Blog post about the dog training stages – ‘watch me‘ protocol By Laure-Anne Visele, dog training in The Hague, June 2012.
The secret recipe for ALL dog training
Training stages are steps through which to take the dog before he masters something new. If you get the principle, you can pretty much teach anything.
In short, the steps are:
- Lure the dog into what you want him to do, reward each time.
- Introduce a command, lure, reward each time.
- Fade out the lure so you eventually only need the command, reward each time.
- Reward less and less often, then unpredictably
- Perfect and generalise for different durations, distractions, contexts, locations
The theory
There’s no getting around it, we need to throw in a bit of terminology. I’ll make it as painless as I can.
Shaping
Taking the dog from early, shaky, attempts, to the target behaviour through gradual stages. You do this by gradually increasing how well the dog has to perform before he gets rewarded. Basically, you are rewarding closer and closer approximations to the desired behaviour. You might want to consider shaping if your dog doesn’t seem to get it, even when lured with the most delicious treats. It beats overwhelming the dog by asking him to do something he simply cannot do yet.
The Cue
It’s trainerese for a command, an order. We prefer to say ‘cue’ as it sounds less like an army drill, which dog training shouldn’t be.
Make sure your cues are:
- dog-catchy (e.g. with a ‘tch’ sound)
- short (‘anticonstitutionally’ is quirky, but you’ll soon tire of it)
- not likely to embarrass the living daylights out of you (‘bang!’ with a pointed gun gesture to get him to lie on his side is a hell of a lot less funny on the vet’s table or at the groomer’s)
- not a word that you commonly use. I, am STILL teaching my dogs ‘no’ to make him stop what he’s doing. So he is always looking quizzically if I drop the word in a casual conversation. ‘What did I do THIS time?’
- not too similar to another training word. e.g. ‘Wave’ to wave goodbye (a nice trick) and ‘Wait’
- not poisoned: as in, make sure you haven’t been using it before, but sloppily, thus blurring the cue’s meaning and dampening the intensity of the dog’s response.
Luring
Guiding the dog into the desired position by getting him to follow a treat. You can imagine the dog’s nose is a magnet, and the treat is metal. Using luring, you can position the dog’s nose (and, by extension, the rest of this body), to where you want them. This treat is not a ‘reward’ yet. It is the promise of one. He will get a reward when he is in the desired position.
Prompting
Prompting is when the dog is not so much guided into position, but is placed into position. An example of prompting would be if you call the dog, then drag him to you with the leash, or ask him to sit, then press his butt down with your hand.
Why I am not a big fan of prompting? Because very little insight is going on in the dog’s head so it takes longer for him to get what you are asking. That, and it’s terribly disrespectful to your dog. He is your companion, not your slave or prisoner of war. You do not own his body. Another reason I like to avoid it, is that it can be painful or unealthy for the dog. I can’t imagine you are doing wonders to your pup’s orthopedic fine-tuning by force-sitting him several times per day.
Reinforcement
This term comes from the operant quadrants.
‘Reinforcement’ means anything that follows a dog’s behaviour that will make him perform the behaviour more often. Put another way: your dog does something -> Something nice happens or something not nice stops happening -> the dog is keen on repeating the behaviour. Both examples of something nice happening and something not nice stopping are examples of ‘Reinforcement’.
- Something nice is happening: You ask your dog to sit -> he does -> you give him a treat for his trouble. As a result, the dog is learning that sit is followed by something cool, so, the theory goes, he will then sit more often.
- Something annoying is stopping: Reinforcement does not mean that the dog is enjoying something. It means the dog will repeat his behaviour more often as a result of it. You can get there by asking the dog to sit, then pressing on his back with your knee (don’t!), only releasing once he sits. Once he has learnt the connection between sitting and your knee going away, he will sit sooner after hearing the command (and, added bonus, you could enter a contest for medieval dog training methods).
When it comes to 1 (reinforcing with something pleasant), people assume some things are rewarding to the dog when they aren’t (e.g. kissing, hugging, tapping on the head are kind of annoying to most dogs and, in some situation, even food can be annoying). Always look at the dog, and be creative to find the thing that really gets his juices going. Not what you assume will.
Schedules of reinforcement
A schedule of reinforcement is a ratio defining how many good performances (by the dog) it will take to get a treat.
When teaching the dog a new, he will be on a continuous-ratio (1:1 ratio of reinforcement, every performance gets reinforced). As he gets better, you want to space this out a little. After all, your dog, when he masters the sit, no longer gets a treat each time, right? You want him on an an intermittent-ratio (as in, not every time any more).
Here are some intermittent-reinforcement pitfalls to avoid:
- If you reinforce a well-mastered every single time (1:1, continuous), the behaviour will all fall apart the second you don’t have your treats handy. The dog has learnt to blackmail you.
- Imagine you are reinforcing the behaviour every 3 times he performs the behaviour (fixed-ratio for nerds). Religiously so. You’ve gone and made the ratio too predictable. The dog will work out when he’s ‘due’ a treat and will skim on the interim performances.
- If your ratio is too stretched too soon (e.g. abruptly going from 1:1 to 1:25), the dog may give up and stop trying.
Variable-ratios
For a well-mastered skill to stick, you want a variable-ratio schedule. You are going to reinforce the behaviour once every few good performances, but you’ll make this unpredictable. And so you’ll gradually stretch your average from 1:1 to 1:10. This is how gamblers get addicted to the slot machines. Once in a while, the dog doesn’t know when, they’ll get rewarded.
An variable-ratio can be a Godsend to make good habit stick, but it is also what makes bad habits stick. It’s the last ratio you want if you need your dog to drop a habit. Imagine you want your dog to stop barking at you to demand you throw the ball. You are determined to wait for 2-3 seconds of quiet before you throw the ball from now on. Only sometimes he drives you insane with the barking, so you give in. Congratulations: you have put demand barking on an intermittent ratio of reinforcement. Your dog knows that it pays to be perseverent and bark for longer bouts.
Proofing
Making a skill ‘life-proof’, honing and practicing it in gradually more life-like conditions to make sure the dog doesn’t break down when applying the skill in a real-life situation (i.e. with flesh and blood distractions, sometimes with you at a distance, ) doesn’t break down in real-life conditions. Proofing is key to working dog training. Imagine a guide dog who breaks character if he smells another dog? Imagine a bomb dog who gets distracted by the smell of sandwiches at the airport? Proofing is also essential for your average family dog, so he doesn’t just listen at the dog graining school, but also in real-life situations.
Like everything in effective dog training, proofing is done gradually and systematically. You are going to GRADUALLY raise your criteria for Distraction, Duration and Distance (we call these the 3 D’s), so that, yes, the dog will stay sitting if you ask from afar and even if there are other dogs present.
The dog training stages
So, I’ve kept you waiting long enough. I am using the example of teaching a dog to look you in the eyes, but you can apply this to most anything.
Stage 0 – Luring the dog into the right behaviour
Introduction
This is the first stage. Right now, it’s not about him understanding the word yet, just getting him to perform the behaviour. So we’re not even going to use the command.
Try to keep this stage as short as possible, or it’ll be really tricky to fade out the props and help.
Instructions
1. Start session: Tell the dog he’s in training. Say ‘start session’ or something similar.
2. Out of start position: Place the dog into a position that is NOT the target behaviour. e.g. for Watch me, use one hand to get the dog to look somewhere that is NOT your eyes. e.g. ruffle up the packaging of a bag of treats and get him to focus on that, or you get it to find a treat on the floor.
3. Lure: Lure him into the target behaviour. e.g. Bring one hand (i.e. the ‘lure hand’) into his field of vision, then bring it to your eye level so that he ends up looking you in the eyes. Make your eyes soft, hey, dogs aren’t exactly fans of staring contests.
4. Praise: As soon as the dog does what you want (e.g. makes eye contact), praise with a chirpy – but not hysterically enthusiastic – voice
5. Reward: ONLY THEN, reward the dog with a treat (from your lure hand). Do this immediately, so he makes the connection.
In the case of ‘watch me’, give the treat following such a trajectory from your eyes to his mouth that he keeps looking you in the eyes as long as possible.
6. Finish: Microseconds before he can’t hold on any more and you feel he’s about to give up (e.g. stop looking at you), say ‘finish’, then if he doesn’t spontaneously stop, lure him away from the behaviour.
Some notes on this:
- ‘watch me’ really means ‘keep watching me’, just like ‘sit’ really means ‘keep sitting’. So it’s important that he doesn’t cease the behaviour before you’ve given the ‘finish’ permission.
- You should be able to hold on for a second or two, tops, before ‘finishing’ at this early stage.
- Remember, keep your eyes soft. Blink like a cat if it helps you remember not to stare.
7. Do it again: Start again about 5 times or so during one session.
8. End session: After a few times, tell the dog the session is over
Say ‘end session’ and casually walk away. Remember: no petting. That means no hugging, tapping on the head, kissing,… You get the gist.
Criterion for moving up
We need to focus on weaning the dog from the lure stage asap. Lots of dogs get hooked at this stage. You know he’s ready to move up when he performs over 90% of the time in at least 3x consecutive sessions.
Stage 1 – Adding a command
Introduction
Now, we’re going to add a ‘cue’ (i.e. a command).
‘Why didn’t we use the cue from the start?, I hear you ask. That’s because if we do, he won’t be immediately successful (unless he is fluent in English, which most, well, aren’t). That means that he’ll learn that the word is either very confusing or totally irrelevant.
‘So why does it matter? I’ll just make up a new cue’ Cues are a precious commodity in dog training, so we have to keep them pristine, error-free – or we have to keep coming up with unspoilt ones. Personally, there’s only so many words I can come up with for ‘watch me’. That, and soon what you want will get really muddled as he’ll just get really confused.
Instructions
1. Start session: Tell the dog he’s in training: ‘start session’
2. Out of position: Get the dog in a position where he is not already performing the target behaviour.
3. Introduce command: Use the command. Remember, come up with something good. In our example: ‘Watch me’.
Important: Do not ‘bark the order’, make your tone of voice is enticing. You are not bossing him about, you are promising him a piece of cheese
4. Lure: Immediately after you said the command, lure the dog into position (like you did previously). It has to be immediate so he makes the connection between the command and the end position. The quicker the better.
5. Praise: As soon as the dog complies, praise him (voice only: no treat, no petting)
6. Treat: Immediately after the praise, reward him a treat.
In this exercise (watch me), the treat is also in your lure hand.
Remember also: soft eyes when you deliver the treat.
7. Finish: Just before he looks like he might tire of the target behaviour (e.g. looks like he might stop looking at you), say ‘finish’
If he doesn’t spontaneously stop, lure him away from the behaviour.
8. And again: Start again about 5 times or so during this session
9. Close session: Say ‘finish session’. Go through about 5x sessions a day, more if you feel like it. Definitely not more often that is fun for you and the dog.
Criteria for moving up
Keep practising the above, increasing the length of time between:
- cue -> lure (‘watch me’ -> bringing the treat to your eyes), and
- praise -> reward (i.e. ‘Good boy’ -> treat).
Also, play around with gradually making the lure less obvious, so that you get a feel for how close you are to being able to do away with it.
This means be increasingly discrete about the fact that you are carrying a treat in your ‘lure hand’, so that the lure doesn’t come to overshadow the command.
Remember, do not let the dog fail when trying to cover new grounds (i.e. increasing the two time intervals + fading out the lure). If performance is shaky, go back a stage/drop some criteria.
If:
- the dog can hold on for 5 seconds or so between cue and lure; and if
- the dog can hold on for 5 seconds or so between praise and reward; and if
- he occasionally offers the behaviour prior with minimal luring…
… attempt moving on to the next stage.
Stage 2 – Fading out the lure
Introduction
We are now going to focus on eliminating the lure altogether. Lures have a nasty habit of hanging around too long, and it makes for sloppy dog training. It’s sloppy because the dog ends up being completely dependent on our clues, and actually doesn’t really get the command.
Instructions
1. Open session
2. Out of target position: Get the dog out of target position if relevant.
3. Command: Use the command (e.g. ‘watch me’)
4. Praise and bring empty lure hand into position: Once the dog complies, praise him.
Keep praising for the duration of the target behaviour – e.g. As long as he looks you in the eyes.
Whilst praising, bring your lure hand into position, and this time, do so with an empty hand.
5. Reward: Reward the dog with a treat from your other, non-lure, hand.
What we’re doing here is making the lure hand irrelevant. It would be too conspicuous to remove it from the sequence suddenly.
Try to keep him in position while rewarding (e.g. by tapping your forehead with your empty lure hand), to compensate for the temptation he’ll have to look at the non-lure hand (which now has the reward).
The beauty of this is that the dog will be starting to think and consciously decide to perform a behaviour, rather than automatically following your lure.
6. Finish: JUST BEFORE you feel he is about to stop performing the target behaviour (e.g. looking at you), say ‘finish’.
7. Repeat: You can start again about 5 times or so. Repeat this for about 5 sessions a day (closing each session with ‘end session’).
Criteria for moving up
Keep practicing until 1/ the lure hand has been completed faded away, and 2/ you can reward him for success from whatever your hand position.
Remember, do not let the dog fail whenever pushing for a higher boundary. If performance isn’t all that, go back a stage/drop a criterion. Do not push on through and get frustrated.
Stage 3 – Focusing on building duration
Introduction
Now that we’ve got rid of the lure, and the dog performs on your voice command alone, it’s time to build up the duration. A ‘watch me’ of a split second is impressive, sure, but we’re going for a dog that will not lift his gaze until released, even if that took hours (OK, a couple of minutes).
Instructions
1. Out of position: Get the dog out of target position if necessary.
2. Cue: Use the command e.g. ‘watch me’. Remember, now we DO NOT use the lure at all. No more hand to eyes for the watch me exercise. He should just do it on your voice command alone.
3. Praise: Once the dog complies (should be instantaneous after hearing the cue), continuously praise him with your voice for the whole duration of his performance.
4. Reward: Start gradually increasing the period of time between praise and treat.
Remember, the reward should be delivered from any hand position, and on-going delivery should not interrupt the behaviour.
5. Finish: You’re also going to extend the length of time between getting the treat, and you saying ‘finish’. This means you could have to give repeat treats to keep his interest alive.
This is useful because we don’t want him to think that receiving a treat means he gets to stop the behaviour.
The best timing to announce the finish is just before he’s about to give up (e.g. stop looking at you), then just say ‘finish’.
At the beginning, it’ll be a microsecond, but you’ll gradually build on that.
Note: The signs that he’s about to give up should be subtle precursors, not a full blown panic. The last thing we want is to wait until the dog whines to get him a treat.
6. Repeat: Repeat this about 5 times a day, more if you have time. Don’t forget to open and end sessions.
Criteria for moving up
Keep working on these two durations until:
- You’ve reached your target duration (i.e. close to what you’ll want in a real-life situation).
- He does not interrupt his behaviour even when getting a treat.
And now, for the infamous ‘going intermittent’. This is where things get hairy. Brace yourself.
Stage 4 – Going intermittent
Introduction
Up to now, the dog got rewarded every single time he performed well – and he should only have performed well as you read my intro (right?). This is called a ‘continuous ratio of reinforcement’. We want to move on from this as soon as we can.
This transition can be a little delicate, as you’ll be walking a fine line between two extremes:
- Going ‘cold turkey‘: One day he gets a treat every time, next day every 10 times. He’ll fall apart, and soon conclude that this trick has stopped paying off.
- Taking the path of least resistance: It’s tempting, but don’t keep giving a treat each time forever. You might, just might, not have treats on a day when it matters that he listens. That, and the fact that it’s akin to the dog blackmailing you.
Intructions
Do exactly the same you’ve been doing, except now, one time out of two, simply say ‘good goy’, then ‘finish’, and start again.
At the beginning, try to alternate between 1:1 and 1:2, then alternate between 1:1 and 1:3, then 1:2 and 1:3, and so on until you’ve reached a ratio that is workable for you.
Stage 5 – Proofing the response
Introduction
So, now your dog:
- performs the behaviour on command (no luring),
- doesn’t need a reward every single time , AND
- can keep doing it until released for a respectable amount of time.
‘Surely we’re done, right?’. Well… Not quite. If you don’t polish this off in more and more life-like contexts, you’ll have yourself a ‘star in the kitchen, lemon on the field’.
Instructions
Take the sessions in different rooms in the house, then in the garden, then try outside, then on the street, then in the dog park. Whenever you see signs of hesitation, go back as many stages as you need to to get back to a stellar performance.
Then take different people through whole hierarchy again (all stages, then all locations, in order of difficulty).
Then try with another (familiar) dog in the vicinity, or whatever distraction he’ll need to be ready for in real life.
When can you ‘go live’?
You are ‘live’ when you can use the command for real-life situations. e.g. Another dog is approaching, your dog is staring at him, you say ‘watch me’ and he looks at you continuously as you U-turn away.
If you’ve gone through all the stages successfully yet the dog falls apart in a real-life situation, break it down to work out what factor is different than during rehearsal, and then gradually add this factor to your proofing stage.
Conclusion
Phew, you did it. And do you know what, you did better than that. You went and understood one of the most central principles of dog training. If you truly understand this, you can teach your dog anything short of flying.
Remember now: things tend to get hairy around
- adding the cue,
- going intermittent, and
- proofing for distractions.
Don’t skip corners and you’ll be getting very impressive results indeed.
Some last advice
About sessions:
- Keep sessions short and frequent (five times a day or more),
- End each session on a good note, even if it means asking him something he knows full well how to do.
About using food:
- Obviously, take account of food sensitivities. Adding (organic) oil to his normal kibble can work wonders to make the food more interesting, even for allergic dogs.
- Plan a ration in the morning for the whole day, and pick from there. If you train a lot and don’t do that, you’ll end up with an obese pet on your hands.
- Chop the treats into tiny pieces, so you’ll have more reward instances for the same calorific buck. Play around to see how far down in size you can go before it starts dimming the response
Avoiding intimidation/discomfort:
- Make sure your dog is not cornered for the exercise, that he has plenty of escape routes;
- Design the exercises to minimize having to tower above/bend over him;
- Do not pet him to ‘reward’ him. Right now, the treat is the prize, anything else is, well, a disappointment. Sure Fido loves you but, you know, ‘Not now’
If he’s not progressing despite the above, try to think:
- Did he really master the previous stage? Go back.
- Is he distracted? Proof
- Is he enjoying himself enough? Relax and enjoy the exercise yourself
- Is he a little intimidated/scared? Check this section for some useful tips on body language to pick up on this.
- Is he motivated enough? Crank it up a little – chopped hot-dog, cheese…
- Are you suuuure you ‘re not petting him when he does well?
- Do double-, nay, triple-check that there is no physical reason. It could be localised pain (e.g. rheumatism, ear infection) or a sensory deficiency (i.e. sight/hearing)
If you’re really not getting anywhere and the behaviour is important enough, contact a private dog trainer in your region.
Comments
I love to read your comments, so leave me your thoughts any time. I’d particularly like to hear from you if you:
- Have any questions on shaping/fading/reinforcing/etc.
- Have been advised to approach this in a different way, and what your results/impressions are;
- Have a success story, or on the contrary, on using these methods;
- Are hitting a snag somewhere along the way.
Further references
Luring vs. prompting:
6 Comments
Oh my goodness. I think that was four blog posts in one! My head is spinning. =P
I had a few “trainer moments” while reading this – meaning I went “but that’s not how I do it!” – so I had to laugh. (We all do things fractionally different, but the emphasis here is on the word fractionally. And yet, those tiny things are what always stand out to us. So, I thought that, in the interest of good humor and sharing, I’d share them.)
1.) I used to do random treating until I went to an Ian Dunbar seminar a few years ago, and he reminded me of DRE – differential reinforcement of excellent behavior. So, when my dog knows a cue, I only reward those behaviors that I would consider “the best” or “the most accurate” or “the fastest”; something of that nature.
The hang-up here is that… if the dog becomes perfect – nails the behavior each time with quick precision, you’d have to move to random treating, wouldn’t you? =P
2.) I love your definition for shaping, but I have to admit that if I didn’t have prior knowledge of learning theory…. Well, my head would have exploded.
3.) You try to stay away from using everyday words – words one might hear in conversation like “no” or “okay”. I know plenty of trainers who recommend much the same, and I’ve heard some clients say they’ve had similar troubles…. but I remain bias because I never have. I actually proof my dogs to “know” when I am or am not talking to them, and also not to respond to words that sound like, but are not, cues. (Especially on the release-from-cue-word, which is “okay”.)
4.)You put luring and physical prompting in the same sentence. o_____o
5.) You don’t give the dog the treat when you lure? Or did you meant that the dog should be given the treat during the actual luring (ie: wait until the dog is in the proper position before he gets the food reward)? It sounded almost as if the dog never got the food at all!
6.) I use Koda’s kibble for her treats. The only time I use something special – store bought treats from a good company (no fillers or gross stuff that will cause gastrointestinal distress) or hot dogs, cheese, chicken, etc – is when the “pressure is on”. Pressure, I guess, would be the vet’s office, a new training club, a class where she is learning something new and difficult, or a training session where I know she will be particularly frustrated.
**I do this because I’m weird. I just wanted to mention it because it’s a different thought. My thinking is that I want the special treats to stay special. Also, I’m bias against fat dogs, and most of the time, the only dogs I run into that won’t work for regular kibble won’t do so because they are overfed and couldn’t care less about food unless it’s string cheese.
I have preconceptions. haha.
Anyway, I loved the article…even if it was a long read. =P
I think you did a great job being concise and to the point (you just made many points!). And your points are well thought out and easy to follow. It all makes sense. I found myself nodding my head most of the time like “yep; it’s all true”. LoL
Glad you’re back to blogging. =] It’s been a while!
And glad you’re back to commenting! I love that you took the time to read it all AND comment.
1/ Yup, the DRE and going intermittent are tough ones to manoeuvre/distinguish. DRE is pure shaping, and I’d use that for more complex behaviours. These stages I set out I would consider the most basic training steps. I wouldn’t advise teaching anything without these steps. For behaviour that requires more subtle shaping, some trainers use DRE. Going from DRE to an intermittent ratio (once you’ve shaped it to your target performance) is every bit as tricky – if not more – than going intermittent from continuous as, if your DRE was effective, you’d really be giving the dog feedback that he has to try harder, which isn’t always appropriate. I only really use DRE’s to mark stellar responses on target behaviour that’s still shaky, so if am on continuous reinforcement ratio, I’d give more (or a better) treat(s) on that star response, but every response is still rewarded.
2/ Definition of shaping. I am so bad at speaking mortal. I swear my kid says words no two-year-old should know. My – think – like – a computer, not good, not good.
3/ Gosh, the ‘okay’ is also one of my words. You’ve proofed your dogs to distinguish between command and everyday? Tips, young lady, tips.
4/ Luring/physical prompting. Oooops, yup, confusing. Changed the sentence and added a vid in the reference section.
5/ ‘You don’t give the dog the treat when you lure?’ When I fade out the lure, sometimes I give him the treat from the non lure hand (definitely in the watch me exercise)
6/ Good point about using the kibble. I actually use kibble for standard stuff with my dog, but, like you say, when something is tricky or needs perfect performance, I go up a notch. I guess I had owners, rather than trainers in mind. Using the better stuff, at least you don’t have to worry about motivation being an issue quite so much. Oh, and I forgot to add one point (if that’s even possible on such a lengthy post!): you can go too yummy, too appetitive. To the point that the dog wants the treat so hard that he’s distracted from performing.
Yup, am definitely too lengthy by a mile, but I’ve decided to follow my own rules and screw blog popularity. As I use these articles to support theory points I make during client appointments, I really want them to get all the important details. I give my clients so much homework sometimes I think I wouldn’t like to be them ; P But seriously, though, my entire model revolves around teaching the clients to be independent from me as quickly as possible, so it takes some hefty knowledge transfer.
Annnnd, wouldn’t you know it, I went and got lengthy on your proverbial again ; P
Wow that was unusual. I just wrote an incredibly long comment but after I clicked submit my comment didn’t appear. Grrrr… well I’m not writing all that over again. Anyways, just wanted to say great blog!
Oh don’t you hate it when that happens? Thanks anyway!
There are various ways for dog training, but broadly it can be categorise in two types one is by punishment method, which is a worst kind of training method and another is by positive reinforcement in which trainer give reward to dog for every good response, but as mentioned above we should not pet a dog for rewarding him so we should always keep things in mind. As there is no age for learning new thing so it same in the case of dog so always try to teach something new to your dog and always keep session short and more interesting by gaming or anything which dog like the most.
Very good point, Michel. You don’t have to reward your dog with food. It could be with a tug-of-war game or a game of retrieve. Whatever the dog finds the most rewarding.
Keeping sessions short and interesting. I cannot say this often enough. Some people ‘overtrain’ their dogs with disastrous effects on their motivations. This is a common problem with working dogs (too much work and too little play makes Fido a very dull boy)…
One Trackback
[…] […]