How to Train a Reliable Recall (and Why Rewards Beat E-Collars)

Why recall is the cue you cannot fake

Recall is the only cue with a floor under it. A dog who ignores "sit" is mildly irritating. A dog who ignores "come" while a gate stands open, or a deer breaks from a hedge, is a dog whose safety rests on a habit you either built properly or did not.

It is also the cue most owners ruin without noticing, and not by picking the wrong treat. The word gets said dozens of times when the dog cannot possibly comply, and arriving at the human keeps turning out to be the dullest option available. Recall is a competition between what you pay and what the environment pays, and early on the environment has deeper pockets.

What the evidence actually shows

The clearest trial here is China, Mills and Cooper, published in Frontiers in Veterinary Science 7:508 (2020). Sixty-three dogs were split into matched groups, some trained with electronic collars, some with reward-based methods only.

On the measure that matters, the reward-based dogs came on the first cue 82.5% of the time against 71.0% (p = 0.015), and were quicker to sit once they arrived, 1.36 seconds against 1.67 seconds. They were not merely happier; they were better at the exact job the e-collar is sold to fix.

Be straight about the limits: 63 dogs is one trial, the gap is real but not enormous, and the tasks were obedience tasks. Nobody has shown this transfers to serious aggression.

The welfare picture is clearer. Vieira de Castro and colleagues (PLOS ONE 15(12), 2020) followed 92 dogs across aversive, mixed and reward-based schools. Lip-licking, a stress signal, averaged 55.9 observations in the aversive group against 4.1 in the reward group (d = 3.9), cortisol was raised, and weeks later those dogs read an ambiguous situation more pessimistically. The mechanism is in why punishment-based dog training backfires.

Build it on a clean cue, in small steps

If you have already burned your recall word, retire it and pick a new one. Karen Pryor's Ten Laws of Shaping, in Don't Shoot the Dog! (Bantam, 1985), are the skeleton, and the first four do most of the work. The full set is in the ten laws of shaping.

• Law 1: raise criteria in increments small enough that the dog always has a chance of being reinforced. If your dog fails more than about one rep in five, the step was too big.
• Law 2: train one aspect at a time. Distance, distraction and location are separate criteria. Add one, never two.
• Law 3: put the current level on a variable schedule before raising criteria. Once the dog nails ten feet in the kitchen, stop paying every rep before you move outside.
• Law 4: when you introduce a new criterion, relax the old ones. First time in the garden, come close, keep it short, pay every arrival again. That is the method, not backsliding.

Keep sessions short. Pryor notes (p. 65) that appetite for reinforcement falls off after roughly a dozen reinforcements, with a ceiling near eighty in a day. Five clean reps beat twenty grinding ones.

Practical take: Say the cue once, then stop talking. Repeating "come, come, COME" teaches the dog the cue is background noise that eventually stops.

A progression you can run this week

Two sessions a day, a handful of reps each, and stop while the dog still wants more.

1. Days 1 and 2, hallway. No distraction. Cue once from two metres, mark the moment the dog turns toward you, pay on arrival. Every rep pays.
2. Day 3, whole house. Same distance, new rooms. Distance holds still while location changes.
3. Day 4, distance. Back to the quietest room. Stretch to five, then eight metres. If a rep fails, the next one is shorter.
4. Day 5, garden on a long line. New place, so criteria relax: two metres again, every arrival paid, three reps and done.
5. Day 6, mild distraction. Garden, long line, a dropped toy ten metres off. Call before the dog commits to the toy, not after.
6. Day 7, variable pay. As day 6, but only some arrivals earn food. All earn something: praise, a thrown toy, release back to sniffing.

Keep the long line on for weeks longer than feels necessary: every recall you lose teaches the dog the cue is optional. Tavi's training module sequences a progression like this and illustrates each rep, and long-line handling follows the rules in our loose-leash walking guide.

Markers and jackpots, used properly

A marker is a sound that tells the dog "that, right there, is what earned this" (Pryor, p. 33, the conditioned reinforcer). For recall, mark the turn rather than the arrival: the turn is the decision you are buying.

The evidence on markers is mixed. Gilchrist and colleagues (PeerJ, 2021) ran two controlled studies with pet dogs on simple tasks: a clicker gave no advantage over food alone, or over a verbal marker. But Lazarowski and colleagues (Frontiers in Veterinary Science, 2025) tested 28 detection dogs on a hard task with a long delay before reward: with a marker they reached the final stage in 36 trials against 68, and alerted with 92% accuracy against 69%. Both fit if a marker earns its keep when the reward cannot arrive immediately. A recall from thirty metres is exactly that gap. A sit in your kitchen is not.

A jackpot is an occasional, out-of-proportion reward (Pryor, p. 31). Save it for the rep that surprises you, the day your dog turns away from another dog mid-approach. More in our guide to jackpot rewards. Once recall is solid, move to variable reinforcement (Pryor, pp. 36-40): the slot-machine argument is that an unpredictable payoff builds a habit harder to extinguish than one paying every time.

Practical take: No established, dog-specific number exists for how fast a marker must follow the behaviour, and Pryor gives none. The one-second target in the Tavi app is our own design choice, not a research finding.

Generalising: distraction and stimulus control

A recall that works in your kitchen is a kitchen trick. What you want is stimulus control, which Pryor defines with four conditions (p. 81): the behaviour always occurs on the cue, never without it, never in response to a different cue, and no other behaviour occurs in response to this cue. The fourth catches people out. If "come" sometimes produces a dog who arrives halfway and sits, the cue is not clean. More in stimulus control, explained.

Build a distraction ladder and climb it in order: empty room, room with a person in it, garden, garden with a toy out, quiet field on a long line, park at a busy hour. Move up only when the current rung is boring; move down the moment two reps go wrong.

Distance from the distraction is a dial you control. Most dogs who cannot recall near other dogs can recall a long way from them. Start where the dog wins, then shrink the gap. If your dog saw little of the world early on, recall is harder, and part of the fix is our puppy socialisation guide.

The commonest way a recall rots

Here is the mistake almost everyone makes, and it is not about treats. You are at the park. Your dog is having the best hour of their week. You call, they come, you clip the lead on and go home.

Do that twenty times and you have taught something precise: the cue predicts the end of everything good. Pryor's Law 8 says do not interrupt a session gratuitously, because the interruption is itself a punishment. Ending the fun the instant the dog arrives punishes the arrival. The dog is not being stubborn later. The dog learned what you taught.

The repair is dull. Most recalls should end with the dog going straight back to what they were doing. Call, pay, release. Occasionally clip the lead on, walk twenty paces, unclip, let them go again.

When a recall that used to work stops working

Adolescence, a house move, one frightening encounter, and a solid recall comes apart. Pryor's Law 9 is the answer: when a behaviour deteriorates, go back to kindergarten. Not halfway back. Work through this before deciding the dog is being difficult:

• Has the cue been poisoned? Count the last ten times you used it. If most ended something the dog enjoyed, retire the word.
• Are you asking for two criteria at once? A new park and a longer distance on the same day is two. Drop one.
• Has the pay quietly stopped? Most people stop reinforcing once a behaviour looks finished. A recall never comes fully off the payroll.
• Are you calling too late? Once a dog has committed to a squirrel you are not competing, you are narrating. Call at the head-turn.
• Is something physical in the way? A dog who has become slow, or ignores you from one side, may be sore or losing hearing. That is worth a vet check, not more training.

Rebuilding usually takes weeks rather than months, provided you genuinely return to the hallway and pay every rep again. Finish each session on a rep the dog got right. Law 10, end on a high note, is not sentimentality. It is what the dog remembers next time you pick up the long line.

Frequently asked questions

How long does it take to train a reliable recall?

A basic indoor recall usually takes a few days of short sessions. A recall you would trust off-lead near other dogs takes months, because the work is not teaching the behaviour, it is proving it at every distraction level one rung at a time. Plan on keeping a long line for most of that period. Rebuilding a recall that used to work is faster, often a couple of weeks, provided you go right back to the easiest version and pay every repetition again.

Are e-collars better for recall than treats?

No. The one controlled trial designed to test that question, China, Mills and Cooper in Frontiers in Veterinary Science in 2020, trained 63 dogs in matched groups. Dogs trained with reward-based methods came on the first cue 82.5% of the time against 71.0% for dogs trained with electronic collars, and sat faster once they arrived. Separate work links aversive training to raised stress signals and cortisol, so the collar costs welfare without buying performance.

Why does my dog come at home but ignore me at the park?

Because the behaviour has not been generalised yet. Dogs do not automatically transfer a cue from one context to another, so a recall learned in a kitchen genuinely is a kitchen behaviour until you rebuild it elsewhere. Work a ladder of locations and distractions, relaxing distance and paying every arrival each time the setting changes. Also check what the park is paying: if coming back always ends the visit, the park is outbidding you and the dog is choosing sensibly.

Should I use a clicker to train recall?

You can, but a word works too. Gilchrist and colleagues, writing in PeerJ in 2021, found a clicker gave pet dogs no advantage over food alone or a verbal marker on simple tasks. Lazarowski and colleagues in 2025 found markers mattered a great deal on a hard task with a long delay before reward: 28 detection dogs reached the final stage in 36 trials with a marker against 68 without. Recall from a distance has that delay, so mark the turn, then pay on arrival.

What should I do when my dog ignores the recall cue?

Do not repeat the cue and do not chase. Go and collect the dog calmly, without scolding, then treat the miss as information. It usually means the step was too hard: too much distraction, too much distance, or both added on the same day. Drop back to a level the dog was winning at, pay every repetition there for a few sessions, then climb again more slowly. If a previously reliable dog starts missing cues generally, a vet check for hearing or pain is reasonable.

Tavi's guides are informational and not a substitute for professional veterinary advice, diagnosis or treatment. When in doubt, call your veterinarian.

Sources

• China, Mills & Cooper, Frontiers in Veterinary Science 7:508 (2020)
• Vieira de Castro et al., PLOS ONE 15(12) (2020)
• Gilchrist et al., PeerJ (2021)
• Lazarowski et al., Frontiers in Veterinary Science (2025)
• Karen Pryor, Don't Shoot the Dog! (Bantam, 1985)
• DK, The Complete Dog Breed Book, 2nd edn (2020)