Our long-term objective is to understand how levels of consciousness moderate the influence of emotionally relevant stimuli on both normal and psychopathological cognitive processes. Notably, much evidence suggests that stimuli are automatically evaluated as pleasant or unpleasant, even when presented unconsciously. Further, because many clinical disorders (e.g., anxiety, depression) exhibit preferential processing of negative stimuli, automatic evaluation (e.g., emotional Stroop) paradigms show considerable promise as markers of, vulnerability to, or recovery from such conditions. Moreover, because presenting stimuli consciously vs. unconsciously mediates such effects, taking consciousness into account could improve marker validity. However, although most assume a single unconscious process, our work (e.g., Snodgrass, Bernat, & Shevrin, 2004a) suggests that two qualitatively distinct varieties objective/deep and subjective/shallow. Accordingly, it may be useful to examine both kinds of potential unconscious markers. Our proposal applies this perspective to a direct task stimulus detection--rather than typical indirect RT measures. Unlike Dijksterhuis and Aarts (2003), who found a negative (vs. positive) word detection superiority under shallow unconscious conditions, our preliminary data found no such effect. Moreover, we found the reverse pattern (i.e., positive detection superiority) under deep unconscious conditions. Further, this positive superiority was mediated by defense-related individual differences, suggesting unconscious inhibitory emotion regulation (cf. Gross, 1998). These counterintuitive deep effects, including initially unexpected RT effects, were replicated several times. Finally, additional preliminary data with clinical phobics suggested that the potential deep marker predicted improvement following laboratory exposure (less defensive phobics improved more; more defensive phobics improved less), suggesting significant translational potential. Additionally, finding no shallow effect is actually consistent with other work (e.g., Anderson, 2005; Zeelenberg Wagenmakers, & Rotteveel; 2006), suggesting arousal but not valence effects under shallow conditions. The proposed experiments specifically aim to: 1) Clarify inhibitory vs. facilitatory mechanisms by adding neutral stimuli (allowing examination of valence vs. arousal effects); 2) Examine the time course of the RT effects; 3) Generalize the effects to facial stimuli; and 4) Further explore translational potential by again including clinical phobics, now examining all deep and shallow unconscious markers, both accuracy and RT. The proposed research may yield laboratory markers for clinical disorders complementary to those made by RT (e.g., emotional Stroop) paradigms. Moreover, even if the current affects ultimately lacked clinical usefulness themselves, definitive demonstration of qualitatively distinct unconscious processes would significantly improve the potential clinical utility of any paradigm examining potential unconscious markers. ? ? Public Health Relevance: Although treatments for anxiety disorders and depression have become more effective, many sufferers either do not respond fully and/or relapse once improved. The proposed studies examine unconscious emotion regulation as a potential predictive marker for treatment response and/or relapse, thus enabling improved diagnosis and treatment. ? ? ?