococci of known genetic relatedness and could resolve >6 billion STs. Among 274 isolates from recent cases of invasive pneumococcal disease in eight countries, 143 STs were resolved, but 12 STs contained at least five isolates (range 5-21 isolates). The repeated recovery of indistinguishable isolates from invasive disease in different countries implies that these STs define strains with an increased capacity to cause invasive disease. The relationship between STs and serotypes suggested that, in the longer term, capsular genes have been distributed horizontally within the pneumococcal population, but in the short term, expansion of clones occurs with only occasional changes of serotype. The multilocus sequence typing scheme provides a powerful new approach to the characterization of pneumococci, since it provides molecular typing data that are electronically portable between laboratories, and which can be used to probe aspects of the population and evolutionary biology of these organisms. A Web site for the molecular characterization of pneumococci by MLST is available (http ://mlst.zoo.ox.ac.uk).