TLDR – Research shows that horses retain long-term memories of learned concepts, familiar people, and human emotional expressions. Their learning relies on classical and operant conditioning, with timing, individual temperament, and prior experience shaping how effectively they respond.

Calm conditions and precise reinforcement support learning, while stress, delayed cues, flooding, and dominance-based methods can create confusion, fear, welfare problems, and safety risks. Understanding these principles can improve training and human-horse relationships.

Ask most horse people whether horses have good memories and you will get an emphatic yes, usually followed by a story about a horse that remembered exactly where it spooked three years ago, or one that recognised a former owner after a decade apart. The folk wisdom on this turns out to be well-founded. But the science behind equine memory and learning is considerably more detailed, more nuanced, and more practically useful than most of us realise.

How horses learn, what they retain, and how emotional experience shapes their behavior over time have been studied systematically for several decades. The answers matter not just as interesting facts but because they have direct implications for how we train, handle, and build relationships with horses. Getting it wrong creates lasting problems. Getting it right changes everything.

What Horses Remember, and for How Long

Horses have genuinely impressive long-term memory capabilities that extend well beyond procedural skills like responding to leg aids or walking through a familiar gate.

A 2025 systematic review of equine cognition research published in Animal Cognition, which assessed studies across discrimination learning, categorisation, concept formation, spatial cognition, and memory, summarised the long-term memory findings from Hanggi and Ingersoll’s foundational experiments. Across three separate experiments with different retention intervals, horses showed no meaningful decline in performance after six years for learned discrimination tasks, after ten years for categorisation rules, and after seven years for the application of a size-based concept. In each case the horse applied previously learned rules not just to familiar stimuli but to entirely novel ones, demonstrating that the retained memory remained flexible and generalisable across new situations.1 These findings remain the benchmark for long-term equine memory and have been replicated in spirit across subsequent cognition research.

Memory for people specifically is well established. A 2020 study published in Scientific Reports tested whether horses could identify familiar humans from photographs of faces alone using a computer-controlled screen. Eleven horses were trained to select from pairs of photographs, then tested with images of their current keeper alongside unfamiliar faces. They selected their keeper’s face significantly above chance. Crucially, when tested with a keeper they had not seen in six months, performance was at least as strong as it was for their current keeper. The researchers concluded that horses have advanced face recognition abilities and a long-term memory for specific human faces.2 The practical implication is direct: your horse is not simply responding to your smell or your footsteps. They are forming and retaining a visual representation of you across extended periods of separation.

Perhaps the most striking memory research concerns emotional experience. Researchers from the Universities of Sussex and Portsmouth showed horses either an angry or a happy photograph of a specific person, then had that same person approach the horse hours later wearing a neutral expression. Horses responded measurably more negatively to people whose angry faces they had seen earlier that day, even though those people were now expressionless and behaving identically to the positive-expression group. The study concluded that horses not only read emotional facial expressions but retain a memory of a specific person’s emotional state and apply it when they encounter that person again.3 This was the first demonstration of this kind of cross-temporal emotional memory in any non-human mammal.

The implication for everyday horsemanship is significant. The way you are with your horse on any given day becomes part of how your horse understands you. They are not simply reacting to who you are in this moment. They are integrating who you are now with who you have been before.

How Horses Actually Learn

Understanding how horses learn is where research most directly meets training practice, and where some of the most consequential misunderstandings between humans and horses tend to arise.

Horses are associative learners. The two main mechanisms are classical conditioning, in which horses form associations between stimuli, and operant conditioning, in which behaviours are shaped by their consequences. Most trained horse behaviours involve operant conditioning in some form, even when trainers do not consciously think of it that way. Pressure and release, reward at the right moment, repetition that builds a habit: all of these are operant conditioning in practice.

A 2023 study published in Anthrozoös examined the relationship between equestrian knowledge of learning theory, horse welfare, and rider safety among Australian riders. The study found that knowledge of learning theory was low across the board, and that training and riding directly affects horse welfare in ways closely tied to accident and injury risk. The authors noted that equitation scientists have consistently advocated for training methods based on a sound theoretical understanding of learning theory principles, including operant and classical conditioning, as the most appropriate scientific foundation for how we work with horses.4 This is not an abstract academic argument. It has observable consequences for the horses in our care.

The critical practical point is timing. When a consequence follows a behaviour too slowly, the association the horse forms may not be the intended one. When a release of pressure is delayed even briefly past the moment of the correct response, the horse is learning something different from what the trainer thinks they are teaching. Applied incorrectly, the same tools produce confusion, anxiety, and behavioural problems that can become deeply entrenched over time.

Research has also demonstrated consistently that stress impairs learning in horses. A study published in PLOS ONE investigated how acute stress affected the acquisition and reacquisition of an instrumental learning task in horses. Horses stressed after an acquisition session did not improve their performance in a reacquisition session eight days later, unlike their non-stressed counterparts. Temperament further shaped this effect: fearfulness specifically impaired performance under stressful conditions.5

A horse that cannot learn in a given session is not being stubborn. Something about that session, whether the level of pressure, the environment, the horse’s physical state, or the cumulative stress load, has taken them past the threshold where learning can reliably occur. Pushing harder does not produce faster learning. It produces stronger fear associations, and those are remarkably durable.

The Problem with Dominance-Based Frameworks

One of the clearest contributions learning science has made to horse training is the challenge it poses to dominance-based frameworks that remain widespread in equestrian culture.

The idea that horses organise themselves into strict linear hierarchies, that trainers must establish themselves as dominant leaders, and that behaviour problems stem from a horse testing its handler has been substantially undermined by both ethological and learning science research. The Luke et al. study cited above found that most equestrians, including qualified coaches, had limited understanding of the learning theory principles underpinning what they do day to day.6 Training built on dominance assumptions rather than learning science creates welfare problems and safety risks because it consistently misidentifies the source of behaviour problems.

What the learning science shows is that most equine behaviour problems, including aggression, evasion, apparent disobedience, and spookiness, are more accurately understood as the predictable outputs of learning systems responding to stimuli in their environment. A horse that bucks when asked for canter is not asserting dominance. It is producing a behaviour that has been associated in its nervous system with escape from pressure, pain, or confusion. That is a learning history problem, and it calls for a learning-based response.

The International Society for Equitation Science has articulated training principles grounded in this research, including avoiding flooding techniques that force a horse to endure aversive stimuli, minimising delays in reinforcement, and recognising that incorrect use of operant conditioning leads to serious behavioural problems including aggression, escape behaviours, and compromised welfare.7 These are not idealistic recommendations. They are conclusions drawn from a substantial body of empirical research into how associative learning systems actually work.

What This Means for Your Horse

The practical implications of equine memory and learning research come down to a few principles that apply regardless of discipline or experience level.

Your horse remembers more than you think, more specifically than you think, and for longer than you think. Every interaction is part of a cumulative record that shapes how they experience you and what they expect from you. Consistently positive, clear interactions build a relationship your horse will carry into every future session. Consistently aversive, confusing, or stressful interactions build associations that also carry forward and that will need to be patiently and systematically overwritten.

Your horse learns best when calm and under threshold. Sessions where a horse is chronically over-aroused are neurologically inefficient at best and actively counterproductive at worst, building associations with exactly the emotions you are trying to move away from.

Timing matters more than most people realise. The moment of reward or release is the moment that defines what is being learned. When that moment is clear, consistent, and prompt, horses learn efficiently. When it is delayed or inconsistent, they still learn something, but it may not be what you intended.

Finally, individual differences between horses are real and significant. Breed, age, sex, prior experience, management conditions, and temperament all shape how a horse learns and what they find easy or difficult.8 The horse in front of you is not a generic member of the species. They are an individual with a specific learning history, specific associations already formed, and specific capacities and limitations. Understanding the science is useful. Applying it thoughtfully to the actual horse in your care is the real skill.

The Takeaway

Horses are not simple stimulus-response machines, and they are not enigmatic creatures who can only be understood through feel. They are cognitively sophisticated animals whose learning and memory systems follow discoverable principles, respond predictably to specific conditions, and carry the weight of their entire relational history into every interaction.

The science of equine memory and learning does not replace good horsemanship. It gives good horsemanship a foundation to stand on. Understanding why certain approaches work, why certain horses struggle, and why your history with your horse matters as much as what you do today makes you a more effective, more humane, and more informed partner for the animals in your care.

Endnotes

  1. Claire Ricci-Bonot et al., “A Systematic Review on the Effect of Individual Characteristics and Management Practices on Equine Cognition,” Animal Cognition 28, no. 1 (2025): 96, https://doi.org/10.1007/s10071-025-02016-2. Full text: https://pmc.ncbi.nlm.nih.gov/articles/PMC12657588/. ↩︎
  2. Léa Lansade et al., “Female Horses Spontaneously Identify a Photograph of Their Keeper, Last Seen Six Months Previously,” Scientific Reports 10 (2020): 6302, https://doi.org/10.1038/s41598-020-62940-w. Full text: https://pmc.ncbi.nlm.nih.gov/articles/PMC7156667/. ↩︎
  3. Leanne Proops et al., “Animals Remember Previous Facial Expressions That Specific Humans Have Exhibited,” Current Biology 28, no. 9 (2018): 1428–1432, https://doi.org/10.1016/j.cub.2018.03.035. Author accepted manuscript freely available: https://pure.port.ac.uk/ws/files/9254379/ProopsMcComb_Post_print_version.pdf. ↩︎
  4. Karen L. Luke et al., “Does a Working Knowledge of Learning Theory Relate to Improved Horse Welfare and Rider Safety?” Anthrozoös 36, no. 4 (2023): 703–719, https://doi.org/10.1080/08927936.2023.2166713. Full text: https://www.tandfonline.com/doi/full/10.1080/08927936.2023.2166713. ↩︎
  5. Mathilde Valenchon et al., “Stress Modulates Instrumental Learning Performances in Horses (Equus caballus) in Interaction with Temperament,” PLOS ONE 8, no. 4 (2013): e62324, https://doi.org/10.1371/journal.pone.0062324. Full text: https://pmc.ncbi.nlm.nih.gov/articles/PMC3633893/. ↩︎
  6. Ibid Luke ↩︎
  7. International Society for Equitation Science, “ISES Training Principles,” accessed September 1, 2026, https://www.equitationscience.com/ises-training-principles. ↩︎
  8. Ibid Valenchon ↩︎