Researchers in Portsmouth set up a game your dog could win by cheating.

Three cups sat on a rolling board. Two had food under them. The dog watched both get baited, and the dog also watched a stranger sitting across the table see only one of them. All the dog had to do was remember which cup the stranger never saw.

Twenty-two dogs played. None of them used it.

The setup came straight from primate research, run at the Dog Cognition Centre in Portsmouth.

During one of those baitings the curtain dropped in front of the woman, so she saw only a single cup get loaded. The dog watched all of it, including the curtain, and then the two of them took turns picking cups until the food was gone.

That is a lot of chances to notice the pattern in what they see when a human looks away.

When the woman picked first, the logic is simple. She can only pick the cup she saw, so that cup is now empty and the other one is a free meal.

The competition study found no difference at all in how dogs behaved based on who picked first.

They were guessing between the two baited cups. They avoided the empty one, they just never worked out that one of the baited cups had already been cleaned out by somebody who could not have known about the other.

This is the part that rules out the easy explanation.

The researchers ran a control where both players watched every baiting with the curtains open, and the dog could see the woman make her pick.

So the dogs tracked her actions cleanly. Watching a hand go to a cup told them that cup was done. Watching a curtain close in front of her face told them nothing at all.

Memory is not the limit either, since dogs hold the location of a hidden object for at least four minutes in other work.

The same paradigm has been run on apes. A 2008 chimpanzee study found that chimps pick the cup only they saw baited when they are second to choose, and do it more in that situation than when they go first. They model the competitor's ignorance and exploit it.

Human adults and children go a step further. They take the shared cup first, because that clears out the one the competitor was going to grab and leaves the secret one safe for round two. Neither chimps nor dogs run that play, which needs a kind of forward probability math that appears to be ours alone.

Dogs land below chimps on this specific test, and that ranking is exactly where the story gets interesting.

The authors think the dogs may have declined the premise.

Free-ranging dogs compete hard over food, and household dogs guard toys and beds from each other. Competing against a human is a different proposition. When wolves and dogs were put on a cooperative rope-pulling task, the wolves led and treated the person as a peer, while the dogs hung back and waited for the human to start.

That pattern reads as something domestication selected for. A dog that defers over resources is a dog that keeps living in the house, and fifteen thousand years of that is not undone by a plate of salmon kibble under a plastic cup. The trait separates dogs from every other animal that ended up near us, including cats, who were never fully domesticated in the same direction.

The task may have been asking dogs to do the one thing they were bred out of.

There is a second force working against the dogs. Human gesture overrides a dog's own eyes, and it has been demonstrated repeatedly.

In this experiment the woman indicated her choice by pointing. Some dogs likely read that point as help rather than competition, which is the most dog thing imaginable and the reason a quarter of choices still went to a cup they had just watched get emptied.

Forty-two dogs started training. Only 24 finished it, with most dropouts coming from dogs that locked onto one side, refused to choose, or methodically searched every cup like the scent games they play at home.

Stop crediting your dog with reading your mind and start crediting what it actually does.

That last one is worth sitting with. A sloppy point or an absent-minded glance will beat whatever your dog can smell. The things dogs do for their favorite human run on that same wiring.

Twenty-two dogs produced a null result, and the null result is the finding.

Dog cognition research has spent two decades documenting how far dogs outrun chimps at reading us. This is the edge of that talent. Dogs win on attention, gesture, and cooperation, and they lose on modeling what somebody else knows, which is a trade the species appears to have made on purpose.

Your dog is not plotting. Your dog is watching your hands, which is a harder job than it sounds and the reason it never misses the sound of a jar opening three rooms away.

1. Noordenbos J, Beerda B, Layzell H, Kaminski J. Do dogs know what humans know? A study into pet dogs' ability to attribute knowledge to an unfamiliar person. Anim Cogn. 2026;29:10. https://link.springer.com/article/10.1007/s10071-025-02034-0

2. Kaminski J, Call J, Tomasello M. Chimpanzees know what others know, but not what they believe. Cognition. 2008;109(2):224-234. https://doi.org/10.1016/j.cognition.2008.08.010

3. Range F, Marshall-Pescini S, Kratz C, Virányi Z. Wolves lead and dogs follow, but they both cooperate with humans. Sci Rep. 2019. https://doi.org/10.1038/s41598-019-40468-y

4. Szetei V, Miklósi Á, Topál J, Csányi V. When dogs seem to lose their nose. Appl Anim Behav Sci. 2003;83(2):141-152. https://doi.org/10.1016/S0168-1591(03)00114-X