dogtrainingisfun
Registered: 04/08/08
Posts: 15
|
|
|
Reply with quote | #1 |
Note: This was written by a fellow trainer friend of mine.
This is the first entry in a series which will be dedicated to the issue of proofing behaviors. I hope that it will help my friends on their training journey working with their canine family member(s).
I plan on cross-posting in my dogster diary, on my website http://www.rewardingbehaviors.com and on Karen Pryor's site (view my previous blog entry It's not Click/Treat, it's Click THEN Treat! Mechanical Skills for Clicker Trainers). You may cross post, but please give credit to the author (Casey Lomonaco) and link back to her website (www.rewardingbehaviors.com) if you chose to do so.
What does proofing behaviors mean, anyway?
If I had a nickle for every time I heard, "but my dog knows how to sit at home!" I might become the world's wealthiest dog trainer. My response is generally, "if your dog knows how to sit, why isn't he doing it?"
It's a rhetorical question. I know exactly why, and you need to also if you want reliable behavioral responses to cues. (Because I'm a nice dog trainer, I'll give you a hint...your dog is not trying to be dominant, assert his authority, he is not "spiteful" or out to embarrass you, and he certainly doesn't want to cause you frustration or grief of any sort.)
So what is the answer? It revolves around two key terms...generalizing and proofing.
If you frequent any dog training communities or clubs, you'll often hear the statement, "dogs don't generalize well." What does this mean?
It means that "sit" in your kitchen doesn't equate with sitting on the sidewalk, in class, when your kids are running around screaming, or your Aunt Ida unexpectedly drops by for a visit.
"Generalizing" in behaviorspeak/jargon means, "the ability to respond to a discriminative stimuli (cue) regardless of environmental influences." (I'm sure there are better or more technical definitions, but for our purposes, this should do just fine.)
When I first attended clicker classes with my chow mix Mokie, my instructor (and now business partner, Abbie Tamber) really brought home the concept of what generalizing means. I was the student who said, "but my dogs knows how to *insert behavior here*..." and she said, 'Has she done it 5,000 times?''
She hadn't, and I said so. 5,000 times?! This woman must be insane. Abbie told me, "then she doesn't 'know' it!" I must admit, I was a bit disgruntled.
5,000 times? Seriously?!
Seriously. Some service dog organizations will cue specific behavior thousands of times (as many as 8,000 times) before they consider a dog sufficiently "proofed," at which time they will have enough confidence to assert "this dog knows the behavior."
The number of repititions is not set in stone, and is in fact somewhat arbitrary. Once your dog is able to generalize a few behaviors through these proofing criteria, you'll find that all subsequently taught behaviors tend to generalize more rapidly. In essence, your dog is learning to learn!
What is not arbitrary is the fact that for a behavior to be learned, it has to meet a number of criteria.
What are these criteria?
In short, they are:
* Distance
* Distraction
* Duration
* Precision
* Latency
* Speed
and in my book, Stimulus Control rounds out the septet of critical factors which will influence your dog's ability to fluently respond to your cues in any environment.
In the series, you can expect a separate entry on each of the proofing criteria. For this introduction, I will provide a quick definition of each. In the later entries, expect more detailed information on how to proof for these aspects of fluency.
Distance: Just because your dog can respond to a cue directly in front of you does not mean he will "generalize" that the cue is still valid when he is ten, fifty, or two hundred yards away from you. If you want fluent responses at a distance, you must teach your dog to do so.
Distractions: While your dog may recall to you in your kitchen, she may not recall to you if she is off leash and spots a squirrel, deer, other dog, or even a leaf blowing in the wind. If you want your dog to respond to your cues in the middle of a construction zone, the dog park, or the pet store, you must proof for distractions!
Duration: Does your dog respond to a "down" cue and then pop right back up to a standing or sitting position? If you want your dog to offer an extended down until released, you must build duration for the behavior.
Precision: What is your vision of the ideal behavior? Proofing behaviors for precision is a fairly advanced process. Are you participating in competition obedience and getting crooked sits? To get that straight sit that you are seeking, you must concentrate on proofing for precision.
Latency: Have you ever cued a dog for a behavior and then waited....and waited...and waited for a behavioral response? You say "sit" and what seems like ages later, your dog's bum hits the floor? Latency is the time lag between the cue delivery and the offering/initiation of a behavioral response from the dog. If you want your dog to sit as soon as you give the cue, you need to proof for latency!
Speed: The criteria of speed in relation to behavior is signified by the time lapse between when the animal starts the behavior and when the animal completes the behavior. Sometimes a student recalls a dog, and you see a dog walking back to them. If we need to pick up that speed, we must proof for it!
Stimulus Control: according to Karen Pryor, there are four fundamental aspects of stimulus control. They are as follows:
1) the dog offers the behavior in response to the cue
2) the dog does not offer some other behavior in response to the cue
3) the dog does not offer the behavior in the absence of the cue
4) the dog does not offer the behavior in response to another cue
I am hoping that this series of entries will help you all through the critical stages of proofing so that you know how to train any behavior your dog is performing to reliability in the environments and situations you and your canine will encounter.
If it all sounds impossible, relax. Not only is it possible, it's probable and even better...it's fun! Until the next entry in the series, happy clicking to you and your canines!
|
| | dogtrainingisfun
Registered: 04/08/08
Posts: 15
|
|
|
Reply with quote | #2 |
SECOND INSTALLMENT: PROOFING FOR DISTANCE
Again, feel free to cross post. If you choose to do so, please credit the author (Casey Lomonaco, KPA CTP http://www.rewardingbehaviors.com).
Now that I've introduced what exactly proofing behaviors is, it's time to break it down into the particulars...elements of fluency.
Many dog trainers speak about the "3 D's". What are the 3 D's, you might ask? Distance, Distraction, and Duration.
It is not imperative that you begin proofing for distance before distractions or duration. It is also not as if you have to proof for distance exclusively until you reach your goal distance, and then go back to working on distractions, then duration.
I only work on one aspect of fluency during any given training session, but may work on another aspect later on that day in another training session.
Let's consider distance further, and how we can work towards proofing our behaviors for this aspect of fluency.
One of my favorite methods, which works for establishing both distance and duration is called "300 pecks".
What is "300 pecks"? It is a method that originally stemmed from behavioral experiments with pigeons in the laboratory, and the term for the method was eventually coined by Alexandra Kurland, a well known clicker trainer who specializes in horses.
"300 pecks" means that we raise criteria in small increments and immediately go back to basics (baseline level of the behavior) if the dog is not able to succeed at the current level of performance.
How can you use "300 pecks" to build distance?
Let's say we're working on sit, and we'd like our dog to sit at a distance, and we'd like to work up to a response at a distance of 100 yards, eventually (this number is arbitrary, you may proof for distance as far away from you as your dog is able to perceive the cue).
While this is our eventual goal, if we have just started training the behavior, it is not likely that we will be setting our dog up for success by expecting that level of performance at this stage in the training.
Most dogs that have just been trained the sit behavior respond directly in front of their trainer. To build duration on the sit behavior using "300 pecks" you would proceed as follows:
1) Cue the sit immediately in front of you. If your dog responds to the cue, click then treat (hereafter referred to as CT).
2) Take one step back, cue the sit. If your dog sits one step away, CT.
3) Take two steps back, cue the sit. If your dog sits two steps away, CT.
4) Take three steps back, cue the sit. If your dog responds with the behavior, CT.
Let's say that your dog didn't respond at three steps but responded at two. If you cue the sit three steps away and your dog tries to approach you to sit in front of you, offers another behavior, or otherwise does not respond to the cue appropriately, do not reinforce.
Automatically go back to the first step - cueing directly in front of you. Then work back up to three steps, four steps, five steps, a hundred steps, etc. If at any time the dog fails, you go back to the first step and work through the incremental raising of criteria all over again.
I know that it sounds slow and tedious, but it's really not. You'll be surprised at how quickly you are able to progress in your training using the "300 pecks" method.
In some circumstances, even that first step back is too big for the dog. If you take a step back/away from your dog and she automatically moves to being directly in front of you, this is one of those occasions.
For these dogs, you have two options:
a) use a tether/gate/other physical barrier: secure the dog in place and proceed with the 300 pecks method. If your dog is skilled at targeting, you can also use a stationary target and send your dog out to target to create distance. HINT: If you would like to use a target, make sure you proof that behavior for distance first!
b) proof for the distraction of handler movement for a bit and then come back to distance work. (More about proofing for distractions - including handler movement - to come in later installments of the series).
If you are working on a moving behavior like a recall, you may need the help of a volunteer who will hold the end of the lead/tether and release it when you give the cue.
Another behavior that many people want to proof for distance is loose leash walking (LLW), or LLW's show-off cousin, the heel. You can use 300 pecks to train this also...one step, reinforce. Two steps, reinforce. Three steps, reinforce. Four steps, reinforce. Five steps - leash goes taught. Return to baseline (one step) and resume the procedure.
It is important to remember that we only raise one criteria at a time, and while we raise any one criteria, we temporarily lower our other criteria. We do not want to introduce distractions when we are first training for distance. So you're going to work on distance first in an environment with very minimal distractions. (LLW is a bit of an anomaly in this respect, as the nature of this behavior is that duration and distance are built simultaneously.)
If you choose to work in another session on distractions, make sure that you aren't asking for much distance while you train for distractions. Once both aspects of fluency are achieved, you can introduce them together (begin building distance in a distracting environment). It is very important that you set a clear standard of achievable goals for the dog within any training session and that your dog is set up for success.
To put this into perspective, let's say that you are learning the art of belly dancing, and you are also learning how to perform complex algebraic equations in your head. Each of these are new tasks to you, and require a fair bit of concentration as you develop this foreign skill. Would it be easier to learn about belly dancing one class and algebra in another class period, or would you like to learn belly dancing and complex mental calculations simultaneously?
Once you are skilled at both belly dancing and mental mathematical feats of genius, you may be able to belly dance while thinking of a simple equation but not a very complex equation. If you want to be able to do both fluently simultaneously, you will need to work on them each separately, and then work on combining them in a way that will allow you to achieve good results in both.
Hopefully, this admittedly silly example will help illustrate the importance of proofing for one aspect of fluency at a time, until there comes a time when they are all individually fluent. Then, and only then, should we begin introducing multiple criteria at a time.
All that said, do your dog a favor. Teach them how you would want to learn, and allow them separate training opportunities/sessions for each new criteria/aspect of fluency.
Until our next installment, happy proofing to you and "woof! woof! *play bow*" to your dog(s)! (They'll know what that means, trust me.) |
| | dogtrainingisfun
Registered: 04/08/08
Posts: 15
|
|
|
Reply with quote | #3 |
DISTRACTION TRAINING...ARE YOU RELEVANT?
Welcome to the third entry in our proofing series. Feel free to cross post, but please give credit to the author - Casey Lomonaco, KPA CTP, APDT, http://www.rewardingbehaviors.com
At this point in the series, you've received a brief explanation of proofing and its components, along with a more in-depth look at training for distance responses to behavioral cues.
Moving on to the second of the "3 D's," it's time to start proofing for distractions
Distractions are one of the aspects of fluency students tend to struggle with the most. After all, it's difficult to compete with the environment when it seems as though there is a squirrel around every corner, an invisible distraction (to you) in the form of a scintillating scent trail, kids on bikes or rollerblades, construction equipment/activity, you name it!
So how do we even begin to gain relevance in our dog's world...a world of heightened sensual experience we could never even imagine?
Systematically, of course. We have to break these criteria down into manageable and achievable steps, just like we did with distance. How do we confront this daunting task?
I like to use a distraction and reinforcement hierarchy. Borrowing from Tamar Gellar, I really like the idea of bronze, silver, gold, and platinum level reinforcers and have taken it a step further by extending these divisions to distractions. As I see it, reinforcers and distractions must go hand in hand for effective training, and using hierarchies makes that happen.
I'm sure you're wondering what the heck I'm talking about, so let's delve into this matter further, shall we?
The success of a positive reinforcement training program directly correlates with the handler or trainer's ability to recognize what is reinforcing to the dog. Distraction and reinforcement hierarchies will vary widely between dogs, as we must remember that the individual learner determines what is reinforcing (and distracting) to him.
My favorite food may be different from yours. To each his own, right?
In any case, the success of distraction training relies on the careful selection of criteria and reinforcers. Differential reinforcement is an important concept to understand and is at the crux of distraction training success.
Differential reinforcement means that exceptional efforts should be reinforced exceptionally. It means that in the face of higher distractions, you will need to draw on higher value reinforcers. Many of you may be familiar with the concept of "jackpotting" which is a prime example of differential reinforcement.
An example which may be relevant to you is overtime or holiday pay. If I am your boss and expect you to work regular work hours, you make regular pay. If I expect you to work overtime, weekends, or holidays, you'll probably want more money (more reinforcement). If I paid you by commission, you would make more money for better efforts/sales. This is a human application of the theory of differential reinforcement.
In order to differentially reinforce behaviors, it helps to know how valuable each of your reinforcers are. This is where a reinforcement hierarchy comes in handy.
There is some debate as to the usefulness of secondary reinforcers as opposed to primary reinforcers. What is the difference, you might ask?
Primary reinforcers are necessary for the advancement/survival of the individual/species. For dogs, primary reinforcers include: food, water, air, sex, and participation in the prey sequence. Without any of these factors, the species would not be able to continue filling its ecological niche.
A secondary reinforcer is a reinforcer that has been conditioned (hence the term "conditioned" reinforcer) - there is nothing in the dog's biological composition that compels the need for this reinforcer, but something in the dog's experience/environment has created a positive emotional response to the stimuli in question.
In dog training, a good example of a conditioned reinforcer is the clicker. It is reinforcing to the dog because we have paired it systematically with access to a primary reinforcer (food).
In the human world, money is the most widely recognized secondary reinforcer. Pieces of green paper with pictures of presidents and numbers are not intrinsically reinforcing to people. Money is reinforcing because throughout our experience with money, it has been systematically paired with access to primary reinforcers (shelter, food, etc.).
A bit off topic, but this is also an illustration of why dogs work for the click as opposed to working for the food, much like we humans work for a paycheck as opposed to a few gallons of gas and a bag of groceries at our desk each Thursday morning.
Since the individual dog determines what is reinforcing to him, and most people believe they are supposed to reinforce with a primary reinforcer, it is not uncommon to hear students saying that their dog finds fetch or tug more reinforcing than a hot dog, and, "is that ok?"
I dispute the belief that fetch and tug are secondary reinforcers. I believe that each represents a part of the prey sequence which is: orient, eye, stalk, chase (fetch), grab/bite (tug), kill/bite (tug - the bite and shake!), dissect. Therefore, I do believe it is quite common to see these reinforcers rather high on the hierarchy for many dogs.
I also dispute the notion that secondary reinforcers do not have their place in a reinforcement hierarchy. My agility instructor/colleague's Jack Russell loves nothing more than the opportunity to jump excitedly into his handler's arms. The cue for a behavior taught with positive reinforcement can also function as a reinforcer.
To create a reinforcement hierarchy, think of as many things as possible which are reinforcing to your dog and break them into categories (Challenge yourself! How many reinforcers can you think of?).
A typical reinforcement hierarchy may appear as follows (I'll be keeping this relatively brief for the sake of the article, but you get the picture):
BRONZE:
kibble
opportunity to go through doorway
opportunity to get into/out of car
cheerios
having leash put on
physical contact (pats)
praise
SILVER:
natural balance/red barn food roll
Wellness Pure Rewards
opportunity to smell interesting scents on ground
cue for an established behavior
fetch
hot dogs
GOLD:
liverwurst
peanut butter
canned dog food
meat balls
opportunity to swim
PLATINUM:
opportunity to greet other dogs
opportunity to greet approaching human
opportunity to chase squirrels
mashed potatoes and gravy
steak
raw liver/hamburger
tug
deer poop
flirt pole
I encourage you to create a much more comprehensive hierarchy, aiming for thirty, forty, fifty or more reinforcers.
Now that we have our reinforcement hierarchy, it's time to begin working on our distraction hierarchy. You're going to see a lot of parallels between the two hierarchies, I'm guessing!
We'll use the same method to create the distraction hierarchy that we used to create our reinforcement hierarchy: bronze, silver, gold, platinum. You will most likely notice that many of your distractions for each level correlate with your reinforcers for each level. For instance, your distraction hierarchy may look as follows (again, a simplified version):
BRONZE:
leaves blowing in the wind
kibble on the ground
open doorway
open car door
handler movement
SILVER:
interesting scents on the ground
cheerios on the ground
traffic
other dog fifty feet away
tennis ball
GOLD:
swimming area
kids on bikes/skateboards
construction
car horns beeping
kitties
kitty litter box
PLATINUM:
other dogs
other people
squirrels
deer
tug toy
deer poop
remote control car
Completing these hierarchies is valuable for a number of reasons. First, it allows you to really think critically about your dog and how he views the world...what is important to him? Second, it provides you with a framework for distraction proofing.
If you are working on a silver distraction level, you'd better not be pulling out your bronze treats. At the silver distraction level, "to par" efforts should receive silver reinforcers, above criteria responses should receive gold level reinforcers, and exceptional, I-can't-believe-you-did-so-well efforts deserve platinum reinforcements.
Using bronze level reinforcers in the face of a platinum level distraction will certainly leave both you and your dog feeling frustrated.
Looking at the provided examples, can you see the parallels?
For example, note that opportunity to chase squirrels is a platinum level reinforcer and the sight of squirrels is a platinum level distraction. When you have worked up to platinum distractions, if you cue a behavior when your dog spots a squirrel and the dog responds, what better reinforcement than providing an opportunity to chase (once the squirrel is safe, of course) the object of your dog's interest?
Sometimes this works, sometimes not. Like my trainer/partner/KPA faculty member says, "if I could stomach carrying deer poop around for reinforcers, I could train my dog to do anything!" Since deer poop is not a practical platinum level reinforcer, feel free to substitute one of the others on the list.
Using the squirrel chase to reinforce focus around the squirrel is an application of the Premack principle, which states that a creature is likely to perform a less favored behavior for the opportunity to engage in a more favored behavior (i.e., if you eat your spinach, you can have this ice cream!).
Start out at the bronze level distractions of your hierarchy. If handler movement is on that list for your dog, begin by cueing the behavior while you are moving slightly (perhaps lifting your left leg to knee level and putting it back down). Then work up to larger movements, jumping jacks, running in place, running back and forth, waving your arms around, etc.
If your dog is unable to respond to cues at a particular level when you are training, evaluate the situation critically...what distractions are you up against? Is your reinforcer valuable enough, or do you need to go up the hierarchy or lower your criteria for distraction? Can you create distance between your dog and the distraction (lowering the distance criteria while raising the distraction criteria)?
I strongly encourage keeping notes of training sessions. A training journal will allow you to spot trends, peaks and valleys in your training journey and help you improve your skills as a trainer.
The bulk of proofing behaviors is generally in the distraction training, which is why this entry is so long. It's tricky because there are an infinite number of distractions and reinforcers which could comprise your hierarchies, and in all likelihood, every dog you ever train will have a totally different list for each.
Prey distractions are one that many people struggle with because they are generally outside of the handler's control - you can generally neither manipulate their appearance or at what distance they appear. That said, remote control cars, other dogs, or other fast moving objects can often stimulate the same chase response that prey animals do, and can function as good distractions which you can manipulate for distance and movement.
I've just discovered there is a deer farm in our area, and you can bet I will be contacting the owners to see if I can take advantage of this as a great opportunity to proof behaviors around predatory distractions!
Although this all sounds overwhelming, really concentrating on creating thorough hierarchies for reinforcers and distractions makes distraction proofing a cinch. These two lists will basically guide you through all you need to know to get reliable behaviors in the face of any distraction you may encounter.
Happy proofing, until next time! |
| | dogtrainingisfun
Registered: 04/08/08
Posts: 15
|
|
|
Reply with quote | #4 |
Without further ado, here is the next entry on proofing for duration. Again, free to cross post with credit (Casey Lomonaco, KPA CTP http://www.rewardingbehaviors.com):
Going the "extra mile": PROOFING FOR DURATION
In the fourth installment of the proofing series, we will discuss duration, the last of the "3 D's". Much of this will be a review of concepts already discussed, but there is some new material to be covered.
The 300 pecks method discussed in the distance article can be applied to building duration also, through withholding the reinforcement. To use 300 pecks to build duration for the sit behavior, I would:
1) Cue the sit. CT (click then treat) when the dog sits.
2) Toss the treat away so that the dog has to break the sit to retrieve the treat.
3) Recue the sit. When the dog sits, count to one, CT. Again, we're going to toss the treat away so that the dog has to get up to retrieve the reinforcement.
4) Cue the sit once again. When the dog sits, count to two, CT. Toss treat to reset, providing the dog with another opportunity to earn reinforcement.
5.) Repeat the sequence, counting to three when the dog sits, and then reinforce.
If at any time you raise the criteria an increment and your dog breaks the sit, return to step one and begin the process again.
Now that you are familiarized with "300 pecks," a question may arise...if I work up to a count of sixty and the dog breaks position, do I have to go all the way back to one in my count? Aidan Bindoff does a nice job of answering this common question in his blog entry on Karen Pryor's site, you can read it for free at: http://www.clickertraining.com
Remember, the click ends the behavior. Gale Pryor has a wonderful article with quotes from Karen Pryor on this particular topic. You may read the full text of the article at: http://www.clickertraining.com
You can also use treat placement to build duration for behaviors by feeding for position without clicking. If your dog is sitting, reinforce on a variable schedule. Let's say your dog is able to maintain the behavior for an average of two counts.
Cue the sit, and when your dog responds, withhold the reinforcement for a count of one, reinforce. Count to three, reinforce. Count to five, reinforce. Count to two, reinforce. Count to one, reinforce. Count to four, click and toss the treat away to reset your dog for another opportunity to be cued for the sit, repeat the process.
This is what is known as a variable interval (VI) reinforcement schedule, as the time interval at which the reinforcement will be delivered varies between trials.
If the dog is able to hold a sit for approximately five seconds, we may reinforce at three seconds, at seven seconds, at one second, at eight seconds, at two seconds, at six seconds. We're averaging approximately five seconds over the course of a designated number of trials.
Work up to higher averages of time your dog is able to maintain the position at a rate that sets your dog up for success. Morgan Spector has called this method "ping ponging" the criteria because the criteria bounces, with highs and lows that are all relative to the baseline time the dog is able to achieve reliably.
Ping ponging the criteria helps keep the game interesting for the dog. If the game always gets harder, the dog may be tempted to give up. Occasionally reinforcing below baseline efforts keeps the dog's confidence level high.
Putting the behavior on a variable reinforcement schedule is often explained by using a slot machine analogy (although a slots player is technically operating on a variable ratio which is based on the number of trials as opposed to a variable interval which is based on the length of time, for the purposes of our discussion, the analogy still functions well).
People do not play slots because they win after every five pulls of the lever. You never know how many lever pulls it will take, but you know that if you try often enough, chances are that it might really pay off!
For anyone who may doubt the power of a variable reinforcement schedule to intensify behaviors, visit a Gambler's Anonymous meeting. That chance at "the big win" can be very addicting (and for those who suffer with a gambling addiction, the reinforcement schedule can become powerful enough to be socially, financially, and emotionally detrimental).
Even when you have worked up to durations measured in minutes rather than seconds, every once in a while you should reinforce a five second stay. After all, sometimes that first quarter pays off with a big win!
Click for you! You've now made it through "the 3 D's" of proofing.
In the next entry, we'll be moving on to proofing for precision and polishing up sloppy behaviors through shaping. I'm sure that until then, you've got plenty to keep you busy working on the 3 D's! |
| |
|
