Teaching Techniques

Reinforcing a behavior your dog offers entirely on their own, with no prompt, no lure and no cue. You watch, you mark the moment it happens, and you pay it. Your dog notices it worked, and does it more.

It is the right tool for behaviors that are difficult or impossible to prompt: a yawn, a stretch, a shake off, actually relaxing. You cannot lure a dog into relaxation, so for that whole family of behaviors capturing is usually the easiest place to start. It is not the only route: relaxation can also be shaped, by paying the small pieces of it like a hip rocking over or a head going down on a mat, and built through classical conditioning, by pairing a place or a sound with calm moments.

It is a poor first choice for anything you could simply lure. Waiting a fortnight for a dog to offer a down that a lure would produce in a minute is not kinder, it is only slower, and slow is what makes owners give up.

It asks something unusual of you: patience, and a willingness to sit there doing nothing with a marker ready. It also asks you to arrange the environment well, because you can only capture what your dog offers. A dog who never settles in your kitchen will not offer a settle in your kitchen, and the fix is a comfortable spot, a boring moment, and a full exercise quota, not a better marker.

In Push Drop Stick

Capturing steps have no fail button, because a behavior your dog volunteered cannot be wrong. If nothing is happening to mark, that is information about the setup rather than about your dog.

Often confused with

Shaping. Capturing waits for the finished behavior and pays it. Shaping pays a rough version and builds it.

Building a behavior your dog has never done by reinforcing rougher versions of it and gradually raising the bar, so that each closer version is paid and the older, rougher ones stop being paid. Also called successive approximation, and the paying-only-the-closer-ones part is differential reinforcement.

You cannot pay a behavior that never happens, so you pay whatever is already pointing in the right direction. Teaching a chin rest, you might start by paying a glance at your palm, then a nose that moves toward it, then a touch, then a touch that lingers, then the chin actually lowering into it. None of those is the goal. Each one is close enough to the last that your dog can find it.

Two things make the difference between shaping and frustration:

Change one thing at a time. If you are raising duration, do not also move your hand.

Raise the bar in small steps. Each increase should be small enough that your dog is still getting paid nearly as often as before. If the success rate drops sharply the moment you raise the bar, the step was too big, and the fix is a smaller one. While your dog is finding a new detail, it is fine to relax the old ones for a moment.

If your dog stops offering anything at all, you raised the bar too fast. Go back to the last version they were confident with, pay a few easy ones, and make the next increase smaller. That is not a setback, it is the method working.

In Push Drop Stick

Every plan in the app runs on the logic of shaping, whatever technique each step uses: small criterion steps written down in advance, with the bar-raising decided by data instead of by feel. When your dog needs a smaller increment than the plan has, that is what a split step is.

Often confused with

The app's Shaping step type, which is a narrower thing: timed rules for tracking two criteria at once, used mostly when adding a verbal cue.

A treat held at your dog's nose and moved to guide them into a position or through a movement.

Lures are excellent and slightly dangerous, and both for the same reason: they work instantly. A dog who has never sat will follow food up and back into a sit on the first try, which gets you a behavior to pay in seconds.

The danger is that the food is doing the work. Your dog is following a treat, not responding to a cue, and if the treat never leaves the picture then your dog has learned "hand with visible food moving up and back" as the cue. Take the food away and the behavior is not there, because the cue you trained is no longer complete.

So lures get faded early. The standard route in this app: lure with food for one set of five reps, and if your dog is succeeding, the next step makes the identical hand movement with nothing in your hand, paying from the other hand or a pouch afterwards. That empty movement is now a hand signal, and a hand signal is a real cue you can keep. If the empty hand turns out to be too big a jump, that is what the plan's hidden split steps are for.

The specific trap worth naming, because it is so easy to fall into: do not negotiate mid-rep. You cue, your dog does not respond, so you produce the food and wave it, and then your dog complies, and then you pay. What you just reinforced is the whole sequence, including the ignoring. A few dozen of those and you have a dog who reliably responds on the second offer, with a treat visible. If your dog cannot do the rep, the rep was too hard. Mark it honestly, and let the app find a smaller step.

Often confused with

Bribery, and reinforcement. A lure comes before the behavior and is antecedent. Reinforcement comes after.

Not a technical term. It is an accusation, and usually it is aimed at food training in general by people selling something else.

There is still a real thing underneath it, so here is the honest version. The distinction that matters is timing: food that appears before the behavior is a lure or a prompt, and food that appears after it is a consequence, and usually reinforcement. Luring is a legitimate teaching tool. Nothing has gone wrong at the moment you show a dog a treat to get a first sit.

What people are right to worry about is the version where the food must be visible before your dog will respond. At that point the food has become part of the cue, which is a fading problem with a specific fix, not a moral failing. And the genuinely damaging version is negotiating mid-rep, described above: cue, no response, produce food, comply, pay. That reinforces the ignoring.

Note that the accusation cuts the other way too. "He only works for food" is a strange complaint from anyone using a tool that only works because the dog wants it to stop.

Often confused with

Using treats at all. Most of what you train depends on its consequences. Paying your dog is not a phase you graduate out of, though the rate thins out enormously once something is learned. See Treats.

Any extra help you add to make a behavior happen, on top of the cue you eventually want, and plan to remove.

Some prompts need no training at all: food moving in front of the nose, a high-pitched noise, you crouching down, you moving away quickly, a hand patting the floor. Your dog just reacts. Others are things your dog has already learned, like a trained hand signal used to help a new word along.

What makes something a prompt is the job it is doing, not whether it was taught. A hand signal is the cue for a behavior when it is the signal you want your dog to answer. It is a prompt when it is temporary help you mean to fade, as it is in jumping the prompt, where the hand signal helps a new word take over. The distinction that actually matters in practice is whether the help gets removed.

Which is the rule for all prompts: add the least you need, and take it out as early as you can. See Fading.

In Push Drop Stick

Prompting steps are the app's main step type, and the name is used broadly there: you cue or prompt the behavior, whichever the step calls for, and score whether you got it.

Removing help gradually until your dog performs the behavior without it.

Every lure, prompt, hand movement, body position, and step toward your dog is help. Fading is the unglamorous half of training that determines whether you end up with a trained dog or a dog who performs beautifully as long as you stand in the kitchen with your hand in your pouch.

Faded well, it happens in small deliberate slices: the food leaves your hand but the movement stays, then the movement shrinks, then it happens with your hand at your side, then from a step further away, then sitting down instead of standing. Each slice is a step in a plan, and each one is small enough that your dog barely notices.

Faded badly, which means all at once, the behavior falls apart, and it looks as though your dog was never trained. They were. The cue they learned simply included your body, and you just changed the cue.

A dog who "only does it when you have food" has not been disobedient. They have been under-faded.

Often confused with

Extinction. Fading removes the help before the behavior. Extinction removes the payment after it.

When your dog stops waiting for the help and responds to the new cue on its own.

To add a verbal cue you pair it with something your dog already knows: say "sit", then give the trained hand signal, then pay. The word reliably predicts the hand signal, and the hand signal reliably predicts a payable sit. After enough repetitions your dog puts it together and sits on the word, before your hand has moved. That moment is jumping the prompt, and you pay it immediately and generously.

Two expectations worth setting. This takes far more than five repetitions, which is why the app uses timed sets for it rather than counting five reps. And you are not counting right and wrong. You are counting how often your dog beat you to it versus how often they still needed the hand signal.

When your dog is jumping the prompt about three times in five, they are ready to be asked for the word alone.

In Push Drop Stick

The Shaping step type, and how nearly every verbal cue goal in the app gets earned.

A calm word that tells your dog this attempt will not be paid, and to try again.

Typically "oops", "whoopsie", or "try again". Said in a neutral or friendly tone. Never "No!", never anything with an edge on it: the goal is information, not intimidation.

It is worth being precise about what it is, because it is often described as pure information. Anything that follows a behavior and makes it less likely is, by definition, a punisher, and that is what a no reward marker does when it works. It is a learned signal that this attempt has lost its chance to pay, which makes it a mild form of negative punishment, however friendly the voice. The tone decides how mild. An edge on it adds something unpleasant on top, which is positive punishment, and then you are training a dog who is anxious about being wrong, which is the opposite of what you need in a dog who is learning. None of that makes it a bad tool. It makes it one to use well: kept light, and saved for the clear-cut mistakes below.

They are genuinely useful, and often the most efficient tool there is, for clear, well-defined mistakes that contrast sharply with the goal: breaking a stay, going for the food on a leave it, taking the wrong end of a retrieve. They are unhelpful for vague or gradual work, like shaping a dog toward a new movement, where "not that" tells your dog nothing they can act on.

They are also optional. Plenty of excellent trainers never use one. If in doubt, say nothing, reset, and run the rep again.

In Push Drop Stick

Not required anywhere. Tapping fail is not an NRM, and your dog should not be able to tell from your face or your voice which button you pressed. See Pass and Fail.

Delivering the treat while your dog is still in the position you asked for.

It sounds fussy and it is one of the highest-value mechanical skills there is, because where the food lands shapes the next rep. A marker tells your dog which moment was right, but the treat still pays whatever your dog is doing when it arrives. Ask for a sit, and if the treat lands while your dog is standing up to take it, you paid the sit with the marker and then paid the stand with the food.

So bring the food to your dog rather than making them come to the food. For a sit, in to the mouth at their head height. For a down, to the floor between the front paws, low, so there is no reason to lift up. For a stand, level and slightly forward. Done well, position feeding also builds duration by itself, because holding still is what keeps producing food.

If your dog pops up before the treat arrives, do not lure them back into position and then feed. That looks like a fix and it teaches a chain: sit, stand, get put back, get paid. If you marked the sit, pay it anyway, since a marked behavior is always paid, and simply deliver it low and quickly. If you did not mark, the treat goes away. Either way you reset and run the next rep with the food arriving sooner. Nine times out of ten the problem was your hand being late, not your dog being wrong. See Pass and Fail.

Often confused with

Hand feeding. This is about where the food arrives, not who holds it.

The gap between your cue and your dog starting the behavior.

Worth naming because it is a quality measure that has nothing to do with whether the rep passed. A dog who sits four seconds after the word has responded to the word, and if your criteria said three seconds, that rep is still a fail. Both facts are useful, and the second one is what stops "he did it eventually" from quietly becoming your standard.

Long latency usually means one of three things: the cue is not yet solid, the environment is harder than you think, or your dog is weighing up whether it is worth it. That last one is a payment problem.

Latency usually shortens with practice, because fast answers keep getting paid sooner. Early on you do not need to chase it, and once a cue is solid you can tighten it on purpose with a latency criterion. Either way, never hurry a dog who is thinking.

Deliberately giving your dog unlimited time to work out the answer after a cue, without repeating yourself.

You say the word once, and then you wait. And wait. As long as your dog is still engaged and thinking, you keep waiting. If they wander off or check out entirely, the rep is over: reset and make it easier.

This feels wrong and it is one of the most useful things you can do while a cue is new. The alternative is cue chanting — "sit, sit, siiit, SIT" — which teaches your dog that the first cue was optional and dilutes the word into background noise. Every extra repetition of the cue costs you a little of its meaning.

The delay usually shrinks within a few reps, because answering keeps being what gets to the treat. See Latency.

A word that tells your dog they are free to move out of a position you asked them to hold. "Okay!", "Break!", "Free!"

It is the bookend of every stay, and without one your dog is deciding for themselves when a position is over, which means the position has no duration in it. With one, a sit lasts until the word comes, and the word can come sooner or later without your dog having to guess.

Pick something you do not say all day, so "okay" is a slightly risky choice for most households. To build it, say the word and immediately make moving happen, with a hand clap, a step away, a tossed treat or a bit of play.

Always release on purpose. If your dog wanders out of position and nothing happens, you have taught a self-release, and every stay you train afterwards will have a hole in it.

Often confused with

A marker. A marker says "that was right". A release says "you are done". Trainers handle the two in one of two ways, and both work. In the clicker tradition, the click ends the behavior: the marker is also the release, so your dog leaving the position to collect the treat is expected, and you build duration by marking later. Others keep them separate, with the marker paid in position and a release word to end it. What matters is choosing one and being consistent, so your dog always knows whether a "yes" means stay or go.

Linking several behaviors together so they run as one sequence.

Go to your mat, lie down, stay there. Or come, sit in front, turn to heel. Each link is taught separately to fluency first, then joined. Trying to teach the whole chain in one go before the pieces are solid is lumping and produces a dog who is shaky at every part of it.

Back-chaining is teaching the last link first, then the one before it, and so on. It is often the preferred method in animal training for anything where the ending matters, because your dog is always moving toward the most familiar, most reinforced part of the sequence, so confidence rises through the chain rather than falling off at the end.

Once a chain is solid, one cue can run the whole thing, and the behaviors in the middle hold themselves up, because each link's cue has become a reward for the link before it: the chance to do the next thing is what pays the last one.

Chains also form when nobody meant to build one. Anything that reliably happens just before the payment becomes part of what your dog learned, including the part you wanted rid of. Your dog jumps up, you ask for a sit, you pay the sit. You believe you taught "sit instead of jumping." What your dog learned is jump, then sit, then cheese. Same on leash: pull, get asked to heel, get paid, and pulling becomes how the routine starts.

The fix is not to stop paying the behavior you want, it is to stop letting the unwanted behavior start the sequence. Pay four feet on the floor as your dog approaches, before the jump. Arrange the setup so the unwanted link cannot be rehearsed. See Management. And watch for your cue becoming your answer to misbehavior: if the only time your dog hears the word is right after they got it wrong, then getting it wrong is part of the cue.

In Push Drop Stick

Automatic goals: the app records when a whole chain runs from a single cue, or when a behavior happens in its context without being asked each time.

How well, how fast, and how reliably a behavior happens. The difference between a dog who can do it and a dog who does do it.

A fluent behavior is quick off the cue, in a consistent form, repeatable without warm-up, and durable in mildly imperfect conditions. A behavior that only shows up on the third try, slowly, in exactly one room, is real learning that is not finished yet.

Fluency is what you build before adding difficulty. Owners who take a shaky behavior somewhere new are usually the ones who conclude their dog is being stubborn, when what actually happened is that they added a hard variable to something that had no margin in it.

Often confused with

Stimulus control, which is about whether the cue is in charge, not how good the behavior is. A dog can be beautifully fluent at a behavior they also offer uninvited all evening.

Your dog's ability to perform a known behavior in new places, with new people, in new situations, rather than only where they learned it.

Dogs are far more context-specific than people expect. A sit that is rock solid in your kitchen can genuinely fall apart at the park, and that is not defiance, it is that the behavior was learned along with the kitchen: your position, the floor, the light, the fridge, the pouch in your hand.

To generalize, practise everywhere, starting easy. Change one thing at a time: same room, but you are sitting down. Same behavior, but in the hallway. Same cue, but your partner gives it. Every new setting is its own small challenge, so lower your criteria at first and let your dog succeed before raising them again.

This is the entire purpose of proofing. A behavior is not reliable until it works somewhere you have never trained it.

Often confused with

Testing your dog. Generalizing is teaching. It should look like a lot of easy wins in a lot of different places.

The other half of generalization: your dog telling the difference between two things rather than treating them the same.

Sit and down are different words. Your mat and the sofa are different places. The cue "touch" and the cue "take it" mean different things. Generalization without discrimination gives you a dog who does their favourite trick whatever you said, and discrimination without generalization gives you a dog who is precise in the kitchen and useless outdoors. You want both.

Discrimination is best built by paying the right answer and simply not paying the other one, rather than by correcting the wrong one: keeping errors few is what makes it fast and keeps it calm. If your dog is guessing between two cues, the cues are too similar or too new: separate them, make each one solid on its own, then alternate them unpredictably.

In Push Drop Stick

Proofing games are a discrimination exercise. Shuffling earned goals from different plans into a random order removes the pattern your dog was leaning on, so the cue itself has to carry the meaning.

How long your dog holds a behavior before you mark and pay.

One of the three Ds trainers use to describe difficulty, along with distance and distraction, and one of the eight kinds of goal a dog can earn here.

Duration is the criterion owners raise fastest and least carefully, usually by counting in their head while the dog is already struggling. Raise it in increments small enough that your dog barely notices, and raise one thing at a time. A dog who can hold a down for thirty seconds in your kitchen cannot necessarily hold it for thirty seconds while you walk away, and asking for both at once is how a solid stay comes apart.

Duration behaviors also need a release cue, because a stay that ends whenever your dog decides it has ended is not a stay.

In Push Drop Stick

Duration steps are timed and the plan raises the number for you. Where the length varies from rep to rep, see Interval: mixing short reps in among the long ones is what keeps the rate of reinforcement liveable on a step that would otherwise pay twice a minute.

Often confused with

Patience, or the lack of it. A dog who breaks a thirty second down is not impatient or defiant. Thirty seconds is simply above what they can currently do, which is a criteria problem with a number attached, and the app will drop it.

How far apart you and your dog are while the behavior happens.

The second of the three Ds, and a goal type of its own. It covers the dog who holds a sit while you stand beside them and melts the moment you step back, and the recall that works at ten feet and not at fifty.

Distance is harder than owners expect, because moving away is itself a cue to follow. You are asking your dog to stay put while everything you have ever reinforced says to come with you. Expect to build it in pieces: half a step, a step, a step with a turn, two steps and back.

In fear work, distance runs the other way. The space between your dog and something that worries them is a margin to protect before it is a criterion to shrink. It does come down over time, but only as fast as your dog stays comfortable, never because the plan says so, and more of it is always easier. A frightened dog needs the room to be big enough that they can watch without needing to react. See Threshold and Desensitization.

Often confused with

Those two uses of the word. Distance from you is a skill you build up. Distance from a trigger is a safety margin you keep. Closing the second one because the number is supposed to go up is how fear work goes wrong.

Anything in the environment competing with you for your dog's attention: another dog, a dropped chip, a child on a scooter, whatever was on that lamp post.

The third D, and a goal type. The reason a behavior that is perfect in your kitchen evaporates in the street is not that your dog is being difficult, it is that your kitchen has nothing in it. See Generalization.

Distraction is a criterion like any other, which means it gets added deliberately and small, and everything else comes down when it does. Add a distraction and shorten the duration, close the distance, go back to the easy version of the behavior. Raise your treat value as well: what your dog will work for in a quiet room is not what they will work for with another dog forty feet away.

In Push Drop Stick

Proofing games are the app's version of this. Skills your dog has already earned, shuffled and called wherever you happen to be, so a cue has to work in a new room, out of order, and next to a behavior it was never practised beside.

Often confused with

A trigger. A distraction competes for attention. A trigger produces an emotional response. They can look alike from the outside, and the difference decides your whole approach: you work through a distraction and you back away from a trigger.

Disobedience. A dog who cannot answer a cue with a squirrel in view has not stopped knowing the cue. The squirrel is worth more than you are at that moment, which is information about your distance and your treats, not about your dog's character.

Rearranging the environment so the unwanted behavior cannot happen in the first place. If it never happens, it never gets reinforced, and it never becomes a habit.

Put the bin behind a door. Baby gate the hallway. Pick the shoes up. Walk at seven in the morning instead of five in the evening. Close the curtains on the window your dog barks out of.

This is the least glamorous entry in the glossary and the one that solves the most problems. It also carries very little risk, which makes it the honest first answer to almost every "how do I stop him" question. Preventing a habit is far easier than breaking one.

Two things to be clear about. Management is not training. Your dog is still learning while you manage, mostly that the old habit has stopped paying, but nothing in it teaches them what to do instead, which is why it works best paired with differential reinforcement — prevent the old option while you pay a new one. And it is not cheating. A trainer who manages a problem for six months while quietly building the alternative is doing the job properly.

In Push Drop Stick

It is the reason a plan's early steps can be so easy. Arranging the situation so the right answer is the obvious one is management, applied to a single rep. See Antecedent, Behavior, Consequence.