Most studies of bacterial pathogen populations have been based on isolates collected from individuals with disease, or their contacts, over short time periods. For commensal organisms that occasionally cause disease, such as Neisseria meningitidis, however, the analysis of isolates from long-term asymptomatic carriage is necessary to elucidate their evolution and population structure. Here, we use mathematical models to analyse the structuring and dynamics of three vaccine-candidate antigens among carried meningococcal isolates collected over nearly 30 years in the Czech Republic. The data indicate that stable combinations of antigenic alleles were maintained over this time period despite evidence for high rates of recombination, consistent with theoretical models in which strong immune selection can maintain non-overlapping combinations of antigenic determinants in the presence of recombination. We contrast this antigenic structure with the overlapping but relatively stable combinations of the housekeeping genes observed among the same isolates, and use a novel network approach to visualize these relationships.