How does a Horse Learn?
Todays horse is required to learn a wide variety of different tasks, most of which go against almost every instinct that evolution has drilled into him. After all, how often is it that we see feral horses jumping over obstacles in preference to going around them, or chasing a ball around a field. To perform tasks such as these, the modern horse must learn to suppress many of its natural instincts as well as learn to discriminate and respond to a wide variety of new stimuli. Every aspect of an individuals behaviour reveals something about what he has learned during his life and how he has learned it.
Horses have an incredible capacity to learn, to learn quickly and to learn well. Coupled with a natural willingness to please, this has made them extremely useful to us. But to be effective the whole process pivots around the ability of the trainer to understand the horse and how he learns, and to set up and maintain clear lines of communication.
Learning can be classified into 5 basic categories: habituation, associative, latent, imprinting and insight.
HABITUATION - occurs when the animal gradually ceases to respond to repeated stimulation, for example getting used to traffic. A gradual introduction to cars and the experience of many cars passing without causing any harm, habituates him to traffic. Natural responses become diluted with habituation - which is just as well because without it horses would be far less trainable. This form of learning can only be temporary though and the original, more primitive behaviour patterns can resurface if the horse is put under stress. For example, one car may alarm the horse by driving to fast and close, this will cause a reversal back to the primitive natural response of fear and flight. Hence, this type of learning only becomes fully fixed when it has been habituated in as many different testing situations as possible. Horses habituate more rapidly to a new stimulus when it is first rehabituated to an already familiar situation. So a suitable set of sessions for a young horse will run through several known stages before a new one is introduced at the end of a session.
ASSOCIATIVE LEARNING There are two kinds of associative learning, CLASSICAL AND OPERANT
CLASSICAL : was first documented by Pavlov in the early 1900's with his dogs. If a bell was sounded immediately prior to the production of food, the dogs would start salivating in anticipation of the food by the sound of the bell alone. An equine training example of this could be if the rider clicks his tongue just as he feels a young horse about to strike off in canter, and repeats this every time the horse goes into canter, the horse will soon begin to associate the click with the canter transition and produce the canter transition on the click signal. These associations will become stronger the more they are reinforced - whether good or bad.
OPERANT : learning was first demonstrated by Thorndike in 1911 and is the basic trial and error type where the animal learns to associate a reward not with a particular stimulus (as in classical) but with its own behaviour, discovered by chance. From the training point of view this type of learning involves presenting the horse with a choice ; if it makes the right choice it will be rewarded, and the wrong choice punished - either directly or by lack of reward. An example of this is the rein back, at first the horse finds the movement difficult, and so a reward, for example a pat or cessation of the aid after one step backwards is sufficient to encourage him to repeat that action when asked. Next time the rider can ask for several steps and progressively the lesson is built up until the horse is rewarded only when the whole movement has been completed without any errors.
LATENT LEARNING : is the ability of the memory to store an experience unconsciously, without having an immediate need to do so. Horses are very good at this and a good example is their capacity for recalling places, routes and locations.
IMPRINTING : is a permanent, non-intentional type of learning that is not dependant on the repetition or association. During imprinting, an early perceptual experience is indelibly marked into the horses brain, though the stimulus itself may not be consciously remembered. Some trainers have latched on to this idea and start working with foals even as early 1 hour after birth. The aim is to desensitise the foal to potentially alarming sensations, sights and sounds and to set up a bond of trust in this time when the foals are at their most receptive.
INSIGHT LEARNING : Horses don't seem to be prone to "flashes of inspiration" - which is probably a good thing. It can be defined as the immediate understanding of, and response to, a new situation without the need for operant learning. In training we must encourage the horse to think for itself by putting it in the right frame of mind - calm and unpressurised - and in a situation with a problem he has the motivation to solve.
Generally, equine learning research seems to favour the assumption that horses learn through the STIMULUS - RESPONSE - REINFORCEMENT theory. This states that the horse perceives a cue, such as the riders legs or seat, and then makes a random response to it. If the response is correct it will receive a positive reinforcement from the rider and if incorrect, the rider can either both ignore the response and repeat the stimulus or apply a negative reinforcement until the horse makes the correct response.
STIMULUS In order for a stimulus to elicit a response, the horse must be able to determine what constitutes a stimulus. Generally horses are very good at this, examples being that of Clever Hans - a horse that seemed to posses the ability to answer mathematical and spelling questions by the number of times he pawed the ground. However, a scientific investigation revealed that Clever Hans could only correctly answer the question if the questioner himself knew the answer. The questioner was unintentionally cueing the horse to make the correct answer by very subtle tensing and relaxing of his facial and body muscles. Dixon (1970) also demonstrated the depth of equine discrimination with her study where a pony was taught to pick one of each pair of these 20 visual patterns to obtain a food reward. Because horses become so adept at discriminating stimuli - trainers must be very specific with their presentation of cues. If a specific cue is not similar in presentation method and timing each time it is used, the horse will respond by generalising.
This is a familiar scenario with the riding school pony who can become so habituated to the accidental stimuli from beginner riders that they become dull and unresponsive. It must also be noted that in order for a horse to respond quickly to a stimulus, the stimulus must be applied at a time when the horse is able to respond - ie, there is no point in asking the horse to rein back when he is trotting on a loose rein.
There are two main types of stimulus, the first being natural or "primary", this could be in the form of food, pain, return to herdmates and the second learned or "secondary" like a pat on the neck or the riders voice. Successful trainers begin horse training utilising simple, natural stimuli. After the horse has mastered these, it can be taught more subtle, or learned stimuli by pairing the new stimulus with the old, already learned stimulus. It is believed that the horse learns best through either delayed or trace conditioning
The idea behind this is that the old cue is being used to show the horse the meaning of the new cue to reinforce the new cue. Two other types of conditioning procedure are also available; these are simultaneous presentation - where both stimuli are presented at the same time, and a backward stimulus, in which the old stimulus is presented and then the new one; have been shown to be ineffective.
So, how do horses respond to these stimuli ? All of the major maneuvers that a horse perform start out as many small responses. We tend to teach the horse to perform each small response to a major manoeuver, then connect them all together for the final, polished manoeuver. Using the example of the rein back, the trainer first teaches the horse to relax its jaw and shift its weight backward in response to pressure from his legs and the bit. Then the trainer requires the horse to make this response, plus take one step backwards. This is then progressively added to with a greater number of steps and rhythm to until the horse can do the entire movement. When the horse is initially learning the meaning of a new stimulus, its responses to that stimulus will be random actions. For example, the horses initial response to a cue to move backwards may be the incorrect response of throwing his head up in the air or stepping sideways; or the correct response, relaxing his jaw and shifting his weight backwards.
During this initial learning phase, we should aim to ignore the incorrect and encourage the correct, random response. This encouragement that connects a correct random response to a specific stimulus is reinforcement. Reinforcements can be in the form of either positive, or negative.
Both positive and negative reinforcement strengthen the connection between a specific stimulus and the desired response, so that when the specific stimulus is re-presented, there is a greater chance of the horse making the correct response. Like stimuli, reinforcements can be either natural or learned. Trainers teach horses secondary reinforcements by pairing them with primary reinforcers in the same way that a new stimulus is paired with an old. That is they present the new, secondary reinforcement (the voice "good horse") then follow it with an old primary reinforcement - such as a handful of grass. After a number of such pairings , the horse will associate the voice praise with the rewarding properties of a mouthful of tasty grass.
AVERSIVE STIMULUS When the trainer applies an aversive stimulus it can take one of two forms punishment or negative reinforcement. Punishment is applied after the horse makes an incorrect response and aims to suppress or eliminate a response, For example if a horse bites or kicks at the trainer, the trainer would apply the use of punishment to eliminate that response. Negative reinforcement is applied to the horse before the horse makes the correct response and terminates when the horse makes the desired response. Its aim is to increase the probability that the response will occur again. eg. Rapping. We must guard against the use of unintentional punishment, like falling back hard in the saddle, or jabbing the horse in the mouth, because the desired behaviour will be suppressed.
FACTORS EFFECTING EQUINE LEARNING ABILITIES Trainers must attempt to apply reinforcements or punishment immediately following the horses response to a stimulus. This time connection between response and outcome enables the horse to know when it has performed correctly and incorrectly, and it gives the horse a sense of order and expectancy about its responses. We must also make sure that the correct response is available to the horse, both physically and mentally before we use aversive stimuli. For example, it would not be fair to punish the horse for not reining back because you had him positioned against a fence. If continuous reinforcement and the correct responses are not provided for the horse, the horses behaviour may become unpredictable and neurotic.
Negative reinforcement may become so intense that the horse may become panicked and the trainer has to stop its application before the desired response is obtained, in this case the horse may learn that unmanageable behaviour stops negative reinforcement. A example of this could be applying pressure to the horse to rein back, and the horse responding by rearing. Trainers also make use of the principles of extinction ( that is, non-reinforced behaviour will decrease in frequency).
When a horse is initially learning a new response, it tends to make many incorrect, and often annoying, responses. If these incorrect responses are ignored and only the correct response reinforced, the incorrect responses will eventually decrease in frequency until they are rarely exhibited. At the same time, the correct responses increase in frequency to replace the incorrect responses.
For those of you that ride the classic example involves those lovely flowers that are always placed around the dressage arena deliberately to spook the horse. If you ignore the behaviour, the horses reaction to them will decrease in frequency, if you get uptight and perhaps even punish the horse - then you've had it!
It has also be stated that a horse will work harder to obtain reinforcements when the reinforcement is not on a predictable or continuous schedule. Most trainers shift to a variable reinforcement schedule after the horse has initially learned the response. This seems to keep the response rates higher and gives the appearance of the horses responding quickly and willingly in the absence of any overt reinforcements.
LENGTH OF TRAINING SESSIONS Most learning researchers agree that concentrating trials in long training sessions leads to inefficient learning. Ruben et al (1980) trained ponies for either 7 days per week, 2 days per week or 1 day per week in a shock avoidance response. The ponies trained one day per week achieved a higher level of performance in fewer training sessions than the other treatments. These results indicate that concentrating trials in long learning sessions is an inefficient method of training, and that fewer training sessions over a longer period of time is the most efficient.
So going back to the reinback, a short session one day per week should achieve better results faster than drilling the horse in the movement for long periods of time every day. Once the horse has performed the movement correctly and a positive reinforcement can be applied, that training session should stop. A common misconception is that you shouldn't 'start' horses until they are three years old. But it has been shown that horses 'learn to learn and it is through this phenomenon that we should encourage the introduction of the young horse to as many new stimuli and situations as possible.
Finally, it has been demonstrated that equine memory is very good, in one study it was likened to the hard disc drive of a computer. Dixon (1970) reported that her pony had an 81% memory retention rate one month after learning 20 pairs of visual patterns, and 6 months later still showed an amazing 77.5% correct responses. Another point that arose through this study is that the pony seemed to "learn to learn", it learned a general solution to a problem that made subsequent problems easier to solve. You might not be able to ride him, but the more you can introduce him to new stimuli to keep his mind alert and interested the better. He does not necessarily need to be learning anything that will be of use to him at a later date: the fact he is learning anything is all that is important. The learning to learn phenomenon also indicates that training should follow logical and progressive steps.
TO SUM UP We are very fortunate in that the horse is so amenable to our strange ideas of what his role is. We can over-ride many of the natural instincts by teaching new , learned associations, something which they have proven to be very adept at. There are various processes by which the horse learns, including; habituation, association, latent, imprinting and insight; all of which can, and should, be applied to the way in which horses are trained. As has been discussed, an effective way of training horses is by the STIMULATION - RESPONSE - REINFORCEMENT theory. Stimuli should be in the form of simple cues that the horse can easily discriminate and respond to. If the response is correct, the reinforcement should be positive, and if incorrect, the stimulus is either ignored and repeated or negatively reinforced. Complex movements are made up of many small responses, that can be progressively built upon over time. Sessions should be kept short and start off with work that the horse is already habituated to, so that he learns the new stimuli in a relaxed and hopefully motivated frame of mind.
We looked at the five types of learning; habituation, associative, latent, imprinting and insight, all of which - except for the insight - the horse is very good at. The stimulus - response - reinforcement theory explains to us how the horse perceives a cue and then makes a random response to it. And we as trainers must immediately indicate to the horse whether his response was correct - in which case we should apply a positive reinforcement - or incorrect - at which time we either reapply the stimulus or give a negative reinforcement. We have seen that aversive stimuli comes in two forms, punishment and negative reinforcement and that neither should be used carelessly or continuously. Training sessions should be kept short and intermittent so that the horse has time to digest what he has learnt. So as soon as the horse has responded correctly to a question, the session should end so that he leaves the session in a good frame of mind, having learnt something he will be only too happy to repeat next time he is asked.
A common misconception is that you shouldn't 'start' horses until they are three years old. But it has been shown that horses 'learn to learn and it is through this phenomenon that we should encourage the introduction of the young horse to as many new stimuli and situations as possible.
Baer, K.L., Potter, G.D., Friend, T.H. and Beaver, B.V. (1983) Observational effects on learning in horses. Applied Animal Ethology 11, 123-129
Dixon, J. (1970) The horse: a dumb animal?...neigh!! Thoroughbred Record 192 (19): 1654
Fiske, J.C. and Potter, G.D. (1979) Discrimination reversal learning in yearling horses. Journal of Animal Science 49 (2), 583-588
Heird, J.C., Lennon, A.M. and Bell, R.W (1981) Effects of early experience on the learning ability of yearling horses. J. Amin. Sci. 53:1204
Heird, J.C., Whitaker, D.D., Bell, R.W. and Ramsey, C.B.(1986) The effects of handling at different ages on the subsequent learning ability of 2 year old horses. Appl. Anim. Behav. Sci. 15:15
Houpt, K.A., Parsons, M.S. and Hintz, H.F.(1982) Learning ability of orphaned foals, normal foals and of thier mothers. J. Anim. Sci. 55:1027
McCall, C.A. (1990) A review of learning behaviour in horses and its application in horse training. J. Anim. Sci. 68:75-81
Mal, M.E. and McCall, C.A (1996) The influence of handling during different ages on a halter training test in foals.