Background and objectives Conditioned fear may emerge in the absence of directly experienced conditioned stimulus (CS) - unconditioned stimulus (US) pairings. Here, we compared three pathways by which avoidance of the US may be acquired both directly (i.e., through trial-and-error instrumental learning) and indirectly (i.e., via verbal instructions and social observation). Methods Following fear conditioning in which CS+ was paired with shock and CS- was unpaired, three separate groups of participants learned by direct experience (Instrumental-learning), were instructed about (Instructed-learning), or observed (Observational-learning) a demonstrator performing an avoidance response that canceled upcoming US (shock) presentations. Groups were then tested in extinction with presentations of the directly experienced CS+ and CS-, and either a novel CS (Instrumental and observational groups) or an instructed CS (instructed-group). Results Similar to instrumental learning, results demonstrate that avoidance may be acquired via instructions and social observation in the absence of directly learning that an avoidance response prevents the US. Retrospective US expectancy ratings were modulated by the assumed presence or absence of avoidance. Overall, these findings suggest that instrumental-, instructed-, and observational-learning pathways to avoidance in humans are similar. Limitations Alternative experimental designs would permit direct comparison between the pathways for stimuli with no prior experience of fear conditioning, and trial-by-trial US expectancy ratings would help track the modulation of fear by avoidance pathway. Conclusions Instrumental-, instructed-, and observational-learning pathways of avoidance are similar. Findings may have implications for understanding the etiology of clinical avoidance in anxiety.