PxPixel
The limits of negative reinforcement - The Horse Owner's Resource

The limits of negative reinforcement

This training technique teaches a horse to obey and respond to commands, but it builds little trust between horse and rider.
Author:
Publish date:

The phrase “negative reinforcement” is usually greeted with wrinkled foreheads and glazed eyes. Yeah, we know the term … but please don’t ask us to define it. So, let’s start at the beginning: Negative reinforcement is a training tool in which painless pressure is applied until a horse responds as desired. When the horse responds, the pressure is removed. Over time, the horse associates that particular cue with the proper response.

A simple example of negative reinforcement occurs when our cars train us to fasten our seat belts. The chime warning that the seat belt is not engaged applies continuing pressure to our ears, stopping only when we buckle up. When you ride, it’s used when you press your legs against the sides of your horse, then release that pressure as he begins to trot. Through association, the horse learns that equal leg pressure on both sides means “trot.”

Equine brains are built to seek associations, whether they are provided consciously or not. Because horses are always looking to make these connections, handlers must give them an idea of what is expected and what is forbidden. If you don’t make those decisions, your horse will make them for you---and his idea of right and wrong probably won’t match yours.

Negative reinforcement is the most common form of associative learning used in horse training, very likely the method you were taught when you began to ride. Because of that long history, it is our natural default mode, one that we fall into with little thought. Let’s take a closer look at its strengths and weaknesses, then in our next installment we will consider a more effective means of training.

Negative reinforcementin practice

Negative reinforcement works best when it’s applied in a form that corresponds to the horse’s nature. Researchers Andrew McLean and Janne Christensen point out that horses use displacement in their natural lives every day---a dominant mare needs only to pin one ear to get a subordinate to move away from her food. A horse displaces you by swinging his head, stepping into you, pushing against you or kicking. In other words, he knows how to move you around if you let him. Because horses use displacement naturally, we can harness that ability for greater learning.

When riding, you use a type of pressure that accommodates the horse’s tendency toward displacement. Take leg pressure, for instance. Why don’t we use eye blinks, bicep curls or words like “hurry up, slowpoke” to get a horse to speed up? Because leg pressure mimics the horse’s natural means of displacement---the horse moves away from pressure to his sides, no matter who applies it. If you apply pressure to a horse’s left side, he will move to the right, and vice versa. Equal pressure on both sides will cue him to move forward to escape pressure. Theoretically, he could choose to move backward (even on a loose rein), but backward motion is less natural and consequently rare in a green horse.

As soon as the horse responds in a way that fits a rider’s desire, like moving from a walk to a trot, we remove the pressure. Ahh … to a horse, pressure release feels good. Horses do not like pressure and will work to avoid it. If you immediately release, the horse’s brain will connect his action with your response. Next time you press with both legs, the horse will speed up again, hoping to achieve the same result.

Nuts and Bolts

Physiologically, the association between pressure and release occurs when two neural networks become linked by simultaneous activation. One group of neurons in the horse’s brain represents his sensation of your leg pressure. You press, he feels, and certain brain cells light up. A different set of neurons represents his movement forward. When the two networks fire at or near the same time, they are linked through a process we call “long-term potentiation.”

Long-term potentiation is a form of priming. Activated neurons remain more fully awake for a few seconds after initial firing. They are primed to fire more quickly and intensely during this brief moment. Releasing pressure during that moment causes the two networks to become connected. Too early, and the first network isn’t activated yet. Too late, and its priming is gone. Equine brain function de-mands careful timing to create the link between your pressure and your horse’s response.

Trainers use negative reinforcement to teach green horses to respond to human pressures of all sorts. A young horse’s first mounted canter is usually confusing for her. She has been ridden at a walk and trot, learning the basics of stop, go, turn, circle and loop. But cantering with a rider is something new---suddenly the trainer is using the pressure of only one leg, yet his upper body position and light rein tell the horse to speed up. If horses used conscious thought, they might wonder, “Hmm, that’s different from the usual trot cue. What does it mean?”

With steady pressure from one leg, and some awkward moments at a jarring extended trot, the young horse will eventually try a canter. Imagine her saying, “OK, it’s not the fast trot he wants because his leg is still pressing only one side. I can try throwing my head; no, that’s not it. How about a stop? No, he pushes with both legs when I do that. Well, let’s try some canter.…” The moment the young horse starts to canter, the trainer releases leg pressure and moves comfortably with the horse. The horse now knows, “Aha! That’s what one leg means.” Her brain uses long-term potentiation to connect the two networks and learn the lesson. Over time, of course, we will sharpen that horse’s perception of our cues in many ways. But at the very beginning---after suitable groundwork---one leg, a light rein and some stick-to-it-iveness is all the mounted canter takes.

Later Stages

Negative reinforcement works best in early stages of horse training, but it is used commonly at later stages as well. The half halt, or downward transition, is a case in point. Suppose you are trotting and wish to walk. You use your seat in a stilling tempo that resists the horse’s back movement. He feels that pressure and responds by slowing down to accommodate your rhythm. When he walks, you release seat pressure and move with him once again. He learns the half halt in this way and will respond more readily each time, as the two neural networks continue to activate simultaneously through practice.

The equestrian seat is a critical source of pressure. A good rider’s seat variations change by many newtons of physical force and in all different directions---up, down, left, right, forward, back, diagonal and circular. Eventually, a highly trained horse will respond to seat variations of every angle and force within 360 degrees per vector. When that happens, a skilled rider is able to place each of the horse’s shoulders, hips and feet at any location with her seat, in real time---at a walk, trot or canter. Squeeze the outer corner of one glute, and the horse will change leads. Lift a quarter circle with the adductors, and she will jump an inch higher to clear the hardest obstacle on the course.

Work is another form of pressure to a horse, and rest is a release from pressure. Suppose Missy occasionally bunny-hops under saddle. If bronc riding is not your forte, hire a trainer to correct this behavior. But if you have the skills to manage it, don’t stop or slow down when the tricks begin---instead, put Missy to work. As soon as she bucks, trot her hard and fast for 60 seconds. When she flows forward easily without bucking, stroke her neck and let her walk on an easy rein. Then try the initial maneuver again. Each time the horse begins to buck, push her forward into harder work. We are not trying to tire the horse; we are teaching her that she will be relieved from the pressure of work as soon as she stops bucking.

The downsides of negative reinforcement

Negative reinforcement is useful with horses who are just learning how to carry riders and interpret their cues. It is also helpful with horses whose relatively mild misbehavior needs to be corrected. Many horses are trained exclusively with negative reinforcement and manage pretty well.

But negative reinforcement can create problems. First, it has to be done in the moment. Most amateurs have trouble coordinating their movements with a horse’s movements. Think back on your first canter, pounding along at a bone-jarring trot that threatened to shake the teeth out of your head, wasted muscles flopping in the breeze. About that time, the instructor says, “Press with your left leg, just behind the girth.” Oh, sure. I’ll get right on that. Many of us at such moments aren’t even sure our legs are still attached to our torsos.

Second, coordination becomes even more difficult while simultaneously timing the application and release of pressure. Because long-term potentiation is so brief, timing must be very precise. Imagine you are teaching a horse the leg yield, in which the horse stays straight head-to-tail but moves diagonally away from leg pressure. To teach this maneuver, the rider applies her outside leg during the swing phase of the horse’s outside hind leg. The “swing phase” is that teensy micro-moment when the horse lifts a leg off the ground but before he sets it down again. So, you apply outside pressure behind the girth as the swing phase begins, and if the horse responds by moving his outside hind diagonally, you release your pressure just as the swing phase ends. This interval lasts for less than half a second at a medium trot. That’s precise timing!

Accidental unintended reinforcement is a third problem. You’re flying along at a gallop and need to touch the left rein lightly to begin a large circle. Whoops! You “touch” too hard, the horse turns on a dime (exactly as you inadvertently requested), and you are sitting on the ground where the horse used to be. By coming off, you’ve relieved pressure in abundance, providing a potent lesson. If the horse could speak, he might say, “Holy wow! She wanted me to turn right out from under her. I did, and she released all pressure instantly. I’ll do that again next time.”

Strong riding skills prevent unintended reinforcement. By riding with weight in your heels, upper body straight, arms and hands soft, legs and seat moving with the horse, you can deliver clear cues. You can repeat those cues consistently time after time, giving the horse practice at a new lesson.

Fourth, riders sometimes mistake “pressure” for “punishment.” The pressure of negative reinforcement might have to be annoying or displacing at first, but it must never be painful or damaging. Punishment, as an educational tool, can cause severe problems and must be used only by highly qualified trainers in the rare event of egregious equine behavior. It is the least effective means of training a horse.

Perhaps most important, negative reinforcement teaches a horse to obey and respond, but it does not build much trust or attachment between horse and rider. It leads a horse to seek, identify, and use human cues---all very important abilities---but it does not offer the added benefit of teaching the horse that you’re on his side.

Correct and release

Pressure release is the most critical part of negative reinforcement. The technique is used incorrectly when riders apply pressure but fail to release it as the horse responds. This is a frequent error. Horses do not respond well, in the wild or in the arena, to constant pressure. Some lose all motivation to try; others become too jittery to perform; a good number act out by bucking, rearing, freezing or bolting. The principle of correct-and-release is helpful here.

Correct-and-release works for all forms of equine performance, from leading to levade. Holding a correction---for example, when a touch becomes a steady pull on the bit---impedes learning, annoys or frightens the horse, and places human strength in competition with equine power. No matter how strong you are, you will never outpull a half-ton horse. Instead, he’ll develop a hard mouth, an inverted neck and a sullen attitude---and you will develop some very sore arms.

To avoid steady pressure, try using a series of touches, releasing the horse from the series when he responds correctly. (Remember the seat belt alarm? It doesn’t need to play continuously to cause us to buckle up; it can exert pressure just by beeping on and off repeatedly.) Add other methods of slowing to help the horse decipher your cues: lowering your body weight, bending your elbows, softening your legs, posting more slowly, adopting a more vertical position.

Negative reinforcement is the most common form of associative horse training, but it requires superb coordination, timing and equitation. It teaches a horse to respond like a good soldier, but it rarely motivates the horse to want to please through excellent performance or to build a bond of trust with his handler. For that, we train by reward. In my next article, we’ll explore the ways in which a horse’s brain learns by reward, then apply that knowledge to every-day horse handling. Now, if only cars would stop beeping and just drop a bite of cheesecake from the ceiling as we fasten our belts!

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 with every horse. Located in Colorado, Jones currently owns a 17.1-hand off-the-track Thoroughbred who makes every day interesting. Reach her at ridewithyourbrain@gmail.com.

This article first appeared in the September 2017 issue of EQUUS (Volume #480)

Related