BRENNEN

Industries / Retail, hospitality and personal services / Pet care

Training

Teaching pets to respond to cues and to live calmly with people, in classes, at home or with a private trainer, and the rules on which methods and tools trainers may use.

7

Milestones achieved, from 1933 to 2020.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1933

    The first obedience test

    Helen Whitehouse Walker, a Standard Poodle breeder, held what the American Kennel Club (AKC) calls the first obedience trial in October 1933 on her father's estate in Mount Kisco, New York, with eight dogs entered. She borrowed the idea from the competitive tests of England's Associated Sheep, Police, Army Dog Society. The AKC approved its first obedience regulations on 10 March 1936, and owners still train and handle almost all dogs entered in its trials.

    akc.org

    Credit

    Helen Whitehouse Walker; American Kennel Club

  2. 1961

    Instinct beats conditioning

    Keller and Marian Breland's company, Animal Behavior Enterprises, took operant conditioning, the laboratory method of shaping behavior with well-timed rewards, into commercial animal training in the 1940s and 1950s; a 2005 history links the later spread of these methods among marine mammal and bird trainers to their work. In "The Misbehavior of Organisms" (American Psychologist, 1961) they reported conditioning 38 species and more than 6,000 animals, and described failures: a raccoon rewarded for dropping coins into a box kept rubbing them together instead, which they called instinctive drift.

    Lessonrewards shape behavior only within the limits of a species' instincts, especially its food-getting habits.

    psychclassics.yorku.capmc.ncbi.nlm.nih.gov

    Credit

    Keller Breland and Marian Breland, Animal Behavior Enterprises

  3. 1992

    Clicker training reaches dogs

    Karen Pryor, who had trained dolphins at Sea Life Park in the 1960s and whose book Don't Shoot the Dog won the Dog Writers Association of America's Maxwell Award in 1984, ran what she calls the first Don't Shoot the Dog clicker training seminar with trainer Gary Wilkes in 1992, for 250 dog trainers in the San Francisco Bay Area. A clicker marks the instant an animal does the right thing, and a treat follows; marine mammal trainers had used a whistle the same way. Pryor's method avoids punishment within training sessions.

    clickertraining.comkarenwpryor.com

    Credit

    Karen Pryor; Gary Wilkes

  4. 1999

    The alpha wolf retired

    Most research on wolf dominance had used captive wolves placed together by people. After 13 summers watching wild wolves on Ellesmere Island in Canada, the biologist L. David Mech concluded that a wild pack is a family led by the parents, not an "alpha" pair that won its rank in contests; he now calls the term outdated, though his own 1970 book helped spread it. The American Veterinary Society of Animal Behavior (AVSAB) cites his work in its position statement against dominance-based dog training, along with a 2008 study in which confrontational methods such as the "alpha roll", pinning a dog on its back, often provoked aggression.

    Lessona model drawn from captive animals in groups people assembled shaped advice on handling pet dogs until field data corrected it.

    api.crossref.orgdavemech.orgavsab.org

    Credit

    L. David Mech (Canadian Journal of Zoology, 1999); AVSAB

  5. 2001

    Certifying dog trainers

    The Certification Council for Professional Dog Trainers (CCPDT), set up in 2001 as an independent not-for-profit body, examines trainers on humane, science-based methods and requires continuing education to keep titles such as CPDT-KA. In November 2014 the National Commission for Certifying Agencies, part of the Institute for Credentialing Excellence, accredited it.

    Lessoncertification stays voluntary; in England and Wales, a Defra-commissioned taskforce noted in 2026, anybody can still call themselves a dog trainer or behaviorist.

    ccpdt.orgccpdt.orggov.uk

    Credit

    CCPDT

  6. 2010

    Wales bans shock collars

    The Animal Welfare (Electronic Collars) (Wales) Regulations 2010, in force from 24 March 2010, made it an offense to attach a collar designed to give an electric shock to a cat or dog, or to be responsible for an animal wearing one, punishable by up to 51 weeks in prison, a fine or both.

    LessonEngland's version, drafted after a 2018 consultation drew 7,334 responses and approved by the House of Lords in June 2023, lapsed twice without a Commons vote, the second time when Parliament ended in May 2024, so shock collars remain legal to use and sell in England, by the Royal Kennel Club's account.

    legislation.gov.ukgov.ukapi.parliament.ukthekennelclub.org.uk

    Credit

    Welsh Ministers (Elin Jones, Minister for Rural Affairs)

  7. 2020

    Rewards beat shocks

    Two 2020 studies scored dogs from video recordings. At the University of Lincoln, trainers who relied on rewards got dogs with poor recall to sit and come after a single command more often, and faster, than trainers nominated by electronic-collar makers who used shocks; the share of commands disobeyed did not differ (63 dogs, 21 per group). In a second study, 92 companion dogs from aversive-based schools showed more stress behaviors, larger rises in cortisol after training and more "pessimistic" choices in a cognitive test than dogs from reward-based schools. AVSAB's 2021 position statement recommends reward-based methods alone for all dog training.

    pmc.ncbi.nlm.nih.govebi.ac.ukavsab.org

    Credit

    Lucy China, Daniel Mills and Jonathan Cooper (Frontiers in Veterinary Science, 2020); A. C. Vieira de Castro, I. A. S. Olsson and colleagues (PLOS ONE, 2020); AVSAB

Watch for

  1. England revisits shock collars

    Defra's Animal Welfare Strategy for England commits the government to consult on banning the use of electric shock collars because of the possible harm to pets, with the strategy to be delivered by the end of 2030.

    Achieved whenregulations banning their use in England are made and come into force.

    gov.uk

    Claimed by

    Department for Environment, Food and Rural Affairs (Defra), December 2025

  2. Rules for UK dog trainers

    The Responsible Dog Ownership taskforce recommends exploring options to regulate dog trainers and behaviorists and names an evidence-gathering exercise on them as a priority. It notes that the sector regulates itself, with no mandatory standards, and that the Animal Behaviour and Training Council (ABTC), a charity with a public register of assessed practitioners, had just been accredited by UKAS.

    Achieved whenEngland or Wales makes registration or minimum standards mandatory for paid trainers and behaviorists.

    gov.uk

    Claimed by

    Responsible Dog Ownership taskforce, September 2026

Open

  1. 01

    Time cue responses from video

    An open set of at least 1,000 home and class videos of dogs given everyday cues (sit, down, come, stay) during ordinary reward-based practice, each labeled with the moment the cue was given, the moment the dog responded and whether the response was correct, plus a model that finds the same moments on its own. The output is a training log of latency and success rate per cue over weeks, the measures that studies of training methods rely on.

    The brief
    Credit · unclaimed
  2. 02

    Test AI training advice

    A public set of 200 dog-training questions written the way owners ask them (jumping up, pulling on the leash, barking at the door, growling over food), with a scoring guide written by three credentialed trainers from the AVSAB 2021 statement: does the answer recommend aversive tools or methods, does it send signs of aggression or anxiety to a vet or behavior professional, and is it specific enough to act on. Run it on general chat models and on assistants built into pet products, and publish the scores.

    The brief
    Credit · unclaimed

Open 01 · the brief

Time cue responses from video

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open set of at least 1,000 home and class videos of dogs given everyday cues (sit, down, come, stay) during ordinary reward-based practice, each labeled with the moment the cue was given, the moment the dog responded and whether the response was correct, plus a model that finds the same moments on its own. The output is a training log of latency and success rate per cue over weeks, the measures that studies of training methods rely on.

Why it is still open

Both 2020 studies of training methods scored dogs from video recordings, the Lincoln team timing responses to "sit" and "come". Open tools stop short of this: DogPoseCV, a 2021 Colorado State project, labels 20,578 still images as standing, sitting or lying and reached up to 92% test accuracy by its authors' figures, to trigger a treat dispenser, and DogMo (2025) records dog movement in 3D; neither times a response to a cue. In 2016 an NC State team detected sits with sensors in a dog harness rather than a camera, and its algorithm rewarded the right posture 96% of the time, against 100% for an expert trainer; it timed the reward, not the dog's response to a cue. CompanionPro, a commercial device that watches for a sit and dispenses a treat, has published no accuracy figures, and a case study that the SF SPCA said in 2020 was under review for publication could not be found in Europe PMC.

pmc.ncbi.nlm.nih.govarxiv.orghuggingface.coarxiv.orgsfspca.orgnews.ncsu.edu

Built with what exists

DogPoseCV images; DeepLabCut, an open pose-tracking toolkit; speech-to-text to time-stamp spoken cues (see the Note "Fewer word errors"); a small vision model fine-tuned on one graphics card (see the Note "One GPU to fine-tune"); ELAN, a free video annotation tool, for the human labels; Cohen's kappa for labeler agreement.

  1. First step

    Film ten volunteer owners practicing sits and downs with treats at home, ten clips each, and have two people label cue and response times without seeing each other's work.

  2. Done looks like

    1,000 clips from at least 100 dogs released under an open license, with the model agreeing with two human coders on whether the response was correct in 90% of held-out clips and placing the response within 0.25 seconds of their time in 90%.

Open 02 · the brief

Test AI training advice

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of 200 dog-training questions written the way owners ask them (jumping up, pulling on the leash, barking at the door, growling over food), with a scoring guide written by three credentialed trainers from the AVSAB 2021 statement: does the answer recommend aversive tools or methods, does it send signs of aggression or anxiety to a vet or behavior professional, and is it specific enough to act on. Run it on general chat models and on assistants built into pet products, and publish the scores.

Why it is still open

Pet products now answer behavior questions with AI; Furbo's Nanny Chat, marked beta, answers health and behavior questions using sources such as PetMD. Published tests of AI on animal questions cover veterinary knowledge, such as PetQA (September 2026), about 18,800 Korean question-and-answer pairs on dogs and cats answered by vets, and a search found no peer-reviewed test of chatbots' training advice.

help.furbo.comarxiv.org

Built with what exists

model APIs and open-weight models such as Mistral Large 4 (see the Note "A trillion open parameters"); a model that scores fixed-answer rubric items such as "does this answer recommend a shock or prong collar: yes or no" (see the Note "A thousand questions, one pass"); the AVSAB 2021 statement as the reference; a Python test harness; Cohen's kappa for scorer agreement.

  1. First step

    Write 40 questions on common problems such as jumping up and leash pulling, and have three certified trainers score two chat models' answers on their own.

  2. Done looks like

    200 questions and the scoring guide released under an open license, the trainers' agreement with each other reported, and at least six assistants scored on how often they recommend aversive methods and how often they refer serious cases, rerun every six months.