Andrew McLean – Working With Elephants

Working with elephants… Andrew discovers it’s just like working with horses…

One of the great things about the teachings of Andrew McLean is that he insists that when we are training horses, we are not dealing with fellow human beings wearing shiny fur coats, but with an animal: an animal with a mental capacity and mind-set that is very different from our own. I don’t know how many times I have heard otherwise sensible people gush ‘Oh my horse loves his work now’, and dear me, do they get uptight when you suggest that they are anthropomorphizing and attributing attitudes to an animal, that it cannot possibly have.

Oh well, I hear you say, why upset them; why not let them enjoy their fantasy? The problem is that if the horse is working because he ‘loves’ it, or ‘understands’ it, and then starts going badly – what can you do? Send him a Valentines Day card? Invite him in for a quiet chat over a glass of chardonnay? I guess what most riders mean when they say, my horse loves his work, is that right now the horse is fit enough and fed appropriately for the level of work, that the rider is using the aids clearly enough for the horse to respond in the desired way, and that the horse has been gymnasticised to the point where he can perform the exercises without stress. Now if it all starts to fall in a heap, we have tools to unravel what went wrong: diet? rider error? Exercises too physically demanding for the horse’s current level of strength and suppleness?

Talking to Andrew McLean is an exhilarating roller coaster ride of ideas and hypotheses, and I wasn’t at all surprised to find myself listening to his experiences helping train elephants in Nepal – and guess what, for Andrew working with elephants, like working with horses, all starts with the legs…

“What I am trying to show people is that the most important thing you can do when you are training, is recognize that you must first get the legs of the horse trained to react to light aids to do the right thing regardless of what the head and neck are doing. Then make sure that you have your speed control organized, so that the horse maintains the speed, and you can also quicken the legs if you want to, and you can also lengthen the stride. Quickening and lengthening are important variables to separate and have very different aids for.”

“You also have to control the horse’s legs so that he doesn’t drift. Once we’ve got the immediacy of the reaction, and the lightness, and the rhythm and then the straightness, then we are right to look at head carriage. Head carriage is so simple – every horse I’ve ever dealt with when people have said to me, oh everybody has tried to get this horse round and no-one can – it’s dead easy once you’ve got the horse light and in a rhythm. I think that’s where it is right in the old classical French way where they talk about the head and neck being a consequence of the legs.”

“I’ve even found that it is exactly the same working with elephants in Nepal. They don’t do a round outline of course but you do have to flex them to get them to turn – that’s their signal to turn, and their legs follow where their head goes. It is quite a natural reaction… with the elephant, you do want to have control of the legs first, and when you do, the trunk becomes much less anxious. You are much safer when the elephant really steps back, steps forward, and you can lengthen his steps.”

Was it difficult adjusting to working with such different animals?
“Not really, no. I must admit I was rather nervous about the idea in the beginning because they are such towering animals – thirty plus hands high –that’s a bit of a shock. The aids are different, but once you know what aids they use and what goals they have, it is really quite easy, you just fill in the gaps with learning theory.”

They don’t just hit them with sharp sticks?
“They do, that’s what I was trying to stop. It’s not that dissimilar to where the horse world is – it’s just that elephants are slightly less reactive to pressure, so the sticks are bigger. They just use bigger sticks and sharper spikes but in terms of the pain, it is similar.”

And the problem of trainers not understanding the mental set of the animal is much the same as in the horse world – the Mahouts believe that the elephants really know what they should be doing, and they are just being naughty when they don’t do it…

“The handlers expect that the animal knows what to do, and then the animal is in trouble for not doing it. That’s when the sticks and the ankus (metal spike) usually are reached for.”

And the answer – as with horses – is to break down the training demands to give horse or elephant, a chance to respond correctly:

“The problem is that the animal is met with too many demands at once rather than separating everything out into blocks of single learned responses. Just start off with a single response and be sure there is nothing dual about it, that it is purely a single response, and then shape that separately through all the steps. Once you get them to do that, the learning is really fast – we were training elephants under a rider in four days, when previously it had taken them weeks to do this.”

Tuikku (trainer from Finland) and I experimented with training the young elephants and then developed a training programme that amalgamated the use of positive and negative reinforcement (pressure-release). With no restraints or enclosures, our training was in a jungle clearing so positive reinforcement helped keep the animals with us. We could then focus on training the negative reinforcement components: the precise use of pressure-release to achieve results of greater duration and magnitude.

“The old training system of Nepal is similar to all of Asia – traditionally they bring the young elephants in, and haul them with an older elephant to a post and they tie them to the post, and weaken them with smoke for five or so days. Then when they are weak they can haul them about. Then they place a lot of demands on them – they get a general result, but it is blurry for the animal and highly dangerous for the Mahout.”

“If you are pulling the elephant forward and it has done 200 steps being pulled, then the animal doesn’t know what it is that finally made the pressure go away. The rope should have been released on that first step. That’s all we did. Teach a single step forward, a single step back. We did use food rewards in the beginning because we had no yards, we were working in a jungle clearing and the elephants could just leave when they felt like it. We had no ropes, so it was just us with sticks basically, to tap them – forward and back, and teach them. So the food rewards really helped a lot to get them to want to be with us.”

“We taught them the basic movement from a little bit of pressure on their ears and neck, which is where the Mahout’s feet go, to go forward and back and turn. Then we translated those into pressure aids and we did away with the food – we could say, ok, you know the pressure aid, now walk forward, then you would use the pressure a bit stronger and say, walk bigger, or walk faster, and the elephant would, and then you would stop the pressure.”

“That was the reason I was there in Nepal to make this conversion to teach them how to use their pressures. These people had been working with elephants all their lives and you would see them use their stick on the back leg to get the animal to go forward and they would keep hitting it for one step, or maybe ten, depending on the Mahout. They think the animal knows what to do, and if it doesn’t, it’s the elephants fault. At first it was difficult to teach the Mahouts to stop tapping the minute the elephant went forward, but when they did, they realised they had a much quieter animal, the elephants were much less aggressive.”

“It’s really similar to people working with horses. I’m always saying to people, when you are teaching your horse to go forward, why do you keep giving the aid when your horse is responding? Make it stronger if you have to – but stop when you get the result. The elephant trainers think if they give the voice command, the elephant should know what to do – that they are born knowing those commands. With the elephant it is important to get it going from the stick but only by tapping – not beating it. Teach it with constant taps faster and faster until the animal responds, and then very quickly shrink that to say, two taps to make the legs go faster, or one tap for making the legs go longer – so the whip is used as a stimulus rather than a general punisher that says do what I told you to do.”

“My first job in Nepal was basically to analyse the work the elephants do – so the first week was just spent going on expeditions, cutting grass – they cut grass, they also chase poachers – our team leader, who had worked with elephants in Africa, told me that elephants have a sense of smell sixty times stronger than a blood hound, they can sniff out the poachers. They haul logs, they basically do the work of bobcats and bulldozers, they can lift a ton with their trunk. Amazing.”

“So much of the work with the elephants so correlates with horse riding and training. The fellow on top is giving the elephants signals all the time. One elephant owner said ‘I’m really worried because the elephant has become quite dangerous. At night it sometimes tries to hook me with its trunk to squash me.’ At the conference we staged, I asked the trainers there, had they heard of people being killed by elephants? They all put their hands up. Training elephants is a very deadly kind of operation.”

Here Dr Kamil Ghari is translating my lecture to the Nepalese trainers. Dr Ghari is a highly intelligent man and quickly understood how learning theory works. He is the top elephant Vet in Nepal.

A young elepahnt learning the ear signals of forward using positive reinforcement. Tuikku implemented a ‘preschool’ for baby elephants beginning at one year where they learned basic in-hand commands using positive reinforcement.

“I asked if I could ride the elephant that was causing the concern. The Mahout sat at the back – I thought he was keeping an eye on what I did, but he was actually just text messaging.”

“I thought, if the elephant is aggressive, there must be some training issue here, or a management issue. I noticed that in this case, the trainer was constantly giving the elephant signals. I asked him, what happens when you don’t give it signals? And he didn’t really understand what I meant, because he was so used to giving them. So I had a ride myself and I found that every time I stopped giving it signals, it slowed down. I showed him how he might retrain that elephant and what he would need to do is use his stick, and rather than beating it over the head with the stick, just tap it sufficiently to produce the reaction, and then stop – and use the stick only for a while. Then when the animal was more sensitive to the go signal, then resume using his toes. As soon as the elephant showed any signs of not responding to the toes, go back to five more repetitions with the stick.”

“I do this with horses if they are lazy. I try to teach riders that a lazy horse is not a personality disorder; it is actually a learned response. Sure horses come to us with different genetic predispositions, but the outcomes are something we can change – particularly with dressage horses that are ‘lazy’, it’s not so difficult teaching them to go again with light aids. We did the same thing with the elephant…”

And it hasn’t killed anyone?
“I hope not, I haven’t heard anything – but while we were on it assessing it, it did bolt, which was fairly disconcerting. It took off with its ears out, then it suddenly spun. Apparently it did this quite often, and it was always scared of imaginary tigers in the forest. I explained in the workshop that this kind of behaviour is a direct consequence of holes in pressure release signals for stop, go and turn. You fix these and the animal is now secure, so he doesn’t get afraid in the jungle anymore. His fear and insecurity was sponsored by his lack of clear dialogue with the trainer. ”

“When you have to keep pressing the go button continually, then ceasing pressing becomes a stop signal, so it is not clear to the animal what is the correct response. It’s the same with the reins and the rider, often letting go the reins means ‘run’ and it shouldn’t be, the release of the rein pressure should be the re-inforcement of the slowing response. You can see how confusing this would be – the ceasing of the aids now result in the dead opposite of what should happen. Stop now means go and vice versa. No amount of voice commands superimposed on top is going to fix it. You have to fix the pressure-release”.

“ “They normally only break in elephants at the age of six. We wanted to start earlier, because like horses, they are really precocial animals, they can learn very early. We thought a four year old was strong enough to ride quietly, and it would be a good idea to get them trained earlier when they are not as strong and they haven’t had so much bad interaction with people, the only inter-action they really have is being chased away with sticks!”

“I went there with a Finnish trainer, she trained various animals, mainly birds for TV advertisements. She had been there the year before, using positive reinforcement, and she had got so far, but elephants do need to be trained using negative reinforcement because the Mahouts are sitting on them, and even if they are using very light pressures, it is still negative reinforcement. I was there to help make the jump from positive to negative reinforcement. You can see in the photos that the elephants are free, and we are in a jungle clearing with an elephant at liberty. Through positive reinforcement we got them started moving forward and from then on, we phased out the food to a large extent and swapped it with pressure-release. We needed to use food first because unlike horses, the elephants were free to run away, we didn’t have a lead rope or anything like that to say ‘comeback, don’t leave’. Our training was at their mercy -once they’d had enough, they would say ‘I’m out of here’ – it gave us a rest. We’d have a tea break (milky tea boiled on a fire) and then the elephant would come back. The food rewards did do that, they made them want to be with us in the initial stages of training. Then the consistent and well-timed use of negative reinforcement, through pressure-release training, also adds to the animal’s security and they always want to be with you. Positive reinforcement can still be used in the form of caressing. I don’t think you need to continue using food, so long as you switch to other positive reinforcement and use it when possible.”

“A well trained elephant is very nice to watch, but what they don’t realise is that the well trained elephant is a product of somebody who despite their haphazard training system, has done the right thing – and also maybe the luck of the draw with the elephant cottoning on to the right response quite early. The rest are left at various levels of confusion and certain amounts of confusion will actually lead to aggression.”

“There was one elephant that we trained stop, turn, and go, his Mahout was a very good trainer, he barely used the stick at all, he could do everything with his toes, he was a talented trainer, somehow he knew through his own experience when to release the pressure, and the other trainers didn’t. The difficulty was that some of the most senior trainers were also the worst trainers, but they had acquired their status thanks to the caste system, or their authority, not because they were good at training. Part of our job was to try and show them what training was as opposed to what was a traditional or religious component, and how to optimize training efficiency by understanding the basis of learning theory. We had to be very sensitive of course regarding the Nepalese religion and culture, but the people were as keen to know how to improve training as they wanted to make their jobs safer.”

“There are many religious aspects to their interaction with elephants. Every elephant’s name ends in either Kali or Prasad, which means goddess or god; there is a strong spiritual aspect to their connection with these animals – which makes you wonder, why do they treat them so badly in some circumstances? But it is only because they believe the elephants were given to them to help them live their lives, and when the elephant does the wrong thing, they think the elephant does know the right thing to do but is just being naughty. It leads to the belief that animals know the commands you are giving them, you just have to make them do it, they are just being willful and naughty.”

“Part of their problem was tradition, their culture is based on doing what their mentors told them to do. They say ‘you can never train an elephant younger than six’ – and it was difficult to say, yes, you can. In the workshops I tried to introduce the idea to them that elephants are precocial learners as opposed to humans – we need care and attention and are more helpless as babies.”

“It was satisfying working with the elephants because once we became clear with our signals, all the elephants became so quiet. One elephant that was aggressive when we started, he did attempt to kick and they kick sideways very powerfully and fast with all legs – they don’t just kick with their back legs. I was glad he never tried to kick me. However I was trying to find pressure spots on his body similar to those on horses – if you press the horse on his brachiocephalic muscle, he steps back, and I was trying to find similar muscles on the elephant, and he would just hold my hand with his trunk trying to keep it away from his body. But once he realised that when I poked him, he had to step backwards, he didn’t mind at all. Once he learned that he could switch my poke off – and in the end my poke was just a touch – by stepping back, he never tried to take my hand away. He became very soft and easy to work with, and I developed a strong rapport with him.”

“We saw so many instances of the trainers continuing to use the stick, for even one second too long, and the elephants became aggressive. Then the trainers would shout and move erratically, and want to beat the elephants. I pointed out to them, that every time they used pressure incorrectly, the animals became aggressive. But when they got a reaction from the stick, then shrunk it to a lighter version as early as possible, and stopped the pressure on the first response, the elephants melted.”

“If you use your whip tap on its own on your horse, you can see how little he does know. For example, if you use your whip tap for forward on his ribcage, and you tap, tap and tap – it might take twenty taps before the horse goes forward. Then the next time, it might take 15, next, 10, then 5 then just two, and we try and reduce it to one or two. But the assumption that the horse knows the whip is a real problem, not just for the riders but also for their horses – and the people dealing with the elephants are under just the same
illusion.”

“The training principles are really very easy to follow:
• We should reward the smallest sign of a learned response, then build on it gradually. Break everything down.
• Start with the lightest pressure and try not to remove it until you get the correct response
• Use as many signals as you like for a response but don’t have more than one response per signal
• Make sure signals are sufficiently different so the animal perceives the difference • Establish each signal on its own
• Build signals as a train where the new signal always comes first
• Don’t alter signals until you get the correct response – for example don’t start using a voice command to help you establish pressure release
• Just like you shouldn’t teach a child to spell by showing it a picture of a cat then punishing it for not writing ‘c-a-t’, break everything down to its smallest components. That’s all I do. That’s what trust is – the development of rapport through effective dialogue. Make it easy for the animal and be big and brave and accept that his problems are yours – look in the mirror and tell yourself that.”

Habituating the elephant to weight on its shoulders before mounting. We shaped this in the same gradual way as we train horses to be mounted at the AEBC.Centre: Mounted for the first time, Kush Prasad is being rewarded for immobility. We gradually reduced the food rewards and relied on both the release of the signals and tactile reinforcements for reward. We aimed to finish a set of training on a train of around 3 improved responses followed by rest of 2 or so minutes. 

On the fourth day, Kush Prasad and Saraswati kali were ridden in the open. Here Kush Prasad is doing a rather nice turn right. Elephants turn more easily by crossing their outside leg over the inside which is different to horses where the inside leg abducts in the swing phase as the first part of the turn. So the trainer must focus on putting those responses under stimulus control.

“The elephant trainers could see that we were making good progress, and achieving results in a short space of time, and when they mastered the techniques and saw the results, that helped a lot. They saw that it was not a ‘bad’ animal, not ‘nasty’, it was simply confused. For instance, if the trainer kept tapping for too long, or chose a different place, or had two people tapping at once, that all produced a confused animal, and the Mahouts saw that. They saw that if you asked correctly, the elephant would produce the right response and not try to run away or get aggressive.”

“It really is the same with horses. We certainly can produce all sorts of training outcomes – some really impressive outcomes – but horse trainers have never really got past that mindset of the ‘willing’ horse, the horse that wants to please you. I don’t think there is any training system for horses on earth that has embraced all the principles of learning. There is room for improvement in every training system I have ever seen – there are always wonderful aspects, but there are still holes you could fix by using learning theory more correctly. That is the future of dressage training, I’m convinced. Of course, the better the rider, the easier it is to teach them to use learning theory, and I am glad that European dressage federations are now embracing it.”

“What I wrote for the WWF Nepal working elephant programme was a training system, it’s really a training recipe that begins with groundwork, acquiring signals, habituation to a rider, then groundwork with a rider and then control taken by the rider. We broke down the steps of getting obedience to the signals, speed control, line control and trunk control in the direction of go, slow (including backwards) and turns. It’s nonsense to say, as horse trainers often insist, that you can’t make a training recipe. Of course you can: the ingredients are standard, it’s the amount of the ingredients that change according to the nature and experience of the horse. Those who maintain that there is no systematic approach are the ones with no tool box for problems – they can ride an FEI test but can’t load a reluctant horse onto a float, or train a needle-shy or clipper-shy horse to lose its fear of needles or clippers.”

“So with the elephants, we established movement with positive reinforcement. The trouble is that it is hard to reward the exact moment of correct response, so we used secondary reinforcers (as in clicker training). We used the Nepalese command ‘Leh! (like ‘good boy’) to define the precise moment of the correct behaviours. The famous human psychologist, B F Skinner’s biggest contribution to training animals such as dolphins and other performing animals showed that if you rewarded the leap in the air with a toot of a high pitched whistle – you rewarded the leap, and then when they finally came up to the surface, you could reward them with a fish – but you can’t give them the fish in the middle of the leap, and if you only give fish after the leap with no marker they don’t know what they did right that made it happen. That is what clicker training is about. I don’t use clicker training per se so much with horses, but I use the expression, ‘good boy’, and then give him a wither caress or maybe food. I do use clicker training though for certain behaviours, it’s powerful and can really help many responses.”

“We got the elephants going forward on the ground, and then habituated the elephants to the rider. First we pressed on the wither, then started to get the Mahout to gradually jump up bigger and bigger and finally get their leg over, then their body, then sit up, and the elephants weren’t disturbed by this. But if they went too fast, too soon, the elephants tried to shake them off, and they could! In the beginning when I said, now we are going to get them used to the rider, as soon as the Mahouts heard the word rider, a Mahout from nowhere would leap on the elephants back. They were very keen – then the elephant would shake, and when they shake their body goes in one direction and their head goes in the other, and it is very hard to stay on. That’s their version of bucking.”

“Once we had the rider on, we still had the person on the ground controlling the elephant with a stick – forward, back, stop, turn. Then the rider began to take control with the person on the ground phasing out, and the rider phasing in…”
“We showed that when you went quietly, it was much better. You should allow the animal to dictate its learning pace, but when you use your training tools correctly, and remove ambiguity and overlapping blurry aids, learning is surprisingly fast. I’m going back to Nepal for another three week workshop in November, it is an ongoing programme and interest is spreading to Sri Lanka, Thailand and India as they too are plagued by the same safety issues and they too hear the calls of the tourists who don’t appreciate the current tough methods.