The value and pitfalls of using treats in dog training
,Food treats are powerful and useful reinforcers in animal training. A common misconception is that treats encourage an animal to only perform for treats. There is some truth in this idea. If treat rewards are continued after an animal has learned a behavior the animal can become accustomed to the regular rewards and refuse to perform if not rewarded.
A helpful analogy would be to think about how we humans behave when putting money in vending machines and slot machines. Let's say I am in the habit of putting money into a vending machine every day to purchase a soft drink. After I put my money in, I hear the chink-a-ching as the quarters fall down into the machine, and then out comes my soft drink. Most people have experienced the frustration of putting money in a vending machine and not getting what they paid for - it's frustrating! The same could be said for paychecks, though the level of frustration is probably heightened! What if you go to work faithfully and perform your job and at the end of two weeks you're told there is no money to pay you? How would you feel? A bit angry? How would you react? Would you continue to work? Would you see what you could do to get your check?
Now consider what happens when we insert money into a slot machine. We put money in knowing that we are not going to be rewarded with a big bucket of coins every time we perform, but we keep on inserting money. Why? Because in the back of our minds, we're hopeful that any time we could win the big prize! Slot machines basically work on a schedule of random reinforcement.
What about random reinforcement, then?
The theory behind random reinforcement is that animals - or people - will continue to be motivated to perform well when randomly reinforced for their behavior, because like the slot machine, they never know when the big reward will come. The problem with this in dog training is twofold. First it is humanly impossible to do anything randomly without the aid of a random number generator.
This is best explained using an example. Say I decide to have my dog sit for me 10 times and I want to randomly reinforce that behavior. I would need to generate a random number list so I know which "sits" to reinforce. Using an online random number generator such as the Research Randomizer, I decide that I would like to reinforce my dog 3 times on a random schedule. I plug in the values for how many random numbers I want, how many sets of random numbers I want, the range of the numbers (1-10 in this case), whether any of the numbers can be repeated (they cannot for this purpose), and whether I want the numbers sorted. Once I have selected all of these parameters, I click on the button, Randomize Now, and presto! I have a list of random numbers. Then I use this list of random numbers to reinforce my dog for sitting. If my dog sits 10 times and I generated a random number list of 1, 2 and 7, I would reinforce my dog on the 1st, 2nd and 7th sits! Needless to say, this is not very practical.
The second problem with random reinforcement is that it doesn't give the dog what Chris Bach calls "useful information." While randomly reinforcing laboratory rats for pressing a lever can teach the rats to continue to press the lever for tasty treats, random reinforcement of human and dog behavior doesn't give either the information they need to improve performance. Just think of the "random" trinkets you have received from an employer. Did these small gifts, provided during an employee recognition day or other event such as new employee orientation serve to motivate you to work harder?
So, instead of "random" reinforcement, a more accurate and effective way to think about reinforcement that is not paired with each and every behavior is what I prefer to call a "variable schedule" of reinforcement. A variable schedule of reinforcement means that the dog is reinforced on occasion without an regularity, just enough to maintain interest and performance but not enough to cause the dog to think s/he will receive a reward for every sit or down.
A "fixed schedule of reinforcement," is one to one pairing of a reward with a behavior. This is essential during training, but once the animal understands the behavior, this fixed schedule of reinforcement should be abandoned. If it isn't, the result is a dog that comes to expect a treat for previously learned behaviors such as sit, and when s/he doesn't get it, runs through his/her repertoire of tricks to see what will elicit the desired response from the owner. Performance suffers and the dog may refuse to perform at all.
Chris Bach recommends reinforcing a dog when s/he experiences a "learning breakthrough" and when s/he re-commits to paying attention to the trainer vs. something else in the environment. I personally have found that this technique really puts a positive spin on training and makes it an enjoyable experience for both dog and trainer.
But my dog isn't motivated by treats...
Since all dogs must eat to survive, all are motivated by food to some extent. However, some dogs are more motivated by toys such as balls, and play can be used very effectively as a reinforcer in training. Food is generally easier to use when teaching new behaviors, but after your dog has learned a behavior, praise and petting can also be used as reinforcers.
Food is also useful during training because it gives us an indication of a dog's mental and emotional status. Sometimes a dog will refuse food because its level of arousal is too high. When a dog is in a high state of arousal, he is not in an optimal state to learn and simply getting the dog to take food can lower his arousal state and put him into an operant or thinking state.
A skilled trainer can help you determine why your dog may be refusing food during training, and can help you determine or even create a motivator for your particular dog.
© 2009-2014 Canine Connection LLC All rights reserved.
Home About Training philosophy Services Contact Links Testimonials
A helpful analogy would be to think about how we humans behave when putting money in vending machines and slot machines. Let's say I am in the habit of putting money into a vending machine every day to purchase a soft drink. After I put my money in, I hear the chink-a-ching as the quarters fall down into the machine, and then out comes my soft drink. Most people have experienced the frustration of putting money in a vending machine and not getting what they paid for - it's frustrating! The same could be said for paychecks, though the level of frustration is probably heightened! What if you go to work faithfully and perform your job and at the end of two weeks you're told there is no money to pay you? How would you feel? A bit angry? How would you react? Would you continue to work? Would you see what you could do to get your check?
Now consider what happens when we insert money into a slot machine. We put money in knowing that we are not going to be rewarded with a big bucket of coins every time we perform, but we keep on inserting money. Why? Because in the back of our minds, we're hopeful that any time we could win the big prize! Slot machines basically work on a schedule of random reinforcement.
What about random reinforcement, then?
The theory behind random reinforcement is that animals - or people - will continue to be motivated to perform well when randomly reinforced for their behavior, because like the slot machine, they never know when the big reward will come. The problem with this in dog training is twofold. First it is humanly impossible to do anything randomly without the aid of a random number generator.
This is best explained using an example. Say I decide to have my dog sit for me 10 times and I want to randomly reinforce that behavior. I would need to generate a random number list so I know which "sits" to reinforce. Using an online random number generator such as the Research Randomizer, I decide that I would like to reinforce my dog 3 times on a random schedule. I plug in the values for how many random numbers I want, how many sets of random numbers I want, the range of the numbers (1-10 in this case), whether any of the numbers can be repeated (they cannot for this purpose), and whether I want the numbers sorted. Once I have selected all of these parameters, I click on the button, Randomize Now, and presto! I have a list of random numbers. Then I use this list of random numbers to reinforce my dog for sitting. If my dog sits 10 times and I generated a random number list of 1, 2 and 7, I would reinforce my dog on the 1st, 2nd and 7th sits! Needless to say, this is not very practical.
The second problem with random reinforcement is that it doesn't give the dog what Chris Bach calls "useful information." While randomly reinforcing laboratory rats for pressing a lever can teach the rats to continue to press the lever for tasty treats, random reinforcement of human and dog behavior doesn't give either the information they need to improve performance. Just think of the "random" trinkets you have received from an employer. Did these small gifts, provided during an employee recognition day or other event such as new employee orientation serve to motivate you to work harder?
So, instead of "random" reinforcement, a more accurate and effective way to think about reinforcement that is not paired with each and every behavior is what I prefer to call a "variable schedule" of reinforcement. A variable schedule of reinforcement means that the dog is reinforced on occasion without an regularity, just enough to maintain interest and performance but not enough to cause the dog to think s/he will receive a reward for every sit or down.
A "fixed schedule of reinforcement," is one to one pairing of a reward with a behavior. This is essential during training, but once the animal understands the behavior, this fixed schedule of reinforcement should be abandoned. If it isn't, the result is a dog that comes to expect a treat for previously learned behaviors such as sit, and when s/he doesn't get it, runs through his/her repertoire of tricks to see what will elicit the desired response from the owner. Performance suffers and the dog may refuse to perform at all.
Chris Bach recommends reinforcing a dog when s/he experiences a "learning breakthrough" and when s/he re-commits to paying attention to the trainer vs. something else in the environment. I personally have found that this technique really puts a positive spin on training and makes it an enjoyable experience for both dog and trainer.
But my dog isn't motivated by treats...
Since all dogs must eat to survive, all are motivated by food to some extent. However, some dogs are more motivated by toys such as balls, and play can be used very effectively as a reinforcer in training. Food is generally easier to use when teaching new behaviors, but after your dog has learned a behavior, praise and petting can also be used as reinforcers.
Food is also useful during training because it gives us an indication of a dog's mental and emotional status. Sometimes a dog will refuse food because its level of arousal is too high. When a dog is in a high state of arousal, he is not in an optimal state to learn and simply getting the dog to take food can lower his arousal state and put him into an operant or thinking state.
A skilled trainer can help you determine why your dog may be refusing food during training, and can help you determine or even create a motivator for your particular dog.
© 2009-2014 Canine Connection LLC All rights reserved.
Home About Training philosophy Services Contact Links Testimonials