[Loisiana]

Okay, I wasn't sure what title to give this thread, as it isn't really the clicker necessarily that I have a question about, but the method of training where silence means keep trying and a click or verbal means you're right.

My question is, how do you transition the idea of silence means keep trying to the obedience ring where you have to keep silent while you are working? I do use a clicker to initially mark some behaviors, but I teach my dogs that if they are wrong I will immediately let them know verbally. My silence means they're doing the right thing. So when I'm silent in the ring, they are getting the message "you're right," followed by praise at the end of the exercise.

I have heard of problems with novice stays with some dogs trained this way: when the handler is silently standing across the ring for three minutes, the dog wonders what it should be doing and will begin to offer behaviors. Same idea for heeling....doing a full heeling pattern, dog never hears affirmation that they are right, so they begin to offer other options. But I'm sure not every dog trained this way has that problem, so I'm curious as to how the change is made clear in the dog's mind.

Even though I don't train this way, I want to have a better understanding of all training methods, especially by those who have proven it to work in the ring.

 

[katieanddusty]

You gradually build up duration so that the dog is okay with waiting longer and longer for a reinforcement (and the dog learns that if they stay long enough they'll eventually get a reward, but if they move they definitely won't get a reward). A dog that literally got up and started offering behaviors in the obedience ring (as opposed to just coming toward the handler or going into the wrong position) would be pretty bizarre, and my guess would be that the dog wasn't very experienced in actually having cues attached to behaviors. Trained dogs know that when you say "sit" you're only going to reinforce them if they sit, when you said "stay" you won't be reinforcing any movement, etc.

The only time a clicker-trained dog might try something like that would be if the trainer decided to shape "roll over" or "crawl" from a down, right before the trial (meaning they've been having the dog down and then waiting for sideways/forward movement and rewarding that). That would be a dumb thing to decide to train right before a trial. After a while once the dog understands the full behavior of rolling over, though, you attach a cue to the rolling over and the dog learns that rolling over will only get a reinforcement if you've said "roll over," while "down" still means that you'll only reinforce for staying down. Same idea with "wave" from sit - at first you might have the dog sit and reinforce any paw movement, but once you have the complete behavior and add the "wave" cue the dog learns that waving the paw will only be reinforced if you've said "wave."

Some people will also use a "no-reward marker" to tell the dog that what they're trying is not going to be reinforced and the dog should try something else. I think that mostly applies when you're trying to shape one thing and the dog just keeps offering something else. Because most of us come from more traditional training backgrounds and have that instinct to use that "NO!", it usually ends up a verbal correction with a fancy name, which is not necessarily wrong but also is not clicker training. Most clicker trainers try to use something like "try again" as opposed to "no" or "wrong," to keep themselves focused on the fact that the dog should be learning primarily through positive feedback and not negative.

This isn't really a scientific-based method I've read about or anything, I just started thinking about it and what I do to let my dogs know that I do or do not want them to offer behaviors. Sometimes my dogs will get "stuck" just standing there offering eye contact, because I reinforce eye contact a lot and don't have a specific cue for it (if they haven't been told anything else to do I want them looking at me). So when I go to shape something, they'll usually stand there staring at me like good dogs. But they know that if I say "what?" they should go find something to offer (most of the time I'm working with interaction with an object - a wobbly board, a target, etc). I don't really remember how that started, I think I was just trying to rev them up and get them more excited about interacting with the object by going "what? what?" and then that sort of became a generic "go do something" cue.

 

[Loisiana]

A dog that literally got up and started offering behaviors in the obedience ring (as opposed to just coming toward the handler or going into the wrong position) would be pretty bizarre, and my guess would be that the dog wasn't very experienced in actually having cues attached to behaviors.

Yeah, if the dog actually started offering other tricks that would be bizarre (and pretty funny if it isn't your dog), so let me clear up that I meant more minute movements such as moving a paw on stays or getting an inch or two out of heel position during heeling (more the small increments that you look for during shaping). Although a more dramatic example for heeling that occurs often is the dog getting stressed from not getting afirmation from its handler, and begins to severely lag due to its stress. But I wasn't actually referring to a dog bowing, spinning, etc (although it would be cute!)

Thanks for responding!

 

[katieanddusty]

I don't see the paw moving as a problem unique to clicker-trained dogs at all. It's a problem with defining what you want the dog to do. Maybe the handler accidentally rewarded the moving-the-paw a few times (like if they walk back to the dog, the dog moves a little and they still click/treat), or maybe the handler just wasn't paying enough attention to the dog (practicing out-of-sight stays without having someone watch their dog or something like that) so the dog has been moving his paws around a lot and still eventually getting rewarded for the stay. Same thing happens a lot in agility - dog is supposed to stay at the start line, one time the dog gets excited and moves a foot but the handler still continues with the run (which hopefully is rewarding for the dog), next thing you know the dog is creeping forward or completely breaking the stay. The problem is just that the handler has failed to define what "stay" really means (regardless of the method they choose to do so).

The lagging while heeling would be a problem with the reinforcement schedule. Ideally the handler would be always varying when the dog gets rewarded, so sometimes the dog gets rewarded for starting out really well and walking two steps, other times the dog doesn't get rewarded for the length of a full heeling pattern or even longer, and everything in between. That gets a consistent enthusiastic response because the dog is always thinking maybe this next step will be when the reward comes.

 

[Loisiana]

So if I'm understanding correctly, the idea of silence meaning the dog needs to keep trying is only used in the beginning of training something, and in a finished product the silence does not have that meaning to the dog?

The varying reinforcement schedule for heeling makes sense, it's pretty much what I do. I was just confused as to how the dog was going from silence to mean one thing in learning to silence meaning another thing later.

I find training theories so interesting.

 

[tippykayak]

I don't do pure clicker training, but I have been pushing myself to transition to more and more positive methods over the years, and I've been very happy with the results. I'll just chime in that the silence I use during the execution of the behavior is a pre-reward silence in the dog's mind. He knows that longer and longer behaviors are necessary before a reward, so he learns that patience.

 

[katieanddusty]

So if I'm understanding correctly, the idea of silence meaning the dog needs to keep trying is only used in the beginning of training something, and in a finished product the silence does not have that meaning to the dog?

I think so. The dog learns that certain behaviors have duration and aren't always going to be reinforced right away. So once the dog is sitting, you might wait for just a second before you click, then 2 seconds, then 3, and so the dog gradually gets the idea that if he sits long enough the click will come. I actually don't have a "stay" command with my younger dog, because the dog learns that "sit" means "sit until you hear a click, a release word, or another command."

The dog can also learn that he might have to do a chain of several behaviors before the reward, but eventually there will be a reward. In this case the cue for the next behavior is sort of a reward for the previous behavior, because it means the dog is one step closer to getting the reward. So every time my dog hits a contact in agility and I give the cue for the next tunnel, that acts as a reward for hitting the contact because my dog likes tunnels and each obstacle is one step closer to the cookie. Ideally I would also do fun matches so that I could sometimes give an actual food reward for hitting the contact in a trial environment.

 

[Jersey's Mom]

Another important thing to note about clicker training is that you're not giving the behavior a name (cue) until it is complete and up to your standards. So in the early learning stages when the dog is experimenting with the behavior (a good example would be the one you used with heeling, where the dog makes minor positional adjustments) that behavior is not called "heel." You click and reward when the dog is doing the right thing (or making some move toward the right thing in the case of free shaping)... you adjust and fine tune the behavior... and then when the dog is consistently performing exactly as you want them to, you introduce the cue. In that way, there is never a question in the dog's mind as to what exact position constitutes heel, or sit, or whatever.... because it's never been associated with anything else.

That, combined with the points Katie made about making sure you aren't inadvertently rewarding sloppy behavior goes a long way towards avoiding the types of SNAFUs you describe.

Oh and Katie... I love that you're not really into the nitty gritty of the science, but you managed to describe the Premak Principle to a tee!! That is where each subsequent command serves to reinforce the previous behavior. It's a wonderful tool... and in a lot of ways it's the key to getting through a full obedience (or agility or whatever sport) routine without losing points (or time) for lavish praise... and of course, for fading out food rewards as is necessary for competition.

Julie and Jersey

 

[FlyingQuizini]

but you managed to describe the Premak Principle to a tee!! That is where each subsequent command serves to reinforce the previous behavior.

I don't mean to be a science nerd, but the concept of commands reinforcing previous behaviors is more like chaining, not the Premack Principle. When behaviors are trained positively, cues become reinforcers b/c of the strong R+ history attached to them. This is what makes chaining work.

Premack is high probability behavior reinforcing low probability behavior... Also known as "Grandma's Rule"... i.e., if you eat your veggies, you can have ice cream. Let's say a dog pulls on leash at the sight of a squirrel.... get one step of loose leash walking in the presence of a squirrel... CLICK... reward with opportunity to chase squirrel.

 

[FlyingQuizini]

I think the issue of dogs offering behaviors in the face of silence is a fluency and stimulus control issue. Like Katie said, duration has to be taught and the dog must also learn that only the eventual CUE is what indicates that a reward is forthcoming.

SO - if a dog understands duration and has behaviors on stimulus control, he'll readily wait for the CUE and will perform the behavior for the required duration.

 

[katieanddusty]

I think opportunity to do a tunnel is a reinforcer for most dogs But yeah, in general what Steph said is more what I was going for. And it's okay, being a science nerd is cool

 

[Jersey's Mom]

but you managed to describe the Premak Principle to a tee!! That is where each subsequent command serves to reinforce the previous behavior.

I don't mean to be a science nerd, but the concept of commands reinforcing previous behaviors is more like chaining, not the Premack Principle. When behaviors are trained positively, cues become reinforcers b/c of the strong R+ history attached to them. This is what makes chaining work.

Premack is high probability behavior reinforcing low probability behavior... Also known as "Grandma's Rule"... i.e., if you eat your veggies, you can have ice cream. Let's say a dog pulls on leash at the sight of a squirrel.... get one step of loose leash walking in the presence of a squirrel... CLICK... reward with opportunity to chase squirrel.

LOL, I could have sworn I had put something in there about making it something the dog really wants to do (i.e. a tunnel, a spin, or some other particular trick that the dog just loves to throw at you).... but on rereading, apparently not. I shouldn't be posting tonight... I'm multi-tasking out the wazoo trying to get packed (I have to leave for the airport in 2 hours! Thankfully I'm finally ready, but now I'm wired). Thanks for picking up my fumble there! Oh, and nothing wrong with being a science nerd... I make my living at it!

Julie and Jersey

 

[Ljilly28]

I do use a friendly no reward marker instead of a long silence if the dog is not on the right track. "Too Bad"- just means try something else