We present investigations of counting statistics for five different luminescence-measuring systems. The dark current background and the counts obtained for a PM tube under constant illumination at various light intensities were measured. Where no divider is present in the counting system, the standard deviation of the number of counts exceeds the value predicted by a Poisson distribution but by different amounts in the different systems. This has implications for analysis of the uncertainties in the estimated equivalent doses and other quantitative analyses of luminescence measurements. We suggest how to account for the observed overdispersion of count numbers in uncertainty estimation.