© 2016 Macmillan Publishers Limited. Classical Pavlovian fear conditioning remains the most widely employed experimental model of fear and anxiety, and continues to inform contemporary pathophysiological accounts of clinical anxiety disorders. Despite its widespread application in human and animal studies, the neurobiological basis of fear conditioning remains only partially understood. Here we provide a comprehensive meta-analysis of human fear-conditioning studies carried out with functional magnetic resonance imaging (fMRI), yielding a pooled sample of 677 participants from 27 independent studies. As a distinguishing feature of this meta-analysis, original statistical brain maps were obtained from the authors of 13 of these studies. Our primary analyses demonstrate that human fear conditioning is associated with a consistent and robust pattern of neural activation across a hypothesized genuine network of brain regions resembling existing anatomical descriptions of the 'central autonomic-interoceptive network'. This finding is discussed with a particular emphasis on the neural substrates of conscious fear processing. Our associated meta-analysis of functional deactivations - a scarcely addressed dynamic in fMRI fear-conditioning studies - also suggests the existence of a coordinated brain response potentially underlying the 'safety signal' (that is, non-threat) processing. We attempt to provide an integrated summary on these findings with the view that they may inform ongoing studies of fear-conditioning processes both in healthy and clinical populations, as investigated with neuroimaging and other experimental approaches.