The Shirley Card: How Kodak Decided What Skin Should Look Like in Color

In 1978, the Mozambican government hired Jean-Luc Godard to make a short film. He turned down the assignment's own film stock. Kodak's color negative film, Godard said, was racist. He wasn't being provocative for its own sake. Point the film at a mixed-race group and expose for the lighter faces, and the darker ones would vanish into shadow, sometimes down to nothing but the whites of their eyes and teeth. The stock had a default, and that default was whiteness.

The reason traces back to a single item Kodak mailed to photo labs for decades: a reference print of a woman named Shirley, used to calibrate every color photograph printed from Kodak film. For the next forty years, "Shirley" was white. The story of how that happened, and how long it took to change, says something ColorSift keeps coming back to: color is never rendered by a neutral machine. Someone decides what counts as correct, and that decision gets built into the hardware.

What was actually in the box

Photo labs in the mid-twentieth century didn't just drop film into a machine and get prints out. A technician had to calibrate exposure time and color balance for every printing run, and they needed something to calibrate against. Kodak supplied that something: a reference card showing a woman in colorful clothing next to blocks of red, yellow, blue, green, gray, black, and white. Technicians adjusted the equipment until her skin, and the color blocks beside her, printed the way Kodak said they should.

One of the earliest models was a Kodak employee named Shirley Page, and the name stuck to the whole product category the way "Kleenex" stuck to tissues. Labs across the industry, not just at Kodak, called these cards "Shirley cards" regardless of who was actually pictured. Every version, for decades, showed a white woman.

That reference card wasn't a side detail. It set the exposure and color balance that every print run in that lab was measured against, for every roll of film that came through, of every person photographed. If the industry standard for "correctly exposed" was calibrated to pale skin, film calibrated to anything else would look wrong by that same standard, even when it was accurate.

Nobody sat down and chose this

It's tempting to look for the moment someone decided film should favor white skin. Communication scholar Lorna Roth's 2009 paper "Looking at Shirley, the Ultimate Norm," published in the Canadian Journal of Communication, argues that's the wrong way to look for it. Roth traces how a light-skin bias got embedded in film emulsion chemistry and camera design without any single deliberate choice, through what she calls a "technological unconscious": an industry made mostly of white engineers, serving a customer base Kodak's own marketing assumed was mostly white, building tools that measured success against the faces already in front of them.

That's a more uncomfortable story than a single villain, because it means the bias didn't need bad intent to survive. It just needed nobody with the power to change it to notice, or to care enough to ask why. Roth proposes a fix she calls "cognitive equity": deliberately building a wider range of skin tones into a technology's design, rather than treating whatever range shipped first as the natural one.

What the camera actually couldn't see

The Godard incident wasn't an isolated complaint. Filmmaker Adam Broomberg, who researched Kodak's color history for a photography project with Oliver Chanarin, put the technical failure plainly: if you exposed the film for a white subject, a Black subject standing next to them could lose all detail, rendered as a dark shape with barely visible eyes and teeth. This wasn't about a photographer's skill. It was baked into how the film's dyes and dynamic range worked, before anyone picked up a camera.

The effect showed up constantly in ordinary photography: family portraits where one relative disappeared into the background, group photos where half the group looked correctly exposed and the other half looked like silhouettes. Anyone outside the film's calibrated range paid for it in the print, every time.

What finally moved Kodak wasn't the argument you'd expect

Complaints from photographers and filmmakers about racial bias in film stock had been around for years by the 1970s. What actually got Kodak to widen the film's dynamic range came from a different direction entirely: furniture manufacturers and chocolate companies.

Wood furniture advertisers found that Kodak film couldn't distinguish a light-grained wood from a dark-grained one in a catalog photo. Chocolate makers ran into the same wall trying to shoot milk chocolate next to dark chocolate. Both looked flat and similar on film that had never been built to register a wide range of browns. Those clients had advertising budgets Kodak wanted to keep, and that leverage did what years of civil rights criticism about photographing people hadn't managed on its own.

Kodak responded by reformulating its film stock to capture a broader tonal range, later marketed under names like Gold Max, with an ad line claiming the film could photograph the details of a dark horse in low light. The industry needed to sell better chocolate photography before it invested in a wider dynamic range, and better representation of Black and brown skin arrived as a side effect of that investment rather than its purpose.

Shirley gets some company, slowly

Kodak started including Black and Asian models on reference cards in the 1970s, expanding what the "correct" standard included. But a genuinely multiracial reference card, showing white, Black, and Asian women side by side (with a Hispanic model added a bit later), didn't appear until 1995. That's roughly four decades after the original Shirley card, and even then, the darker-skinned models chosen still sat toward the lighter end of their respective ranges. Widening the standard didn't mean throwing it out.

The bias didn't stay in the darkroom

Film stock eventually gave way to digital sensors, and a reasonable assumption might be that a fresh start in silicon would fix an old chemistry problem. It didn't, because the same pattern, technology tuned against whichever faces its builders tested it on, showed up again in new hardware.

In December 2009, a viral video showed an HP webcam with face-tracking software following a white user's face smoothly while failing to track a Black colleague's face at all. HP's explanation was that the algorithm measured contrast between the eyes and the surrounding cheeks and nose, and struggled under lighting conditions where that contrast was lower for darker skin. The company framed it as a lighting edge case. It was also, functionally, the webcam version of the Shirley card: a default tuned to one kind of face, failing outside it.

The same story repeated in 2017, when Chukwuemeka Afigbo, then Facebook's head of platform partnerships for the Middle East and Africa, posted a video of an automatic soap dispenser. It released soap instantly under a white hand and a white paper towel, and released nothing at all under a Black hand held the same way, for over ten seconds. The dispenser used infrared light, triggering when a hand reflected enough of it back to a sensor. Darker skin absorbs more of that light instead of bouncing it back, so the sensor never saw the signal it was built around.

By 2018, the pattern had a number attached to it. Researchers Joy Buolamwini and Timnit Gebru tested three commercial gender-classification systems for their paper "Gender Shades," presented at the first Conference on Fairness, Accountability, and Transparency. The systems misclassified darker-skinned women at error rates up to 34.7 percent. Lighter-skinned men topped out at 0.8 percent. The datasets used to train those systems, like the Shirley card before them, were built mostly from faces that looked like the people building the systems.

Building a different default on purpose

The clearest sign of what a deliberate fix looks like arrived in October 2021, when Google launched the Pixel 6 with a feature called Real Tone. Instead of retuning one algorithm after a complaint went viral, Google worked with more than 60 photographers and image-equity experts of different ethnicities to rebuild the camera's exposure and color decisions around a fuller range of skin tones from the start.

Real Tone is worth noting less as a triumphant ending than as a contrast. Everything before it, the chocolate and furniture complaints, the multiracial card that took forty years to arrive, the webcam and soap dispenser failures, happened because someone outside the design process had to notice the gap and make enough noise about it. Real Tone happened because Google built the correction in before shipping, rather than waiting for the next viral video to force it.

The default is never neutral

None of this is really a story about film chemistry or infrared sensors. It's a story about what happens when a technical standard gets treated as neutral just because nobody wrote down who it was calibrated for. Kodak's engineers weren't trying to build a racist product. They built a product measured against whoever happened to be in the room when they decided what "correctly exposed" meant, and that decision outlived the people who made it.

The Shirley card is gone now, replaced by digital calibration targets and machine learning models trained on datasets that carry their own version of the same question: whose face did you test this against? Every color tool ColorSift has ever covered, from a UI palette to a print process, has a default baked in somewhere. The Shirley card just happens to be the one where you can point to the exact card, the exact woman's name, and forty years of prints that all measured themselves against her.