My Shelter Dog “Failed” a Temperament Test. What Does Science Actually Tell Us?

A dog growls when someone pushes a fake hand toward his food bowl. What have we learned? We have learned that this dog growled, under those conditions, in response to that sequence of events. We have not yet established that the dog has a particular “temperament.” We have not established why he growled. We have not established what he would do if a family member walked past his food bowl at home. We have not established that he would bite. And we certainly have not established that he is a dangerous dog. Those distinctions are not philosophical hair-splitting. They come directly from the foundations of animal-behavior science.

 

Long before animal shelters developed temperament tests, researchers were asking how we can legitimately move from an animal's observable actions to explanations of those actions. Charles Darwin's The Expression of the Emotions in Man and Animals in 1872 carefully compared observable expressive movements across species. C. Lloyd Morgan later cautioned comparative psychologists against attributing more elaborate psychological explanations when simpler explanations could account for the evidence. Thorndike experimentally studied how consequences altered animal behavior. Pavlov demonstrated how previously neutral events could acquire behavioral significance through experience. Lorenz, Tinbergen, Hinde and the classical ethologists studied the organization, causation, development and function of natural behavior. Scott and Fuller then spent roughly two decades studying the development, genetics and social behavior of dogs themselves. Darwin Online The scientific problem, therefore, is much older than the shelter temperament test.

 

What did the animal do, what conditions produced it, how did that behavior develop, what behavioral system was operating, and how far can we generalize from the observation we made?

 

Those questions should come before the label.

 

Temperament Is a Claim About Consistency, Not Something Revealed by One Performance

 

The word temperament implies more than behavior occurring once. Research on canine temperament and personality has repeatedly examined whether individual behavioral differences remain sufficiently consistent across time and circumstances to be considered traits. Jones and Gosling's major review found evidence for meaningful canine temperament dimensions, but also important limitations in reliability and validity across methods. A later meta-analysis of 31 studies found moderate overall temporal consistency in dog personality, with consistency affected by such things as the dog's age, the interval between measurements and whether researchers used the same measurement instrument. ScienceDirect That produces an important conclusion that is sometimes missed on both sides of the temperament-test controversy.

 

Dogs are not behaviorally random. Individual differences exist. But neither are dogs machines that express an invariant trait identically regardless of where they are, who is present, what preceded the interaction, what they have learned, or what physiological state they are in.

 

If I want to claim that a shelter dog possesses a stable behavioral characteristic, I should expect some evidence of repeatability or generalization, not merely one response elicited in one contrived situation. A test performance can contribute evidence toward a temperament hypothesis. It cannot make the hypothesis true by definition.

 

Tinbergen Showed Why “He Is Aggressive” Is Not an Explanation

 

One of the most important contributions in twentieth-century animal behavior came from Nobel laureate Nikolaas Tinbergen. Tinbergen argued that a full biological explanation of behavior involves distinct questions concerning its immediate causation, its development during the individual's lifetime, its function or adaptive significance, and its evolutionary history. His framework remains foundational because answers at one level cannot simply substitute for answers at another. Wiley Online Library Apply that to a shelter dog that freezes and growls when someone interferes with his food.

 

At the mechanistic level, we might investigate sensory information, physiological state, distance, restraint, pain, arousal and the immediate sequence of events.

 

At the developmental level, we might ask about previous food competition, deprivation, handling experiences, learning history, early socialization and maturation.

 

At the functional level, we might ask what the behavior accomplishes within the interaction: does another individual stop approaching, relinquish an item, increase distance or terminate contact?

 

At the evolutionary level, we can consider the behavioral systems from which conflict, resource competition, defense, communication and avoidance arose.

 

Those are different questions.

 

Writing “food aggressive” on the dog's record answers none of them. Neither does “reactive.” “Reactive” may be a convenient conversational description for conspicuous behavior, but it is not a causal mechanism. A barking and lunging dog might be attempting to increase distance, attempting to decrease distance, frustrated by restraint, participating in an agonistic interaction, engaging with prey-like movement, responding to learned contingencies, or showing some combination of processes. The visible output does not tell us the mechanism automatically.

 

The Behavior First Belongs in a Behavioral System

 

Ethology provides another distinction that is especially important when discussing what shelters call aggression. Researchers commonly describe agonistic behavior as a broader family of behaviors involved in social conflict. It can include threat, attack, defense, retreat, escape, submission and other behaviors that regulate conflict. Those events are related, but they are not interchangeable. McGlone's review, for example, explicitly distinguishes threats, aggression and submission within the broader agonistic system. PubMed That distinction matters because many shelter forms collapse dramatically different events.

 

A dog freezes. A dog growls. A dog raises its hackles. A dog snaps without making contact. A dog moves away. A dog lunges against a kennel gate. A dog crosses twenty feet of open space, makes contact, punctures another animal repeatedly, maintains the attack and attempts to re-engage after interruption.

 

Those observations should not become equivalent merely because somebody places all of them under one heading called aggression. Scientific publications themselves use aggression in different operational ways. Some include threat displays; others distinguish threats from physical attack. For practical case assessment, I therefore prefer to record the actual behavior before arguing about the adjective.

 

Did the dog growl? Did the dog retreat? Did the dog snap? Was contact made? Was skin broken? Did the dog bite once or repeatedly? Did the dog maintain contact? Did the dog disengage spontaneously? Did the dog pursue after the other individual withdrew?

 

Those facts preserve information.

 

“Aggressive” compresses it.

 

The Shelter Is Not a Neutral Background

 

A temperament test also cannot be separated from the environment in which it is performed. That does not mean every shelter dog is “stressed” in exactly the same way or that stress explains every behavior. That would simply replace one catch-all explanation with another. But the physiological effects of shelter entry are real.

 

Hennessy and colleagues found that dogs in their first several days in a county shelter had higher plasma cortisol concentrations than dogs housed there longer, and dogs tested on their first day had higher cortisol than the same dogs several days later or pet dogs tested in their homes. Later work has continued to demonstrate that sheltering can affect physiology and behavior, while also showing that individual responses vary and that no single biomarker provides a complete explanation of welfare or future behavior. ScienceDirect

 

Beerda and colleagues likewise demonstrated that social and spatial restriction can alter canine behavior and physiology. In their experimental work, restricted housing was associated with changes in posture, vocalization, repetitive behavior and other measures. ScienceDirect

 

That matters when somebody says: “This test showed us who the dog really is.” There may be no scientifically defensible “really is” hidden beneath all context. The dog being tested is an organism interacting with the conditions present at that time. A shelter evaluation may therefore tell us something important about the dog in that environment. The error occurs when the observation is silently converted into a context-free trait.

 

Learning History Matters Too

 

Animal behavior cannot be reduced to ethology alone. More than a century of experimental learning research demonstrates that animals learn relationships among events and that behavioral probabilities change as a consequence of experience. Thorndike's late nineteenth-century experiments, Pavlov's conditioned-reflex work and Skinner's experimental analysis of behavior established different but complementary parts of that scientific foundation. Nature Suppose a dog has learned that human hands approaching a particular object predict removal of that object. That history is relevant. Suppose growling has repeatedly caused approaching people to back away. That consequence is relevant. Suppose the dog has never experienced anyone reaching into his bowl before. That is relevant too. But learning history is one analytical silo. It should not be permitted to swallow the rest of the dog.

 

Learning does not eliminate genetics, development, physiology, perception, injury, social relationships or species-typical behavioral organization. Nor does saying “the behavior was reinforced” tell us everything we need to know about why the underlying interaction existed in the first place. A scientifically serious case formulation keeps those questions separate long enough to determine what evidence actually supports each one.

 

Development Matters

 

A dog's present behavior also has a developmental history. Scott and Fuller's classic work on canine social development, along with the related experimental work of Freedman, King and Elliot, showed that early experience can have enduring effects on social behavior. In one famous study, puppies isolated with their mother and introduced to people at different ages showed increasing withdrawal from humans after approximately five weeks; puppies not socialized to people before fourteen weeks could show profound and persistent difficulty forming normal human relationships. PubMed

 

Modern reviews have refined the terminology and details, but the central principle remains: behavior develops. It emerges through interactions among biological maturation and experience rather than appearing fully formed as an immutable adult property. PubMed Central (PMC) That means age matters. History matters. The point in development at which something occurred matters. A behavior seen in a four-month-old puppy cannot automatically be interpreted the same way as superficially similar behavior in a socially mature adult dog. Again, topography alone is not enough.

 

Genetics Matter, But Genetics Are Not a Shortcut Around Individual Assessment

 

Scott and Fuller also demonstrated that heredity matters. Contemporary behavioral-genetic research confirms that genetic differences contribute to behavioral variation among dogs and among breed lineages. Large studies have identified heritable breed-associated differences in multiple behavioral traits. PubMed Central (PMC) But this does not permit the opposite mistake:

 

breed → presumed behavior → verdict about the individual dog.

 

Modern research also finds extensive variation within breeds, and a large genomic study published in Sciencefound that breed ancestry explains only part of behavioral variation among individual dogs. Environmental history, development and individual variation remain substantial. PubMed Central (PMC) Breed can therefore provide prior biological context. It is not a substitute for observing the dog.

 

Medical Causes Belong in Their Own Silo

 

Behavioral assessment also requires a medical boundary. Pain, sensory impairment, neurological disease, endocrine disease, gastrointestinal discomfort, dermatological disease, medication effects and other physical conditions can alter what a dog is capable of tolerating or how the animal responds during handling. That does not mean every growl has a medical explanation. It means a behavioral conclusion should not silently rule one out.

 

This is particularly important in shelters because medical handling itself can change the situation. The ASPCA's current shelter-assessment position explicitly notes that behavior during an intake examination may be influenced by procedural discomfort, previous medical experiences and pain, particularly when the examination occurs soon after admission when shelter stress may be greatest. ASPCA

 

A valid behavioral investigation therefore does not ask whether the problem is medical or behavioral as though those categories cannot interact. It asks what each domain contributes.

 

Why Food-Bowl Tests Have Been Such an Important Scientific Warning

 

Food-related shelter testing gives us an unusually clear example of why context and prediction cannot be collapsed.

 

Marder and colleagues followed 97 adopted shelter dogs. Twenty had been classified as showing food-related aggression during the shelter evaluation. Eleven of those 20 were subsequently reported to show comparable behavior in their adoptive homes. But 9 of the 20 did not. Meanwhile, 17 of 77 dogs who had not shown the behavior during the shelter assessment were later reported to show it at home. ScienceDirect

 

A separate study followed 96 dogs that displayed food-bowl guarding during shelter assessment. Guarding was rarely reported after adoption, and food-bowl guarding in particular was uncommon in the homes during follow-up. PubMed

 

Another study of 139 adopted dogs found statistically significant associations between shelter resource-guarding classifications and some forms of guarding in the adoptive home, but the positive predictive values were low. Depending upon the scenario, the shelter result correctly predicted the corresponding post-adoption behavior only 10% to 47.4% of the time. MDPI

 

This is an excellent example of why “statistically associated with” and “accurately predicts this individual dog” are not synonymous. The test may contain information. That does not mean the test warrants a categorical conclusion about the individual animal.

 

Fake Dogs and Dolls Present the Same Problem

 

Artificial stimulus devices make the generalization problem even clearer.

 

Barnard and colleagues compared responses to a child-like doll, a fake dog and other stimuli. They found some relationships between dogs' histories and responses to the models, but the correspondence was incomplete. The authors pointed out a fairly fundamental biological limitation: fake dogs and dolls do not smell, move or interact like the living beings they supposedly represent. ScienceDirect

 

A later study comparing shelter dogs' responses to fake and real dogs found good agreement for friendly behavior and moderate agreement for fearfulness, but inadequate agreement for the aggressive trait. ScienceDirect

 

That does not make a fake dog useless for every purpose. It means the validity claim has to match the evidence. A model dog can tell us how a dog responded to that model dog. Generalizing from that observation to: “This dog will attack real dogs” requires evidence that the test has not reliably supplied.

 

Modern Shelter Research Confirms That Different Behavioral Dimensions Predict Differently

 

One of the dangers in this debate is going too far in the other direction and saying that shelter assessment tells us nothing. That is not what the literature shows either.

 

Clay and colleagues prospectively studied 123 shelter dogs and compared standardized shelter assessments with adopter reports one month later. Friendly/social behavior, fear and anxiousness showed significant correspondence with related behavior after adoption. But aggression, food guarding and separation-related problems were not reliably predicted. PubMed Central (PMC)

 

That is important. The scientific conclusion is not: Shelter behavior is meaningless. The conclusion is: Predictive performance differs depending on exactly what behavior is being measured, how it is measured, and what later outcome somebody claims it predicts. That is exactly what we should expect from a serious behavioral science.

 

Why the False-Positive Problem Matters

 

Patronek and Bradley brought another scientific discipline into this discussion: diagnostic-test mathematics. If the severe outcome we are trying to predict is uncommon, even an apparently respectable test can produce large numbers of false positives. Their 2016 analysis argued that shelter behavioral evaluations used to predict future serious aggression can perform extremely poorly at the individual level for precisely this reason. Patronek, Bradley and Arps then reviewed more than 25 years of shelter behavior-evaluation research in 2019. They concluded that none of the evaluated batteries or individual subtests had demonstrated the level of evidence necessary to justify broad claims that they were validated for routine predictive use. Across the studies they reviewed, the mean reported false-positive rate was 35.1%; they estimated it could reach 63.8% under more typical shelter-population conditions. ScienceDirect This does not prove that every dog who fails a shelter test is safe. It proves something narrower and much more important: a positive result cannot be treated as though it establishes the future outcome in the individual dog. That is a basic distinction between population statistics and individual prediction.

 

Behavior After Adoption Is Not Necessarily Static Either

 

There is another reason not to imagine that the shelter contains a hidden permanent version of the dog waiting to be discovered. Behavior continues to change.

 

A 2023 prospective study followed 99 shelter dogs for six months after adoption using the C-BARQ. Several measured behaviors changed significantly across the follow-up period. Some increased; others decreased. Owners nevertheless reported high satisfaction with their dogs. PubMed Central (PMC) That study should not be interpreted to mean every adopted dog follows the same trajectory. It demonstrates the opposite. Time itself is part of the behavioral record. An evaluation that captures ten minutes of behavior cannot automatically substitute for longitudinal observation.

 

Recovery Is Part of Behavior Too

 

This point is particularly important in my own assessments. I don't only want to know whether a dog reacted to something. I want to know what happened before, what the dog did during the event, what caused the event to end, and what happened afterward.

 

Consider two dogs that both growl when a stranger approaches. One increases distance, stops growling as soon as space is restored, investigates the environment, accepts normal social contact later and does not reproduce the pattern under less intrusive conditions. The other remains highly activated after the person retreats, repeatedly searches for the person, escalates when exposed again at greater distance and attempts to re-engage after interruption. The word growled accurately describes part of both events. It does not make the events behaviorally equivalent. Trajectory, termination and recovery are evidence. A snapshot loses them.

 

State Is Not the Same Thing as Skill

 

This also explains why obedience performance should not be treated as a temperament test. A dog may know “sit.” That does not mean the dog can perform it in every biological state.  Conversely, a dog may sit on command while still being in a state from which unsafe behavior remains possible.

 

Training history, behavioral capacity, momentary access to a learned response and actual performance are different questions. A shelter dog that fails to respond to a cue during intense environmental disturbance has not thereby demonstrated a defective temperament. A dog that performs obedience exercises perfectly has not thereby demonstrated public safety. Skill testing and behavioral assessment answer different questions.

 

Risk Is Not Located Entirely Inside the Dog

 

Perhaps the most important mistake comes at the end of the chain. Suppose we eventually establish that a dog has a repeatable pattern that deserves concern. We still have not finished the assessment.

 

Risk is not itself a behavior.

 

Risk depends on the dog, the probable behavior, the potential consequence, the target, frequency of exposure, physical environment, available barriers, handler capability, management reliability and other factors.

 

A dog with a particular behavioral repertoire living with one experienced adult under reliable management represents a different risk system from the same dog living with several toddlers, frequent visitors, uncontrolled access to the front door and owners unable to maintain separation. The dog may be the same. The exposure system is not. That is why behavior assessment and placement decision are not the same operation.

 

What Would a Scientifically Defensible Assessment Look Like?

 

Instead of asking whether a dog is “aggressive,” “reactive,” “dominant” or whether he “failed,” I would reconstruct the case across separate lines of evidence:

 

1. Observation: What exactly did the dog do, in behavioral terms, and what did he not do?
2. Sequence: What happened before, during and after the event?
3. Environment: Where was the dog, what distance was available, was there confinement or restraint, and what alternatives could the dog perform?
4. Ethology: What behavioral system or systems plausibly organize the sequence?
5. Learning: What prior experiences and consequences could be influencing the behavior?
6. Development: What is known about age, social development and life history?
7. Biology: What do genetics, breed lineage and individual predispositions contribute without treating them as destiny?
8. Medicine and physiology: Could pain, illness, sensory change, drugs or physiological state materially affect the behavior?
9. Relationship: Who was interacting with the dog, and what history existed between them?
10. Trajectory: Does the pattern repeat across people, places and time? How does it terminate, and how does the dog recover?
11. Uncertainty: Which conclusions are observed, which are inferred, and which remain unknown?
12. Risk: If the behavior does occur again, who or what is exposed, how likely is exposure, what consequence could result and how reliable is management?
13. Decision: Only after those questions should somebody decide what training, management, placement or other intervention is justified.

 

The purpose is not to make every assessment impossibly complicated. It is to prevent conclusions from outrunning evidence.

 

So What Does It Mean If Your Dog “Failed” the Shelter Test?

 

It means you should obtain the original observations. Not merely the score. Not merely the checkbox. Not merely the adjective. Ask what the dog did.

 

If the shelter recorded that your dog “failed food,” ask what happened during the food test. If the paperwork says “dog aggressive,” ask whether the dog barked behind a barrier, growled, avoided another dog, snapped without contact, attempted to escape, made contact, injured another dog, pursued another dog or did something else entirely.

 

If it says “reactive,” ask for the behavior that word represents. Then compare those observations with what happens now. Does the pattern reproduce at home? Under what conditions? What changes it? What stops it? Does it generalize? Does it escalate? How quickly does the dog recover? Those are behavioral questions.

 

Do Not Recreate Provocative Shelter Tests at Home

 

You do not need to stick a fake hand into your dog's food bowl to find out whether the shelter was right. You do not need to repeatedly remove food. You do not need to crowd the dog. And you do not need to manufacture confrontations simply to obtain another data point. The goal is not to make the dog pass the test. The goal is to understand the dog.

 

Current ASPCA guidance reflects much of the same evidentiary concern. It recommends collecting information from multiple sources, warns that standardized assessments have not shown high accuracy for predicting post-adoption aggression, and cautions against making major outcome decisions from a single assessment except in unusually serious circumstances. ASPCA Ordinary behavior across real situations often gives us better comparisons than deliberately producing conflict.

 

A Shelter Test Can Be Evidence Without Becoming a Verdict

 

There is a temptation in controversial subjects to choose one extreme.

 

One side says: The dog failed the test, therefore the dog is dangerous. The other says: Temperament tests are unreliable, therefore the result means nothing. Neither position reflects the complete science.

 

A shelter observation can contain valuable information. A serious bite history can contain extremely important information. A repeated cross-context behavioral pattern can contain much stronger information than an isolated test response. A medical finding can change interpretation. A foster home can reveal behavior that never appeared inside the shelter. An adoptive home can reveal behavior that never appeared during either.

 

This is exactly why a scientifically responsible assessment integrates information without allowing one source to silently become the whole dog. The deepest problem with the traditional temperament-test model is therefore not simply that some particular tests have disappointing predictive statistics. It is more fundamental.Animal behavior is produced by an organism with an evolutionary history, a genotype, a developmental history, a nervous system, a physiological state, a learning history and a current environment. The behavior unfolds through time and within relationships.

 

A ten-minute test can sample that system. It cannot erase it. So when somebody tells me that a dog “failed a temperament test,” my first question isn't:

 

What was his score?

 

It is: What actually happened? From there, we can begin doing behavioral science.

 

Selected Scientific References

 

Darwin, C. (1872). The Expression of the Emotions in Man and Animals. John Murray. Darwin's original text includes comparative observations of expressive behavior in dogs and other animals. Darwin Online

 

Freedman, D. G., King, J. A., & Elliot, O. (1961). Critical period in the social development of dogs. Science, 133, 1016–1017. doi:10.1126/science.133.3457.1016. PubMed

 

Hennessy, M. B., Davis, H. N., Williams, M. T., Mellott, C., & Douglas, C. W. (1997). Plasma cortisol levels of dogs at a county animal shelter. Physiology & Behavior, 62, 485–490. doi:10.1016/S0031-9384(97)80328-9.PubMed

 

Jones, A. C., & Gosling, S. D. (2005). Temperament and personality in dogs (Canis familiaris): A review and evaluation of past research. Applied Animal Behaviour Science, 95, 1–53. ScienceDirect

 

Marder, A. R., Shabelansky, A., Patronek, G. J., Dowling-Guyer, S., & D'Arpino, S. S. (2013). Food-related aggression in shelter dogs: A comparison of behavior identified by a behavior evaluation in the shelter and owner reports after adoption. Applied Animal Behaviour Science, 148, 150–156. doi:10.1016/j.applanim.2013.07.007. ScienceDirect

 

Mohan-Gibbons, H., Weiss, E., & Slater, M. (2012). Preliminary investigation of food guarding behavior in shelter dogs in the United States. Animals, 2, 331–346. PubMed Central (PMC)

 

Patronek, G. J., & Bradley, J. (2016). No better than flipping a coin: Reconsidering canine behavior evaluations in animal shelters. Journal of Veterinary Behavior, 15, 66–77.

 

Patronek, G. J., Bradley, J., & Arps, E. (2019). What is the evidence for reliability and validity of behavior evaluations for shelter dogs? A prequel to “No better than flipping a coin.” Journal of Veterinary Behavior, 31, 43–58. doi:10.1016/j.jveb.2019.03.001. ScienceDirect

 

Pavlov, I. P. (1927). Conditioned Reflexes: An Investigation of the Physiological Activity of the Cerebral Cortex. York University

 

Scott, J. P., & Fuller, J. L. (1965). Genetics and the Social Behavior of the Dog. University of Chicago Press. Google Books

 

Skinner, B. F. (1938). The Behavior of Organisms: An Experimental Analysis. Appleton-Century. Google Books

 

Tinbergen, N. (1951). The Study of Instinct. Clarendon Press. Google Books

 

Tinbergen, N. (1963). On aims and methods of ethology. Zeitschrift für Tierpsychologie, 20, 410–433. doi:10.1111/j.1439-0310.1963.tb01161.x. Wiley Online Library

 

Clay, L., Paterson, M. B. A., Bennett, P., Perry, G., & Phillips, C. J. C. (2020). Do behaviour assessments in a shelter predict the behaviour of dogs post-adoption? Animals, 10, 1225. doi:10.3390/ani10071225. PubMed Central (PMC)

 

Bohland, K. R., Lilly, M. L., Herron, M. E., Arruda, A. G., & O'Quin, J. M. (2023). Shelter dog behavior after adoption: Using the C-BARQ to track dog behavior changes through the first six months after adoption. PLOS ONE, 18, e0289356. PLOS

Does My Dog Really Have a Behavior Problem? What the 99% Study Got Wrong

Not every unwanted or commonly reported dog behavior is a behavior problem. A reanalysis of more than 43,000 dogs shows why frequency, severity, development, welfare and safety have to be considered separately.

 

You may have seen the claim that more than 99 percent of dogs in the United States have at least one moderate to serious behavior problem. That statistic came from a real scientific study using data from more than 43,000 dogs. But a subsequent reanalysis of the same dogs found that the way the original study grouped and classified behaviors substantially exaggerated the apparent prevalence of behavior problems.

 

There is a larger lesson for dog owners here.

 

A dog doing something you do not like is not automatically a dog with a behavior disorder. A behavior being common does not make it abnormal, and a behavior appearing on a questionnaire does not tell us whether it needs professional treatment.

 

Before deciding that your dog has a "behavior problem," I want to know what the dog actually does, how often it happens, how intense it is, when it happens, whether anyone is at risk, whether the dog's welfare is impaired, and, especially with puppies, where the dog is developmentally.

 

An Example

 

Years ago, I was hired because the student’s boyfriend didn’t like her dog looking at him. That wasn’t a dog behavior problem, he had a problem. And they had a relationship problem that no dog trainer could fix.

 

Do 99 percent of American dogs really have behavior problems?

 

No scientific evidence currently establishes that 99 percent of American dogs have meaningful behavior problems.

 

In 2024, veterinary behaviorist Bonnie Beaver analyzed mini C-BARQ questionnaire data from 43,517 dogs enrolled in the Dog Aging Project. The paper reported that 99.12 percent had at least one behavior problem classified as moderate to serious. It also reported 55.6 percent for "aggression," 49.9 percent for fear and anxiety, and 85.9 percent for separation and attachment behaviors. (Beaver, 2024). James Serpell and Lauren Powell subsequently reanalyzed the same dataset using the established factor structure of the mini C-BARQ. Their conclusion was very different. They argued that the original analysis grouped behaviors that should not have been combined and used thresholds that tended to inflate apparent prevalence and severity. (Serpell & Powell, 2026).

 

The reanalysis does not tell us the exact percentage of American dogs that truly have behavior problems either. Serpell and Powell point out that a representative population sample would be necessary to establish that number.

 

That is important. The correction to a bad estimate is not to invent another overly precise estimate.

 

What percentages did the reanalysis find?

 

The answer depends greatly on what is measured and where the threshold is placed. Serpell and Powell calculated prevalence using both a score of 2 or greater and a more stringent score of 3 or greater. The difference is substantial.

 

Behavior measured by mini C-BARQ	Score 2 or greater	Score 3 or greater
Stranger-directed "aggression"	10.0%	1.5%
Owner-directed "aggression"	1.2%	0.3%
Dog-directed "aggression"	30.4%	11.0%
Familiar-dog rivalry	10.5%	3.2%
Nonsocial fear	15.6%	2.4%
Dog-directed fear	28.2%	9.0%
Stranger fear	14.1%	4.2%
Separation-related behavior	9.3%	2.3%
Attachment/attention-seeking	88.9%	55.7%
Escaping or attempting to escape	29.8%	14.1%
Urinating against objects indoors	3.7%	0.9%
Urinating while alone	6.2%	1.8%
Defecating while alone	4.4%	1.1%
Tail or hind-end chasing	11.0%	2.5%
Persistent barking when alarmed or excited	46.9%	22.7%

 

These are questionnaire scores. They should not be read as percentages of dogs with diagnosed behavioral disorders. That distinction is the point.

 

How did separation behavior go from 85.9 percent to 9.3 percent?

 

The original analysis combined two different things: separation-related behavior and attachment/attention-seeking behavior. When those factors were separated in the reanalysis, 9.3 percent of dogs met the score-2 threshold for separation-related behavior, while 88.9 percent met that threshold for attachment and attention-seeking. At the stricter score-3 threshold, separation-related behavior fell to 2.3 percent.

 

That makes biological sense. Dogs are social animals that have been selected for thousands of years to form relationships with humans. A dog following you around the house, wanting proximity, seeking attention or choosing to lie near you is not equivalent to a dog experiencing serious distress when left alone. If we combine ordinary social attachment with abnormal separation behavior, we can manufacture an alarming statistic without necessarily identifying an alarming population. This is why a behavior label is not an explanation.

 

Does 55.6 percent "aggression" mean more than half of dogs are violent?

 

No. This is another place where terminology can mislead the reader. The mini C-BARQ category called "aggression" includes behaviors such as growling, barking and baring teeth at intermediate levels. The highest end of the scale includes snapping, biting or attempting to bite. Serpell and Powell specifically point out that a dog that growls or barks when a delivery person approaches the house can therefore contribute to an "aggression" score.

 

For practical assessment, I do not consider a threat display and a physical attack to be the same event. A growl is behavior. Baring the teeth is behavior. Barking at a stranger approaching the property is behavior. Those signals may be important and may indicate that further assessment is warranted, but they are not interchangeable with a bite or an attack.

 

The more useful questions are: What exactly did the dog do? To whom? Under what circumstances? At what distance? How often? What happened next? Was contact attempted or made? Was anyone injured? Once those questions are answered, the word used to label the behavior becomes much less important than the behavior itself.

A 2026 study by Normando and colleagues demonstrates just how much the question itself can manufacture an apparent prevalence rate. Among 569 dog owners, only 32.7% spontaneously reported at least one behavioral problem when asked an open-ended question. When owners were instead presented with a list of behaviors to consider, 99.0% identified at least one. Even the order of the questions affected their answers. The study does not tell us that 32.7% is the “true” prevalence of behavior problems. It tells us something more important: a prevalence figure can partly measure the questionnaire rather than the dog. Before calling a behavior a problem, we still have to ask what the dog actually does, under what circumstances, how often, how intensely, with what consequences, and whether it meaningfully compromises the dog, the household, or public safety. (ScienceDirect)

 

We do not currently have a credible basis for saying that 99% of dogs have behavior problems. The percentage depends heavily on what researchers call a problem, how behaviors are grouped and scored, what cutoff is chosen, how questions are framed, and which population is sampled.

 

When is a dog behavior a problem?

 

I think owners get better answers if we stop trying to put everything into one box called "behavior problems." For practical purposes, I would separate at least four possibilities.

 

1. Normal canine behavior that humans find inconvenient. Dogs bark, dig, chew, chase, sniff, seek social contact and respond to things happening around their homes. Normal does not mean desirable in every household. It means that the presence of the behavior alone does not establish pathology.

2. Developmental behavior. Puppies and adolescents are not finished dogs. House-training capacity, oral behavior, attention, emotional responses, social behavior and impulse control change as dogs mature. Some problems diminish with maturation while others emerge or become clearer during adolescence. Rather than a behavior problem, these combine a lack of maturity with a lack of proper training repetitions.

3. A training or management problem. A normal behavior may still interfere substantially with life in the household. Jumping on visitors, pulling on leash, counter surfing, barking or poor household manners may require training and management even when nothing is wrong with the dog. Dogs need training, they don’t come trained “out of the box.” Owners need training, too. I had no clue what to do with my first dog, and I had to enroll in classes to figure out solutions. My dog didn’t have a behavior problem, nor did I have a behavior problem. We had a lack of training.

4. A welfare or safety problem. Severe distress, inability to function normally, repeated escape attempts, injurious behavior, serious fear, self-injury, biting and attacks belong in a different category. These are not equivalent to a puppy chewing your shoe.

 

The purpose of assessment is to determine which problem you have before deciding what to do about it.

 

Which puppy problems might improve as the puppy develops?

 

Some puppy behaviors genuinely change as the nervous system, body and social behavior mature.

 

Researchers who filmed 32 puppies home alone three times during the first months after adoption found that only three showed a cluster of stress-associated behaviors such as substantial vocalization, lip licking and oral behavior. Those behaviors decreased substantially across subsequent observations and did not persist at the same level. (Cannas et al., 2010).

 

Young puppies also commonly vocalize when socially isolated. Separation protest has been described as an adaptive juvenile response, and nighttime disturbance and house soiling in newly adopted puppies often decline as puppies mature and adjust to their new environment.

 

That does not mean you should ignore puppy problems.

 

House-training still requires management. Chewing still requires supervision and appropriate outlets. Puppies still need to learn skills. Preventable experiences can affect later behavior. What maturation means is that I would be cautious about interpreting every difficulty in a ten-week-old puppy as evidence of a permanent adult behavioral disorder.

 

Which puppy problems should I not assume the dog will outgrow?

 

Do not use "he's just a puppy" as an excuse to ignore a worsening trajectory. Behavior does not uniformly improve with age. In a large study of prospective guide dogs, measurable behavioral changes occurred between six and twelve months. Some characteristics decreased while others increased, and the changes differed among breeds. German Shepherd Dogs in that population, for example, showed increasing stranger-directed "aggression" scores between six and twelve months. (Serpell & Duffy, 2016). (Please recall: “aggression” isn’t the same as posturing and vocalizations, but are often lumped in together as one. Truly, scholars should know better, but often even they need more training.)

 

That should not be interpreted as a prediction about every German Shepherd. It demonstrates the larger developmental point: the puppy you see today is not necessarily a miniature version of the adult dog.Genetics, breed tendencies, maturation, learning, environment and experience continue to interact.

 

This is why I am more concerned about direction than about one isolated observation. Is the behavior diminishing, remaining stable, becoming more intense, occurring in more situations, or becoming harder for the dog to recover from?

 

Which behavior problems can I reasonably work on myself?

 

Many owners can begin with ordinary household and puppy-training problems themselves.  That can include house-training, age-appropriate chewing, puppy mouthing that is not causing a serious safety concern, jumping, basic leash skills, mild attention-seeking, common household manners, and mild developmental difficulties that are already improving.

 

The goal is not necessarily to suppress every normal dog behavior. Often the job is to provide an appropriate outlet, prevent rehearsal of unwanted behavior, teach an alternative, and set expectations that fit the dog's developmental stage.  Track progress rather than demanding perfection.

 

If accidents are becoming less frequent, the puppy can remain alone comfortably for gradually longer periods, recovery is getting faster, or an unwanted behavior is becoming easier to interrupt and redirect, those trends provide useful information. You are on track. Some puppies are very easy right from the start. 

 

Some.

 

When should I get professional help?

 

Professional assessment makes sense when the stakes, frequency, intensity, severity or uncertainty exceed what can reasonably be handled by trial and error by a novice dog owner. Or when you are in over your head with a particular dog.

 

I would be especially concerned about biting or attacks; repeated attempts to bite; behavior creating a meaningful risk to children, adults or other animals; severe or escalating fear; panic or sustained distress when left alone; self-injury; dangerous escape behavior; sudden major behavioral change; inability to find any situation in which the dog can work or interact successfully; behavior that continues to deteriorate despite reasonable management and training; or signs suggesting pain or another medical contribution.

 

Separation-related behavior deserves the same kind of careful definition. Barking a little when you leave, following you from room to room, and prolonged distress with pacing, destruction, uncontrolled elimination, and extreme vocalization can’t all be lumped in together as the same problem. Research on separation-related behavior also shows why context matters: the behavior must be associated with separation rather than simply occurring at another time.

 

If a behavior appears suddenly, particularly in an adult dog, veterinary evaluation may need to be part of the assessment.

 

So does my dog have a behavior problem?

 

Start by describing the behavior instead of naming the problem. Write down what the dog does, what happens immediately before it, who is present, where it happens, how often it occurs, its intensity, what ends it, how quickly the dog recovers, and whether the pattern is changing. Then ask a second question:

 

Is this primarily normal behavior, development, a training and management issue, a welfare concern, or a safety concern?

 

That distinction matters far more than whether your dog helps push a population survey above or below some statistical threshold. And if you don’t know, it is time to make a call.

 

The lesson from the 99 percent controversy is not that canine behavior problems are rare. Some are common, some are serious, and some can profoundly affect dogs and the people who live with them.

 

The lesson is that measurement is not diagnosis, a behavior is not automatically a disorder, and prevalence depends heavily on how we define what counts as a problem.

 

If your dog is having a behavior problem and you are not sure which category it belongs in, an individualized assessment can help identify what the dog is doing, the circumstances in which it occurs, and what kind of intervention is justified.

 

You can learn more about working with me at SamTheDogTrainer.com.

 

Glossary

 

Aggression: Violence.

Behavior problem: A broad and partly subjective term for behavior considered undesirable, harmful, disruptive or concerning. The term does not itself identify cause, severity or appropriate treatment.

 

C-BARQ: The Canine Behavioral Assessment and Research Questionnaire, a standardized owner questionnaire developed to measure behavioral and temperament characteristics in dogs.

 

Mini C-BARQ: A shortened form of C-BARQ used in the Dog Aging Project and other research.

 

Developmental behavior: Behavior influenced by the dog's current stage of physical, neurological and social maturation.

 

Separation-related behavior: Behavior occurring specifically in association with separation from human companionship, which may include vocalization, destruction, pacing, elimination or other behaviors.

 

Attachment/attention-seeking: Social behavior involving proximity to or interaction with familiar people. Its presence by itself does not establish separation distress.

 

Threat display: Signaling behavior such as growling, barking or baring teeth. A threat display can be important for risk assessment but should not automatically be equated with physical attack.

 

Prevalence: The proportion of a population displaying a specified characteristic during a defined period or under specified measurement criteria.

 

Threshold: The score selected by researchers as the dividing point between categories. Changing the threshold can substantially change the percentage of subjects placed into a category.

 

Bibliography

 

Beaver, B. V. (2024). The prevalence of behavior problems in dogs in the United States. Journal of Veterinary Behavior, 76, 34–39. doi:10.1016/j.jveb.2024.11.001.

 

Cannas, S., Frank, D., Minero, M., Godbout, M., & Palestrini, C. (2010). Puppy behavior when left home alone: Changes during the first few months after adoption. Journal of Veterinary Behavior, 5(2), 94–100. doi:10.1016/j.jveb.2009.08.009.

 

Dale, F. C., Burn, C. C., Murray, J., & Casey, R. (2024). Canine separation-related behaviour at six months of age: Dog, owner and early-life risk factors identified using the Generation Pup longitudinal study. Animal Welfare, 33, e60. doi:10.1017/awf.2024.56.

Normando, S., Canori, C., Borlini, M., Zanetti, R., Walsh, E. A., Meers, L. L., & Contalbrigo, L. (2026). How questions on dog behavioural problems may shape caretaker answers. Applied Animal Behaviour Science, 304, 107134. https://doi.org/10.1016/j.applanim.2026.107134 (ScienceDirect)

 

Serpell, J. A., & Duffy, D. L. (2016). Aspects of juvenile and adolescent environment predict aggression and fear in 12-month-old guide dogs. Frontiers in Veterinary Science, 3, 49. doi:10.3389/fvets.2016.00049.

 

Serpell, J. A., & Powell, L. R. (2026). Prevalence and severity of behavior problems in dogs in the United States: A re-assessment. Journal of Veterinary Behavior, 83, 8–13. doi:10.1016/j.jveb.2025.11.005.

 

Wilkins, V., Evans, J., Park, C., Dog Aging Project Consortium, Fitzpatrick, A. L., Creevy, K. E., & Ruple, A. (2024). Validation of the shortened version of the Canine Behavioral Assessment and Research Questionnaire (C-BARQ) using participants from the Dog Aging Project. PLoS ONE, 19(4), e0299973. doi:10.1371/journal.pone.0299973.