Support Hyperallergic’s independent arts journalism. Become a Member »

Support Hyperallergic’s independent arts journalism.

Color could impact image-sharing (image via Wikimedia)

Last week, a dress went viral because no one could decide whether it was white and gold or black and blue. But what might have happened if the dress had appeared red, purple, or even pink?

New research suggests that it might have actually broken the Internet. Published this month in PLoS ONE, Red, Purple and Pink: The Colors of Diffusion on Pinterest found that images with those colors tend to be shared much more than images without them.

Researchers Saeideh Bakhshi and Eric Gilbert, from Georgia Tech’s comp.social lab, developed a web crawler to sample 1 million images pinned to Pinterest in the months spanning between 2009 and 2011. They analyzed each image pixel-by-pixel using the Munsell color system to determine its dominant hue, average saturation,‹ and brightness. Next, they applied what’s called “negative binominal regression” to characterize how colors affected repins (while controlling for network structure and activity).

The results revealed that red, purple, and pink promote image sharing online, while green, blue, black, and yellow suppress it. “Although we don’t claim that every image draped in Red or Purple will be shared significantly more, on average, warm and exciting colors seem to affect the recipient’s likelihood of sharing the image,” the authors write. The results also showed that highly saturated images perform well, while black-and-white ones fail to win as many clicks.

As the researchers note, many other studies have shown color can impact human behavior; one 1995 paper published in the Journal of Consumer Research suggested that color accounts for 60% of whether or not we accept an object, place, or person. Studies have also explored what makes things shareable online, with one 2012 study concluding that New York Times readers tend to share stories that inspire awe. But this seems to be the first study that actually connects the two, looking at how color influences online behavior.

And obviously, it’s just a start. Since the data was culled from Pinterest — an image-sharing website we all know to be frequented by crafty DIY-ers, bakers, and their ilk — the study can only tell us what colors such click-happy hobbyists seem to respond to. Had the data come from a site with its own unique demographic like Tumblr, the results might have been different.

Also, the data doesn’t explain why these colors inspire sharing on Pinterest. Are the cookies and cakes in all those pins frosted in mouth-watering, pink icing? Or is purple the latest hip trend in home decor? “Our approach is useful for describing what factors affect [image sharing], but without a corresponding qualitative approach, we can only speculate,” the authors write.

For this study, the important takeaway is that color does appear to impact online behavior — a fact that makes perfect sense, though it’s still always strange to realize just how much are behavior is guided by subconscious instincts we barely realize exist. Just try not to think too much about it next time you’re clicking away.

The Latest

Did Judy Chicago Just Troll Us?

Nowhere in the museums’ advertising blitzkrieg for the performance were we told to bring our wildfire-season masks as well as our covid masks, and covid masks don’t prevent smoke inhalation.

Laura C. Mallonee

Laura C. Mallonee is a Brooklyn-based writer. She holds an M.A. in Cultural Reporting and Criticism from NYU and a B.F.A. in painting from Missouri State University. She enjoys exploring new cities and...

One reply on “Study Finds Color Impacts Online Image Sharing”

  1. This is relevant to a study published last summer out of MIT titled ‘What Makes an Image Popular’ It broke down data about images on Flickr by pageview to find a cross section of commonalities between the most popular images. The paper is very interesting and easy to digest from an artists perspective. Their website also has a tool to tell users by percentage points how popular their images might be based on the algorithms used in the study. http://popularity.csail.mit.edu/

Comments are closed.