CNN misreports “classic study on race”

CNN“In a classic study on race, psychologists staged an experiment with two photographs that produced a surprising result.”

—CNN’s John Blake, in a Nov. 27, 2014 story titled “The new threat: ‘Racism without racists’.”



CNN and the expert it sourced, Howard Ross, both provide badly distorted accounts of a research project.

The Facts

The new threat: ‘Racism without racists,’” a story from CNN published on Nov. 27, 2014, caught our attention thanks to the fact check we published in April about racism and the tea party. The story by John Blake overlapped with our fact check to some degree, but the opening example of contributing research triggered a “too good to be true” warning flag for us:

(CNN) — In a classic study on race, psychologists staged an experiment with two photographs that produced a surprising result.

They showed people a photograph of two white men fighting, one unarmed and another holding a knife. Then they showed another photograph, this one of a white man with a knife fighting an unarmed African-American man.

When they asked people to identify the man who was armed in the first picture, most people picked the right one. Yet when they were asked the same question about the second photo, most people — black and white — incorrectly said the black man had the knife.


We started digging.

Analyzing the Rhetoric

We were concerned but not especially surprised when the “classic study” proved elusive. We tried a few combinations of appropriate search terms and found nothing. We used Twitter to query the author [Ross responds in the comments section below, Feb. 27, 2015] but before receiving a response we finally found a widely cited study involving a white man, a black man and a knife.

allport and postman originalThe Psychology of Rumor

The study we located, “The Psychology of Rumor” by Gordon W. Allport and Leo Postman, was not a study of race. It studied the spread of rumor.

It used illustrations instead of photographs.

There was no knife fight involved.

So far as we can discover, the illustration in the study consistently featured a central black character.

The experiment in “The Psychology of Rumor” used an illustration with a number of details like clothing, postures and ads to study how a description of the image changed as one person described it to another, and so on. After the story was told several times, the black man often ended up with the razor. Brent Staples summarized the experiment for The New York Times in 2009:

After being shown the slide, subjects were asked to describe it to others who had not seen it. These people then described it to others, who then passed on their descriptions as well. Those who had heard the story secondhand were then asked to recount it. More than half the time, the razor was said to be held not by the white man but by the well-dressed black man, who was sometimes described as brandishing it wildly.


During our research we also ran across a journal article about the way “The Psychology of Rumor” often gets misused when applied to witness testimony in the courtroom. The abstract for Effects of racial stereotypes on eyewitness performance: Implications of the real and the rumoured Allport and Postman studies by Molly Treadway and Michael McCloskey sums up the problem:

The Allport and Postman (1945, 1947) study of rumour has been widely cited in support of the claim that expectations based upon racial stereotypes can cause eyewitnesses to make dramatic errors in perceiving or remembering an event. However, this claim is founded on inaccurate accounts of the study that have appeared throughout the eyewitness testimony literature. In this article we explore the implications of the actual Allport and Postman study, and the study as erroneously described, for questions about eyewitness performance.


So apart from featuring a razor blade, confusion about whether a white man or a black man possessed the blade and a degree of renown, the Allport and Postman study didn’t fit Blake’s version. Though at the same time, the study achieved some notoriety for being misinterpreted and misreported.

Using Twitter, we conveyed our concerns to Blake, asking him to name those responsible for the study he described in his article. Blake pointed out that his article said the example came from Howard Ross’ book “Everyday Bias.”

The version of the study in Ross’ book was a close match for Blake’s version:

One of the most popularly known studies on implicit bias and eyewitness identification involves a photograph of two men fighting; one man was holding a knife while the other was unarmed. When both men in the photograph were Caucasian, subjects generally correctly remembered which man was holding the knife. When the Caucasian man was armed and the African American man was unarmed, the majority of subjects, both African American and Caucasian, wrongly thought the African American man was holding the knife.


The paragraph was marked with a footnote referring to “The Nature of Prejudice” by Gordon W. Allport. But that book contained no description of any study involving a knife. It did not even mention the research Allport conducted earlier with Postman. Ross’ citation didn’t match.

Did Ross create a rumor about Allport and Postman’s study of rumor?

Rumors Greatly Exaggerated: Allport and Postman’s Apocryphal Study

As we moved to tie the elements of our fact check together, we happened on another study involving images of men with knives. Rumors Greatly Exaggerated: Allport and Postman’s Apocryphal Study, by researchers Julian C. Boon and Graham M. Davies, made a much better match for Ross’ description (bold emphasis added):

All subjects were shown slides depicting a violent knifing incident between two men on the London Underground. Half were shown the altercation as occurring between two white men and half between a black and a white man. In all cases, the critical slide in which the knife was presented showed it to be in the hand of the white man. After a distraction interval of 45 minutes subjects were given forced-choice recognition and recall tests. The results indicated that those subjects who had seen the slides with the black and white protagonists were significantly less accurate than those who had seen only white protagonists though only where the recognition test preceded that of recall.


Boon and Davies specifically tried to craft an experiment that would test ideas implied in professionals’ false accounts of the Allport and Postman study. After recounting the error-filled version presented by Robert Buckhout, the researchers chided their colleagues for their carelessness:

In fact, the only sense in which this reported eyewitness paradigm accords with that of the original’s serial report is in the chain of misreports it has generated in the psychological literature, law journals, and textbooks, and even in expert psychological testimony delivered in court.


While details of this study better matched the respective reporting from Blake and Ross, discrepancies remained. Yes, the pictures used in the experiment depicted a knife fight of sorts. Yes, the researchers used a control group showing a conflict between two whites. But the images were line drawings, not photographs, and both Blake and Ross misreported the results. Neither mentions the recognition phase of the experiment that accounted for the memory interference. For the group recalling events from memory before viewing the recognition slides, not one reversed possession of the knife from the white man to the black man. Eighteen of the 24 rightly recalled the white man wielding the knife while the remaining six did not notice the knife at all. It was the group viewing the recognition slides before trying to recall details that wrongly recalled the black man with the knife.

The results of the study by Boon and Davies, therefore, showed no bias effect with immediate recall, contrary to the presentations by Blake and Ross. The study addressed the way subsequent consideration of an event may influence memory for a population predisposed to stereotyping.

Boon and Davies noted the irony that people altered a study of rumor through rumors. Their work has suffered a similar fate.


The Allport and Postman, Boon and Davies studies were distorted over time, likely owing to some of the same dynamics Allport and Postman aimed to study.

“In a classic study on race, psychologists staged an experiment with two photographs that produced a surprising result.”

icon False

The study used line drawings, not photographs. There were more than two line drawings involved in the study. The results of the study didn’t match those the author, John Blake, apparently found surprising. The study wasn’t about race, but the effect of stereotyping on memory. Was the study a classic on something besides race? Maybe. But it’s a stretch to credit Blake for that potential grain of truth.

“They showed people a photograph of two white men fighting, one unarmed and another holding a knife.”

Again, no photographs were used. The statement is true to the extent that one of the line drawings showed two white men confronting each other while one of the two wielded a knife.

“Then they showed another photograph, this one of a white man with a knife fighting an unarmed African-American man.”

We detect no evidence that Blake is trying to describe the recognition phase of the Boon and Davies experiment. Otherwise, we’d apply charitable interpretation and call this statement from Blake true. But we believe Blake is trying to describe the same experiment Ross describes, and Ross’ version has no hint of the recognition phase. It’s true that the Boon and Davies experiment showed, to at least some of its participants, a picture of a white man with a knife confronting an unarmed black man.

“When they asked people to identify the man who was armed in the first picture, most people picked the right one.”

True Statement Missing Context

Blake probably fails to realize the experiment used four different groups of people, but he’s right that most people correctly identified the armed man in the picture of two white men confronting each other. They picked the right white guy 67.9 percent of the time. That percentage dropped to 58.6 percent of the time when the recognition phase preceded the recall phase.

“Yet when they were asked the same question about the second photo, most people — black and white — incorrectly said the black man had the knife.”

icon False

Only 29.4 percent identified the wrong person as the knife-wielder when the white man confronted the black man with the knife, and Boon and Davies did not record the responses of blacks and whites in their various groups. Most, 52.9 percent, picked the right one. But in the group where the recognition phase preceded the recall phase, 55.5 percent picked the wrong one and only 33.3 percent picked the right one. Describing the results of the experiment without describing the recognition phase, as Blake did, counts as misleading.

As mentioned above, when the recall phase preceded the recognition phase nobody incorrectly identified the black man as the knife wielder. Nobody. The Blake and Ross accounts of the experiment greatly mislead on that point.

Supplement: The pathway to distortion?

We found a number of inaccurate versions of the Allport study on the Web. We’re highlighting this one from Journalism and Mass Communications Quarterly since it contains some of the misleading elements we see from Blake and Ross (bold emphasis added):

Similarly, Boon and Davies showed subjects some photos depicting blacks as victims of a robbery with a white perpetrator and other photos with whites as victims. They used race schemata to explain why subjects incorrectly remembered the blacks as perpetrators significantly more often than as victims. Likewise, a classic study by Allport and Postman showed that when a black person is portrayed as the victim of a crime perpetrated by a white person, the black person is most often remembered as the perpetrator instead of as the victim, particularly by whites.


Supplement: The citation in “Everyday Bias”

Lacking a hard copy of Howard Ross’ book, we used Google Books to confirm the reference Ross cited for the knife-fight experiment.



Blake, John. “The New Threat: ‘Racism without Racists’CNN. Cable News Network, 27 Nov. 2014. Web. 30 Nov. 2014.

Allport, Gordon W., and Leo J. Postman. “The Psychology of Rumor.” Google Books. Google, n.d. Web. 30 Nov. 2014.

Staples, Brent. “Even Now, There’s Risk in ‘Driving While Black’.” The New York Times. The New York Times Company, 14 June 2009. Web. 30 Nov. 2014.

Treadway, Molly, and Michael McCloskey. “Effects of Racial Stereotypes on Eyewitness Performance: Implications of the Real and the Rumoured Allport and Postman Studies.”  Applied Cognitive Psychology 3.1 (1989): 53-63. Wiley. Web. 30 Nov. 2014.

Treadway, Molly, and Michael McCloskey. “Cite Unseen: Distortions of the Allport and Postman Rumor Study in the Eyewitness Testimony Literature.” Law and Human Behavior 11.1 (1987): 19-25. Springerlink. Web. 01 Dec. 2014.

Ross, Howard J. “Like Water for the Fish.” Everyday Bias: Identifying and Navigating Unconscious Judgments in Our Daily Lives. London: Rowman & Littlefield, 2014. 88. Web. 01 Dec. 2014.

Ross, Howard J. “Howard Ross: “Everyday Bias: Identifying and Navigating Unconscious Judgments.” Talks at Google. 01 Dec. 2014. Web. 01 Dec. 2014.

Allport, Gordon W. The Nature of Prejudice. Cambridge, MA: Addison-Wesley Pub., 1954. Web. 01 Dec. 2014.

Boon, Julian C., and Graham M. Davies. “Rumours Greatly Exaggerated: Allport and Postman’s Apocryphal Study.” Canadian Journal of Behavioural Science/Revue Canadienne Des Sciences Du Comportement 19.4 (1987): 430-40. APA PsycNET. Web. 01 Dec. 2014.

Grimes, T., and R. Drechsel. “Word-Picture Juxtaposition, Schemata, and Defamation in Television News.” Journalism & Mass Communication Quarterly 73.1 (1996): 169-80. Web. 02 Dec. 2014.


  1. Mark

    Great treatment of the subject. I read Blake’s article and, intrigued, tried to find the study myself online. I couldn’t but came across your fact check. Very much appreciated!

    1. Bryan W. White (Post author)

      Thank you, Mark, and thanks for reading. Please share.

  2. Howard Ross

    Thanks so much for doing such a thorough job on this. I have to admit to being a neophyte on Twitter so I never saw your inquiry, but I appreciate the correction and will be sure to excise the study from future editions of the book. In the process of looking through literally hundreds of studies I obviously made a mistake about this one! Again, thanks so much for the correction! And please keep up the great work you are obviously doing!

  3. cjknecht

    Thanks. I too attempted to fact check Blake’s claims and could not locate this “study,” albeit I did come to your website. Thanks again.


Leave a Comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.