How to get perfect obedience
by Adam Beral
This article will give you the recipe for perfect obedience from your dog. In fact, it is a recipe that can be used for almost any behaviour in almost any animal. The reason I can start with such a confident statement is because what I am writing about is nothing new; it is good old fashioned, tried and tested learning theory. The principles of learning theory, if applied correctly and consistently, will never fail to work.
As a scientist by education and by mentality, I look for proof in everything and accept nothing without challenging it. This is why I find many traditional methods of dog training unsatisfactory, because I know them to be inherently less effective than modern methods. How do I know this? Because the studies have been done, and continue to be done, which prove the effectiveness of modern methods under controlled conditions. I am not referring explicitly to clicker training now (although the science that explains the efficacy of clicker training is as solid as any). Rather, any methods which allow associate learning to occur using positive reinforcement. Where I refer to clicker training throughout this article, I could just as well be referring to a trainer who says ‘good’ instead of clicking.
The first step in teaching a behaviour is normally to build an association using operant conditioning. To ensure that a behaviour is repeated, it needs to be reinforced. The most common reinforcement we use as dog trainers is food and because this involves the presentation of something that the dog finds rewarding, this is termed positive reinforcement. This is where the term ‘positive trainer’ comes from when applied to modern methods but actually, ‘positive trainers’ utilise a fair degree of negative punishment too, so the term is erroneous.
How you get the dog to perform the behaviour in the first place so that it can be rewarded depends on the situation. You may be a fairly hands-on trainer, or you may find the idea of manipulating the dog into position cruel. You may decide to lure the dog into position using a treat, as is commonly taught for basic positions such as the sit, stand and down. If the behaviour is more complex, you may need to break the behaviour into pieces and build it up (known as shaping). An alternative, known as free shaping, involves leaving the dog to work out for themselves what is required and rewarding approximations of it until you have the full behaviour. This is a technique used frequently by clicker trainers who can pinpoint very precise behaviours using the click and whose dogs are used to working for a reward (dogs trained using traditional methods often do not dare to show such variety of behaviours, and using free shaping can therefore cause them much stress).
Once your dog is reliably performing the behaviour for a reward, it is time to put it on cue. Now we use good old fashioned classical conditioning (as made famous by Ivan Pavlov). Timing here is crucial, as the strongest association will be formed if the cue is given fractionally before the behaviour is performed and followed immediately by a reward. Be careful here: there is a danger that if you always reward your dog with a food treat and then say ‘good boy’ that he will form a negative association with the phrase ‘good boy’ because it signals the end of reinforcement. Be sure to verbally praise before giving a food treat. This stage needs to be repeated until your dog is reliably responding to the cue 100% of the time. Do not be tempted to move on too soon by removing the reward too early thinking that your dog has ‘got it’. They will soon ‘lose it’ if the behaviour is no longer reinforced!
The big mistake that a lot of trainers make is that once the dog is reliably responding to a cue, they remove the reward completely. And what happens? Their dog stops performing the behaviour! This is obvious when you apply learning theory, as "any behaviour that is reinforced will be repeated, any that are not will disappear". So how do you make sure that a dog responds reliably to a cue for the rest of his life without the need for administering a treat every time? The answer is that once your dog is reliably performing the behaviour 100% of the time, you put the reward onto a variable ratio. Perhaps you reward once every three, four or five times (make it random though – dogs are good at spotting patterns!)
This method also has another extremely important benefit. If you have always rewarded your dog for performing a behaviour and then one day stop doing so (because your dog has ‘got it’) you will encounter a problem. Your dog is now expecting a reward and doesn’t get one; this situation is very punishing. There is potentially a deeper issue here as well because suddenly dropping the reward may lead to frustration. This is a big criticism of clicker trained dogs who are hardwired to work for rewards, to problem solve, and potentially get very stressed when that reward is not forthcoming. This can be prevented by increasing the amount of communication you give to your dog by teaching a ‘training is over’ cue, a ‘just be a dog’ cue and a ‘that won’t earn you a reward’ cue.
The final stage of training for perfect obedience is to make sure that your dog can discriminate the cue and that it has become generalised to any situation. This is done by getting into the usual scenario and saying ‘sit’ then rewarding your dog. Next, get into position and say ‘banana’. Your dog may well sit, but he will not get anything for it. What is happening now is that we are making it clear that the sit behaviour is only rewarded when we ask for it, so there is no point in presenting it otherwise (this also helps to eradicate that other bugbear of clicker cynics, the dog who has become addicted to the click and spends all day feverishly playing through a series of learned behaviours desperate for a reward).
Further, it becomes important to generalise the cue to different situations, different locations, different people and so on. Depending on what the behaviour is you are training, it is also possible at this stage to selectively reward the better examples and therefore polish the behaviour. For example, if you require a super-fast sit, in any training session you only reward the sits that happen in under a second. Ones that take longer go unrewarded. Or perhaps you want neat sits by your left leg; the same applies; only the best ones get rewarded. Your dog will soon figure out what is required, assuming you are consistent with your requirements!
Nothing in this article should be new to you; even if the terminology is foreign the methods will not be because to have achieved success with your own dogs you will have had to use some approximation of what I have written. However, it is worth reminding ourselves of these principles from time to time. Next time you pick up the ‘next best training tool’ in the local pet shop, or get told the ‘best way’ to train your dog to do this that or the other, think through how it might work. Application of the above methods will always result in learning taking place, no matter whether they are packaged in a little box that goes ‘click’ or not.
For any aspect of dog training, when we are faced with a dog that is not performing, all that is required is to go back to the basics of learning theory, figure out where we went wrong and build the association up from scratch. Yes, it sometimes requires imagination to engineer a scenario that will allow the dog to perform the behaviour we are trying to reinforce, but it will work. If you ever doubt the validity of learning theory, try clicker training another species. You will find that, without all the preconceptions involved in training dogs, you will have much success as long as you follow learning theory. Ask my two mice, Fievel and Moriarty – they will do anything for a click!