A number of years ago my horse was standing in the cross-ties having his foot soaked for a suspected abscess. Something spooked him and he pulled back, taking the tub with him and flooding the groom stall. This precipitated a sideways leap onto the staff person who also leapt sideways, caught her pony tail on the tack cleaning hook, and remained bound beside by a panicking horse who was convinced the rubber tub and encroaching water were lethal.
Someone came to the rescue, detangled the staff member from the tack hook, and my horse was left with a terror of anything that happened in enclosed spaces where water was involved. With much Positive Reinforcement training he eventually tolerated wash stalls, but he never really relaxed and held on to this bad experience for as long as I owned him. (I don’t know about the staff person; possibly she still looks at tack cleaning hooks with apprehension).
Out on the plains you don’t want to be a slow learner, or you may not be around for the next lesson.
From an evolutionary point of view, it makes sense for horses to remember bad things, to learn about them quickly (unlike most of the things we train horses to do that require many repetitions to acquire, fear responses may be learned in a single trial), generalize broadly (if this groom stall is a bad place, undoubtedly any enclosed space is also bad), and retain that information for a lifetime. Out on the plains you don’t want to be a slow learner, or you may not be around for the next lesson.
But do horses have an equally impressive memory for good things that happen to them? A burgeoning field in equine science about horses’ memory for humans and horse-human interactions suggests that they do.
Horses remember faces
Horses are known to have sophisticated social cognition, recognizing and remembering the relationship, relative status, and varying degree of social affiliation for each and every member of their group. For example, after a one-year separation, stallions were immediately able to distinguish and sort their mares from a much larger herd of mares (Feh, 2005). Horses are also able to recognize and remember recorded whinnies of familiar herd members and distinguish these whinnies from those of horses they do not know (Lemasson, et al. 2009).
Recent research on equine face recognition indicates that horses’ social cognition skills extend across species – namely with humans. Horses, having been introduced to a novel person only through a photograph of their face, are able to later identify that person in real life, and respond to that person (with approach or avoidance) according to the emotion that was portrayed on the photograph in the original exposure (either joyful or angry). (Proops et al., 2018; Trösch et al., 2020).
Sherril Stone (2010) found that not only could horses learn to distinguish between photographs two sisters, but counter to the authors’ predictions, they could outperform humans in differentiating between identical twins! They were then able to apply this information to a real-world setting, choosing to spend more time with and offer more nuzzling toward the person whose photograph had been associated with receiving a food reward.
Horses are also able to reverse this process, using the information they know about a real live person to a two-dimensional image. Lansade and colleagues (2020) found that horses could distinguish – and show a preference for – a photograph of a known caretaker over an unknown person, even when they had not seen that caretaker for over six months. They successfully chose the target face even when it was presented in profile, in black and white, with eyes hidden, or with a changed hairstyle (2020).
Horses’ ability to recognize and respond consistently to photos of human faces is particularly striking since normal recognition cues (odour, gestures, behaviour, or depth of visual information) are unavailable. Even dogs, who do very well on facial recognition of live people, perform quite poorly on these tasks (Mongillo et al., 2017). These studies suggest that horses do not simply learn to discriminate between two abstract two-dimensional images (one that brings a reward and one that does not), but that they process the information as a real human face that corresponds to a live person.
Horses remember places
Whishaw and colleagues (2021) suggest that animals process information related to safety, personal needs and their position within their environment according to a dual processing theory: Allocentric – a spatial map that remains relatively unchanging such as the riding arena, and Egocentric – the objects within the space and the animal’s position within that space, which changes often. They note that a horse in a familiar environment still needs to go back and reinvestigate the space every day because even though the space has not changed, the objects within it (such as dung, pylons, or jumps) may well have changed.
We often allow horses the opportunity to view new surroundings, but seldom allow them to sniff.
These researchers allowed ridden horses a free rein to explore a variety of objects placed in the arena (straw, fur, paper, clumps of debris and dung from other horses) before their regular training session began. Horses sniffed all the objects once, but once sniffed did not go back to explore them even when given another opportunity after the training session one hour later. However, they did reinvestigate the same objects the following day, seemingly to readjust and update their environmental representation.
The authors note that we often allow horses the opportunity to view new surroundings, but seldom allow them to sniff. Sniffing, however, enables the horse to update their spatial frame of reference from a horse-centred perspective, rather than the purely visually based human-centred perspective. The authors were sufficiently convinced of the value of this pre-training sniff that they now allow their own horses to sniff their environment as part of their training and show warm-up routine.
Horses remember human interfaces
Sankey and colleagues (2010) trained 20 yearling horses to remain motionless while undergoing various handling procedures (picking up feet, fitting a surcingle, taking temperature, etc.) with either positive reinforcement (PR: rewarding desired behaviours with treats) or no reinforcement. Compared to horses trained with no reinforcement, PR horses learned the tasks more quickly, held their immobile stance longer, and were more likely to remember these commands six months later with no training in between.
When faced with a new, unfamiliar handler introducing a novel procedure (paste worming) eight months later, PR horses responded cooperatively. Control horses took twice as long to be wormed and exhibited numerous evasive behaviours – lifting their heads, backing up, and pushing into the handler. PR horses apparently recalled the specific bond they had developed with their trainer, and used this knowledge to build and maintain a longer-term memory of humans in general.
In another study Sankey’s group trained 21 ponies to back up over a very brief training period (one- to three-minute sessions over five days) using either PR (treats) or negative reinforcement (NR: presenting a whip in front of the pony which was lowered when the pony took a step backward) (2010b). Compared to NR ponies, PR ponies were more likely to seek out and spend more time with the trainer, sniff and nibble them, and retain this “humans are terrific” perspective five months later without intervening training.
Conversely, NR trained ponies showed elevated heart rates during training, did not seek human contact, and showed more negative behaviours including laying back ears, head tossing, and hollow high neck. These differences are particularly astounding when one remembers that the five training sessions (of 1–3 minutes) amounted to a maximum of 15 minutes of interaction.
Sankey’s results might prompt us to question our current training practices which are almost exclusively based in NR. Not to be confused with punishment, NR involves applying a mildly unpleasant stimulus (such as leg pressure) which is released when the horse guesses the appropriate response (moving forward). As Sankey notes, “Positive reinforcement training experience created a deep, long-lasting trace in the ponies’ memories … [that] not only influences learning and memorization of the task itself, but also has a major impact, both on the short and long terms, on the animals’ perception of humans” (2010b).
Horses remember you
Research over the past 20 or more years continually supports the fact that horses’ cognitive skills are probably much more advanced than we previously believed, and that they have excellent short and long-term memories. Evelyn Hanggi (2009), for example, found that horses who learned a concept or category rule (eg. always choose the smallest object, or always choose the round object) can apply this same rule with unfamiliar objects up to 8 and 10 years later with no practice in between and no deterioration in performance.
Not only do horses remember lessons learned (both intentional and unintentional – recall the groom stall train wreck) for a very long time, like Santa Claus, they remember who was naughty and who was nice. Since they are such incredibly good sports, they will undoubtedly continue to perform for us even when we are not nice, but be assured that they know the difference between the person who is consistently associated with a good experience and the one who is not.