Food rewards – a matter of technique

There has been another comment on the discussion about use of rewards. Bette writes:

Although I have been training dogs for over forty years, and had a dog training facility, I have never encouraged the use of food rewards. I recently acquired a new dog (two years old), and started him at an obedience school that stresses food and clickers. He is a (conformation) show dog with strong basic obedience skills. Before we started school, I worked him with only praise, as has been my custom. Once we began classes, I incorporated food (I refuse to use the clicker). I have seen a marked reduction in the quality of his responses, and he is constantly looking for the food. His recall sits are no longer straight, as he aims for the pocket where the treats are. He is now in a hurry to complete the entire exercise so he can get the morsel. I have discontinued the food, and he is much better. Like Carol Lea Benjamin says in Mother Knows Best, look at the bitch. Her puppies obey her despite her not having food, clickers, rolled-up newspapers (heaven forbid), etc. From now on, my dog will work for me because of our mutual bond, not because I have to rely on some external motivator. (Bette, September 25 2013)

I replied (at length, as I usually do … sorry folks, bear with me)

Hi Bette,

Interesting comments. I am familiar with the behavior that you describe. However, I have to say that the problems you describe are the result of faulty technique. They are not the result of using food rewards as such.

I can explain that in a moment. However, first I want to simply make the point that the best training method is reward-based. This is not the same as “food-based”. There is no point in an instructor saying “you should use food” and you saying “I want to use praise”. The point is, if you are reinforcing behavior, the reinforcer is whatever your dog finds most rewarding, at that time, in those circumstances.

One dog’s general preferences might be toys, food and praise, in that order, and another’s might be something completely different.

Also in different environments, your dog’s preferences will change. No point in offering a “foodie” Lab a piece of fillet steak if her top level reward right now is to jump in the creek. A motivational trainer will use all the available rewards, choosing the most appropriate ones for the circumstances.

The other issue is that you said you refused to use a clicker. I wonder why? The clicker is an important precision tool which, when used correctly, identifies for the dog exactly what behavior is being rewarded. So it is a communication device. It speeds up the learning process because it helps the dog to know what exactly is being rewarded.

The clicker is far quicker and therefore more precise than using a verbal bridge, which some people use. Clicker trainers learn to break behavior down into very small slices. Many other trainers try to teach behavior in chunks that are too big. The result is that the whole process takes longer, and also the resultant behavior is “broad brush stroke” rather than being built up from a series of “micro behaviors”.

For example, in the show ring, your criterion for clicking and rewarding will gradually increase. The timing of the click determines whether you are rewarding your dog for going into the stand position, for example, or staying in position for a given length of time. For a stay you can build the time up to whatever duration you need. Then you can add the next criterion, which might be “look in this direction” and when the dog has mastered that, you might want to click “and while doing all that, bring your ears forward”. If you are only using a reward (whether it is food or praise or anything else), it has to be delivered in a timely fashion, but it is usually a bit late – e.g. you dog pricked up her ears for a moment, but the timing of the reward was when the ears had come down again. The wrong behavior gets reinforced.

When you train using these methods, you always have a criterion. This means “what precise behavior does the dog have to offer in order to earn a reward?” If your dog presents crookedly, don’t reward. A clicker trainer would click when the dog was in the correct position, not otherwise.

When you use a clicker, she learns that in order to earn the food (or any other) reward, she has to offer the desired behavior. No amount of looking at your treat bag will make food materialise in her mouth. Offering the behavior will. This is what the dog comes to realise. The click marks the behavior, the primary reward reinforces it. If the click is timely, you have a bit of time to deliver the primary reward. In some cases I put a bowl of food on the table, a couple of meters away, click (or use a verbal bridge, a special word) to mark the behavior, then go to get the treat.

The other issue of technique is that food should not generally be used as a stimulus, a means of getting the dog’s attention or getting the behavior. There are exceptions, such as lure-reward training for puppies, but that is a special case. If you do lure a bit with an older dog, it is done briefly to “kickstart” a behavior, then phased out, and then you refine the exercise further by shaping.

So I might do a bit of a lure to initially get the dog to present front, and the rest is done by shaping – selectively rewarding the better responses, and gradually raising the criterion, lifting the bar, so to speak, as the dog improves, until you eventually get a very accurate position, performed consistently. Dogs can “get” this very quickly if the trainer is communicating clearly.

I teach Treibball, and one of the exercises is “handler orientation”. This means that from wherever the dog is on the field, she has to go behind the ball, and face the handler. Initially I generally lure bit to get the dog facing the handler. After that, the handler turns in different directions, and the dog has to figure out that re-orientation is required, without being lured or guided. If the handler does this in small increments, the dog comes to understand very quickly. I give a food reward “in place”, which means not only do you click when the dog is in the correct position, you also deliver the primary reward when the dog is still in the correct position.

When the dog is in position, she has to wait (so the click would come after she has taken up her position and waited, not immediately after taking up the position). Then the handler cues the dog to push the ball. For most dogs, being allowed to push is a reward in its own right, which is given only if the dog completes the previous steps. The click would then come after the dog has successfully driven the ball towards the handler and goal.

One final point. Most novice handlers use voice, clicker and delivery of reward at the same time. For example, you say “stand”, and before the dog has even come into  position, your hand is gravitating to the treat pouch. This means the dog’s attention goes to the treat pouch, and this over-shadows the word “stand” and takes the dog’s attention away from the task of offering the correct behavior. People do not realise they are moving their hand. It is hard for handlers to learn to assume a neutral body posture before giving a cue, wait until the dog has completed the behavior, then click or bridge, then deliver the reward – in that order.

If a dog is very keen and excited about food, you can use that, but you have to use it correctly. The reward must be timely, only given contingent upon the desired behavior (not just given when the dog looks at the food, or does anything else), and it must be given when the behavior meets your current criterion.

On the question of rushing through an exercise, if it contains several components, as all obedience exercises do, we teach each element then put them together. As I described with Triebball, you don’t go on to the next step until the dog has performed the first one correctly. Then each step becomes a reward for the previous one, and the dog is being rewarded along the way as well is being given a primary reward on completion. In addition to this, a sequence is back chained, so for example, the “present correctly to front”,  being the last behavior, is taught first, and each other component is added on and leads to it. This motivates the dog to get each step right, and to find each step rewarding, and gradually builds motivation to complete the exercise correctly.

I hope this explanation helps to clarify that it is not simply a matter of shoving food in the dog’s mouth. It is a matter of knowing how to use reinforcement of all kinds to build the desired behavior.

Advertisements

About transformational1

I have many interests and I have had a varied career. I am a semi-retired professional dog trainer, specialising in the use of positive reinforcement. I do some consultations, I run instructor workshops and I am setting up a Dogs and Psychotherapy Research Project. I have a law degree from Melbourne University (but have never practiced) and I am passionate about Human Rights. My first degree was in Sociology. I worked as a social researcher on issues such as low income housing, women's refuges and women in the workforce. I live in an inner suburb of Melbourne, Australia, and I have a German Shepherd called Chance.
This entry was posted in clicker training, Dog training, Dogs, food rewards, Learning theory, Training methods and tagged , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s