Horses trained to ask for a blanket, and why the design matters most

Whether to rug a horse is a genuine argument among people who keep them. Horses are large, well insulated and evolved for cold, so a blanket that a human would find obviously comforting may be unnecessary or actively unhelpful. The animal has an opinion. There has never been a good way to collect it.

In 2016 Cecilie M. Mejdell, Turid Buvik, Grete H. M. Jørgensen and Knut E. Bøe published a method for doing exactly that, in Applied Animal Behaviour Science. The horses were taught to ask.

The vocabulary

Three symbols, each a neutral visual shape with no natural resemblance to its meaning:

• one indicating blanket on
• one indicating blanket off
• one indicating no change

The arbitrariness is the point. If the symbol for “blanket on” looked like a blanket, a horse choosing it correctly could be responding to a picture rather than to a learned convention. Neutral shapes force the animal to hold a mapping between an unrelated mark and an outcome.

Training was incremental: the horse learns that touching a symbol with its muzzle produces a specific consequence, then that different symbols produce different consequences, then that it can choose between them. Twenty-three horses completed the training.

Why the weather result is the real finding

That an animal can be trained to touch a target for a reward is not news. Pigeons do it. The question that separates a trained response from communication is whether the animal uses the response to report something the trainer cannot already see.

This is where the study earns its place. Once trained, the horses’ choices tracked the conditions. In cold, wet and windy weather they asked for the blanket. In warm sunshine they asked for it to come off. The pattern was not random with respect to temperature, and it was not a fixed preference.

That correlation does two pieces of work at once. It rules out random pressing, and it makes handler cueing much harder to sustain as an explanation, because a horse reading unconscious signals from a person would reproduce the person’s judgement rather than a coherent response to its own thermal state.

The methodological problem this design is fighting

Every study of animal communication with humans has to survive the same objection, and the history of the field is largely a history of failing to.

Animals are extremely good at reading people. A horse in particular is a large prey animal that has spent millennia in close proximity to humans, attending to posture, breathing and attention. If the handler knows the right answer, the animal has a route to it that has nothing to do with the task.

The standard defences are to keep the handler ignorant of the correct answer, to randomise, and to look for patterns that a cueing account cannot generate. The third is the strongest, and it is what this design uses: nobody has proposed a plausible cueing mechanism by which a handler would systematically transmit “it is cold today, ask for the rug” while believing they were transmitting nothing.

What it does and does not establish

It establishes that a horse can acquire an arbitrary symbol-to-outcome mapping, retain it, and deploy it to report a preference that varies with its own condition. That is a real cognitive result with an immediate welfare application: a stable can, in principle, stop guessing.

It does not establish language, syntax, or generalisation beyond the trained domain. The vocabulary is three items about one topic. The broader question of what counts as animal communication, and what separates a signal from information that merely leaks, is the subject of our animal communication explainer.

The practical value is more modest and more usable than the cognitive claim. Anyone who works with horses in a Norwegian winter has to make a rugging decision daily, on incomplete information, for an animal that cannot be asked. This is a protocol for asking.

The rugging argument this was built to settle

The reason a Norwegian group asked this question rather than an American or Spanish one is climatic. In a country where horses are routinely kept outdoors through a cold, wet winter, whether to rug is a decision taken daily and defended vigorously in both directions.

The case for rugging is that a domesticated horse, clipped for work and kept in wet conditions, loses the insulating loft of its coat when the hair mats down, and cannot dry out. The case against is that a horse’s thermoneutral zone extends far lower than a human’s, that rugging suppresses natural coat growth, and that a covered horse cannot shed heat when it moves. Both arguments are sound in the situations they describe, which is precisely why the disagreement never resolves: the correct answer depends on the individual animal, the weather that day, and how much work it is doing.

What a three-symbol vocabulary offers is not a resolution of the argument but a way of bypassing it. Instead of inferring the animal’s thermal comfort from general principles, the handler can ask this animal today. That is a smaller claim than the coverage suggested and a considerably more useful one.