This work presents an analysis of the consequence of using data sets containing small numbers of disc diffusion inhibition zones to calculate epidemiological cut-off values. Twenty replicate data sets that contained 10, 20, 40, 60, 80 and 100 zone size measurements were obtained by randomly sampling a previously published data set that consisted of inhibition zone measurements for discs containing 30 µg florfenicol, 30 µg oxytetracycline and 2 µg oxolinic acid against 217 isolates of Aeromonas salmonicida. Application of normalised resistance interpretation allowed epidemiological cut-off values to be estimated from all of these 360 data sets. The precision associated with the replicate cut-off values was a linear function of the log of the number of data points used to generate them. The frequency with which false sensitive and false resistance errors would have resulted from the application of the cut-off values generated from small data sets using normalised resistance interpretation were compared with those that would have resulted from the application of the breakpoints that have been reported as being currently in use. This comparison demonstrated that most laboratories could improve their interpretation of the disc diffusion data they obtain for any species by performing normalised resistance interpretation of their own data for that species. This would be true even if they possessed inhibition zone data for as few as ten members of that species.