A longitudinal pilot study was designed to investigate variability of test results at commercial dairy farms in New York State. ELISA and fecal culture results of individual cows seem to vary considerably over time, which complicates the determination of the infection status of a cow or a farm. However, it is of great importance, for example for eradication of the disease, to be able to correctly identify animals that might be shedding or are likely to develop clinical disease. The objective of the study was to determine the causes and amount of variation in ELISA and fecal culture results of individual cows. Sixteen cows in each of six herds were tested monthly with a kinetics ELISA (KELA) and fecal culture was done bimonthly during 2001. Cow- and herd-level data were collected at every sampling date from the management information system (DAIRYCOMP), which was used by all farms. The KELA results were modeled with the MIXED procedure in SAS with a random cow-nested-in-herd effect with a CS covariance structure. The hazard rate till a cow had a positive fecal culture result was investigated with a Bayesian model in WinBUGS 1.3. Random effects were included in the survival model to correct for repeated observations on cow-level and the potential risk factors were investigated. The KELA model showed that cows in second or higher lactation have increased KELA values compared with heifers. Cows that calved at most 15 days ago had the lowest KELA value that first rapidly increased and from 60 days in lactation decreased. Moderate and heavy shedders had significantly higher KELA values than fecal culture negative cows. In the second model for fecal culture, the hazard rate of becoming test positive was significantly affected by a random cow-effect and by season. KELA was not a significant factor in the model and had a limited predictive value for a fecal culture positive test result.