lizzybeth, can you chime in on how to help break the chain? I'm not sure how to approach it. I'll tell him down, he downs, and before I can say "good boy" and treat he's rolling over. After he rolls over he's in the down and just stares at me expectantly. When he doesn't get a treat for that he gets back up and starts offering other behaviours, but as soon as I cue him to 'down' again he repeats with the 'down' and then automatic 'over.' =P I'm not totally sure what to do. I've taken to downing him somewhere that he physically can't roll over (he won't roll into a wall, for example, so if I down him right next to a wall he's stuck in the down) and marking and treating for that, but it doesn't seem to be helping break the chain.
I would start by working somewhere where he can't do the roll, like you said next to a wall. Preventing the behavior will go a long way in stimulus control work.
While you're working downs next to the wall, I'd also work on rewarding him for a default position... maybe sit or stand. The default position is where you want him to be when you didn't cue anything. You can have more than one default position - for example, when I'm working on loose leash walking my default is a sit, but when I'm working cues like sit, down, retrieve, go to crate, etc. my default is stand. DON'T CUE the default or click it, but treat your dog when he does it. When your dog knows what to do when he doesn't know what else to do (lol), he will be better prepared to listen to your cue and do what you are asking, rather than guessing and throwing out a whole bunch of different behaviors.
So, you're practicing next to a wall so that he doesn't offer roll overs, click/treating downs, and treating the default position. Your goal is for the dog to go into his default position when you release him and he's waiting for another cue (rather than offering different behaviors), and doing a down - and not offering any other behavior - when you cue down.
Next, you can move away from the wall and practice the same thing. I'd work on rewarding the default before cueing anything. Just stand there and wait for the default, then treat. When he's offering nothing but the default - he's calm and relaxed and focused on you - then cue the down. If he does any behavior besides down, look away until he stops offering behaviors, and wait for him to offer the default; then TREAT the default; you might even give several treats to reward staying in the default for longer durations. Then cue down again. Continue until he starts doing the down when you cue down. When he does downs, release him and encourage/reward him for going into the default position, then cue down again.
Ok, so at this point you should be getting downs when you cue downs - and no other behaviors - and the default when you release him from the down. You should not be getting roll overs because you should not be cueing roll overs; if he offers them he will learn that they don't "pay," that only downs and defaults pay.
Next do the same session, but instead of down you're going to cue roll over. Now only roll overs and defaults will pay.
Now comes the tricky part - teaching the dog to listen to the difference between roll over and down, and choose the right behavior.
Start by rewarding the default. Then cue either down or roll over, and click/treat if he chooses the right behavior. Then release, and reward the default again. He is more likely to do the behavior that just got rewarded, so it will be easier on him (and I suggest to do in the beginning) to cue the same behavior 3-5 times in a row before switching and cueing the other behavior. After each cue, release and reward the default behavior.
If you cue and he chooses the wrong behavior, look away from him for 3-5 seconds, then reward the default behavior and cue again. It will be easier on him (and I suggest to do in the beginning) to cue the same behavior until he gets it right; for example, if you cue roll over and he downs, look away, reward the default, and then cue roll over again.
Keep in mind at this point that cueing a behavior is rewarding. So if you cue down and he rolls over, simply recueing down is a reward for the roll over (especially if he's ever been introduced to a variable schedule of reinforcement). That's why it's important to give him a mini time-out, so that he learns that he chose wrong. I also prefer this method (guess what? It's a No Reward Marker) to using a negative marker, such as "no" (which I guess is also, technically, a No Reward Marker).
Eventually you can progress to making the cues more and more random. Some trainers like to choose ahead of time and make a list of which cues they're going to give, randomly, by flipping a coin or something. If they cue and the dog chooses the wrong behavior, you simply move on to the next cue on your list.
Does this make sense??