Clicker Training - Is it a Gimmick?

Despite what many cynics have to say, positive reinforcement (or clicker) training is based on well-founded principles of learning theory. That is to say that dogs do things for one of two reasons, to achieve a benefit (reward) or avoid a dis-benefit (punishment). It goes without saying that, in the context of obedience training, it is much more motivating for the dog to work for a benefit than avoid a punishment (or correction).

Pavlov’s experiments with dogs involved presenting dogs with food and measuring their salivary response (how much they drooled). Then he began ringing a bell just before presenting the food. At first, the dogs didn’t start drooling until the food was presented, however, after a while, the dogs began to drool at the sound of the bell. They learned to associate the sound of the bell with the presentation of food.

The primary stimulus (that dogs respond to instinctively) is the food – the secondary (or conditioned/learned) stimulus is the sound which has become associated with the primary stimulus (or reward).

How do we use these theories in training?

The main principle of positive reinforcement is to get the dog to do something specific in order to gain a reward. The dog needs to understand that if he does something “right” he will be rewarded. If he doesn’t get it right he will not get the reward – no punishment, no correction. Punishment has its place but should be reserved for dealing with unacceptable behaviour.

Which reward?

Each dog is motivated by different things to different degrees. The reward needs to be potent to be successful.

If food is the reward, your dog must be interested in it – try making the food reward more tasty (ham, cheese, liver, sausage are all good, not only are they tasty but because they also smell) Food will be more interesting if your dog is hungry (try delaying his dinner…)

Toys and games are highly rewarding to very active dogs, probably more so than food. Toys also encourage speed and activity whereas food rewards tend to be more sedentary. Toys can be a much more potent reward than food although the play must always be on the handlers terms and should be short bursts of fun, timed carefully. Other rewards can work. Some dogs will love a cuddle or other form of affection although I would suggest that toys or food are more motivating.

What behaviour is being rewarded?

The biggest difficulty in training your dog is to communicate to the dog exactly which behaviour you are rewarding/praising. What would you like your dog to repeat next time? For example, if your dog is doing a recall and you encourage the dog all the way, then praise it at the end, how does the dog know which bit you liked? What if the dog had done a slow, plodding recall but had sat dead straight? You might know you are praising the dog for the straight front present; he may think you were praising him for his slow recall. Next time, he may go slower…

So how do we communicate what we liked? We introduce a “marker” used at the exact moment we want to mark/praise. This sound creates a “bridge” between the behaviour that is being reinforced and the receipt of the actual reward. This is where many people use a clicker. It is a specific, easily recognisable sound, which is easy to time and is not associated with other day to day activities. It is not necessary to use a clicker. Some people actually use a word or other sound (such as a tongue click) so that they do not have yet another thing to carry. It really doesn’t matter what you use, if your dog has been trained to respond to that stimulus. My advice, however, is the make the sound something that your dog only hears in that context. Some people use the word “good” or “yes” although if you already use these words frequently, they would lose their effectiveness.

Why should the dog respond to the marker?

In actual fact the dog still does not know what the sound of the “click” or other marker means. We have to train the dog (as in Pavlov’s experiment) that when we click (mark) a reward will follow.

This is easy to train. Simply click then reward (food/game/other) a few times. You will be amazed how quickly the dog comes to expect the reward. This is an important step. In order for the click to maintain its potency you need to enter into a contract with your dog. Every time you click/mark (even by accident) a reward must follow.

Next your dog needs to understand that he needs to do something (offer a behaviour) in order to get the click and treat (reward). This means that the dog has to start to use its brain. He has to think “what was I doing when I got the click/treat (C/T)?” in order to offer the same behaviour again. These thought processes make it much easier for the handler to communicate with the dog. It means we can ask the dog to go faster, straighter, tighter, higher etc and get a much more accurate response.

Where do we start?

Positive reinforcement is all about the dog offering a behaviour which is reinforced by the handler. If the dog understands what he is being rewarded for, he will repeat the behaviour (or offer a better one). Initially the dog has to understand the contract. First, simply place some treats on the floor and click every time the dog eats one. Then change the exercise and click first and follow up with the food. Once the dog has understood that the reward is delivered after the click you can start clicking for easy behaviours that the dog offers naturally (e.g. the sit). Much of my work is done off lead – this way I am not tempted to pull or encourage the dog into position. Purist clicker trainers advocate simply waiting for a desirable behaviour then C/T (click and treat). In reality it is quicker to “lure” the behaviour you want then C/T. For example, if you want the dog to sit, holding a treat above his head and moving it backwards is likely to encourage the dog into the sit position – at which instant you C/T

Shaping

Most obedience exercises can be broken down into numerous elements. Positive reinforcement allows the exercise to be broken down even further. For example, when training the recall at its most basic level, we want the dog to come to the handler (forget the wait, call, present, finish etc) but even this basic behaviour has variations. The dog that comes in on the arc, the dog that strolls over to the handler, the dog that runs straight past etc etc.

When we “shape” a behaviour we build the exercise up in minute stages. For example, if we wanted to teach the dog to twist around in front of us we could lure the dog’s head around to the left by holding a treat in the hand – C/T if the dogs turns its head to look at the hand. Repeat this 2/3 times. Then do the same but do not C/T until the dog has moved a foot towards the hand. Repeat several times. Then do the same but do not C/T until the dog has moved his body around a quarter of the circle. Repeat this a few times. Continue to wait for the dog to offer more and more until the circle is complete.

It could take you several training sessions to achieve this behaviour reliably. The first time you lure the behaviour but do not C/T the dog will probably be a bit confused (“this always worked before…”) but by simply waiting and not helping the dog, he will probably try out a few different behaviours before he offers the correct one – he may even walk off appearing disinterested – he will be back , offering you new behaviours, probably stronger than before. Have the courage to wait… He will work it out.

Having been through a barrier like this – finish your session on a high note. Give a handful of treats (or have a big game) and try again later.

Keep “upping” the criteria

It is important that the dog does not become complacent. Once the dog is reliably offering the desired behaviour, ask for more. Don’t C/T unless the behaviour offered is quicker or neater etc. or ask for the same behaviour but give less help (i.e. fade the lure). Also start to reduce the rewards – so only C/T after 2 or more repetitions. The C/T should become random over time so that it can eventually be faded altogether. Do not rush this stage. If you encounter problems, go back to C/T every time…

When do we introduce the command?

Until now, all of our clicker work has been lured or offered with no commands. When the dog knows the exercise fairly reliably, it is time to introduce the “cue” word. The word is a “cue” for the dog to offer the behaviour, it is not a command as you will not correct the dog if he doesn’t comply. Introducing the cue is easy – lure the behaviour as normal and add the cue word. If the desired behaviour follows C/T. (If the dog doesn’t offer the desired behaviour, you have added the cue too early.) From then on you only C/T if the behaviour is offered in response to a cue. The dog will stop offering the behaviour when you have not given the cue as the behaviour is not rewarded.

TIPS

Whilst you are practising the techniques, train some tricks. Avoid confusing your existing obedience exercises – teach the dog to twist, touch a box, roll over, shake hands, beg, bow etc.etc.

Always click whilst the desired behaviour is happening, then treat. We need to “mark” the exact behaviour that the dog is being rewarded for. What he is doing when the reward is delivered, is irrelevant

Only click once. If you want to indicate a particularly good behaviour/effort give a bigger/better reward rather than more clicks.

Once the dog has learned a new behaviour, don’t click for every correct response. You should vary the clicks – every 3rd, 10th, 5th response etc. This will make the dog work harder as he won’t be sure when he is going to get the C/T. A good analogy is a slot machine. If you won every time it could get boring. It is the uncertainty – the not knowing when you will win, that keeps you wanting more.

In one training session, don’t train more than one completely new exercise. You need to introduce a variety of exercises to keep the training interesting, but don’t confuse the dog by trying to shape more than one new exercise in any one training session.

Undesirable behaviours. If the dog is offering behaviours that are wrong, save the dog some effort (and yourself some time) by introducing a way of communicating to your dog that he needs to try something else. Some people use the word “wrong” or “Uh-oh” (“NO” should really be reserved for domestic purposes and is a bit strong in this context). Use sparingly!

Barriers. Dogs often create mental barriers that don’t occur to us humans. For example, if we shape a sit with the dog in front of the handler the dog may start to refuse to sit anywhere else (e.g. heel position). This is because the dog has worked out that he gets the C/T when he sits in front. Some dogs believe that they only get the C/T in the back garden and will not offer the behaviour in another location. You may need to go back a stage and re-shape the behaviour in another location (it will be quicker this time).

Other dogs. Don’t attempt to clicker train with another dog running around – put the other dog in another room. If they hear the clicks, they then will not expect the treat.

Clickers in a class environment. Dogs become very attuned to the tone, direction and distance of the clicker. Whilst they may be distracted by other clickers in the first instance, they soon realise that the treat only follows one specific click.

What happens if I make a mistake and click at the wrong time?. Laugh and try again. So what? If you reinforced a wrong behaviour, your dog may offer that behaviour again but the next time you will not C/T and eventually your dog will work out that it was wrong. Don’t forget your contract with the dog and follow up the mistaken click with a reward.

Should I never correct my dog? There are times when the reward on offer is simply no competition for another, self-rewarding behaviour (e.g. chasing squirrels) With the best will in the world, a bit of cheese is never going to lure a dog away from the adrenalin rush of the instinctive chase. In these instances (and other safety related circumstances) a good “old fashioned” reprimand is probably more effective.

Extinction. The above instances aside, there is no need to correct your dog as he will always be seeking the C/T. He will always be working out how to give you a bigger, better, faster, neater behaviour. If you don’t like the behaviour, don’t C/T. The dog will eventually stop offering the behaviour and it will become extinguished.

Please note that this article reflects my own opinions and experiences. If you act on this information, you do so at your own risk. Please read my disclaimer.

This page was updated on 29th May 2007