Most racetracks in America keep a screen. It is a few panels of fabric on a portable frame, often green, and it is kept close enough to the track that a crew can have it standing within minutes. Most fans will never notice it. The ones who do learn quickly what it means when it starts to move, because the screen goes up for exactly one reason: so that the grandstand does not have to watch what happens next to a horse who cannot stand up.
On the second afternoon of the new Belmont Park, the screen went up. Belmont reopened on September 18 after a three-year reconstruction reported at more than half a billion dollars, financed largely by a $455 million loan from New York State, to a sold-out crowd of six thousand (Finger Lakes 1, 2026). The next day, a mare named Heat Missile — the betting favorite — fractured both sesamoid bones in her left front leg and dislocated the fetlock joint at the three-sixteenths pole. She was euthanized on the track, and workers screened the view (Kail, 2026; New York State Gaming Commission, 2026). Six days later, a horse named Main Thing injured his right hind leg after the fifth race and was euthanized on the course (New York State Gaming Commission, 2026). On September 28, PETA asked the New York State Gaming Commission to suspend racing on Belmont's inner turf course, where it says both horses went down — two dead, by its count, in nine races on that surface (PETA, 2026).
The state's own ledger is longer. The Gaming Commission publishes every equine death on New York grounds, and in the first three weeks after the grandstand opened it recorded five at Belmont: Heat Missile and Main Thing in races; a horse named High Tide Honey, who fractured the same two bones as Heat Missile galloping out after a workout on October 7; and two more horses, Kadena and Freeze, lost to colic and to pneumonia complicated by laminitis (New York State Gaming Commission, 2026). Two days before the reopening, a horse named Punchmaster had already been euthanized after breaking a cannon bone on the training track.
So — what are we to make of a list like that? The racing industry has a ready answer, and it deserves to be heard in its best form before anyone answers it. Catastrophic injury rates in American racing have fallen for years. The Jockey Club's Equine Injury Database put the 2025 national rate at 1.07 fatalities per 1,000 starters, the lowest since the database began in 2009 (Vespe, 2026). New turf courses take time to settle. Two racing deaths in a small number of races is a thin basis for statistical conclusions about a surface, and a barn of several hundred horses will lose some to colic and pneumonia whether or not anyone ever races them. All of that is true, and none of it should be waved away.
But notice what the reassurance presupposes. It does not claim that horses stop dying. It claims that they die at an acceptable rate — and 1.07 per 1,000, across the roughly 235,000 starters the database counted in 2025, is about 250 horses (Vespe, 2026). The debate the industry wants is a debate about the price. The question underneath it — whether there should be a price at all, and who agreed to pay it — is the one the screen is there to keep out of view.
Here is where the psychology becomes interesting, because racing is not an industry of people who are indifferent to horses. It is, by and large, an industry of people who love them. Backstretch workers live a few steps from the stalls. Grooms grieve the horses they lose. Fans know bloodlines by heart. And yet the sport, by its own accounting, kills hundreds of its athletes every year. That combination — sincere affection for animals alongside participation in practices that harm them — is precisely the structure that moral psychologists have spent the last decade mapping under the name of the meat paradox. The motivational account is that the conflict is uncomfortable, and that people resolve it less by changing what they do than by changing how they think about what they do: minimizing the harm, denying the animal's experience, or diffusing their own responsibility for it (Bastian & Loughnan, 2017).
The tools are familiar. When people are asked to justify eating meat, their reasons cluster into four families: that it is natural, normal, necessary, and nice (Piazza et al., 2015). Racing has its own translation of every one. Horses were born to run. They have been racing on Long Island since the 1600s. Without racing, there would be no thoroughbreds, and no money to care for them. Have you ever seen anything as beautiful as a field turning for home? Each of these is partly true, which is exactly why each of them works. A rationalization does not need to be false. It needs only to end the conversation before it reaches the horse.
Then there is the language, which does a great deal of quiet labor. Horses are not killed; they are pulled up, vanned off, euthanized on the course. Injuries are catastrophic, a word borrowed from the weather, as though the fracture arrived from somewhere outside the sport. Bandura (1999) described this kind of euphemistic labeling as one of the core mechanisms of moral disengagement — the means by which people with intact moral standards participate in harmful conduct without registering it as harm — alongside the mechanisms that blur who, exactly, is responsible. In racing, the answer is usually the surface. The turf needs time. She took a bad step. Nobody chose this. The ground did.
Indeed, what is striking about racing is that it defeats one of the most reliable effects in the literature on compassion. People respond far more strongly to a single identifiable victim than to statistics about many (Small et al., 2007). Most animal industries are protected by anonymity: no one knows the name of a broiler chicken. But every racehorse has a name. It is printed in the program, called over the loudspeaker, shouted from the rail. Heat Missile was not a statistic; a grandstand watched her break from the gate. And still, the name did not do what names usually do. It was printed beside her odds, which is to say beside a price. The sport has found a way to make an animal fully visible as a competitor and almost invisible as a victim, and the screen is simply the physical form of that trick.
None of this should be oversold. One bad month does not prove that the new Belmont is uniquely dangerous, and PETA's request does not mean the Gaming Commission will act. Some of the deaths on the state's list would have happened in a pasture. Racing nationally is measurably safer than it was a decade ago, and that progress is real. But the reconstruction was paid for with a public loan, and the public, by and large, has been invited to celebrate the building rather than to ask what happens inside it. To the extent that New Yorkers lent the money for that track, the horses dying on that track are not someone else's business.
The moral circle, it turns out, is not a fixed line that some animals are inside and others are outside. It is something people actively draw and redraw, and people differ reliably in how far they extend it (Crimston et al., 2016). Recent work suggests that what pushes the line outward is a felt sense of obligation — the conviction that we owe a being something (Crimston et al., 2026). Racing keeps horses inside the circle as objects of admiration while keeping them outside it as bearers of interests. That is a more comfortable place for a horse to stand than a factory farm. It is not the same as being counted.
What would counting look like? At a minimum, it would mean taking the request seriously: the Gaming Commission has the authority to suspend racing on a surface while it investigates, and it can be asked to use it. It would mean reading the state's equine death list, which is public, rather than waiting for a headline. And it would mean noticing the screen the next time it moves — and remembering that it is not there to protect the horse. Heat Missile was already beyond protection when it went up. The screen is there to protect us, so that the race after hers can still feel like a good day out.
The least we owe her is to look.