Everything this site rests on.

• T3avsab2008Professional-body statement

  American Veterinary Society of Animal Behavior (institutional author; no individual bylines) (2008). Position Statement on the Use of Dominance Theory in Behavior Modification of Animals. American Veterinary Society of Animal Behavior (AVSAB)

  AVSAB position statement (no printed date; attributed 2008) · open source ↗

  What it actually establishes. The American Veterinary Society of Animal Behavior advises veterinarians not to refer clients to trainers or behaviour consultants who advocate dominance-hierarchy theory and the confrontational training that follows from it, recommending reward-based methods instead.

  Commonly overstated as. It is routinely cited as scientific proof that 'dominance does not exist in dogs' or that canine social hierarchy is a myth. The document says the opposite about existence: it explicitly defines dominance as a real relationship and accepts that dominance-submissive relationships occur. Its argument is that the framework fails to explain or treat most pet behaviour problems, not that the phenomenon is fictional. A second overreach is treating it as evidence rather than as opinion: it is a professional body's recommendation, not a study.

  Limits: A consensus position document, not primary research, a systematic review, or a peer-reviewed paper; it presents no new data and reflects one US professional body's recommendations. CORRECTION TO THE YEAR: no publication date appears on the document, and AVSAB's own current position-statements page lists it with no year at all. The 2008 attribution rests on external evidence (it was being publicly excerpted in December 2008), while the hosted filename ('10-3-14') indicates a 2014 re-upload or revision and AVSAB simultaneously hosts a second copy under a 2019 upload path. Cite it as undated or 'c. 2008' rather than asserting 2008 flatly. CORRECTION TO THE PRIOR REVIEW: the earlier verification cited a '2025 AVSAB board communication'; the clarification actually on AVSAB's page is dated December 9, 2024, and it concerns a different, since-rescinded document that had suggested aversive tools may sometimes be appropriate. The dominance statement itself is not the rescinded document and remains listed and hosted by AVSAB.

• T3avsab2021Professional-body statement

  American Veterinary Society of Animal Behavior (institutional author; no individual bylines) (2021). AVSAB Position Statement on Humane Dog Training. American Veterinary Society of Animal Behavior (AVSAB)

  AVSAB position statement, 2021 · open source ↗

  What it actually establishes. In its 2021 position statement, the American Veterinary Society of Animal Behavior recommends that only reward-based methods be used for dog training and behaviour modification, stating that it found no evidence aversive training is necessary.

  Commonly overstated as. Popular writing upgrades this into 'science has proven aversive training does not work.' AVSAB's actual claims are narrower: that aversive methods are unnecessary, that they carry welfare and relationship risks, and that there is no evidence they are more effective. CORRECTION TO THE PRIOR REVIEW: the earlier verification asserted that AVSAB's objection is to the cost of aversive tools 'not to their capacity to change behaviour.' That is wrong. AVSAB does make a comparative-effectiveness claim, stating that reward-based methods are both safer and more effective and pointing to survey studies showing higher obedience. The defensible caveat is different: that comparative claim rests substantially on owner-survey and observational data, which cannot isolate training method from the kind of owner or dog involved, so a site should report it as AVSAB's reading of the evidence rather than as a demonstrated causal fact.

  Limits: A professional-body recommendation, not primary research and not a registered systematic review; no protocol, no pooled effect estimate, no risk-of-bias assessment. It synthesises a literature widely described as limited by small samples, inconsistent definitions of 'aversive,' reliance on owner report, and study populations (police, laboratory, and shelter dogs) that may not represent companion dogs. It reflects a single US body's consensus and is advocacy-adjacent in tone. CORRECTION TO THE PRIOR REVIEW: the earlier verification dated AVSAB's reaffirmation to a '2025 board communication.' The clarification published on AVSAB's position-statements page is dated December 9, 2024; it states that an outdated, since-rescinded AVSAB document suggesting aversive tools can sometimes be appropriate had been circulating, and directs readers to this 2021 statement as current. The substance of the prior claim holds; the year does not.

• T1bradshaw2009Peer-reviewed research literature

  John W. S. Bradshaw, Emily J. Blackwell, Rachel A. Casey (2009). Dominance in domestic dogs—useful construct or bad habit?. Journal of Veterinary Behavior: Clinical Applications and Research, vol

  doi:10.1016/j.jveb.2008.08.004 · open source ↗

  What it actually establishes. Bradshaw, Blackwell and Casey (2009) argue that dominance describes a relationship between two individuals rather than a fixed trait of one dog, so calling a dog 'a dominant dog' names nothing measurable about that dog - an argument, not a settled finding, and one other researchers have contested in print.

  Commonly overstated as. The most common overreach is 'this paper proved dominance does not exist in dogs, so dominance-based framing is scientifically refuted.' It proves nothing of the kind: it is chiefly argument, its evidential core is 19 dogs, and it makes the narrower claim that dominance is not an individual trait and is a poor motivational explanation for aggression - not that dominance relationships never form. A second overreach, which the ORIGINAL VERIFICATION being checked committed, is the claim that 'Bradshaw et al. published a response in 2016 that modified some of their original conclusions.' That is wrong. The follow-up (Bradshaw, Blackwell & Casey, 'Dominance in domestic dogs - a response to Schilder et al. (2014)', Journal of Veterinary Behavior 11:102-108, doi 10.1016/j.jveb.2015.11.008, online December 2015) rebuts the critics and reaffirms the original position, arguing that dogs lack three features central to wolf social organisation. Do not describe it as a partial retraction. A third overreach is quoting the paper as consensus when it is one side of a live disagreement.

  Limits: The original data component is one group of 19 dogs, all male and all neutered, in an artificial rehoming setting - no females, no intact animals, no household or free-ranging dogs, no replication. The rest is argument. It is directly contested by Schilder, Vinke & van der Borg (2014, Journal of Veterinary Behavior 9(4):184-191, doi 10.1016/j.jveb.2014.04.005), who conclude formal dominance is present in domestic dogs; that exchange is unresolved. Independent quantitative work reports measurable formal dominance signals and linear hierarchies in group-housed and free-ranging dogs (van der Borg, Schilder, Vinke & de Vries 2015, PLOS ONE 10(8):e0133978, doi 10.1371/journal.pone.0133978; Cafazzo et al. 2010; Bonanni et al. 2017). A site should present this as an open scientific debate. Note also that rejecting 'dominance' as an explanation for aggression is a separate question from whether confrontational training methods are harmful - do not use this paper as the evidence base for the latter; cite training-method outcome studies for that.

• T2bradshaw2011Authored scholarly or clinical text

  John W. S. Bradshaw (2011). Dog Sense: How the New Science of Dog Behavior Can Make You a Better Friend to Your Pet (US) / In Defence of Dogs: Why Dogs Need Our Understanding (UK). Basic Books, New York (US) / Allen Lane, London (UK, 352 pp.)

  ISBN 9780465019441

  What it actually establishes. In Dog Sense (Basic Books, 2011; UK title In Defence of Dogs), biologist John Bradshaw argues that training methods built on captive-wolf "pack" models misrepresent domestic dogs and that confrontational techniques are not justified by the evidence — an argument about training practice, not a demonstration that social hierarchy is absent in dogs.

  Commonly overstated as. The standard overreach is the flat claim "science has debunked dominance — alpha hierarchies don't exist in dogs or wolves, so hierarchy is a myth." Bradshaw's actual argument is narrower: hierarchy-based models are the wrong explanatory frame for most owner-dog problems, and training built on them is unjustified. The strong version is contested in the peer-reviewed literature. A second, smaller error is his affiliation: he founded the Anthrozoology Institute at the University of SOUTHAMPTON in 1992; it moved to the University of Bristol in 2003. Describing him simply as "founder of the Anthrozoology Institute, University of Bristol" compresses that inaccurately.

  Limits: A trade book with no methods section, so it cannot be checked at the level of a paper. The empirical plank most often cited from it — Bradshaw, Blackwell & Casey, "Dominance in domestic dogs: useful construct or bad habit?", Journal of Veterinary Behavior 4(3):135-144, 2009, DOI 10.1016/j.jveb.2008.08.004 — has been directly challenged by van der Borg, Schilder, Vinke & de Vries, "Dominance in Domestic Dogs: A Quantitative Analysis of Its Behavioural Measures," PLOS ONE 10(8):e0133978, 2015, DOI 10.1371/journal.pone.0133978, which found a near-linear hierarchy (h' = 0.94, p < 0.003) once lowering of posture was included, and explicitly faulted the 2009 study for excluding posture. Crucially, and omitted from most summaries of that rebuttal: van der Borg studied a single group of 10 core dogs (16 across the study) of mixed breeds, all sexually intact, living together in a university kennel at Utrecht — a small captive group that is a poor proxy for pet dogs living with people, so it does not settle the question for household dogs either. A related critique is Schilder, Vinke & van der Borg, "Dominance in domestic dogs revisited," Journal of Veterinary Behavior (2014) — cited here without a DOI because I did not verify one. Honest position: the anti-confrontational-training conclusion is well supported; "no hierarchy at all" is disputed, and both sides rest on small single-group studies.

• T1byosiere2016Peer-reviewed research literature

  Sarah-Elizabeth Byosiere, Julia Espinosa, Barbara Smuts (2016). Investigating the function of play bows in adult pet dogs (Canis lupus familiaris). Behavioural Processes 125:106-113. Companion paper on puppies: PLOS ONE 11(12) e0168570.

  doi:10.1371/journal.pone.0168570 (puppy study); PMID 26923096 (adult-dog study) · open source ↗

  What it actually establishes. Play bows in adult pet dogs most often occurred after a pause in play, and were followed by both partners resuming play, which is consistent with the bow functioning to restart or maintain play rather than to apologise for a rough bite. The original meta-signal proposal is credited to Bekoff (1995).

  Commonly overstated as. Commonly presented as proof that a bow always means the dog is friendly and that everything afterwards is safe play. It signals playful intent within an ongoing interaction; it does not guarantee that arousal will stay in a safe range.

  Limits: Small observational samples of dyads at play, with different results in wolf puppies, where bows did not reliably entice partners to play. The bow is one of the least ambiguous canine signals, but its precise function is still debated.

• T1castro2020Peer-reviewed research literature

  Ana Catarina Vieira de Castro, Danielle Fuchs, Gabriela Munhoz Morello, Stefania Pastur, Liliana de Sousa, I. Anna S. Olsson (2020). Does training method matter? Evidence for the negative impact of aversive-based methods on companion dog welfare. PLOS ONE, volume 15, issue 12, article e0225023

  doi:10.1371/journal.pone.0225023 · open source ↗

  What it actually establishes. In a 2020 PLOS ONE study of 92 companion dogs at seven training schools near Porto, Portugal, dogs attending aversive-based schools showed more stress-related behaviour and larger post-training cortisol increases than dogs at reward-based schools; the authors state their design cannot establish cause and effect.

  Commonly overstated as. It is almost always reported as proof that aversive training causes stress and makes dogs pessimistic. The authors rule that out in their own words: 'because this was a quasi-experimental rather than an experimental study, we cannot infer a true causal relationship between training methods and dog welfare.' Owners selected their own training school rather than being randomly assigned, so pre-existing differences between owners and between dogs cannot be excluded. Any site sentence containing 'causes,' 'makes dogs,' or 'proves' misreports this paper.

  Limits: n = 92 in the training phase (Aversive 28, Mixed 22, Reward 42) and n = 79 in the cognitive-bias phase (24 / 18 / 35); non-randomised and quasi-experimental. The authors document baseline group differences in owner gender, presence of children in the household, and the information owners used to choose a school, and state that 'a volunteer bias cannot be excluded and hence any generalization of the present results must take this in account.' The cognitive-bias test used absence of reward rather than any aversive stimulus, for ethical reasons, limiting what it demonstrates. Single region: seven schools in metropolitan Porto, Portugal. CORRECTION TO THE PRIOR REVIEW: the earlier verification flagged an unverified '2021 Scientific Reports matched-pair study.' I have now confirmed it — Casey, Naj-Oleari, Campbell, Mendl and Blackwell, 'Dogs are more pessimistic if their owners use two or more aversive training methods,' Scientific Reports volume 11, article 19023 (2021), 100 dogs in 50 matched pairs. It does replicate the direction of the cognitive-bias result, but it does NOT upgrade the evidence to causal: those authors likewise describe the association as correlational rather than causal and note that dogs with pre-existing pessimistic tendencies might prompt owners to adopt punitive techniques. Two correlational studies pointing the same way remain correlational.

• T2dogfacsAuthored scholarly or clinical text

  Bridget M. Waller, Catia Correia Caeiro, Kate Peirce, Anne M. Burrows, Juliane Kaminski (2013). DogFACS: The Dog Facial Action Coding System (manual). Manual / book. University of Portsmouth; repository record at University of Lincoln (eprint 17556, dated 15 March 2013, item type: Book). Distributed through the AnimalFACS project.

  No DOI or ISBN exists for this manual · open source ↗

  What it actually establishes. DogFACS is an anatomy-based system that gives each visible dog facial movement its own numbered code so researchers can record precisely what a dog's face is doing. Its own documentation states that it makes no inference about any underlying emotion or context causing the movement.

  Commonly overstated as. The dominant overreach is presenting DogFACS as an emotion decoder - an 'expression chart' mapping coded movements to feelings ('AU101 = sadness', 'this combination = guilt', 'DogFACS proves dogs smile'). The manual states the opposite in its own words: it is not an ethogram of facial expressions and makes no inference about underlying emotion. It is a measurement instrument, not an interpretation key. A second overreach is citing DogFACS as a source for a claim about dogs; it is a method, and any substantive claim must be attributed to the individual study that used the method. A third is implying anyone can apply it from a web summary - valid use requires passing a certification test.

  Limits: Not peer reviewed, no DOI, no ISBN, so it cannot support a finding - only a methodology. Access requires registration; valid coding requires certification, so uncertified coding carries no reliability guarantee. The official site publishes no version number or date, so the 2013 year derives from the repository record rather than the document itself; cite it by title and year, not by a precise imprint. Author order differs between sources (repository/Semantic Scholar list Waller, Correia Caeiro, Peirce, Burrows, Kaminski; the official AnimalFACS site lists Waller, Kaminski, Burrows, Caeiro, Peirce), so do not treat any one ordering as canonical. There is no single canonical reliability figure - individual studies report their own inter-coder agreement, generally moderate to good. CORRECTION TO A COMMONLY REPEATED SCOPE CLAIM: it is often asserted that Sexton et al. 2023 (Animals 13(14):2385, doi 10.3390/ani13142385 - DOI verified) documents that brachycephaly, wrinkling, long facial hair, ear carriage and coat colour degrade DogFACS coding validity. That is not what the paper reports. It examined superficial facial MARKINGS (eyebrow markings, patches, widow's peaks) in 103 pet dogs and found a statistically significant but weak negative correlation between facial complexity and DogFACS-coded expression rate (r = -0.326, p <= 0.001), i.e. plainer-faced dogs coded as more expressive. The authors frame this as a question about human-dog communication, and the paper does not establish whether the effect is a coding artefact or a real behavioural difference. Do not cite it as a validated list of coding-confound phenotypes.

• T1farago2010Peer-reviewed research literature

  Tamás Faragó, Péter Pongrácz, Friederike Range, Zsófia Virányi, Ádám Miklósi (2010). ‘The bone is mine’: affective and referential aspects of dog growls. Animal Behaviour, 79(4), 917–925 (Elsevier / ASAB)

  doi:10.1016/j.anbehav.2010.01.005 · open source ↗

  What it actually establishes. When dogs reached for an unattended bone and a hidden speaker played a growl, they were most likely to back off if that growl had originally been recorded from a dog guarding food — 11 of 12 withdrew within 15 seconds, versus 2 of 12 for growls recorded at a threatening stranger and 4 of 12 for play growls (Faragó et al. 2010; 36 dogs, 12 per condition).

  Commonly overstated as. Almost universally reported as "dogs can tell a play growl from a serious/aggressive growl," often as reassurance that a growling dog at play is read as harmless by other dogs. The behavioural test does not show that: play growls actually deterred slightly MORE dogs (4 of 12) than threatening-stranger growls (2 of 12), and the abstract's headline contrast is food-guarding versus threatening-stranger. The play-versus-agonistic difference is a production finding about how growls sound, not a demonstrated listener discrimination, and the claim also gets stretched to dog–human perception, which this study never tested. A further caution the popular retelling never mentions: the 'referential' reading rests on the two agonistic growls being acoustically indistinguishable, and the same team's larger 2017 study later DID find threatening growls have significantly lower formant dispersion than food-guarding growls — so the acoustic premise of the referential argument may simply have been a power limitation.

  Limits: The number of dogs whose growls were recorded is reported inconsistently across sources I could reach (20 in one secondary summary, 25 in Faragó's own dissertation, 36 in a blog account); I could not read the paywalled methods, so do not quote a recording sample size. Between-subjects design, one trial per dog, because a within-subject approach produced order effects. Faragó's doctoral dissertation (ELTE, 'Dog (Canis familiaris) growls as communicative signals', teo.elte.hu/minosites/ertekezes2011/farago_t.pdf) states the growl sample was too small for a permuted discriminant function analysis. Contexts staged; only one receiver context (a bone) was tested, so 'referential' rests on a single behavioural contrast. UNVERIFIED and removed from the prior version: the claims that the playback stimuli came from 11 dogs (6 per condition) and that the play and threatening-stranger conditions did not differ statistically from each other — I could not confirm either, and the raw counts (4 vs 2 of 12) are consistent with no difference but do not establish it. No independent direct replication by a separate group was found; follow-ups are largely by the same Budapest/Vienna team, e.g. Faragó, Takács, Miklósi & Pongrácz 2017, 'Dog growls express various contextual and affective content for human listeners', Royal Society Open Science 4(5):170134, doi:10.1098/rsos.170134 (verified via Crossref), in which humans classified growls at 63% against 33% chance, with food-guarding and threatening growls the pair most often confused. Referential signalling in dog vocalisation is contested in principle by Lord, Feinstein & Coppinger 2009 (doi:10.1016/j.beproc.2009.04.008), though that critique targets barking rather than growls.

• T3fearfree-fasProfessional-body statement

  Fear Free, LLC. The handout text is credited to Kenneth Martin, DVM, DACVB and Debbie Martin, LVT, VTS (Behavior). (The original entry's attribution of Fear Free's founding to a named individual was not verified here and has been dropped.) (Version referenced in the peer-reviewed evaluation: Fear Free® 2022. The fearfree.com handout page currently carries "© 2026 Fear Free. All rights reserved."). Spectrum of Fear, Anxiety and Stress (FAS Spectrum / FAS Scale). Fear Free, LLC — proprietary handout distributed through the Fear Free Certification Program (fearfree.com), with species-specific versions for dogs, cats, horses, birds and rats

  Fear Free FAS Spectrum (proprietary; do not reproduce the chart) · open source ↗

  What it actually establishes. The Fear Free FAS Spectrum is an expert-consensus observational scale (Fear Free® 2022) that scores a dog's visible signs of fear, anxiety and stress from 0 to 5 during veterinary handling; a 2025 study in the Journal of Veterinary Behavior reported strong inter- and intra-rater reliability and concluded the scale is a valid and reliable tool for assessing acute stress at the vet, while recommending minor design adjustments and rater training — and finding that untrained owners were significantly less accurate than behaviour experts on mid-range scores.

  Commonly overstated as. The dominant overreach is presenting the FAS score as a validated, objective measure of how a dog feels that any owner can apply anywhere. It is an observational scale of visible behaviour, not a physiological or emotional readout; it was designed and benchmarked for the veterinary-visit context, so exporting it to walks, dog parks or home life is off-label use; and untrained owners are demonstrably worse at it than experts in exactly the middle of the range where it matters most. Note also a correction in the opposite direction: it is an overreach to summarise the 2025 paper as concluding the scale is not yet fit for clinical use. The authors' stated conclusion supports the FAS Spectrum as a valid and reliable behavioural assessment tool while calling for minor design adjustments and sufficient training. A separate error is attributing the scale to "veterinary consensus" or a professional body; it is the intellectual property of a commercial certification company.

  Limits: Single evaluation study: 79 participants (owners, veterinarians, behaviour experts, trainers) rating 14 video recordings of MOCK veterinary examinations — raters, not dogs, are the sampling unit, so the dog sample is very small and the setting simulated. Reported figures: inter-rater ICC = 0.99 [0.99-1.00], intra-rater ICC = 0.83 [0.80-0.86], Spearman rho = 0.811 (p < 0.001) against the Lincoln Canine Anxiety Scale. Owners and behaviour experts differed significantly on FAS 3 (moderate signs, p = 0.0025) and FAS 4a (flight, p = 0.0355); extreme signs were easiest to score and mild escalating signs hardest. The authors flagged overlapping categories and a confusing numbering system (eight items mapped onto 0-5). Not independently replicated. The accepted manuscript is hosted on fearfree.com; I could not retrieve the funding or competing-interests declaration, so independence from the scale's owner should be checked before leaning on the paper. CORRECTION to a comparison often made: Mercier, Honeckman, Jokela, Dunham & Overall, Journal of Veterinary Behavior 62:12-17 (2023), DOI 10.1016/j.jveb.2023.02.004, did NOT evaluate the FAS Spectrum — it assessed two five-point Likert scales plus a benchmarked eight-component scale, using four raters (three early-stage residents and one experienced specialist), and reported fair-to-moderate inter- and intra-rater agreement. The "35 dogs" figure attached to it elsewhere could not be confirmed and should not be repeated. Finally, the handouts are copyrighted Fear Free material: link to the official page rather than reproducing the chart or artwork.

• T1herron2009Peer-reviewed research literature

  Meghan E. Herron, Frances S. Shofer, Ilana R. Reisner (2009). Survey of the use and outcome of confrontational and non-confrontational training methods in client-owned dogs showing undesired behaviors. Applied Animal Behaviour Science, volume 117, issues 1–2, pages 47–54

  doi:10.1016/j.applanim.2008.12.011 · open source ↗

  What it actually establishes. In a survey of 140 owners of dogs referred to a university veterinary behaviour clinic, owners who had tried confrontational techniques such as hitting, growling at, or alpha-rolling their dog often reported that the dog responded aggressively.

  Commonly overstated as. It is popularly rendered as 'research proves alpha rolls make dogs aggressive,' with the percentages quoted as though they described dogs in general. Two distortions matter. First, the percentages are denominated on the owners who actually attempted each technique, not on all 140 dogs, and not on pet dogs at large. Second, this is a retrospective, owner-recalled survey of a self-selected clinical population already referred for behaviour problems, many specifically for aggression; the design cannot separate 'the technique provoked aggression' from 'the dog was already aggressive, which is why the owner escalated to confrontational handling.' The direction of causation is genuinely undetermined.

  Limits: 140 completed questionnaires out of 219 distributed, a 64% response rate, collected in two waves between April 2007 and May 2008 at a single institution. Retrospective owner recall with no direct observation, no control group, no blinding, and no standardisation of how forcefully or competently each technique was applied. Some per-technique percentages rest on small numbers of owners who had actually attempted that technique, so the precision implied by two-digit percentages is misleading. Widely cited but never a randomised or controlled trial, and explicitly not a general-population sample of pet dogs. I did not independently verify the '190+ citations' figure asserted in the prior review, so no citation count is claimed here.

• T1horowitz2009Peer-reviewed research literature

  Alexandra Horowitz (2009). Disambiguating the "guilty look": Salient prompts to a familiar dog behaviour. Behavioural Processes, 81(3), 447-452

  doi:10.1016/j.beproc.2009.03.014 · open source ↗

  What it actually establishes. In a 14-dog experiment, the behaviours owners call the "guilty look" appeared more when the owner scolded the dog and no more often when the dog had actually eaten a forbidden treat than when it had obeyed (Horowitz, 2009).

  Commonly overstated as. Commonly flattened to "science proved dogs cannot feel guilt." The study recorded only video-coded behaviour, with no physiological or affective measure, so it cannot speak to whether a guilt-like state exists. Horowitz says so explicitly in the paper - the results do not indicate that dogs do not experience guilt. The defensible claim is that the display is an unreliable indicator of a misdeed. A second, subtler overreach is treating it as proof the look is "just" learned fear of scolding; the study identifies owner cues as the proximate trigger but did not test a learning history.

  Limits: n=14, single owner-dog treat-refusal paradigm, repeated within-subject trials on the same dogs (order and habituation effects are not controlled away). Outcome is owner-identified behaviours coded from video. Note the design confounds scolding with owner belief: owners were sometimes misinformed about what the dog did, which is what produces the headline result but also means "scolding" and "owner's mistaken knowledge state" are not fully separable. Converging independent evidence is genuine and I confirmed both records: Hecht, Miklosi & Gacsi 2012, Applied Animal Behaviour Science 139:134-142, DOI 10.1016/j.applanim.2012.02.015 (owners could not tell from greeting behaviour whether the dog had transgressed); and Ostojic, Tkalcic & Clayton 2015, Behavioural Processes 111:97-100, DOI 10.1016/j.beproc.2014.12.010, whose abstract states greeting behaviours "were not affected by their own action or the presence or absence of the food." I found no published study reversing this result.

• T2horowitz-bookAuthored scholarly or clinical text

  Alexandra Horowitz (2009). Inside of a Dog: What Dogs See, Smell, and Know. Scribner (Simon & Schuster), New York — first Scribner hardcover September 2009, 368 pp.

  ISBN 9781416583400

  What it actually establishes. Alexandra Horowitz's Inside of a Dog (Scribner, 2009) popularised interpreting dog behaviour from within the dog's own sensory world — dominated by smell rather than sight — rather than by mapping human emotions onto it.

  Commonly overstated as. The single most common overreach is "Horowitz proved dogs don't feel guilt." Her actual experiment — Horowitz, "Disambiguating the 'guilty look': Salient prompts to a familiar dog behaviour," Behavioural Processes 81(3):447-452, 2009, DOI 10.1016/j.beproc.2009.03.014 (verified via Crossref) — found no difference in guilty-look behaviours between dogs that disobeyed and dogs that did not, but more such behaviours when owners scolded, with the scolding effect strongest in dogs that had in fact obeyed. That is a finding about what causes the behaviour. It is silent on whether dogs have any guilt-like internal state, and Horowitz says so. A close second is quoting the book's sensory passages as hard numbers ("dogs smell 10,000 times better than humans"); the book is illustrative rather than metrological and such multipliers lack a stable published basis.

  Limits: A 2009 popularisation; cite it for framing, never as the source of a quantitative claim. The guilty-look study behind the most-quoted passage used 14 dogs and their owners in a single setup — small, with a thin conceptual-replication landscape. Roughly seventeen years of olfaction and cognition research postdate the book, and several sensory figures in circulation have been revised since. Note the book is not the study: cite the Behavioural Processes paper directly for the guilty-look result.

• T1kaminski2019Peer-reviewed research literature

  Juliane Kaminski, Bridget M. Waller, Rui Diogo, Adam Hartstone-Rose, Anne M. Burrows (2019). Evolution of facial muscle anatomy in dogs. Proceedings of the National Academy of Sciences of the United States of America, vol

  doi:10.1073/pnas.1820653116 · open source ↗

  What it actually establishes. A 2019 PNAS study dissected six dogs and four grey wolves and found the inner-eyebrow-raising muscle as a discrete muscle in five of the six dogs, while in the wolves it was present only as scattered fibres in connective tissue; separately, 27 shelter dogs raised their inner brows more often than 9 captive wolves when a stranger stood in front of them. Later dissections of coyotes and other wild canids indicate this muscle is not unique to domestic dogs, so the domestication explanation is now contested.

  Commonly overstated as. The dominant popular overreach is 'dogs evolved a special muscle that wolves simply do not have, in order to manipulate humans / to deliberately look sad.' Four separate errors. First, the wolves were not muscle-free - they had scant fibres of the same muscle, just not organised as a discrete muscle - and one of the six dogs lacked the discrete muscle entirely, so 'uniformly present in dogs, absent in wolves' overstates the specimen data (and the paper's own abstract commits this overstatement). Second, the selection-by-human-preference story is a stated hypothesis, not a finding, and nothing in the study demonstrates intent, deliberate signalling or manipulation by the dog. Third, and most serious for a public-facing page, the domestication-uniqueness claim has since been contradicted by independent dissections of other canids and should be presented as contested rather than settled. Fourth, do not use this paper to claim the inner brow raise means the dog is sad or is asking for something - a later behavioural study found the movement is tied to eye movement and occurs more when no person is engaging the dog.

  Limits: Very small dissection samples (n=6 dogs, one specimen per breed; n=4 grey wolves from two populations). The behavioural comparison pits 27 shelter dogs - 20 of them a single breed, Staffordshire Bull Terrier - against 9 wolves in two wildlife parks, so breed, life history and lifetime human habituation are uncontrolled confounds between groups, and the 2-minute exposure is a very thin behavioural sample. The domestication interpretation has NOT replicated; all four contradicting sources verified: Cunningham, Shankar, vonHoldt, Brzeski & Kienle 2024, Royal Society Open Science 11(10):241046, doi 10.1098/rsos.241046 (PMID 39359465, PMCID PMC11444785) dissected 10 coyotes plus 1 dog, found a well-developed LAOM in coyotes, and concluded the trait is probably basal within Canis and lost or modified in grey wolves rather than created by domestication. Sexton, Diogo, Subiaul & Bradley 2024, 'Raising an Eye at Facial Muscle Morphology in Canids', Biology 13(5):290, doi 10.3390/biology13050290 (PMID 38785773, PMCID PMC11118188), examined 10 canid specimens and concluded LAOM/RAOL are not unique to domesticated dogs. de Souza-Junior et al. 2026, Anatomia Histologia Embryologia 55(3):e70121, doi 10.1111/ahe.70121 (PMID 42068130), dissected 12 specimens across four Neotropical canid species and found both muscles consistently present in all taxa. CORRECTION: Smith et al. 2024, Anatomical Record, doi 10.1002/ar.25441 (PMID 38597113), 'Adaptations to sociality in the mimetic and auricular musculature of the African wild dog (Lycaon pictus)', is a dissection of a SINGLE adult male African wild dog - it is n=1 and must not be described in the plural or treated as population evidence. Separately on function: Bremhorst, Mills, Stolzlechner, Wurbel & Riemer 2021, Frontiers in Psychology 12:568935, doi 10.3389/fpsyg.2021.568935, studied 21 Labrador Retrievers and Labrador crosses and found the inner brow raiser occurred independently of eye movement in only 6% of cases (13 of 211), that eye movements were never observed without the inner brow moving or remaining tensed, and that the movement was shown MORE in a non-social context than a social one - concluding it is an integral feature of eye movement rather than a communicative signal.

• T1mech1999Peer-reviewed research literature

  L. David Mech (1999). Alpha status, dominance, and division of labor in wolf packs. Canadian Journal of Zoology, vol

  doi:10.1139/z99-099 · open source ↗

  What it actually establishes. The biologist whose 1970 book popularised the term 'alpha wolf' later argued the label was misleading: watching free-living wolves on Ellesmere Island, he found packs were normally families led by the breeding parents rather than by rank won in contests among unrelated adults (Mech 1999, Canadian Journal of Zoology).

  Commonly overstated as. The dominant overreach is 'science has proved dominance does not exist, so dominance is a myth in dogs.' The paper is about wolves and never studied domestic dogs. It does not claim dominance and submission behaviours are absent - Mech reports that they occur (uncommon except around food), describes rank order among offspring, and expressly keeps 'alpha' as valid for multi-litter packs. A second overreach, present in the ORIGINAL VERIFICATION being checked, is describing this as pure family harmony with no asymmetry; Mech's objection is terminological and descriptive, not a denial of social asymmetry. A third is presenting it as a retraction: Mech has continued publishing on wolf social behaviour throughout.

  Limits: Observational and synthetic, not experimental; no statistical hypothesis testing. The field component is a single habituated high-arctic population on Ellesmere Island over 13 summers, so pack composition, prey base and habitat are uncontrolled, and the wider claim rests on re-reading others' work. The captive studies it critiques (notably Schenkel 1947) used unrelated wolves confined together, where overt contests genuinely are more frequent - the paper explains those observations rather than erasing them. Critically for a dog site: nothing in this paper licenses inference to domestic dogs or to dog-human relationships. Independent work on free-ranging dogs reports significantly linear dominance hierarchies (Cafazzo, Valsecchi, Bonanni & Natoli 2010, Behavioral Ecology 21(3):443-455, doi 10.1093/beheco/arq001; Bonanni et al. 2017, Behavioral Ecology 28(4):1004, doi 10.1093/beheco/arx059). Mech's 1970 book The Wolf remains in print despite his requests to the publisher to stop, which is why the outdated framing still circulates.

• T1mendl2010Peer-reviewed research literature

  Michael Mendl, Oliver H. P. Burman, Elizabeth S. Paul (2010). An integrative and functional framework for the study of animal emotion and mood. Proceedings of the Royal Society B: Biological Sciences, vol

  doi:10.1098/rspb.2010.0303 · open source ↗

  What it actually establishes. A 2010 theoretical paper in Proceedings of the Royal Society B proposes describing animal affective states along two dimensions - valence (positive/negative) and arousal (activation level) - and argues that longer-term mood can be understood as accumulated short-term emotional states. It is a framework for designing research, not a study of dogs.

  Commonly overstated as. Three overreaches, in ascending seriousness. (1) Dog-training and dog-behaviour writing routinely converts the valence-arousal quadrant into a real-time diagnostic grid read off a dog's posture ('stiff body = high-arousal negative = the dog feels X'). The paper offers no mapping from observable behaviour to quadrant position. (2) The paper contains a figure placing discrete emotion LABELS inside the core-affect space, and this figure is frequently reproduced as if it were an emotion-identification chart. It plots words in a conceptual space; it is not an instrument and carries no observational criteria. (3) The paper is cited as evidence that dogs experience specific named emotions. It contains no dog data whatsoever and makes no species-specific claim. Also avoid describing it as a 'study' or 'research showing' anything - it is indexed as a Review and reports no new data.

  Limits: Theoretical review, zero new empirical data, zero dog-specific data; the scope is animals generally. Its principal derived measurement tool, the judgement-bias task, yields heterogeneous results across studies and species and its validation remains contested. Named conceptual criticism verified: Victor Carranza-Pinedo (2024), 'Rethinking core affect: the role of dominance in animal behaviour and welfare research', Synthese, doi 10.1007/s11229-024-04591-2, argues two dimensions are insufficient and that dominance should be reintegrated - specifically because the standard operationalisation (positive valence = approach, negative valence = avoidance) breaks down where 'liking' and 'wanting' dissociate. A separate recurring methodological criticism is that physiological indicators track arousal far more reliably than valence, so the valence axis is the weaker of the two in practice.

• T2miklosi2016Authored scholarly or clinical text

  Ádám Miklósi (2016). Dog Behaviour, Evolution, and Cognition. Oxford University Press (Oxford Biology series) — 2nd ed

  ISBN 9780198787778

  What it actually establishes. Ádám Miklósi's Dog Behaviour, Evolution, and Cognition (Oxford University Press, 2nd ed. 2016) is a widely used academic synthesis of research on dog domestication, behaviour and cognition, useful for tracing a popular claim back to the underlying studies — with the caveat that its coverage stops around 2015.

  Commonly overstated as. The dominant overreach is citing Miklósi as if he had proven that dogs possess a unique, human-like social intelligence — usually rendered as "dogs understand pointing as well as a two-year-old, so dogs think like toddlers." His own treatment is more cautious, presenting point-following as a specific domestication-shaped sensitivity to human communicative cues while flagging small samples, rearing-history confounds, and competing domestication hypotheses. The second and more common error is citing the book as primary evidence: it is a review, and any specific claim should be attributed to the study he cites, not to him.

  Limits: Secondary synthesis, not original data — every claim inherits the sample size and design limits of the study it summarises, and many foundational dog-cognition experiments used fewer than 30 dogs. Literature cutoff roughly 2015, so it predates later large-sample and multi-laboratory work in comparative cognition; I did not independently verify which specific pointing or social-referencing results have since been qualified, so do not attach named replication failures to this citation without checking them separately. No third edition was located in searches as of August 2026, which means anyone citing it for "current consensus" is citing a decade-old snapshot. First-edition dating is genuinely messy (2007 UK hardback, 2008 US, 2009 paperback), so cite the ISBN rather than a bare year.

• T4mills2020Documented clinical consensus

  Daniel S. Mills, Isabelle Demontigny-Bédard, Margaret Gruen, Mary P. Klinck, Kevin J. McPeake, Ana Maria Barcelos, Lynn Hewison, Himara Van Haevermaet, Sagi Denenberg, Hagar Hauser, Colleen Koch, Kelly Ballantyne, Colleen Wilson, Chirantana V. Mathkari, Julia Pounder, Elena Garcia, Patrícia Darder, Jaume Fatjó, Emily Levine (2020). Pain and Problem Behavior in Cats and Dogs. Animals (MDPI), vol. 10, issue 2, article 318, published 18 February 2020. Article type as labelled by the journal: Commentary.

  doi:10.3390/ani10020318 · open source ↗

  What it actually establishes. In a 2020 commentary in the journal Animals, nineteen veterinary behaviourists reviewed dog cases from their own referral clinics and reported that they suspected a painful condition in a substantial share of them — the authors offer roughly a third as a conservative estimate, though individual clinics ranged from 23% to 82%. On that basis the authors argue that pain should be actively ruled out before a change in behaviour is treated as purely behavioural.

  Commonly overstated as. Popular dog-training writing routinely inflates this into 'up to 80% of dog behaviour problems are CAUSED by pain' or 'a third of all dog behaviour problems are pain.' Four distortions, the fourth of which the previous verification missed: (1) the 80% is the single highest clinic in the table (Lincoln, 82%), quoted as if it were the headline result rather than the top of a 23–82% range; (2) these are specialist REFERRAL caseloads — dogs already escalated past their regular vet, heavily enriched for complex and medical cases — so the percentage cannot be applied to dogs at large; (3) the paper says pain is 'involved in' or 'associated with' the case, which the authors explicitly note is often an observed correlation, not a demonstrated cause — they write that 'some of these associations are simply observed correlations, and the exact relationship (if any) with pain needs further elucidation'; (4) the table's own column header is 'Proportion (%) of behavior cases where a painful condition was SUSPECTED' — clinician suspicion, not confirmed diagnosis. Any site wording that upgrades 'suspected' to 'involved', 'found', or 'caused by' is overstating the source's own language. A further trap: this is a commentary about BOTH cats and dogs, and the percentage table is dogs only.

  Limits: Self-described commentary, not a prevalence study: no random sampling, no control group, no blinding, and clinicians assessed their own cases. Data come from six referral clinics over 2017–2019 (Bristol 68%, Lincoln 82%, Pennsylvania A 23%, Pennsylvania B 29%, Missouri/Illinois 79%, Barcelona 28%) — a roughly fourfold spread showing there is no single defensible headline number. Confirmation of the suspected pain diagnosis ranged from 16% (Barcelona) to 79% (Bristol and Pennsylvania B), and confirmation methods were often physical exam or response to analgesia rather than imaging, so much of the attribution rests on clinical judgement rather than a demonstrated lesion. The abstract's phrase 'the caseloads of 100 recent dog cases of several authors' is genuinely ambiguous in the paper itself as to whether 100 was per contributing author or in total — do not state a precise total sample size on the site. Not independently replicated as a prevalence estimate.

• T1morrill2022Peer-reviewed research literature

  Kathleen Morrill, Jessica Hekman, Xue Li, Jesse McClure, Brittney Logan, Linda Goodman, Mingshi Gao, Yinan Dong, Marjie Alonso, Elena Carmichael, Noah Snyder-Mackler, Jacob Alonso, Hyun Ji Noh, Jeremy Johnson, Michele Koltookian, Charlie Lieu, Kate Megquier, Ross Swofford, Jason Turner-Maier, Michelle E. White, Zhiping Weng, Andrés Colubri, Diane P. Genereux, Kathryn A. Lord, Elinor K. Karlsson (2022). Ancestry-inclusive dog genomics challenges popular breed stereotypes. Science, vol. 376, no. 6592, article eabk0639 (American Association for the Advancement of Science), 29 April 2022

  doi:10.1126/science.abk0639 · open source ↗

  What it actually establishes. Across owner-reported behaviour scores for 18,385 dogs, breed accounted for roughly 9% of the variation between individual dogs, so breed is a weak predictor of how any one dog will behave - even though most of those behaviour scores were moderately heritable and breed-level differences were still detectable (Morrill et al., 2022, Science).

  Commonly overstated as. The most common overreach is 'science has proved breed has nothing to do with behaviour' or 'it is all how you raise them.' The paper reports the opposite about genetics: most behavioural factors were substantially heritable, breed-level differences were detectable, and breed ancestry carried real signal for traits such as biddability. The finding is about weak prediction for one individual, not absence of breed-linked genetic influence. A second overreach - one the ORIGINAL VERIFICATION being checked partly repeated - is citing it on aggression or bite risk. The nearest factor, agonistic threshold, was among the LEAST heritable and least breed-differentiated traits, and the study measured owner-reported provocation thresholds, not bites or bite severity; the paper is weakest exactly where a safety page would most want to lean on it. A third overreach is calling the 9% a per-trait figure; it is an aggregate across the eight behavioural factors.

  Limits: Behaviour is owner-reported and the cohort is self-selected: the authors warn that survey responses are susceptible to rater bias including breed stereotypes, and that survey data alone cannot separate environmental effects (including stereotype effects on how owners perceive and raise their dog) from genetic effects. The cohort skews toward US pet dogs and is not a random or working-dog sample. Only 2,155 dogs had genetic data and only 1,967 entered heritability estimation, which limits GWAS power - the 11 loci need independent confirmation. The ORIGINAL VERIFICATION's claim that 'no independent replication of comparable scale has been published' is wrong and should be removed: Salonen et al. (2023, iScience 26(5):106691, doi 10.1016/j.isci.2023.106691) surveyed over 11,000 dogs and reported breed as the single most important factor behind personality differences, and Pongracz & Dobos (2025, Biological Reviews 100, doi 10.1111/brv.13128) review 97 studies and argue breed-comparison designs in this field are frequently mismatched to their research questions. Those results are compatible with Morrill et al. - group-level breed differences and individual-level prediction are distinct questions - but a page citing the 9% figure must say so rather than imply the matter is closed. Contemporaneous peer commentary (Udell, Learning & Behavior 2022, doi 10.3758/s13420-022-00543-3) was broadly supportive while stressing the same individual-versus-group distinction.

• T2overall2013Authored scholarly or clinical text

  Karen L. Overall (MA, VMD, PhD, DACVB, CAAB) (2013). Manual of Clinical Behavioral Medicine for Dogs and Cats. Elsevier, St. Louis, MO — 1st edition, 2013, with companion DVD of questionnaires and client handouts. Imprint is listed inconsistently across catalogues: the Australian Veterinary Journal review and most print listings say Elsevier Mosby; some e-book listings say Saunders. Page count also varies by source — 812 pp. per the AVJ review, 832 pp. per publisher e-book listings. Use 'Elsevier, 2013' on the site and avoid asserting a page count.

  ISBN 9780323008907

  What it actually establishes. Karen Overall's Manual of Clinical Behavioral Medicine for Dogs and Cats (Elsevier, 2013) is a widely used single-author clinical reference in veterinary behavioural medicine, setting out diagnostic criteria and treatment protocols for behavioural conditions in dogs and cats.

  Commonly overstated as. The commonest overreach is citing 'Overall 2013' as though it were an experimental study — 'research shows...' or 'a study by Overall found...' — when it is an authored clinical textbook. In practice this most often attaches to her named protocols (the Protocol for Relaxation and the Protocol for Deference), which get described online as 'clinically proven' or 'evidence-based' treatments; they are expert clinical protocols, and the book does not present controlled-trial efficacy data for them. A second version: her well-argued clinical position against confrontational and dominance-based handling is quoted as if it were a meta-analytic finding produced by the book, rather than a clinician's reasoned position (a position separately supported elsewhere in the literature, but not by primary data in this volume). A third, subtler one to avoid: do not write that this book is what veterinary behaviourists 'work from' or that its protocols are 'the standard of care' — that is a claim about professional practice worldwide that the book cannot itself establish. 'Widely used reference' is defensible; 'the standard' is not.

  Limits: Single author, single edition — 1st edition, 2013, now over a decade old; the psychopharmacology chapters, drug dosing, and cognitive-dysfunction material are the parts most likely to be superseded. No systematic-review methodology and no formal evidence grading. Largely US clinical context. Reviewed in the Australian Veterinary Journal by P. Tucak, 2014, vol. 92 issue 7 p. 234 (doi:10.1111/avj.12192) — that is a book review, not independent validation of its clinical claims, and it is a single short review rather than a body of critical appraisal. Cite this book for definitions, clinical frameworks and description of practice, never as the evidence base for an efficacy claim.

• T1pongracz2005Peer-reviewed research literature

  Péter Pongrácz, Csaba Molnár, Ádám Miklósi, Vilmos Csányi (2005). Human Listeners Are Able to Classify Dog (Canis familiaris) Barks Recorded in Different Situations. Journal of Comparative Psychology (American Psychological Association), 119(2), 136–144

  doi:10.1037/0735-7036.119.2.136 · open source ↗

  What it actually establishes. Asked to match recorded barks to one of six listed situations, Hungarian listeners were right between 23% and 58% of the time depending on the situation, against a 17% chance rate (Pongrácz et al. 2005; barks from 19 dogs of one breed).

  Commonly overstated as. Routinely rewritten as "humans can reliably understand what a dog's bark means" or "dogs evolved a bark vocabulary for humans." The study shows above-chance guessing on a closed six-option menu from a single breed, and listeners were wrong more often than right in every context. The frequently quoted "about 40%" is an average across six contexts, not a headline result; two of the six (walk 23.15%, ball 25.00%) sat barely above the 16.7% chance line. Nothing here shows a listener can interpret a bark heard cold, without being handed the list of possible answers. Lord et al. 2009 state the point directly: guessing context at 23% or even 58% may be statistically significant but does not indicate dog–human co-adaptation.

  Limits: Barks from 19 Mudis only; three listener groups (Mudi owners, other dog owners, non-owners); forced choice among the six contexts the barks were actually drawn from, so open-ended interpretation was never tested; no independent measure of the dogs' internal states. Partially replicated in Portugal — Silva, Faragó, Pongrácz, Romeiro, Lima & de Sousa (2021), Animal Behavior and Cognition 8(2):107–123, doi:10.26451/abc.08.02.02.2021 verified via Crossref, n=51 (26 owners, 25 non-owners): the same confusion pattern appeared, but overall accuracy was significantly lower than the Hungarians'; only 'stranger' and 'alone' beat chance, 'Schutzhund', 'walk' and 'play' did not differ from chance, and 'ball' was categorised significantly BELOW chance. Directly criticised by Lord, Feinstein & Coppinger (2009), 'Barking and mobbing', Behavioural Processes 81(3):358–368, doi:10.1016/j.beproc.2009.04.008 (verified via Crossref), who argue the contextual variation is entirely accounted for by Morton's motivational-structural rules and arousal level, and gives no support to a referential hypothesis.

• T1pongracz2006Peer-reviewed research literature

  Péter Pongrácz, Csaba Molnár, Ádám Miklósi (2006). Acoustic parameters of dog barks carry emotional information for humans. Applied Animal Behaviour Science, 100(3–4), 228–240 (Elsevier)

  doi:10.1016/j.applanim.2005.12.004 · open source ↗

  What it actually establishes. When bark sounds were spliced into sequences varying in pitch, roughness and repetition rate, people's emotion ratings tracked those sound properties in the direction predicted for mammal calls generally (Pongrácz et al. 2006).

  Commonly overstated as. Marketed as "science has decoded dog barks," and used to justify bark-translation apps and pitch rules of thumb ("a low bark always means aggression"). The paper measures how humans rate artificially constructed sound sequences; no independent measure of any dog's internal state was taken, and the sequences were not natural utterances. It is a finding about human perception being governed by cross-species acoustic rules, which is close to the opposite of dogs possessing a species-specific bark code. A specific caution against the tidy three-way rule: the paper's own abstract states that high-pitched sequences with long inter-bark intervals were rated happy and playful INDEPENDENTLY of their tonality, so the popular clean split of tonal=fearful versus atonal=playful is not cleanly supported by the source.

  Limits: Same single-breed (Mudi) source recordings as the 2005 study, so the two papers are not independent evidence. Ratings are subjective Likert scores on anthropocentric emotion labels ('despair', 'happiness') whose mapping to a motivational continuum is questioned by Lord et al. 2009 (doi:10.1016/j.beproc.2009.04.008). CORRECTION to the prior verification: the claim that this study used 'the same three Hungarian listener groups' as 2005 could not be confirmed from any source I could reach — the paper describes participants only as people with differing levels of dog experience. Full text is paywalled; I confirmed metadata via Crossref and design/results via the paper's own abstract plus descriptions in Silva et al. 2021 and Lord et al. 2009, so treat fine methodological detail as secondhand.

• T1quaranta2007Peer-reviewed research literature

  A. Quaranta, M. Siniscalchi, G. Vallortigara (2007). Asymmetric tail-wagging responses by dogs to different emotive stimuli. Current Biology, 17(6), R199-R201 (Correspondence; indexed by PubMed and Europe PMC with publication type "Letter")

  doi:10.1016/j.cub.2007.02.008 · open source ↗

  What it actually establishes. In one 30-dog laboratory study, frame-by-frame video measurement found a group-level rightward bias in tail-wag amplitude when dogs viewed their owner, an unfamiliar person or a cat, and a leftward bias when they viewed an unfamiliar dominant dog (Quaranta et al., 2007). The bias is a statistical average extracted from video, not something a person can read off a live dog.

  Commonly overstated as. Almost universally sold as a real-time owner-usable test: "wag right means happy, wag left means scared." Three separate problems. First, the effect is a group-level amplitude bias recovered by frame-by-frame analysis and is not visible to the naked eye. Second, the construct is approach versus withdrawal motivation, not happiness versus sadness - the cat produced the same rightward bias as the owner, so "right = happy about you" is not what was measured. Third, and fatal to the simple valence reading, the no-stimulus control also produced a significant left bias, meaning an empty aperture patterned the same way as the threatening dog. Any site presenting this as a wag-direction decoder is misrepresenting it.

  Limits: n=30 (15 intact males, 15 intact non-oestrus females), mixed breed, ages 1-6, one laboratory, one research group, short Correspondence format. Replication is weaker than the prior entry implied and I am downgrading the confidence accordingly. Simon, Guo, Frasnelli, Wilkinson & Mills 2022, Neuroscience & Biobehavioral Reviews 143:104950, DOI 10.1016/j.neubiorev.2022.104950, a critical review of exactly this literature, concludes that previous studies "have not been able to discriminate well between competing hypotheses and tended to focus on confirmation bias." The same group's Simon et al. 2024, Animal Behaviour 216:155-173, DOI 10.1016/j.anbehav.2024.08.005, found no population-level tail-wagging asymmetry during positive anticipation of food or a toy, only individual-level biases. Leonetti, Cimarelli, Hersh & Ravignani 2024, Biology Letters 20(1):20230407, DOI 10.1098/rsbl.2023.0407, restates the asymmetry claim approvingly but is a narrative review by rhythm researchers, not an independent test, and does not engage the null results. Treat the production-side asymmetry as suggestive and context-dependent, not established.

• T1romero2013Peer-reviewed research literature

  Teresa Romero, Akitsugu Konno, Toshikazu Hasegawa (2013). Familiarity Bias and Physiological Responses in Contagious Yawning by Dogs Support Link to Empathy. PLoS ONE, 8(8), e71365

  doi:10.1371/journal.pone.0071365 · open source ↗

  What it actually establishes. Dogs do yawn more after watching a person yawn than after a control mouth movement - a contagion effect since supported by a Bayesian reanalysis of 257 dogs across six studies - but the same reanalysis found no familiarity, gender or prosociality bias, so the empathy explanation that Romero et al. proposed is not supported (Romero et al., 2013; Neilands et al., 2020).

  Commonly overstated as. Routinely presented as settled proof that dogs catch yawns because they empathise with their owners. The specific finding that carried the empathy interpretation - yawning more for a familiar person - has failed to replicate in a larger pooled analysis. Cite Neilands et al. 2020 for the contagion effect itself, which survives; do not cite Romero et al. 2013 as evidence of canine empathy. A site that does is citing a contested and largely superseded interpretation.

  Limits: SERIOUS - do not cite unqualified. n=25 dogs, 21 with valid heart-rate data, and only 13 of the 25 yawned at all. Effects are tiny: mean 1.0 yawns (sd 1.5) in the yawn condition versus 0.2 (sd 0.4) in control, with borderline p-values (contagion beta=1.309, P=0.025; familiarity beta=0.034, SE=0.015, t=2.197, P=0.032). The familiarity finding is directly contradicted by Neilands, Claessens, Ren, Hassall, Bastos & Taylor 2020, Proceedings of the Royal Society B 287(1920):20192236, DOI 10.1098/rspb.2019.2236, which reanalysed 257 dogs across six studies - Romero et al. 2013, Joly-Mascheroni et al. 2008, Silva et al. 2012, Madsen & Persson 2013, O'Hara & Reeve 2011, Buttner & Strasser 2014 - plus a new 32-dog prosociality experiment, and found robust support for contagion but none for the familiarity or gender biases the empathy hypothesis predicts. Earlier nulls: Harr, Gilbert & Phillips 2009, Animal Cognition 12:833-837, DOI 10.1007/s10071-009-0233-0, which found no contagion at all (superseded on that specific point by the Neilands pooled result); and O'Hara & Reeve 2011, Animal Behaviour 81:335-340, DOI 10.1016/j.anbehav.2010.11.005, where familiarity, similarity and bonding all failed to predict yawning.

• Prugaas2006Practitioner framework, not peer-reviewed

  Turid Rugaas (2006). On Talking Terms With Dogs: Calming Signals. Dogwise Publishing, Wenatchee, WA — 78 pp.

  ISBN 9781929242368

  What it actually establishes. The term "calming signals" was coined by Norwegian dog trainer Turid Rugaas in On Talking Terms With Dogs (Dogwise, 2nd ed. 2006) for about thirty low-arousal behaviours she observed in dogs during social tension; the list is a trainer's observational taxonomy, not a validated instrument, and only two small studies have since tested whether these behaviours behave like signals.

  Commonly overstated as. The dominant overreach is treating the list as scientifically validated and as a one-to-one emotional decoder — "a yawn means your dog is trying to calm you down." Note also a correction to a claim often made in the other direction: it is NOT true that no dog-human study exists. Firnkes et al. (2017) ran a standardised behavioural test on 116 dogs with unfamiliar humans and found looking away and lip licking occurred significantly more in socially direct situations (friendly greeting, threatening stare), which the authors read as appeasement in dog-human communication. What remains undemonstrated is the calming EFFECT on the receiver — both studies measure when the behaviours are produced, not that they reliably change the other party's state. Separately, most of these behaviours appear in peer-reviewed ethology under non-equivalent labels (displacement behaviours, appeasement signals, stress indicators) where the assumption is that they reflect the sender's internal conflict rather than a deliberate message. Reading one signal in isolation rather than as a cluster in context is the practical failure mode.

  Limits: The book has n=0 formal subjects and no controls. Mariti et al. 2017 is explicitly a pilot: 24 senders (12 male, 12 female), each observed across four 5-minute off-leash encounters (96 encounters), dog-dog only. In 67.0% of cases (N=73) at least one calming signal followed an aggressive act by the recipient, and in 79.4% of those (N=58) the recipient's aggressive display de-escalated — correlational, not causal, since de-escalation could follow from elapsed time or the recipient's own state. Firnkes et al. 2017 is larger (116 dogs) but is a standardised test with muzzles and leads, measures signal production rather than receiver response, and found lip licking did NOT rise as expected during threatening screaming or physical threat. Neither has been replicated at scale. Rugaas's ~30-signal set has never been validated as complete or discrete. Do not conflate Mariti et al. 2017 with the earlier 2014 Mariti conference paper of a similar title.

• T2shepherd2009Authored scholarly or clinical text

  Kendal Shepherd (chapter author); Debra F. Horwitz and Daniel S. Mills (book editors). Ladder illustration by the late Maggy Howard. (2009). Behavioural medicine as an integral part of veterinary practice. Chapter 2 in BSAVA Manual of Canine and Feline Behavioural Medicine, 2nd edition, eds D.F

  doi:10.22233/9781905319879.chap2 · open source ↗

  What it actually establishes. The Ladder of Aggression is a teaching diagram published by veterinary surgeon Kendal Shepherd in the BSAVA Manual of Canine and Feline Behavioural Medicine (2nd edn, 2009). It illustrates the range of behaviours a dog may show under escalating perceived threat, from appeasement signals such as lip licks and head turns through to growling, snapping and biting, and makes the clinical argument that punishing or ignoring the quieter signals can teach a dog to stop using them. Shepherd describes it as a context- and response-dependent sequence, not a fixed order every dog follows.

  Commonly overstated as. By far the commonest overreach is presenting the ladder as a validated, fixed sequence every dog climbs rung by rung — 'dogs always warn you in this exact order, so no bite is ever out of the blue.' Shepherd's own framing is explicitly context- and response-dependent: dogs can skip rungs entirely and go straight to a bite, and a dog showing many lower rungs may never escalate at all. The second, more dangerous overreach is using the ladder as a predictive risk instrument — reading a dog's current rung as a measure of how close it is to biting. It was never designed or validated for that, and a site that implies it can be used that way is giving readers a false safety signal. A third overreach specific to how it is often published: describing it as 'the BSAVA ladder' or otherwise implying it is a professional body's official guidance rather than one clinician's diagram that BSAVA published. A fourth: the widely repeated claim that 'dogs punished for growling bite without warning' is a clinical inference from this chapter, not a controlled finding.

  Limits: No sample size, no dataset, no statistics — it is a conceptual diagram, and there is little published empirical validation of the ladder in its current form. It was developed in a veterinary-clinic handling context, so its ordering should not be assumed to transfer to other contexts. Documented criticisms include that intense aggression can be preceded by a single signal, that a dog displaying many signals may never escalate to a bite, and that a snap sits low on the ladder yet is not characteristic of seriously aggressive interactions. Signal legibility also varies with morphology — brachycephalic, cropped, docked or heavily coated dogs cannot produce some of these signals visibly. On the frequently cited 2025 counter-evidence: the paper is real — Shnookal J., Clay L., Howell T. et al., 'Exploring trainer insights and understanding of on-lead dog aggressive behaviours and behavioural modification advice', Discover Animals vol. 2, article 59, 2025, doi:10.1007/s44338-025-00119-1 — but it is a QUALITATIVE INTERVIEW STUDY OF TEN AUSTRALIAN DOG TRAINERS, and the specific claim previously attributed to it (that trainers' accounts never mentioned mid-ladder items such as sitting/pawing or lying down/leg up) could NOT be verified in the accessible text. Ten trainers' recollections are in any case far too thin a basis to cite as evidence about how dogs actually sequence behaviour. Do not use this paper on the site as a refutation of the ladder; at most it shows practitioners describe on-lead aggression in their own terms.

• T1siniscalchi2013Peer-reviewed research literature

  Marcello Siniscalchi, Rita Lusito, Giorgio Vallortigara, Angelo Quaranta (2013). Seeing Left- or Right-Asymmetric Tail Wagging Produces Different Emotional Responses in Dogs. Current Biology, 23(22), 2279-2282

  doi:10.1016/j.cub.2013.09.027 · open source ↗

  What it actually establishes. When 43 dogs watched video and silhouette images of another dog wagging with a leftward bias, they showed higher cardiac activity and more fear-related behaviour than when the same figure wagged rightward (Siniscalchi et al., 2013) - a difference in arousal to an artificial stimulus, not evidence that dogs read wag direction during real encounters.

  Commonly overstated as. Repackaged as "dogs read each other's tail wags to tell friend from foe," then sold to owners as a left-bad/right-good decoder for live dogs. Elevated heart rate indexes arousal, not valence. The stimuli were videos and silhouettes, not live conspecifics, so nothing here shows wag direction functions as a signal in natural interaction. And nothing in the design licenses a human observer applying it - the discriminating party in this experiment was a dog watching a screen.

  Limits: n=43. Same research group as the 2007 original, so this is an extension by the original authors, not independent replication. The one independent test of the perception side reached a differently patterned result, and I verified its abstract verbatim: Artelle, Dumoulin & Reimchen 2011, Laterality 16(2):129-135, DOI 10.1080/13576500903386700, used a life-size robotic dog replica across 452 interactions and found "a significantly greater proportion of dogs approaching the model continuously without stopping when the tail wagged to the left, compared with a right wag," concluding that "the responses are not readily integrated into the predicted behaviour based on hemispheric dominance." That is the opposite direction from the naive decoder reading. Add the broader methodological critique in Simon et al. 2022, Neuroscience & Biobehavioral Reviews 143:104950, DOI 10.1016/j.neubiorev.2022.104950. Present the perception claim as unsettled and never as an owner-facing tool.

• T1yin2002Peer-reviewed research literature

  Sophia Yin (2002). A New Perspective on Barking in Dogs (Canis familiaris). Journal of Comparative Psychology (American Psychological Association), 116(2), 189–193

  doi:10.1037/0735-7036.116.2.189 · open source ↗

  What it actually establishes. Barks recorded when a stranger rang the doorbell were on average lower-pitched and longer than barks from the same dogs when isolated or playing (Yin 2002, 10 dogs, 4,672 barks).

  Commonly overstated as. Cited as proof that dogs have distinct barks with distinct meanings, and as licence for pitch-reading rules ("a low bark means threat, a high bark means play"). Two things break that. The study could not separate isolation barks from play barks at all — statistically indistinguishable, so the same high-pitched bark came from both a lonely dog and a happy one. And the author herself writes that confirming any function requires playback studies of listener response, which this paper does not report. It is an argument that barking deserves study, not a demonstration that barks are decodable.

  Limits: 10 dogs, six breeds, one lab; only two acoustic variables analysed (mean dominant frequency, duration). Significant dog × context interactions were found for both measures: 1 of 10 dogs varied opposite the group pattern on frequency and 2 of 10 on duration, though the deviation was described as relatively small. Contexts were staged, with no independent measure of the dogs' internal states. The fuller analysis is the companion paper Yin & McCowan, 'Barking in domestic dogs: context specificity and individual identification', Animal Behaviour 68(2):343–355, doi:10.1016/j.anbehav.2003.07.016 (verified via Crossref) — cite that one, not this paper, for any discriminant-function classification claim; note Yin 2002 itself cites it as 'Yin & McCowan 2002' because it was then in press. Lord, Feinstein & Coppinger (2009, doi:10.1016/j.beproc.2009.04.008) accept that Yin showed barks vary with context but argue the variation reflects motivational state under Morton's rules and a general mobbing function, not context-specific reference. Sophia Yin died in 2014; cattledogpublishing.com hosts the free full-text PDF.

• T2yin2009Authored scholarly or clinical text

  Sophia A. Yin, DVM, MS (2009). Low Stress Handling, Restraint and Behavior Modification of Dogs & Cats: Techniques for Developing Patients Who Love Their Visits. CattleDog Publishing, Davis, CA — approx

  ISBN 9780964151840

  What it actually establishes. Sophia Yin's Low Stress Handling, Restraint and Behavior Modification of Dogs & Cats (CattleDog Publishing, 2009) is a widely used technique manual for handling animals with minimal restraint and food-based counter-conditioning during veterinary and grooming procedures; a fully revised second edition appeared in 2024.

  Commonly overstated as. The common overreach is citing Yin as the evidence base — "studies show low-stress handling reduces cortisol" attributed to this book. The book is a manual with no controlled outcome data. IMPORTANT AUTHOR CORRECTION: the physiological trial usually paired with this claim is NOT by "Edwards et al." It is Squair, C., Proudfoot, K., Montelpare, W. & Overall, K. L., "Effects of Changing Veterinary Handling Techniques on Canine Behaviour and Physiology Part 1: Physiological Measurements," Animals 2023, 13(7):1253, DOI 10.3390/ani13071253 (authorship verified via Crossref and PMC). Attributing it to Edwards is a fabricated citation and would fail any reviewer's check. Cite Squair et al. for the physiology and Yin for the technique.

  Limits: No sample, no control group, no statistics — it is a clinical text. The supporting trial is small and its result is narrower than usually reported: 28 dogs randomised into 14 intervention and 14 control across four visits; BOTH groups' serum cortisol fell by visit four, with the intervention group showing a greater decrease from visit one to visit four (p < 0.04), and the composite stress response index differing significantly across visits for the intervention group (p = 0.03) but not controls (p = 0.288). Habituation across repeat visits is a live alternative explanation for part of that, and the paper describes "low-stress handling techniques and collaborative care" without naming Fear Free. Date accuracy matters here: Yin died in September 2014, so the 2009 text is fixed in time; CattleDog issued a second edition on 12 November 2024, retitled Dr. Sophia Yin's Low Stress Handling® and Behavior Modification of Dogs and Cats (note "Restraint" dropped) and updated by Pat Koven, LVT, KPA-CTP and Christine Calder, DVM, DACVB, moving further toward minimal restraint, cooperative care and choice-and-control, with revised feline guidance. Citing a 2009 restraint technique as current best practice is a real scope error. "Low Stress Handling" is a registered trademark, which constrains commercial use of the term.

• T1ziv2017Peer-reviewed research literature

  Gal Ziv (2017). The effects of using aversive training methods in dogs—A review. Journal of Veterinary Behavior: Clinical Applications and Research, volume 19, pages 50–60

  doi:10.1016/j.jveb.2017.02.004 · open source ↗

  What it actually establishes. A 2017 narrative review of 17 studies in the Journal of Veterinary Behavior concluded that aversive training methods can jeopardise dogs' physical and mental health, and found no evidence that they are more effective than reward-based training.

  Commonly overstated as. It is commonly described as a meta-analysis that 'proved aversive training harms dogs.' It is a narrative review: it pooled no effect sizes, computed no summary statistic, and applied no formal risk-of-bias tool. Most included studies were surveys or observational rather than controlled experiments, so the review inherits their inability to establish causation. Its conclusion is that harm CAN occur, not that it invariably does. A further common distortion is dropping Ziv's own concession that positive punishment can be effective — the finding is about relative effectiveness and cost, not about aversives failing outright.

  Limits: 17 studies, many with small samples, inconsistent operational definitions of 'aversive,' frequently absent effect sizes, possible coding bias in the observational work, and little long-term follow-up — limitations the author himself notes. Cited roughly 127 times (SciSpace), not 190+. CORRECTION TO THE PRIOR REVIEW: the earlier verification described a 'competing review' in Applied Animal Behaviour Science 2017 (volume 196, pages 1–12, 'Do aversive-based training methods actually compromise dog welfare?: A literature review') as drawing 'markedly weaker conclusions' from overlapping literature. I verified that paper's existence, title, journal, volume and pages, and its authorship: Guilherme-Fernandes, Olsson and Vieira de Castro. Two things in the prior framing are wrong. First, it is not an independent competitor — it is by the same Porto research group that then produced the 2020 PLOS ONE study cited elsewhere in this set. Second, its conclusion runs in the SAME direction, not against: it concludes that while scientific evidence on training method and welfare is limited, the existing literature indicates that at least at some level aversive-based methods generate stress in dogs. The honest caveat is that both reviews describe the evidence base as thin, not that the field is split. I did not verify that paper's DOI, so it is deliberately not supplied here.

Identifiers are printed as text as well as linked, so this page remains useful offline and if a link rots. Proprietary materials are linked rather than reproduced. Spotted an error?