Theory and Background

This section contains articles about training theory and the theory of behaviour modification.

Classical Conditioning and the Clicker

How do you get paid? 

Does your boss stagger into your office groaning under the weight of fourteen cabbages, a lemon, six rolls of toilet paper, a jar of coffee, a few cans of petrol and twenty kilos of dog food? 

Probably not.

When you pay your bond, do you deliver three chickens, a pair of shoes and ten metres of fencing to the bank manager?

I doubt it.

I would love to deliver a cow, four goats, six snoek and a ton of good-quality, well-rotted manure to the Receiver of Revenue to pay my taxes.

But somehow I don’t think it would work.

 

So how do you get paid?

You probably get given a piece of paper.  Or possibly a lot of pieces of paper.  You work your butt off for an entire month, and you get a piece of paper in exchange.  And you keep on doing it!  Are you nuts?

Possibly not.

The piece of paper may be a cheque, or notification of a bank transfer, or you may be paid in cash, but whichever way you are paid, the piece of paper you receive represents the goods you can buy with it.  This is the function of money in our society.  It is a medium of exchange with which we can purchase the goods and services we require.

Why are you satisfied to receive money and not the actual goods and services you have earned?  Because you trust money as a medium of exchange.  Why do you trust it?  Because you have always been able to exchange it for what you wanted in the past.  In psycho-speak, you have been classically conditioned to form an association between money and the things you can buy with it.

Classical conditioning as a means of learning owes its discovery and terminology to the  Russian physiologist Ivan Pavlov (1848-1936).  Pavlov performed extensive studies on the basic reflexes of animals and in particular, the salivary reflex in dogs.  In his experiments, he would place food in a dog’s mouth and observe the salivary response.

One day, he noticed that his dogs started salivating at the sight of food.  This observation led him to formulate his famous bell-ringing experiment, in which he discovered the process of classical conditioning.  It goes something like this:

First, Pavlov noted that salivation was a basic reflex (as opposed to a learned behaviour).  He then noted that no learning (conditioning) needed to take place for food to cause salivation.  He therefore called the food the unconditioned stimulus (US) and the salivary response the unconditioned response (UR). 

(A stimulus is simply something external which acts as a signal to an animal, and a response is what the animal does after experiencing the stimulus.  For example, if you stick a pin (stimulus) into someone’s arm, he will flinch (response).  The pain response is another example of a reflex, or unconditioned response; you don’t have to learn that having a pin stuck in your arm is painful.  It just is.)

In psychology textbooks, which are all nearly unreadable, this is usually written as follows:

            US ————————-> UR

In English, the unconditioned stimulus (e.g. food) causes the unconditioned response (e.g. salivation).

Pavlov then added a stimulus, or signal, which meant nothing to the animal, so he called it a neutral stimulus (NS).  In his experiments, he used a bell.

To start with, he rang the bell and the dog, quite sensibly, didn’t salivate.  This showed him that the bell was in fact neutral (scientists have an obsession with proving things that are obvious to everyone else).  In psycho-speak:

            NS ————————–> no observable response.

Next, he rang the bell and gave the dog the food at the same time (in fact, the bell was rung immediately before the food was presented).  He did this several times.  Each time, the dog salivated.  Schematically:

            NS + US ——————–> UR

The next bit is the interesting bit.  Once Pavlov had paired the bell with the food a few times, he rang the bell without presenting the food and observed the dog – and the dog salivated.  The dog had made the association that the bell meant that food would surely follow.  The bell had thus become a conditioned stimulus (CS) and the salivary response had become a conditioned response (CR):

            CS —————————–> CR

Pavlov called salivation in response to the sound of a bell a conditioned response because the dog had to learn (or be conditioned) that the bell was associated with food.

Only a dog, you may be thinking, would be stupid enough to learn to drool at the sound of a bell.  Indeed.  It goes without saying that you, of course, have never felt even the slightest twinge of pleasure at the sight of a $100.00 bill – which is, after all, only a piece of paper!

Classical conditioning is an extremely powerful phenomenon which is at the heart of many of our instinctive emotional reactions, irrational fears and superstitions.  Properly understood and applied, though, it is also a very powerful therapeutic tool, as it tackles the ‘gut’ reaction underlying many problem behaviours.  It is particularly successful in the rehabilitation of aggressive dogs – and aggressive people! 

But how do we apply it to training?

One of the characteristics of the conditioned stimulus, e.g. the bell, is that the dog appears to perceive it as a cue that the unconditioned stimulus, e.g. the food will soon appear.  After very little conditioning, if the dog hears the conditioned stimulus, he will prick his ears up and look for the unconditioned stimulus.

Try it.  Do you have a ballpoint pen of the kind that has a nib which clicks up and down?  (A Parker jotter or similar pen will do very well).  Line up 20 treats on the counter in your kitchen (a frank cut into small pieces should work.)  Click the pen and immediately give your dog a treat.  Repeat this – click, treat, click, treat, click, treat – until you’ve used up 10 of your treats.  Remember to click first and then treat.

Now click once, but don’t give the treat immediately.  What does your dog do?

In all probability, he pricks up his ears, looks at you and says: “where’s my treat, then?” You have just classically conditioned a ballpoint pen as a conditioned stimulus.  To strengthen the association, carry on clicking and treating until you’ve used up all the treats.

Now remember that a positive reinforcer is anything which makes it more likely that a behaviour will be repeated (See Training without Pain on this site).  In other words, if the dog is reinforced after performing a behaviour you want, such as a sit, it’s more likely that he will sit again in the future.  Reinforcers such as food, water or sex constitute primary reinforcers because they meet a basic physiological need. 

Because you have conditioned the ballpoint pen as a conditioned stimulus which means ‘treat coming’, you can now ask, lure or coax Fido to sit, click the pen and then treat, and Fido will understand perfectly well when he hears the click that he is being reinforced for sitting, and that the treat (the primary reinforcer) is on its way.  The pen has become a secondary, or conditioned reinforcer (CR).

The most powerful conditioned reinforcer in our own lives is, of course, money.  People have been known to kill for it.  The association between money and reward is so strong for us that we work quite happily for money without ever thinking about precisely which primary reinforcer we are going to translate it into.

With persistence, the click takes on the same sort of meaning to the dog that money has for us, and develops tremendous reinforcing power in its own right.  It is important in training to maintain that association, and we do that by honouring the promise made by the click and treating every time we click.  After all, how would you feel if you walked into a shop and they refused to take your $100.00 bill?

In practice, we don’t use a ballpoint pen.  We use a clicker, which is basically a strengthened version of a child’s metal cockroach, but in fact we could use anything – a light, a bell, a whistle (deaf dogs can in fact be trained very successfully using a flash of light as a conditioned reinforcer).  There is no magic at all in the clicker; it does, however, have the advantages of

  • producing a sound which is dissimilar to most other sounds in the dog’s environment (take this one with a pinch of salt; Slug turns up looking for a treat every time I try to clip my toenails!)  

  • producing the same sound every time

  • producing a sound which is clearly audible to the dog at quite a distance

  • being quick and easy to operate (a reflex click will come out a lot faster than ‘good boy’ will)

These characteristics are important if the dog is not to become confused, and are the reasons for the clicker taking off as the conditioned reinforcer of choice in modern dog training.  It is also the reason that training using operant conditioning has, somewhat unfortunately, become known as ‘clicker training’.  I repeat:  there is nothing magic about the clicker, and adding a clicker to your normal training repertoire will buy you little or nothing.  Using the clicker correctly as part of a training system based on operant conditioning will, however, bring about astonishing levels of speed and accuracy, as well as improving your dog’s mental health (and his relationship with you) out of recognition.

At this point you may be wondering why anyone would bother with a conditioned reinforcer when it has to be paired with a food treat (primary reinforcer) anyway.  Why not just give the treat and have done with it?

If there are three things that are critical to successful training, they are: timing, timing and timing.  And there are three reasons for using a conditioned reinforcer (CR) such as a clicker: timing, timing and timing.

Although the concepts of operant conditioning were discovered in the laboratory by B. F. Skinner and his students, one of the first major applications in the real world was in the training of dolphins in aquaria.  It’s difficult to punish a dolphin if he does something you don’t like; he just swims away from you.  Choke chains don’t work on dolphins.

Furthermore, if a dolphin does something the trainer does want, such as a jump or a splash, by the time the trainer manages to get the treat (usually a fish) to him, he will probably have done several other things in between and may not even associate the reward with whatever it was the trainer liked.  He may eventually learn through trial and error that jumping will earn him a fish, but fine points like ‘jump high and to the left’ will be impossible to train. 

The use of conditioned reinforcers (a high-pitched whistle in this case) revolutionized dolphin training and made possible the almost unbelievably precise exhibitions we have come to expect from them.  The desired behaviour could be precisely marked using the CR at the moment it occurred, and because the dolphin had been conditioned to the whistle, he knew that he had been rewarded, and that the fish would follow; the likelihood of him repeating that behaviour thus increased. 

Yes, that’s all very well, you may be saying, but my dog isn’t swimming around under water when I train him; he’s right here next to me.  I beg to disagree.  First of all, any training which goes beyond the most basic involves some distance work.  Secondly, even when you’re right next to your dog, you have very little time to respond to a behaviour before he produces the next one; dogs move fast.  Studies show that for a dog to associate a reinforcer with a particular behaviour, the reinforcer needs to follow the behaviour within one second, and preferably within four-tenths of a second.   By the time you’ve mumbled ‘good boy’ and grabbed for the cheese, Fido could be over the hills and far away!

This becomes particularly important when shaping fine distinctions in behaviour; you might want to reinforce Fido for having a foot in a certain position, or being halfway into a sit.  Being able to mark the correct behaviour at the precise instant it occurs is probably the biggest advantage offered by a CR. 

The clicker thus has two very important functions in its role as conditioned reinforcer:

  1. as a cue that the treat is on its way, and

  2. as an event marker which marks the instant the desired behaviour occurred

The latter usage is critical when shaping behaviours – you may wish to mark a slightly straighter sit, a slightly faster trot, a raised head, pricked ears, you name it.  With the clicker, you can mark anything the dog is physically capable of doing, and this is what gives clicker training its astonishing accuracy and precision.

Can your dog chase his tail on command?  If not, try this.  (Tricks are a good place to start clicker training so you can hone your skills without adding some unwanted…um…variations to your obedience exercises!)

Get out plenty of treats.  Decide which way you want your dog to spin – left or right.  Let’s assume you’ve chosen the left.

Say nothing.  I repeat, say nothing.  This is not command-based training.  Your tone of voice is irrelevant.  In fact, you don’t even need a voice!

Start by clicking and treating (C/T) every time the dog looks to the left.  An eye movement is enough.  Keep going until he’s looking to the left at least 8 times out of every 10 trials. 

Up the criteria slightly.  Now you want him to turn his head slightly.  C/T for a slight head turn, don’t C/T for just an eye movement.  Count the number of successes and failures.  If he’s getting less than 2 out of 10 right, you’ve raised the criteria too sharply and he doesn’t know what to do – go back a step.  If he’s getting between 2 and 8 out of 10 right, he’s learning, but he hasn’t got it yet.  Keep going at this level.  If he’s getting more than 8 out of 10 right, he knows what you want and you can up the criteria again.  Perhaps you can C/T for a slightly sharper turn of the head.

Build your steps up gradually, asking for a sharper and sharper head turn, then a paw movement, then both paws, then a body bend and so on.  It’ll probably take a while and will seem quite slow compared to conventional training; but the important thing is that true learning is taking place.  There will be a point where your dog realizes that what he does influences whether you C/T or not; this is really exciting and you can expect to be jumped all over, several times.  His behaviours have now become truly operant; he deliberately operates on his environment in order to obtain a benefit.  Suddenly, your dog is training you, and just how intelligent he really is becomes astoundingly obvious!

Once he’s spinning away like a top, you can put the behaviour under stimulus control.  Add a verbal cue (we don’t call it a command any more) such as ‘Spin’ just before you C/T.  Gradually introduce the cue earlier and earlier.  The dog will associate it with the reward and will start offering the spinning behaviour whenever you say ‘Spin’.   Once this is completely reliable (at least 9 out of 10 successes in several trials), you can stop rewarding freely offered spins.  The dog will learn that you only C/T if he spins after you have given the cue, and that offering spins without the cue is pointless.  The ‘spin’ cue has become a discriminatory stimulus.   OK, you can call it a command if you really, really want to.   

Clicker training seems slow and painstaking at the beginning, but speeds up dramatically as the dog gets the idea.  Watching someone free-shape an experienced clicker dog is quite an experience; the dog starts experimenting freely to find the desired behaviour, and new exercises emerge with remarkable speed – and are remembered!  Morgan Spector, author of Clicker Training for Obedience, estimates that it takes about a third as long to put a clicker-trained dog through its obedience titles as it does a conventionally trained dog.  

The clicker training movement is busy revolutionizing the dog training world.  Suddenly it is possible to train accurate, reliable behaviours with no punishment or coercion; in fact, the training is done entirely hands-free, and the lead has become obsolete except as a safety measure.  The biggest operant conditioning success story comes straight from Skinner’s labs:  Robert Bailey and Marion Breland Bailey, two of Skinner’s most influential students, formed a company called Animal Behaviour Enterprises which over the last forty years has trained something like fifteen thousand animals of every species from cockroaches to elephants, including many dogs.  To improve your training skills, they offer a chicken training camp where the objective is to train a chicken to play a four-note tune on a xylophone – and believe me, it can be done.  In all their vast experience, they estimate that they have used punishment between six and nine times, and then only because their clients (who include the United States Department of Defence) insisted on it.  It raises some uncomfortable ethical questions about our obsession with punishment-based training, doesn’t it?

The clicker dog training movement was pioneered by Karen Pryor, a marine biologist and dolphin trainer who recognized that the principles of operant conditioning could be applied to dogs as easily as to any other species.  Her book, Don’t Shoot The Dog!, is a must-read for anybody wanting more information.  Amongst other things, she points out that dolphin trainers, who are accustomed to using positive reinforcement correctly every day, usually have exceptionally nice, well-behaved children! 

Clicker training has a very sound basis in scientific theory and uses a lot of scientific terminology, which may seem daunting to some people; yes, it is important to understand at least the basic concepts of operant conditioning.  It doesn’t really mix with traditional methods, and requires you to abandon much of what you have done in the past.

But once begun, clicker training, an art, a science and a sport, is a journey which is so enjoyable and rewarding for both you and your dog that I have never heard of anyone turning back!

Training without Pain

(Note: this article was written for a magazine called Dobe Capers during a period when I was the consulting behaviourist for the Dobermann Club of the Cape, so it refers a lot to problems training Dobermanns.  The theory which it attempts to explain is, however, completely relevant to other breeds, and in fact other species!)

If you are reading this article, you have probably trained (or tried to train) a Dobermann at some stage in your life.  Perhaps it was easy and enjoyable.  On the other hand, perhaps it was a constant battle of wills, a battle between you and a powerful, intelligent, strong-willed animal who loved you, but did not particularly want to do what you wanted him to, and resisted all (or most) of your efforts to make him.  Anyone who has been towed along by a Dobe supposedly obeying the command to “heel” will, I think, recognise himself or herself in this description.

Perhaps you have a dog who is loving and affectionate at home, but bored and resistant in class.  Perhaps he avoids you when it’s time for practice.  If he gets bored and resistant enough, perhaps you eventually lose your temper with him, shout at him, and try to force him to respond.  He becomes even more resistant, and has to be physically hauled into any posture you want him to adopt, which he abandons as soon as you let him go.  If you’ve gotten tough enough with him, you may even have been bitten.

Perhaps you’ve tried to do competition obedience with a Dobermann, and have watched your (naturally!) superior animal being easily outstripped by Border Collies, GSDs and other, well, nice dogs.  Frustrating, isn’t it.

Training like this can quickly become an unpleasant and distressing task, which is easily abandoned.  (Any trainer can tell you what the dropout rate from obedience classes is like).  When your dog loves you and is so affectionate at home, why go through the misery of fighting with him week after week?  It’s much easier to find an excuse not to go to class.  The fact that you end up with an unreliable and disobedient dog from a guarding breed with a high potential for aggression is just the cross you have to bear.  Dobermanns are difficult, and that’s all there is to it.  

If this has been your experience, don’t feel alone.  There are many others like you.  I have vivid and embarassing memories of…er…training my first Dobermann, a rather plain but fortunately extremely good-natured chap called Billy.  I was young at the time (this was 30 years ago) and Billy was very intelligent indeed; he certainly outwitted me every time he tried!  I did basic and advanced obedience, tracking, agility (which in those days was aptly named obstacle work) and eventually manwork with him.  It was uphill most of the way.  He learned everything under considerable duress, except the odd exercise which he enjoyed; those he learned quickly and happily.

At the end of his training, Billy was like the little girl in the nursery rhyme; when he was good, he was very, very good, and when he was bad, he was worse than horrid.  When he was good, he would work on hand signals from 100 metres.  When he was bad, he wouldn’t walk at heel.  When I did a right turn, he would do a U-turn and go and lie down somewhere comfortable.  He was trained, but he certainly wasn’t obedient.  Ring any bells?

Will power, stubbornness and resilience to punishment are characteristics of the Dobermann.  If its superb intelligence can be harnessed, it is capable of being an outstanding working dog – one of the best in the world – but it is by no means an easy dog to train, and this has led to the German Shepherd being preferred as a police dog in many countries, including South Africa.  In 1956, the New Zealand Police Dog Unit was established, using German Shepherds as the dog of choice.  Its founder and Chief Trainer, Inspector Frank Riley, actually kept two Dobermanns as pets (clearly a man of taste), but had this to say about their use as police dogs, having worked with them in the UK:

“This dog makes an excellent police dog, but matures slowly and for the best results needs an experienced handler who may have to experiment a little in his training methods. “

(from: Born to Obey, by Valerie and Colin Salt, Collins, 1972)

Ring any bells?

Top Cape Town obedience trainer Sandy Lombard says that the Dobermann is far more resilient to punishment than the GSD.  A Dobermann will stubbornly resist a series of harsh corrections which would permanently traumatise a GSD, and come bouncing back for more.  This determination and hardness of temperament is a wonderful characteristic for a police dog to have, but is offset by the difficulty of training such a dog. 

Although individual Dobermanns have performed exceptionally well at obedience, the breed does not dominate in the competitive obedience world, largely because of its stubbornness.  Dogs such as GSDs and Border Collies are far easier to get results with, and are thus often the choice of competitive handlers.     

Dobermanns can be managed after a fashion.  As I have grown older, I have become more authoritative, and am better able to persuade my dogs that I mean what I say.  They listen a bit better.  But I have to admit that I don’t really enjoy doing obedience work with them.  I don’t like speaking forcefully, correcting sharply, being in any way harsh with my lovely, affectionate dogs.  In fact, I’m really rather half-hearted about practicing obedience with Slug (my current male), and so is he.  It’s a frustrating state of affairs, because I really love my dogs and enjoy spending time with them.  Wouldn’t it be wonderful, wouldn’t it be marvellous, I have often thought wistfully, if Dobermanns were like Border Collies, always looking for work, always waiting for the next command, always eager to do what you ask them to?

Well, actually, they are. 

Actually, it is easy to harness all that will power, determination and superb intelligence and persuade your Dobermann (yes, that stubborn, recalcitrant you-know-what) to apply it in looking for work, working out what you want him to do, and doing it eagerly, just like a Border Collie, and with no harsh corrections whatsoever.

Does your dog have a reliable “sit”?  By a reliable “sit”, I mean that the dog is commanded to sit, sits promptly and does not move until given the next command, no matter what the distraction.

Slug doesn’t have a reliable sit.  He sits when told, but moves as soon as something distracts him.  As he’s extremely affectionate, if I tell him he’s a good dog in a pleased, excited tone of voice, he leaps around me in great excitement, trying to hurl himself into my arms and lick my face.  So I decided to teach him to sit until released.  As I have a fairly strong background in psychology, I designed the exercise myself.

Within three minutes of starting the exercise, he was sitting, while I praised him, told him what a good boy he was, how clever he was, sending him dilly with delight.  His entire bottom was waggling, his eyes were shining, his ears were up – and he was quivering with the effort of holding himself in the sit!  Being a Dobermann, of course, with all that intelligence, will power and stubbornness, he succeeded.  I used no correction at all.  In fact, he didn’t have his lead on, and I didn’t touch him once during the exercise, except to praise.  For the next 15 minutes, he followed me around begging for the next command!  Even one of his archenemies barking in the road outside my house failed to distract him.  I could not believe my eyes!

Just like a Border Collie?  Streets ahead!

At this juncture, you are possibly harbouring a suspicion that I was under the influence of some or other interesting substance while writing this article.  Not at all.

The method used to achieve this happy state of affairs is called operant conditioning, and has been used by animal trainers for centuries.  You have probably heard the term positive reinforcement somewhere along the line.  It is a term used extremely loosely and casually, but applied in its strictest sense as a training method in conjunction with the other concepts of operant conditioning, it gets results that seem little short of miraculous.

Needless to say, I am by no means the first person to have thought of training a dog this way.  (In fact, it took rather a long time for the penny to drop with me.)  Dr C.W. Meisterfeld, an American canine psychoanalyst who is the first ‘dog psychologist’ to have been certified as an expert witness by the US judiciary, has been developing a dog training method based on positive reinforcement since 1944.  In 1957 he entered the competitive world of American Kennel Club obedience to prove that these principles could be successful in training a German Shorthaired Pointer which others considered (at that time) a breed that was far too stubborn (ring any bells?) for competitive obedience. On November 10, 1957 at the Southern Michigan Obedience Training Club show, Meisterfeld’s bitch "Baroness Meisterfeld" received her third leg and the Canine Distinction Award for AKC obedience for earning an average score of 196-1/2 at three consecutive shows inside of seven days.  In 1962 "Baroness Meisterfeld C.D.X." won the National German Shorthair Pointer Retriever Championship with a (considered impossible) perfect score of 500 points. She retained the championship for 1963 and 1964 where she also won the 1964 National All German Pointing Breeds Championship.

And using similar methods, other trainers have achieved such titles and awards as:

  • Delta Society National Service Dog of the Year
  • Australian Police Dog of the Year
  • America’s Most Versatile Collie
  • World Record for Speed Weave Poles (set right here in South Africa!)

…and many more.

Beyond belief?  Not at all.

End of part one.  For a discussion of how the method works, see the next issue of Dobe Capers, which will appear in about three months time, or possibly later, depending on how busy the committee are……….

Continue reading