Not so long ago I wrote an article on the problems and high levels of frustration that clicker trained horses were facing, due mainly to the ways in which clicker training was being implemented. Not unreasonably the article was criticised in some quarters due to being too negative and being ambiguous as to whether or not I was in favour of clicker training at all. So here is my attempt to make amends; a constructive article about how we can keep clicker training a positive experience for horses. I really believe in the potential of clicker training to improve the lives of many horses. But I don't think that we are yet tapping into that potential; the use of rewards in training does not necessarily ensure a positive training session for the horse. So how can we improve on what we are doing?
Where to start? Not in front of your horse. It is so easy to get hold of a clicker and start clicking before you know enough about what you are doing. Knowing to mark the behaviours you like and ignore the ones you don't want is good but just isn't enough. A crucial but, bizarrely, often overlooked, factor is shaping. Karen Pryor outlined a wonderful list of "rules of shaping" in her book Don't Shoot the Dog (or see http://www.clickertraining.com/node/299) and it never ceases to amaze me that some professional trainers don't always pay attention - and, in my opinion, always to the detriment of their training. I am really not a fan of rules and normally see them as something to bend, stretch and generally play around with. But this is my one exception. The horses really need you to pay attention to these rules, otherwise you introduce sources of confusion and frustration before you even start.
Once you are ready to get back to your horse, the most obvious place, perhaps, is where the horse starts when he first experiences clicker training. We want to make this easy for him. We want the first behaviour we try to teach him to be intuitive, easy to achieve and hard to get wrong. For this reason I would always start with a non-threatening target. It appeals to the horse's natural curiosity; he can bump into it accidentally and earn treats before he even realises earning treats is an option. There are caveats to this for very nervous horses who might be afraid to touch the target, particularly if they believe that the click is coming from the target (like an electric fence - for example, you might want to delay the click until after the touch and/or use a much quieter click). But for the majority of horses the target exercise works very well.
"What about anti-mugging exercises?", I hear you say. The trouble is that mugging is a healthy part of the horse learning to clicker train. If the horse knows that you have treats and that he is supposed to be working out how to get them, he needs to explore his environment and try different methods of obtaining those treats. Mugging might be something he tries out, until he learns that other behaviours work better. As soon as we use any form of prevention or punisher (including inadvertent negative punishment - as perceived by the horse) we are giving the horse a very conflicting message - "offer behaviours spontaneously to earn treats, but only the right behaviours - or else". And once the horse is receiving conflicting messages, he is going to feel worried or frustrated or both. The target exercise IS an anti-mugging exercise. It simply provides the horse with a task which makes earning treats incredibly easy. The horse may attempt some mugging but as long as you are there with that target, he has all the information he needs to achieve the task. When dedicated anti-mugging exercises are used prior to starting target exercises we run the risk of making the horse's first clicker exercise much harder - "look away" is a much more ambiguous task which horses seem to find harder to understand. Look where? How long for? Leave head looking away or return to neutral? Swing head violently? And once the horse understands to "look away", he still does not necessarily equate that with "not mugging". Has he learnt a fairly advanced concept in this first clicker lesson, or merely an incompatible behaviour?
Of course, we need to stay safe and if you have the sort of horse who is going to mug you dangerously the moment he realises treats are available then, provided you feel clicker training is still an appropriate tool for you to use (and it may not be, at least not without the help of an experienced behaviourist), starting your task from behind a barrier is still a good thing to do. It will help the horse achieve the desired behaviour. But to actually remove the mugging from the equation, at some point you will need to remove the barrier and allow the mugging to extinguish. When we use extinction to remove an unwanted behaviour we can expect the behaviour to worsen temporarily before it improves; an "extinction burst" such as this can easily be confused with "training simply going wrong" so again, help from a behaviourist is advisable. Even once the mugging has extinguished, we can expect episodes of "spontaneous recovery" in the future as the horse "checks" to see if old and formerly reinforced behaviours still work. It can be a long hard process; alongside the "ignoring unwanted the behaviour" you need to be making every possible attempt for the horse to learn alternative ways of obtaining those treats. This is the difference between the horse feeling positively reinforced and negatively punished.
So now we have a horse who can perform a few clicker trained exercises and no longer mugs you (routinely, at least). But maybe you feel that clicking and treating for the rest of your life is perhaps not the point of clicker training. And rightly so, it is a tool to learning a behaviour; surely once the behaviour is learnt you should be able to remove the teaching aids? If a horse has learnt that every time he perform a behaviour he is going to receive a treat, he can become pretty resistant to the introduction of a variable schedule of reinforcement. And don't even think about trying to fade out those clicks and treats! The minute you try, you have a frustrated horse on your hands, feeling negatively punished and confused about the change of rules. As humans we feel very tied to the "one click one treat" approach to clicker training. We have an anthropomorphic sense that variable schedules are unfair. But tying ourselves into this obsessive clicking and treating is not doing the horse as many favours as we might imagine. If we can't make use of variable schedules then we can't really progress with our training. And if we continue to train at the same level then life gets pretty repetitive. And for "repetitive" you can read tedious, over-controlling and frustrating for your horse. To avoid this downward spiral of resistance when we try to incorporate variable schedules we can simply incorporate them from the beginning. That way we don't have a sudden change of rules. We can use intermediate bridges in the form of treat-less clicks; we can build in patience by having slightly longer pauses between click and treat. In fact, to save me repeating everything Ben Hart says in his book "The Art and Science of Clicker Training For Horses", I strongly suggest you read it.
There's something about clicker training which is pretty addictive - for the human. You have a horse, apparently begging you to be allowed to work more and more. It must be positive, after all the horse is choosing to do this. So the more we become reinforced by our horses' apparent enthusiasm, the more we decide to clicker train and soon we have treats in our pocket all the time and the horse is thinking he needs to be performing throughout all of our interactions with him. What might appear as enthusiasm is much more akin to highly conditioned default responses. Can you imagine the amount of pressure he must be under? His every move has suddenly become controlled by the human, deemed worthy or not worthy of a treat. Especially if this is a horse who has ever had food issues, such as long-term stabling without free access to ad lib forage. Or a horse whose treats are actually making up a significant portion of his daily food ration. All of a sudden an animal who has evolved to graze freely for 60-80% of every day, has had his food intake made contingent on performing certain arbitrary behaviours. Again, a very significant source of stress for him. We can resolve this by making our clicker sessions short and well-defined. We don't need lots of repetition or continuous sessions - just occasional clicker sessions, preferably with grazing always an option, to emphasise that the horse can make choices and explore his environment, without fear of consequences for "wrong" choices. Occasional free-shaped sessions like this give the horse a very powerful message, that he is in control of his own behaviours and can opt in or out of our sessions. There is nothing positive about feeling controlled all the time, even if you are being controlled via rewards. Having an increasing sense of autonomy is what really counts. The aim here is not to "resort to aversive techniques" in between clicker sessions, but to avoid the clicker sessions themselves becoming aversive. Mild, non-escalating pressure outside clicker sessions really isn't so aversive for the normal, well-adjusted horse (although yes, there are many horses who are sufficiently traumatised that they do need their human interactions to be pure positive reinforcement for a while, but still not ad nauseam) and is often actually less aversive than never-ending "positive" sessions - it's less about the treats and more about the psychological message of control which you are giving the horse.
If you do find your horse is becoming frustrated during clicker training sessions then the most advisable thing to do about it is stop. There are so many potential sources of frustration that we can't be sure of avoiding all of them. But we can use the suggestions above to get the horse to a point where he is able to learn to problem-solve and resolve any minor frustration before it becomes significant. What we want to avoid is using clicker training symptomatically to train away any signs of frustration. It's relatively easy to train "be calm"; at least, it is easy to train "standing still", with little regard for how the horse is actually feeling. Training "ears forward" is not the same thing as the horse not feeling frustrated in the first place. But, for the sake of argument, even if we believe that training the horse to engage in these "happy-looking" behaviours will result in him feeling happy (which personally I do not), we still run the risk of hiding information about how he is feeling in the early stages. If clicker training is a source of frustration then we need to know about it. If it is a source of sexual arousal (as in many geldings "drop" or become erect and/or masturbate - that's not a normal training response we want to encourage...) then we need to know about it. Clicker training has so much potential but it is very much in its infancy when it comes to application to horses. We can't afford to train away information about the effects it has on horses, we need to know what they are thinking for the sake of their welfare and our safety.
Of course, there are plenty of clicker sceptics. But writing an article about the problems that clicker training can cause, even when performed by experienced and/or professional trainers, does not make me one of them. Quite the opposite, I truly believe in the positives that clicker training can bring to the equine world. However, we have a long way to go before we can rely on a clicker session being positive. We need to stop being defensive about anyone who criticises the method and engage in open debate about why so many clicker trained horses become frustrated and/or sexually aroused, look so miserable and/or suppressed and perform behaviours so robotically. Then we will stop doing the process a disservice and finally find a way of training horses so that they genuinely feel positively reinforced.
Copyright Catherine Bell 2013