Traditional horse training relies on teaching horses to move away from discomfort. They learn to move from whips, from our legs, from the bit, from pressure on their halters, and from martingales and side reins. Most of our training relies on some form of negative reinforcement to produce the behavior we want.
Suppose we want to turn the picture around and have the horse learn by linking things he wants to the behavior we want to shape. How would that change our training? First we would have to find things our horse wants. The list would include things like a chance to run and play, a chance to stand still after hard work, time with a favorite pasture mate, a pat, a vigorous neck rub, a chance to roll in a sand pit, and best of all for many horses - food.
The problem with this list is obvious. It's hard to use these things in a training session. You can't let your horse drop and roll every time he gives you a right answer. Even petting can be a problem. If you give him a massage every time he does something you like, your hand is going to get tired.
Also, for best results rewards need to be delivered exactly when the behavior occurs. That way the horse can clearly mark what it was doing and repeat it again for another reward. Delays between behavior and reward can lead to confusion. You think you're rewarding your horse for dropping his head. He thinks it's for swishing a fly with his tail. So how do you resolve the problem? Very simply. You introduce a secondary reinforcer.
Food, or a pat on the neck is the primary reinforcer. It's the thing the horse wants. The secondary reinforcer, or bridging signal as it is also called, is a conditioned signal which becomes linked to rewards. It tells the horse "You are about to get a treat."
Whistle every time you give your horse a carrot, and you'll start to notice that whenever you whistle, he'll be looking around for his treat. If you only give him carrots when you whistle he won't be checking out your pockets at other times. This is the key. It allows you to use food without it creating problems. The horse learns to expect a treat only when he is presented with a particular signal. I work around clicker-trained horses with my pockets stuffed full of grain, and I am never mugged. Instead, the horses are listening to me and waiting for the magic sound of the clicker.
Once the horse understands that a whistle or a click means food, I can link that sound to behavior. The horse begins to learn that the only time he hears the click is when he presents certain behavior. Now it's the horse who thinks he has me trained. He's aware of the power of his own actions. Present behavior. Get treat. What a wonderful system!