It’s “the dress” and “yanny versus laurel” all over again. A new study published by Science found that the context in which we see colors—and, by extension, judge threats—can change our perception of their prevalence.

Researchers at Harvard University and other institutions showed participants 1000 dots ranging in color from very blue to very purple, and asked them to sort the dots into either category. For the first 200 trials, participants divided up the dots pretty evenly: Roughly half were blue and half were purple.

However, as the trials continued, researchers made sure that fewer and fewer blue dots were shown. Participants’ responses didn't align with this change, though. They continued to see roughly the same number of blue and purple dots, and dots they previously labeled purple they now perceived as being blue.

This happened even after researchers warned participants that blue dots would become more infrequent, and that they would be given cash rewards for judging the colors correctly. The study wasn’t just about color, either. Similar results were observed when participants were asked to judge whether faces were threatening or non-threatening, and whether research proposals were ethical or unethical.

Even when the rate of “problems” (threatening faces/unethical proposals) went down, participants continued to see them in areas they previously considered benign. Researchers believe this “prevalence-induced concept change” could explain why so many people take a pessimistic view of the world, even if some of our social ills have lessened over the years.

“Social problems may seem intractable in part because reductions in their prevalence lead people to see more of them,” researchers wrote in their abstract. To learn more about this phenomenon, check out Science’s video.

[h/t Science]