For the past 30 years, research on the amygdala has largely focused on the genesis of defensive behaviors as its main function. This focus originated from early lesion studies and was supported by extensive anatomical, physiological, and pharmacological data. Here we argue that while much data is consistent with the fear model of amygdala function, it has never been directly tested, in part due to overreliance on the fear conditioning task. In support of the fear model, amygdala neurons appear to signal threats and/or stimuli predictive of threats. However, recent studies in a natural threat setting show that amygdala activity does not correlate with threats, but simply with the movement of the rat, independent of valence. This was true for both natural threats as well as conditioned stimuli; indeed there was no evidence of threat signaling in amygdala neurons. Similar findings are emerging for prefrontal neurons that modulate the amygdala. These recent developments lead us to propose a new conceptualization of amygdala function whereby the amygdala inhibits behavioral engagement. Moreover, we propose that the goal of understanding the amygdala will be best served by shifting away from fear conditioning toward naturalistic approach and avoidance paradigms that involve decision-making and a larger repertoire of spontaneous and learned behaviors, all the while keeping an open mind.