Humans, like most organisms, are built to learn what may cause us harm so that we avoid it in the future. I learned the truth of this statement as a young boy when I decided to touch the burner of an electric stove just after it had been turned off. As I touched the burner I noticed a light illuminated on the stove panel, thereby learning to associate the panel light with the potential for a painful burn. For many weeks after my painful initiation to the ways of electric stoves the sight of the panel light made me feel apprehensive and wary. This simple, but essential, form of associative learning has for decades been studied using the conditioned fear (CF) paradigm in which participants learn to ‘fear’ a previously neutral stimulus (the conditioned stimulus or CS+, e.g. a simple shape or stove panel light) through its repeated pairing with an intrinsically aversive stimulus (the unconditioned stimulus, or UCS, e.g. a mild shock, or in my case, a hot burner). Over time, presentations of the CS+ alone comes to elicit responses like those initially elicited only by the UCS (e.g. a transient rise in skin conductance). In recent years, the CF paradigm has been a powerful tool in both animal and human studies for elucidating the neural systems supporting learning to ‘fear’ that which may cause us harm. One brain structure in particular—the amygdala—has emerged as the key site for forging the link between stimuli and their potentially aversive consequences (LeDoux, 2000). Learning through direct experience is not, however, the only or even most common way we learn what has threat value. Indeed, the friends who watched me touch the stove did not need to repeat my mistake in order to know the meaning of an illuminated panel light. This ability to learn vicariously from the actions of others has been captured by the observational fear (OF) paradigm, in which an individual watches another person undergo the CF procedure. Although studies in both animals and humans have shown OF and CF learning may show similar behavioral characteristics (Olsson and Phelps, 2004), until now, no studies had addressed the underlying neural mechanisms. In this issue, Olsson and Phelps report the first functional imaging study of OF learning. Using a human analog of a behavioral paradigm first developed in animal studies, Olsson and Phelps scanned participants both while watching a video clip of an individual in a CF experiment and afterwards during presentations of the CS+ and a control CS− that had not been paired with shock in the video. Strikingly, amygdala activity was observed both during observational learning and during expression of a CR. This result is important for at least five reasons. First, it provides the first examination of the neural systems involved in a form of social, vicarious learning. Second, it demonstrates a viable method for studying social learning that could be adapted for addressing various questions about how we learn from others. Third, it demonstrates that vicarious learning of fear may rely upon mechanisms like those involved in learning to fear through direct experience, thereby joining a growing literature suggesting that common or shared representations support the direct perception of pain, emotion or action as well as the perception of the same in other people (Wicker et al., 2003; Singer et al., 2004; Decety and Grezes, 2006). Fourth, Olsson and Phelps observed activity during observational learning in a network of brain systems not typically associated with CF learning, but rather associated with social cognition, including medial prefrontal cortex and the superior temporal sulcus (Gallagher and Frith, 2003). Although activity in these regions previously has been associated with drawing inferences about the mental states of others, and taking their third person perspective, this is the first study to show that these regions may play a role in learning. Fifth, it highlights the usefulness of animal models for social cognitive and affective neuroscience research. The OF paradigm used by Olsson and Phelps used methods similar to those used in animal models of observational learning, that in turn were based upon animal models of conditioned fear (Mineka and Cook, 1993). Cross-species comparisons may both illuminate underlying mechanisms and highlight what is unique about vicarious learning in humans. That being said, by taking steps in new directions, the present experiment—like any impactful study—may raise more interesting questions than it answers. Perhaps foremost among them is the question of what specific roles in observational learning are played by the various neural systems activated in addition to the amygdala, and what factors mediate the learning of observational fear. Now that we know it depends upon activity both in the amygdala and regions involved in social cognition, the question naturally arises as to what role social cognition plays in OF learning. Might perspective taking, for example, play an essential role in this learning, supported by activity in regions associated with mental state inference, and could activity in these regions mediate amygdala-based fear learning? Other questions concern the generalizability of the effects observed here—would other forms of vicarious or observational learning depend upon the specific systems identified here, and/or more generally follow the principles observed here and in other studies of shared representations—i.e. will similar systems always support direct and vicarious learning? Whatever the answers to these specific concerns and questions, it is clear that part of what makes this study important, and a likely citation classic, is its Homer Simpsonesque, ‘Doh!’ factor. Upon hearing about the study and its results, we might immediately make this exclamation and wish we had thought of doing it ourselves. May we all learn through observation to conduct studies of similar impact.