The Thinking Horse: Cognition and Perception · PDF fileThe Thinking Horse: Cognition and Perception Reviewed Evelyn B. Hanggi, MS, PhD Cognition and perception in horses has often

The Thinking Horse: Cognition and PerceptionReviewed

Evelyn B. Hanggi, MS, PhD

Cognition and perception in horses has often been misunderstood. Not only in the past but eventoday, people proclaim that horses react only by instinct, that they are just conditioned-responseanimals, that they lack advanced cognitive ability, and that they have poor visual capabilities (e.g.,acuity, color vision, depth perception). Until relatively recently, there was little scientific evidenceto address such beliefs. Change, however, is underway as scientific and public interest in all aspectsof equine learning and perception intensifies. A review of the scientific literature, as well as practicalexperience, shows that horses excel at simpler forms of learning such as classical and operantconditioning, which is not surprising considering their trainability when these principles and prac-tices are applied. Furthermore, horses have shown ease in stimulus generalization and discrimina-tion learning. Most recently and unexpected by many, horses have solved advanced cognitivechallenges involving categorization learning and some degree of concept formation. A comprehen-sive understanding of the cognitive and perceptual abilities of horses is necessary to ensure that thisspecies receives proper training, handling, management, and care. Author’s address: Equine Re-search Foundation, PO Box 1900, Aptos, CA 95001; e-mail: [email protected]; website: www.equineresearch.org. © 2005 AAEP.

1. Introduction

Traditionally, horses (Equus caballus) have rarelybeen classified as intelligent, and even today, gaps inknowledge, myths and misconceptions, and limited re-search affect how horses are understood or misunder-stood by the public, the horse industry, and even thescientific community. Common beliefs maintain thathorses have a brain the size of a walnut; horses do notthink; horses are merely conditioned-response ani-mals; horses cannot generalize; horses have no senseof concept; horses are colorblind, have poor acuity anddepth perception, and cannot transfer informationfrom one eye to another.

In reality, horses manage not only ordinary dailycognitive tasks but mental challenges as well. In the

wild, they must cope with food and water of inconsis-tent quality or unpredictable distribution, predatorsthat change locations and habits, and a social systemin which identities and roles of individuals must bediscovered and remembered.1 Domesticated horsesmay face even more potentially bewildering condi-tions. In addition to dealing with similar situa-tions encountered in nature, many domesticatedhorses must live in largely unsuitable environ-ments, must suppress instincts while learning tasksthat are not natural behaviors, and must co-existwith humans who sometimes behave bizarrely, atleast likely from an equine standpoint. Horses,both feral and domesticated, are faced with variedconditions that require an assortment of learningand perceptual capabilities.

246 2005 ! Vol. 51 ! AAEP PROCEEDINGS

IN-DEPTH: BEHAVIOR

NOTES

Research is gradually providing insight into thecomplex and fascinating nature of cognition and per-ception in horses. This presentation will reviewlearning theory and key scientific studies to date onthese abilities. The findings are not only interest-ing and useful for the education of clients but alsorelevant to efficient veterinary interaction withhorses. Moreover, this information may be drawnupon to enhance the training and behavior modifi-cation of horses.

2. Habituation and Desensitization

Habituation is a learning process whereby, afterrepeated exposure at the same intensity to an incon-sequential stimulus, an animal becomes accustomedand its reaction to that stimulus either diminishesor disappears. A stimulus may be anything thatincites a response. Although this is an extremelysimple form of learning, it is, nonetheless, extremelyimportant because it allows an animal to subcon-sciously filter out nonvital information in its envi-ronment and focus instead on what is significant.From birth to old age, horses learn through habitu-ation. In a new environment, many horses tend tobe much more reactive, at first paying attention toany and all stimuli perceived by any of the senses.However, when particular stimuli turn out to beinsignificant, the response diminishes.

Desensitization is a process used to extinguish aresponse to a stimulus in a sensitized or hypersen-sitized animal. For instance, a horse may becomeunusually fearful or sensitive to being bridled be-cause of rough handling of its ears or the bit bangingagainst its teeth. This leads to head shyness andthe need for retraining. Such training will be moreproblematic because the horse will attempt to avoidthe bridle; therefore, the handler will need to ap-proach the horse’s head in increments and retreatwhen necessary until the horse willingly acceptsnormal, gentle bridling. This is an example of de-sensitizing a horse that had become sensitizedthrough experience, but this process also works withhorses that are by nature overly sensitive.

Good trainers take advantage of these learningabilities by exposing horses in a positive manner toall sorts of sights, sounds, and contacts (Fig. 1).When done correctly, horses become habituated anddesensitized, even to potentially overwhelmingstressors, and are much more capable of handlingnovel events calmly. It is imperative, however, fortrainers to understand the principles behind thepractice before attempting to desensitize horses.Done incorrectly, for example, removing an alarm-ing stimulus too soon, can have the reverse effect—the horse becomes sensitized and more fearful.Sensitization, nonetheless, cannot be overlooked, be-cause it is the means of achieving responsivenessand lightness to riding and handling aids.

3. Classical and Operant Conditioning andReinforcementHorses, like most organisms, learn effortlesslythrough classical or Pavlovian conditioning, whichoccurs when initially unimportant stimuli or eventsare regularly paired with stimuli that initiate sometype of response. As a result, a new stimulus–re-sponse association forms in which an animal gives aresponse upon presentation of the original inconse-quential stimulus (now the conditioned stimulus),even though the unconditioned stimulus is no longeravailable. Under this condition, the animal’s be-havior does not affect the occurrence of later events.

Horse trainers use classical conditioning regularlywhen they place a word onto a behavior, such aspairing the initially meaningless word “trot” withthe flick of a whip (previously associated with invok-ing a flight response or pain) immediately before thehorse changes gait during an upward transition.Done consistently, it does not take long before thehorse responds with the appropriate action whengiven only the verbal cue.

Researchers, likewise, employ classical condition-ing to facilitate their experimental methodology.Hanggi2 uses the verbal conditioned stimulus “good”to indicate correct responding during cognition test-ing, thus informing the horse that a food reinforceris forthcoming.

To many a veterinarian’s dismay, horses quicklylearn that the sight of a syringe is associated withpain or discomfort. The unconditioned response(escape) then occurs whenever a horse that has notbeen trained to accept such handling catches sight ofa syringe.3

Fig. 1. This horse is being positively exposed, without restraint, toa variety of visual, auditory, olfactory, and tactile stimuli. Stimuliinclude a bouncing beach ball, multicolored shiny mylar balloons,a bobbling umbrella, a plastic tarp, pylons, a barrel filled withrocks dragged rapidly around the horse, and strangely behavinghumans. Photo credit: Evelyn B. Hanggi.

AAEP PROCEEDINGS ! Vol. 51 ! 2005 247

IN-DEPTH: BEHAVIOR

Stabled horses learn sounds associated with feed-ing, such as the opening of hay room doors or grainpoured into a bucket, or recognize visual cues, suchas the arrival of a feed person or vehicle. Within ashort time and to the frustration of owners, theydisplay anticipatory behaviors, e.g., vocalizing, paw-ing, or kicking stall doors, which, when reinforced,become conditioned behaviors.

Unlike classical conditioning, during operant con-ditioning, an animal manipulates its environment toobtain reinforcement—either positive (receivingsomething desired) or negative (removing somethingunpleasant). Both types strengthen the relation-ship between a stimulus and the sought after re-sponse, so that when the same stimulus is givenagain, there is an increased chance that the animalwill repeat the response.4

When a horse begins to learn the meaning of anew stimulus, it will respond randomly, and throughtrial and error chance on the desired response.Reinforcement of the response at the correct mo-ment will cause the animal to repeat the behavior,albeit imperfectly at first. Horses excel at this typeof learning, especially when positive reinforcementis available. For example, when presented with anenvironmental enrichment device called a foodball—which dispenses pelleted feed as it rolls around onthe ground—horses will approach and investigate.5Because of olfactory cues, they push the foodballwith their noses or legs, causing it to roll and droppellets. Most horses rapidly learn to manipulatethe foodball in this manner, thereby receiving rein-forcement and gaining control over this element oftheir environment. It is believed that such controlis important with regard to animal welfare, andmore attempts should be made to allow stabledhorses some degree of instrumental control overtheir surroundings.1

A good understanding of positive reinforcement isvery useful in working with equids. For instance,on hearing grain hitting a bucket, a horse may in-advertently kick the stall door, perhaps out of impa-tience. If a person then hurriedly feeds the horse(in a misguided hope of quieting it down), the kick-ing behavior will have been positively reinforced,and the person can quickly become a prompt forkicking the stall door. Reinforced this way, itdoes not take long before the horse becomes anavid ruckus raiser capable of training humanseffectively.

Operant conditioning is a horse training standard,and negative reinforcement has been the primarymeans of shaping behaviors. Horses typically aretrained to perform actions in order to avoid some-thing aversive. For example, under saddle, theymove forward when leg pressure from the rider isapplied to both sides; on the ground, they yield theirhindquarters when pressure is applied to a flank;they back up when pressure is applied to the bridgeof the nose; and they enter a trailer to avoid pressurefrom ropes or whips. The judicious trainer works

on refinement: taking advantage of the principlesof shaping and extinction by reinforcing the correctand ignoring the incorrect, so that, over time, onlythe slightest pressure produces the desired action,which makes the human/horse partnership appeareffortless.

As popular as negative reinforcement is withinthe industry, research employing avoidance condi-tioning in horses is limited. In one study, poniesthat learned better in a positive reinforcementsingle-choice point maze also learned better in ashock avoidance test. This indicates that somelearning abilities are similar under positive and neg-ative reinforcement conditions.6 In another study,some horses performed better through positive rein-forcement, whereas others performed better whenthey were required to avoid an aversive stimulus.7Two other studies used sound or visual cues assignals to avoid electric shocks for the purpose ofdetermining optimum number of conditioning ses-sions8 and optimum number of trials per session.9In these tests, horses reached criterion in fewer tri-als when given only one avoidance session per weekinstead of two to seven, but varying the number oftrials per session did not affect their learning per-formance. In other words, in these negative rein-forcement studies, length of session was not critical,but frequency was. This may not be true for alltypes of training, as can be seen from anecdotalreports citing greater success with shorter sessions,especially with young horses.

Training for research purposes is predominantlybased on positive reinforcement. Horses learn torespond to a wide variety of stimuli in tests rangingfrom simple discrimination to concept learning, andtraining of basic experimental procedures flows eas-ily. The use of positive reinforcement may be re-stricted to rewarding correct test choices, or it maybe made the most of by chaining behaviors so thatthe horse, in effect, works completely on its own,thereby minimizing inadvertent cueing by a han-dler. In many experiments, a handler leads a horseinto the testing area and turns it loose so that it maymake a selection10,11; in others, a handler leads thehorse all the way up to the test apparatus where itmakes a choice.12–14 The latter runs the risk of thehorse picking up unintentional signals from the han-dler15 and is, therefore, open to criticism. To avoidsuch pitfalls, Hanggi2,16–18 incorporates chaining(linking a number of behaviors into a series) intoexperiments ranging from vision testing to categori-zation and concept learning (Fig. 2). Horses learnto stand quietly in a “station” (a waiting area con-sisting of pylons with a bar spanning the front)during intertrial (Fig. 2A) and stimulus exposure(Fig. 2B) intervals, to walk forward after the bar islowered and select a stimulus presented in one oftwo or more windows by touching it with its nose(Fig. 2C), to find a food reinforcer in a feeder locatedon the bottom of the test apparatus, to walk awayfrom the apparatus (Fig. 2D) then halfway down the


IN-DEPTH: BEHAVIOR

length of the stable breezeway where they turnaround (Fig. 2E), and walk back into the station toawait the next trial (Fig. 2F). Horses learn thischain within two or three training sessions and re-tain it indefinitely.

Practical applications of positive reinforcement havebeen researched with respect to the use of these prin-ciples to facilitate trailer loading behavior in horses.19

Trailer loading the reluctant horse is a common prob-lem and can take hours to accomplish, with hazard toboth horse and human. Resistant or frightenedhorses rear, pull back, kick, paw, and even fall overduring the ordeal: these behaviors are reinforcedwhen owners fail to load them and give up. Tradi-tional loading methods are based on negative rein-forcement (often with a measure of punishmentthrown in). Ferguson and Rosales-Ruiz19 found that,with positive reinforcement and target training,

horses learned to load willingly, improper behaviorsdisappeared, and these effects generalized to novelsituations. Although those of us who regularly usepositive reinforcement and target training—for trailerloading (Fig. 3) and unloading, lifting feet for hoof care,groundwork, standing quietly for grooming and veter-inary handling, overcoming fear or resistance, and re-search procedures—advocate this approach, moreresearch and public demonstrations are needed to ed-ucate horse handlers on the techniques and efficacy ofthese methods. Ideally, trainers and handlers shouldincorporate intelligent use of both positive and nega-tive reinforcement into a well-balanced program.

4. Discrimination LearningDiscrimination learning in horses has been reportedsince the 1930s and is still regularly used today inan array of tests. In discrimination tasks, horses

Fig. 2. Chaining of behaviors for re-search purposes at the Equine ResearchFoundation. Horses are trained to workindependently of a handler to minimizeinadvertent cueing. (B and C) The rel-ative size concept study: for this horse,“larger than” was the correct response.Photo credit: Evelyn B. Hanggi.


IN-DEPTH: BEHAVIOR

must learn that one stimulus, and not another, willresult in reinforcement. That stimulus then beginsto control behavior, such that the horse acts in aspecific manner in the presence of one stimulus butnot the other.20 Gardner21 found that horses coulddiscriminate between a feed box covered with ablack cloth and boxes that were not. This simplediscrimination was retained for over 1 yr. Horsesshowed a standard learning curve—errors de-creased as number of trials increased. However,these horses did not appear to generalize when thecloth was located above or below the box. Anotherstudy showed that one horse was able to learn 20pairs of discriminations. This horse displayed theability of “learning to learn” by using a general so-lution (one pattern in each pair was always re-warded) to more easily solve subsequent tests andwas able to retain 77.5% of the discriminations after6 mo.22 The learning to learn phenomenon hasbeen noted in numerous other studies2,23–25 and is aworthwhile tool in training. Too often, horses, es-pecially show horses, are limited to performing onlywithin a particular discipline. Thus, western plea-sure horses do not jump, and dressage horses rarelyset foot on a trail. This is unfortunate because suchrestriction keeps the horse from learning about agreat variety of stimuli, which creates an animalthat cannot deal with novel situations as comfort-ably as one involved in a broad range of activities inmany different surroundings. Researchers, equinewelfare advocates, and good horse trainers agreethat the more positive stimulation a horse experi-ences the more easily it learns in new situations andthe better adjusted it is in a variety of environments.

Based on speed of acquisition and the extent towhich discriminations can be reversed, spatial cues

(those relating to area) seem to be more discrim-inable than other stimulus features to horses.26

Vigilant horse owners who have remarked on howwell horses find their way around in areas that theyhave only visited infrequently support this claim.Horses also react noticeably when objects in theirsurroundings have been moved, indicating that theyrecognize that something has changed spatially.Nonetheless, horses are quite adept at discriminat-ing visual stimuli: real life objects that have beentested include buckets,11 doors,27 toys and photo-graphs of toys,17 objects for everyday human use,18

abstract stimuli include striped patterns and col-ors,28 and two-dimensional black figures.2,16,29

5. Discrimination and Visual PerceptionThe ease with which horses discriminate stimuli facil-itates not only cognition testing but also the measur-ing of perceptual abilities, which in turn aidsresearchers in studying equine vision. For example,in a study of depth perception, horses learned to dis-criminate between two patterns using random-dot ste-reograms—one with no form visible and anothercontaining a square visible to individuals with stere-opsis.30 Along with other depth perception studies,28

this experiment provided evidence that horses pos-sessed true stereopsis. Visual acuity was also testedin this manner. Horses were trained to choose be-tween stimuli made up of vertical black on whitestripes of different widths.12 Discrimination testingcontinued until the horses could no longer differentiatethe stimuli: results showed that a horse’s acuity is"20/30 on the Snellen scale. This is not quite as goodas that of a “standard” human (20/20) but better thandogs (20/50 or higher), cats (20/75 to 20/100), or evenwhat the Department of Motor Vehicles requires fordriving.

Color vision in horses has also been testedthrough the use of discrimination learning.14,31–33

This topic of great interest has yet to be satisfacto-rily resolved. Some studies showed that horsescould discriminate red and blue from gray, whereasothers showed that they could discriminate not onlythese colors but green and yellow as well. How-ever, confounding factors, such as brightness, mayhave played a role to the degree that horses learnedto discriminate based not on color but on anotherstimulus characteristic. Research recently com-pleted by Hanggi and Waggonera confirmed thathorses are color-deficient compared to humans withtrichromatic vision. Using discrimination tests,they showed that horses respond to color vision test-ing in the same manner as some red/green color-deficient humans.

A myth that has surfaced repeatedly is that horsescannot recognize with one eye what they have seenonly with the other eye. This notion is used toexplain why horses startle at the same object whenviewed from different directions (such as whenriding out and then coming back on a trail or revers-ing directions in an arena). Anatomical examina-

Fig. 3. The author loading a horse into a trailer after positivereinforcement training. Traditional methods use whips, ropes,negative reinforcement, and sometimes punishment, until the re-luctant horse relinquishes. Horses do learn to load under negativeconditions but some can become dangerous when stressed. Horsesthat have learned to load through positive reinforcement do soeagerly, even with only a visual or verbal signal and even whenunrestrained. Photo credit: Jerry Ingersoll.


IN-DEPTH: BEHAVIOR

tion has confirmed that the horse’s cerebralhemispheres do have a functional pathway for theconveyance of information (belief contrary to thiswas another misconception), and a behavioral studyshowed that they do indeed have sufficient intero-cular transfer.16 This study once again used mul-tiple two-choice discrimination tests in which horseswere trained to respond to one stimulus and notanother while blindfolded over one eye (Fig. 4).Once the discrimination was learned, the blindfoldwas switched to the other eye. Horses immediatelyresponded to the same stimulus, clearly indicatinginterocular transfer.

Discrimination learning was also used to furtherstudy why horses startle at objects they had appar-ently already seen. One intriguing hypothesis holdsthat it is not a matter of recognizing the same objectduring a return trip but more a matter of the objectappearing different from another perspective. Asmentioned earlier, horses perform very well on spatialdiscrimination tasks, and anecdotal evidence is vastwith reference to horses noticing when objects in theirsurroundings have been relocated. Therefore, it ispossible that horses do not always realize that anobject is the same one when viewed from alternateangles. Using children’s toys and a two-choice dis-crimination paradigm, Hanggib initially trainedhorses to choose one of two objects, with the front ofboth always positioned to the left. Once the discrim-ination was learned, the objects were rotated front toright, front forward, front backward, upside down, etc.The horses accurately chose objects presented in cer-tain rotations but failed with others, indicating thatrecognition of rotated stimuli is good under some butnot all conditions.

6. GeneralizationUnder stimulus generalization, a behavior previ-ously conditioned to one stimulus transfers to othersimilar stimuli. This adaptive trait permits an an-imal to form associations with a wide range of stim-ulus features rather than with only one element.Under generalization testing, gradients (a series ofmeasured changes) are acquired that show how di-rectly behavior is controlled by a given stimulus.Responding is highest to the training stimulus(which predicts reinforcement) yet still occurs, but toa lesser degree, in the presence of stimuli possessingcertain features of the original stimulus.

Horses tested for stimulus generalization using cir-cles showed symmetrical gradients: this is contraryto gradients found in pigeons where generalizationtended toward the larger stimuli.34 Generalizationin horses was also examined using tactile stimuli—repetitive tapping by solenoids along the horse’sback—with results showing that behavior was effec-tively controlled by the training stimulus.35 Horsesresponded most often to the training stimulus withbehavior decreasing as the stimuli were moved fartheraway from the original location.

Lesson horses draw on generalization regularly,understanding the assorted and inexact hand, leg,and seat cues from numerous riders of unequal skilland ability. On the other hand, generalization isdiscouraged in dressage horses, which must discrim-inate highly precise cues from their riders.

Many horses could benefit from opportunities forgeneralization. As mentioned earlier, horses inspecific riding disciplines are frequently not allowedto participate in activities other than what intereststheir riders. As a result, they go through mechan-ical motions that rarely enhance any cognitive skill.Evidence of this can be seen in a recent study thatshowed that, compared with horses involved in otherdisciplines, high-level dressage horses displayed thelowest level of learning performance in simpletests.36 It was hypothesized that because thesehorses are trained to perform highly sophisticated,precise behaviors, riders give them minimal free-dom; therefore, they are inhibited from learning tolearn or generalizing.

Horse trainers, from backyard to professional, canenhance their horses’ generalization abilities by in-corporating variety into their programs, both fromthe ground and in the saddle. This type of traininghelps keep interest up, aids the horse when it comesacross something new, and also gives human andhorse an alternative to riding, especially during in-clement weather when horses are otherwise left totheir own devices in stalls (Fig. 5).

7. Observational LearningHorses are social animals most comfortable in thecompany of other horses (Fig. 6). For many ani-mals, social interactions can facilitate the learningof new behaviors. Thus, it would not be unusualfor horses to learn by observing others, and in

Fig. 4. This Equine Research Foundation horse, while blind-folded on the right eye, learned that the plus stimulus was cor-rect. Evidence of interocular transfer (IOT) was shown when,after the blindfold was switched to the other eye, it again cor-rectly chose the plus. IOT was maintained regardless of stimu-lus shapes, their location, or which eye was blindfoldedfirst. Photo credit: Evelyn B. Hanggi.


IN-DEPTH: BEHAVIOR

fact, many horse owners believe this to be true.McGreevy et al.37 reported that 72% of #1000 own-ers thought that abnormal behaviors were learnedby observation. Abnormal or stereotypic behaviors,including cribbing, weaving, head bobbing, pacing,and self-mutilation, are most often exhibited by sta-bled horses. Unfortunately, owners mistakenly as-sume that when groups of horses display the samestereotypies it is because they have learned by ob-serving one another. This leads to the detrimentalact of removing a horse from any social contact.In reality, the appearance of stereotypies in horsesliving near each other is more likely caused by ge-netic relatedness or to the stress of existing in thesame, inappropriate environment.

To date, there is no research supporting observa-tional learning in horses. Horses that observed dem-

onstrator conspecifics solve discrimination tasksbetween buckets showed no sign of superior learningover controls.25,38 Discrimination learning tests un-der more rigorous conditions showed similar negativefindings, but observer horses did approach goal areasmore rapidly—leading to the assumption that somelearning had occurred.39 Observational learning alsodid not seem to assist acquisition of a foot-press re-sponse.40 Nevertheless, it is difficult to accept thathorses cannot learn by observation in any situation.More likely, the proper experimental procedure hasyet to be developed. Perhaps a design more suited tothe nature of the horse might better reveal the extentof observational learning in horses.

8. Categorization and Concept LearningMany horse people believe, some even vehemently ar-gue, that the learning abilities of horses do not gobeyond the scope of associative learning and memory.Although a large amount of cognitive behavior can beexplained by these mechanisms, it is critical for thewell being of horses to study whether they possessmore advanced learning abilities. If the cognitiveabilities of horses are misunderstood, underrated, oroverrated, their treatment may also be inappropriate.Equine welfare is dependent on not only physical com-fort but mental comfort as well. Confining a thinkinganimal in a dark, dusty stable with little or no socialinteraction and no mental stimulation is as harmful asproviding inadequate nutrition or using abusive train-ing methods. Therefore, it is in the interest of bothhorses and humans to understand more fully the scopeof equine thinking.

In comparison with the cognition work with otheranimals, little research into advanced equine learninghas been completed, which is astounding consideringthe importance of horses to humans. Fortunately,this is now changing as researchers design experi-ments that center on more complex cognitive skills.For example, the ability to categorize provides thebasis for substantial higher cognitive function.41

Categorization through the identification of similarphysical characteristics may involve stimulus general-ization. Nicol1 noted that this should be functionallyvaluable because it would allow animals to acquirebroad categories (food, predator, surroundings) andreact quickly in novel or unpredictable situations.For example, endurance racehorses are confrontedwith a great diversity of stimuli during their trainingand competition. Rather than having to learn abouteach object or event separately, they may make in-stant classifications of new stimuli and adjust theirmovements accordingly. Developing techniques toincorporate categorization learning into everydaytraining would undoubtedly be beneficial and wouldgive trainers another practical tool to use.

Categorization learning in horses has been exam-ined in only a handful of studies. Using two-dimen-sional triangles, Sappington and Goldman29 foundthat horses could discriminate triangles from othershapes; however, they were not able to provide con-

Fig. 5. Nonriding activities can include generalization training,which teaches the horse to accept unusual and sometimes alarm-ing, stimuli in a relaxed manner. Photo credit: Jerry Ingersoll.

Fig. 6. Being social animals, horses depend on conspecifics forsafety and companionship. This is true for domesticated horsesas well as these free-roaming Mustangs in Nevada. Photo cred-it: Evelyn B. Hanggi.


IN-DEPTH: BEHAVIOR

clusive evidence of categorization with novel trian-gles. Therefore, it is possible that the horsesresponded correctly because of associative learningprocesses.1 The ability to categorize was more ev-ident in a study by Hanggi,2 in which horses weretrained to discriminate between two-dimensionalblack figures with open centers and solid black fig-ures (Fig. 7). Horses learned the initial discrimi-nations by operant conditioning but then acquiredsubsequent discriminations with fewer errors, indi-cating that they had learned to learn. Evidence ofthe formation of a category came with the introduc-tion of novel stimuli. Horses immediately chose theopen-center stimuli in most instances, which wouldhave been at chance levels had they been function-ing only by associative learning.

Interestingly, under certain circumstances,horses have difficulty transferring learned stimuli tonew tasks.42 In one case, horses failed to use fa-miliar stimuli (a striped board or white bucket pre-viously used as cues for lever pressing) to solve anovel, unrelated task (maze). This may have beenbecause of experimental design and insufficient gen-eralization training before the start of the experi-ment or this specific type of transfer may indeed bedifficult for horses in general. Further study isneeded before conclusions can be reached regardingthis detail of equine learning.

Another topic barely discussed with respect tohorses, except from a negative perspective, is conceptformation. Unlike categorization, where stimuli re-semble one another physically, conceptualization in-volves responding to certain stimuli because theyrepresent the same idea, regardless of whether theyphysically resemble one another.

Concept learning may involve either absolute orrelational concepts, with the latter divided into (1) a

complex level of conditional relations or relationsbetween relations or (2) a more simple level of un-derstanding relative class concepts43 (e.g., bigger,darker). This second level, although less compli-cated, is still an effective measure of cognitive abil-ity, because as Pepperberg and Brezinsky,43 whenparaphrasing Thomas,44 state:

Responding on a relative basis requires a subjectto compare stimulus choices and then derive anduse an underlying, more abstract (and thus gen-eral) concept: in contrast, learning an absolutestimulus value requires only that a subject form asingle association (p. 286).

In a discussion of concept learning in horses,Nicol1 points out that even though Sappington andGoldman29 refer to concept formation in theirhorses, what they actually tested was categorizationlearning. Likewise, Flannery’s horses,13 in anidentity match-to-sample (sample and correct re-sponse stimulus look identical) attempt, were nottested using novel, unrewarded stimuli; therefore,her findings can be explained by conventional asso-ciative learning.

Hanggi18 studied the ability of horses to form con-cepts based on relative size (larger than, smallerthan). After learning to respond to the larger oftwo stimuli for six sets of two-dimensional trainingexemplars, two horses were tested for size transpo-sition that used novel larger and smaller stimuli aswell as three-dimensional objects (sizes: large, me-dium, small, tiny; Fig. 2). Horses chose the largerof two (or more) stimuli, regardless of novelty ordimension. A second phase of this study tested“smaller than” and produced the same results, thatis, the horse selected the smaller stimuli based onrelative size. Moreover, the horses generalizedacross situations that varied from black two-dimen-sional shapes to three-dimensional objects of differ-ent materials and colors, including yellow foamballs, green plastic flowerpots, and red PVC connec-tors. All horses responded as if they had developeda concept of relative size, marking this study as thefirst to provide evidence of relative class conceptformation in equids.

This is not to say that horses possess the sameconceptualization abilities as humans, nonhumanprimates, or other so-called advanced species, but itis an indication that they possess more cognitiveability than what was known. Much more focusshould be placed on these types of abilities, not justto satisfy scientific curiosity but also for practicalpurposes. As with the application of categorizationlearning and generalization ability to training pro-cedures, simple concept learning may be applied totraining and handling. Equine welfare with re-spect to care and management may also benefitwhen more is discovered about abstract concept for-mation. It has been suggested that, if horses do nothave a concept of “sameness,” we cannot expect

Fig. 7. This horse is making the correct choice in an EquineResearch Foundation categorization study. Stimuli were eitheropen-centered or solid two-dimensional black figures of variousgeometric shapes, letters, or icons of U.S. states, people, orplants. Photo credit: Evelyn B. Hanggi.


IN-DEPTH: BEHAVIOR

them to instantly accept new situations that are, tous, just the same as previous ones. For example,humans classify all horse trailers as the same be-cause of their function and so might expect horses toreadily enter any type after they have been trainedwith just one.1 According to Nicol,1 horses may notpossess such a shared functional concept and willhave to learn to enter each type individually. Thisidea has merit in that it should make handlers con-sider the environment more from the horse’s point ofview. However, at least in the trailer example,there is more to the issue than concept capability.

In many instances, horses refuse to enter trailersbecause of poor training, negative experiences, orlack of opportunity to generalize. Horses that havebeen properly exposed to myriad stimuli rangingfrom stables to loafing sheds, one-horse trailers tostock trailers, crowded places to wide-open spaces,and chaotic traffic to quiet countryside are remark-ably adaptable in novel situations. On the otherhand, horses cloistered in isolated stalls and onlybrought out for repetitive training are apt to reactadversely when novelty arises, and many tend to-ward nervousness, flightiness, and neurotic behav-ior. Perhaps it is not a matter of possessing acognitive ability or not, rather it is a function ofupbringing, training, and opportunity to learn.

9. More Than the Sum of the PartsAs is generally the case in research, only particularaspects of an organism are studied. This occur-rence arises from the interests of the researcher orsome other entity and is not, in itself, a problemunless the item of interest overshadows everythingelse. Take, for example, the attention given toequine color vision and visual acuity. Substantialeffort has been made to discern what and how theequine eye sees. This is admirable, that is, untilassertions are made regarding equine behavior thatmay or may not have anything to do with vision.For example, in one case, a small group of visionresearchers claimed that a racehorse crashed into agreen fence because of the color, which according totheir research could not be seen by horses. How-ever, when horses were observed behaving natu-rally, as well as under test conditions, they easilyavoided green objects even while moving at highspeed.c This led Hanggic to further test this phe-nomenon by placing green objects of various shapesand sizes against a similarly colored green back-ground (a plastic tarp) set against green trees andobserving whether the horses could locate them.A horse was stationed 18 feet from the backgroundand then allowed to walk forward and make contactwith the object with a nose-touch. The horse hadno difficulty noticing the stimuli, regardless of theirlocation on the background. Most notably, it couldstill find an object that was only 1 $ 2 $ 5/8 in andcovered with the same material as the background.This simple test showed that, even though horsesmay not pass certain color vision tests, they still

have the capability of seeing colored objects in theirenvironment, even when those objects are nearlycamouflaged and even under conditions when theirreported visual acuity would not suffice. Ratherthan considering the whole horse, the vision re-searchers made judgments based on only a few fea-tures and consequently came to the wrongconclusion. When all details of the accident wereexamined, it seemed much more likely that thehorse crashed into the fence because of a combina-tion of equine behavior and rider error.

10. Conclusion

In the not too distant past, little consideration wasgiven as to why horses behaved as they did or whatpossibilities existed to ensure adequate care andwelfare. However, during the past decade, an ex-plosion of sorts has occurred within the horse indus-try. Scientific conferences, research articles, horseexpositions, clinics, the Internet, television, maga-zines, books, videos, and worldwide tours haveserved the public by making available educationalprospects never before seen on such a large scale.From skilled horse handlers to wide-eyed novices,everyone has the opportunity to advance theirknowledge about equine cognition, behavior, train-ing, and care. Unfortunately, as is human nature,some equine authorities take advantage of thoseeager for information, creating a persona of nearthaumaturgy. It is up to the individual to differen-tiate between the sincere and the artificial, to searchfor truth among unsubstantiated declaration, and tobecome an eclectic in the world of horses.

Research into equine cognition and perception hasmade noteworthy advances, and greater interest isnow being paid to the improvement of trainingmethodology and management. Nonetheless,much more still needs to be learned before scientistsand laypersons alike can make unequivocal claimsas to what it is to be a horse. Studies combiningequine learning, perception, and behavior are thenext step in understanding this remarkable animal.

The Equine Research Foundation is supported bygrants and public donations. I thank Betty andMeinrad Hanggi and Jerry Ingersoll for their enthu-siastic support and efforts. Research conducted atthe ERF is assisted by student volunteers andparticipants of the ERF learning vacations andinternships.

References and Footnotes1. Nicol CJ. Learning abilities in the horse. In: Mills DS,

McDonnell SM, eds. The domestic horse: the evolution, de-velopment, and management of its behavior. Cambridge, MA:Cambridge University Press, 2005;169–183.

2. Hanggi EB. Categorization learning in horses (Equus cabal-lus). J Comp Psychol 1999;1113:243–252.

3. Houpt KA. Learning in horses. In: The thinking horse.Guelph, Canada: Equine Research Centre, 1995;12–17.

4. Tarpy RM. Basic principles of learning. Glenview, IL:Scott, Foresman and Co., 1975.


IN-DEPTH: BEHAVIOR

5. Winskill LC, Waran NK, Young RJ. The effect of a foragingdevice (a modified Edinburgh Foodball) on the behaviour ofthe stabled horse. Appl Anim Behav Sci 1996;48:25–35.

6. Haag EL, Rudman R, Houpt KA. Avoidance maze learningand social dominance in ponies. J Anim Sci 1982;50:329–335.

7. Visser EK, van Reenen CG, Schilder MBH, et al. Learningperformances in young horses using two different learningtests. Appl Anim Behav Sci 2003;80:311–326.

8. Rubin L, Oppegard C, Hintz HF. The effect of varying thetemporal distribution of conditioning trials on equine learn-ing behavior. J Anim Sci 1980;50:1184–1187.

9. McCall CA, Salters MA, Simpson SM. Relationship betweennumber of conditioning trials per training session and avoid-ance learning in horses. Appl Anim Behav Sci 1993;36:291–299.

10. McCall CA. The effect of body condition of horses on dis-crimination learning abilities. Appl Anim Behav Sci 1989;22:327–334.

11. Sappington BF, McCall CA, Coleman DA, et al. A prelimi-nary study of the relationship between discrimination rever-sal learning and performance tasks in yearling and 2-year-oldhorses. Appl Anim Behav Sci 1997;53:157–166.

12. Timney B, Keil K. Visual acuity in the horse. Vision Res1992;32:2289–2293.

13. Flannery B. Relational discrimination learning in horses.Appl Anim Behav Sci 1997;54:267–280.

14. Macuda T, Timney B. Luminance and chromatic discrimi-nation in the horse (Equus caballus). Behav Processes 1999;44:301–307.

15. Pfungst O. Clever Hans (the horse of Mr. Von Osten).Thoemmes Continuum, Bristol, 2000.

16. Hanggi EB. Interocular transfer of learning in horses(Equus caballus). J Equine Vet Sci 1999;19:518–523.

17. Hanggi EB. Can horses recognize pictures?, in Proceedings.Third International Conference on Cognitive Science 2001;52–56.

18. Hanggi EB. Discrimination learning based on relative sizeconcepts in horses (Equus caballus). Appl Anim Behav Sci2003;83:201–213.

19. Ferguson DL, Rosales-Ruiz J. Loading the problem loader:the effects of target training and shaping on trailer-loadingbehavior of horses. J Appl Behav Anal 2001;34:409–424.

20. McCall CA. A review of learning behavior in horses and itsapplication in horse training. J Anim Sci 1990;68:75–81.

21. Gardner LP. The responses of horses in a discriminationproblem. J Comp Psychol 1937;23:13–33.

22. Dixon J. The horse: a dumb animal?. . . Neigh! Thorough-bred Rec 1970;1655–1657.

23. Fiske JC, Potter GD. Discrimination reversal learning inyearling horses. J Anim Sci 1979;49:583–588.

24. McCall CA, Potter GD, Friend TH, et al. Learning abilitiesin yearling horses using the Hebb-Williams closed field maze.J Anim Sci 1981;53:928–933.

25. Baer KL, Potter GD, Friend TH, et al. Observation effectson learning in horses. Appl Anim Ethol 1983/84;11:123–129.

26. Nicol CJ. Equine learning: progress and suggestions forfuture research. Appl Anim Behav Sci 2002;78:193–208.

27. Mader DR, Price EO. Discrimination learning in horses:effects of breed, age and social dominance. J Anim Sci 1980;50:962–965.

28. Timney B, Macuda T. Vision and hearing in horses. J AmVet Med Assoc 2001;218:1567–1574.

29. Sappington BF, Goldman L. Discrimination learning andconcept formation in the Arabian horse. J Anim Sci 1994;72:3080–3087.

30. Timney B, Keil K. Local and global stereopsis in the horse.Vision Res 1999;39:1861–1867.

31. Grizmek VB. Versuche uber das Farbsehen von Pflanze-nessern. Tierpsychology 1952;9:23–39.

32. Pick DF, Lovell G, Brown S, et al. Equine color perceptionrevisited. Appl Anim Behav Sci 1994;42:61–65.

33. Smith S, Goldman L. Color discrimination in horses.Appl Anim Behav Sci 1999;62:13–25.

34. Dougherty DM, Lewis P. Stimulus generalization, discrim-ination learning, and peak shift in horses. J Exp Anal Be-hav 1991;56:97–104.

35. Dougherty DM, Lewis P. Generalization of a tactile stimu-lus in horses. J Exp Anal Behav 1993;59:521–528.

36. Hausberger M, Bruderer C, Le Scolan N, et al. Interplaybetween environmental and genetic factors in temperament/personality traits in horses (Equus caballus). J Comp Psy-chol 2004;118:434–446.

37. McGreevy PD, French NP, Nicol CJ. The prevalence of ab-normal behaviours in dressage, eventing and endurancehorses in relation to stabling. Vet Rec 1995;137:36–37.

38. Baker AEM, Crawford BH. Observational learning inhorses. Appl Anim Behav Sci 1986;15:7–13.

39. Clarke JV, Nicol CJ, Jones R, et al. Effects of observationallearning on food selection in horses. Appl Anim Behav Sci1996;50:177–184.

40. Lindberg AC, Kelland A, Nicol CJ. Effects of observationallearning on acquisition of an operant response in horses.Appl Anim Behav Sci 1999;61:187–199.

41. Nicol CJ. Farm animal cognition. J Anim Sci 1996;62:375–391.

42. McCall CA, Salters MA, Johnson KB, et al. Equine utiliza-tion of a previously learned visual stimulus to solve a noveltask. Appl Anim Behav Sci 2003;82:163–172.

43. Pepperberg IM, Brezinsky MV. Acquisition of a relativeclass concept by an African Gray parrot (Psittacus erithacus):discrimination based on relative size. J Comp Psychol 1991;105:286–294.

44. Thomas RK. Evolution of intelligence: an approach to itsassessment. Brain Behav Evol 1980;17:454–472.

aHanggi EB, Waggoner TL. Unpublished data. 2005.bHanggi EB. Unpublished data. 2005.cHanggi EB. Unpublished data. 2005.


IN-DEPTH: BEHAVIOR

Documents

The Thinking Horse: Cognition and Perception · PDF fileThe Thinking Horse: Cognition and Perception Reviewed Evelyn B. Hanggi, MS, PhD Cognition and perception in horses has often