Friday, September 30, 2011

Treats - Why Treats?

Why use food?
Just as we need encouragement and payment in exchange for doing a job, your puppy also needs a payment for a job well-done. She is not able to think and plan for the future.  She will not learn a new skill just because it may help get a more interesting job in the future. Your dog lives in the moment and will focus where ever it finds a payoff--right now! Your puppy's environment provides lots of payoffs--interesting smells, movements to follow and tastes to try.

Gimme here:  If your dog training instructor tells you not to use treats to train your dog - well, "nuts" I say.  As in, he or she is a nut.  You just ask that stingy so-and-so how many times you can come to dog training class without paying for the class.  If he expects to be paid to do the work of teaching you to train your dog, then your wonderful dog should also be paid for working for you.  Dogs work much cheaper than humans.  I'm just saying.

It is a challenge for humans to be more interesting to their puppies than the environment is.  But humans are clever! We can provide our puppies with great smells, unpredictable movement of toys and extremely tasty treats. We can introduce activities that provide a puppy with puzzles to solve. When you are unpredictably fascinating, your puppy will find it worthwhile to pay attention to you.

Do remember:  If you are also unpredictably punishing, your puppy may not want to be around you, as the uncertainty of unpleasant experiences outweigh the possibility of wonderful experiences.

A puppy or dog who loses interest during a training session is probably finding the payoff to be too little for the amount of effort required. Lower your standards so that your puppy is successful and receives payment more often. As puppy gains confidence through many successful trials, you will be able to slowly raise your standards again.

Which Treats?
When shaping behaviors it is best to be able to have as many trials as possible, with very little time between the trials. Small bits of food can serve as convenient payments during a shaping session. You want your dog to eat the treat quickly so that you can go on with the next shaping trial. You also want the treat to be valued enough by your puppy so that they continue exploring the learning puzzle.

Some guidelines for food treats
Soft, moist treats can be swallowed more quickly than crunchy ones.
Smelly and tasty treats are more valuable.
You can use less of a treat that your puppy really loves.
Use the most valuable treats for the most difficult learning tasks and situations.

Possible treats
String cheese
Refrigerator jerky
Nuked Hot Dog slices
Liver Brownies
Nuked or boiled beef liver
Nuked or boiled chicken

Beyond Food
Food is not the only thing you can use as payment for a job well-done. Tummy rubs, a game of fetch or tug and other activities can also be used as a payoff. As you get to know your puppy you will become more and more aware of what is motivating to her. As your training relationship develops, knowing what motivates your puppy will be very valuable for you.

Remember – the puppy always decides what is motivating, i.e. they don’t have to like it just because you think they should...

Rewards 102

Primary reinforcers
"Primary" just means something that is hardwired into the dog to want; usually related to survival needs, such as food, drink, some touch, and sexual stimulation.

Some breeds have hardwired certain drives that are primary reinforcers, such as vermin for terriers, birds for some hunting breeds, visual stimulation for sight hounds and some sporting breeds, and sheep for herding breeds.  These breeds are often highly sensitive to and easily distracted by these "special" primary reinforcers or things that are similar to them.  For example: movement is distracting to all dogs, but especially distracting to sight hounds and herding breeds.

Some secondary reinforcers can get so strong that they become primary reinforcers, such as tugging.  Dogs aren't born knowing the tugging game.  When they learn to tug it starts as a secondary reinforcer; later it can become primary.

Secondary reinforcers
"Secondary" means something starts out as neutral and had no meaning on its own, but then it is taught or "conditioned" to have value.  Secondary reinforcers depend on you teaching (conditioning) the dog to mentally "pair" it with other primary reinforcers.

For dogs, these conditioned reinforcers can include smiles, praise, attention, clapping, toys, tennis balls, Frisbees, petting and the sound of a clicker.  For humans this includes praise, smiles, thumbs-up gestures, an "A" on homework, and money.

Gimme here:  I love it when my person laughs at me, so I've learned to do a lot of funny things to make her laugh.  Any time I do something that makes her laugh, I remember and do it again and again.  She tells people that I am, "in touch with my inner-clown".   Really I just know when she is laughing, she is happy.  And when my person is happy, I'm happy because I know fun is about to erupt.  Its a great system we have going.

Some conditioned reinforcers, such as the clicker, act as both an "event marker" and a "bridge".  An event marker precisely marks a moment at which the dog is doing a behavior that earns a reward.  It then bridges the gap in time between the moment when the behavior happens and the time when the reinforcement can be given.  Even though you have a bridge, it is still important to deliver the reinforcement as soon as possible or you could also be reinforcing unwanted behavior that happens in the meantime.

Tertiary Reinforcement
"Tertiary" is a special class of reinforcement.  It is when a cue for a known behavior acts as a reward for whatever behavior came before it.  Dogs in agility are rewarded in this way, where cuing the next obstacle reinforces completion of what came before.  It is the opportunity to continue on course and get closer to the reward at the end that makes the next cue reinforcing.

A well trained dog that works for rewards will enjoy working and tertiary reinforcement is highly effective for them.  A dog that is trained with much punishment cannot benefit from tertiary reinforcement.  The idea is that:

      ·   the reward for behavior one is the cue for behavior two
      ·   the reward for behavior two is the cue for behavior three
      ·   the reward for behavior three is the cue for behavior four
      ·   the reward for behavior four is the cue for behavior five
      ·   the reward for behavior five is click/treats

There is a flip side to tertiary reinforcement that you need to know about to keep from unintentionally rewarding undesirable behaviors.

When the dog does something bad, many people will cue it to do something else and then reward the dog for doing that instead.  Since the cue for the new behavior gives the dog the opportunity to earn a reward, it acts as a reward for what the dog did moments before hearing the cue.  I call this I'll Be Bad So I Can Be Good And Get Rewarded syndrome, which is very common with extra smart dogs.

Gimme here:  Just recently I met a horse that was being trained by his person.  When they took a break from training, they came over to where we were.  I barked at the horse to tell him not to come too close (and to tell my person that the horse was close enough).  The lady with the horse asked me to sit and then gave me a treat.  I quickly realized the game possibilities and barked again.  Sure enough, she asked me to sit and then gave me another treat for doing so.   After that I barked a lot and each time she told me "sit" and then rewarded me when I did.  Boyoh was I ever training that lady to give me treats.

Unfortunately, my person knows how that game works and she doesn't want me to bark and she really doesn't want me to be rewarded for barking.  So she made the nice lady stop playing that game with me.  My person explained to her that she should only ask me to sit when I was being quiet, so I didn't play the be-bad-to-be-good game.  After that when I wasn't barking, the nice lady asked me to sit and then rewarded me when I did.  I know that game too, so I was happy to play it.  Then everyone was happy.  My person, nice lady and the horse were happy that there was no more barking; I was happy because I was getting lots of yummy treats.  What a great deal for me.  Life is good if you are trained with rewards.  I'm just saying.

Expand Your Knowledge
There are times when something a dog normally finds rewarding isn't going to be rewarding.


      ·   Being warm is rewarding when its chilly, but not when it's
           too hot already.
      ·   Food is a common reward, but when your dog is too full it
           won't work well.
      ·   Dogs that love tennis balls, may lose interest when distracted
           by a female in season.

The same principle works for humans too.

      ·   Being admired by someone is great, but not when that
           someone is creepy.

As important as it is to know what things your dog thinks are rewarding, its also important to know when those rewards aren't rewards at all.  Examine your list of your dog's faves and think about the times when they might not work for you.

Thursday, September 29, 2011

Rewards 101

What is Reinforcement?
Reinforcement, or a reward, is anything your dog will work to get.  If a behavior is happening more often, then it is being rewarded in some way.  Many behaviors are self-rewarding and they will happen more often over time, since the dog simply enjoys doing that behavior or enjoys what it gets from doing that behavior.  An example:  Many dogs enjoy barking and over time will bark more often.  Another example: dogs generally like food and free food even more, so counter-surfing is self-rewarding.  Dogs love to smell good stuff, so even the chance to smell what is on the counter will reward counter-surfing behavior.

Reinforcement is done in two ways - adding a good thing or taking away a bad thing.

If you give your dog a treat when she walks with you on a loose leash that will happen more often.  Likewise if you only move forward when the leash is loose, that behavior will happen more often.  Both are examples of adding a good thing to reward behaviors.  Reward based trainers use this kind of reinforcement most often.

If the dog's behavior makes something "bad" go away (taking away a bad thing) then that behavior tends to happen more often.  Reward based trainers are very careful about using this kind of reinforcement, because you have to add a "bad" thing before you can take it away.  So reward based trainers only use this type of reward in very special cases, such as when the bad thing is already there.

For example:  One special case would be teaching your dog to not fear something, since the scary thing must be present in order to work on becoming less afraid of it.  To reward good behavior in the presence of the scary thing, either the dog can move away from it or the scary thing can move away from the dog (plus give treats at the same time).  Both ways take away stress and scared feelings, so the calm not-afraid behavior is rewarded and will happen more often.

Please be very careful if you try this by yourself.  Working with dog fears and anxiety can be tricky.  Any work dealing with fear or anxiety must be done below the dog's fear threshold.  So, if your efforts seem to be making the matter worse or if the dog can't eat the treats when the scary thing is there, please find a skillful trainer that uses only reward based training to help you.  

Gimme here:  When I was afraid of horses, my person never made me go close to them.  Forcing me to get close to something I was afraid of would only have made me more scared.  On top of that, I would have worried whether my person could be trusted; I need to know my person will keep me safe.

Instead, we watched the horses from far away where I wasn't afraid.  My person also gave me lots of peanut butter to lick up while I was watching the horses.  Peanut butter is the BEST!  I sure love that stuff.  Every time I saw horses I got to eat peanut butter, so I very quickly learned when horses are in sight, peanut butter is on my lips.   It took many months, but now I like horses a lot better.  Just recently I sniffed a horse's nose and let him sniff me.  I'm getting braver all the time because my person doesn't force me to do something I am not ready for.

The one who is receiving the reward (in this case the dog) is the one who "decides" what is rewarding and what is not and how much so.  What one dog loves another may not like at all.  Some dogs like food the most and others are more interested in toys.  Many things that are in the dog's world can be rewards.  If the dog wants it enough to work for it and if you can control access to it, then it is usable as a reward.

My dog Michael was stressed by strange places and wasn't so sure he wanted to do agility.  Every time I tried to get him to do something, he ran off to the edge of the field to pee on bushes.  I knew this was a sign of how worried he was and that he wasn't being disobedient, he was trying to relieve his stress.  I also knew that I couldn't fight him on it if I wanted him to get over it and to one day enjoy agility.  So I put him back on leash and got him to jump one very low jump and then I ran with him to the bushes so he could sniff and pee.  After he'd finished, we went and did two jumps and then ran to the bushes.  We repeated this sequence, sometimes harder and sometimes easier.  It only took a couple of sessions before he realized that this place was okay.  By respecting his needs and using what he wanted (and needed) to do as a reward, he came to love agility.  Learning agility made him more confident and before long he was going to all kinds of new places without any concerns at all.  Michael was halfway to his Masters Agility Championship before we had to retire him because he was too old to jump any more.

I heard of a woman who trained her terrier in agility by using access to nearby trees as rewards.  Her dog didn't want to pee on the trees, he wanted to run around them looking for "little buddies" (squirrels) in the branches.  Instead of fighting that huge distraction, she used it to great advantage.  A dog who can win access to something that distracts them, won't be distracted by it because they know they can have it if they work for it.  Dogs that learn how the training game works are willing to work for stuff.

You and your dog will have different ideas about what is reinforcing.  In this situation, the dog's ideas win.  If you observe your dog closely you will find out what they want.  How many things other than food and toys does your dog want?  Make a list and then be creative and find a way to use those wants to your advantage. 

Gimme here:  I learned on the first day I had my new person that she is very likely to give me what I want if I sit pretty first.  So if I want to go out the door, I sit.  If I want to get out of my crate, I sit.  If I want out of the car, I sit - then she puts my leash on and off we go.  If I want her to play tug with me, I bring a toy and sit.  If I want her to get off the computer and pay attention to me, I sit.  Sometimes you have to nudge them with your nose so they see you are sitting pretty.  Still, most people simply cannot resist a sitting puppy.  It may be most powerful thing you can teach your person.  I'm just saying.

And remember, things that are rewarding to a dog on one day or at one time, may be aversive or neutral at another time.  A dog that has just eaten may not work for treats.  A dog that is really tired, may not work for the chance to play tug with you.  But both of them may work for a belly rub.

Rewards are not as effective when a dog is under stress.  In fact, if a dog is really stressed or fearful, learning cannot happen.  Still you are likely to get more from a stressed dog with rewards than with punishment.  Punishment only increases the anxiety making it unlikely that she can perform or learn. 

Punishment 201

"Fallout" from using punishment

If not used perfectly, and sometimes even when perfect, punishment can have unwanted side affects.  These side affects can damage your relationship with your dog and may cause other behavior or emotional problems.  The Troublesome Thirteen fallout categories are:

Fallout Number One   Punishment differs from reward because it has no reliable effect on future behavior.  Punished behaviors aren't actually gone, they are only suppressed and can return at any time.  Punishment undermines confidence and suppresses all behavior, especially the dog's problem solving behavior.  This then is the reason guide dogs and bomb detection dogs aren't trained using punishment - they need to be problem solvers.  Punishment often suppresses a dog's natural warning cues, such as growling or posturing.  This is why behaviorists so often hear, "he just bit with no warning". 
Fallout Number Two   The one who is receiving the punishment is the one who "decides" what it is being punished for. Thus punishment may result in avoiding something you didn't intend (i.e. the dog still chases rabbits but now it won't go near peach trees) or an increase in the behavior if it was from fear (such as barking/growling at strangers or other dogs).

Fallout Number Three   The unwanted behavior may not happen when you are there and actually increase when you aren't there (i.e. when you can't enforce it). Punishment often creates smarter and more clever misbehavior. 

Gimme here:  If you punish your puppy for pottying in the house, you haven't taught them where they should potty, but you have taught them not to potty when you are around to see them.  Then that puppy may potty behind furniture or in another room and then you get to feel it squishy in between your toes. 

Even if you do get through the lesson about where-to-potty, you may discover that she doesn't want to potty when she is on leash because you are there and looking.  That can be a real nuisance when you are traveling, especially when its raining.  So, if you feel a burning need to hit someone with a rolled up newspaper for a house training accident, hit yourself.  I'm just saying.

Fallout Number Four   Poorly timed punishment breaks the dog's trust and creates harm to your  relationship by teaching the dog that you are sometimes bad to them for no reason.  If the dog hasn't learned the behavior you want and is punished during learning, the same harm to the relationship occurs.

Fallout Number Five   Even with perfectly timed punishment the trainer is associated with the punishment.  Dogs trained with punishment learn to do things to avoid the punishment... they also learn to avoid any situation that leads to punishment - they can learn to avoid you.

Fallout Number Six   When punishment is applied inconsistently and the unwanted behavior is self-rewarding, you create a situation of variable rewards, which has the effect of increasing the bad behavior.  Dogs trained by punishment learn that absence of punishment means they are doing right - "false positives" for a behavior being okay. Thus any time you cannot punish a bad behavior they are getting a reward that says that it is okay (on a variable schedule - the strongest of all). This is confusing, to say the least, and complicates the process of reducing bad behaviors.

Fallout Number Seven   When the behavior is "necessary" or self-rewarding, the punishment will only suppress it for a short while and then survival drives will cause the bad behavior to return, usually stronger than before.

Fallout Number Eight   When the trainer isn't ready with an good replacement behavior, the dog may create its own behavior to fill the void, which can be worse or no better than before.

Fallout Number Nine   Punishment is aggressive in nature and aggression begets aggression.  You may unintentionally teach your dog that it needs to defend itself - from you.  

A dog's physical response time is 5 times faster than a human response. Also, dogs have a jaw strength of 1500 pounds of pressure per square inch. So what we have is a dog who can respond to a situation very rapidly with the potential to do a lot of damage. Add to the mix the fact that he knows what he is going to do before you do. Why would you ever want to teach your dog that aggression is an good way to handle conflict between the two of you?

Fallout Number Ten   "Mild" punishment can gradually become tolerated until more severe punishment is necessary and needed more often to keep the behavior suppressed. In competition with self-rewarding behaviors, punishers eventually lose effectiveness unless extremely intense. And yet, punishers that are strong enough can even put a stop to survival-based behaviors.

Fallout Number Eleven   Punishment that only suppresses behavior without teaching a good replacement behavior can lead to obsessive behavior. In the case of punishment that suppresses behaviors innate to a specific breed, those drives become stronger and stronger until the reward for satisfaction overcomes the avoidance of the punishment.

Fallout Number Twelve   Punishment can become a conditioned positive reinforcer if your timing is imperfect or if you "show them you still love them" following punishment. Both cases will increase the frequency of the behavior you don't want.

Fallout Number Thirteen   Dogs trained with punishment may learn they only have to "behave" for someone who can punish them; i.e. to only respect someone who enforces rules with force. For those who can't enforce the rules they may be set up to be "at risk", at the very least they are set up to be ignored by the dog and unable to control it. Being "at risk" includes other family members, children, elderly people, neighbors and guests.

Punishment 102

When punishment "works"

It may "work" - if the punishment is strong enough and lasts long enough. Then the unwanted behavior may, for all practical purposes, disappear. You should remember that the affect of punishment is to suppress behavior, so you can never be sure if or when the behavior will come back.

It may "work" - if the punishment is used carefully to interrupt a chain of behavior allowing the trainer to teach and reward a better behavior, while never allowing the dog to "practice" the misbehavior. Of course, management without punishment will accomplish the same goal.

It may "work" - if the rules for using punishment are followed precisely.

Rule Number One  The dog identifies the correct behavior causing the punishment and there must be no chance that the dog will link the punishment to some other "cause". This requires exquisitely perfect timing and complete control of the environment.

Remember:   The same perfect timing will make you a superb reward based trainer.   The same complete control of the environment will make you a great reward based trainer.

Dogs are "associational" learners and they are highly visual in their perception.  So when punishment is delivered, the dog will tend to associate the punishment with whatever she is looking at or what she is near.  For example: if you yell at your dog every time she barks at people walking by your yard, she is likely to think those people walking by cause her to "get in trouble", so now she has even more reason to want to bark at them and make them go away.

A study of programs to teach people to stop smoking found, for those programs that used any form of punishment (even very mild ones) the most common reason people gave for leaving the program was "not liking their counselor".  One such program had the client put a rubber band around his or her wrist and snap it (causing mild discomfort) any time they saw or thought of a cigarette or smoking.  At every session, the counselor shows the clients a slide show (cigarettes, smokers, advertisements, ashtrays) and the clients practice snapping themselves.

Even though this was a very mild punisher, in addition to discouraging thoughts of smoking, it discouraged the clients from wanting to be near the person who was most often there when it happened, the counselor.  Your dog may learn to avoid you as well as the bad behavior.

Rule Number Two   The punishment must be something the dog will work to avoid every time.

You must know exactly which punishment will be the most effective at any given time and you must never be wrong.  Punishment based training is not very forgiving.

Rule Number Three   The punishment must be strong enough to interrupt or suppress the bad behavior.  It must be a great deal more punishing than the bad behavior is rewarding (most bad behavior is self-rewarding).  At the same time it must not shut down other behavior, much of which you may want now or later.

A punishment that is strong enough to do all this has a greater likelihood of being associated with more than the bad behavior you are aiming for and it is more likely to be associated with the person that delivers it, you.

Rule Number Four   The punishment must be consistently applied and perfectly timed EVERY time the misbehavior occurs.

Remember:   The same perfect timing will make you a superb reward based trainer.

Any time the dog does the bad behavior and isn't punished, the absence of punishment is the same as a reward for the bad behavior.  Also, that random non-punishment "reward" is on a variable schedule, the strongest kind of reward, which makes the bad behavior even stronger than if you'd never punished it in the first place.  See reward or reinforcement labels.

Gimme here:  Food on the counter is very tasty.  I only have to get it once to KNOW how great it is.  You have to be perfect at putting ALL food away EVERY time and you would have to punish your dog for even looking at food on the counter EVERY time she looked. 

Admit it, you don't even know what your human friends are looking at ALL the time.  Its much easier to teach your dog to look away from food on the counter and then reward her with even yummier food from inside the refrigerator.  I'm just saying.

Rule Number Five   A good behavior must be put in place of the bad one.

Nature abhors a vacuum.  You must have identified a new behavior to replace the bad one and you must teach it carefully and completely.  If you don't teach a new and better behavior, then the dog is likely to develop a new and equally bad replacement behavior.  It is much easier to teach the one good behavior you do want than it is to one-by-one punish away all the myriad of bad behaviors that might crop up by themselves.

See problem solving label.

Punishment 101

What is Punishment?
Punishment is anything your dog will work to avoid.  If a behavior is happening less often, then its being punished in some way, or extinction is in effect.  Extinction is what happens when a behavior doesn't get rewarded over a long period of time -- we'll talk more about that another time.

Punishment causes a suppression of all behavior, not just the one you want to get rid of.  In a way, suppression is like depression.  You might be depressed because your boyfriend dumped you, but you stop doing more than just boyfriend things, you stop doing much of anything.  A dog that doesn't do much of anything may seem to be better behaved, but he's probably not very happy.  Also, like depression, suppression doesn't last forever.

Punishment is done in two ways - adding a bad thing or taking away a good thing.  If you jerk or pop the leash when the dog pulls on the leash - that's adding a bad thing. 

Gimme here:  You don't have to be a doggie psychologist to know that its not any fun getting punished for being bad when you haven't learned how to be good.  I'm just saying...

If your dog pulls on the leash and you stop moving -- that's taking away a good thing.  In this case the good thing is getting closer to what she wants to sniff and when you stop you are taking away the chance to get closer.  If you wait until she loosens up on the leash and start moving forward again, you are rewarding her for keeping the leash loose.

The one who is receiving the punisher (in this case the dog) is the one who "decides" what is punishing and what is not and how much so.  What one dog hates another may like.  Squirting water at dogs is icky and yet most Labrador Retrievers love it.  Also, something a dog likes at one time, may be bad at another time.  A dog that normally loves people might find them scary when its dark outside or when they suddenly appear coming around a corner.  Sometimes that new scary impression can last a long time and affect the dog every time they see that thing.

Gimme here:  I loved horses when I was little; they were big and warm and they smelled nice.  Then when my person and I were walking in the woods, more than once the horses suddenly appeared on the road ahead of us.  I didn't know they were coming from side paths, so it seemed evil-bad to me.  For a long time after that I was afraid of them, until my person taught me that they really are okay. 

Primary punishment
"Primary" just means that its innate for the dog to avoid it.  It usually has to do with survival.  Such as: loud, harsh, or sudden sounds, a high-pitched-high-intensity stimulus, sudden loss of support, excessive heat or cold, anything that causes pain, some "bad" or "offensive" odors, and some tastes. Some dogs are especially fearful about things that aren't usually in this category.  For them, their extra sensitivity makes these "special" primary punishers and they won't respond normally to them.  For example, some dogs are especially noise sensitive and they often act fearful to very normal sounds.

Secondary punishment
"Secondary" means that in the beginning something is neutral, but then the dog learns to see it as a punisher.  Secondary punishers are taught to the dog by pairing it with other primary punishers.   Examples for dogs include: a word such as "NO!", an upraised hand, stomping your foot, certain postures, and grabbing toward the dog.   Humans have secondary punishers too, such as: frowns, insults, tone of voice, name calling, thumbs-down gestures, an "F" on homework, loosening the belt, and police lights and sirens.

Wednesday, September 28, 2011

Welcome to Our Training Advice Blog

Hi there,

My name is Gimme and I'm really tired of all the mean things I hear about that are done to doggies in the name of training.  I'm really lucky because my person doesn't do that stuff any more.  Many years ago she used to train in the old Koehler style and was proud of how hard she could jerk the leash.  But she didn't know any better and now she does. 

So we are writing this blog so people who want to teach their dogs to be good can have some place to go to see how to train without being mean.  You can train and still be your dog's best friend.  Be sure to check out the list of labels down the right side to help you find blog entries that are about the topic you are interested in.

We hope you will like our blog and come often to see what is new. 

Love, Gimme