Resource Guarding in Dogs: Definition, Prediction, Assessment and Treatment

1. What the Term Means

1.1 The Definitional Problem

Possessive aggression and resource guarding have been used interchangeably in clinical and research writing, with different authors specifying different things — some by the item involved, some by the behaviours shown, some separating food-related from object-related guarding as distinct categories (Jacobs, Coe, Widowski, Pearl & Niel, 2018c).

Inconsistent terminology affects communication, treatment success and the comparability of research (a recurring measurement problem).

1.2 The Consensus Definition

A structured expert-opinion study produced a working definition: the use of avoidance, threatening, or aggressive behaviours by a dog to retain control of food or non-food items in the presence of a person or other animal (Jacobs et al., 2018c).

1.3 Why That Wording Matters

Three features carry weight. Avoidance is included, which places running off with an item or eating faster inside the definition rather than outside it. Retain control identifies the function rather than the emotion, which keeps the definition usable without inferring how the dog feels. And in the presence of a person or other animal makes it social by definition — a dog alone with a bone is not guarding.

The authors present the definition as provisional and invite modification as research accumulates — a qualification most canine definitions arrive without.

1.4 Resource Value Is Not Fixed

One implication is easy to miss: the dog determines what counts as a resource, and the assessment can be counter-intuitive.

Value varies with hunger, scarcity, novelty, competition and learning history. A stolen tissue can be functionally more valuable than a full bowl — partly because it is novel, partly because previous attempts to retrieve such items taught the dog what an approaching hand predicts.

This is why "he only guards high-value things" is not a stable description: it describes what the owner considers valuable.

1.5 Value Is Relative, Not Absolute

The stronger version of the point concerns variation within the same dog and the same item. A chew that produced no reaction on Monday can be defended on Thursday, without the chew having changed.

Hunger shifts it. So does the presence of a competitor, a period of poor sleep (with sleep affecting emotional regulation), physical discomfort, and whether anything has recently been taken. Value is a relation between the dog's current state and the item, not a property of the item.

That is the central reason guarding works poorly as a stable trait description — and, as section 8 shows, the reason a single test occasion carries less information than it appears to.

2. What Dogs Actually Guard

2.1 Beyond Food and Toys

The research base concentrates on food, because food is what shelter assessments test and what surveys ask about. The definition is broader, and so is the clinical picture.

Dogs guard chews and long-lasting items, stolen objects, resting places, doorways and thresholds, spaces under furniture, and proximity to a particular person (with attachment shaping which person). Some of these look like resources from outside; several do not.

2.2 Where the Evidence Is and Is Not

Food guarding is by far the best documented — the shelter literature, the assessment work and much of the survey data address it directly. Object guarding is covered in the owner surveys. Guarding of resting places and of proximity to people appears in clinical description and expert consensus but has almost no dedicated empirical work.

This matters for how confidently anything can be said. A statement about food guarding rests on multiple studies; a statement about a dog guarding its owner's lap rests on clinical experience and inference from the same functional account.

2.3 Why the Distinction Is Practical

Interventions and management transfer unevenly. Removing food and toys from a daycare group removes two categories of trigger and leaves resting places, doorways and human proximity untouched — which is management of the measurable subset rather than of the behaviour.

The same applies to assessment: a test built around a food bowl says nothing about a dog that guards the sofa (with the general problem of tests measuring something narrower than intended documented separately).

2.4 The Multi-Dog Household

One environmental factor emerged clearly from the survey data: dogs living in multi-dog households were more likely to display guarding aggression, avoidance and rapid ingestion than dogs living without other dogs (Jacobs et al., 2018b).

The finding is cross-sectional and open to more than one reading. Competition raises the value of a contested item, which would raise guarding directly. Alternatively, households that already contain several dogs may differ systematically in feeding arrangements, space and supervision. And selection runs the other way too — owners of a guarding dog may be less likely to add a second.

What can be said is that the presence of other dogs is associated with more guarding across all three behaviour classes, not only the aggressive one — which is consistent with a competition account and inconsistent with the idea that guarding is primarily about the human relationship.

3. The Behaviour Range

3.1 What It Covers

At the low-visibility end: freezing, stiffening, body-blocking, moving away with the item, rapid ingestion. In the middle: hard staring, lip curling, growling. At the visible end: snapping and biting.

Read functionally, these are not different problems. Moving away with a chew and growling over it can serve the same end — retaining control — through very different costs to the dog (which is why topography alone does not identify function).

3.2 What Owners Actually Recognise

Owners were shown video of shelter assessments and asked which behaviours were present. Aggression was identified accurately: 85 percent noticed it when present, 96 percent correctly reported its absence. Threatening behaviour — a category combining growling, freezing, body tension and teeth-baring — was recognised reasonably well when present but poorly when absent, and avoidance and rapid ingestion were consistently under-identified (Jacobs, Pearl, Coe, Widowski & Niel, 2017). The category structure matters: growling and freezing were not scored separately, so the finding concerns threat behaviour as a whole.

3.3 The Consequence

An owner who recognises guarding only when a dog snaps first learns about it at the point of injury (a general problem with reading canine signals). This offers one explanation for a common history: the dog that "bit without warning". On that reading the warnings were given and not seen — and if earlier warnings were punished, they may have been suppressed rather than absent (with the fallout documented).

It also means reported prevalence is likely to underestimate the full behavioural range. If the early forms are missed by owners, they are missed by the surveys that ask owners — which affects how much of the definition the figures actually cover, not only how many dogs they count.

4. Learning History and the Expectation of Loss

4.1 Two Processes, Not One

Guarding is usually described as either emotional or learned. It is both, and the components do different work.

The classical component concerns what an approach comes to predict. If a hand reaching toward a bowl has reliably preceded the bowl being removed, the approach itself acquires meaning — and the emotional response occurs before any decision is made. The operant component concerns what the dog's behaviour achieves. If stiffening, growling or moving away has reliably ended the approach, that behaviour is reinforced by the outcome (with both systems described separately).

4.2 Why Separating Them Matters

A dog whose growl has repeatedly worked will growl earlier and more readily. Suppressing the growl removes the operant component without touching the classical one — the approach still predicts loss, and the dog still needs an answer to it. The answer that remains is the one that was never punished.

That is the mechanism behind the standard warning against punishing a growl, stated in learning terms rather than as an appeal to communication ethics.

4.3 Escalation Is Not a Personality Change

An owner who reports that guarding "came out of nowhere" or "got much worse suddenly" is often describing an ordinary learning trajectory that became visible at the point where the cost to the dog rose. The early forms were present and unread.

4.4 What Functional Analysis Showed

A single-case in-home study is the only work to have tested the functional account directly. Functional analysis of one dog's human-directed guarding indicated that the behaviour was multiply controlled — maintained by escape, attention and tangible consequences simultaneously (Mehrkam, Perez, Self, Vollmer & Dorey, 2020).

That result is worth carrying forward for a specific reason: it means asking "what is the dog getting out of this" may have more than one answer in the same dog, and an intervention addressing only one contingency may leave the others intact.

4.5 How the Trigger Set Widens

A dog that has learned what an approaching hand predicts around a bone has not learned something about bones. Fear learning generalises to stimuli resembling the original, and to the context in which it occurred (as the conditioning literature sets out).

That is a plausible route from a narrow beginning to the picture owners describe as the dog "guarding everything": first the chew, then any long-lasting item, then the space where such items are consumed, then the approach itself regardless of what the dog has. Each step is ordinary generalisation.

It is worth marking as an account rather than a finding. No canine study has tracked the widening of a guarding repertoire over time, and the trajectory is inferred from what is known about fear learning generally.

5. What Predicts It

5.1 Guarding From People

A cross-sectional survey of dog owners examined factors associated with guarding in the presence of people. Dogs with higher levels of impulsivity and fear were significantly more likely to display resource guarding aggression (Jacobs, Coe, Pearl, Widowski & Niel, 2018a).

Both findings point away from the popular framing. Impulsivity relates to inhibitory control (which has its own developmental and physiological basis), and fear is an emotional state, not an assertion of rank.

5.2 Guarding From Other Dogs

A parallel survey screened 3,068 owners for their ability to identify guarding forms from video and analysed 3,589 dogs from the 2,207 who passed. Dogs in multi-dog households were more likely to show guarding aggression, avoidance and rapid ingestion; impulsivity and fear again predicted aggression; and neutered males were more likely to be aggressive than dogs of other sexes and neuter statuses (Jacobs, Coe, Pearl, Widowski & Niel, 2018b).

The authors describe patterns around other dogs as relatively more fixed across resources and situations than patterns around people — a group-level observation, not a prediction about an individual.

5.3 What This Does to the Dominance Account

Neither survey found what a dominance model predicts. Guarding is associated with impulsivity and fear, differs in structure depending on who the dog is guarding from, and concerns access to a specific item rather than status (with the dominance construct addressed in full elsewhere).

Neither survey assessed pain either — which matters, because the clinical data in section 7 suggest it belongs in the model (including internal sources that present as irritability).

5.4 Prediction Is Not Explanation

Fear and impulsivity predict guarding statistically across large samples. Neither explains why a particular dog guards a particular item in a particular situation.

Three levels are worth keeping apart. Risk factors — impulsivity, fear, multi-dog household, neuter status — shift probability across a population. Immediate triggers — an approach, a specific person, spatial confinement — determine whether the behaviour occurs now. Learning history determines what the approach has come to predict for this dog.

A cross-sectional survey establishes the first and is silent on the other two, which matters clinically: an individual case is solved at the second and third levels.

6. Why the Same Dog Guards Inconsistently

6.1 The Observation

Owners describe it constantly. The dog guards from one person and not another, guards in the kitchen and not the garden, guards on some evenings and not others. This is often read as unpredictability, and it is not.

6.2 Who Is Approaching

Guarding is directed, and the dog's history with each individual differs. A person who has repeatedly removed items is not equivalent to one who has not, regardless of how either behaves now.

Children are a distinct case for reasons that go beyond history: they move unpredictably, approach at the dog's eye level, and are less likely to interpret or respond to early signals (which the bite data in section 7 bear out).

6.3 Where It Happens

Spatial confinement removes the low-cost options. A dog that can leave with an item does not need to defend it; a dog in a corner, a crate or a narrow corridor cannot (which is the approach–avoidance situation in a specific form).

That is why guarding often appears more severe in confinement, and why the same dog may show avoidance at home and aggression in a kennel — a point that returns in section 8.

6.4 What Else Is Going On

Competition raises value. Fatigue, illness and pain lower the threshold at which defence becomes the available option (with pain heavily represented in behaviour caseloads). Accumulated arousal from earlier in the day narrows the behavioural repertoire (with the arousal constraint documented).

6.5 Why This Matters for Everything Else

If guarding varies systematically with approacher, location, physical state and recent history, then any single observation samples one combination of those. That is the reason a shelter test generalises poorly, and the reason a household intervention has to specify conditions rather than prescribe a technique.

7. Bite Risk

7.1 The Clinical Series

A retrospective case series examined records of dogs referred to a university veterinary hospital for aggression that had bitten a child, covering 111 cases (Reisner, Shofer & Nance, 2007).

7.2 Who Gets Bitten Over What

Children under six were most commonly bitten in association with resource guarding — 44 percent of cases — while older children were most commonly bitten in association with territory guarding, at 23 percent.

The split by familiarity is sharper still: food guarding was the most common circumstance for bites to familiar children, at 42 percent, against territory guarding at 53 percent for unfamiliar children (Reisner et al., 2007). In this clinical series, food guarding was associated primarily with bites to familiar children rather than unfamiliar ones.

7.3 Why Young Children Specifically

The study reports the association; the mechanisms are inference from what is known more generally, and are worth stating as such.

Small children are at the dog's height and often approach face-first. They move unpredictably and reach without signalling. They are present in exactly the locations where guardable resources sit — the floor, the dog's bed, the space under the table. They handle food themselves and drop it. And they are poor readers of canine warning signals, which removes the function of the early behaviours entirely: a dog that stiffens at an approaching adult may achieve distance, while a dog that stiffens at a toddler achieves nothing.

That last point connects the two findings in this section. The behaviours that work with adults do not work with small children, which leaves the dog with the ones that do.

7.4 What the Screening Found

Behavioural screening of the 103 dogs examined identified resource guarding in 61 percent and discipline measures in 59 percent as the most common stimuli for aggression. Anxiety screens revealed abnormalities in 77 percent of the dogs (Reisner et al., 2007).

In this population, how the dog was handled was almost as common a trigger as what it was guarding (consistent with what confrontational methods produce).

7.5 The Medical Finding, and What Did Not Prevent the Bites

Potential contributory medical conditions were identified or suspected in 50 percent of the dogs (Reisner et al., 2007) — which makes veterinary assessment a first step rather than a formality.

Two further results undercut the standard reassurances. Sixty-six percent had never previously bitten a child and 19 percent had never bitten any human — most bites came from dogs without a bite history. And the measures owners rely on had not prevented them: 93 percent of the dogs were neutered and 66 percent of owners had attended obedience classes. The authors state the conclusion directly — common calming measures were not routinely effective deterrents (Reisner et al., 2007), consistent with what the neutering literature shows about behavioural effects (set out separately).

8. The Assessment Problem

8.1 What Is at Stake

Most animal shelters test for food guarding, around half consider dogs identified as food aggressive unadoptable, and only 34 percent attempt behaviour modification — with food bowl guarding reported as one of the most common reasons for euthanasia (Mohan-Gibbons, Weiss & Slater, 2012). A test carrying that weight should predict the behaviour it is testing for.

8.2 What Happened When It Was Checked

Ninety-six dogs identified as guarding their food bowl were placed in homes on a modification programme, with follow-up at three days, three weeks and three months. Six adopters reported any guarding incident in the first three weeks, one of which involved the food bowl; by three months there were no reports except one new occurrence involving a rawhide. The return rate was 5 percent, against 9 percent for adult dogs not identified as guarding (Mohan-Gibbons et al., 2012).

8.3 The Independent Check

The finding is not isolated. In an earlier study, twenty-one behavioural tests were administered to 81 shelter dogs and compared against adopter reports: for aggression over food or a bone, shelter tests agreed with adopter reports about 43 percent of the time (van der Borg, Netto & Planta, 1991).

Forty-three percent agreement on a test whose result can be fatal is a serious problem.

8.4 What the Authors Concluded

The convergent recommendation is that shelters reconsider or discontinue food guarding assessments. Where shelters stopped assessing, there was no significant increase in returns, while more dogs were adopted because fewer were flagged — an unusual shape of result, in which removing a safety test did not produce the harm it was designed to prevent.

8.5 Why the Test Fails

Everything in section 6 applies at once. Shelter kennels are resource-scarce, socially unpredictable and stressful, and animals commonly guard more intensely when resources are scarce (with chronic stress altering the system it is measured in). Confinement removes the avoidance option. The approacher is a stranger with no history. And a plastic hand on a pole removing a bowl mid-meal is not something the dog will encounter again.

The test therefore samples one unusually adverse combination of the variables that determine guarding, and generalises it to all others.

None of this means guarding never persists, or that adopters need no information. It means a single standardised test taken during stressful confinement is a weak basis for a permanent decision — and the honest version of the problem is that no better test exists, which is why the recommendation to discontinue has been contested despite the data supporting it.

8.6 What Makes a Good Test

Three questions are routinely collapsed into one, and separating them explains why the shelter finding is not paradoxical.

Reliability asks whether the same dog produces the same result under the same conditions twice. Construct validity asks whether the situation measures resource guarding as defined, or something narrower. Predictive validity asks whether the result forecasts behaviour elsewhere, months later.

The three are independent. A test can be perfectly repeatable and still be a poor predictor of behaviour outside the test situation, because repeatability describes the test and prediction describes the world.

A further problem the data cannot solve: predictive value depends on how common the behaviour is in the tested population, and no prevalence estimate exists using the consensus definition. The proportion of positive results corresponding to guarding at home therefore cannot be calculated in advance, only observed afterwards.

9. What Makes It Worse

9.1 Confrontational Handling

Owner-reported data on training methods bears on this directly. Confrontational interventions — physically forcing an item from a dog's mouth, staring it down, hitting or kicking — produced aggressive responses in a substantial share of the dogs on which they were used (Herron, Shofer & Reisner, 2009).

Applied to a dog already guarding, these methods provide the evidence the dog needed that the approach predicts loss.

9.2 Taking Things Away to Practise

The routine advice to remove food repeatedly so the dog "gets used to it" runs opposite to the learning it is meant to produce.

The survey data are consistent with that prediction: removing the food dish during meals was associated with more severe or more frequent guarding, while adding palatable food to a meal the dog was already eating was associated with milder forms (Jacobs et al., 2018a).

A growl, likewise, is the last low-cost signal before contact. Suppressing it removes the warning without changing the state that produced it (with suppression regularly mistaken for resolution).

9.3 Removing the Resources Instead

Removing everything a dog guards manages the situation without changing anything, and the trigger set cannot be fully removed — resting places, doorways and proximity to a person remain.

10. What Treatment Can Target

10.1 The State of the Evidence

This section should be read against a specific fact: there are no randomised or controlled group trials of resource guarding treatment in dogs. What exists is one controlled single-case experiment combining functional analysis with treatment evaluation (Mehrkam et al., 2020), the shelter modification programmes reported alongside adoption outcomes, and a large body of untested practice.

The principles below are therefore graded by what supports them, not presented as equivalent.

10.2 Change What the Approach Predicts

The target is the classical component: an approaching person should stop reliably predicting loss. Adding palatable food to a meal in progress is the version of this with the most direct empirical support in the surveys, where it was associated with milder forms of guarding (Jacobs et al., 2018a).

The general method — graduated exposure paired with something better than the current outcome — is the best-evidenced intervention in canine behaviour therapy generally, though not tested for guarding specifically (with its own evidence and limits set out separately).

10.3 Work Below the Point of Defence

Guarding that occurs during a session is guarding being rehearsed. The distance, duration and approach speed at which the dog does not need to defend is where the work happens (with arousal setting the boundary on what can be learned).

This follows from the exposure principle rather than from a guarding-specific study, and it is the practical reason the early behaviours in section 3 matter: they mark the threshold.

10.4 Exchange Is Not Confiscation

Both end with the item in the human's hand, and they establish opposite contingencies. In a forced removal, the dog's behaviour has no effect on the outcome. In an exchange, releasing the item produces something better and the dog's behaviour determines what happens.

The distinction is operant and follows from the functional analysis result: if guarding is maintained partly by tangible consequences, then arranging for release to produce the better tangible outcome addresses that contingency directly (Mehrkam et al., 2020).

10.5 Address More Than One Contingency

The same study found guarding to be multiply controlled — escape, attention and tangible functions operating together (Mehrkam et al., 2020). Function-based treatments reduced guarding to zero for each maintaining contingency, and effects held with a different handler, in an untrained setting, with a highly preferred item, and two weeks later without training.

That is a strong result and it is one dog. Its value is directional: it indicates that "what is the dog getting out of this" may have several answers at once, and that treating only the obvious one may leave the behaviour intact.

10.6 Management Is Not Treatment

Separating dog and child at mealtimes, feeding behind a closed door and preventing access to contested items are necessary and they change nothing about the behaviour. They buy the safety within which treatment can happen.

Confusing the two produces a household that has been stable for a year and a dog that guards exactly as it did — with the risk reappearing the moment the arrangement lapses (much as suppressed behaviour returns when conditions change).

10.7 What Is Not Known

Whether any of these principles outperforms another, how long effects persist, which dogs respond, and whether treated guarding recurs under stress are all open. The strongest claim available is that the approaches with evidence behind them are consistent with the learning account and that the ones without it are associated with worse outcomes.

10.8 Generalisation and Maintenance

Two questions decide whether treatment holds outside the session, and neither has a canine answer specific to guarding.

Does the change transfer? In the single-case study it did — effects held with an untrained handler, in an untrained setting and with a highly preferred item (Mehrkam et al., 2020). One dog is one dog, and transfer is exactly what fails most often in applied work (with context-bound learning being the norm rather than the exception).

Does it last? The same study tested at two weeks. Nothing has followed treated dogs for months, which means the relapse question — central to every other fear-related intervention — is unaddressed here.

Both gaps point the same way for practice: treat generalisation as work to be done deliberately rather than as something that follows from the initial change.

11. Summary at a Glance

There is now an agreed definition — Avoidance, threatening or aggressive behaviours used to retain control of food or non-food items in the presence of a person or other animal (Jacobs et al., 2018c).

Value is relative, not fixed — The same item can be defended on one day and ignored on another, depending on hunger, competition, physical state and what has recently been taken.

Owners miss the early signs — Accurate on overt aggression, substantially worse on threatening behaviour, avoidance and rapid ingestion (Jacobs et al., 2017).

Impulsivity and fear predict it — Both significantly associated with guarding aggression toward people (Jacobs et al., 2018a).

Dog-directed guarding appeared more consistent across contexts — Relative to people-directed guarding, at group level (Jacobs et al., 2018b).

It was the leading circumstance for bites to young children in this clinical series — 44 percent of bites to children under six across 111 cases, and 42 percent of bites to familiar children involved food guarding (Reisner et al., 2007).

Half the biting dogs had a possible medical contributor — Potential contributory medical conditions in 50 percent, anxiety abnormalities in 77 percent (Reisner et al., 2007).

Shelter tests predict home behaviour poorly — Of 96 dogs identified as food guarders, one food-bowl incident in the first three weeks; return rate 5 percent against 9 percent (Mohan-Gibbons et al., 2012), with about 43 percent test–report agreement in an independent comparison (van der Borg et al., 1991).

Guarding can be multiply controlled — Escape, attention and tangible functions operating together in the one dog studied under functional analysis (Mehrkam et al., 2020).

There are no controlled group treatment trials — One controlled single-case experiment (Mehrkam et al., 2020), shelter programmes reported alongside adoption outcomes, and a large body of untested practice.

12. Research Gaps and Critical Appraisal

The prediction evidence is owner-reported and cross-sectional. The Jacobs surveys establish associations across large samples and cannot establish that impulsivity or fear cause guarding rather than accompanying it.

The prediction surveys did not assess pain. The clinical series found potential contributory medical conditions in half its dogs (Reisner et al., 2007), while the large owner surveys that model prediction did not measure physical status at all.

Non-food guarding is barely studied. Resting places, doorways and human proximity appear in the definition and in clinical description, and almost nothing in the empirical literature addresses them directly.

The definition is new and not yet operationalised for research. A consensus definition is a starting point; measurement instruments built on it, and prevalence data using it, do not yet exist.

Prevalence is unknown. Without a base rate, the predictive value of any assessment cannot be calculated — and since owners under-identify the early forms, reported figures are likely to cover only part of the defined behavioural range.

Adopter reports are subject to the same recognition problem. Follow-up surveys asking adopters whether guarding occurred will under-count it. The shelter-test finding may be partly a measurement artefact on both ends.

Treatment evidence is one dog. Mehrkam et al. (2020) is methodologically careful, produced maintained effects across handlers and settings, and has an n of one. Nothing larger has been published.

Shelter modification programmes are confounded. In the Mohan-Gibbons work, adopters did not comply with at least one aspect of the programme, which makes it unclear why so little guarding was reported — the authors say so explicitly.

13. Conclusion

Resource guarding is unusual among canine behaviours in how directly the measurement problem translates into consequences. There is now an agreed definition that includes avoidance and rapid ingestion alongside growling and biting, and good evidence that owners recognise only the visible end of that range — accurate when a dog snaps, substantially less so when it freezes, stiffens or simply eats faster. What predicts guarding toward people is impulsivity and fear rather than anything resembling status, and guarding toward other dogs behaves differently again. Underneath both sits a variable that neither survey measured and the clinical data put at fifty percent: physical discomfort. The finding that should have changed practice most is the one about testing. Of ninety-six dogs identified as food guarders in a standardised shelter assessment, a single food-bowl incident was reported in the first three weeks of adoption, and an independent comparison found about forty-three percent agreement between shelter tests and adopter reports. That is a weak basis for a decision that can determine whether a dog is offered for adoption at all — and section 6 explains why: the test samples one unusually adverse combination of approacher, location, confinement and physical state, then generalises it to every other combination the dog will meet. Treatment is where the gap is widest. The principles that follow from the learning account — changing what an approach predicts, working below the point at which defence is needed, exchanging rather than confiscating, and addressing more than one maintaining contingency — are coherent, consistent with the survey associations, and supported by a single-case study and nothing larger. That is not a reason to abandon them; it is a reason to state their status honestly, particularly in a field where the confidently delivered alternative is the handling that the evidence associates with making things worse (much as behavioural suppression is mistaken for resolution elsewhere).

Key Insights (Takeaways)

• Owners recognise the end of the sequence, not the beginning. They accurately identified overt aggression and the absence of guarding, but were substantially worse at identifying threatening behaviour, avoidance and rapid ingestion (Jacobs et al., 2017). This provides one plausible explanation for "bit without warning" histories, and it means reported prevalence is likely to underestimate the full behavioural range.

• Value is a relation, not a property. The same item can be defended on one day and ignored on another as hunger, competition, physical state and recent history shift — which is why guarding works poorly as a stable trait description and why a single observation carries less information than it appears to.

• It was the leading bite circumstance in a referral population, not a general bite statistic. Across 111 children bitten by dogs referred for aggression, resource guarding accounted for 44 percent of bites to children under six and food guarding for 42 percent of bites to children the dog knew (Reisner et al., 2007). Two-thirds of those dogs had never bitten a child before, 93 percent were neutered and 66 percent had attended obedience classes.

• Shelter food-bowl tests predict home behaviour poorly, for identifiable reasons. Of 96 dogs identified as food guarders and adopted with a modification programme, one food-bowl incident was reported in the first three weeks, and the return rate was lower than for dogs not identified as guarders (Mohan-Gibbons et al., 2012). The test combines stranger, confinement, scarcity and an artificial approach — a combination the dog will not meet again.

• Treatment rests on principle and one case. There are no randomised or controlled group studies of resource guarding treatment. One single-case functional analysis found the behaviour multiply controlled and reduced it to zero with function-based treatment, maintained across handlers, settings and two weeks (Mehrkam et al., 2020). Meanwhile, removing the food dish during meals was associated with more severe or frequent guarding while adding food was associated with milder forms (Jacobs et al., 2018a) — cross-sectional, consistent in direction, and not causal evidence.

References

Herron, M. E., Shofer, F. S., & Reisner, I. R. (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, 117(1–2), 47–54. https://doi.org/10.1016/j.applanim.2008.12.011

Jacobs, J. A., Pearl, D. L., Coe, J. B., Widowski, T. M., & Niel, L. (2017). Ability of owners to identify resource guarding behaviour in the domestic dog. Applied Animal Behaviour Science, 188, 77–83. https://doi.org/10.1016/j.applanim.2016.12.012

Jacobs, J. A., Coe, J. B., Pearl, D. L., Widowski, T. M., & Niel, L. (2018a). Factors associated with canine resource guarding behaviour in the presence of people: A cross-sectional survey of dog owners. Preventive Veterinary Medicine, 161, 143–153. https://doi.org/10.1016/j.prevetmed.2017.02.005

Jacobs, J. A., Coe, J. B., Pearl, D. L., Widowski, T. M., & Niel, L. (2018b). Factors associated with canine resource guarding behaviour in the presence of dogs: A cross-sectional survey of dog owners. Preventive Veterinary Medicine, 161, 134–142. https://doi.org/10.1016/j.prevetmed.2017.02.004

Jacobs, J. A., Coe, J. B., Widowski, T. M., Pearl, D. L., & Niel, L. (2018c). Defining and clarifying the terms canine possessive aggression and resource guarding: A study of expert opinion. Frontiers in Veterinary Science, 5, 115. https://doi.org/10.3389/fvets.2018.00115

Mehrkam, L. R., Perez, B. C., Self, V. N., Vollmer, T. R., & Dorey, N. R. (2020). Functional analysis and operant treatment of food guarding in a pet dog. Journal of Applied Behavior Analysis, 53(4), 2139–2150. https://doi.org/10.1002/jaba.720

Mohan-Gibbons, H., Weiss, E., & Slater, M. (2012). Preliminary investigation of food guarding behavior in shelter dogs in the United States. Animals, 2(3), 331–346. https://doi.org/10.3390/ani2030331

Reisner, I. R., Shofer, F. S., & Nance, M. L. (2007). Behavioral assessment of child-directed canine aggression. Injury Prevention, 13(5), 348–351. https://doi.org/10.1136/ip.2007.015396

van der Borg, J. A. M., Netto, W. J., & Planta, D. J. U. (1991). Behavioural testing of dogs in animal shelters to predict problem behaviour. Applied Animal Behaviour Science, 32(2–3), 237–251. https://doi.org/10.1016/S0168-1591(05)80047-4