Introduction: Previous studies have suggested a direct relationship between the nasopharyngeal carriage of potential middle ear pathogens and the development of middle ear disease. It has been shown that otitis-prone (OP) children tend to be colonized more often than non-OP children. To study the turnover of nontypeable Haemophilus influenzae (NTHI) in the nasopharynx of OP children, arbitrarily primed PCR was applied to NTHI strains isolated nasopharyngeal swabs collected prospectively during a 2-year study period from 35 OP children under 4 years of age at fixed intervals. Methodology/materials: In 20 patients, H. influenzae could be isolated from different sites (left and/or right ear and/or nasopharynx) or at different occasions during follow-up. Forty-eight H. influenzae isolates of different sites (left and/or right ear and/or nasopharynx) of the same patient as well as from siblings were typed using arbitrarily primed PCR with primer ERIC2 and RAPD Ready-to-Go beads (Pharmacia Biotech, Uppsala, Sweden). Results: Typing with arbitrarily primed PCR enabled to differentiate 29 genotypes among the 48 H. influenzae isolates. Sixteen of these fingerprints were observed only once. Thirteen of these fingerprints appeared on two or more occasions. In the three pairs of siblings the same strain was identified at one moment. Genetically identical NTHI strains from unrelated individuals were never identified. In 11 of 14 cases for which isolates were obtained simultaneously from different sites (throat and/or left ear and/or right ear) or from three pairs of siblings, identical fingerprints were observed. In nine cases whereby isolation was separated by a period of more than 4 weeks (maximum 28 weeks) the fingerprints of the isolates were different from the original isolate. Conclusion: Typing with arbitrarily primed PCR indicated substantial genetic diversity among the H. influenzae isolates studied, since for a total of 48 isolates of 20 different patients, 29 different genotypes were observed. Since simultaneous isolation for different sampling sites (ear or nasopharynx) as well as for both of siblings, resulted mostly in identical fingerprints, and since sampling of the same site of the same patient, separated by an interval of more than 1 month, almost always resulted in different genotypes, one could conclude that both cross colonization (between sampling sites within the same patient and between siblings) and turnover are high. The relation between acquisition and development of disease needs further attention.