Prior neuroimaging studies have shown that the amygdala is activated robustly by fearful facial expressions. This response is apparently so automatic that it can be observed even when fearful faces are presented using a technique called backward masking, where amygdala response is observed to fearful faces despite the fact that subject's report that they have not seen any fearful faces. The way this works is that a fearful face is presented for a very short period of time (e.g., 17 ms). It is then followed by a slightly longer presentation of another face that is neutral in expression (e.g., 183 ms). If the neutral face is presented immediately following the fearful face presentation, subjects' will report that they saw only the neutral face. (In the past, such presentations have been called 'subliminal' and this is probably an OK way to think about what is being done here). When we do these studies, we will often have two masked expression conditions so we can compare amygdala response to each of them (e.g., masked fear vs. masked happy). Thus, in our studies the subject thinks they have seen numerous neutral face presentations, but they have really seen numerous neutral face presentations, where half were preceded by fearful faces, and the other half by happy faces.
Studies of this sort show that the amygdala is still responsive to fearful faces, even when subjects do not know that that these expressions have been presented. These data suggest that this ability is probably so important to the human condition that it has become highly automated.
We argued then that the amygdala is interested in fearful faces for two reasons: What's there, and what's NOT there. That is, first the amygdala is interested in this expression because of its predictive information value (i.e., it has predicted negative outcomes in the past). But, the amygdala is also interested in the source of the event that you are dealing with. Fearful faces (unlike angry faces, for example), when presented alone in a neuroimaging experiment, give you no information about the source of the environmental event that evoked them. So while part of the amygdala is responsive to the negative valence of fearful faces, another part is responding to their predictive uncertainty (i.e., what has happened that evoked that expression?). To make this point, in a subsequent study we showed that this part of the amygdala was also responsive to surprised faces. Surprised faces make the point nicely because like fear, they do not give information about the source of their eliciting event, but, unlike fear, they are not necessarily negatively valenced. A surprised face could predict an oncoming car or a birthday party.
If the amygdala is responsive to fearful and surprised faces, then one begins to consider what features these two expressions have in common. The most striking similarity between the two is the eyes. Indeed, we refer to these expressions as "wide-eyed" information gathering expressions. When you see them, part of you might say to yourself "That person has detected an important event in our immediate environment and I would do well to find out what that is." If the size of the eyes is a basic signal of predictive significance, then an experiment that manipulated the size of the eye white should show that this simple rule is sufficient to produce amygdala activity.
When designing this study, the first problem you encounter is that if one sets off to simply present eye stimuli to human subjects, you run into the problem of the abnormal nature of the stimuli themselves. 'Eyes floating in space' are weird and these sort of thoughts going on in other areas of the brain connected to amygdala, would likely influence activity there, potentially confounding your ability to measure an effect there.
As noted above, one can use backward masking to present stimuli in a way that subjects will not be aware of their presence (thus mitigating competing cognitions concerning their abnormal nature). So, in the present study, we essentially replicated our earlier study using backward masking of fearful vs. happy faces, but this time we presented only the eye white information within fearful and happy expressions in a backward masking paradigm. As can be see in our published report, the amygdala was responsive to eye white information alone. Given that facial expressions are very complex configural stimuli, one may have wondered how the amygdala could have handled all this information during the short stimulus presentations used in backward masking. The eye white finding offers a simple rule that the amygdala may have been using even when the whole face was presented.
I think the paper tells you how automatically your brain picks up on environmental stimuli that predict biologically relevant outcomes for you. In this case the potential outcomes might be dangerous. Oversensitivity of such systems in some people could contribute to hypervigilance for threat (i.e., anxiety disorders).
Others will want to emphasize the fact that this is a likely evolutionarily conserved rule that you were born with. They will suggest that this is the case since it works subliminally. The caveat is that subliminal does not equal innate and this could still be a learned skill. There are many examples of our brains automating important tasks that we have done a thousand times so that we can carry them out without explicit attention to every detail (e.g., driving a car). Thus, it is not so heretical to suggest that facial expressions are so important to us that we have automated the processing of some of them. Also, if one wants to think of this ability in an evolutionary or innate sense, then it must be a more recent development in the human because the sclera of most monkeys and apes is not light, but dark colored.
Finally, anatomists are embroiled in an argument concerning whether there are shorter subcortical anatomical pathways to the amygdala that can bypass more detailed processing within sensory cortices. If such a pathway exists it would likely only be able to communicate very simple information (i.e., a simpler rule than something as complex as a whole fearful face). The present eye white finding offers evidence for such a rule (though it cannot speak to the existence of the short pathway). More generally, we designed the present study because we suspected that the amygdala is not that bright (i.e., in terms of the neuronal machinery it can bring to bear when processing environmental stimuli). The eye white finding offers a simple rule that the amygdala could handle. In the amygdala's defense, if we are to think of it as an idiot, then I would say it is more of an idiot savant. The amygdala may not know much, but what it knows is very important - so important that it will react to this information in a very fixed and ritualistic way, and it will resist all attempts to dampen its efforts. It is interesting that the amygdala is implicated in the disorder of autism, because in this sense, the amygdala itself is a bit autistic.