It does matter a Toss

Kay Laurence
 October 2008

I was a dog trainer before I was a clicker trainer – and this colours my perception of how I apply the technology. It didn’t take me long to remove myself from the error-punishment strategy and explore way of using intellect to motivate accuracy, desire and commitment rather than avoidance of the opposite.

My first contact was through Don’t Shoot the Dog in the early 90’s, and the lack of any videos or presentations or established styles of applying clicker training resulted in the bare information fertilising my current training strategies. In the late 80’s I was teaching distance control – or signals exercises, through attaching a different command to the different movements, not the outcomes. The movement to “sit” from standing was a different movement to “sit” from the down position. I did not think it was the dog’s responsibility to translate my commands by having to choose the correct action to be successful. The majority of failures in this exercise are the down to sit (do what?) movement, and the dogs will often twitch their hips muscles on hearing the “sit” cue – the unconscious response to sit from standing. Many skilled trainers understand this at an innate level and use a different vocal pitch to indicate “sit-back” from “sit-up”, which are two different cues.

The quantity of competing dogs drove the need for error-free movement to separate the top level competitors from the average. The dog could not afford to just “end up in the down position”, it needed to be the cleanest, most efficient movement to that position.

Using a clicker became the next layer on this teaching strategy, and my understanding was that it was more than just a marker to improve communication, but it attached itself to the neurological/muscle patterns. Naturally I used the clicker for movements rather than just “you done good” communication.

If I am teaching a dog that the cue “sit” requires a specific muscle action then only that action gets  the click. If the trainer clicks for the outcome - a sit position, the dog can sit with any one of several muscle patterns to achieve this. This variable chain, of sit, style A, B, C, N ,G will get reinforced. The variation of the actions will result in variations of the sit position, some degree of hesitation: “hmmm…. which way shall I be most comfortable this time?”

When seeing teaching as actions, not outcomes we begin to use the clicker very effectively to enable the dog with a wide array of skills to achieve our outcomes. The essential contributor to the actions and movements is the reinforcer. What you feed is not nearly as important as how you deliver the food and where the dog collects the food.

Activity begets activity

To teach a movement, broken down into the smallest of time frames, the dog needs to be in an active state. Not highly aroused, but certainly not lethargic. I like to teach a stress free action that will become more fluent with repetition as the muscles develop fluency and strength, and the dog gains confidence in their skill. This requires a acquisition phase that promotes movement, interest, clarity and rhythm. The more prolonged that acquisition phase the faster the movement becomes.

To teach the “action of sitting”, (I find myself regular cueing “sitting” instead of “sit”) the dog will be rewarded from the palm of my hand. To get to this location the dog must take a step forward, out of the sit. When repeating the behaviour the dog can step forwards from standing and backwards to sit with only minor movements of their feet. This ensures that very similar collections of muscle patterns get the reinforcement. In later stages when the behaviour is on cue, the reward will be tossed to several different points around the dog to generalise the action from many opening positions.

If the dog is learning a sit with reinforcement in the outcome there are side effects that can burden the behaviour:

Firstly the dog has to decide which way to sit. Some dogs will be very consistent and some will have tremendous variation. I have seen many adult dogs adopt the puppy posture of sitting with a loose pelvis to achieve the reinforcer, yet when not under cue these dogs sit as adults, with a controlled pelvis.

Secondly the dog will probably relax the muscles in the outcome position, as soon as they hear the click, they avoid movement since the reward is “served in situ”. This is a useful strategy when relaxed outcomes are required.

Thirdly the dog needs to be lured, or cued, to move out of the place of reinforcement so they can repeat the movement for practice. This can become slightly stressful and contradictory and 10 repetitions become more laborious rather than more fun. ALL muscle based learning needs repetition for fluency, and the style of using the cues, clicker, reinforcers needs to be conducive to high numbers of repetition, actions, and not counter an essential part of the learning process. 

The reinforce-for-outcome style of teaching leaves the dog, and trainer, wide open to errors. If the trainer is aware that they will only reinforce for a specific style of sit position, then the dog does not get the critical information until after the behaviour chain is complete. Some dogs are excellent at puzzle solving chains, and can change actions to affect outcomes. Many cannot. For these dogs they begin with a successful sit, and through repetition, appear to repeat the outcome but loose the reinforcer, it is not the “right sit”. Very depressing and the response to cue will slow down, making high numbers of repetitions almost impossible to achieve.

If we require a behaviour to be carried out in a particular fashion, or with a particular attitude, we need to teach the dog how this is done, and not leave it their responsibility to puzzle it out.

Dogs can also learn that the location of reinforcement is the desirable outcome. Focussing on the outcome can leave them without the skills of achieving the outcome. For instance the heel location can receive a high number of reinforcers, but unless the dog has the skill to acquire the heel position they would have difficulty finding it. The dog is near the location but not skilled to adjust their position a hand span to be accurate, or the dog will flare with the hips, and not have the skill to move into a straight line. Additionally the delivery will reinforce a static heel position, whereas 90% of the exercise is a moving position.

Choose the right strategy

The strategy of Outcome Reinforcement should only be employed when the trainer can be sure that the dog is skilled to achieve the outcome without error. If you teach the skills successfully then the outcome will look after itself.

Other examples of teaching actions not outcomes:

Run to the mat. The action of running towards you is highly reinforced and then the target (mat) is placed in the path of the action.

Picking up a retrieve article. The dog must learn the action of moving into the article to take hold, without over shooting, using paws, or excessive hesitation. This is taught through teaching a hold in movement, not stationary, where the dog first learns to move towards an open hand, and then the  article is placed in the path of this movement. 

Behaviours are in a cycle

The whole behaviour cycle is reinforced by the process. The cycle begins as soon as the dog has taken delivery of the reinforcer. I hesitate to say “eaten”, as some dogs are still eating well into the next behaviour, and other dogs will not begin the repetition until they are empty mouthed.

The cycle begins with the dog preparing for repetition (which also allows you to prepare to click), repeats the behaviour, gets a click for success, looks at you for delivery information, goes to collect the food. The cue will be inserted when the “preparing for repetition” part of the cycle is without hesitation.

• The “preparation” is reinforced by the success of the following behaviour.
• The “behaviour” is reinforced by the click
• The “looks at you” is reinforced by the anticipation of collecting the food
• The “collect the food” is yummy, or a fun chase

Between the “gets the click” and “looks at you”, the action determined by the delivery style can set the dog up for easy or complex repetition. It is your choice what behaviour or action your click cues:

Delivery affects behaviour:

1. The dog is clicked, you begin to move towards the dog for delivery in situ:

outcome: relaxing in the position when clicked is reinforced
perfect for teaching maintenance of the relaxed position

2. The dog is clicked and moves towards your hand for reinforcer:

outcome: coming to you after the click is reinforced
very hard for the dog to learn to maintain position or distance from you, the click acts as a “return to trainer” cue

3. The dog is clicked and the food is tossed behind the dog:

outcome: checking position on the click is reinforced

What is regularly reinforced is going to maintain or get stronger.

Tossing Rules:

1. Where possible the food is easily visible on the floor.

Do not use brown treats on brown carpet. If only available to train on grass, then place a tossing area at a suitable spot, a large tray or tarpaulin sheet, and practising tossing to this area – I’m sure the dog will join in just for the game.

Do not use shatter on impact treats, this encourages hunting.

2. What gets tossed stays where it lands.

Rolling, or bouncing food can be a nuisance at the wrong time. If you are tossing food to achieve a perfect set up for repetition, then the dog chasing the food out of the optimum area can set the dog up with too difficult a chance at successful repetition. Some dogs will mark where the hear the food land, and not following it visually.

But rolling food at the right time, can add excellent value to an otherwise boring reinforcer.

3. Throw with follow through.

Chasing treats is a great game, it adds arousal to the training, it makes repetition more enjoyable, but this can all be lost when someone says “look a mouse” and you have no idea where to look. Make very sure that when you toss, your dog is able to clearly read your hand and go straight to the location of the food. Random, land-in-the-tree throws are fun once, and seriously depressing over and over again – to the point where your reinforcer is no longer a reinforcer, or the dog has needed to spend so long searching that they have forgotten what to prepare for.

4. Clean ground, that is not overly attractive.

I sometimes think the Barn carpet is rather like a newspaper, and some stories just catch the dogs’ nose and they have to stop and have a read. Personally my dogs have not found flooring more interesting than training opportunities.

If the dog spends a lot of time sniffing the floor there could be reasons:

1. It is looking for the treat, or has the scent of a lost treat that it is searching for
2. It finds the ground scent more reinforcing than training.
   Make sure you teach re-connection after collection. This is a foundation skill that teaches the dog to absorb the environment, notice the distractions, but stay connected to you.
3. It is trying to avoid returning to training.
   The dog may be tired, may not understand what you are asking for, or just too full. Avoidance should be listened too, as much as you should notice the way the dog prepares for repetition. It should always be with enthusiasm and joy and this will be built into the behaviour.

Tossing is not appropriate when you want muscle relaxation, maintaining of position or holding a specific balance.

Planning

When making a plan for teaching a behaviour what your reinforcer is, how you deliver it and where you place it all have impact on the behaviour and need to be part of the error free ease of constructing the learning, not making it harder.

Examples of intelligent placement

Teaching paw wave: feed the dog with their balance over the paw that is not going to be used.

Recall to hand: food thrown out for chase and running, the dog can run back to touch the hand again. If the food is always thrown behind the dog the dog will hold back as they approach, so the food but the thrown is a variable direction around the trainer.

Sit from moving: throw the treat away behind the dog so that the dog is travelled into a sit, not standing still before the sit.

Settled down: place the food between the dog’s front and back feet, slowly but frequently, so that the dog maintains a rolled hip position.

Going round an object: food sets the dog up at a tangent to the object so that they can pass by.

Learning Games by Kay Laurence contains many training ideas and includes clear instruction of where the food should be placed to set up successful acquisition of the behaviour. The gradual change of this location can increase the criteria in an error-free style, where the dog achieves increasing criteria without hesitation.

An example of this variable delivery style is on You tube:
 http://uk.youtube.com/watch?v=EP7Soem31qg

This is a silent extraction from copy, the full video is: Perfect Foundation DVDs