Field Notes · An evolving learning library from Angier K9

Play, reward and food · Might beats will

Should I reward my dog every time?

Once your dog knows the job, reward it some of the time rather than every time - 'might get it' keeps him keener than 'definitely will' - but never let the reward disappear altogether.

The answer

No - and not never either. Four dogs' trainers say the same thing in different words. On a trail walk with Apollo, a dog who works for a ball or a tug, the trainer explains that he is now paying only some of the time, because 'might get it is more powerful than definitely will get it'. Kelton gives Penny's family the same rule with an analogy: paying someone every single time for the dishes makes the habit less stable; sometimes-yes, sometimes-no is 'the gambling', and the dog stays keen. At Penny's go-home he adds the other half: every command is a battery that must be recharged, by the handler or by the behaviour itself, or it runs down.

Connor states the same pair of rules for two dogs. To Stormie's family: the house does not rely on treats, because a treat-trained dog starts checking for the treat before it sits - so if treats are used, they must be random and sporadic - but you cannot train with pressure and nothing positive either, so find what makes her happy (for her, a rub and a sniff). To March Zeus's family: give the treat after the release word, about half the time, so the sit is never contingent on seeing it - but not nothing, or the training slows and gets spotty.

Stated on film: the rule and its reason. Kelton adds a brain explanation (dopamine peaks in anticipation, not on receipt); that is his gloss, not something the footage tests.

What to do

Reward after the release, not before; do it randomly - roughly half the time is the figure Connor gives Zeus's family - and never stop entirely (watch 08:48). If you use treats at all, keep them random so the dog does not start looking for one before obeying (watch 42:37).

What the trainers say

But when you start doing what's called an intermittent or a variable reinforcement schedule,

the trainer · Apollo (the Sharkie) · Watch at 28:36 ↗

So then he's really charged up to do it because he's like, oh, I might get it, and might get

the trainer · Apollo (the Sharkie) · Watch at 28:44 ↗

um but either way you don't stay on a reward them every time for everything forever because

Kelton · Penny · Watch at 40:28 ↗

portion um but then if like sometimes you reward them and sometimes you don't that's actually a

Kelton · Penny · Watch at 40:49 ↗

right so think about it this way it's like a battery all of your commands and all of your um

Kelton · Penny · Watch at 41:05 ↗

for good out there okay then you're fine um we want to make sure there's some kind of reward

Connor · Stormie · Watch at 42:04 ↗

sit for treats you know what i mean you get that kind of thing so for her you can no so you can use

Connor · Stormie · Watch at 42:46 ↗

treats for instance say okay and then after you say okay then give the treat and i wouldn't i

Connor · Zeus (March 2026) · Watch at 8:48 ↗

it's like maybe a 50 50 thing where he's like sometimes i get a treat sometimes i don't

Connor · Zeus (March 2026) · Watch at 8:59 ↗

How it looks, dog by dog

Apollo (the Sharkie)

A toy-motivated dog whose heel is now paid with a ball or tug only some of the time. The trainer points to how happy the heel entry looks and puts it down to the chance of a reward.

On film

You can see he's happy to do it because he knows that there's a chance I'm going to throw

Watch at 28:30 ↗

Penny

Given to the family twice by Kelton: mid-program as the dishes analogy and 'the gambling', and at go-home as the battery - behaviours have to be reinforcing somehow, by the handler or by themselves, or they run down.

On film

up i'm owed things now um and so that's not ideal that's the gambling though they might

Watch at 41:08 ↗

but eventually i put the battery to charge um behaviors are like that too for people and for

Watch at 41:38 ↗

Stormie

The house position on treats, given at the home lesson: the program is balanced, not all pressure, but treats risk a dog that checks for one first; if used, random and sporadic. Stormie's reward is a rub and a chance to sniff.

On film

because you really can't train without some kind of reward right so we do use like negative

Watch at 42:10 ↗

thing. So I use just like I rub out like a rub like when she does a good job I rub on her and I

Watch at 43:02 ↗

Zeus (March 2026)

The most concrete version: the treat comes after 'okay', about half the time, and not never - drilled commands that bring nothing slow down and get spotty.

On film

i do listen i get what i want and so that's really important is that if you're drilling these commands

Watch at 8:28 ↗

Why it works

Stated: an intermittent reward charges the dog up because 'might get it' is more powerful (watch 28:44); a reward every time makes the behaviour less stable, because the dog comes to depend on it (watch 40:28); treats every time teach the dog to check before complying (watch 42:37); but behaviours with no reinforcement run down like a battery (watch 41:05; watch 08:32). Kelton's dopamine-in-anticipation explanation (watch 41:14) is his gloss, not tested on film.

Common owner mistakes

  • Rewarding every rep forever.
    On film

    that makes the it actually makes the behavior less stable and strong like doing the dishes

    Watch at 40:35 ↗
  • Drilling commands with nothing ever in return.
    On film

    on him all the time and every time he's like i do it and then i don't get anything in return

    Watch at 8:32 ↗

Questions owners ask

How often should I give a treat?

Connor tells Zeus's family roughly half the time, at random, and after the release word so the sit is not contingent on seeing it.

On film

would say like do it kind of randomly because you don't want it to be like contingent on the treat

Watch at 8:52 ↗

Do your trainers use treats?

Connor tells Stormie's family the house does not rely on them, because a treat-trained dog checks for the treat first - but they can be used, randomly. Play, praise, a rub or a sniff are the usual rewards on film.

On film

treats but it needs to be very random sporadic and not like a every time or like every other time

Watch at 42:55 ↗

If I stop rewarding, will the training fade?

Kelton's battery: commands must be recharged somehow, by you or by being rewarding in themselves.

On film

dogs it's like you could have like a it has to be reinforcing somehow um whether it's self-reinforcing

Watch at 41:46 ↗

The lesson

The goal

Move from continuous to intermittent reward without letting it reach zero.

Learn first

The drill

As handed to Zeus's family: after the release 'okay', give the treat at random, about half the time (watch 08:48).

Watch out for

  • Reward every time forever
  • Reward never
  • Treat before the release

Check yourself

Why does a dog paid every single time become less reliable, not more?