So here we are with a new board to discuss the science behind the system. Here is another topic for discussion Do we really need to treat? Or is it the click that counts - this is a topic raised on Patricia McConnells excellent blog and being discussed on the positive gun dogs group at the moment. Here's the article: Click and Treat? Or not?
Re: Click and always treat? I have just read that Pippa! A lot of information, and that is without following any of the links. I found it fascinating. Thanks for adding this into the forum, I won't lose it now
Re: Click and always treat? That's the big advantage of the forum I find - much harder to lose stuff. On FB everything disappears so quickly!
Re: Click and always treat? [quote author=editor link=topic=7960.msg112178#msg112178 date=1411492917] That's the big advantage of the forum I find - much harder to lose stuff. On FB everything disappears so quickly! [/quote] Exactly!!
Re: Click and always treat? Really interesting article!! I must say I reward after every mark/click, and I use a high rate of reinforcement. But the concept of the secondary reinforcer (mark/click) being a powerful reward in itself make a lot of sense. Along those lines, is knowing that they've got it right enough of a reward at times? Clicker training is used to teach humans too (eg to teach people gymnastics or a whole other host of physical tasks) and the click (the knowledge of success....and the public kudos that comes with it) is all they usually need. Is the glow of success (getting a click) sometimes enough for dogs too?
Re: Click and always treat? Interesting! I know I am guilty of treating too much - I need to be more intentional about fading out the rewards. It makes perfect sense that the uncertainty of the reward is a strong motivator, which I guess this is what she is saying?
Re: Click and always treat? [quote author=Lisa link=topic=7960.msg112188#msg112188 date=1411495697] Interesting! I know I am guilty of treating too much - I need to be more intentional about fading out the rewards. It makes perfect sense that the uncertainty of the reward is a strong motivator, which I guess this is what she is saying? [/quote] Lisa, I think she is saying that some dogs actually find the click more rewarding than the treat. So some people sometimes just click to mark the behaviour but don't follow through with the treat, something that goes against one of the first rules of classic clicker training, click = treat!
Re: Click and always treat? As a uuuman I wouldn't work for uncertain rewards - if my salary hadn't turned up at the end of the month, then neither would I! But I am sure it must be different for dogs. There is no such thing as delayed gratification for dogs - is there?
Re: Click and always treat? But also that a treat in response to the click doesn't need to be a food treat - could also be a ball throw, a tummy rub,or even a kind word. I suppose it depends how your dog is programmed Lilly would hate a head rub as a reward. But I don't know whether she thinks me saying "good girl" is just a marker, as she often looks up to me when I say it, and ?expects a treat, although COULD it be a reward in itself (probably not! but made me think)
Re: Click and always treat? Yeah, I was thinking the same as I say 'good boy' all the time, mostly with no treat. I'd say it's a weak secondary reinforcer (weak because I use it in a casual, sloppy sort of way). But basically it's like a click with no treat. It's definitely a bit of an attention getter and it has motivating properties - definitely seems to encourage Obi to keep trying.
Re: Click and always treat? Before I read more about it I boldly assumed that the click was like Pavlov's dog where the bell causes the dog to salivate in the expectation of food. I thought the click equated to the treat in the end causing the same happy feeling as the food...... I always treat if I click now.
Re: Click and always treat? Oh, yes, I use other things instead of food rewards, but when I click I always follow through with a reward, either food or a retrieve or a chance to hunt.
Re: Click and always treat? [quote author=Boogie link=topic=7960.msg112194#msg112194 date=1411496180] As a uuuman I wouldn't work for uncertain rewards - if my salary hadn't turned up at the end of the month, then neither would I! But I am sure it must be different for dogs. There is no such thing as delayed gratification for dogs - is there? [/quote] ;D Partial Reinforcement Effect is actually demonstrated in humans and exploited in human psychology too. I agree though, that you would only work one month without salary - wouldn't risk a second!
Re: Click and always treat? This brings me to a question - how and when to fade out the food treats? I'm assuming food is the dog's primary motivator here, obviously. I assume you only start to do it once the behaviour is as consistent and guaranteed as you could possibly expect. Would you then start, maybe, using the occasional low-value treat, say one in five? Then two in five? Then in every five, have three high-value treats, one low-value treat and one head scratch? How long would this process take? Once you've completely "phased out", what would your reward pattern look like? Using the gambling principle, would you reward with a treat once in five, or less? I guess it would still be a good thing to use a very high value reward once in a while to keep the dog on its toes?
Re: Click and always treat? There are no hard and fast rules here, but you clearly understand the principle In both cases it will depend to an extent on your dog, and on the skills you are teaching. Yes do keep using the occasional jackpot, and don't fade out rewards entirely. Other than that, it is often a question of trial and error. if you find a behaviour becoming unreliable then you have gone too far. Fade from high to lower value rewards before you begin eliminating rewards for a proportion of behaviours. And don't forget to use the non-food rewards available to you, such as opportunities to run, fetch etc. Some clicker trainers don't fade to 'no rewards' much at all. That is to say, they may carry on rewarding every time on certain skills (recall for example) and believe that this may deliver more reliable behaviours. Though laboratory experiments do support the use of intermittent rewards, they would argue that there is not a great deal of evidence to support intermittent rewards being more effective 'in the field'
Re: Click and always treat? Thanks, Pippa! I think I would be very comfortable rewarding on every recall as that is such an important command. I guess "suck it and see" is the name of the game for everything else
