Evidence for implicit aversive learning effects has been criticized for its lack of experimental rigor and statistical reliability. Here we examine whether attentional emotional responses to aversive conditioned stimuli can occur in the absence of stimulus-outcome contingency awareness, and use a novel Bayesian tool to reliably perform a post hoc categorization of awareness. Across two experiments (n = 40 and 69) participants completed an aversive conditioning task. A novel Bayesian awareness categorization tool was applied to sensitively measure contingency awareness. Finally, attentional and subjective responses toward conditioned stimuli were measured. For participants unaware of contingencies, conditioned stimuli generated attentional avoidance, but only aware participants showed subjective learning effects. For both experiments, awareness scores for unaware participants did not regress above chance level on a subsequent awareness check, revealing a reliable determination of unawareness states. These findings provide evidence for the existence of aversive learning in the absence of contingency awareness, as demonstrated via conditioned attentional responses, and build an analytical framework that can be extrapolated to other implicit paradigms.