The Verification Venue · pointed at a thing everyone gets wrong
It Was Always the Motion, Never the Red
A bull does not charge because the cape is red. It cannot even see red the way you do: cattle are dichromats, two cone types instead of your three. Red is duller to a bull, not invisible, and here is the part the internet's own correction gets wrong: red is not a bull's weak colour. Blue is. The trigger was never a wavelength. It is the cape's motion.
Everyone knows the folk version: matadors wave red to enrage the bull. Every fact-check blog then serves the same tidy fix: "bulls are red-green colourblind, they cannot see red, it is just the motion." The first half of that fix is real. The second half is itself half-wrong, and that is the interesting part. Below, the render collapses your three cone channels down to a bull's two and desaturates the muleta live. Then you swap the cape's colour, and finally you toggle its motion, the only variable that actually moves the animal.
Cape colour vividness (chroma)
–
human vs bull
Cape vs sand, by colour alone
–
how much it stands out
Cone types
2 not 3
S ≈451 nm · M/L ≈555 nm
Swap the hue and watch the right panel. Red and green both go muddy against the sand. In this standard model the one colour that survives is blue, which is exactly where the model and the real animal disagree. Hold that thought.
A still cape of any colour is nearly inert. Set it moving and the stimulus appears, and it is the same whether the cloth is red, green, white, or black. The hue never entered into it.
The two cones a bull has, and the third it is missing
A Gaussian model placed at the electroretinogram peaks measured by Jacobs, Deegan and Neitz (1998). With no dedicated long-wavelength cone, the red-green chromatic signal that makes a muleta vivid to you is the exact signal a bull is short of, so red arrives desaturated. Not grey. Not gone. Duller.
The correction everyone repeats is itself half-wrong
Red is not even the colour a bull struggles with. Blue is.
The fix on every blog is "bulls are red-green colourblind." That borrows a human term (deuteranopia) and points it at the wrong band. When researchers actually conditioned the real fighting breed to colours (Riol et al. 1989), the cattle discriminated long and medium wavelengths (550–700 nm, red included) well, and did worst on short wavelengths (400–500 nm), the blues and violets. So the honest correction has two layers the fifty explainers miss:
1. Red is desaturated, not erased. The missing third cone flattens red's vividness, it does not turn the cape to static grey. A bull sees a duller red, and can still tell it from the sand by brightness.
2. Even the "colourblind" meme is still arguing about colour. That is the real ruler-twist. Whether you trust the tidy cone model (which flattens red) or the behavioural data (which flattens blue), the verdict is identical: no hue is what triggers the charge. Swap the red for green, white, or black and wave it the same way, and the bull comes the same.
If the question is the colour
Answer both ways and it still fails. The cone model says red is the desaturated one; the behavioural data says blue is the confused one. Red is nobody's trigger under either reading.
If the question is the motion
Now it resolves. The muleta only appears in the final third of the fight; earlier stages use a magenta-and-gold cape. What draws the charge in every stage is the sweep of the cloth, a moving threat, not its wavelength.
Same animal, two honest readings of the colour question, and both send you to the same answer: it was the movement. The colour was a story we told ourselves in a language the bull does not read.
Every cape colour, checked
The table recomputes each swatch through the standard trichromat-to-dichromat projection and reports two numbers: the colour's own vividness (CIELAB chroma) retained, and how far it stands out from the arena sand by colour alone. Red keeps a duller red (chroma about two-thirds) but loses most of its contrast against the sand. Blue is the survivor of the model, the colour that still pops, which is precisely the colour the live animal is actually worst at.
| cape | chroma (you) | chroma (bull) | chroma kept | vs sand (you) | vs sand (bull) | contrast kept |
|---|
The check: every number recomputed in front of you
Nothing here is a stored figure. For the cape you have selected, the page recomputes the colour transform, the chroma, and the cape-vs-sand contrast, live, from the published cone peaks and the standard dichromat matrix:
The offline gate recomputes all of this two independent ways: node research/do-bulls-hate-red/verify-do-bulls-hate-red.mjs. Free choices & uncertainty: the cone peaks are the midpoints of the Jacobs 1998 ranges (S 444–455 nm, M/L 552–555 nm). The render is the Machado, Oliveira & Fernandes 2009 deuteranope projection at severity 1.0, a standard model of losing the third cone; it is a model of reduced saturation, not literal bovine qualia. The arena sand colour (sRGB 194,154,96) and the swatch RGBs are fixed representative choices; the ordering (red dulls and loses sand-contrast, blue survives the model) is robust to reasonable changes.
What's exactly true here, and what's a model
Exactly true (physiology). Cattle have two cone photopigment types, not three: they are dichromats. Electroretinogram measurements put the short-wavelength (S) cone peak near 451 nm (measured range 444–455) and the single medium/long (M/L) cone peak near 555 nm (552–555). With no dedicated long-wavelength cone, the red-green chromatic dimension carries less signal, so a red muleta is a less-saturated red to a bull. It is not invisible and it is not grey.
Exactly true (behaviour, the ruler-twist). Conditioning studies of the actual fighting breed (Riol et al. 1989) found the cattle discriminate long and medium wavelengths (550–700 nm, red included) well and are poorest at short wavelengths (400–500 nm, blue/violet). So "red-green colourblind" as a mechanism is a misapplied human term; a bull's genuinely weak band is at least as much blue as red-green. And the muleta enters only in the final third of the fight; the earlier capote is magenta and gold. The charge tracks the cape's motion, not its hue: a green, white, or black cape waved the same way draws the same response.
A model, not a measurement (the render). The two panels use the Machado, Oliveira & Fernandes (2009) deuteranope simulation matrix at full severity, applied in linear sRGB, then CIELAB chroma and cape-vs-sand chromatic distance are computed. This is the standard way to visualise the loss of one cone; it models reduced discrimination and saturation, not the private experience of a cow. Note the honest tension: this deuteranope-shaped model flattens red while the behavioural data flags blue. Both point to the same conclusion, so we show the render and name the disagreement rather than hide it.
Commonly cited, not a lab result. The idea that red cloth was chosen to hide blood is a frequently repeated tradition, not a measured finding. We mention it as folklore, not fact.
Framing. Bullfighting is ethically and culturally charged. This page is about vision and stimulus, not endorsement.