Welfare Science and Quality of Life in Dogs
Michael Sauerwein · September 27, 2026
Welfare and quality of life get used as if they meant the same thing. In the research literature they do not: one describes a dog's whole state across conditions and resulting mental states, the other is usually a score from an instrument, most often built around a disease.
This article sets out the framework that organizes modern welfare assessment, what the canine quality-of-life instruments actually measure and how well they have been validated, why positive states remain harder to measure than negative ones, and what all of it changes in the life of an ordinary pet dog.
1. Two Words That Are Not Interchangeable
1.1 Welfare, Quality of Life, and Why the Distinction Matters
Welfare and quality of life get used as synonyms, and in casual conversation that is harmless. In the literature they are not the same thing. Welfare is the broader concept: the dog's state across nutrition, environment, health, and behavioral opportunity, together with the affective experiences that arise from those conditions. Quality of life, as veterinary research uses the term, is usually narrower and more instrument-bound: a score or profile, most often reported by the owner, describing how the dog is faring — frequently in the context of a disease, a treatment, or an end-of-life decision.
The distinction matters because the two are measured by different means and answer different questions. "Does this dog have a life worth living?" and "Has this dog's arthritis treatment improved its day-to-day experience?" are both legitimate, but a tool built for the second will not answer the first.
1.2 What Went Wrong With "Good Welfare Means No Suffering"
For decades, welfare assessment was essentially a hunt for deficits. The Five Freedoms formulation — freedom from hunger, discomfort, pain, fear, and freedom to express normal behavior — was a huge advance over what preceded it, but it is framed almost entirely in negatives. A dog that is not hungry, not in pain, not frightened, and not obviously restricted passes. Whether anything good is happening in its life is not asked.
That gap is what the modern frameworks were built to close. Mellor and Beausoleil set out explicitly why an assessment of enhancement needs different indices from an assessment of compromise: you cannot detect contentment by confirming the absence of fear. Their 2015 extension of the Five Domains Model added structured assessment of positive states alongside negative ones, and argued that the two have to be graded separately and then combined, because absence of acute suffering does not by itself demonstrate positive welfare.
1.3 What This Article Covers
This article sets out the framework most widely used to assess canine welfare, what the quality-of-life instruments actually measure and how well they have been validated, why positive states remain the harder half of the problem, and what all of this changes in the life of an ordinary pet dog. It is deliberately an umbrella article: many of the individual mechanisms have their own treatment elsewhere, and are linked at the points where they belong.
2. The Five Domains Model
2.1 The Structure
The Five Domains Model, in its 2020 form, organizes welfare assessment into four domains of physical and functional condition plus one domain focused on mental state arising from those conditions:
Domain 1 — Nutrition. Water and food: availability, quality, quantity, variety.
Domain 2 — Physical Environment. Temperature, air quality, substrate, light, noise, space, hygiene.
Domain 3 — Health. Disease, injury, functional impairment, body condition, fitness.
Domain 4 — Behavioral Interactions. The opportunities the animal has to interact with its environment, with other animals, and with humans.
Domain 5 — Mental State. The positive and negative mental states inferred from the conditions described in the first four domains.
The logic runs in one direction. Domains 1 to 4 describe the conditions and influences acting on the animal; Domain 5 describes the mental state inferred from them. The other four domains matter because they provide the basis for that assessment.
2.2 Two Kinds of Feeling
Mellor and colleagues draw a distinction that is easy to skip and worth keeping. Domains 1 to 3 generate what they call survival-critical affects: hunger, thirst, pain, breathlessness, nausea, weakness. These arise from disruptions to internal stability and produce compelling motivations to engage in genetically embedded responses. A thirsty dog does not decide to seek water in any interesting sense; the state itself drives the behavior.
Domain 4 generates situation-related affects: frustration, fear, loneliness, boredom — and, on the positive side, comfort, contentment, playfulness, affectionate sociability. These arise from external circumstances and are associated with motivated, situation-dependent behavior. This is the domain where agency lives, and it is the domain where most of what a trainer or an owner can influence sits.
The practical consequence is that the two categories respond to different interventions. You cannot fix boredom with food, and you cannot fix pain with enrichment.
2.3 What Changed in 2020
The model has been revised repeatedly — updates are documented for 2001, 2004, 2009, 2012, 2015, and 2017 before the current version. The 2020 revision made two changes that matter for dogs in particular.
First, Domain 4 was renamed from "Behavior" to "Behavioral Interactions," with three explicit subcategories: interactions with the environment, with other animals, and with humans. The renaming was not cosmetic. It shifted the domain from cataloguing what behavior the animal performs to asking what the animal is given the opportunity to do.
Second, human–animal interaction was brought inside the framework as something to be systematically assessed rather than assumed benign. The 2020 paper provides detailed guidance on evaluating both negative and positive welfare impacts arising from handlers, veterinarians, trainers, owners, and researchers. For a companion dog, whose entire environment is arranged by humans and whose most significant social relationship is with a human, this is not a minor addition. It means the quality of the human's behavior is a welfare input, not a background condition.
2.4 Why the Model Is Useful Even Though It Produces No Number
The Five Domains Model is a structure for thinking, not a scoring system with a published cutoff. Mellor's operational account describes how grades are assigned within domains and aggregated, but the output is a profile, not a diagnosis. Its value is that it forces coverage: it is very difficult to work through all five domains and still conclude that a dog whose only problem is that nothing ever happens to it has good welfare.
3. Agency: Choice, Control, Challenge
3.1 What Agency Means Here
Littlewood, Heslop, and Cobb define agency as the capacity of individual animals to engage in voluntary, self-generated, goal-directed behavior that they are motivated to perform. The definition does three useful things. It requires the behavior to be voluntary, which excludes behavior compelled by discomfort. It requires it to be self-generated, which excludes behavior produced by a cue. And it requires motivation, which means the dog has to actually want the outcome, not merely be capable of the action.
Their argument is that exercising agency produces what they term positive affective engagement — a state of pleasant occupation with rewarding activity — and that this constitutes genuine positive welfare rather than merely the absence of harm.
3.2 Three Components
The authors separate three agentic qualities, and the separation is worth keeping because interventions tend to supply one and not the others.
Choice is selecting between available options. A dog that can move between a sunny spot and a cool one, or between the company of the household and a quiet room, has choice.
Control is predictably achieving a motivated outcome — acting and having the action work. This is where the experimental literature on controllability and predictability connects directly: the difference between an animal that can influence what happens to it and one that cannot is one of the most robust findings in the stress literature, and the physiological consequences of losing it are substantial.
Challenge is difficulty that can be surmounted. Not absence of difficulty — surmountable difficulty, the kind that produces competence. This is the component most often missing from enrichment that consists of making things easy.
3.3 Where Agency Fails Quietly
The failure mode is not usually dramatic. It is a dog whose every outing is on a short lead along the same route, whose food arrives in a bowl, whose social contacts are chosen entirely by the owner, and who has no way to end an interaction it does not want. Nothing on that list is cruel and none of it would register as a welfare problem in a deficit-based assessment. What it produces, cumulatively, is an animal with very little to do and no means of affecting anything.
The extreme end of losing control is documented: learned helplessness is what happens when an animal's actions stop predicting outcomes altogether. Most dogs never get there. The point is that the dimension is continuous, and that most of it lies well short of the extreme.
4. Quality-of-Life Instruments: What Exists
4.1 The State of the Field in 2015
Belshaw and colleagues ran a systematic, replicable search for published, peer-reviewed canine quality-of-life instruments and then assessed their quality. The results are worth stating in full because they set the baseline for everything since.
They obtained 52 instruments. A further 27 instruments that had been used in peer-reviewed publications could not be included, because they had not been fully described and the authors did not supply them on request. Of the 52, 48 were disease-specific rather than generic. Only four publications provided a definition of quality of life or wellbeing at all. And only 11 of 52 instruments demonstrated any evidence of reliability or validity testing — with the quality of even those varying considerably.
Their conclusion was blunt: many novel, unvalidated instruments had been generated and applied as clinical outcome measures before anyone knew whether they measured quality of life.
4.2 Generic Tools
Fulmer, Laven, and Hill scoped the generic instruments — the ones intended to work for any dog or cat rather than for a specific disease — and found nine. Nine, across the whole literature, for the general case. The specific case is much better served than the general one, which is the reverse of what an owner asking "how is my dog doing?" would expect.
The best-developed of them is the instrument from Reid and colleagues. They reduced an initial 109-item pool to 46 items on a seven-point scale through expert evaluation and statistical reduction, tested it on 125 dogs, and extracted four health-related domains — vitality, pain, distress, and anxiety — accounting for 64.1% of the variance. It showed good discriminative properties and strong reliability, and was presented as the first valid and reliable companion-animal health-related quality-of-life instrument in web format.
Two things about that are worth noticing. It is genuinely validated, which puts it in a small minority. And its four domains are all about health and negative affect. Vitality is the closest thing to a positive dimension, and it is not the same as "has a good time."
4.3 A Screening Questionnaire for the Ordinary Pet Dog
One early attempt deserves separate mention, because it aimed at exactly the case the rest of the literature neglects: the healthy pet dog arriving at a veterinary practice for something else.
Mullan and Main built a screening questionnaire around four kinds of item. Owners supplied biographical information. They assessed the resources they provided — comfort, exercise, diet, mental stimulation, companionship — from the perspective of their dog's wellbeing and choice. They scored a range of behavioral and medical signs on visual analogue scales. And they rated, also on a visual analogue scale, how willing they would be to take on the life their dog was living. They were then asked what changes would improve their dog's quality of life.
Twenty-seven owners completed it on consecutive days. All parameters correlated well between the two days, and of 16 expected correlations testing internal consistency, nine were significant. The authors concluded it was repeatable, feasible, and internally consistent enough for practice use. With 27 participants, that is a pilot rather than a validation, and the authors present it as preliminary.
Two of the results are more interesting than the psychometrics. First, 26% of owners scored in the top 10 mm of the scale for willingness to take on the life their dog was living — which is a way of asking the welfare question that no disease-specific instrument approaches. Second, the single most commonly suggested improvement, named by 14 of the 27, was more exercise. Owners reaching for a way to make their dog's life better reached first for the one intervention whose behavioral claims turn out to be the least well supported (section 6.4).
4.4 Disease-Specific Tools, Ten Years On
Rhein, Klee, Albrecht, and Krämer repeated the exercise for disease-specific instruments in 2025. They analyzed 41 publications. Complete instruments were available for 30. Information on how items were developed appeared in 24. Any form of evaluation was described in 12.
Within those 12: test-retest reliability was examined for three instruments, of which two reached an appropriate threshold; seven showed acceptable internal consistency; four were tested for face validity and seven for content validity; and nine had at least one form of construct-validity hypothesis testing. Their overall verdict was that none of the instruments had been thoroughly evaluated across all the psychometric aspects needed for proper application and interpretation of results.
That is a decade after Belshaw. The volume of instruments has grown; the validation has not kept pace.
4.5 A Welfare Grid Rather Than a Questionnaire
One tool takes a different approach and is worth separating out. Malkani, Paramasivam, and Wolfensohn adapted the Animal Welfare Assessment Grid for dogs: 16 factors across four parameters — physical (mobility, body condition, clinical assessment, eating and drinking), psychological (aggression toward caregivers, aggression toward unfamiliar people, frequency of fear and anxiety, reaction to stressors), environmental (choice, control and predictability; enrichment; social interactions), and procedural (behavior during assessment, changes to daily routine, stress during handling, procedure-related pain).
It was tested on 96 dogs — 41 healthy, 47 with chronic conditions, 8 undergoing neutering. Content validity was assessed by seven subject-matter experts (I-CVI 0.99, S-CVI/Ave 0.94). Test-retest was examined in 19 dogs with correlations of 0.89 to 0.97 and no statistical difference between repeated scores, and the grid discriminated between healthy and sick dogs. The authors describe this as preliminary validation and are explicit that broader testing is needed.
What makes it interesting is the content rather than the psychometrics. "Choice, control and predictability" appears as a scored environmental factor. Enrichment and social interaction are scored separately. This is a tool that would notice the under-stimulated, fully agency-deprived, physically healthy dog — which most health-related quality-of-life instruments would not.
4.6 What the Instrument Landscape Means in Practice
If a dog has osteoarthritis, atopic dermatitis, epilepsy, cancer, or heart disease, there is a reasonable chance that a disease-specific instrument exists, and a smaller chance that it has been properly validated. If a dog is simply a dog and the question is whether its life is going well, the options are few and mostly health-framed. And in either case, the underlying measurement problem is the one that runs through all welfare work: the dog cannot report, so everything is either an owner's inference or an observer's inference (why behavior is not the same as emotion).
5. The Hard Half: Measuring Positive States
5.1 Why Negative Is Easier
Almost every physiological measure in wide use responds to threat, exertion, or disruption. Cortisol rises under challenge. Heart rate rises. Heart-rate variability falls. None of that distinguishes a dog that is frightened from a dog that is thrilled, because the arousal component is shared and the valence component is where the difference lies (what the stress measures can and cannot tell you).
Mendl, Burman, and Paul's framework makes the structure of the problem explicit: emotional states are usefully described along two axes, arousal and valence, and most available measures track the first far better than the second. An elevated cortisol value tells you something happened. It does not tell you whether the dog wanted it to.
5.2 The Experiment That Tested This Directly
Flint and colleagues ran the comparison properly. Sixty healthy adult dogs across three breeds — Labrador Retrievers, Beagles, Norfolk Terriers — went through six ten-minute scenarios designed to produce four combinations of valence and arousal: toy play and treat throwing (positive, high arousal), petting and a long-lasting chew (positive, low arousal), car travel (negative, high arousal), and social separation (negative, low arousal). Measures included cortisol, ACTH, heart rate, heart-rate variability, eye, nose and ear temperature, serotonin, secretory immunoglobulin A, panting, whining, body shaking, posture, activity, and qualitative behavioral assessment.
The result was informative in an awkward way. Cortisol, ACTH, heart-rate variability (RMSSD), panting, whining, and body shaking showed arousal-related differences primarily within the negative conditions. Several other measures — heart rate, secretory IgA, standing and lying — varied with arousal across both valences. Qualitative behavioral assessment was the best single indicator of positive valence. The authors' recommendation was that assessment should use a combination of parameters covering both valence and arousal, and they note that food provision, handler expectations, exercise, and ambient temperature all influenced results.
That is the current honest position: there is no single physiological readout that independently identifies positive emotional experience in a dog. The best candidate is a structured human judgment.
5.3 Behavior as the Positive-Welfare Indicator
De Winkel and colleagues systematically reviewed the observational behaviors and emotions used to assess dog welfare. They found 39 studies and grouped the indicators into nine themes. Five studies reported any form of validity assessment; 23 reported inter-rater reliability; and only six explicitly linked the observed behaviors to a welfare conclusion. The field uses behavioral indicators constantly and validates them rarely.
Cobb, Jiménez, and Dreschel make the parallel argument on the physiological side, and identify construct validity as the central unsolved problem: measures are routinely treated as indicators of welfare without the demonstration that they actually index the construct claimed. The recommendation in both cases is the same — multiple measures, interpreted jointly, with the limits stated.
5.4 What Owners Notice, and What They Miss
Buckland, Volk, Burn, and Abeyesinghe surveyed 445 owners in the UK and Ireland about how they recognize positive emotional states in their dogs. Owners described two distinct clusters. High-arousal positive states — "happy and excited" — were linked to anticipation of walks or food and described in terms of barking, raised head, wide bright eyes, pricked ears, tail wagging, open mouth, and active play. Low-arousal positive states — "happy and relaxed" — were linked to resting and gentle stroking and described as silence, head down, eyes closed or nearly closed, ears down, mouth closed.
The finding that matters: low-arousal positive states were described in considerably less detail than high-arousal ones, which the authors suggest indicates they are less well understood. Breed, age, owner experience, and level of attachment all influenced which indicators owners reported.
This is a practically important asymmetry. Excitement is conspicuous and gets counted as evidence of a good life. Contentment is quiet, easy to overlook, and easy to mistake for nothing happening. An assessment that relies on visible enthusiasm will systematically undervalue the states that make up most of a well-functioning dog's day (reading canine body language accurately).
5.5 Cognitive Measures
One class of measure does get closer to valence, by asking the animal to make a judgment rather than by reading its body. Judgment-bias testing presents an ambiguous stimulus and records whether the animal treats it optimistically or pessimistically, on the reasoning that mood biases expectation. It is slow, requires training, and is not a field method. But it is the approach that most directly targets the question the physiological measures cannot answer.
6. What Actually Changes a Dog's Welfare
6.1 Health First, Because It Is Underdiagnosed
Domain 3 is the domain most likely to contain an unrecognized problem. Pain in particular is systematically missed, because dogs do not reliably display it in ways owners recognize and because gradual-onset pain is absorbed into what the household considers the dog's personality (how pain is masked and missed). Musculoskeletal pain is associated with behavioral change including irritability and aggression (chronic pain and aggression), and visceral discomfort produces a different and even less legible set of signs (visceral pain and behavior).
Any welfare assessment that has not asked the health question properly is provisional. This is also the point at which age matters: in older dogs, changes that look like temperament or training loss may be cognitive dysfunction, and the distinction changes what should be done.
6.2 Sleep
Sleep belongs in Domain 2 and is almost never assessed. Dogs in busy households, dogs with no undisturbed resting place, and dogs whose rest is repeatedly interrupted accumulate a load that shows up as reduced tolerance rather than as tiredness (sleep, memory, and emotional processing). It is one of the cheapest welfare improvements available and one of the least often made.
6.3 The Noise Environment
Domain 2 also covers sound. Noise sensitivity is common, frequently unrecognized, and frequently comorbid with pain. On the intervention side, the evidence on auditory enrichment is more modest than the marketing suggests, and habituation to a played track happens faster than most protocols assume.
6.4 Behavioral Opportunity, Done Honestly
Domain 4 is where most owner effort goes and where the most confident claims are made. Some of them do not hold. The idea that more exercise produces a calmer dog is not supported in the form in which it is usually stated (exercise and behavior), and physical exertion is not interchangeable with the kind of activity that produces engagement.
That matters here because exercise is the default answer. When Mullan and Main asked owners what would improve their dog's quality of life, more exercise was the most frequent suggestion by a clear margin. It is the intervention that comes to mind first and it is easy to increase, but its effect on behavior problems is less clearly established than the simple formula “more exercise makes a calmer dog” suggests. An assessment that stops at "could he get more walks?" has asked the easiest question in Domain 4 and skipped choice, control, and challenge entirely.
What does appear to matter is variety in the kind of opportunity rather than volume of one kind. Olfactory activity uses the sensory channel dogs are actually built around (canine olfaction). Play has developmental and social functions beyond energy expenditure (the neurobiology of play). And social contact with other dogs is a real need for some individuals and a real stressor for others, which is the reason to assess the individual rather than apply a rule (how dogs behave socially when free-ranging).
6.5 The Human Interaction, Now Formally In Scope
The 2020 revision brought human behavior inside the assessment, which puts training method squarely in Domain 4. The evidence that aversive methods carry costs beyond their immediate effect is not marginal (punishment fallout, the neurological picture), and under a framework that treats human interaction as a welfare input, method choice is a welfare variable rather than a matter of preference.
The same applies to handling. Veterinary care, grooming, and husbandry are unavoidable, and how they are conducted is assessable (cooperative care). So is the arrangement of a multi-dog household, where resource access and the ability to withdraw are the variables that matter most (multi-dog households).
6.6 Load That Accumulates
Welfare is not assessed one event at a time. Repeated activation without recovery has consequences that single-event measurement misses (chronic stress), and the short-term version of the same phenomenon — several manageable stressors arriving close enough together to stop being manageable — is a routine feature of ordinary dog life (trigger stacking). A good day-to-day arrangement that never allows recovery is worse than it looks on any single day.
7. Assessing an Individual Dog
7.1 The Domains as a Checklist
For a working assessment, the five domains function as a structure that prevents omission. Nutrition: is intake appropriate, is feeding an event the dog engages with or a bowl that appears? Physical environment: temperature, noise, surfaces, and above all whether there is a place to rest that is not interrupted. Health: when was pain last actively looked for, rather than assumed absent? Behavioral interactions: what can this dog choose, what can it control, what is it allowed to decline, and what does it get to be competent at? Mental state: what is the balance of what this dog appears to experience across a normal week?
7.2 Observation Has to Be Operationalized
"He seems happy" is not an assessment, and neither is "she's stressed." Both are conclusions. What can be assessed is described behavior in described situations, at a level of specificity another person could replicate (operationalizing behavior for measurement). The same requirement applies to welfare assessment as to a behavior case (behavioral assessment): the value of the record is that it can be compared with a later record.
7.3 Baselines Are Individual
Dogs differ substantially in how they respond to the same conditions, and the differences are stable enough to be described as temperament and coping style (temperament and coping styles). A quiet dog and a reactive dog under the same arrangement may be having very different weeks. Assessment against the individual's own established pattern is more informative than assessment against a population norm, and the same holds for life stage: adolescence produces genuine, temporary change that is easy to misread as deterioration (adolescence in dogs).
7.4 The Relationship Is Part of the Assessment
How a dog uses its person under uncertainty is informative and has been studied as attachment (attachment styles). A dog that cannot settle in its owner's presence and a dog that ignores the owner's return are giving different information, and neither is captured by a health-focused instrument.
7.5 When Treatment Enters
Where anxiety is severe enough to prevent a dog from engaging with anything, welfare is not going to be improved by enrichment alone, and medication is a welfare intervention rather than an admission of failure (psychopharmacology in dogs). The assessment question is whether the dog's capacity to have positive experiences is being blocked, and by what.
8. Where the Assessment Actually Gets Made
8.1 The Dissemination Gap
Roberts, Blackwell, Roe, Murrell, and Mullan surveyed 110 UK veterinary professionals — 90 surgeons and 20 nurses — about canine quality-of-life assessment tools. Only 29.1% were aware that formal tools existed. Fewer than 4% reported actively using any of the three tools listed. Over half, 55.5%, said they would certainly or probably adopt such a tool. The dominant barrier reported was lack of time — for completion, for discussion, and for documentation — with perceived owner resistance second.
The authors' conclusion was that the tools are not well disseminated to people working in practice, creating a gap between the available evidence-based resources and what actually happens in the consulting room.
This reframes the problem. The literature's difficulty is validation; practice's difficulty is that even the validated instruments largely are not used. A tool nobody knows about does not improve any dog's welfare.
8.2 The Same Finding in Germany, With Better Reasons
Rhein and colleagues examined the same question in German small-animal practice with a mixed-methods design: a survey of veterinarians and veterinary assistants on awareness and use, plus interviews on how they perceive quality-of-life measurement. Their finding matched the UK picture — standardized quality-of-life assessment was rare, and where it happened it was often tied to research rather than to routine care.
The interviews add something the survey data alone does not. Four barriers emerged: lack of time, lack of awareness, a perceived lack of need, and resistance to fitting the individuality and subjectivity of quality of life into a schema. On the other side, the veterinarians named two concrete benefits: the ability to document and track quality of life over time, and an improvement in the relationship between professional and client.
The fourth barrier is the interesting one, because it is a real argument rather than a logistical complaint. Quality of life is individual and partly subjective, and an instrument that averages across dogs will misdescribe particular dogs. That objection is correct as far as it goes. What it does not support is the conclusion that no instrument should be used — the appropriate response to individuality is to compare a dog against its own earlier record rather than against a population norm, and that requires exactly the documentation an instrument provides. Without a record, the comparison being made is against the assessor's memory, which is not less subjective for being unwritten.
The authors' own conclusion is that research should stop concentrating solely on developing tools and start addressing their dissemination. Given that there are now two independent studies in two countries finding the same disuse, that seems right.
8.3 Who Ends Up Deciding
In the absence of instruments, quality-of-life judgments are still made — continuously, and by owners. They are made when a treatment is chosen, when a behavior problem is tolerated or addressed, when an older dog's mobility declines, and at the end of life. The owner's inference is the instrument, whether or not it is written down.
That is an argument for structure rather than for deference. An owner working through five domains and a set of described behaviors will reach a better-founded judgment than one asking whether the dog seems all right, and the structure is available even where the validated questionnaire is not.
9. Summary at a Glance
Welfare and quality of life are not synonyms. Welfare is the broad state across conditions and resulting mental states; quality of life in the veterinary literature is usually a narrower, instrument-based and often disease-framed measure.
The Five Domains Model organizes assessment into four condition domains and one mental-state domain. Nutrition, physical environment, and health produce survival-critical affects; behavioral interactions produce situation-related affects; mental state is inferred from those four domains and is the level at which welfare is assessed.
The 2020 revision brought human interaction inside the framework. Domain 4 became "Behavioral Interactions," covering environment, other animals, and humans — which makes handling and training method welfare variables.
Agency has three components that are not interchangeable. Choice, control, and challenge. Enrichment that supplies novelty without control, or activity without surmountable difficulty, is supplying one of three.
Validation of quality-of-life instruments is the field's weak point. Belshaw found 11 of 52 instruments with any reliability or validity evidence in 2015; Rhein found, a decade later, that of 41 publications only 12 described any evaluation and none were thoroughly evaluated.
Generic tools are scarce. Nine, by Fulmer's count. The best validated is Reid's web-based instrument, whose four domains — vitality, pain, distress, anxiety — are almost entirely health and negative affect.
No single measure reads positive emotion. In Flint's controlled comparison, most physiological and behavioral measures differentiated arousal mainly within negative conditions; qualitative behavioral assessment was the best candidate for positive valence.
Quiet positive states are underrecognized. Owners in Buckland's survey described high-arousal positive states in far more detail than low-arousal ones, suggesting contentment is less well understood than excitement.
The tools that exist are barely used. Under a third of surveyed UK veterinary professionals knew formal canine quality-of-life tools existed; under 4% used one. A German mixed-methods study found the same disuse, with time, awareness, perceived lack of need, and objection to schematizing something individual as the reasons given.
10. Research Gaps
Positive valence has no validated stand-alone indicator. Qualitative behavioral assessment is the leading candidate and is a structured human judgment rather than an objective readout. Whether any physiological measure can index positive states specifically remains open.
Construct validity is assumed more often than tested. Both the behavioral review by De Winkel and colleagues and the physiological critique by Cobb and colleagues identify the same problem: measures are used as indicators of welfare without the demonstration that they measure the construct they are claimed to measure.
Generic instruments remain thin. The healthy pet dog whose owner wants to know whether its life is going well is the least well served case in a literature dominated by disease-specific tools.
Nobody has quantified the weighting between domains. The Five Domains Model provides structure, not exchange rates. Whether a dog with mild chronic pain and rich behavioral opportunity is better or worse off than a pain-free dog with none is not answerable within the framework.
Agency has been argued more than measured. Choice, control, and challenge are conceptually well specified. How much of each a companion dog needs, and how to score what it has, is not established. The Animal Welfare Assessment Grid scores choice, control, and predictability as a single environmental factor — an advance, but a coarse one.
Implementation is a research question in its own right. Two independent studies, in the UK and Germany, found that quality-of-life instruments are largely not used in practice, and Rhein and colleagues conclude explicitly that dissemination now needs as much attention as development. Given that time is among the dominant reported barriers, the useful instrument may be the short one that gets used rather than the thorough one that does not.
The generic welfare question has barely been asked since 2007. Mullan and Main's screening questionnaire — including the item asking owners whether they would be willing to live their dog's life — was a 27-owner pilot. Nothing of comparable ambition for the healthy pet dog appears to have been scaled up since.
11. Conclusion
Welfare science has moved from asking whether an animal is suffering to asking what its life contains. That shift is well represented in the frameworks and much less well represented in the measurement tools. The Five Domains Model, in its 2020 form, gives a structure that covers positive as well as negative experience and explicitly includes the human as a welfare input. The instruments available for scoring a real dog against that structure are fewer, mostly disease-specific, largely under-validated, and — where they are validated — rarely used in practice.
The honest position is therefore neither that canine welfare assessment is solved nor that it is guesswork. The framework is sound and usable without any instrument at all: five domains, worked through deliberately, with health actively investigated rather than assumed, with agency broken into choice, control, and challenge, and with described behavior rather than conclusions as the record. What that will not give is a number. It will give something more useful — a profile of where a particular dog's life is thin, which is the thing that can be changed.
And it will guard against the most common error in this area, which is not cruelty. It is concluding that a dog whose needs are met and to whom nothing happens is a dog with good welfare.
Key Insights
Absence of suffering is not welfare. The Five Domains Model exists because deficit-based assessment passes dogs whose lives contain nothing good. Positive states have to be assessed separately, with different indices.
Domain 4 is where owners have the most leverage. Nutrition, environment, and health produce states that drive behavior automatically. Behavioral interactions produce states that depend on opportunity — which is the part an owner arranges.
Choice, control, and challenge are three things, not one. Most enrichment supplies novelty. Control — acting and having it work — is the component with the strongest supporting stress literature, and the one most often absent.
Ask the health question before any other. Pain is the most frequently missed welfare problem in dogs, and it is routinely absorbed into the household's description of the dog's character.
Rest is a welfare variable. Undisturbed sleep is among the cheapest improvements available and among the least often made.
No measure tells you a dog is enjoying itself. Cortisol, heart rate, and heart-rate variability track arousal, largely within negative contexts. The best current indicator of positive valence is a structured human judgment.
Learn to see quiet contentment. Owners describe excitement in detail and relaxed positive states vaguely. An assessment built on visible enthusiasm will miss most of a good life.
Write down behavior, not verdicts. "Settles within two minutes after the postman, eats normally afterward" can be compared with next month. "Seems happy" cannot.
Validated tools exist and are mostly unused. If a formal instrument is wanted, it has to be asked for; under 4% of surveyed UK veterinary professionals were using one, and German practices reported the same.
"More exercise" is the reflex answer, not the best one. It was the most common improvement owners named in Mullan and Main's screening study, and it is the Domain 4 intervention with the weakest behavioral evidence. Choice, control, and challenge are harder to arrange and more likely to matter.
One question is worth borrowing. Mullan and Main asked owners how willing they would be to take on the life their dog was living. It cuts through more than most scores do.
The framework works without the instrument. Five domains, health actively investigated, agency broken into its components, behavior described rather than judged. That is available to anyone, today.
References
Belshaw, Z., Asher, L., Harvey, N. D., & Dean, R. S. (2015). Quality of life assessment in domestic dogs: An evidence-based rapid review. The Veterinary Journal, 206(2), 203–212. https://doi.org/10.1016/j.tvjl.2015.07.016
Buckland, E. L., Volk, H. A., Burn, C. C., & Abeyesinghe, S. M. (2014). Owner perceptions of companion dog expressions of positive emotional states and the contexts in which they occur. Animal Welfare, 23(3), 287–296. https://doi.org/10.7120/09627286.23.3.287
Cobb, M. L., Jiménez, E. Y., & Dreschel, N. A. (2025). Beyond cortisol! Physiological indicators of welfare for dogs: Deficits, misunderstandings and opportunities. Journal of Applied Animal Welfare Science. Advance online publication. https://doi.org/10.1080/10888705.2025.2572616
De Winkel, T., van der Steen, S., Enders-Slegers, M.-J., Griffioen, R., Haverbeke, A., Groenewoud, D., & Hediger, K. (2024). Observational behaviors and emotions to assess welfare of dogs: A systematic review. Journal of Veterinary Behavior, 72, 1–17. https://doi.org/10.1016/j.jveb.2023.12.007
Flint, H. E., Weller, J. E., Parry-Howells, N., Ellerby, Z. W., McKay, S. L., & King, T. (2024). Evaluation of indicators of acute emotional states in dogs. Scientific Reports, 14, 6406. https://doi.org/10.1038/s41598-024-56859-9
Fulmer, A. E., Laven, L. J., & Hill, K. E. (2022). Quality of life measurement in dogs and cats: A scoping review of generic tools. Animals, 12(3), 400. https://doi.org/10.3390/ani12030400
Littlewood, K. E., Heslop, M. V., & Cobb, M. L. (2023). The agency domain and behavioral interactions: Assessing positive animal welfare using the Five Domains Model. Frontiers in Veterinary Science, 10, 1284869. https://doi.org/10.3389/fvets.2023.1284869
Malkani, R., Paramasivam, S., & Wolfensohn, S. (2022). Preliminary validation of a novel tool to assess dog welfare: The Animal Welfare Assessment Grid. Frontiers in Veterinary Science, 9, 940017. https://doi.org/10.3389/fvets.2022.940017
Mellor, D. J. (2017). Operational details of the Five Domains Model and its key applications to the assessment and management of animal welfare. Animals, 7(8), 60. https://doi.org/10.3390/ani7080060
Mellor, D. J., & Beausoleil, N. J. (2015). Extending the ‘Five Domains’ model for animal welfare assessment to incorporate positive welfare states. Animal Welfare, 24(3), 241–253. https://doi.org/10.7120/09627286.24.3.241
Mellor, D. J., Beausoleil, N. J., Littlewood, K. E., McLean, A. N., McGreevy, P. D., Jones, B., & Wilkins, C. (2020). The 2020 Five Domains Model: Including human–animal interactions in assessments of animal welfare. Animals, 10(10), 1870. https://doi.org/10.3390/ani10101870
Mendl, M., Burman, O. H. P., & Paul, E. S. (2010). An integrative and functional framework for the study of animal emotion and mood. Proceedings of the Royal Society B: Biological Sciences, 277(1696), 2895–2904. https://doi.org/10.1098/rspb.2010.0303
Mullan, S., & Main, D. (2007). Preliminary evaluation of a quality-of-life screening programme for pet dogs. Journal of Small Animal Practice, 48(6), 314–322. https://doi.org/10.1111/j.1748-5827.2007.00322.x
Reid, J., Wiseman-Orr, M. L., Scott, E. M., & Nolan, A. (2013). Development, validation and reliability of a web-based questionnaire to measure health-related quality of life in dogs. Journal of Small Animal Practice, 54(5), 227–233. https://doi.org/10.1111/jsap.12059
Rhein, F. F., Klee, R., Albrecht, B., & Krämer, S. (2025a). Assessment of the awareness and use of quality of life tools in small animal practices in Germany. Animals, 15(24), 3617. https://doi.org/10.3390/ani15243617
Rhein, F. F., Klee, R., Albrecht, B., & Krämer, S. (2025b). Instruments to assess disease-specific quality of life in dogs: A scoping review. Animals, 15(12), 1780. https://doi.org/10.3390/ani15121780
Roberts, C., Blackwell, E. J., Roe, E., Murrell, J. C., & Mullan, S. (2023). Awareness and use of canine quality of life assessment tools in UK veterinary practice. Animals, 13(6), 1105. https://doi.org/10.3390/ani13061105