Coloured Australian Pony Stud

So what does elephant training have to do with horses or science for that matter, and how could “feel” ever be cold or hard?

I was fortunate to attend the RPSCA’s annual scientific seminar recently and the topic of this year’s meeting was “All work and no play; Modifying the behaviour of animals”.  A variety of speakers presented the results of their research into the training of animals. The talks covered dogs, horses, elephants, sheep, cattle, Tasmanian devils and exotic animals like giant pandas, rhinos and chimpanzees.  The speakers worked at universities, commercial settings, wildlife parks, zoos or as private consultants and all held masters or PhD level qualifications.  As well as conducting original research into their areas of expertise, all were accomplished animal trainers in their own right- no ivory towers here; bruises aplenty and manure on the boots!

In spite of the range of species trained by the speakers, they spoke unanimously on the qualities that make a good animal trainer and these are, timing and consistency. “Feel” wasn’t really mentioned,  though there was a fair bit of discussion of anthropomorphism (attributing human characteristics to non human things) and the dangers it poses to good training outcomes and ultimately animal welfare.  Dr Paul McGreevy noted in his opening address that just because you love your horse doesn’t guarantee that you will safeguard its welfare and that many of things we do to our horses in the name of ‘love’ can actually cause harm.  

The other key point of agreement was that in “training” an animal, what we are  really doing is putting an aspect of their innate behaviour (sitting, jumping, rolling over, cantering etc) under our control so that they perform it each time they are cued by us.  When we train an animal we are conditioning (teaching) it to respond to our cues, even when what we ask it to do is contrary to what it would rather be doing (eg being ridden in a 20m circle instead of loafing under the trees with a belly full of grass).   We don’t teach a horse how to perform a flying change, it already knows how to do that, what we teach it is to perform a flying change when we ask it to.

 The keynote address was given Dr Andrew McLean on his work with elephant trainers in Nepal and India. Dr McLean is an internationally respected trainer of horses and has an enviable record of success in retraining problem horses world wide.  He has been awarded Australia’s first PhD in equine cognition, so is very well placed to put theory into practice.

Applying learning theory principles, and in partnership with an experienced elephant trainer from Australia Zoo, Laurie Pond and other trainers from the WWF, Dr McLean developed a reliable and humane system for training young Nepalese elephants to be ridden, despite having no prior experience with elephants before he started.  

Working with the Nepalese mahouts and elephant trainers, over four days, previously unridden and largely unhandled elephants were able to be mounted and controlled with light signals delivered by the bare toes of the mahouts.  This training was completed without the elephants being restrained or tied up and even though  they would regularly take themselves off for a break from the training, they always returned of their own volitation and recommenced the training when they were ready.

So what’s this got to do with horses? We’re getting there….

 Like the horse world, elephant training has a long tradition of several thousand years of accumulated knowledge.  And just like we do with our horses, mahouts have strong opinions about the character and motivations of their charges based on anthropomorphic notions of respect, arrogance, submission and that the animals know what is right and when they don’t do as required they are being deliberately disrespectful.  Just like with horses, the key to being able to ride and handle an elephant is to have its innate behaviours (locomotion) under the control of the rider.

The traditional method of elephant breaking is just that- breaking the animal until it submits and the methods used to achieve this vary from country to country, but all involve rendering the elephant immobile, weakening it by deprivation of food and subjecting it to various painful and aversive stimuli until it stops reacting.  Dr McLean showed some footage of some of these methods and they were very difficult to watch, though not dissimilar in technique, if not level of force, to those used on horses today (hobbles, sacking out, dropping to the ground, tying up for hours and hours, tying heads to tails etc).  The result of this treatment is that the elephant learns that no response it can make stops the pain and discomfort, so its stops trying.  It develops what is called learned helplessness, where it becomes dull and unresponsive.  

At this point, the elephant is deemed to be suitably "respectful" of its human handler and is mounted and taught via aversive stimuli, the commands for go, stop, back turn etc.   The problem with this training method, aside from its obviously deleterious effects on the elephant is that the safety of the mahouts is seriously compromised.  Many of the elephants will go on to display what is called latent hyper aggression, which is an expression of aggressive behaviour later on and out of context.  Something happens that triggers an aggressive response in the elephant and it will then turn on the mahout or other people in its vicinity.  It is actually reliving the pain and fear it experienced when being “broken in” and being free of the restraints this time, can express that fear and aggression, often with tragic results for the mahout and the elephant.  In India 8% of all mahouts are killed by their elephants each year and no doubt you will have seen footage on the news or youtube of domestic elephants running amok.  

So what’s this got to do with horses? A little more about the elephants, because this next bit is key!  Dr McLean applied the scientifically validated principles of learning theory to train the elephants, demonstrating that when applied correctly, the outcomes of the training are predicably reliable and for any species of animal that can be trained. No whispering, respect or leadership required.

The point about learning theory is it’s not a system or method, but a set of principles which describe how animals learn new behaviours or responses to cues, what will motivates them to change their behaviour or response and also explains what has gone wrong when outcomes don’t match expectations.  Learning theory can be applied to any animal, any training system or method.

To get the elephants’ behaviour under the control of the mahouts, Dr McLean used a combination of positive reinforcement (food rewards), negative reinforcement (pressure signals and their removal) and classical conditioning (making an association between two events because one event predicts that the other event will occur- eg verbal command predicts food reward, animal responds to verbal command to get food reward).  

After four days of applying these techniques and carefully shaping the responses from a small try to whole movements (eg, standing still, to being mounted, to moving at the right speed and direction) the elephants were calmly and obediently under the control of their riders and the cues the mahouts were using were very light (toe pressure from bare feet).  Now these are not small animals and they aren’t ridden with any kind of bridle, saddle or restraint.  The results from the first visit have now been replicated at several other training centres, the Nepalese government has mandated its use for all elephant training and it is now being adopted in India.  

So what’s this go to do with horses?  We’ve finally arrived!  We use exactly the same techniques when we train our horses and using them, we can get the same kind of results.  

When riding and handling horses we primarily use negative reinforcement, which isn’t bad, it’s simply a maths term, which means that something is removed or subtracted to reward or reinforce (make more likely) a behaviour.

 Let’s take using your legs to signal to the horse to go- you apply a pressure on their sides, the horse doesn’t like the pressure on its side and wants it to stop, it moves forward and you stop pressing.  The removal of the pressure is the reward, next time you apply the pressure the horse is likely to move forward because that’s the behaviour that made the pressure go away last time:

Pressure of legs on horse’s side= cue

Horse walks forward=response

Pressure removed =reinforcement (reward)

So long as that pressure does always go away, the horse’s world is predictable and you have his go forward behaviour under your control.  The pressure motivates the horse to respond, the release tells it what it needs to do to make the pressure go away. You haven’t taught the horse to walk or gained its respect, what you have taught the horse to do is to start walking when you signal it to.  Its walking is now under your control.   Whether it’s walking forward from leg cues, performing passage, cutting out a beast, clearing a fence, or leading onto a float the horse is responding to pressure cues by doing something in order to make that pressure go away.

 It really is that simple and whether it’s your seat, your reins, the leadrope, the bosal,  a whip it’s all the same principle at work.

 Positive reinforcement involves adding something, usually food, to reinforce or encourage a behaviour.  It works in exactly the same way as negative reinforcement except that something is added instead of being subtracted:

Hand signal=cue

Horse nods its head= response

Piece of carrot= reinforcement (reward)

The head nodding in response to the cue has been reinforced by the addition of the carrot. Next time the horse sees that cue, it is likely to nod its head again, because that’s what made the carrot appear last time.  Because food is a highly motivating reinforcer- eg they love it(!), horses trained with positive reinforcement will usually pick up new responses or behaviours very quickly because they really want that food!  

 Both of these methods of conditioning (training) actually put the animal in control of its responses- it learns what it needs to do to get the outcome it wants (either food or relief from a pressure).  When we get it wrong and don’t reward it for the correct response it gets confused and tries new ways of getting its reward and often those new behaviours are ones we don’t appreciate so we call them problem behaviours and blame the horse for getting it wrong or believe that the horse is being disrespectful of our leadership.

Classical conditioning involves the animal making an association between two cues in which the first cue predicts the second one.  We use it in combination with either positive or negative reinforcement.  In positive reinforcement we might make a sound, a click,  then offer the horse some food.  Very, very quickly (normally about two goes!) the horse starts to expect the food each time it hears the click.   We have positively reinforced the clicker as a predictor of food.  Now when we want to let the horse know that a behaviour it has offered in response to a cue (such as a hand signal or voice command) is the right one, we can use the clicker(secondary reinforcer) to let it know that it got it right and that the food (primary reinforcer) is coming:

Hand signal=cue

Horse nods its head= response

Click= food is coming because you made the right response (secondary) reinforcer

Food= (primary) reinforcer

 By pairing the clicker (secondary reinforcer)  with the food (primary reinforcer) and then using the clicker to tell the horse it got it right we can get its behaviour responding to our cues even when it is working at a distance or we are riding it. The click allows us to mark the correct response without having to give the food reward straight away.

 We can use the same classical conditioning principle with negative reinforcement by pairing a low volume version of a stimulus- eg a light squeeze with our calves, with a high volume one (stronger squeeze):

Light squeeze with calves= soft cue

Horse doesn’t respond

Stronger squeeze with calves= stronger cue

Horse walks forward =response

Pressure removed = reinforcement

After a couple of repetitions...

Light squeeze with calves= soft cue that now predicts that the strong cue is coming

Horse moves forward=response

Pressure removed= reinforcement

Horse has avoided the strong cue by responding to the light cue.

This is exactly how the elephants were taught to respond to the pressure of the mahout’s toes. This pairing of the light cue to a stronger one allows us to train the horse to respond to light signals and thus we can largely avoid having to use strong cues at all.  Horses which are soft to ride or respond to subtle seat cues demonstrate classical conditioning in action.

So back to the elephants, instead of tying them up, subjecting them to beatings and using force and strength to control them, applied learning theory allowed the mahouts to use humane, gentle but consistent signals which were not harsh or punishing in their effect to gain complete control over the mobility and locomotion of 2 tonne elephants.  The ultimate result has been calmer elephants, calmer trainers and a much safer working environment for mahouts and elephants.  

As Paul McGreevy pointed out in his address, horses kill and injure more people in Australia than any other animal (except snakes and most people who are killed by snakes are tying to kill the snake at the time).  At the very least, from a safety perspective we owe it to ourselves to train our horses in ways that produce calm relaxed and obedient mounts, who don’t trial random hyperactivity (bucking, rearing, shying etc) or random aggression (biting, kicking) and who are not confused and in conflict because they are being subjected to pressures they can’t escape from (strong rein contact, using legs and rein aids simultaneously, strong punishments, harsh gadgets and bits, tight nosebands).  If a little learning theory can get control of an elephant without fear, force or intimidation, imagine what it can do for you and your horse.

There is a wealth of information about applying the very simple techniques of learning theory to horse training, a great place to start is www.aebc.com.au, the website of the Australian Equine Behaviour Centre which is run by Dr Andrew McLean, and Equitation Science, a comprehensive introduction to applying learning theory to horse training, written by Drs Paul McGreevy and Andrew Mclean.  And the beauty of it is, you won’t need to give up on your current methods, be they Parelli, Lyons, Monty Roberts, the German Training Scale, but you will be equipped to understand why they work and what’s going wrong when they don’t.  Whether its an elephant or a horse, learning theory gives us the tools to humanely gain control over our mounts so we stay safe and they stay relaxed.   Warm hearted science in action.

For video of Dr McLean's team in action in Nepal check out http://www.abc.net.au/catalyst/stories/2533487.ht