Four methods of enumeration were compared by monitoring levels of probiotic bifidobacteria in fermented oat drink during storage. Strains of Bifidobacterium longum and B. lactis were quantified by plate counts, fluorescent in situ hybridization (FISH), quantitative real-time PCR and commercial LIVE/DEAD BacLight bacterial viability kit, and the methods were further developed to suit the enumeration of bifidobacteria in fermented foods. Plate counts of both B. lactis and B. longum were lower than the PCR and FISH counts. The LIVE/DEAD counts of B. lactis were comparable to PCR and FISH counts. The plate counts of B. lactis were slightly but significantly lower than LIVE/DEAD counts, suggesting that the cells that were not able to grow on plates may have become dormant. The plate counts of B. longum were several log units lower than LIVE/DEAD counts, suggesting that a remarkable part of the cells were dormant. Real-time PCR and FISH were shown to suit the quantification of the total amount of probiotic bifidobacteria in foods. Plate counts and LIVE/DEAD counts provided conflicting information on viability especially in the case of B. longum. We conclude that the choice of enumeration method for probiotic bacteria may have significant effect on the results of the analysis. The strain-specific properties and the objects of the analysis should be taken into account when enumeration methods for different probiotic strains are chosen.