senses and cognition
and in society
Training with positive reinforcement involves rewarding the correct movements with small pieces of food in the beginning of training. At later stages of training, the food is gradually left out. The daily work with the fully trained elephant does not require food rewards.
The benefits of positive reinforcement also extend to the mahouts. Their work becomes remarkably safer and more pleasant as compared to the prevailing punishment-based training techniques.
Rewards can also be gentle touches. Getting touched - especially on the trunk, temples or forehead - is an important part of the friendly social interactions betweern elephants.
Food rewards can be given either to the trunk tip or directly to the mouth. They can be, for example, small pieces of fruit, or tiny slices of the sweet core of banana stalks. Another approach is to select the preferred part of the daily diet of the elephant and use part of that during training. The picture shows the most favoured part of the ordinary elephant diet in Nepal: the kucchi, bundles of rice, salt and molasses wrapped in grass. For training purposes, they have to be made as small as possible, in order to minimize the time the elephant stops the action and focuses on chewing.
An elephant's skin is very sensitive to touch. Light touches are sufficient in training and work, when used skillfully. This calf is being trained to walk backwards when feeling a light touch in front of its shoulders.
An adult elephant will usually be ridden by a mahout touching the back of the ears by his toes and the area in front of the shoulders by his heels. During training, the same spots in the calf will be touched by hand.
For a calf too young to be ridden, the meaning of touches behind ears is trained by using fingers to imitate the touch of a mahout's toes later on.
Shaping means gradually forming the action towards the final form the trainer has in mind. For example, lying down can be trained by first rewarding a tiny accidental movement downwards, then expecting an increasingly more extensive movement before rewarding. This calf is already at the stage of fully going down; the next part is to continue shaping to train it to lie on the side.
One of the uses for shaping is training an elephant for a test for tuberculosis. For a reliable mucus sample from deep inside the trunk, an elephant needs to blow powerfully into a plastic bag or bucket. In the beginning of training, the trainer blows gently into the nostrils in order to get the elephant to blow back. This is rewarded, repeated and associated to a cue word. On the next step, the trainer will no longer need to blow himself, but will wait for a gradually stronger and stronger "sneeze" from the elephant before rewarding. The elephant will also be gradually habituated to feel comfortable with a plastic bag or bucket around the tip of the trunk, after which the sneeze on command can be directed into this and collect the mucus sample.
Presenting the foot on a convenient place for foot care is one of the tasks that is easiest to train with target training. The sole of the elephant's foot is touched with the target, rewarding the elephant at the same time. After some repetitions, the elephant will start touching the target with the sole of its foot. The trainer can then start modifying the action, such as placing the target where he wants it.
Habituating a calf to accepting a rider is started by first just gently touching the area where the mahout's weight will later be. It is important to ensure the calf is in a relaxed state of mind. (Continued below)
Gradually, the pressure is increased, until a mahout can climb onto the back for a brief moment. At the precise time when the calf feels the weight of the mahout, it is given a piece of favoured food. (Continued below)
The reward can also be a gentle touch in a body part where the elephant likes to be touched, such as on the forehead. The mahout comes down at a moment when the elephant is still calm.
If the elephant will have ropes around its body later in life, for example to carry a howdah or other type of saddle, or to wear a harness when in rescue work, it is a good idea to habituate it to ropes as a calf already. At first, the calf is allowed to sniff and examine a rope. (Continued below)
Once the calf loses interest in examining the rope, the next step is habituating it to the feeling of a rope on its skin. This is done by first just letting the rope hang loosely on the calf. Once the calf is relaxed with this, the trainer can proceed and tie a rope loosely around its body.
With calves of any age, it is important to maintain a calm and focused state of mind during training. One of the most efficient ways is the presence of the mother.
If an elephant is tense durning training, one way to help it relax is to gently stroke the trunk.
Elephants can be trained for any type of work by positive reinforcement, habituation and other animal-friendly techniques. When used correctly, these result in better, more reliable performance than the ordinary pain-based methods. This is why they have enormous potential in replacing pain-based training across south Asia.
This page gives an overview of the basic principles. However, it is not a training manual. To actually learn to train elephants requires more knowledge of details. This is best achieved by getting tutoring from an experienced trainer.
The overview below comprises four sections:
1. Changing the training system: why and how?
2. Frequently asked questions
3. Elephant-friendly training in practice
4. Handling of elephants at work
1. Changing the training system: why and how?
The benefits of elephant-friendly training
When used skilfully, elephant-friendly training techniques typically attract great interest in mahouts, elephant owners and managers. The main reason, in our experience, is that they see elephants learning faster, working with more precision and becoming safer for people to work with, as compared to the commonly used pain-based training.
An example of faster learning comes from our Nepal programme: according to mahouts who attended our workshops, elephants now take less than ten days to reach the same level of learning that previously took two to three months with the ordinary pain-based method.
Safety of people is improved in two ways: immediately while training, and during the following decades of working life. When training involves no pain and injury, the elephant stays calm and focused. This prevents injuries to trainers that sometimes occur in pain-based training when an elephant panics. The long-term improvement in safety is mainly due to better everyday handling. In those tragedies in which mahouts in the prevailing handling system have been killed by their elephants, a very typical case has been a mahout beating an elephant, which then defended itself by attacking. In an animal-friendly system, mahouts have means to control elephants without inflicting pain, resulting in a remarkably safer attitude of elephants towards people.
Another important factor is human dignity - the ability to live according to one's own core values. Many mahouts feel troubled by the pain caused in the prevailing system of elephant training. They only use it because it is the only way they know. After our courses and workshops, we have received heartwarming feedback from several senior mahouts, telling how good they feel now that they can train elephants without hurting them anymore.
Professional-level skill is needed
In order to be effective, an elephant-friendly training system has to be learned in detail from skilled professionals. There have been some previous and current well-intended attempts by both Western and Asian individuals, who are not professional animal trainers themselves, to introduce positive reinforcement to mahouts. Despite the good intentions, such attempts typically fail. Imperfect training has resulted in elephants learning slowly or not at all. Such cases have led some mahouts to believe that positive reinforcement does not work. The elephants have ended up in the ordinary pain-based training, and the mahouts have become less likely to get interested if someone else tries to later introduce actual positive reinforcement. Therefore, it is crucial to seek professional collaboration for such projects, to prevent them from becoming counterproductive.
2. Frequently asked questions
Where has this system been used before?
Elephant-friendly training is used in good-quality zoos, some of the African elephantback safari businesses and a few Asian elephant sanctuaries. Our Nepal programme is the first in the world that has also successfully taught it to mahouts who train elephants for more complex work, such as patrolling in national parks.
In many animal species, the use of animal-friendly training, especially positive reinforcement, has increased rapidly during the past decades. The training of police dogs and police horses in several countries has been changed from pain-based systems to animal-friendly ones. The main reason is the better work performance achieved with the animal-friendly systems. Pet dogs are also increasingly trained by positive reinforcement, as are zoo animals that need to be trained to cooperate with a veterinarian.
Is positive reinforcement the same thing as elephant-friendly training?
The elephant training system that we use and teach is a combination of several animal-friendly methods, all of which are based on the science of how animals learn. These methods include positive reinforcement, gradual habituation, painless forms of negative reinforcement and classical conditioning. The meanings of these terms are explained further below, under "Elephant-friendly training in practice".
In elephant training, this combination of elephant-friendly techniques is sometimes called "Positive Learning Method". This term was coined by WWF Nepal in the beginning of our Nepal programme, when seeking a term that is easy to translate to the local languages.
There is no fixed official term for this combination of animal-friendly techniques. Some people talk about positive reinforcement when they also use habituation and classical conditioning; others just call it elephant-friendly training.
What is not positive reinforcement?
As the idea of training animals humanely has become more popular around the world, especially with positive reinforcement, there have been some misunderstandings on the meaning of the term. The most common misunderstanding is to think that positive reinforcement means anything that is less cruel than the ordinary system. It is not rare to hear someone saying he trains his elephant with positive reinforcement, but a closer look shows he actually uses just a little milder form of pain and punishment.
Practices that are not positive reinforcement include pushing, pulling, poking and inflicting any type of pain. It also is not positive reinforcement to explain to an elephant, "if you do what I want, you will get this piece of food". Elephants cannot understand sentences like this.
Such misunderstandings are typically unintentional, and stem from well-meaning attempts. They simply reflect the fact that there has been not been enough information available.
How is obedience ensured?
One question that is often asked when discussing elephant-friendly training is that of obedience. Will an elephant do what the mahout wants if it is not afraid of the mahout?
With skilful training, positive reinforcement actually makes elephants more reliable and precise than pain-based training. Reliability is achieved by repetition. When an animal has responded to the same words by the same movements for thousands of times, and when the animal associates this to a positive motivation thanks to rewards received earlier in training, reacting to that word by that action becomes virtually automatic, a "second nature" to the elephant.
An important component of training is a "stop" signal. This is trained by positive reinforcement, just as any other task: the elephant is trained with rewards to respond to a specific word by stopping anything it it doing at the moment. It will then be used in everyday occasions, such as stopping walking, but also when the elephant is doing something it is not wanted to do. Regular repetition of this signal will make it as reliable as any other trained signal. In other words, it is not necessary to use punishment when training an animal to obey a word that means "no".
Does a trainer need to show dominance?
Some people assume that elephants used in especially difficult and dangerous work, such as commanded to perform mock attacks at armed illegal hunters during arrests, would have to be trained roughly to become "tough" enough. However, even in this case the opposite is true. When an elephant is not afraid of the mahout, it is easier for the elephant to stay calm and focused on the tasks, frightening situations included. If an elephant is afraid of both the mahout and the surrounding events, the elephants is more likely to shake off the mahout and run in panic. This sometimes does happen with elephants from pain-based training. In police horses, the same benefit has been seen: horses trained by an animal-friendly system work better in dangerous situations such as street riots.
The usual pain-based training is based on the belief that a person has to show he is dominant over an elephant, otherwise he cannot control it. In reality, such dominance is not needed. A better control over an animal is achieved if the trainer knows the answers to two key questions: how to motivate the animal, and how to form strong positive associations in its mind between specific words and actions.
One of the reasons why animals learn faster by positive reinforcement is that a motivated animal focuses its full attention to the task. If the animal is afraid or in pain, it focuses part of its attention on its own fear and pain. Once the task has been learned, another difference is seen in the speed of performing. The typical slow, reluctant movements of elephants that have undergone the ordinary pain-based training are partly due to the painful memories they associate to the command words.
At which age should training start?
These training techniques work on elephants of any age (calves, juveniles as well as adults). If one can choose when to start training a calf, the best age is very early, already at a few months. Calves start eating solid food when they approach the age of half a year, so one can train with food rewards from then on, but already before that one can train by using gentle touches as rewards and by associating words to the calf's movements by classical conditioning (see further below).
The earlier the training is started, the faster and smoother the whole process will be. One of the reasons is that training also involves "learning to learn", such as understanding that words can mean specific actions and that one can get rewarded by them. Another reason is that training at a young age provides an efficient way of channelling a playful calf's play attempts. In the punishment-based system, it is very common for one-year-old calves to get hit or beaten when they try to play head-butting games with people. But if the calf has already been trained before, it is easy to redirect its playfulness to whatever the trainer wants it to engage with.
Do elephants understand language?
In many mahout cultures, there is a belief that elephants understand the local language but not foreign languages such as English. Such a belief in elephants' language ability has been maintained by the fact that when a mahout says a long sentence to a trained elephant, the elephant often does what he wants. However, even in such cases the elephant actually recognizes an individual word in the sentence - a word it has previously learned in training - and responds to that word. Or the mahout always accompanies a command with a specific gesture or body posture, and this is what the elephant uses as the cue.
The belief that elephants understand human language is probably the most important reason for the existence of the pain-based training system. When a trainer says "agat" or "boit" to an elephant calf, and the calf does not react by starting to walk forward or sitting down, the mahout interprets this as deliberately challenging his authority and punishes the elephant for it. Mahouts do not know they are actually punishing the calf for not understanding the words.
Elephants do learn to associate specific words to specific actions, objects, people or animals. But the association has to be formed separately for each word, and with enough repetition to get it established in the elephant's mind. These associations take place during both the pain-based and elephant-friendly training. They just are achieved via different routes in the two training systems.
3. Elephant-friendly training in practice
How does positive reinforcement work?
In positive reinforcement, the trainer establishes an association in the animal's mind between three things: 1) a specific action (for example, turning left or walking directly forward until told otherwise); 2) a specific word, gesture or touch that will mean that action and nothing else, and 3) a reward, which is something the elephant wants, such as a piece of food or a gentle touch.
There are slightly different ways to apply positive reinforcement in practice. The following is a brief overview of one version.
Because the elephant does not previously know what is the action the trainer wants, the trainer has to create a situation in which the elephant does the first moment of that action (for example, starting to turn left), so that he can reward it with a piece of food or a gentle touch. In the beginning, the trainer just waits until the animal happens to spontaneously do a little movement to that direction (such as shifting the weight a little bit to the left), and rewards it.
Timing of the reward is crucially important. If the elephant turns to the left, then moves the trunk a bit, then flaps the ears, and then gets rewarded, it is impossible for the elephant to know that it was the turn to the left that was rewarded.
Once the elephant has learned to do the first part of a specific movement in the hope of a reward, the trainer will start shaping the movement into a more and more advanced form. He will also associate it to a specific cue, which can be a word, gesture or touch. The action will later be rewarded only when performed as a response to that cue. These techniques are briefly explained further below.
Secondary reinforcer: how to mark the right moment
In some cases, the trainer is standing near the elephant and can give the reward immediately. But there will be situations when the trainer is further away. Therefore, the trainer needs a way to "mark" a specific moment: a signal to let the elephant know that precisely this movement is the one that will be rewarded soon.
This signal, also called a secondary reinforcer, can be any distinct sound, gesture or touch. In so-called clicker training, the sound is produced by a handheld device called a clicker. But it can also be any other clear and distinct sound, as long as it is always the same. It can even be a distinct word, as long as the word is not used for any other purpose.
The click, or whichever secondary reinforcer is being used, first needs to be associated to a reward. The trainer gives the elephant a piece of food immediately after clicking. After some repetitions, the next stage is to gradually separate the clicker and reward in time. The trainer clicks, waits a few seconds and gives the reward. The trainer clicks, waits a few more seconds and gives a reward; and so on. As a result, the elephant knows the click means a reward will come soon. After this, the elephant can be rewarded with a click from a distance, and without interrupting the action it is carrying out. The reward is given a moment later.
Shaping: how to increase precision and complexity
The reason for starting the training by first rewarding the first little part of a movement is the ease of making this happen, so the trainer gets an opportunity to reward it. The next step is to wait for a little more extensive movement before rewarding. To return to the example above: an elephant has now learned that taking a small step to the left results in a reward. Once this has been repeated successfully several times, the trainer will no longer reward the first step. Instead, he will wait until the elephant, still hoping for a reward, takes a second step, which the trainer will reward. This will be repeated a few times, so that the elephant knows two steps are needed for a reward. After this, the trainer will wait for three steps, and so on. At this point, when done correctly, the training will start proceeding very fast.
In a similar way, the performance of the elephant in any other task is modified into a more and more precise version of the set of movements the trainer has in mind. For example, when training an elephant to pick up a hat from the ground, the elephant is at first rewarded for any accidental little trunk movement towards the general direction of the hat. Once this has been learned, the trainer will wait for the elephant to extend the trunk further towards the hat before rewarding. Then he will reward only when the elephant touches the hat, etc.
Very complex sequencies of movements can also be trained by shaping, by training each part separately and then linking them together. Depending on the nature of the task, the parts can either be trained in chronological order as in the hat example bove, or in the reverse order: training the last part of the sequence first, then the part before the last, etc.
Cues: linking a word or touch to an action
In addition to letting the elephant know that performing a specific action results in a reward, the trainer also needs to associate a cue to that action. The action will then be rewarded only when the elephant performs it a response to a person giving this cue. Some trainers prefer to first establish a specific action by rewarding it, and then add the cue afterwards. Other trainers use the cue right from the start of training a new task.
The cue can be a word, a gesture or a touch to a specific body part. In most mahout cultures, when the mahout wants the elephant to go forward, backward, stop, turn left or right, or go slower or faster, the mahout signals these by using his feet to touch the back of the elephant's ears or the front side of its shoulders. Mahouts do this either instead of a command word, or in combination with using a word. In elephant-friendly trainig, where training can be started with very young calves that cannot be ridden yet, such touches can easily be trained by trainers standing on the ground and touching with their hands in the same spots where the mahout's feet will later touch.
When the cue is associated to a behaviour the elephant does spontaneously, the trainer says the cue word precisely at the moment when the elephant starts the desired action. Once this has been repeated enough many times, the elephant will associate the word to this action. It is also possible to associate new cues to existing ones. For example, one way to train a calf to move to a apecific direction on command is to first use a touch as a cue, and later also use a word at the same time with the touch.
Target training means that the elephant is trained by positive reinforcement to use its trunk or some other body part to touch a target, which can be any object; one commonly used multi-purposetarget is a ball fixed at the end of a stick. Once it has learned this, it can be trained to follow the target as the mahout moves it; in other words the elephant can be led anywhere. Another version of target training is to train the elephant to present its feet, one at a time, at a specific place for foot care.
Target training is not a necessary component of training elephants for work, but it does make some tasks remarkably easier and faster to train.
Later stages of training
When written out like this, the training process may look very slow. But a skilled trainer can take the elephant through the stages very fast. During the first five or ten minutes, an inexperienced passer-by may get the impression that nothing is happening. But once the elephant has got the idea and is well underway on the path of shaping, it can learn basic movements in hours and complex combinations of movements in a few days, sometimes faster. The experience by us and others has shown that the training process for elephants from start to finish takes only a fraction of the time as compared to the common punishment-based system.
While the training process advances, it is necessary to occasionally repeat the same responses in different places. If an animal is always trained in the same place, it may not know that the cue has the same meaning in other places too.
Phasing out rewards
One of the concerns often raised when discussing positive reinforcement is whether the mahout will have to keep rewarding it throughout life when at work. This is not needed. Rewards are used in the beginning of training, after which they are gradually phased out. Once the elephant has learned a task and performs it well, the trainer starts rewarding only every second time it responds correctly; and later, every third time, etc. In later ordinary work, rewards are needed only very seldom. An occasional reward during routine work will be sufficient to maintain the positive association between the cue and response. If possible, it is also good to reward the elephant if it has performed correctly in unusually difficult circumstances.
Gradual habituation: getting used to new situations
Gradual habituation means introducing new things little by little, so that the elephant does not get frightened of them. In the prevailing system of training, habituation is instead usually carried out in the form of flooding: suddenly exposing the elephant to something new in full form, such as loud noise by banging kitchen pans in front of the elephant, and forcing the elephant to stay there until it no longer fights back or shows any other extarnal response.
In both methods, the final outcome is that the elephant no longer responds to those events, but the experience and internal state of the elephant is different. After gradual habituation, the elephant is genuinely calm, because its threshold of getting frightened was not crossed during the habituation process. After the flooding-type desensitation, the elephant may still be internally afraid or tense. In some cases, this may result in the elephant panicking out of control in a similar situation later in life.
An example is getting the elephant used to a rider. If a mahout simply jumps onto the neck of a previously unridden calf or young elephant, it will usually try to shake the person off vigorously. (In the wild, the only situation in which something would suddenly jump onto an elephant calf is a tiger trying to kill it.) A better, gradual approach is to first lightly touch the calf's neck with a hand for a moment. At exactly the same time, the elephant is given a piece of food. Gradually, the hand is applied for a longer time and with more pressure, after which the trainer can start leaning onto the elephant. When someone finally mounts the elephant, he first stays on it for just a few seconds, while another trainer gives the elephant a food reward, and comes back down. Then the duration of sitting on the elephant is gradually increased.
It is crucial for success that the trainers proceed slowly enough, so that the elephant feels comfortable all the time. It also is important that when a mahout starts sitting on the elephant, he comes down before the elephant starts shaking. If, however, the elephant does start shaking with the mahout still on, it is important that the mahout stays up there until the elephant stops shaking even for a little moment, and comes down only then. If the mahout comes down while the elephant is shaking, the elephant will learn that it can get rid of a mahout by shaking.
Another situation where gradual habituation is needed is getting the elephant used to the noises and events it will encounter in its working life: for example traffic, barking dogs and other village noises. These are also introduced gradually. For example in the case of traffic, the elephant is first brought to a distance from a road, and allowed to do something pleasant such as graze. Gradually day by day, it is taken closer. The key is again to proceed so slowly that the elephant remains calm all the time.
Gradual habituation is also the solution to a problem that sometimes arises with elephants used in patrolling national parks: the park guards find a poached tiger, it has to be transported to the park headquarters, and the only way to do so in the roadless jungle is on an elephant's back. Even a dead tiger still smells like a tiger, and an elephant is very afraid of having such a predator on its back, so the mahouts in the pain-based system typically have to handle such situations with a lot of violence in order to force the elephant to comply. With a combination of habituation and positive reinforcement, elephants can be prepared for such situations so that no punishment is needed. Many park headquarters possess confiscated tiger skins. With the permission of authorities, the trainer can cut one skin to pieces of various sizes: from small handkerchief-sized pieces to half a life size. Then the elephant is first shown the smallest piece from a distance, then gradually brought closer then rewarded for sniffing it, and finally trained to lift the little piece with its trunk onto its back. This can then be repeated and rewarded with the other pieces, moving gradually to bigger and bigger ones and finally to a full tiger skin.
Classical conditioning: using direct associations
Classical conditioning is a simple form of learning in which an animal forms an association between a neutral cue (such as a sound that previously had no meaning to this animal) and an event it already is familiar with. This happens all the time in everyday life. For example, an elephant can learn that the sound of a specific motorbike means that a specific person is arriving.
In training, classical conditioning can be used for many purposes, for example in forming associations between words and actions. One way of letting an elephant know that, for example, "agat" means walking forward, "ra" means stopping, "imher" means turning, "sut" means lying down and so on, is to just watch an elephant calf when it is moving freely, and to pay attention to the moments when the calf changes what it is doing, for example starts walking forward or stops. At the precise moment when the calf starts walking, the trainer says "agat". At the moment it stops, he says "ra". When it happens to start turning, he says "imher", and so on. At this stage of training, the trainer is not trying to get the calf to do anything on command. He is just attaching these words as "name tags" in the elephant's mind for each activity. When the calf is later undergoing the actual training as described further above, it will be very fast in learning to associate cues to actions, as it already knows which word is related to which action.
This is especially useful in training those movements that occur seldom in an elephant's daily life, but are performed in the full form whenever the elephant does spontaneously do them. One example is lying down on the side, which in the pain-based system takes quite a lot of time and violence to train. With classical conditioning, the trainer can create a situation in which the elephant is known to spontaneously lay down, such as going to a river. Elephants often lay down for a moment when entering water. When the elephant starts doing this, the trainer immediately says the word for laying down. It is important to associate the word to the moment the elephant starts to lay down, instead of saying it when the elephant already is lying down, because starting to lay down is the key part of the action that will be needed in training too. Once this has been repeated on several days, the elephant has learned to associate that word with going down, and the rest of the training will proceed swiftly when using positive reinforcement as further above.
One of the most common ways to use classical conditioning in elephant-friendly training is adding a new cue to something the elephant already performs on a different cue: for example, starting to use a word as a cue for something the elephant has learned to do as a response to a touch. In doing this, it is crucial that the new cue is given immediately before the known one, in order to get associated to it. After enough many repetitions, the word can replace the touch. If the elephant does not initially respond to the word, the trainer can follow up with the familiar touch, until the association is fully established.
Negative reinforcement in a painless form
Negative reinforcement (which also is known by the lay synonym pressure-release) means doing something unpleasant to the animal until it starts moving to the direction the trainer wants. This is one of the aspects of training animals with pain. However, there also is a painless version, which can be used as a part of elephant-friendly training.
Negative reinforcement is not the same thing as punishment. Negative reinforcement means that something unpleasant happens until the animal does what the trainer has in mind, at which point the unpleasant thing stops. Punishment means that something unpleasant happens after the animal has done something that the trainer does not want. With punishment, the trainer can only teach the animal what it is not supposed to do, but not what it is supposed to do.
The skin of an elephant is very sensitive to touch. If a trainer wants to use negative reinforcement, it is enough to use very mild touches, such as fingers tickling or vibrating on the skin.
For example, the cues that a mahout will later give with his feet for moving forward, backward, turning etc., can be trained either by positive reinforcement, by painless pressure-release or by a combination of both. The painless pressure-release way to train, for example, turning to the left when a mahout touches the back of the right ear, would go as follows:
A trainer stands by the right side of a calf and keeps his hand behind the right ear, selecting the spot a mahout will touch with his foot when the elephant is older. The trainer starts gently vibrating his fingers, causing a mildly irritating feeling to the elephant, as if there is an insect walking on the skin. If the calf stands still, or moves to any other direction than left, he continues the vibration, keeping the intensity of vibration the same all the time. A calf never stands perfectly still fora long time; sooner or later, it will shift to its weight to one direction or another. Once it happens to shift its weight a little to the left, the trainer immediately takes his fingers off the elephant's skin. After this, the trainer puts his hand back behind the ear and starts vibrating again, until the calf again happens to shift its weight to the left, and the trainer takes his fingers off. After some repetitions, the calf will notice that it can get rid of that tickling feeling by shifting its weight to the left, and will shift immediately when the vibration starts again. The trainer will repeat this for several times to make sure it is well established in the calf's mind. The next time he starts vibrating and the calf shift its weight, he still keeps vibrating, until the calf shifts its weight a bit further, and stops vibrating only then. It is the the same process of shaping as described in the context of positive reinforcement further above. The trainer again repeats this for several times, and after that he will expect a still further movement before stopping the vibration. As the calf leans further and further to the left, at some point it will have to also take one step to the left to hold the balance. This way, the trainer will continue until one light touch behind the right ear will be enough to send the calf walking towards the left.
The importance of a calm mindset
One of the reasons why the common pain-based training is so slow is because the calf in training is very afraid and in pain. It has been separated from its mother; in some cultures it also has been allowed very little food, water and sleep during the past week; and it has bleeding wounds, while people around it continue to inflict pain on it.
Animals are easiest to train when they are feeling calm, safe and comfortable. This is why we recommend training elephant calves in the presence of the mother, and making sure that the calf is not hungry or thirsty. It is also important that there is nothing in the surroundings that could frighten the calf.
Training is the most efficient when it is carried out in short sessions, with breaks in between. The younger the calf, the shorter the sessions; even five-minute sessions can be used with those under half a year. Older calves also perform the best when trained in short bouts. There is no exact rule on the duration, but we have used, for example, 15-minute sessions for two- and thee-year-olds. The duration of the pauses can also vary; we have used pauses that are approximately the same duration than the sessions, but they can also be shorter or longer. If the trainers remain near the elephant during the pause, the elephant itself will often express when it has rested enough, by starting to approach the trainers.
4. Handling of elephants at work
Training only lasts for a short period, after which there are decades of work for the elephant. Occasionally there may still be a need to train some new task, but mostly it is now a question of handling in everyday work.
Handling involves carrying out the tasks the elephant has learned in training. If some of the tasks are needed very seldom, it is a good idea to do them sometimes anyway, for the sake of rehearsal. And if the performance of some task seems to start deviating from the original form, a short period of training it again will fix the problem. This is rare, though; for the most part, elephants trained with elephant-friendly methods are a lot easier to handle in everyday work than those trained by pain.
The difference between training backgrounds
When an elephant is performing its daily work, a skilled observer can see whether it has been trained by positive reinforcement or by using pain. Elephants trained with animal-friendly methods work more precisely, giving the impression of doing it willingly. Elephants from pain-based training are often more slow and reluctant, sometimes requiring several repetitions of a command before acting (though part of the stiffness of many captive elephants is also due to health problems; see further below). In some cases, an elephant from pain-based training can also be nervously alert, wary of punishment, especially when still young. As elephants under the pain-based system grow older, they often enter a state known as learned helplessness, in which they inhibit almost all of their actions because of fear of punishment, which often makes their behaviour appear sluggish. Otherwise, the work looks the same despite the training system, except for that the elephant trained by positive reinforcement will very occasionally receive a piece of food or a gentle touch, in order to keep up the motivation.
Can adult elephants be re-trained?
We are often asked this question, as the vast majority of captive elephants have undergone the prevailing type of training that mostly relies on painful punishment and painful forms of negative reinforcement.
If an adult elephant has been trained by the pain-based system, but is performing its tasks satisfactorily, there is no need for actual re-training. Instead, it is sufficient to focus on ensuring that the mahout is gentle with the elephant and does not use pain in controlling it. Some mahouts are already like this, even if they have only been taught the pain-based system. They are usually the ones with most skill and most compassion. If, however, the mahout does use pain in everyday work even when the elephant it performing correctly - as many mahouts do - the mahout will benefit if he can get an access to a course or workshop on understanding elephant behaviour and handling.
The situation is different if the elephant is reluctant to work, or sometimes only obeys after the mahout has hit or prodded it. In such cases, it is beneficial to re-train both the mahout and elephant with elephant-friendly techniques. However, this will require a very skilled and experienced trainer to supervise the process. It is more difficult to re-train a previously punished elephant than to start with a calf that has no bad experiences of people.
If the mahout has been very abusive, it is safer to get a new mahout for the elephant at the same time as the re-training starts. This is because when a frequently punished elephant finds itself in a situation where it is no longer beaten, some elephants first react by experimenting on what they can do now. If an elephant has lived with a very abusive mahout, the foremost thing on its mind may be a desire to attack that mahout. For safety reasons, it is better to assign such a mahout to another job, such as a driver or gardener, and hire a new, skilled and gentle mahout to start forming a positive relationship to the elephant during the re-training. In any case, it is necessary to know as much as possible about what the old mahout did: which commands he used, when and how he punished etc. in order to know in which situations one needs to be especially wary.
In any case, it is important to bear in mind that elephants that have undergone the currently common pain-based training are permanently somewhat different mentally from those elephants that have only been in elephant-friendly training. An elephant with traumatic memories of people injuring it will never be able to fully forget that, no matter how good its re-training may be at a later age. It may become very motivated and precise in its work, and relaxed around people, but it may still get frightened at some sudden incident that reminds it of the painful memories. Therefore, such previously punished elephants are never as safe for people as are those trained with positive reinforcement only. It is necessary to be a bit more alert around them, and to consider carefully whether and when there can be bystanders nearby. However, such re-trained elephants are safer to people as compared those elephants that continue to be controlled with pain by their mahouts. A usual situation in which an elephant kills a mahout is precisely at the moment when an abusive mahout is beating it.
When considering re-training of an elephant that is slow or reluctant to respond to commands, it is important to also check whether these are symptoms of physical health problems. The elephant may have chronic pain in the spine or nerve damage in the back, due to a poorly fitting howdah (saddle) or an excessive workload; or chronically painful inflammations in the ankle, knee or elbow joints due to having stood chained in unnatural positions. Some other potential causes for slow and reluctant performance include parasites, malnutrition and infectious diseases such as tuberculosis. A veterinarian with expertise in elephants is needed to do the necessary examinations and tests to see if the reluctance and slowness is partly due to some of these.
Copyright © Elephant Experts 2014-2016
Photo copyright © Elephant Experts and the photographers:
Minna Tallberg, Helena Telkänranta, Marc Pierard, Karpagam Chelliah,
Nirvay Sah, Sudhir Yadav and Ramesh Belagere