Sunday, January 23, 2011

An introduction to dog training, pt 2

The schools of training - old and new school

There are two basic schools of training when it comes to dogs: old and new.

Old school, or traditional dog training tends to be very hierarchically based, where a trainer tends to take a very no-nonsense approach with their dogs. They expect a dog to listen and obey commands lest risking punishment for non-compliance. This is a very corrections-based method. You'll often find a lot of people using terminology like “be the alpha” in conjunction with these training styles and the two ideologies are closely related.

Rewards also exist in this methodology, but often they're pats or verbal praise in lieu of food rewards. There is the concern that reliance on food to train results in reliance on food for compliance for the rest of the dog's life, and that a dog may not obey in situations where no food is present.

Ever pop your dog's leash? That's an traditional training method. It is a correction for the dog misbehaving in a situation where it can be expected that the dog knows what it should be doing, but is doing something else instead. The general idea behind these methods is that you first teach a dog to do something, then you punish it for not complying if necessary. (Personally I feel that the teaching is often rushed, and a dog does not always know why it is being reprimanded.)

These tactics can and do work, but they can also be misused and abused. In soft-tempered dogs, or already anxious animals you might find the animal shutting down when faced with the stress added by the handler.

Popular old school dog trainers are Cesar Milan, Brad Pattison, and Don Sullivan. Certain training schools also adhere to these methodologies. One of the most popular and widely spread schools is Bark Busters.

The schools of training - new school

The new school of dog training is reward based, founded on the basic principle that you reward desirable behaviours to increase the frequency at which they're offered. You use reinforcers like food, play and praise as rewards to lay the groundwork for positive associations and future training.

This new school of training was developed to work with a wide range of animals, from bears and lions to whales and rats. I'm sure you're all aware, but it's not recommended you alpha roll a bear, and leash popping a whale is difficult, to say the least. A better method of communication and training had to be developed, as physical coercion simply cannot occur with some species.

This new school of dog training normally goes hand in hand with a tiny handheld device called a clicker. Prior to its use, there was some discussion that methods of delivering praise and reward did not inform the animal of successes with enough promptness and precision to create the required cognitive connections for speedy learning. The response was the clicker. It is a unique sound that the animal only ever encounters as a mark for good behaviour that is more reliable and precise than a human voice can ever be.

These techniques have been used with whales, bears, lions, as well as domestic dogs and cats. It's a faster, more reliable way to communicate to your animal. The click is used to mark a split second of a behaviour and communicate that a reward is on its way. I'll speak a bit more about using a clicker in the “Tools”section.

The positive reinforcement that this methodology is based on is excellent for working with quiet dogs, or dogs lacking self confidence.

Popular new school dog trainers are Victoria Stillwell, Pat Miller, Jean Donaldson and a bunch of other people you've probably never heard of.

Check out this video as an introduction to clicker training. What Is Clicker Training? by Kikopup (youtube link)

Here are some popular misconceptions about new school training:

The dog won't perform without food present: This is a popular problem people have. The issue is that the handler is not fading out the food lure fast enough when teaching a new behaviour. The idea is to fade out the lure quickly, and to reward on a variable rate of reinforcement so the dog is never sure when it will receive its reward.

The dogs become fat with too many treats being provided: Yes, you'll be giving your dog a lot of food. But you don't always have to use treats. For 95% of the training I do with my dog I use kibble, and if I'm doing an intense training session I'll skip my dog's meal and feed it to her one at a time as rewards.

Dominance and dog training

Have you ever heard someone refer to themselves as their dog’s “alpha”? How about being told that if your dog does X then it’s a sign that it’s exerting dominance over you? Now, if you hear people saying as much you can quietly roll your eyes for the following reasons:

Dominance theory is based on flawed studies performed on captive wolves in the 1960s, and basic brute force prior to then. There have been many studies that display that dogs do not operate under a dominance hierarchy. Dogs' social structure is more fluid, and not typically hierarchical.
Association of Pet Dog Trainers says:
One of the biggest misconceptions we find ourselves faced with is the definition of "dominance." Dogs are often described as being "dominant" which is an incorrect usage of the term. Dominance is not a personality trait. Dominance is "primarily a descriptive term for relationships between pairs of individuals." and moreover, "the use of the expression 'dominant dog' is meaningless, since "dominance" can apply only to a relationship between individuals. (Bradshaw et al., 2009)

The idea is that dominance is never elicited forcefully, rather it is voluntarily given. What we often consider “dominant” traits are often anxiety-based behaviours that can and do intensify if addressed with force.

So, forget about dominance theory – regardless of whether it exists or not it’s completely irrelevant. Your goal, when addressing behavioural issues, is not to understand a dog’s emotional state, but rather to focus on what the dog is currently doing, and what we want the dog to do instead. Then we help the dog understand what’s expected of it by rewarding desirable behaviours.

“Alpha rolls” and “scruff shakes” are not only unnecessary, but they can damage a dog’s trust in its handler and can intensify the already-present anxiety.

This article says it better. If you have anyone who’s a bit stuck on dominance theory it’s a good thing to show them:

However, dogs need structure and routine, and they need to know what’s expected of them. Just leave the alpha-dominance theory at the door where it belongs.

Learned helplessness

Learned helplessness is a term that means a condition of a human or animal in which it has learned to behave helplessly, even when the opportunity is restored for it to help itself by avoiding an unpleasant or harmful circumstance to which it has been subjected. In short, the animal has given up on helping itself. It's a sad state of affairs.

Learned helplessness is sometimes the result of harsh training and over-correction. If a dog is constantly punished for not behaving accordingly (and remember that the dog might have no idea what “accordingly” is) it might give up on ever trying to figure it out.

That is the primary reason why I infinitely prefer clicker/positive reinforcement training since it rewards the dog for trying new things. If you have a creative, enthusiastic dog there's really no limit on what you can teach it.

Here's one of my videos of Cohen when she was one year old to show the fruits of positive dog training:

Selective disobedience

One major criticism of dogs trained primarily with positive reinforcement is that they’re selectively disobedient. It makes sense – when few reprimands are given the dog is not scared of making a mistake and incurring punishment.

A lot of people who train improperly can’t get their dogs to behave without waving a treat in front of their noses. And to these people I say, YOU’RE DOING IT WRONG.

The goal of positive reinforcement training is to lay the groundwork for good behaviour when you’re out in the real world and away from your bait bag. You do this by quickly fading lures and utilizing a varied rate of reinforcement so the dog’s never sure when it’s getting a reward. When you’re beginning you can maybe get 3-5 behaviours for a treat, but veterans can expect to get 50+ without pausing to feed/reward.

The other key aspect of maintaining obedience without threatening punishment is that you have to do some work to convince the dog that they want to do what you want them to do. Convince them that it’s their idea. I’m really not too sure how to describe this, but in order to achieve success with this you have to know your dog very well, and have a good working relationship with them. You really must go out of your way to relate well to your dog.

Training methods:

There are a couple basic training methods: luring, capturing, shaping and targeting.

Luring: This is when you have a piece of food in your hand and you slowly move your hand in such a way for the dog to follow it. For instance, slowly moving a treat above your dog's head to get him to sit, or moving the treat to your left side to get the dog into heel position. This is normally the fastest way to achieve a new behaviour, but is not the best for retention.

Capturing: This is when you capture a behaviour your dog offers or does in day to day life. For instance, you can capture a bow that your dog might do while stretching after a long nap. You do this by marking the behaviour you want with either a click or a marker word (followed by a treat) to communicate to your dog that it was desirable. This is probably the slowest way to teach a new behaviour, and the dog retains the training moderately well.

Shaping: Shaping is when you mark successive incremental changes in behaviour as they approach your end goal. Say you want to teach your dog to pull a lever. You can do this by shaping it from the ground up. Start by rewarding your dog (with a click or marker) if he looks at the lever. Do this a few times. Then hold out on the treat when he looks at the lever and wait for him to approach it. Repeat. Then wait for him to touch it. Repeat. Mouth it. Repeat. Etc until you've built the behaviour you were looking for. This can be very time consuming, but the dog will retain the information much better since he's actively taking part in the learning process.

Targeting: Targeting works similarly to luring, however instead of asking your dog to follow a piece of food, you're asking your dog to follow a target, which can be your hand, a target stick, a mousepad etc. Targeting works on the principle that "If I touch the hand I will get a treat", compared with luring works like "I want that treat so I will follow it". The dog has learned previously that a reward will be provided after touching the target, so the target isn't a primary reinforcer (see the section below on information on primary/secondary reinforcers). A great usage of this behaviour is if you have a dog who's been known to bite if disturbed from the couch (bite inhibition and/or resource guarding), you can teach him to move from your seat on command by asking for him to target something on the other side of the couch so you can manipulate the dog both without bribery and without touching it then rewarding for good behaviour. This is a very fast way to teach positioning once you've taught your dog how to target the desired object.

Primary vs Secondary reinforcers:

Reinforcers are, as you know by now, a response to a dog operating on its environment that will increase the likelihood that a behaviour will repeat itself. Reinforcers can be broken down into multiple categories:

Primary: A primary reinforcer is providing something that the dog wants in reaction to a desirable behaviour. These can be food, toys, access to other dogs, getting to go outside. Not all dogs value the same things, but food is the most universal of all primary reinforcers, and is most commonly the strongest. This is why food is used so often.

Secondary: A secondary reinforcer is a click or marker word which signals the imminent arrival of a primary reinforcer to which the dog has been classically conditioned.

Tertiary: A tertiary reinforcer (or tertiary bridge) is a signal that the secondary reinforcer is on its way: "Good, keep doing that and you will get a click."

An accomplished trainer will likely use all of the above reinforcement techniques. Primary reinforcers are instantly gratifying and very strong. Secondary reinforcers are intrinsically less gratifying and slightly weaker, and so on. However once you move into secondary/tertiary reinforcers you're able to achieve much more precise, advanced behaviours that would otherwise not be available to you if you focused entirely on primary reinforcers.

No comments:

Post a Comment