Introduction: In France, surveillance of meningococcal infections (MI) is based on mandatory case reporting. In 1989–1990, the completeness of mandatory reporting was estimated at 51% using the capture-recapture method. To determine changes in the completeness of MI surveillance, we conducted a new capture-recapture study using data collected in 1996. Method: A case was defined as a hospitalized patient diagnosed with meningococcal infection from whom a strain of Neisseria meningitidis had been isolated from blood or cerebrospinal fluid. Three data sources were compared: 1) admission records completed by clinicians; 2) cases for which the strain was sent to the National Meningococcal Reference Center (CNRM); and the network of hospital laboratories, EPIBAC, which reports cases occurring in the hospital on a monthly basis. After identifying common cases across the sources, we estimated the number of reported cases in each source using the log-linear regression method. Estimating the total number of cases occurring in France in 1996 allowed us to calculate the completeness of the DO, the CNRM, and EPIBAC in 1996. Results: A total of 415 cases of MI were identified from the three sources (DO: 285, CNRM: 329, EPIBAC: 231). 220 cases were common between the DO and the CNRM, 192 between the CNRM and EPIBAC, and 145 between the DO and EPIBAC. 127 cases were common to all three sources. Analysis of log-linear models revealed a dependency between the CNRM and EPIBAC, interactions between serogroup and DO and serogroup and CNRM, and between geographic area and EPIBAC. The best model (lowest BIC) estimated 460 cases (95% CI: 452–468) occurring in France in 1996. Since 1990, the coverage of the DO has increased from 51% to 62% (95% CI: 61–63), and the coverage of the CNRM has increased from 53% to 72% (95% CI: 70–73). The completeness of EPIBAC was estimated at 49% (95% CI: 45–54) in 1996. Conclusion: The completeness of the DO and the CNRM improved significantly between 1990 and 1996. Greater participation by clinicians and laboratory technicians in reporting MI is largely responsible for this improvement. This study demonstrates the value of the capture-recapture method for evaluating surveillance systems and their evolution. The assessment of trends over the years must take this evolution into account. (R.A.)