I wonder if this remains true even if the test subject is just carrying a gun, but not holding it in their hands (ie, if it applies to cops with holstered guns):

The researchers varied the situation in each experiment – such as the having the people in the images sometimes wear ski masks, changing the race of the person in the image or changing the reaction subjects were to have when they perceived the person in the image to hold a gun. Regardless of the situation the observers found themselves in, the study showed that responding with a gun biased observers to report “gun present” more than did responding with a ball. Thus, by virtue of affording the subject the opportunity to use a gun, he or she was more likely to classify objects in a scene as a gun and, as a result, to engage in threat-induced behavior, such as raising a firearm to shoot.

Science Daily: Holding a Gun Makes You Think Others Are Too, New Research Shows

I’ve seen some skeptical comments questioning whether firearm training would reduce this effect. But Scientific American reports in its coverage of this study that 1/4 of all police shooting involve unarmed suspects. Police receive more than a little (though some argue not enough) firearm training.

The AP also covered the story and notes: “Past research suggests that people can be more likely to perceive a poorly seen object as a gun if it’s held by a black person than by a white person, experts say.”

See also: Priming.