Ahhh but sometimes in convention we find wisdom. Having said that, not being one to merely follow convention, I think we can agree to disagree on the value of a variable reward. I'm not familar with the research you are referring to, however I am familiar with the research conducted by places such as Southwest Research Institute, the University of Maryland, Auburn University, Seaworld etc etc etc. I do admit, I'm a dog trainer, not a behavioralist. My concern is, as previously stated, will the dog perform to the standards.

DFrost

Lee,
I can't find the article, but I know the info is references in Morgan Spector's book on clicker training (now don't boo me on the clicker stuff, I'm not real big on it though I have used it in detector work and see quite a few applications hin that context).

I've got that book and don't remember that discussion, but I'll have to go hit the books tonite. <img src="/ubbthreads/images/graemlins/smile.gif" alt="" />

I don't have years in doing search work, but I do know that in the few dogs we have worked it is always better to keep the reward constant and vary the length and duration of the search.

We have achieved the best results that way. When using variable reward in searches I have found it dulls the dog's intensity.

Constant reward is still a pretty good motivator, and if the search is variable and the location/time of the reward is actually variable. . .then aren't we using the same theory. . .kinda?

yes, you are still using intermittent reinforcement when you vary the duration of the response that is required in order to earn the reinforcer. The dog can still get a 1-for-1 reward (one search; one reward) but by varying the duration of the behavior, you build persistence (resistence to extinction).

An effective schedule of reinforcement is the variable ratio. As you all know the reward is given after a certain number of behaviors have occurred and that number varies.

Variable Interval - This occurs when the reward will occur after a certain amount of TIME has passed.

Both are well documented and if you see a drop in motivation trying to implement either it would be my guess your timing is off or perhaps you have simply over trained/stimulated the dog and the reward value itself has dropped.

With a fixed or continuous reward schedule you can and will reach the same fatigue range but even faster.

Some situations where, by definition of rewards, it would make little difference if the dog made a find or not. Why? Time. The interval of time is to long for many things a trained dog does so rewarding a dog in a building search each and every time he looks is nto a problem. Where you get into trouble is when you sequence an exercise and the dog is doing in his mind the “same” job not another job that is the same as the last search. At least this is what I thinks and I hope I explained it so it makes sense.

Thanks Lee

Some situations where, by definition of rewards, it would make little difference if the dog made a find or not. Why? Time.

We probably don't ask a dog to do searches that are long enough to where an intermittant reward schedule in training is necessary. If you use an intermittant reward schedule then you will see a lessening of the intensity of the behavior you are looking at, but it will be more durable. It is questionable whether you attain a point where this accomplishes anything other than a lowering of the intensity of the sought after behavior.
If you are asking a dog to do huge searches this may become important. But, most of the time what you see when training is done this way is that the dog stops searching at peak intensity. They may come into drive again when they are presented with a stimulating event(like sticking their face into the target odor). Problem is you have seen a decrease....dramatic one often.... in the most desireable of behaviors which is a dog actively seeking out the substance, person, etc. It has been theorized (Auburn University for those that are into this sort of thing) that it is not the dogs great sensitivity that makes them so precious to us in detection work but the fact that they are hunting creatures, actively seeking out their prey.
I often see decreasesin the hunting behavior in dogs which have been presented with a less than advantageous reward schedule and system. Dogs that indicate when presented with an odor, but do not actively seek out their goal. Dogs that in patrol work you have to use on leash, if you take 'em off they are pissing on bushes and sniffing everything they can, and moving only 'cause they are told to. Detector dogs that I coin as substance indicators and not substance seekers. If you stick their nose into the scent cone they'll get it but don't ask them to work the areas well themselves. Proficiency in street operations drops. This is due to the construction of the program in which the dog is grounded. Unfortunatly this discribes a vast majority of police dog programs.
Is the intermittant reward system wrong? Nope. Is consistant rewards wrong? Nope. But, like many many things in dog training variable reward systems, non-variable reward systems, varying, time, difficulty, distance, rewards, changes from a simple exercise in theory to one of application, standards, and goals (and in police dog training often the standard is way below what the goal should be).
In dog training we often overlap and use various programs at the same time or sequentially.
Like a mix of primary reinforcers and secondary reinforcers. Operant conditioning and classical conditioning overlap and in practical dog training you cannot, and should not seek to prevent it in most cases. Everyone does this outside of a laboratory. It pays to keep in mind that variations on the theme are imperative in the craft of dog training, particularly in police service dog work.

Kevin I agree with you on most areas,

Still, I would question the drop of motivation to handler error and not on variable scheduling. Not that most handlers consciously use variable schedule anyway. Often what is “best” is not what is practiced yet by error we have a sort of weak variable scheduling. To much error = drive loss.

The theory placed forward by Auburn is sound for those looking for application to dogs and trying to make the peg fit. Yet when we apply it other animals scent work, they can be taught also to “hunt” for the reward even if, by nature, they are scavengers with little or no prey drive. Rats for example.

Some would say because rats need to hunt for food beacuse thet search for food so the theory holds up.

Yet how far do we go with it?

All amimals hunt for food and water with or without prey drive. Still it is an interesting concept worth further examination I think.

I've found this an interesting, stimulating and very civil discussion. Thanks guys.

DFrost