Field Notes · An evolving learning library from Angier K9

The session: dose, venue and proofing · Raise the criterion in steps

How fast should I raise the bar for my dog?

Start with a rough version the dog can win, raise the bar only a little at a time, and when he keeps failing drop it until he genuinely succeeds - but never fake a success by stopping him just before he fails.

The answer

The trainers do not ask for the finished behaviour on day one. Baxter's trainer, teaching Tyson's recall on the owner's street, says it directly: the finished picture is the maximum criterion, and "i start with an unfinished picture and i teach that first." At first 'come' means come to my general area. The sit on arrival comes later, once coming back is a habit.

The same logic appears across ten dogs. Lucia's recall is new, so her trainer will not "have the criteria so high that she loses every time". He takes an imperfect rep as a step on the way, and holds one thing firm: a sit facing him with a second of eye contact. Oscar's trainer accepts a sloppy finish because "I just want to make sure I get him recalling perfectly and Then we can work on the finish after that." Drago's first down is "not exactly what I want it to look like at the end", but it is "an approximation of it", to be shaped over sessions and made happier. Dolly stalls on her first down, and her trainer lowers the criterion and marks a bend of the front legs. Mazi stares a second and a half too long after 'let's go'. His trainer says that will be held against him one day, "but like right now i'm not holding that to him at all", because he is new.

When a dog keeps failing, the bar comes down. At Whiskey's session a staff member who feels bad about not giving her any success is told the house rule: "If a dog's failing at it, just lower the criteria as much as you need to to get them to succeed." Then raise it only as much as makes sense. Josephine's trainer puts it in gym terms. A dog who can hold a stay at some maximum cannot do fifty maximum reps in twenty minutes. So once you find the failure point, drop a little below it, do spaced "working level sets" that succeed, and test the maximum again later; "their one rep max improves." Neo's handler adds that yesterday's best is not today's starting point: a new session opens lower and warms up.

There is one condition on all of this. The success has to be real. Whiskey's coach: "As long as you don't smoke and mirrors it, as long as they actually succeed." Stopping the dog just as she starts to fail and calling it good teaches nothing.

What is inferred: this is the early-stage face of a standard that tightens later. Elsewhere the same trainers refuse partial responses outright (see dont-accept-the-almost). The film does not state when one gives way to the other, beyond 'when he is new' and 'once the coming is reliable'.

What to do

Decide what the finished behaviour looks like, then start well short of it. When your dog fails twice, make it easier (shorter, closer, quieter) until he really succeeds. Then raise it a little. For duration and distance, practise a little below his best and test the best only now and then. Don't release him just before he breaks and call that a success.

What the trainers say

like the maximum criteria this is exactly what i want right the finished picture i start with an unfinished picture and i teach that first

Baxter's trainer (not named on film) · Baxter · Watch at 29:00 ↗

So what I wanted to give her there is, one, it's a successive approximation, so I'm not forcing her to do, you know, like a straight to me recall right away

Lucia's trainer (not named on film) · Lucia · Watch at 3:51 ↗

And so, because her focus is so all over the place, the only thing that I am doing that's a little bit extra is that I am requiring that she finish correctly, which is that she sits and i want like one second of eye contact

Lucia's trainer (not named on film) · Lucia · Watch at 5:23 ↗

I'm not being super specific about the finish because I just want to make sure I get him recalling perfectly and Then we can work on the finish after that. We don't have to get a perfect rep yet

Oscar's trainer (not named on film) · Oscar · Watch at 8:31 ↗

So we'll do successive approximations with that, and we'll make it look a little happier, make it look nicer, improve the emotion around it, because obviously he didn't like that very much.

Drago's trainer (not named on film) · Drago · Watch at 4:07 ↗

her to succeed and make approximations so I'm gonna do this see her front legs I'm gonna just

Dolly's trainer (not named on film) · Dolly · Watch at 39:52 ↗

then he kept it rolling with me and that's what i want he's uh he's fairly new when it comes to

Mazi's trainer (not named on film) · Mazi · Watch at 11:52 ↗

Yeah, lower the criteria. Get her a couple wins, and then slowly raise the criteria.

a staff member coaching Whiskey's handler (not named on film) · Whiskey · Watch at 15:56 ↗

That's kind of the rule. If a dog's failing at it, just lower the criteria as much as you need to to get them to succeed.

a staff member coaching Whiskey's handler (not named on film) · Whiskey · Watch at 16:02 ↗

As long as you don't smoke and mirrors it, as long as they actually succeed.

a staff member coaching Whiskey's handler (not named on film) · Whiskey · Watch at 16:25 ↗

but not by maxing it out every time and trying to push it push it push it push it further

Josephine's trainer (not named on film) · Josephine · Watch at 7:50 ↗

So, that's never a big deal because all we had to do is lower the criteria slightly.

Josephine's handler (not named on film) · Josephine · Watch at 5:16 ↗

you're probably you know let's say you trained yesterday right your peak is your the start of your training the next day is not going to be equivalent to the peak of your training a day before

Neo's handler (not named on film) · Neo · Watch at 2:53 ↗

How it looks, dog by dog

Baxter

At a private lesson on the owner's street, Tyson (Baxter's housemate) is taught the recall. The trainer names the finished picture as the maximum criterion and says he teaches an unfinished picture first. For now 'come' means come to my general area; the sit on arrival is added once returning is a habit. Reps are spaced because pressure is stressful.

On film

like the maximum criteria this is exactly what i want right the finished picture i start with an unfinished picture and i teach that first

Watch at 29:00 ↗

pressure on you it's a little bit stressful right so because it's a little bit stressful

Watch at 27:36 ↗

Lucia

Lucia's recall is brand new. Her trainer accepts an imperfect rep as a successive approximation rather than forcing a straight-to-me recall, so she does not lose every time. He holds one criterion firmly because her focus is all over the place: the finish is a sit facing him with a second of eye contact, so she learns that engaging with him is how she gets to go again.

On film

So what I wanted to give her there is, one, it's a successive approximation, so I'm not forcing her to do, you know, like a straight to me recall right away

Watch at 3:51 ↗

she's just starting to learn the recall so i'm not going to have the criteria so high that she loses every time and gets huge punishments or not huge punishments but that i have to force too much and you can see she's giving to the end of the leash a lot better than she was at first

Watch at 4:36 ↗

and not facing away from me because that's very common with dogs who get really like amped up is that they will face away from me um so i want her to face me and that way i can teach her that engaging with me is how she gets to go again

Watch at 5:43 ↗

Oscar

In an evening park session, Oscar's trainer accepts a recall without a precise finish. He wants the coming itself perfect first; the finish can wait. In the next part, when the finish is worked, the pressure into the sit is aimed at making the sit more likely on the next rep, not at rescuing this one.

On film

I'm not being super specific about the finish because I just want to make sure I get him recalling perfectly and Then we can work on the finish after that. We don't have to get a perfect rep yet

Watch at 8:31 ↗

i'm negatively reinforcing or making it uh more likely on the next rep for him to to sit

Watch at 11:08 ↗

Drago

On Drago's first down the trainer catches himself using the word too early, since Drago has never done it. He stops. Drago tries different things and lands in a position that is not the final picture but an approximation of it. The plan is successive approximations, with the emotion part of the target: make it look happier, because he clearly did not like it.

On film

actually i shouldn't be using the word down here i guess he's never done this before

Watch at 3:13 ↗

but it's an approximation of it.

Watch at 4:02 ↗

So we'll do successive approximations with that, and we'll make it look a little happier, make it look nicer, improve the emotion around it, because obviously he didn't like that very much.

Watch at 4:07 ↗

Dolly

Dolly is a nervous dog, and pressure is new to her. When her first down stalls during a private lesson, her trainer does not press harder. He lowers the criterion a little to help her succeed and marks the bend of her front legs toward a sphinx down. He suggests the owners try it mid-play, or with a food lure.

On film

I'm gonna lower the criteria I'm gonna lower the criteria just a little bit and make it easier for

Watch at 39:41 ↗

her to succeed and make approximations so I'm gonna do this see her front legs I'm gonna just

Watch at 39:52 ↗

Mazi

Walking Mazi through a park full of people, the trainer lets a second-and-a-half stare after 'let's go' pass. Mazi is fairly new, so a perfect response is not expected yet. At some point it will be; right now it is not held against him.

On film

after i said let's go he stared for like like like you saw maybe like a second and a half second and

Watch at 11:48 ↗

go good job um at some point would that be the case yes but like right now i'm not holding that

Watch at 12:18 ↗

Whiskey

A handler working Whiskey on an implied stay off the cot says she feels bad about not being able to give Whiskey any success. A colleague gives the rule on camera: lower the criteria as much as you need for her to succeed, get a couple of wins, then raise it slowly. The one condition is that the success must be real. Cutting the dog off as she starts to fail and calling it good is 'smoke and mirrors'.

On film

I know. I just feel bad because I'm not able to give her any success.

Watch at 15:46 ↗

That's kind of the rule. If a dog's failing at it, just lower the criteria as much as you need to to get them to succeed.

Watch at 16:02 ↗

Like you see them starting to fail, so you quickly say, okay, don't do that.

Watch at 16:31 ↗

Josephine

Josephine breaks her first few stays at distance. Her handler lowers the criteria slightly ('never a big deal'). Later a trainer explains it with the one-rep-max analogy. Once you find where she fails, drop the duration a little at the same distance so she succeeds, do spaced working-level reps through a long walk, and test the maximum later. It will have gone up. Maxing out every time does not get you there.

On film

So, that's never a big deal because all we had to do is lower the criteria slightly.

Watch at 5:16 ↗

shorter um on the duration same distance but a little bit less duration so that she was successful

Watch at 9:13 ↗

but not by maxing it out every time and trying to push it push it push it push it further

Watch at 7:50 ↗

they were at before and you should see that their max so to speak their one rep max improves

Watch at 7:25 ↗

Neo

The day after Neo first downed on the word alone, his handler goes back to the same place and does not expect perfection. The next session starts below the previous day's peak and has to warm back up. He also finds Neo downs readily on grass but not on gravel, so he builds on the easy surface and lets it carry over.

On film

you're probably you know let's say you trained yesterday right your peak is your the start of your training the next day is not going to be equivalent to the peak of your training a day before

Watch at 2:53 ↗

but once i proofed it enough in there then it started to carry over into the gravel

Watch at 9:41 ↗

Why it works

As the trainers explain it: a criterion the dog cannot meet produces repeated failure and repeated pressure, so the dog 'loses every time' and has to be forced (Lucia); successes, even small ones, are what the dog learns from, so the bar is set where he can win and raised from there (Whiskey's coach, Baxter's trainer). Spaced sub-maximal repetitions raise the maximum, as working sets do in a gym, where repeated maximal attempts do not (Josephine). Performance fluctuates between sessions, so each session is started below the last peak (Neo). The honesty condition is what keeps it from collapsing into bribery: a rep cut off before it fails is not a success and teaches nothing (Whiskey).

On film

she's just starting to learn the recall so i'm not going to have the criteria so high that she loses every time and gets huge punishments or not huge punishments but that i have to force too much and you can see she's giving to the end of the leash a lot better than she was at first

Watch at 4:36 ↗

instead you're doing working level sets working level repetitions and then by doing that they

Watch at 7:56 ↗

As long as you don't smoke and mirrors it, as long as they actually succeed.

Watch at 16:25 ↗

Where the trainers differ

Accepting a partial response

One view

Start with an unfinished picture and raise it in steps; accept approximations while the behaviour is new

On film

like the maximum criteria this is exactly what i want right the finished picture i start with an unfinished picture and i teach that first

Watch at 29:00 ↗

Another view

What you accept is what you will be offered next time; the partial is re-asked, not paid (concept dont-accept-the-almost)

On film

if you want it to like look correct and you want him to do it right and come all the way to you, you can't accept less.

Watch at 42:42 ↗

How to read it

The map's conflict. Laid out, not resolved. The film suggests the difference is stage (new dog, new behaviour) and whether the command was formally given, but no trainer puts the two positions side by side.

Which part of the recall to hold firm first

One view

Loose on the approach, firm on the finish

On film

And so, because her focus is so all over the place, the only thing that I am doing that's a little bit extra is that I am requiring that she finish correctly, which is that she sits and i want like one second of eye contact

Watch at 5:23 ↗

Another view

Firm on the coming, finish later

On film

I'm not being super specific about the finish because I just want to make sure I get him recalling perfectly and Then we can work on the finish after that. We don't have to get a perfect rep yet

Watch at 8:31 ↗

How to read it

Lucia's trainer loosens the approach and holds the finish (sit and eye contact); Oscar's trainer holds the coming and lets the finish go. Different dogs, different reasons (Lucia's scattered focus), not a stated disagreement.

Common owner mistakes

  • Pushing for the maximum every time - longest stay, furthest distance - instead of working just below it.
    On film

    but not by maxing it out every time and trying to push it push it push it push it further

    Watch at 7:50 ↗
  • Faking a success by releasing the dog just as he starts to break.
    On film

    Like you see them starting to fail, so you quickly say, okay, don't do that.

    Watch at 16:31 ↗
  • Expecting today's session to start where yesterday's best ended.
    On film

    you're probably you know let's say you trained yesterday right your peak is your the start of your training the next day is not going to be equivalent to the peak of your training a day before

    Watch at 2:53 ↗
  • Naming a behaviour before the dog can do it.
    On film

    actually i shouldn't be using the word down here i guess he's never done this before

    Watch at 3:13 ↗

Questions owners ask

Should I accept a sloppy sit or recall?

Early on, the trainers often do, on purpose. Baxter's trainer starts with an unfinished picture; Oscar's accepts a rough finish until the coming is reliable. Later the standard tightens and the partial is no longer paid.

On film

like the maximum criteria this is exactly what i want right the finished picture i start with an unfinished picture and i teach that first

Watch at 29:00 ↗

I'm not being super specific about the finish because I just want to make sure I get him recalling perfectly and Then we can work on the finish after that. We don't have to get a perfect rep yet

Watch at 8:31 ↗

My dog keeps failing. What do I do?

Make it easier until he genuinely succeeds - shorter, closer, quieter - then raise it a little at a time. The one thing not to do is cut the rep short just before he fails and call it good.

On film

That's kind of the rule. If a dog's failing at it, just lower the criteria as much as you need to to get them to succeed.

Watch at 16:02 ↗

As long as you don't smoke and mirrors it, as long as they actually succeed.

Watch at 16:25 ↗

How do I build a longer stay?

Josephine's trainer uses the gym: find where she fails, then do several spaced, successful stays a little below that on a long walk, and only occasionally test the maximum. It will have gone up.

On film

instead you're doing working level sets working level repetitions and then by doing that they

Watch at 7:56 ↗

He did it perfectly yesterday and badly today. Is he going backwards?

Neo's handler expects it: the next day starts below the previous day's peak and warms back up.

On film

you're probably you know let's say you trained yesterday right your peak is your the start of your training the next day is not going to be equivalent to the peak of your training a day before

Watch at 2:53 ↗

The lesson

The goal

The owner can set a starting criterion below the finished picture, lower it on repeated failure, raise it in small steps, and tell a real success from a faked one.

Learn first

The drill

[object Object]

Watch out for

  • Testing the maximum every rep.
  • Cutting a rep short to avoid a failure.
  • Raising two things at once (duration and distance).
  • Expecting yesterday's peak at the start of today.

Check yourself

Your dog broke a 60-second stay three times in a row. What do you ask for next?