Roses are red, violets are blue. Except they’re not. They’re, er, violet.

True blue flowers are exceedingly rare, and not for lack of effort. Plant breeders have repeatedly tried to nudge roses and chrysanthemums into blueness, but doing so is really hard (at least, without the use of dyes). These flowers get their colors from pigments called anthocyanins, which typically look pink or red. A flower must chemically tweak these pigments to make them bluer, and even if they did, the results are essentially purple.

Only a few flowers, like cornflowers and Himalayan blue poppies, have achieved true blue, and all by using special chemical tricks like adding metals to their pigments, or making their petals more alkaline. “All of this is chemically quite difficult and not many species have evolved the enzymes to do it,” says Beverley Glover from the University of Cambridge. “Even with genetic modification, people have managed to make purple, bluish roses, but true blue isn’t happening.”

So imagine her surprise when she found that many flowers have secret blue halos in their flowers.

The halos are rings at the bases of the flowers’ petals. Sometimes, they’re visible to us, especially if the petals are dark. But in most cases, they’re so faint that we can’t see them. Glover only detected them with the aid of laboratory equipment. And yet, they’re there—and they’re visible to bees, whose eyes are more sensitive to blue wavelengths of light than ours are.

Glover discovered the first of these halos back on a sunny day in 2009. While walking through the Cambridge University Botanic Garden, she came across Hibiscus trionum—a white flower whose petals have dark-red bases. And those bases, Glover noticed, were shiny. If you moved them around, they would take on blue, yellow, or green metallic sheens on top of the red undercoats.

The cells on the surface of these petals are mostly flat and smooth, but those at the base have many microscopic ridges, like the grooves on a vinyl record. When light hits each ridge, part of it reflects away, and the rest passes through to the other ridges. And because the ridges are regularly spaced, the reflected beams amplify each other to produce colors that are unusually vivid from certain angles. That’s iridescence. It’s color produced not by chemical pigments, but by microscopic structures. “We wanted to know how common this was,” Glover says. “Was this just one or two weird plants—or more?”

More, it turned out. Her team, including Edwige Moyroud and Tobias Wenzel, found at least 11 families of flowers in which at least one species has the same microscopic ridges, and the same iridescent halos. At first, they were puzzled because the iridescence is inconsistent—strong in some species, but weak in others. But weak or strong, it always has a blue component. “If we had just looked at one plant, we wouldn’t have thought anything of it,” Glover says. “It’s only when we looked at a dozen species, and they were all making the same blue, that we thought: Maybe it’s the blue we should be looking at.”

And when they took a closer look at the petals producing the blue halo, they found something stranger.

To create a vivid sheen, those microscopic ridges should all have the same dimensions, and be equally spaced apart. But biology is messy, and plants can’t manufacture structures to such exacting specifications. So the actual ridges come in a variety of heights, widths, and spacing. Glover’s team found that the degree of this variability, which they term “disorder,” is the same across flowers. And for reasons that are still unclear, this particular amount of disorder scatters blue light at specific angles away from the petals. Hence the halo.

The team tested this by creating artificial ridges of their own, with varying degrees of disorder built in. If the ridges are massively disordered and very different from each other, they don’t produce any interesting optical effects. If they’re all the same, they’re iridescent, but there’s no blue scattering. But in the Goldilocks zone, where the ridges are disordered but not too disordered, they create a blue halo. “It’s a different way of making blue,” Glover says.

If the halo lies over a black, purple, or red base, it’s possible for humans to see it. If it lies over white or yellow, we can’t. But Glover’s team showed that bumblebees can see the pattern no matter the background, and in lab conditions, they find flowers more quickly if they have a blue halo.

“When the same trait evolves over and over again, it is strong evidence that it’s adaptive,” says Lena Hileman from the University of Kansas. And since the blue halo is “very widespread in flowering plants, that suggests it is a reliable signal to pollinators about the nectar reward of the flower.”

But “being able to perceive the pattern doesn’t necessarily mean the pollinators actually care about it, or prefer it to simple, pigment-based colors,” says Yaowu Yuan from the University of Connecticut. “The importance of this blue halo to pollinators is still an open question.”