Discriminating Cues - part 4

I recently had a project (part 1, 2, 3) where I was working on both dogs' ability to touch a target with their nose or paw on verbal cue only.  As I updated last week both Lance and Vito have a 100% rate (on most days) with the target now and I am pretty much bored with it except for some occasional transfer of their cues to different objects.  Well for Christmas I ordered myself some old back issues of Clean Run Magazine and picked a few in the Linda Mecklenburg's series on her handling system (which is still way over my head but I'm getting there!).  How excited was I when I just finished the November 2008 issue and found an article dealing on this very topic of discrimination learning!  Pamela Reid neatly summarizes several studies dealing with discrimination tasks and I found several great ideas that I wish I knew a couple weeks ago during my project.  Here is a further summary of the summary:

1.  Errorless learning is the way to go, especially if you are teaching your dog to only ever choose one response.  So if you only want your dog to retrieve a bottle of dr. pepper and never the can of coke you would first start out with making the bottle a very exciting object and then slowly add in the boring can to your training.  You could even start out with the can being super tiny while the bottle is slathered with peanut butter, but the goal is to have the dog's choice be so easy to start with that he never even thinks about there being a second option.  In a discrimination study with pigeons trial and error learning produced 2000+ mistakes and errorless learning produced about 25 mistakes in shorter learning time.

This point didn't really seem to apply to my experiment since I want the dogs to have multiple ways of interacting with the target.  But it is a good point to keep in mind when teaching other behaviors, perhaps like teaching scent articles?!

2.  If the dog eventually has to choose between multiple behaviors it is best to alternate the two behaviors in each session.  A study looked at  3 options of teaching a discrimination task 1)training one behavior to proficiency and then start to train the other 2)train one behavior per session but alternate behaviors each day 3)alternate the two behaviors each session.  While all options worked, #3 worked faster since switching between cues forced the animal to pay attention to the differences.

To recap what I did was first teach the nosing of target, then taught pawing the target, then spent a session where I did both behaviors but not mixed yet, and then finally started asking for them randomly in each session.  The article didn't really go in depth into the options so I'm not sure if the options imply that the dog already knows the basic behavior.  I am sure the dog would have to know how to retrieve before discriminating between names of objects to retrieve.  But I am not sure if teaching a dog how to paw an object and then teaching him how to nose an object would be what they are describing in their first choice of the 3 options.   So I may have done a combination of option 1 and 3 or could have just done 3.   

3. Make the discrimination easier first.  Animals can learn both an easy discrimination task AND the hard one faster then animals who just were taught the hard one.  Make sure though that when making the task easier that the animal still has to pay attention to the actual relevant feature and isn't focusing on something different instead.  For example, if the dog was supposed to discriminate between two shapes make sure that he can't also do the task by instead focusing on color, texture, size, etc. 

I feel like this is the most relevant when you are teaching the dog to always pick one object in a discrimination task.  However this may also have been what Arwen was getting at when she suggested that I teach pawing to one target (like a post it note) and nosing to another target.  Then when putting the behaviors together in one session, start with the respective targets at a distance and gradually move them closer until they overlap. 

4.  Make each choice have it's own reward!  Nosing the object might get cheese and pawing it might get a piece of hot dog.  This actually had a huge effect on learning since when the dog hears "whack it" he now thinks of hot dog and that makes it easier for him to choose.

Now this point is what I really needed to do! I can't quite grasp HOW it makes it easier for the dog but I'm guessing it has something to do with classical conditioning!

5.  Give the dog time to think about his mistake.  In the can vs bottle discrimination example you could remove the correct choice of "bottle" if the dog chose the can instead. This allows the dog to see what choice didn't earn him a reward.

Since my project involved multiple behaviors with only one object this point isn't quite as helpful as if I had multiple objects and one behavior.  However I did get much better results when I put the target behind my back after each attempt instead of allowing the dog to correct themself/guess the other behavior right away.  I felt that the consequence of having to wait also made it more important that the dog get it right on the first try.

Anyways, I learned a lot from this short article and really wished I had it when I first started out teaching the nose vs paw task to my dogs.