| Author | Messages | |
agilitynut
Champion
Posts:56
 | | 05/04/2008 11:25 AM |
| Hi all, just thought I'd write up a little intro to clicker training and include some resources for anyone who's interested. Clicker training is the use of a marker signal (a short word or sound, an actual clicker, a whistle, a snap of the fingers, etc) to mark the behavior that you like, and is then followed up by a reinforcer of some kind, usually a food treat (just a small taste), but can also be a quick game of tug or some other game, or occasionally some kind of life reward (such as permission to go through the door, though I usually reserve life rewards for learned behaviors where I no longer use the clicker.) So maybe that answers the first question I often get - do I have to keep clickers and treats in my pockets forever? Nope. For well established behaviors, a quick "good girl" or an appropriate life reward is all that's required. Asking for a sit while putting on the leash reinforces the sit behavior, while putting on the leash on a puppy acting like a whirling dervish reinforces acting like a whirling dervish. Making yourself aware of all the times that you are reinforcing your dog (and what for!) makes you a much better trainer! One misconception is that clicker training is only good for "tricks" and special behaviors like what we do in agility. Nothing could be further from the truth! Yes, clicker training is used to teach dolphins and killer whales and walruses etc do their fabulous routines in the marine mammal shows we like to watch- but did you know it is also used to teach those same animals to pee in a cup, have their teeth brushed (do you really want to brush a killer whale's teeth- they do!), allow people to draw their blood and put tubes down their throats, and other much more mundane husbandry behaviors that could be very dangerous for the animal and the handler if not trained well? It has also been used to train animals such as chickens, cats, and squirrels to perform for TV and even live. And US Navy live-caught dolphins (caught as adults) were taught to do open-ocean work where they were out of sight and control of their handlers for hours- and return despite having the full resources of the ocean (fresh fish!) available to them. And a NASA chimp was taught to perform its behaviors using clicker training- that HAD to be reliable! Cats were even taught to trail a target and used by the CIA (hey, I don't have to like the fact that they surgically implanted them with microphones!) to spy on people in Moscow. I kid you not! So if you're interested in learning more, check out these resources: www.clickersolutions.com (and it's yahoogroups list) www.clickertraining.com www.shirleychong.com youtube has some good video of clickertraining of a variety of species- just put clicker training in the search. There are also some videos on the clickertraining.com site. Best of luck and feel free to post questions! Agilitynut, Brodie, Kyp!, and Arie | | | |
| DPFrank
Best of Breed
Posts:327
| | 05/04/2008 3:21 PM |
| a Marker training is a must for teaching behaviours , i agree. Tell me , when a dog knows a behavior, hands down, but does not do it the first time you tell him(her) what is the concequences? Frank | | | |
| agilitynut
Champion
Posts:56
| | 05/05/2008 8:50 AM |
| Thanks for asking!
Consequences of behavior fall into several categories, often referred to as quadrants. They are broken down into consequences where something is applied or given (+) or something is removed or taken away (-) and something that increases the likelihood of a behavior (reinforcement) or decreases the likelihood of a behavior (punishment). In general, clicker trainers prefer to think whenever possible in terms of behaviors they do want, rather than behaviors they don't. For example, jumping up on people is a behavior, and so is sitting. Rather than punishing jumping up, we teach (and reinforce) the dog to sit for attention.
+R (positive reinforcement)= the application of something such that the dog is more likely to do the behavior (something like a treat or play)
+P (positive punishment)= the application of something such that the dog is less likely to do the behavior (something like a shock or leash pop)
-R (negative reinforcement)= the removal of something such that the behavior increases (this is the tough one for many to get- removal of something aversive to increase the desired behavior) An example is an ear pinch which stops when the dog opens his mouth to take the object, or a shock which stays on until the dog starts coming.
-P (negative punishment)= the removal of something to decrease a behavior (for instance, leaving the room when the dog jumps on you or removing the opportunity to continue playing agility when the dog breaks a start line stay)
For the most part, I stick with +R and -P. I do use a No Reward Marker (NRM) with my older dog, but not with my puppy. I don't consider that she knows ANY behavior "hands down" at this point, even though she is very good in many settings. My definition of knowing a behavior "hands down" (what I refer to as fluency) is that the dog does the behavior the first time asked in a wide variety of settings, with a virtually zero latency (immediately). If a dog doesn't do that, it raises a number of questions in my brain:
Is something physically wrong? A 10 year old dog who stops sitting immediately when he has always done so deserves a vet check and possibly some pain meds or anti-inflammatories.
Have I really trained the behavior I'm asking for in an environment as complex as this one? Agility trials are full of sights and sounds that a dog may never have heard before. So when I took my pup to a trial, I didn't get the quick responses that I was used to. My response was to work her first at a distance, then bring her closer to the ring.
Is there something in the environment my dog is reacting to? A dog who has never seen a horse or a llama before may stop and stare and be completely oblivious to you even if otherwise fluent in a given behavior. This is especially true of very visually stimulated breeds such as sighthounds and border collies.
As for your actual question- it depends  I try VERY hard not to put a dog or pup in a situation it isn't prepared for, to avoid having the situation arise in the first place. It does take being very pro-active to set the dog up for success. I take my older dogs offleash in the woods, and only very occasionally reinforce their recalls with a treat. My pup doesn't yet have that privilege, and will be on a much higher reinforcement schedule when we start (and hey, I'll give the older dogs some extras too- can't hurt!)
In many cases, I simply get in the dog's space and wait for him to respond. Occasionally, I'll repeat the "command" (I refer to it as a cue.) Sometimes a privilege is lost or the dog gets put away instead of being able to play. A dog or pup who doesn't come when called gets put back on leash. A dog who spaces out and doesn't respond quickly gets put away and loses the opportunity to earn treats and tugs. And often has to watch as another dog does! A dog who doesn't target at the end of the plank when asked loses the opportunity to play on the plank (and get treats for correct execution!)
Think about the terms "command" and "cue" and how they differ in their inferences. A command generally has the inference of "or else something you don't like will happen." A cue, OTOH, has an inference that now you have an opportunity do something and you may be reinforced for it. In many cases, the behavior has been strongly reinforced, and has become a secondary reinforcer itself. The dog loves doing the behavior- wow, how great is that? Now we can use that behavior to reinforce other behaviors! In agility, we often use a cue for the next obstacle as a reinforcer for the current behavior. This is especially evident on contacts, so the dog performs the correct contact behavior, and is then either given a cue for the next obstacle or a release, which is also a cue to leave the contact. If the correct contact behavior is not performed immediately, the dog realizes the lack of the cue, and then performs it, or the dog is removed from the contact.
Okay, long enough answer!
Agilitynut, Brodie, Kyp!, and Arie
| | | |
| DPFrank
Best of Breed
Posts:327
| | 05/05/2008 9:01 AM |
| A
My appologies, I actually wanted to know what you do. Ive been training dogs a long time and wanted another clicker trainers opinion when a dog does not comply the first time we ask. ( provided they have clearity on the command)I'm familiar with all the book stuff, I was hoping for your laymans technique. Remember, the dogs I train are not always starting from pups, I train for protection as well as obedience. I am as positive as possible when teaching, but where I seperate is when a dog does not comply.I never repeat. thanks,
Frank | | | |
| agilitynut
Champion
Posts:56
| | 05/07/2008 8:39 AM |
| So what do you do when the dog doesn't comply? The other things I mentioned aside from repeating ( mentioned several)? An aversive of some kind? Is it always the same one? What affects your choice? What is your definition of when the dog has clarity (or fluency) of a behavior? Do you want to punish a dog that may be communicating pain or extreme distress (after all, if this is a fluent behavior, then by definition the dog performs it quickly and in a wide variety of settings and amid distractions, so there may be some quite legitimate reason that the dog doesn't perform. Many people are quicker to assume that their dog "knows" a behavior than I am.
Regardless of the venue, regardless of the age of the animal, you can get reliability without resorting to positive punishment. I'm not saying its not an option- plenty of people use it effectively. And some situations make it more or less justifiable if you're not willing to put the work into using alternatives. I know trainers who will put a shock collar on a dog at his second agility class (I kid you not!) But it is one that the trainers who trained the behaviors I mentioned in my first post on this thread deemed not necessary, and possibly damaging. That's true of adult caught dolphins, cats, birds (who also performed spy tasks, really cool stuff!), chickens- and even (gasp!) dogs.
agilitynut, Brodie, Kyp!, and Arie | | | |
| agilitynut
Champion
Posts:56
| | 05/07/2008 9:46 AM |
| Ttraining a chicken...
People give me odd stares when I mention that I spent two weeks at "Chicken Camp" a few years ago, and would like to go back for more. Bob Bailey is a phenomenal instructor, with an incredible depth and breadth of experience. His wife Marian (who passed away a few years ago) and her first husband (Keller Breland who passed away quite a while ago) were graduate students of Skinner's- yep, the same Skinner you read about in Psych 101. They decided to start a company training animals using Skinner's techniques, and over the years, the company trained over 140 different species, and over 10,000 individual animals. These were used in the military and for spying, as well as for TV commercials, movies, and a variety of uses. Ever seen a chicken play tic-tac-toe? Or do a mini agility course? Although they used chickens in some of their enterprises, they also found that chickens were good for training new trainers.
Chickens move quickly but are large enough to watch, without being so large that they are unwieldy. They also have good appetites! And they have visual acuity similar to a human's vision. So my first task was to train the chicken to peck the middle of a circle target that was about 1.5 inch diameter. Not too hard- but really honed my observation skills to get the middle and not the edge! Although clicker training is rather forgiving when considering a general behavior, it is not so forgiving when trying to get something very specific. You don't necessarily get the behavior you want- you get the behavior you click- even if you clicked pecking the edge. Guess what I did? Right- I clicked the edge! Then I had to extinguish that behavior- NOT click it until it went away. One thing that often happens when working on extinguishing a behavior by not reinforcing it, is that you get an extinction burst- this is when the animal throws a bit of a tantrum and does the behavior (that you don't want) even more forcefully (hey, this worked, I'm doing it, pay attention and click me- alright a bit anthropomorphic, but hey, its what I see!)- and guess what happens if you click then? REALLY hard to get rid of the behavior! Fortunately, I was on the lookout for it, didn't click when it happens, and it extinguished- phew!
So now that I had the chicken pecking the middle of the circle, I had to have the chicken click ONLY the middle of the CIRCLE- not a square, or triangle, or even an octagon. Believe it or not, this part actually went smoothly. Sure the chicken pecked the other shapes a few times, but this was easier to observe than pecking the edge vs the middle, since at this stage I only had one shape in front of the chicken. Then two shapes, the circle and one of the others. Then the circle and a different other shape. And then, finally, three shapes, then four.
And then I had to reverse it- make my "hot" target (the circle) "cold" and choose another shape as "hot" and go through the same process!
Now that we had trained a behavior that was very discrete- the chicken either pecked or he didn't, there wasn't a lot of variability in the behavior- we got to work on training the chicken to pull a rubber band, and to keep pulling further and further. We used large loose rubber bands, and apparently this behavior is easy to teach because chickens are disposed to peck at long squiggly things that look like worms
Then we got to teach the chickens to dance. They didn't look like Fred and Ginger, but they did twirl either left or right. This actually involved shaping- getting the behavior by clicking for successive approximations. So we started with clicking for scraping their feet (an innate behavior that is actually hard to get rid of, as it is related to food gathering) and then clicked for slight variations that eventually ended up as a twirling in place behavior. We didn't use a target stick or lure of any kind to get the behavior- just the three Bailey Basics(timing, criteria, and rate of reinforcement, see below) and chicken feed. And then we got to add a cue! But not until we had first added a cue to just pecking behavior (presence of a red light)- and that was hard. Not hard to get the chicken to peck on cue- but very hard to get the chicken NOT to peck when the cue wasn't presented! Anyone know any dogs like that? So we presented a little baton, and the chickens were supposed to continue dancing until we lifted it.
Another thing we did was teach the chickens to go in either an oval or figure eight pattern around two cones on a table. This is where it really became evident that I was using my body to provide an additional cue! Teaching my body to be quiet was hard, and I'm still somewhat sloppy about this.
I learned a lot from Bob Bailey, and I loved hearing his stories of what they taught the animals- and what the animals taught them. One little kernel that has been very helpful is that most training problems come down to an issue with one of the Bailey Basics: Timing, Criteria, or Rate of Reinforcement. Timing refers to when you click- it is very easy to be late or just imprecise. Criteria refers to the specific piece you want to click- hopefully a small increment. We talk in terms of lumping (jumbling large blocks of criteria together, making it harder for the animal to figure out exactly what you are looking for, as well as harder to achieve it. the better alternative is to "split" each piece into extremely small slices. And finally, the rate of reinforcement simply refers to how often the animal is getting reinforced for behaviors.
Food for thought? Any questions? Wondering at its relevance to dog training?
agilitynut, Brodie, Kyp!, and Arie
| | | |
| Newfs
Best in Show
Posts:2172
| | 05/09/2008 5:15 PM |
| Posted By agilitynut on 05/07/2008 9:46 AM
Ttraining a chicken... Gilly, when you have it down PAT, how to "train a man" please post. If you can train a chicken you most certainly should be able to clicker train a man  | | | |
| DPFrank
Best of Breed
Posts:327
| | 05/09/2008 7:08 PM |
| | CLICKER TRAINING WORKS GREAT ON A MAN, BUT THE REWARD AFTER THE CLICK SHOULDNT BE FOOD , PERSAY! | | | |
| UBS_Mom
Best in Show
Posts:1710
| | 05/09/2008 7:16 PM |
| "CLICKER TRAINING WORKS GREAT ON A MAN, BUT THE REWARD AFTER THE CLICK SHOULDNT BE FOOD , PERSAY! "  Whatcha ya want Frank? A scratch under the chin or a belly rub???
| | | |
| Newfs
Best in Show
Posts:2172
| | 05/09/2008 8:36 PM |
| Posted By UBS_Mom on 05/09/2008 7:16 PM Whatcha ya want Frank? A scratch under the chin or a belly rub??? hahahahhaaaaaaa  | | | |
| agilitynut
Champion
Posts:56
| | 05/10/2008 12:17 AM |
| Actually, the book "Don't Shoot the Dog" by Karen Pryor covers various aspects of training humans, including husbands and kids, and touches only lightly on actual dog training. Its a great read, and full o both humor and common sense. There's also a new book out, but I can't remember the title or the author ;(
agilitynut, Brodie, Kyp!, and Arie | | | |
| Dobeluvr
Champion
Posts:81
| | 05/26/2008 8:31 AM |
| | I always found clicker training frustrating because i don't have an extra hand. One for the clicker, one of two for the lead to keep the dog under control and then one to give the treat. or even in the house when you tell the dog to do something, sit, there's not always a clicker or treat in your hand. I found praise better. | | | |
| agilitynut
Champion
Posts:56
| | 05/27/2008 8:45 AM |
| Arg! I just wrote a long reply, and it disappeared! Hi Dobeluvr! Many people have trouble getting started with clicker training for the reasons you mentioned, and I know I felt like I had all left thumbs when I started. Using a marker makes things a lot clearer for the dogs, but it doesn't have to be a clicker- many people use a short sound such as yip, yes, or zap. Another way to make things easier is to get rid of the leash I don't bother inside, and now that my pup is older (10 months) I can have her off leash for her actual training sessions, and just leash her up in between. She is actively engaged in the training, so doesn't need to go explore the 11 unfenced acres with woods, fields, deer, geese, and horses, on which we train our agility stuff. Of course, I didn't start that way! I trained inside, and did short sessions outside on leash, and then some short sessions offleash. Now we have a relationship, and I can work with her outside and she'd rather work with me than go do something else. How are you using praise? It is generally a pretty weak intrinsic reinforcer for dogs, as they don't have a verbal center in their brains. I use praise by making it a strong secondary reinforcer, pairing it with a primary reinforcer such as treats. This allows me to reinforce the dog in situations where I can't give a treat. Another way is for it to become a safety signal- it lets the dog know that what he is doing means he won't get an aversive of some kind, such as a leash pop or electric shock. This can be very powerful, but just isn't something I want to do to my dog. Try taking your dog someplace where you can take the leash off, and train the dog to do something new (a trick of some kind, such as rolling over, or going out around a cone) using only praise. If you need your leash, just be aware that you aren't training your dog with praise, you're training with your leash, and the dog is probably working to avoid the leash corrections you may inadvertantly be giving. Another point I want to make is that the clicker and treats are for teaching new behaviors, and sometimes for proofing in new environments. Asking the dog (or even my older pup) to sit in the house doesn't require a click or treat. Asking my pup to sit while 5 feet from an agility ring while a fast dog that she knows is running the course, does. Hey, she comes from screaming lines that are very easily visually stimulated. Agilitynut, Brodie, Kyp!, and Arie | | | |
|
|