Photon counting has been used in TL analysis for more than 40 years, with evidence that response to light sources show conformity to Poisson statistics at least for short term variations. Recent papers however have drawn attention to excess variations in some photon counting used in more recent instruments 1,2, and have suggested that the errors estimated for luminescence dating would be underestimated if such excess variations occurred. To investigate this studies of 10 photon counting systems were conducted at SUERC examining the ratio of observed to Poisson variance at very low phosphorescence levels and for dark signals. The results indicate that while phosphorescence is normally detected with variance consistent with Poisson uncertainties some of the systems, and particularly the more modern one showed excess dark signal variations. Further examination suggested that multi-event bursts, which were more prominent in some systems than others, were associated with the majority of the excess dark count variations. Accordingly, a procedure was
developed to identify and correct for such events based on retrospectively evaluating the Poisson probability of obtained the observed results. This procedure was then applied to a single grain data set from a Neolithic chambered tomb in Corsica, where significant proportions of grains gave signals which fell below significance levels based on Poisson rejection criteria. When compared the results from corrected data sets (26+-5 Gy) and the original data (25+-4 Gy) lead to similar equivalent doses, from which we conclude that the impact on age determination is modest. Nonetheless it is appropriate to characterise detection systems and to take account of excess variance in dark signals when assessing signal significance.