Saturday, January 24, 2009

49. Marker-based Training without food

We recently received this question! Thought we would post the answer in case it was useful to anyone else.


"I cannot train using treats, and I cannot find any information on using other rewards with clicker training, specifically with loading the clicker. I do plan on using praise as the reward since my boy responds well to that. Any in site (sic) you could provide would be most appreciated."


Primary Reinforcer
The reason food is used, then later faded or switched to other types of reinforcers, is that it is a primary reinforcer (therefore has intrinsic meaning to dogs-food, sex, air, water, chasing, barking, digging etc and the dog does not have to learn to love it), is an easy choice for most people, can be tossed from a distance and allows for quick delivery and therefore many repetitions in short order (key to marker-based training). This allows for faster learning. Treats used are small (we are rewarding the dog, not feeding him), soft so it can get eaten quickly with no crumbs and the dog must value them. To avoid weight gain, simply remove the same amount of food from his feed dish as you use for treating each day. Some people use the kibble itself if their dog will work for it (and if they feed kibble).


Timing (of the marker),
Rate of reinforcement and
keeping training Criteria small enough for the learner to succeed are the three fundamentals of marker-based training.


A Few Considerations of using Secondary Reinforcer for Training New Behaviors
You could use toys (a secondary or learned reinforcer) but that can slow the process down significantly. For example, if you throw a ball, it takes more time for the dog to chase and catch the ball and bring it back (assuming he already knows how to do that or you must go get the ball). Using a bean bag limits how far it can move and would be a better choice. Similarly, using a toy often teaches the dog to be ready to move, which may not be the best choice in early learning stages of a stationary or relaxed behavior such as down, sit or stay.


When using praise, it is usually paired with stroking/physical affection and this may limit you to how far away the dog can work or the position where you are in relation to the dog as you always have to either go to the dog to deliver it, or have the dog move to you. Doable, but again, slows the process down. Most people can quickly throw treats for long distances from any position once he learns to catch them so the dog can stay at and work at a distance.


Once a behavior is understood well by a dog (i.e. is on cue and dog is able to perform it in a variety of environments) that is typically when secondary reinforcers are brought in.


Training Secondary Reinforcers
You can train a secondary reinforcer (pretty much anything else the dog learns to love), but that will likely involved using food for at least part of the training as you have to pair the new reinforcer with a primary one many times so it now has a new meaning for the dog. Periodically, you may have recharge it as well as sometimes they lose meaning/value to the learner. Some examples: a high-pitched voice, a scratch on the back end, a neck massage -anything that becomes meaningful to the dog. Having said that, some secondary reinforcers can come to be more reinforcing than primary ones, if you find the right one! Think of a ball crazy dog, for example.


How to Pair them:
Introduce the Secondary Reinforcer (click or other sound, toy, affection)
Follow it quickly with a Primary Reinforcer x50 to 100


Do this many times until the secondary reinforcer clearly has meaning. The dog should be looking for the primary reinforcer when the secondary is presented.


Now you can use the secondary reinforcer  after your click but will probably have to go back and re-charge it periodically if it loses it's appeal.
Can You Pair the new Secondary Reinforcer with a Click?
(using a secondary reinforcer with a secondary reinforcer).
The short answer: Yes, if it works for the animal. Remember that the animal defines what is reinforcing so they are the factor that makes the decision if it works or not.
 
If you say "Good dog" and pair it with a belly rub, these are both secondary reinforcers. If your dog will work for them, then it works. If not, try something else. If you can remember to say "Good" in a short quick way, you should maintain the benefits of using a precise behavior marker. Studies have shown that the metallic click does speed learning by 45% or more.

If you don't care that you might dilute the effect of the clicker sound, then you can try pairing your new secondary reinforcer with the click.


There is actually a discussion on a facebook page where the horse trainers are experimenting with secondary reinforcers only but conditioning a different marker sound so they don't muddy it in case it doesn't work.

Here's a link for those that are on facebook.
http://www.facebook.com/pages/Clicker-Training-Horses/205152254918?v=wall#!/topic.php?uid=205152254918&topic=16107


You will have to 'Like' the page before you can comment but you should be able to read it without. I also bookmark it on my Internet Explorer Favorites so I can find it again.


Other Reinforcers:
For less formal behaviors such as waiting to go through the door, going through the door becomes the reward. Going for a car ride (if the dog likes doing this) greeting a human friend, another dog, sniffing, chasing squirrels (often called "life rewards" in case you want to Google it) etc can all be creatively used as rewards and reinforcers. Even things and events in the environment can become reinforcers if you take the time to train (the pairing is called conditioning) them.


It is interesting to note that having to train this process means praise means nothing to a dog unless it is first paired with something else of great primary value to the dog-usually food. We train this inadvertently when we feed from the table or pair our voice with food after the dog performs a trick etc.


If you train without food, you will have to be more observant than the average trainer to see what the dog is showing you what is meaningful to him and use your creativity to build on that. Normal reactions are around food preparation-the sound of the fridge opening, the can opener, the tinkle of a spoon on the bottom of a bowl are all learned (or conditioned) like Pavlov's dogs.


Many people think dogs should and want to work for us. In reality, most dogs have no interest in pleasing us, they work for themselves and have to learn that working with us is rewarding in some way (or work to avoid correction or aversive stimuli like our unapproving voices. That is why people who train using traditional correction-based methods have to resort to punishment-they believe the dogs should want to work for us , then correct them when they choose not to.


My previous dog loved to perform agility because of the reaction he got from the crowd-he loved their laughing, clapping and 'oohing and awing' as he performed. He learned this inadvertently. I was not something I taught. However, I did notice that for him in dog class, if someone laughed at something goofy he did, he would repeat it. He was an incredibly sensitive dog to human emotion. He would often stop at the top of the A-frame to make sure everyone was watching. He was quite sensitive yet confident-in short a showman. He was a rare dog though as most aren't that self-aware, clown-like and eager to please. This is a really rare combination.


Exploring New Ground
So far everyone I know has started their animals with food as the reward for the secondary reinforcer, before changing to secondary reinforcer with secondary reinforcer so you will be exploring new ground if you try this. I'd love to hear how it works for you! Even better would be if you could video your progress and show us! This would be great for people who refuse to use the clicker on the basis they don't like to use food. Would love to have an alternative to tell them so they and their animal can benefit from this approach!

No comments:

Post a Comment

Questions? Comments? Ideas for others? Please Share them here.