A generic study designed under the classic fear conditioning paradigm, typically presents the experiment's participant with an object, or the picture of an object, succeeded by some negative stimulus, such as an electroshock. After doing this a couple of times, humans, like Pavlov's dog, learn to associate the object with the occurrence of the electro shock, and whenever they see the picture, their body - as a conditioned response - prepares for the expected shock. Pavlov's dogs, of course, started dripping with saliva at the sound of a bell in expectancy to food, and humans show certain shifts in skin conductance (due to "fear") in expectation of the electroshock.

As mentioned, this type of study has shown humans to be quicker at learning to fear certain dangerous objects than at learning to fear more neutral objects. We also learn to fear angry, snarling faces more readily than we learn to fear laughing faces, and evolutionary stories have been told for most of these findings.

So what to make of a study, such as the one by Olsson et al. conducted in 2005 that showed White people to be better at learning to fear Black people, and Black people to be quicker at learning fear for White people than for other Black people?

As Columbia University's Tiago V. Maia put it in a recent article for the journal in cognitive science: Put it down to simple statistics and prior exposure!

Here is why:

To the above mentioned study, Maia explains that

"Olsson et al. (2005), as well as others, interpreted this finding by appealing to the idea of ‘‘preparedness,'' which proposes that animals (including humans) are genetically programmed to more easily learn to fear stimuli that were dangerous in evolutionary history."

However,

"The timing and pattern of differentiation of human groups into what are commonly called human ‘‘races'' (a concept of questionable biological validity) make it unlikely that humans could have evolved mechanisms specifically to learn to fear different races. Human populations differentiated into different races relatively recently in evolutionary history, and more importantly, different groups evolved different characteristics because they were relatively isolated from each other.
Being genetically prepared to learn to fear individuals from different races is therefore unlikely to have provided any selective advantage."

Since Olsson et al were themselves aware of this inconsistency, they hypothesized instead that humans evolved a general preparedness to learn fear of

‘‘others who were dissimilar to them or who otherwise appeared not to belong to their social group''

However, this is also not a straight forward way of reasoning, since

"The definition of a social ingroup (and, by exclusion, of a social outgroup) depends on experience and familiarity: Those in my social ingroup are those around me and possibly others similar to them."

and, as Maia points out, standard theories of conditioning that do not rely on genetic arguments are much better equipped to explain findings that crucially depend on prior exposure and familiarity.

Besides a number of further arguments that make the genetic hypothesis unlikely, Maia also points to data supporting the idea that familiarity and exposure may be the main driver for Olson's effect.

For one, in Olson's original data set

"contact with members from the other race was the only factor that mediated the superior outgroup conditioning effect. Second, [...] there was an apparent trend for the superior outgroup conditioning effect to be stronger for White than for Black participants. [...]. The latter finding is probably simply a consequence of the former, as Black participants reported more prior exposure to White individuals than White participants did to Black individuals. [...] These two findings are direct predictions of the latent inhibition hypothesis: More contact with members of the outgroup means more latent inhibition for the outgroup, which implies a reduction in the superior outgroup conditioning effect."

Tiago Maia then offers a statistical model of Bayesian learning, that does a great job in explaining the finding based simply on previous contact between the groups; i.e. on the latent inhibition hypothesis, which you might want to look up if you're interested in this type of statistical modeling.

His model and analysis seem a better account of the findings presented in Olsson (2005) since the model requires fewer assumptions to explain the main effect, is able to explain additional findings in the original data set, and - possibly most importantly - the model fits in neatly with standard learning theory, which is based on decades of research, rather than needing to assume

"a new mechanism-a socially specific genetic bias-for which there is no independent evidence."

There is a current pessimism in which many social ills, are quickly put off as being "simply part of human nature", but at least when it comes to racial prejudice, it seems we cannot blame our genes.

While we're on the topic, you might also be interested in this interesting study on how we perceive political candidate's skin color: How dark is Barack Obama?

Main Reference:

Tiago V. Maia (2009). Fear Conditioning and Social Groups: Statistics, Not Genetics Cognitive Science, 33 (7), 1232-1251 : 10.1111/j.1551-6709.2009.01054.x

You might also enjoy this post, in which I discuss another instance of evolutionary reasoning not having been applied with the necessary caution.