Estimating the precision of serial dilutions and viable bacterial counts.

The propagation of error in serial dilutions was investigated theoretically and by means of computer simulations. The principal aim of the study was, given only the pipette manufacturer's specification, to estimate the variance of any step in a dilution series both of pure solutions and of homogeneous bacterial suspensions by means of simple formulae. The study was extended to include bacterial plate counts by both the standard and the Miles and Misra methods. It was found that such estimation was possible and that the distributions approximated the normal sufficiently for the construction of confidence intervals (Cls) by the usual method. Such intervals can be regarded as minima which could be inflated by other, possibly undetermined, factors. It is suggested that laboratories could construct tables such as that reported here for pipettes and methods in common use to facilitate estimation. While replication of the final sampling step of a plate count increases the precision of estimation, averaging across dilutions may decrease precision and is not recommended for the standard pour-plate count.