Sniffy Training

Before I started training Monica, I had to get some practice with a virtual rat, Sniffy.  Sniffy Lite version 3.0 is a computer program that replicates the behavior of a rat in an operant box.

The first step in training Sniffy (and Monica) is to train it to associate the light blink and the noise with the food in the magazine.  This type of training is called magazine training.  I started by reinforcing her behavior when she was near and facing the magazine.  I reinforced multiple times in a row, so that I could give her the most reinforcement when she was paying attention to the magazine.  Sniffy preformed a wide variety of behaviors, from rearing up against the walls to sniffing herself.  I magazine trained Sniffy for 35 minutes before she learned to associate the food with the noise. 

The next step in training is to shape the rats' behavior to press the lever.  To teach Sniffy to press the lever, a behavior that would not occur very often by chance, I had to reinforce behaviors that would lead to pressing the lever.  This form of training, based on successive approximations, is called shaping.  I started reinforcing her behavior when she turned her head toward the lever.  Then I reinforced that behavior less often as I required her to get close to the lever before being rewarded.  I broke down the shaping process into a series of tasks before I began the training.  As Sniffy understood each step, I would require more from her and lead into the next step of the training.  She went from turning her head, to walking to the lever, to sniffing the lever, to rearing up on the wall, putting her paws on the lever, and eventually, to pressing on the lever.  During shaping Sniffy was on a continuous reinforcement schedule, meaning that she was reinforced everytime she pressed the lever.  Shaping is more difficult than magazine training, but it is more rewarding to see the rat understand what you are trying to teach it to do just through a reinforcement of an approximation of the goal behavior.  Shaping is beneficial because it speeds up the process of learning, but you must pay very good attention to the behavior of the subject and the time you reinforce its' behavior.

After spending time with Sniffy, I felt confident in my abilities to train Monica.  It turned out that Monica was easier and more cooperative than my virtual rat!

When I placed Sniffy on a fixed ratio schedule, FR3, its' behavior differed from when I was shaping it.  The bar presses were more consistent because it understood that the pressing the bar resulted in food.  During the same amount of training time, the occurance of the bar pressing behavior was much greater during the fixed ratio than the shaping.