Fear conditioning is one of the prime paradigms of behavioural neuroscience and a source of tremendous insight in the fundamentals of learning and memory and the psychology and neurobiology of emotion. It is also widely regarded as a model for the pathogenesis of anxiety disorders in a diathesis-stress model of psychopathology. Starting from the apparent paradox between the adaptive nature of fear conditioning and the dysfunctional nature of pathological anxiety, we present a critique of the human fear conditioning paradigm as an experimental model for psychopathology. We discuss the potential benefits of expanding the human fear conditioning paradigm by (1) including action tendencies as an important index of fear and (2) paying more attention to “weak” (i.e., ambiguous) rather than “strong” fear learning situations (Lissek et al., 2006), such as contained in selective learning procedures. We present preliminary data that illustrate these ideas and discuss the importance of response systems divergence in understanding individual differences in vulnerability for the development of pathological anxiety.