For unconscious perception research, Bayesian statistics are more appropriate for assessing null awareness of masked stimuli than traditional (frequentist) statistics. This assertion is based mostly upon the theoretical features of Bayesian statistics and modeling studies. To further assess the potential advantages, we compared frequentist and Bayesian statistical tests in a masked Stroop priming experiment in which the prime stimuli were presented at varying degrees of visibility. A novel contribution was to compare a null awareness dissociation approach (i.e., stimulus awareness = 0) to a relative sensitivity approach (indirect or priming effects > direct effects) for the same data. From a null awareness perspective, the frequentist t-tests for the Stroop effect (i.e., perception) for the briefest display conditions had non-significant outcomes. Similar Bayesian t-tests were inconclusive. In contrast, the relative sensitivity dissociation approach was more interpretable, with strong evidence against unconscious perception from a single Bayesian t test. For the longer display conditions, both statistical approaches suggested large conscious perception effects. We conclude that the utility of Bayesian statistics is highly dependent upon the type of dissociation approach, with a relative sensitivity approach being more straightforward to interpret than a null awareness approach.