Variable Rewards And Conditioned Reinforcers

[corvus]

There's something that I've been wondering about and not quite understanding since I started training Kivi's recall over a year ago.

We know that dogs are gamblers and will work harder for unpredictable rewards. I trained Kivi's emergency recall following Leslie Nelson's Really Reliable Recall dvd. Nelson teaches that you ALWAYS reward the emergency recall with something really amazing. As I understood it, she was aiming for a conditioned response, I guess the way most people teach to ALWAYS reward after a click.

So what I am getting at is are recalls more reliable if you have tried to build recalling as a conditioned response by rewarding big every time, or is it better to exploit the gambler in a dog by varying the rewards and reward rate?

I probably should watch the independent breeds section of the RRR dvd, actually.

[poodlefan]

So what I am getting at is are recalls more reliable if you have tried to build recalling as a conditioned response by rewarding big every time, or is it better to exploit the gambler in a dog by varying the rewards and reward rate?

I would and have opted for the former. I can't express it in training-speak other than to say the bigger the payoff (and the more frequent), the greater the motivation for the dog to obey. The more rewarding a behaviour, the greater the frequency of response. It's the WIFM principle at work.

Edited September 9, 2009 by poodlefan

[Vickie]

So what I am getting at is are recalls more reliable if you have tried to build recalling as a conditioned response by rewarding big every time, or is it better to exploit the gambler in a dog by varying the rewards and reward rate?

I don't use either method, but if you are using a reward based system, isn't it inevitable that the rewards & rates are varied at some point?

[poodlefan]

So what I am getting at is are recalls more reliable if you have tried to build recalling as a conditioned response by rewarding big every time, or is it better to exploit the gambler in a dog by varying the rewards and reward rate?

I don't use either method, but if you are using a reward based system, isn't it inevitable that the rewards & rates are varied at some point?

I don't vary the reward schedule on recalls when walking offlead.. they get rewarded every time. When they are looking at the kangaroo and hear the whistle, I don't want them weighing up whether they'll be rewarded or not for coming.

Edited September 9, 2009 by poodlefan

[Vickie]

I don't vary the reward schedule on recalls when walking offlead.. they get rewarded every time. When they are looking at the kangaroo and hear the whistle, I don't want them weighing up whether they'll be rewarded or not for coming.

so do you ever whistle for a recall anywhere else? around the house, at training, at a trial etc...and do you always have a similar value reward for each of those situations?

[poodlefan]

I don't vary the reward schedule on recalls when walking offlead.. they get rewarded every time. When they are looking at the kangaroo and hear the whistle, I don't want them weighing up whether they'll be rewarded or not for coming.

so do you ever whistle for a recall anywhere else? around the house, at training, at a trial etc...and do you always have a similar value reward for each of those situations?

No, I only use the whistle off lead walking in the bush. They have a "come" cue for other situations where the reward is more variable. When I use the whistle, every dog is rewarded regardless of how far away they are when called.

Edited September 9, 2009 by poodlefan

[Monah]

Mine get rewarded EVERY time they come when called.

[corvus]

Well, the way Nelson explains it, if you reward big every time, then when you don't have treats on you and your dog is running off towards a busy road, it doesn't matter that you have no treats. They'll come running and the fact that one time out of 200 they came and didn't get a big reward won't make much of an impact. We nearly always have something like roast meat or steak or rissole on us when we are going anywhere Kivi will be off leash. We certainly vary what we have, but recalls typically earn cooked meat of some sort. Sometimes raw meat or leftovers at home.

I think maybe it needs a lot of maintenance, though. A month ago Kivi's recall was fantastic, but we haven't kept up the practice over the last few weeks, especially after getting the puppy, and today OH said he was not exactly galloping over like usual. OH treated this by giving him a lower value reward than usual, because his reasoning is that he didn't perform as well usual. I'm not sure that that's really the point. Does he even know that his performance was not up to scratch when a conditioned recall is basically a conditioned reinforcer, and a conditioned reinforcer is only as good as the last time it was reinforced? If he's thinking before he comes running, then it's not really conditioned and we need to work on it some more to boost it back to habit. Maybe OH should have gone over the top and rewarded really big for a slow recall to encourage a faster one next time?

I should point out that the ER is not the same as a regular recall. The regular recall is treated more like regular training, with variable rewards. Sometimes a big one, sometimes a game, sometimes a rub, sometimes just a little tidbit. The reason you split them up is so you can have a conditioned recall you have the luxury to spend a long time building up without poisoning its effectiveness by calling when you don't have any decent rewards or when your distractions are too high for the level your dog is currently at. It took about 12 months for Kivi's recall to be fantastic in just about any situation.

[poodlefan]

All I know is really basic operant conditioning theory:

Behaviours that result in positive consequences tend to increase in frequency

Behaviours that result in negative consequences tend to decrease in frequency

Examples in the recall context for the second proposition include calling your dog to punish it or calling it in the offlead park only when you're putting them back on lead to leave.

[Kelpie-i]

I read and watched Bill Campbell's version of the emergency recall and he would use a different word/cue for it and would teach it by throwing a very small pouch filled with chains (to make the rattle noise) used to interrupt the dog if he/she was distracted. He would pair this with the recall cue.

To be honest with you, I cannot really understand why you wouldn't just teach a reliable recall each and every time. To come is to come, not matter what, where or how....but I suppose this might probably be the topic of another thread...sorry Corvus

If off lead, reward profusely I say...the coins should fall out of the slot each time your dog recalls to you. I also add a highly conditioned Name Recognition to the equation which works brilliantly to ensure that dog responds to his/her name immediately and without hesitation. The recall follows on from that.

[corvus]

To be honest with you, I cannot really understand why you wouldn't just teach a reliable recall each and every time. To come is to come, not matter what, where or how....but I suppose this might probably be the topic of another thread...sorry Corvus

No no, feel free to discuss.

I just did it like Leslie Nelson said.

I say my dogs' names quite a lot. Usually to get their attention, but not always. Sometimes I just say it to greet because I'm a verbal creature by nature. Inevitably, my dogs' names lose their sharp edge over time, even though periodically I might sit there with a pile of treats and go "Kivi" *treat* until they are all gone. I just say that name too often without a good reward. I see the ER as an easy way to preserve a good recall, although as I said, I think it still needs maintenance. At least I'm not saying it a dozen times a day just to give my dog a pat on the head or have them look at me or something, though. I talk way too much around my animals. I don't really need or want them to pay attention to me every time I say their name.

Kivi's name IS his ER, but just said in a certain way with a certain pitch. I don't need him to pay attention to me every time I say his name, but he can differentiate tone, and I would like him to respond to his name when I say it in a particular way. I have a "look at me" way I say his name and a "we're going to do something exciting" way to say his name and a "I have roast beef" way to say his name as well as a "just wanted to touch base" way to say his name and so on.

[Kavik]

To be honest with you, I cannot really understand why you wouldn't just teach a reliable recall each and every time. To come is to come, not matter what, where or how....but I suppose this might probably be the topic of another thread...sorry Corvus

No no, feel free to discuss.

I just did it like Leslie Nelson said.

I say my dogs' names quite a lot. Usually to get their attention, but not always. Sometimes I just say it to greet because I'm a verbal creature by nature. Inevitably, my dogs' names lose their sharp edge over time, even though periodically I might sit there with a pile of treats and go "Kivi" *treat* until they are all gone. I just say that name too often without a good reward. I see the ER as an easy way to preserve a good recall, although as I said, I think it still needs maintenance. At least I'm not saying it a dozen times a day just to give my dog a pat on the head or have them look at me or something, though. I talk way too much around my animals. I don't really need or want them to pay attention to me every time I say their name.

Kivi's name IS his ER, but just said in a certain way with a certain pitch. I don't need him to pay attention to me every time I say his name, but he can differentiate tone, and I would like him to respond to his name when I say it in a particular way. I have a "look at me" way I say his name and a "we're going to do something exciting" way to say his name and a "I have roast beef" way to say his name as well as a "just wanted to touch base" way to say his name and so on.

IMO you are creating confusion by doing this. Either you want the dog to respond EVERY time you use a command (in this case the dog's name) or you don't want them to respond. I DO want my dogs to respond to their name when I say, so I do not just talk to them using their name when I don't want a response.

ETA: it is responses that these that make you tend to sound inconsistent.

Edited September 9, 2009 by Kavik

[Aidan]

We know that dogs are gamblers and will work harder for unpredictable rewards.

That's not quite correct. Variable Schedules of Reinforcement build behaviours that are more resistant to extinction.

I use (and teach, although compliance is low) Continuous Reinforcement [Differential]. Responses at the required criteria are reinforced on a continuous schedule. Any sub-par performances are not reinforced. If you reinforce sub-par responses you get more... sub-par responses!

[Kelpie-i]

say my dogs' names quite a lot. Usually to get their attention, but not always.

You can use absolutely any word to get the dog's attention but it must be the same word and it MUST be conditioned and reinforced (continuously in the initial stages and then intermittently) having said that however, just like clicker training, where the food always follows the click, same should occur with Name Rec exercise. For those who use a clicker, you will notice the dog's eyes light up at the sound of the click, this is because the click sound is now classically conditioned. The aim with Nam Rec is to create a very similar conditioned response but with the dog's name. Therefore, I usually suggest using a nickname for the 'everyday' name and use the dog's real name for when you need the attention pronto.

The reason I prefer to use the dog's name (and remember I teach companion dog training classes, not trial classes), is that the very first thing that comes out of the owner's mouths should the dog happen to shoot through the door is the dog's name.

Others will most certainly have a different view to this but I find this works well and really helps the recall process. I use the Name Rec exercise for reactive dogs as well...helps interrupt the dog if he/she is about to eye lock onto another dog.

[corvus]

Kivi's name IS his ER, but just said in a certain way with a certain pitch. I don't need him to pay attention to me every time I say his name, but he can differentiate tone, and I would like him to respond to his name when I say it in a particular way. I have a "look at me" way I say his name and a "we're going to do something exciting" way to say his name and a "I have roast beef" way to say his name as well as a "just wanted to touch base" way to say his name and so on.

IMO you are creating confusion by doing this. Either you want the dog to respond EVERY time you use a command (in this case the dog's name) or you don't want them to respond. I DO want my dogs to respond to their name when I say, so I do not just talk to them using their name when I don't want a response.

Okay, sorry, but that is rubbish and here's why: the ER is his name three times in quick succession at a high pitch. What matters is the pitch and the repetition of the "e" sounds. I can call the same pitch and just say his name once and he doesn't think it's his ER unless he reckons I'm going to use it at any moment and is waiting for it, in which case he'd come running if I used his regular recall, which is his name once. Sorry, should have pointed that out. His ER sounds nothing like the way you would say his name to get his attention, but it's still his name. He does respond to the ER EVERY time. I have said that it's a very good recall. So obviously if it created confusion it wouldn't have worked so well.

If you want your dogs to respond to their name every time, then you don't need an ER, do you? I can't help talking to them. It just comes out. No point in me denying my basic nature for the sake of my dog when it's so easy to just split them all up.

Kelpie-i, I am a little confused. You said why not just use the dog's name, but then you said you teach people to use a nickname for "everyday" things and the dog's full name for a request for attention? And I am assuming that the name recognition then encompasses the recall? That sounds like just a variation on what Nelson teaches to me. What's the difference between having an "everyday" name and a "pay attention" name and having an everyday recall and an emergency recall?

[Kavik]

That is very different to what you said corvus, it is easy to see why I got confused You should have been more clear. ETA: Just not sure what else I was supposed to think when you said his name IS his emergency recall (exactly what you said) but that you also talk to him using his name in conversation just because you like to talk and you don't always want him to respond to his name?

You said you use his name for his recall, in everyday speech for no reason except to talk to him and his emergency recall, so that is what I assumed you did as that is what you said. I'm sure you know that to the dog using the word 3 times sounds is effectively a different word to saying it once (sit as opposed to sitsitsit). The recall is no different to any other command in that respect.

I do not use an emergency recall command. I have their name to get their attention and 'informal' check in type recall and their formal recall command (Diesel's is Come, Kaos's is Here) where I expect them to come in quickly to a particular position.

Edited September 9, 2009 by Kavik

[corvus]

We know that dogs are gamblers and will work harder for unpredictable rewards.

That's not quite correct. Variable Schedules of Reinforcement build behaviours that are more resistant to extinction.

I love having you around.

I use (and teach, although compliance is low) Continuous Reinforcement [Differential]. Responses at the required criteria are reinforced on a continuous schedule. Any sub-par performances are not reinforced. If you reinforce sub-par responses you get more... sub-par responses!

I was reading Grey Stafford's book and he uses a Least Reinforcing Scenario in which he just waits for a few seconds, then either asks something different, asks the same thing again, or ends the training session depending on what he thinks would be best at the time.

I think my confusion mostly lies in not knowing whether a slow recall is worth reinforcing because I'm meant to be reinforcing every recall to get a conditioned recall, or if because my dog is capable of and usually performs a beautiful recall at a dead run, we should consider a slow recall not a recall?

Kavik: I do not use an emergency recall command. I have their name to get their attention and 'informal' check in type recall and their formal recall command (Diesel's is Come, Kaos's is Here) where I expect them to come in quickly to a particular position.

That's almost the same thing. The only difference is it doesn't matter where Kivi ends up for his ER as long as it is in a good position for receiving the goodies. A lot of Nelson's students use "come" or "here" as their ER, just put their dog's name in front to get their attention. We only opted for something else because we are both as bad as each other when it comes to talking to the dogs and we just needed something we didn't say in regular conversation. I should have been clearer about that, yes. I do use Kivi's name in all of those situations, but I'm careful to make sure it sounds different to him if I expect him to respond in a particular way to it.

[Kelpie-i]

Kelpie-i, I am a little confused. You said why not just use the dog's name, but then you said you teach people to use a nickname for "everyday" things and the dog's full name for a request for attention? And I am assuming that the name recognition then encompasses the recall? That sounds like just a variation on what Nelson teaches to me. What's the difference between having an "everyday" name and a "pay attention" name and having an everyday recall and an emergency recall?

Sorry to confuse you Corvus, what I meant was this...

1. You can use any word you wanted to really, as long as THAT word is conditioned

2. If you are going to condition the dog's name, then you must NOT use the dog's name generally (ie..just for every and home lovey dovey speak), you use (no teach) a nickname or something else instead to preserve the conditioned name.

3. The only reason I prefer to use the dog's name is because this is usually the first word that will come out of your mouth (subconsciously) if your dog is heading straight for the busy road.

Yes the name recognition precedes the recall but both are taught separately.

Nothing wrong with teaching an ER, provided you are very clear with the criterion for each type of recall.

[Vickie]

I think my confusion mostly lies in not knowing whether a slow recall is worth reinforcing because I'm meant to be reinforcing every recall to get a conditioned recall, or if because my dog is capable of and usually performs a beautiful recall at a dead run, we should consider a slow recall not a recall?

Just out of curiosity...why do you (or anyone else) want a recall at a dead run?

To be honest with you, I cannot really understand why you wouldn't just teach a reliable recall each and every time. To come is to come, not matter what, where or how....but I suppose this might probably be the topic of another thread...sorry Corvus

no, I think it's right on topic & totally agree which is why my recall is not based purely on rewards.

I will say though that there have been a couple of times where the panic in my voice has elicited a quicker response than normal but it's not a "trained emergency recall"

I am having serious recall issue with one of my old boys at the moment...silly old bugger has gone almost deaf (literally). He still hears a whistle depending which way he is facing, but we are now having to go through a process of teaching hand signals.

Edited September 9, 2009 by Vickie

[Kavik]

Just out of curiosity...why do you (or anyone else) want a recall at a dead run?

Well it does look impressive in competition

http://www.youtube.com/watch?v=AhGtMLs3GaY