The brain reacts differently to the faces of people from different races, research shows.
Looking at faces sparked activity in the amygdala
When volunteers looked at pictures of African-Americans, the brain area that processes emotions became active, a study in Nature Neuroscience found.
When they looked at photos of Caucasian faces, the activity was much less.
This held true regardless of the race of the observer, which the authors say could mean the patterns reflect learned cultural responses to racial groups.
But experts criticised the way the study was carried out and said it was impossible to draw any definite conclusions.
Previously, researchers had shown that pictures of African-American faces activated the amygdala in Caucasian people.
It had been suggested that this might be due to lack of familiarity with other races.
But the latest findings, from the University of California, Los Angeles, suggest novelty is not important, since African-American volunteers also had increased activity in the amygdala when they looked at faces of strangers of the same race as themselves.
Dr Matthew Lieberman and his team said it was possible that this reflected negative cultural attitudes toward African-Americans.
When they asked the 22 volunteers to choose the verbal label for the race of the person in the picture that they were shown, they said they found this dampened down the effect.
They said this may have been useful during evolution "to allow controlled processing responses to threat to override automatic responses."
But the same process may now contribute to the development of racial stereotypes they said.
But Professor Mark Johnson of the Centre for Brain and Cognitive Development at Birbeck College, London, said: "While the authors suggest that this result is the consequence of cultural learning, it is very hard to completely exclude the possibility that low-level differences between different race faces cause these differential patterns of brain activation.
"For example, the higher contrast between skin tone and sclera (white part) of the eye may enhance the viewers processing of the eyes in African-American compared to Caucasian-American faces."
Professor Tim Valentine of Goldsmiths University of London said the experiment had many flaws that made it impossible to draw definite conclusions.
"Neuroimaging work of this nature rests on the interpretation of brain activity in one experimental condition which is subtracted from brain activity recorded in another.
"African-Americans are likely to be more familiar with Caucasians than Caucasians are with African-American faces, because the culture is dominated by Caucasians. Therefore, the subtraction is not symmetrical."
He also said the comments about negative stereotypes were pure conjecture.