The Ultimate Attribution Error

In his excellent blog, Glenn Greenwald made the following point about media coverage of the massacre of Afghan civilians by an American soldier:

Here's a summary of the Western media discussion of what motivated U. S. Staff Sgt. Robert Bales to allegedly kill 16 Afghans, including 9 children: he was drunk, he was experiencing financial stress, he was passed over for a promotion, he had a traumatic brain injury, he had marital problems, he suffered from the stresses of four tours of duty, he “saw his buddy's leg blown off the day before the massacre,” etc.

Here's a summary of the Western media discussion of what motivates Muslims to kill Americans: they are primitive, fanatically religious, hateful terrorists.

Greenwald also points out that when the media try to understand possible causes of the bad actions of American soldiers, we don't confuse them with excuses for their actions. But when anyone suggests that Afghans who commit crimes may be responding to the illegal and increasingly violent occupation of their country, they are immediately accused of attempting to justify or defend terrorism.

This illustrates what social psychologist Thomas Pettigrew called the ultimate attribution error. Attribution refers to the process of identifying the causes of events. Humans are known to make a number of systematic errors when assigning causality. The ultimate attribution error refers to a systematic bias in the way we analyze the behavior of ingroups vs. outgroups, or friends vs. enemies. Good and successful actions by ingroup members are attributed to their stable internal dispositions, such as ability or motivation. But our own and our friends' bad or unsuccessful behavior is seen as a temporary product of a difficult situation or environment.

When analyzing the behavior of an outgroup, especially a disliked one, this tendency is reversed. Their bad behavior is seen as a product of their evil nature, while their good behavior is a product of a temporary, favorable, external situation. (Think of the tendency of prejudiced whites to attribute black success to affirmative action.) Needless to say, any unfavorable stereotypes we have of outgroup members will be maintained by these attributions, even in the face of behavior that might appear to contradict them.

Greenwald's point reminded me of a pair of related studies by Michael Morris and Kaiping Peng. In the first, they analyzed newspaper accounts in the New York Times and the Chinese-language World Journal of two mass murders that occurred in the United States, one committed by a Chinese man and the other by an American man. Each unit, or clause, that gave an explanation for the crime was classified according to whether it referred to a permanent disposition of the actor or a temporary situational influence.

The Times article showed the ultimate attribution error. American reporters attributed more causality to situational factors when describing the American murderer than the Chinese murderer. On the other hand, more dispositional traits were attributed to the Chinese than to the American killer. The Chinese-language newspaper did not show this bias.

The second study was an experiment. Chinese and American participants were given an article about either the American or the Chinese murderer to read. The articles were balanced in that they referred to an equal number of personal and situational causes of the crime. Afterwards, the participants were asked to rate the importance of the personal and situational explanations mentioned to in the article. As predicted, the American participants gave more weight to dispositional factors when rating the Chinese than the American murderer, and more weight to situational factors when rating the American than the Chinese killer. The Chinese participants did not show this bias.

This is a relatively conservative test of the ultimate attribution error, since there was no particular reason to think the American participants were hostile to individual Chinese citizens. In this study, the Chinese are an outgroup rather than an enemy. You would expect the contrast to be greater when we are at war with the outgroup.

The study shows that the ultimate attribution error occurs both in our individual judgments and in the mass media, but it doesn't address the chicken vs. egg problem. Do the media influence our judgments, do they pander to what they know are our preferences, or both?

Finally, you'll notice that in this study the ultimate attribution error is uniquely characteristic of Americans. I hope to discuss how and why Americans' attributions differ from those of most of the rest of the world's people in a future post.