"THE ART OF POSITIVE REINFORCEMENT
DOG TRAINING"
[Clicker
Training][Respondent][Operant][Relationship][How does it work] [The
"click"][When do you click][What do you click][Will you
always need a clicker][Rewards][Other dogs][Starting][Food][Getting behaviour][Adding a "cue"][Fluency][Generalising][Why Tricks][Summary]
Dog Training is not new but as old as the laws of learning, in fact, learning is what Dog Training is all about. There are many ways a dog may learn to exhibit a behaviour without the constraints of ethical or moral values being imposed, nor indeed any greater ideal to serve us, apart from satisfying its own immediate needs for survival as an individual and as a species. Clicker training allows you to apply these laws of learning to train your dog not only with maxium efficiency, but also in a fair and humane way. It also allows you to build a positive relationship based upon mutual respect, trust and understanding with your dog.
The effect of the Laws of Learning on your dog’s behaviour that you observe may vary in efficiency depending upon which laws are predominantly present at any given time, however, the resultant conditioning of your dog’s behaviour has one common, consistent and simple fact:
"It is rewarding for your dog" in some manner or form, sometimes obscure, but "always rewarding" for your dog.
The Laws of Learning
The Laws of Learning that operate on your dog’s behaviour at any given time are influenced in degree by the environmental conditions that exist at that time, the associations with previously learned and to a lesser extent inherited responses. The most important are the Laws of Operant and Respondent Conditioning, (some other forms of learning are Observational and Cognitive conditioning, these are usually only seen in higher order animals). To understand these Laws and apply them to your dog training is not as difficult as it might seem.
- Respondent Conditioning: This is where the environment acts upon your dog resulting in a
(rewarding) reflexive response from your dog.
- Operant Conditioning: This is where your dog acts upon the environment resulting in rewarding consequences for your dog.
Both these forms of learning can exist at any one time, the methods used to train your dog will usually determine which is the strongest. Traditional dog training methods predominately use Respondent Conditioning, which require multiple repetitions over a long space of time to learn, creating a need to keep correcting the dog for making mistakes.
Clicker Training predominately uses Operant learning and Positive Reinforcement which is a far more efficient training method that teaches your dog which of his/her own behaviours are rewarding by catching (rewarding) your dog doing it right, this sets your dog up to succeed, it also allows your dog to learn how to learn.
Dog Training is not about teaching Good v. Bad behaviours, (dogs do not have a "bad behaviour" concept). To your dog "all behaviours are good" to some degree, training is about which good behaviour is rewarded most, when cued.
Behaviours whose consequences are reinforceing to your dog may increase in either frequency, intensity or duration. Behaviours that do not get reinforced will eventually extinguish.
As a Dog Trainer, you are consistently part of your dog’s environment, how you interact with your dog should always be in a positive manner if you are to build a good positive relationship with your dog. It is this positive relationship that increases the bonding and teamwork between trainer and dog that leads to excellence in the performance of learned behaviours.
Why do you need to have a relationship?
A good relationship is, in my opinion, what gives you the "right to train" with your dog. Clicker Training allows you not only to build that relationship, it also opens up a reliable "two-way communication" with your dog, this builds a mutual trust and respect between dog and trainer. This 'positive' relationship aspect of Clicker Training is one of its most important and useful characteristics.
"Clicker Training" is not a tool, a piece of spring steel with a dimple in a plastic box, nor is it just a sound, a command, or a signal, it is not reward nor food training, while all these may be used in dog training, it is "how" they are used that makes the method "clicker training". The focus of clicker training is allowing the dog to operantly learn the behaviours you want through the process of reward (positive reinforcement) and non-reward.
The actual clicker is a convenient and novel "identity" for this form of training that is known worldwide as "Clicker Training". It is not limited just to dogs, many different animals, sea-lions, dolphins, fish, birds, cats, horses, and even disabled children have benefited from this method. It allows communication across the barriers of language, species, and disability, it allows new levels of positive interaction and most importantly, it builds and repairs relationships.
Clicker Training relies upon the proven scientific principles of the Laws of Learning. The "click" is firstly conditioned using Respondent Conditioning by pairing the "click" with a reward, this makes the "click" become a conditioned reinforcer of behaviour, the "click" also predicts the reward. This is done with a powerful or unconditioned reinforcer namely food, because of its practicality in use, (sex and scent could also be used but difficult to apply) In fact anything that your dog finds rewarding or will strive to get can be used as a reward for wanted behaviour.
Your dog hears a "click", it means nothing to your dog, (unless previous associations, or sound sensitivity causes a response, in which case you have to neutralise this response) The "click" is followed a second or two later by a reward, just for doing nothing, this is repeated a number of times (respondent conditioning) till the sound of the "click" acts upon your dog to cause a response to the food, salivation. [ This is the same process (also known as Classical Conditioning), with which Pavlov at the turn of the last century discovered that his dogs had been conditioned to salivate upon hearing the sound of a bell, even when no food was present]. Soon the "click" has the same if not more powerful effect on your dog of predicting a reward, than just the food on its own.
When you now "click" a behaviour your dog is doing, the "click" tells your dog at that instant in time "that his/her own behaviour" is rewarding. When conditioned, the "click" will also cause an alert response in your dog and it also ends the behaviour. Ending the behaviour will not effect the learning as it is what behaviour you "click" that gets reinforced. The "click" is now an "event marker" and a conditioned reinforcer for behaviour.
When the "click" is a conditioned
reinforcer it communicates to your dog that at that instant in time, that behaviour is
rewarding, the "timing" of the "click" is probably the most important
aspect to consider, good timing will give the dog a consistent 'signal of reward' for a
particular behaviour. Sloppy timing may cause confusion reinforceing many different
behaviours and resulting in a slower rate of progress in training your dog. It is useful
to liken the "click" to the click of the shutter in a camera when taking a
photograph, what behaviours you see on the developed photo is what the "click"
reinforced.
Before each training session you can "charge up" the clicker by doing a quick re-conditioning sequence of pairing the "click" with food a number of times. This is optional, as once conditioned the clicker (CR) can be used at any time to reinforce wanted behaviour, the main advantage of "charging up" is that it also gets your dog to focus on you for a training session.
Only give one "click" to mark a behaviour or event, you may vary the reward for different degrees of excellence. But always reward every time you "click", this is very important.
When you "click" a wanted behaviour, to reinforce that behaviour, the "click" may cause your dog to stop the behaviour and seek the reward, this will not effect what actually got reinforced by the "click" as the "click" also ends the behaviour.
What do you use the Clicker for?
The Clicker is mainly used to shape behaviour, to reinforce small changes in a behaviour in order to change that behaviour, bit by bit, by slowly raising the criteria of what behaviours earn a "click" and reward, changing it gradually into the behaviour you want. This selective reinforcement process is called "shaping" a behaviour.
A Clicker can be used to shape any behaviour that your dog is mentally or physically capable of doing, limited by your imagination, patience and understanding.
Will you have to always use a Clicker?
No, only for shaping behaviour, you can have many conditioned reinforcers, you can use, a tongue "click", a short word like "yes" or "bing", any sound, object or movement, all can be conditioned and used exactly the same way. Some conditioned reinforcers may be more consistent and precise in giving your dog information than others, but you need never be without a "clicker" (conditioned reinforcer).
When a behaviour is "shaped" to what you want, you can put a "cue" (a command, in old money) on that "known" behaviour. There is no need to further use a clicker to shape that behaviour, but remember you still need to keep a good history of reinforcement for all wanted behaviours.
What rewards can you use with a Clicker?
Anything your dog finds rewarding, food, toys, play, attention, a pat, a tickle and a scratch to name a few, the best approach to this is to have a reinforcement grading, from a verbal "good boy" at the bottom to what your dog finds most rewarding at the top. This reinforcement variability will increase the response from your dog not knowing what the next reward may be. Always reward good behaviour.
Will other "clicks" effect your dog?
At first it may, but your dog will soon be able to differentiate between who is "clicking" because of the other important aspect of the communication between the dog and trainer when clicker training that of your Positive Relationshp. In Clicker Training Classes the close proximity of dogs and trainers has never been a problem.
The best and most efficient way to train using a clicker is to "plan" in advance what you are going to train, what aspect of the behaviour you are going to shape, i.e., intensity, duration, frequency, distractions.. etc.,
Start at home, in a safe area, have your dog off lead so as not to inhibit or lead your dog. Elsewhere, a flat collar and lead (for safety) should suffice.
Try to keep treats/rewards off your person, in a cup on a table or chair, even hanging on a fence nearby, this will introduce a slight delay between the "click" and the following reward.
Keep training sessions short, your dog also learns from a process called Latent Learning, the break between training sessions allows for this form of learning to occur which will speed up the rate of progress in training.
Many Dog Trainers misuse the use of food in training dogs, in many cases food is used as a lure to "bribe" the dog to simulate a wanted behaviour. The resultant behaviour soon becomes dependant upon the presentation of the food before the behaviour occurs. In fact, many "lure" trained dogs do not associate getting the "lure" with the behaviour the handler thinks is being rewarded, "My dog will not work if I do not have food" is an all too common result. The reward should only follow after the wanted behaviour. Using a "click" allows you to reinforce instantly without the need to have food ready to put into your dogs mouth, which would make the food also a stimulus for the behaviour. Apply food only as a reward "after" the behaviour has occurred, apply it as a stimulus makes it a "lure" or "bribe".
A dog that is focused upon following a reward is learning to take rewards, this "see it, take it" attitude can cause many problems if not addressed early in training. A reward earned in training should be "given" to the dog, not taken by the dog, this also applies to any reward in the dogs environment, other dogs, play, people, animals, etc., To train this very important aspect of training you teach your dog the "Leave-it" behaviour, or a signal of non-reward, this tells the dog "try to get a reward and it will disappear, give me your attention instead will earn you a reward".
There are a number of ways of "getting" the behaviour you want:
Reward - Operant behaviour
First get the behaviour, wait for any example, or nearest example of "offered" behaviour that you can shape to the wanted behaviour.
Reward - Lured behaviour
If you cannot, or due to time constraints wait for a "offered" behaviour, you may have to "lure" the nearest example of a behaviour that you can reinforce. Repeat this three or four times, "click" and reward each time, then wait for the behaviour to be "offered" (operant behaviour). Because the use of the "lure" causes an immediate history of reinforcement of a similar behaviour, the behaviour you want has a high probability or operantly occurring.
Reward - Target-lured behaviour
Targeting can also cause the behaviour to happen, this is where the dog is taught to touch a target and will then produce an example of the wanted behaviour in doing so. Just like "lure" training the target needs to be faded out to allow operant behaviour to occur.
Reward - Modeled behaviour
Sometimes the use of modeling or physical conditioning a muscle pattern may help in getting a behaviour started, this should be used sparingly as it inhibits your dogs ability to "think" and "work it out" without your assistance. If you need to use it, do so only two or three times, "click" and reward each time, then wait for the dog to operantly offer the behaviour. With modeling many unwanted signals (or stimulae) can become attached to the behaviour, touch, pressure, direction, body posture, and also environmental, that will need to be faded out, so only use if you have to.
When you have the behaviour captured and shaped in your training environment you then have to change to many different environments as you can to generalise that behaviour so that the environmental effects on the behaviour are reduced as low as possible.
When you have the behaviour shaped to what you want it to be, you can proceed to attach a "cue" to that behaviour the process is simple but very effective.
First, you only give the "cue" when you know the behaviour is going to happen. The "cue" is given just before the behaviour happens so it becomes a stimulus for the behaviour, this is then followed by the behaviour itself, and lastly, the behaviour is reinforced with a reward, a positive consequence.
The "cue" predicts that the behaviour that follows the "cue" will be rewarding. Only when you have the behaviour as you want it do you attach a cue to it.
When you have a behaviour with a "cue" attached you must be selective in not reinforcing examples of that behaviour that occur without the "cue", this is called "stimulus control".
If you want to attach a "new" cue to a behaviour that already has a "cue", the new "cue" is followed by the old "cue" which is then followed by the behaviour and rewarded. Soon the new cue will also predict a reward for that behaviour, and the "old" cue can be faded out.
Fluency of behaviours on "cue"
Fluency or speed trials are a good way to train your dog to be fluent in a behaviour. These are simple and fun to do, first set out a fixed number of treats or rewards, (say twenty), and time how long it takes to get that many examples of the behaviour (rewarding each time). As the fluency increases the time it takes to perform the behaviour gets shorter for that fixed number of examples of that behaviour. Be careful to only accept examples of that behaviour that meet the criteria for that "cued" behaviour.
Whenever you train a behaviour in a particular environment the environment itself also becomes part of the stimulus for that behaviour, the behavioural response may change if the environment changes. To reduce this effect on trained or learned behaviours you have to train in many different environments to allow your dog to generalise the behaviour so that the "cue" remains the only consistent predictor of reinforcement for that behaviour.
Most training problems are the result of the lack of generalisation in training, how many times have you heard "But my dog does it perfect at home".
When first starting to learn "Clicker Training" it is sometimes better to start with a behaviour that has no "baggage" attached from previous training concepts and methods. Trick training allows for this to happen, start with any behaviour your dog naturally does, for example a stretch behaviour after a sleep can be shaped into a "bow", raising a paw can be shaped into a "wave", a head turn can be shaped into a "spin". The skills learned in shaping simple tricks can then be applied to more complicated tricks, for example; correctly negotiating the obstacles in Dog Agility, teaching formal Obedience exercises "Heelwork", "Retrieving", "Scent discrimination", to teaching your dog to "Track" a scent, to name a few, they are all behaviours.
see [BEHAVIOUR] [TARGET] [TRICKS] [AGILITY] [OBEDIENCE]
If you have a good relationship with your dog the only limit to what you train is limited by your imagination, understanding, and your dog’s operant behaviour and ability.
In Summary
- Allow your dog to discover which behaviours are rewarding.
- Catch your dog doing it right.
- Only one "click", and always give your dog a reward.
- When shaping behaviour use small increments or steps.
- Shape only one aspect of a behaviour at a time.
- Raise the criteria slowly, use small steps.
- Only reward behaviours that meet the criteria.
- Always reward good behaviour.
Never punish your dog
Ignore "bad" behaviour, teach a "good" behaviour instead.
When a behaviour is on cue, ignore un-cue’d behaviour.
The "click" ends the behaviour.
Have many conditioned reinforcers to use with your dog.
Plan your training sessions, keep them short.
Better to "click" early than too late.
One "click" and jackpot rewards for excellence.
See each step when shaping not the finished behaviour.
Having a good relationship with your dog is your reward.
Clicker Training is fun for you and your dog!
Clicker Training is not a Command based training system try not to lure, or influence your dog’s behaviour before it happens, allow your dog to make the behavioural decision that has the most rewarding consequences. You can do this by providing the rewarding or non-rewarding consequences for your dog's behaviour. Any behaviour that is non-rewarding will eventually diminish and disappear.
"Clicker training is fun",
Robert Loftus
İRLoftus1998