Train­ing BY Re­ward

Use be­hav­ioral perks for fast re­sults that last. You’ll build a stronger bond with your horse at the same time.

EQUUS - - In Brief - By Janet L. Jones, PhD

May I give your horse a treat?” Friends ask this ques­tion out of kind­ness, and it’s hard to say “no.” But edi­ble re­wards are such high-test tools that they can hin­der a horse’s train­ing when of­fered in­dis­crim­i­nately. Suc­cess with treats de­pends on ex­actly what the horse is do­ing at a given mo­ment, and on which be­hav­iors you want to en­cour­age or dis­cour­age.

The equine brain learns largely by as­so­ci­a­tion---link­ing an ac­tion to a con­se­quence. Let’s sup­pose Monty is bob­bing his head up and down over the stall door. Our new friend hands him a chunk of ap­ple be­cause she wants to be nice. Yum! A men­tal as­so­ci­a­tion has formed: Bob­bing equals ap­ples. Next thing you know, Monty’s smack­ing his head up and down harder, more of­ten, and in other con­texts, coax­ing us to fork over some ap­ple strudel.

Edi­ble treats are like race cars---they work ex­tremely well, but only if you know what you’re do­ing. One or two in­stances can be enough to cre­ate a new be­hav­ior. When it’s a be­hav­ior we want, as­so­cia­tive clout is won­der­ful. But we’re less pleased when it’s a be­hav­ior we don’t want, like a horse bob­bing his head into some­body’s face.

To ac­com­mo­date a horse’s as­so­cia­tive tal­ents, pro­vide the right re­wards at the right time, and only for be­hav­ior you want to en­cour­age. Usu­ally, the right re­wards are non-edi­ble. Un­for­tu­nately, we’ll have to tell our treat-wield­ing friends “no” more of­ten, but our horses and our safety---and theirs---will be bet­ter for it. If we un­der­stand how horses form as­so­ci­a­tions, we can use be­hav­ioral re­wards to teach them ground man­ners, un­der-sad­dle per­for­mance and the oc­ca­sional mounted mas­ter­piece. Re­ward is the most ef­fec­tive form of as­so­cia­tive learn­ing be­cause it cre­ates a phys­i­cal con­nec­tion that is as real as the spark that moves a pis­ton.


Most mam­mals learn by as­so­ci­a­tion. Some are much bet­ter at it than oth­ers, and horses land near the top of the class, thanks to their neu­rol­ogy. Ba­si­cally, a group or net­work of neu­rons rep­re­sents an ac­tion. Think of a sim­ple equine ac­tion, like bob­bing a head. The brain con­trols

the mo­tion of lift­ing the head up slightly then al­low­ing its weight to fall as the neck stretches down­ward. Each time the net­work fires, the head bobs.

When neu­rons fire an elec­tri­cal im­pulse, their ca­pac­ity to re­ac­ti­vate re­mains strong for a few sec­onds after­ward. The as­so­cia­tive en­gine is primed. Dur­ing the ini­tial ac­ti­va­tion, or the primed state that fol­lows, any ex­ter­nal event can be linked with the ini­tial net­work. If an ap­ple ar­rives while the head-bob­bing net­work’s po­ten­tial to fire is still high, the equine brain will form a con­nec­tion be­tween the two. Long-term po­ten­ti­a­tion, as this phe­nom­e­non is called, is the Fer­rari of horse train­ing---but to train well, you have to learn how to gun the en­gine with­out mak­ing it back­fire.

Cog­ni­tive sci­en­tists re­fer to this learn­ing process with an easy ver­bal re­minder coined by Don­ald Hebb: “Neu­rons that fire to­gether, wire to­gether.” How? When two net­works fire at or very near the same time, chem­i­cals made in the brain are re­leased to strengthen the bond. They shoot across tiny gaps be­tween the two net­works and are ab­sorbed by the re­ceiv­ing neu­rons. They’re like glue that strength­ens a new link while it’s un­der construction.

But wait, there’s more! When the sec­ond net­work rep­re­sents some­thing very pleas­ant---like a be­hav­ioral re­ward or an edi­ble treat---dopamine wells up in a clump of brain cells called the nu­cleus ac­cum­bens. Dopamine is a “feel-good” sub­stance for hu­man and equine brains. Food when we’re hungry, drinks when we’re dry, laugh­ter when we’re sad, safe com­ple­tion of a risky ma­neu­ver---the happy feel­ing of these sat­is­fac­tions comes from dopamine re­lease. Such feel­ings are so pow­er­ful that hu­mans can con­sciously

de­stroy their health with in­sa­tiable ap­petites for food, al­co­hol, drugs or ex­treme sports that kill.

So, let’s re­view: Neu­rons rep­re­sent­ing an ac­tion (head-bob­bing) fire when a dif­fer­ent set of neu­rons rep­re­sent­ing a re­ward (ap­ple) also fire. The brain’s nat­u­ral chem­i­cals glue the new link to­gether while it’s form­ing. Prac­tice strength­ens the bond. Add re­ward for a quick shot of dopamine, and you’ve got a level of sat­is­fac­tion that’s off the charts. Horses re­mem­ber those lessons and want to learn more.


Hu­mans learn the same way. But hu­man minds trowel many lay­ers of mod­i­fi­ca­tion over the top of sim­ple as­so­cia­tive learn­ing. These lay­ers in­clude cul­tural stan­dards, so­cial mores, se­man­tic knowl­edge, per­sonal hopes and ex­pec­ta­tions, episodic and pro­ce­dural mem­o­ries, per­cep­tual fil­ters, ethics and val­ues, cog­ni­tive con­trols, at­ten­tion, peer pres­sure, the weather, and all the mus­cle of vary­ing emo­tions. OK, maybe not so much the weather, but you get my point: Our minds are stir­ring a thick brew that of­ten over­rides the au­to­matic force of a re­ward.

Adults are mo­ti­vated more by a sense of self-reg­u­la­tion than a de­sire for ac­co­lades. We want con­trol of our fates, re­spon­si­bil­ity for our suc­cesses and fail­ures. Even when per­sonal con­trol is lim­ited, we are mo­ti­vated by be­liev­ing it ex­ists. Horses don’t care if they are ma­nip­u­lated into a be­hav­ior by their de­sire for a spoon­ful of su­gar. Hu­mans do. Ex­ces­sive praise back­fires with us, es­pe­cially when prof­fered for medi­ocre skills or easy tasks. Over-praised rid­ers, for ex­am­ple, are of­ten dis­cour­aged by mi­nor fail­ures and stop try­ing to im­prove. An anal­y­sis of 128 stud­ies on hu­man mo­ti­va­tion shows that too many ex­trin­sic re­wards can dam­age peo­ple’s sense of re­spon­si­bil­ity for achiev­ing their own sat­is­fac­tion.

Abun­dant re­wards back­fire with horses as well. They have less power to shape an an­i­mal’s be­hav­ior if they are doled out fre­quently. Treats be­come run-of-the-mill, more like an aid to nutrition than an op­por­tu­nity for learn­ing. It’s the sur­prise of un­ex­pected re­wards that pro­duces the greatest re­lease of dopamine in the brain. Pre­serve your horse’s ed­u­ca­tional in­ter­est by com­mend­ing when a new ac­tion needs re­in­force­ment or a com­plex ma­neu­ver is fi­nally achieved. (The def­i­ni­tion of “com­plex” de­pends on your horse--some­thing that is dif­fi­cult for that horse to do, at that time, in that set­ting.) Up the ante as your horse learns: Two steps of back­ing are praise-wor­thy at first, but even­tu­ally it should take 20 steps--smoother, straighter, rounder---to earn the same prize.


Most horses find wither scratch­ing and neck stroking much more pleas­ant than pat­ting. Stroking a horse’s neck re­duces his heart rate.

Most peo­ple as­sume that a re­ward is some­thing very spe­cial. But in fact, it’s any­thing de­sir­able. Horses de­sire rest, known lo­ca­tions, quiet, equine buddies, fa­mil­iar peo­ple, calm voices, gen­tle han­dling, soft hands, clear di­rec­tion, strokes on the neck, wither scratches, sooth­ing words, con­sis­tency, rou­tine, down­ward tran­si­tions … the list goes on. The bet­ter the re­ward, the deeper the as­so­cia­tive learn­ing. In other words, we can

ad­just the value of pos­i­tive re­in­force­ment by al­ter­ing the type of re­ward.

To horses, edi­ble re­wards---treats--are crème brulee in Water­ford crys­tal. They be­come as­so­ci­ated with good or bad be­hav­ior very quickly. When a horse per­forms a rare and muchde­sired act that is com­pli­cated, goes against equine na­ture or has been deeply re­sisted, he de­serves a treat. The rest of the time, train with noned­i­ble re­wards. The treat is too pow­er­ful to use of­ten.

Treats have another down­side: The horse al­ways wants more. He’ll mug for a sec­ond tid­bit, then gently touch your arm to re­mind you of your obli­ga­tion. There’s noth­ing so bad about a soft touch, but horses can es­ca­late quickly at the prospect of a scrump­tious nugget. Over the course of time, they can go from nudg­ing to knock­ing you down, and from cute baby-nib­bles on your sleeve to bone-crush­ing bites through your arm. Horses don’t know they can hurt you---they just want the candy ma­chine to work.

Make a list of all the non-edi­ble re­wards you can of­fer your horse. Rank the items based on your horse’s de­sires---ev­ery horse is dif­fer­ent---and on your abil­ity to de­liver the re­ward at the right time. Most horses find wither scratch­ing and neck stroking much more pleas­ant than pat­ting. Stroking a horse’s neck re­duces his heart rate. It can be done within a sec­ond or two of good be­hav­ior us­ing the back of your fin­gers while hold­ing both reins---a dis­tinct ad­van­tage when you need to re­ward at a hand gal­lop and pre­fer to avoid a face plant.

Ver­bal praise is ef­fec­tive dur­ing mounted feats be­cause you don’t have to change your body po­si­tion to pro­vide it. Pair it first with bet­ter re­wards, so the horse learns that “good job” is a form of praise. After that, the words alone will do the trick. Just avoid the com­mon er­ror of as­so­ci­at­ing ver­bal praise with slower mo­tion---you don’t want your speed jumper to put on the brakes ev­ery time she hears an atta­girl.


The big­gest chal­lenge with re­ward­based train­ing is get­ting the tim­ing right. Equine re­search demon­strates that the re­ward must en­sue within 10 sec­onds of some tasks for an as­so­ci­a­tion to be formed. In­ter­ven­ing be­hav­iors can­not oc­cur dur­ing that pe­riod be­cause mam­mals as­so­ciate a re­ward with what­ever hap­pened im­me­di­ately pre­ced­ing it. Mis­tim­ing is es­pe­cially detri­men­tal when the re­ward is a de­li­cious treat. The im­por­tance of tim­ing changes with species as well: A dog can be re­in­forced ef­fec­tively many sec­onds after a be­hav­ior. Not so with a horse. If you have to fumble around in your pocket for a treat, or were slow to no­tice a de­sired be­hav­ior, it’s too late. The spark has missed the pis­ton.

Another prob­lem with tim­ing is the fail­ure to dis­crim­i­nate among var­i­ous ac­tions that the horse is per­form­ing at the in­stant of long-term po­ten­ti­a­tion. If Monty is standing qui­etly, mind­ing his man­ners, then a re­ward will re­in­force that good be­hav­ior. But if he’s wig­gling, paw­ing, bob­bing, nudg­ing, touch­ing, nib­bling, push­ing---any man­ner of mis­con­duct---the same re­ward will re­in­force his bad be­hav­ior. Once is enough for quick learn­ers, though most horses re­quire two or three

as­so­ci­a­tions for the les­son to stick.

The most ro­bust in­stants of a train­ing ses­sion (what I call “power mo­ments”) oc­cur just be­fore cool­ing the horse out, just be­fore you dis­mount, and just be­fore you re­turn the horse to her stall or pas­ture. Why? Be­cause rest cre­ates great com­fort in the horse. Use power mo­ments wisely. When you get the perfect slow lope, end your ses­sion and cool the horse out. He’ll want to lope like that next time, in hopes of win­ning the cool-out lot­tery again. If your horse dislikes back­ing, have him back one or two steps just be­fore you dis­mount. Your dis­mount is the berries. If your horse gets ahead of you on a lead, stop him at his stall door, wait, then put him away. En­ter­ing his stall is the No­bel Prize. What­ever sim­ple ac­tion your horse avoids, get him to per­form a bit of it dur­ing a power mo­ment, and you will be teach­ing a po­tent les­son.

An equally po­tent les­son oc­curs when we ac­ci­den­tally re­ward neg­a­tive ac­tions dur­ing power mo­ments. If you quit a train­ing ses­sion in dis­gust at your horse’s re­fusal to lope slower than 50 mph, you have just re­warded him for his speed. If you dis­mount while the horse is danc­ing around, you have re­warded him for unsafe be­hav­ior. If the horse hur­ries you into his stall, you have taught him that push­ing is ac­cept­able. Think about how the horse is be­hav­ing and whether you are inad­ver­tently re­ward­ing bad con­duct.

Pair­ing the right re­ward for the task with split-sec­ond tim­ing might seem like an im­pos­si­ble standard of per­fec­tion. Just try your best, and with prac­tice you will im­prove. Even­tu­ally, as­so­cia­tive train­ing be­comes sec­ond na­ture. The best train­ers man­age such unions all day long with lit­tle con­scious at­ten­tion.

The most ro­bust in­stants of a train­ing ses­sion (“power mo­ments”) oc­cur just be­fore cool­ing the horse out, just be­fore you dis­mount, and just be­fore you re­turn the horse to her stall or pas­ture. Why? Be­cause rest cre­ates great com­fort in the horse.



With the well-timed use of re­wards, horses learn many pos­i­tive be­hav­iors. Ground man­ners are taught by re­ward, so the horse learns to stand in crossties, to defe­cate or uri­nate in ac­cepted lo­ca­tions, to al­low ears and un­der­bel­lies to be clipped, to re­ceive in­jec­tions and to lift all four feet. Us­ing as­so­cia­tive tim­ing, we teach horses to walk safely on a lead, learn­ing to stop, turn, slow down, speed up, wait or back just by notic­ing hu­man body lan­guage. Ground man­ners pre­vent in­juries to horses and han­dlers, and they al­low veterinarians and far­ri­ers to pro­vide care.

Mounted train­ing is also im­proved with pos­i­tive re­in­force­ment. An un­started horse has a lot to learn. He doesn’t know that the sad­dle is heavy, or that the bit feels cold, or that you won’t hurt him. He doesn’t know how to walk a straight line or turn a corner in the arena when he reaches it. He doesn’t know how to keep his bal­ance while car­ry­ing you around that corner. With re­ward-based train­ing, we ease the un­started horse’s process of learn­ing, re­liev­ing his anx­i­ety and earn­ing his trust.

As­so­cia­tive learn­ing by re­ward works long after ba­sic train­ing is com­plete. Hun­ters and jumpers don’t just au­to­mat­i­cally trot through ground cav­aletti. They’ve been taught through re­ward not to touch the poles. Later they will be taught by re­ward to hop

crosspoles, and some­day with tal­ent they might learn to leap ob­sta­cles that stand five feet tall and seven feet wide. In train­ing for any dis­ci­pline, the horse must come to un­der­stand what we want. The best way to con­vey that is through re­ward.

Horses learn to gen­er­al­ize as­so­ci­a­tions, too. Sup­pose you have taught your horse to avoid touch­ing ground poles in the arena. With­out gen­er­al­iza­tion, a horse some­times as­so­ciates place or time, rather than ac­tion, with a given be­hav­ior. He could come to be­lieve, for ex­am­ple, that he is not to touch poles when they are painted yel­low or lo­cated in the cov­ered arena. He’ll need to learn that the les­son holds for all poles in all places. Gen­er­al­iza­tion oc­curs by prac­tic­ing new moves in other lo­ca­tions, or with dif­fer­ent equip­ment, us­ing mild re­wards. It’s easy to in­duce.

All horses learn through re­ward that hu­mans feed and wa­ter them. When food is slow to ar­rive, they put up a fuss. If ir­re­spon­si­ble hu­mans for­get to fill a wa­ter trough, horses will stand near it, whinny, hop, kick, buck, push on the fence, and protest un­til they get our at­ten­tion. Zoo-raised horses don’t even know they need to seek food and wa­ter when re­leased into the wild. Cap­tive horses have learned, through the as­so­cia­tive re­ward of pre­vi­ous feed­ings and wa­ter­ings, that we hold the keys to life.

Gold stars don’t al­ways have to come from us. Many horses learn to es­cape their stalls to seek edi­ble re­wards. Note that this is still re­ward­mo­ti­vated be­hav­ior: There is some­thing de­sir­able out­side the stall that the horse wants---a bas­ket of treats, a path to the pas­ture, or a cute mare down the aisle. It’s also dan­ger­ous be­hav­ior: If your horse es­capes to the grain bin, the re­sult­ing binge puts him at risk of lamini­tis or worse.


Any­one who has al­lowed a horse to bite at long grass while out on the trails knows that once taught, a given be­hav­ior can be very dif­fi­cult to change. Ex­tinc­tion refers to the ef­fort to elim­i­nate a pat­tern of as­so­ci­a­tion. Re­warded mis­be­hav­ior is es­pe­cially dif­fi­cult to ex­tin­guish, for two rea­sons. First, mis­be­hav­ior is of­ten based on in­stinc­tual move­ments de­vel­oped over mil­len­nia of evo­lu­tion. Paw­ing, for ex­am­ple, is a nat­u­ral ac­tiv­ity for a bored, hungry or thirsty horse who is con­fined. Teach­ing the horse not to fol­low this in­stinct will be a chal­lenge.

In train­ing for any dis­ci­pline, the horse must come to un­der­stand what we want. The best way to con­vey that is through re­ward.

Sec­ond, horses rarely for­get a les­son that was once ac­com­pa­nied by an ex­tremely de­sir­able out­come. When horses paw in cross-ties, what of­ten hap­pens next? Yep, their own­ers be­come ex­as­per­ated and put them away. Food, wa­ter, rest, com­fort and buddies await---a five-fac­tor re­ward!

The moral of the story? Be care­ful what you teach your horse by re­ward, be­cause he might re­mem­ber it for­ever.

Re­ward is the most ef­fec­tive type of as­so­cia­tive learn­ing in horses. The brain phys­i­ol­ogy of long-term po­ten­ti­a­tion means that we must rate the value of re­wards care­fully and of­fer them at the right mo­ment. Hand­ing a horse treats be­cause he’s cute, or al­low­ing a friend to of­fer him de­li­cious good­ies, usu­ally leads to trou­ble. In­stead, save spe­cial treats for spe­cial per­for­mance. In the mean­time, train with non-edi­ble re­wards. In­tro­duce your new friend to your horse and let her stroke his neck or shoul­der. Both par­ties will ap­pre­ci­ate that higher level of care.

About the au­thor: Janet L. Jones, PhD, is a cog­ni­tive sci­en­tist who ap­plies brain re­search to the train­ing of horses and rid­ers. Now pro­fes­sor emerita, she taught the neu­ro­science of per­cep­tion, lan­guage, mem­ory and thought for 23 years and is the au­thor of three books. Jones be­gan rid­ing at age 7 and has schooled hun­dreds of young horses, com­pet­ing in hunter, jumper, hal­ter, rein­ing and Western plea­sure classes through­out the west. She uses ba­sic prin­ci­ples of dres­sage in train­ing ev­ery horse. Lo­cated in Colorado, Jones cur­rently owns a 17.1-hand offthe-track Thor­ough­bred who makes ev­ery day in­ter­est­ing. Reach her at ride­with­y­our­

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.