Date of Degree


Document Type


Degree Name





Tracy Dennis-Tiwary

Committee Members

Ekaterina Likhtik

Glenn Schafe

Evelyn Behar

Catherine Hartley

Subject Categories

Biological Psychology | Clinical Psychology | Cognitive Neuroscience


Safety Learning, Fear Learning, Anxiety, Overgeneralized Fear


Anxiety disorders are among the most common mental health diagnoses, affecting about a third of the population in their lifetime. However, approximately a third of individuals with anxiety do not respond to current treatment approaches, highlighting the need to identify additional potential therapeutic mechanisms. Safety learning is one such mechanism, but methodological challenges and a dearth of research have prevented the field from advancing the understanding of the role of safety learning in the etiology and remediation of anxiety disorders. Animal research, using single-cued safety learning paradigms, has yielded promising early findings, demonstrating that safety learning directly reduces anxiety-related behaviors and cognitive processes, such as overgeneralized fear (OGF), or indiscriminate fear responses to non-threat. The goal of the present dissertation was to extend these methods to apply them to research on human anxiety. We tested the hypothesis that safety, versus fear learning, will reduce subjective and objective measures of anxiety and OGF in a sample of low to severely anxious individuals (N = 57, Mage = 24.05; SD = 6.07). We also explored whether these predicted associations were moderated by anxiety-related individual differences, such as anxiety sensitivity, and whether the saliency of safety cues facilitated these predicted effects. Participants were randomly assigned to one of four learning conditions completed on Visit 1 (Day 1): fear learning, safety learning without added saliency (Safety Learning 1), safety learning with a salient cue co-terminating with the CS presentation (Safety Learning 2), or safety learning with a salient cue co-occurring with the CS presentation (Safety Learning 3). Participants also completed a set of questionnaires to assess their baseline anxiety and anxiety-related differences. During the learning phase, skin conductance response (SCR) was measured. For Visit 2, the participants returned 24 hours later and were tested on the following: 1) fear or safety learning retrieval in the context in which they were trained, and 2) fear or safety learning retrieval in a novel anxiogenic environment in order to assess anxiety-related behaviors in relation to the learned cues. During these tests, SCR, brain activity, and behaviors were observed. On the same day, fear generalization task was completed to measure OGF, during which SCR, brain activity, and behavioral risk ratings were assessed. We found that there was a marginally higher SCR during the period in which the CS stayed off compared to when the CS stayed on for safety learning, whereas an opposite pattern emerged for fear learning during in-context retrieval, indicating successful attainment of learning. However, the findings from the fear generalization task did not support the hypothesis that safety learning would reduce OGF. Yet, a moderation analysis revealed that relative to participants in the fear learning condition, those in the safety learning condition who also reported high levels of anxiety sensitivity at baseline showed higher OGF measured via difference scores between risk ratings to CS+ and perceptually similar generalization stimulus. Limitations and future directions are discussed. Taken together, results highlight the importance of applying translational methods to the study of safety learning in human anxiety in order to clarify etiological and treatment implications that may be unique from those of fear learning, and to inform the development of novel treatment approaches.