Therapy Dog Welfare: What Animal-Assisted Interventions Mean for the Dog
Michael Sauerwein
Therapy, visiting and school dogs are expected to help people, but what does the work mean for the dogs themselves? Studies measuring cortisol, heart rate and behavior during sessions mostly report no statistically significant increase in the stress markers examined, and some find signs of relaxation by the end of a visit (Ng et al., 2014; Clark, Martin, et al., 2020; Shih et al., 2025). At the same time, the evidence rests largely on small samples of experienced, certified dogs, on few control conditions, and partly on handlers rating their own dogs (Glenk & Foltin, 2021; Haven-Pross et al., 2026).
This article reviews what physiological and behavioral studies show during sessions, which factors are associated with stress, what is known about positive welfare and enjoyment, and where the methods reach their limits. It closes with what the findings mean for designing programs that protect the dog.
1. Why the Dog's Side Is a Research Question
1.1 Terms and Settings
Animal-assisted interventions (AAI) is the umbrella term for structured work in which animals are deliberately included to benefit people. The international organization IAHAIO distinguishes goal-directed animal-assisted therapy (AAT), animal-assisted education (AAE), which includes school and reading programs, animal-assisted activities (AAA), such as informal visits to care homes, hospitals or universities, and animal-assisted coaching or counseling (AAC) (IAHAIO, 2018). A Dutch field study analyzed these four formats separately (Haven-Pross et al., 2026).
For the dog, these formats differ considerably: a visiting dog may meet many strangers in a waiting room, while a reading dog lies beside one child (what studies show about dogs in schools). A review proposing selection standards emphasized that AAT places different demands on the dog than AAA or assistance work, and that a mismatch between dog, job and handler can produce stress, disinterest or behavior problems (Winkle et al., 2020).
1.2 Welfare Is More Than the Absence of Stress
Most studies ask whether dogs are stressed during sessions. That is a necessary question, but welfare also includes positive states: interest, relaxation, social pleasure, play (how welfare science defines quality of life in dogs). A perspective paper argued that research has concentrated on stress and that positive-welfare indicators, such as play, leaning or nudging, and temperament deserve equal attention (Miller et al., 2022).
The IAHAIO guidelines state that AAI should only be performed with animals that are in good physical and emotional health and that "enjoy this type of activity", and that the animal "is not simply a tool" (IAHAIO, 2018). Whether a dog enjoys the work is an empirical question, and a hard one.
1.3 What Kind of Evidence Exists
The literature consists mainly of small observational studies and a few larger field datasets. A review of therapy dog welfare studies published between 2017 and 2021 found that eight of twelve studies had fewer than ten dogs, with a median of 7.5 dogs (Glenk & Foltin, 2021). All included studies concerned visitation programs, with sessions from 10 to 90 minutes.
An earlier review of about ten studies, with sessions from 10 minutes to 8 hours and indicators from salivary, fecal and hair cortisol to handler perceptions, concluded that session frequency and duration, novelty, controllability, and recipients' age and familiarity modulate welfare indicators, but that clear conclusions are lacking because programs differ, samples are small and methods are limited (Glenk, 2017).
2. How Welfare Is Measured During Sessions
2.1 Salivary Cortisol
Cortisol is the most frequently used indicator, because saliva can be collected without a needle. But salivary cortisol is affected by more than the session. A meta-analysis of 61 studies, with raw data from 31 datasets covering 5,153 samples from 1,205 dogs, found a mean of 0.45 µg/dL with a range from 0 to 33.79 µg/dL and significant effects of sex and neuter status, age, living environment, time spent in the environment before testing, the testing environment, owner presence and collection media (Cobb et al., 2016). The authors warned that large intra- and inter-individual variability makes comparisons between studies difficult (how stress in dogs is measured and where the methods fail).
The collection method itself matters: collection material, saliva volume and stimulants such as beef flavoring or citric acid can distort results (Dreschel & Granger, 2009). Time-of-day effects may occur, but a uniform circadian pattern has not been consistently demonstrated in dogs. In 23 dogs, serum cortisol was higher between 10:00 and 13:00 than between 19:00 and 22:00, but a daily rhythm appeared only in the control group, not in the working dogs (Kolevská et al., 2003). Serum is not saliva, but the finding shows why sampling time should be controlled.
A further problem is valence. When ten female owners stroked and talked to their male Labrador Retrievers for three minutes, the dogs' oxytocin rose, and so did their blood cortisol at 15 and 30 minutes (Handlin et al., 2011). Cortisol can therefore rise after social touch and interaction; whether the dogs experienced it as pleasant was not measured. In service dogs for veterans with PTSD, an indirect comparison, salivary cortisol after free play was higher than after training, and the authors noted that heightened cortisol can also reflect positive arousal (van Houtert et al., 2021). Cortisol indicates activation, not whether the dog experiences it as good or bad (how acute and chronic stress differ in the body).
2.2 Heart Rate, Heart Rate Variability and Temperature
Heart rate (HR) and heart rate variability (HRV) can be recorded continuously with chest belts; higher HRV is generally read as stronger parasympathetic influence. Tympanic membrane temperature has been added as a possible marker of emotional state. Each comes with technical problems: in a single-case study, the HR signal was usable for only half of the sessions (Palestrini et al., 2017). A preprint, not yet peer-reviewed, analyzed ambulatory ECG recordings from two certified therapy dogs across 38 canine-assisted psychotherapy sessions and found that the amount of usable data depended strongly on preprocessing, with 10,075 one-minute windows retained under permissive and 4,529 under strict criteria (Suttkus et al., 2026). The windows were shifted by 10 seconds and overlap heavily, so they are not independent observations.
2.3 Oxytocin
Oxytocin is discussed as a possible indicator of positive social states (what oxytocin does in dogs and what it cannot show). After positive interaction between people and dogs, oxytocin, beta-endorphin, dopamine and other neurochemicals increased in both species, while cortisol decreased in humans only (Odendaal & Meintjes, 2003). These were pet dogs, not therapy dogs. In therapy dog studies, salivary oxytocin has so far mostly remained stable across sessions (Clark, Martin, et al., 2020; Shih et al., 2025).
2.4 Behavior
Behavioral data are collected with ethograms, lists of defined behaviors that are coded from video or live. Typical stress-associated behaviors include panting, lip licking, yawning, shaking off, turning away and lowered posture (what lip licking and yawning do and do not mean). Many of these are ambiguous: lip licking and body shaking can also accompany positive arousal (Glenk, 2017), and panting depends on temperature and exertion. Coding is only as reliable as the observers and definitions (how behavior is defined and measured in research).
2.5 Handler Ratings
Handlers know their dogs and attend every session, so their ratings are practical, but they are not independent observers, and their match with physiology varies (section 7).
3. Physiological Findings During Sessions
3.1 Early Work: Higher Cortisol on Therapy Days
In the first systematic studies, from Austria, owners collected saliva on control days without work and directly before and after each session over three consecutive months (Haubenhofer & Kirchengast, 2006). Cortisol was significantly higher on therapy days than on control days and rose with the number of sessions in the sampling period. Dogs working in the morning had higher values after than before sessions, while dogs working in the afternoon had lower values after sessions; values were higher in short than in long sessions. The authors concluded that the work was physiologically arousing but that it remained "an open question" whether this reflected negative stress or positive excitement.
A related paper with 13 handlers and 18 dogs also found higher cortisol on therapy days in both dogs and handlers, with dog cortisol rising with the number of sessions per week (Haubenhofer & Kirchengast, 2007). The emotions handlers attributed to their dogs did not match the cortisol results. The two papers probably draw on overlapping data and should not be counted as independent confirmation. The morning and afternoon pattern suggests that time of day also shaped the results.
3.2 Mental Healthcare and the Lead
In weekly group sessions in adult mental healthcare, 21 certified therapy dogs and therapy dogs in training were sampled over the course of sessions and on non-working days (Glenk et al., 2013). Cortisol did not indicate that either group was stressed by the sessions, but in certified dogs it varied significantly between working on lead and off lead. These were distinct, established handling groups, not randomly assigned, so the study does not establish a causal effect of leash use. In her later review, the first author reported that cortisol was lower in experienced dogs working off lead and free to move, with sessions of 50 to 60 minutes (Glenk, 2017). The authors called for further study of lead use and of ways to let dogs approach or avoid contact.
3.3 Activity, Novel Room and Home: A Within-Dog Comparison
One of the best-controlled designs compared the same dogs across three 60-minute conditions: an animal-assisted activity for college students in a residence hall, an unfamiliar room without stranger contact but with the handler present, and the handler's home (Ng et al., 2014). Fifteen healthy adult dogs registered with an AAA organization in the United States completed the study. Saliva was sampled every 30 minutes before, during and after each condition, and five-minute petting segments were coded from video.
At 30 minutes, cortisol was higher in the unfamiliar room (0.397 µg/dL) than during the activity (0.257 µg/dL; P = 0.01) and at home (0.213 µg/dL; P = 0.03). During the activity, dogs stood more (59% versus 0% at home, P = 0.008) and moved around more (5.6% versus 0%, P = 0.001) than at home. Over the 60-minute activity, cortisol and stress behavior did not differ from home. The authors concluded that a 60-minute AAA "conducted in a safe and controlled manner" did not appear to elicit significant activation of the stress axis or observable stress-associated behaviors, and they emphasized the predictability of the environment and the role of the handler (how dogs respond to novelty).
The authors named a small, heterogeneous sample and uncontrolled cues from students and other dogs as limits. Because the settings differed in other ways too, the higher cortisol in the unfamiliar room cannot be attributed to novelty alone. The study was funded by a pet food company.
3.4 Hospital Studies: Visits, Frequency and Relaxation
Several studies come from one large medical center in the United States. In a pilot study with four therapy dog teams in an outpatient nursing unit, a 4 × 4 Latin square compared four visit schedules, from two visits per week to one visit in four weeks (Clark et al., 2019). Only the most frequent schedule showed a significant pre-to-post difference (P = 0.04), and it was a reduction after the visit. The youngest dog had the highest post-visit cortisol in three of the four schedules. With four dogs and cortisol as the only measure, this is a pilot finding.
A larger study measured 19 therapy dogs during five 20-minute visits each with outpatients with fibromyalgia, embedded in a randomized trial with 222 patients (Clark, Martin, et al., 2020). Dogs could move freely and interact or not. HR and HRV were recorded continuously; tympanic temperature, cortisol and oxytocin were measured before and after. HR was lower at the end of sessions than at the beginning (p = 0.0003), and right tympanic temperature was lower afterward (p = 0.003); all other indicators, including cortisol and oxytocin, remained stable. The authors read this as some evidence of a more relaxed state at the end of a session; a lower heart rate is not proof of enjoyment or of reduced stress. Behavior was not recorded, because video was not permitted to protect patients.
A pilot study followed nine new teams whose dogs, mean age 3.7 years, had no prior visiting experience, across their first three visits in the same institution: an orientation walk, an outpatient waiting room and inpatient rooms (Clark, Smidt, et al., 2020). Visits averaged 47 minutes, with about nine people per visit. Dog cortisol did not differ significantly across time points or visits. Handlers' ratings of their dog's stress correlated with dog cortisol (P = 0.02): dogs rated as medium or high stress had higher post-visit cortisol. The most frequently reported behaviors were panting, lip licking and yawning. Limitations included cortisol as the only biomarker and uncontrolled visit times.
In an inpatient stroke rehabilitation program, patients were randomized to standard rehabilitation with or without animal-assisted therapy; 14 dog and handler pairs worked in 30-minute active sessions with walking, fetching and brushing, with a veterinarian present (Shih et al., 2025). HR decreased and HRV increased significantly after sessions, right tympanic temperature increased, and cortisol, oxytocin and stress behavior did not change. The randomization concerned the patients; for the dogs, there were only pre- and post-session measurements. The authors cautioned that the highly trained, certified teams "may not be representative for most therapy dogs".
3.5 Work with Children
In pediatric oncology, 26 therapy dog and handler teams visited newly diagnosed children with cancer at five children's hospitals in the United States over four months (McCullough et al., 2018). Saliva was collected at five baseline time points and 20 minutes after the start of sessions, and all sessions were video-coded. Cortisol did not differ between baseline (0.51 µg/dL) and session (0.44 µg/dL; p = 0.757). Higher cortisol was associated with more stress behaviors per session (p = 0.039). The authors found no significant physiological stress response and no predominance of stress-related over affiliative behaviors. A later review summarized that in this study dogs showed more stress behaviors and fewer affiliative behaviors when recipients acted directly on the dog (Glenk & Foltin, 2021).
Nine certified therapy dogs each took part in six sessions with groups of three to four children with ADHD (Melco et al., 2020). Each session had five activities: the first and last with only dog and owner as control periods, and three in between, including reading to the dog. HR and saliva were collected after each activity. Stress behaviors occurred only occasionally, and cortisol and HR did not differ between activities. Because the "control" periods were part of the same session day, the study cannot show how the dogs compared with a day without work.
Two single-case studies offer detailed observations. A seven-year-old experienced Golden Retriever worked in 20 sessions of 20 minutes with 20 children aged 3 to 17 years recovering from surgery in Italy (Palestrini et al., 2017). The dog panted for 28% of the time on average, almost never withdrew, interacted more with its handler (8.6%) than with the child (4.9%), and showed no correlation between stress behaviors and the children's interactions; HR stayed between 60 and 110 beats per minute in the ten sessions with usable data. In Australia, a three-year-old labradoodle with two years of experience worked with eleven autistic children aged four to six in occupational therapy over seven weeks, on workdays with up to six 50-minute sessions and breaks of at least ten minutes (Hill et al., 2023). Across 21 saliva samples and six hours of coded video, there were no significant differences between home and treatment days. The dog's handler and owner was the first author, which makes observer bias possible. The work-day schedule describes this one dog and is not a recommended load.
In a group intervention near Naples, four experienced dogs worked weekly for one hour with ten autistic children over 90 days (Giuliano et al., 2025). Cortisol and oxytocin did not change significantly, although oxytocin tended to increase in half of the dogs and cortisol tended to increase in all dogs at the end compared with before. Handlers reported no stress signs, but the handler questionnaire was the only behavioral source.
3.6 Adults, Dementia and Reading Programs
Four handler and dog pairs worked with adults with dementia and psychomotor dysfunction in five 45-minute sessions each (Pirrone et al., 2017). Behavioral synchrony between handler and dog was highest during sessions, and dogs followed 74% of handler cues (how dogs and people synchronize their behavior). Handlers' cortisol was higher before than after sessions, with the same non-significant trend in dogs. Dog HR was higher on activity days than on control days but within the physiological range; the authors attributed this to positive arousal, an interpretation the data cannot confirm. One dog refused guided physical contact more often and had higher cortisol even on control days, an individual difference that group averages hide.
For reading-to-dog programs, data on the dogs are scarce. One study followed two dogs during 30-minute reading sessions with four children with pervasive developmental disorders (Uccheddu et al., 2019). One dog had high cortisol on arrival, but neither showed behavioral or physiological stress during or after sessions. A systematic review of 48 papers on children reading to dogs assessed human outcomes only and did not address the dogs' welfare (Hall et al., 2016).
3.7 Reading the Physiological Studies Together
Overall, most studies found no statistically significant increase in the stress markers examined during sessions, and several found lower HR, higher HRV or falling cortisol by the end (Ng et al., 2014; Clark et al., 2019; Clark, Martin, et al., 2020; Shih et al., 2025). The early Austrian work stands apart with higher cortisol on therapy days (Haubenhofer & Kirchengast, 2006, 2007); the 2021 review found working cortisol mostly similar to non-working baselines (Glenk & Foltin, 2021). This is reassuring, but non-significant results in small samples do not prove the absence of stress, and they apply to dogs that had already passed selection, in programs willing to be studied.
4. Behavioral Findings
4.1 Observed Behavior in Sessions
In Italy, nine dogs working with patients with mental and psychomotor disabilities were observed for 174 hours before, during and after sessions with a 53-item ethogram (Corsetti et al., 2019). No aggressive or stereotypic behavior was recorded, and anxiety-related behavior was low and similar across contexts. During sessions, dogs showed more attention, affiliative behavior, play and sniffing than before or after, and their attention was directed mostly at the handler. The ethogram also contained a category labeled "dominance", which was more frequent during sessions; this construct is scientifically contested (why dominance fails as an explanation for dog behavior).
Four registered therapy dogs on six-foot leashes were coded during 50 interactions with adolescents completing a social stress task (Dowling-Guyer et al., 2024). Lying lasted longest and shifting position was most frequent; 14 of 28 behaviors, including panting and yawning, had a median of zero. Dog behavior varied with whether and how the adolescent touched the dog.
4.2 A Different Picture at Outdoor Events
Six dogs were video-coded during seven AAI sessions at two outdoor events (Erichsmeier et al., 2026). Appeasement behaviors were the most frequent category, at 10.89 per minute (SD 3.91), followed by posture changes at 5.95 per minute (SD 2.36). Every dog showed avoidance at least once per session. Most owners rated their dog's stress as 1 or 2 on a four-point scale. The authors found positive and negative indicators side by side and concluded that handlers may need education to advocate for their dogs.
4.3 The Largest Dataset: 63 Dogs, 837 Sessions
A Dutch field study collected behavioral scores from 837 sessions with 63 dogs and 30 professionals across AAA, education, coaching and therapy (Haven-Pross et al., 2026). The handlers themselves scored 19 behaviors after completing a six-hour online training and a competency test, and dog personality was assessed with a 45-item Dog Personality Questionnaire. Separate principal component analyses for each intervention type (AAA, AAE, AAC, AAT) yielded different behavioral components, not one set of factors valid across settings. Older dogs were less playful and more settled; experience was linked to positive affect in AAA and therapy but not in coaching; female dogs showed more uncertainty or arousal in AAA and education; the effect of session length depended on context; unfamiliar handlers increased tension in coaching; and younger clients were associated with more uncertainty and tension.
The important caveat: 70% of handlers worked with their own dogs, and another 14% with both their own and other dogs, so many dogs were scored by their own handlers. The authors name self-assessment bias, retrospective scoring with possible recall bias, rater drift and the lack of repeated inter-rater checks as limitations.
A companion study derived welfare thresholds for the same 19 behaviors from ratings by 13 experts and compared them with the same field data (Visser et al., 2026). Stress behaviors were generally within acceptable ranges, though some dogs exceeded upper thresholds, for example for low posture and sniffing. Positive behaviors such as play, voluntary lying down and broad tail wagging were infrequent and often below the experts' minimum. Overt stress seemed largely managed, but opportunities for positive affect may be limited. The thresholds are expert opinion, not independently validated. Both papers use the same dataset of 63 dogs and 837 sessions, so they are not independent replications.
In practice: Many experienced handlers end a session early when their dog repeatedly turns away, licks its lips, yawns or stops engaging with the client. This is practice experience, not a validated stopping rule.
5. Factors Associated with Stress
5.1 Session Length and Frequency
The findings on workload do not point in one direction. In the early Austrian work, cortisol rose with the number of sessions per sampling period and per week, but was higher in short than in long sessions (Haubenhofer & Kirchengast, 2006, 2007). In the hospital pilot with four dogs, the most frequent visit schedule went with a cortisol reduction after visits (Clark et al., 2019). In the Dutch field data, the effect of session length depended on the context (Haven-Pross et al., 2026).
Guideline values rest on expert judgment. IAHAIO recommends that animals "must not be overworked or overwhelmed" and that sessions be limited to 30 to 45 minutes, a guideline recommendation rather than an empirically established safe duration (IAHAIO, 2018). In a survey cited by Glenk (2017), respondents suggested no more than 60 minutes per visit and no more than three sessions per week. A survey of US therapy dog organizations (4 of 6 large national and 24 of 33 regional organizations responded) found that about half set no limit on visit length and left it to the handler, while several with limits allowed 1.5 to 2 hours (Serpell et al., 2020). The authors stated that reliable evidence on safe time limits is sparse.
5.2 Setting, Novelty and Transport
In the within-dog comparison, the unfamiliar room, with the handler but without stranger contact, was associated with higher cortisol at 30 minutes than the AAA or home (Ng et al., 2014). Novelty was not isolated as the cause. Whether familiarity with the setting contributed remains unresolved (how repeated exposure can calm or sensitize).
Transport is often overlooked. The 2021 review found that transport of 50 to 60 minutes was associated with higher arousal (Glenk & Foltin, 2021). In an exploratory study with five kennel-housed shelter dogs taken weekly to AAI sessions in a prison, cortisol after about one hour of transport was higher than at the kennel and at the end of the session (both p = 0.043) (d'Angelo et al., 2021). With five dogs and no control group, this is an indication that the journey can be more demanding than the session (how successive stressors add up).
5.3 Recipients: Age, Number and Behavior
The age and familiarity of recipients was among the factors identified in the 2017 review (Glenk, 2017). In the Dutch data, younger clients were associated with more uncertainty and tension (Haven-Pross et al., 2026). What recipients do also matters: in the pediatric oncology study, dogs showed more stress behaviors when recipients acted directly on them (Glenk & Foltin, 2021), and in a qualitative study of service dogs for children with autism, stressors included unintentional maltreatment and prodding by children, unpredictable routines, too little recreation and lack of rest after working (Burrows et al., 2008). These were assistance dogs living with families, not therapy dogs, so this is indirect evidence.
IAHAIO lists dressing animals up, physically challenging tasks, and children jumping on or bending over animals as inappropriate (IAHAIO, 2018). No study has varied the number of participants as a factor.
5.4 Leash, Choice and the Option to Withdraw
Controllability is one of the most consistent themes in stress research (why control and predictability reduce stress). In therapy work, evidence comes from two studies. In mental healthcare, cortisol differed between established, non-randomized on-lead and off-lead groups of certified dogs (Glenk et al., 2013; Glenk, 2017).
A direct experiment compared two three-minute conditions with 18 certified therapy dogs and 44 participants: forced touch, with the dog on leash and touched continuously, and free choice, with the dog roaming in a pen and touched only if it approached (Sarrafchi et al., 2025). Dogs showed more ears-back during forced touch (p = .0115) and more sniffing and avoidance during free choice (both p < .0001), and in the free-choice condition they stayed within reach 77.9% of the time. Avoidance can only occur where it is possible, so its higher frequency in the free-choice condition reflects the design as much as the dog. The authors recommended building choice and agency into therapy programs (how dogs respond to different kinds of touch).
A practice paper argued that dogs express "assent and dissent", an interpretive and ethical concept rather than a validated scale, that humans often miss subtle stress signals, and that an advocate for the dog matters (Townsend & Gee, 2021). The 2021 review called the dog the "least controlling participant" in AAI and questioned whether consent can be assumed (Glenk & Foltin, 2021).
In practice: Many teams set up a retreat area in the room, such as a mat or crate in a quiet corner that clients are asked not to approach, and plan breaks in advance rather than waiting for the dog to need one. Practice experience suggests that a dog that can leave contact at any time is easier to read, because withdrawal becomes visible instead of being suppressed by the lead.
5.5 Handler Factors
Several studies associate handler stress with dog stress. In a university stress-reduction program with 754 students, 40 handlers and 40 dogs, observational dog stress increased from home to the end of sessions, but not from start to end of sessions; handlers with higher self-reported stress had dogs with higher end-of-session stress (Silas et al., 2019). The authors suggested emotional contagion or spillover from handlers rather than from students. In a prison AAT program with trained pet dogs, dog stress was associated with handlers' self-reported stress, and cortisol decreased after sessions in most cases (Koda et al., 2015). In 58 pairs of Shetland Sheepdogs or Border Collies and their owners, outside AAI, owners' long-term hair cortisol correlated with their dogs' (Sundman et al., 2019) (how human stress can transfer to dogs).
These are correlations; none of these studies tested whether reducing handler stress reduces dog stress. Handlers may also be a resource: in two observational studies, dogs attended or interacted more with their handler than with clients (Palestrini et al., 2017; Corsetti et al., 2019) (how a familiar person can buffer stress).
An online survey of 111 French handlers reporting on 165 dogs named the spatio-temporal framework of sessions, interactions with beneficiaries and the handler as risk factors; the authors identified a potential positive bias in handler representations (Mignot et al., 2022).
In practice: Experienced handlers often describe their main job during a session as watching their own dog rather than the client, and stepping in before the dog has to. Treating the handler as the dog's advocate, with explicit permission to interrupt or end a session regardless of the therapeutic plan, is a principle several authors also call for (Townsend & Gee, 2021; Erichsmeier et al., 2026).
5.6 Dog Factors: Age, Experience and Sex
In the Dutch field study, age, experience and sex were associated with behavior in some formats (Haven-Pross et al., 2026). In the hospital pilot, the youngest dog had the highest post-visit cortisol (Clark et al., 2019), and in the free-choice experiment, male dogs avoided more (Sarrafchi et al., 2025). These findings are not consistent enough to recommend a minimum age or a preferred sex; they mainly show that individuals differ (how temperament and coping style differ between dogs). Evidence for retirement criteria is lacking (Ng & Fine, 2019).
6. Positive Welfare and Enjoyment
6.1 What Points to Positive States
Some findings are compatible with positive experiences: more play and affiliative behavior during sessions (Corsetti et al., 2019), lower HR and changes in tympanic temperature by the end of sessions (Clark, Martin, et al., 2020; Shih et al., 2025), higher HRV after sessions (Shih et al., 2025), and higher HR on activity days attributed by the authors to positive arousal (Pirrone et al., 2017). In the free-choice condition, dogs spent most of the time within reach of the participant without being made to (Sarrafchi et al., 2025).
6.2 Why Enjoyment Remains Unproven
None of these measures, including play during sessions, distinguishes reliably between relaxation, positive engagement and fatigue. Lower HR may reflect relaxation, habituation or reduced activity; rising HR may mean positive arousal or tension (what play signals reveal). The expert-threshold study found positive behaviors infrequent and often below what experts considered desirable (Visser et al., 2026). In 2022, a perspective paper noted that no study had yet combined positive and negative welfare indicators with temperament assessment (Miller et al., 2022). The Dutch field study partly addressed this by combining behavior scores with a personality questionnaire (Haven-Pross et al., 2026), but validation of positive affect remains limited.
A review of 116 papers on dogs and horses in animal-assisted services found ethical frameworks that were diverse and often under-specified, with positive- and negative-welfare perspectives frequently conflated (Zenker et al., 2026). How researchers frame the animal shapes what they measure.
7. Methodological Problems
7.1 Cortisol Timing
Cortisol differs widely between individuals, and time-of-day effects may occur without a consistently demonstrated circadian pattern (Cobb et al., 2016; Kolevská et al., 2003). Studies sampled at different times: 20 minutes after the start of a session (McCullough et al., 2018), about five minutes after the last interaction (Clark, Smidt, et al., 2020), or every 30 minutes (Ng et al., 2014). In the visit pilot, home baselines were taken between 06:00 and 08:00, while visit times could not be controlled (Clark, Smidt, et al., 2020). The reversed morning and afternoon pattern in the early Austrian study illustrates how time of day can produce or mask effects (Haubenhofer & Kirchengast, 2006).
7.2 Novelty and Transport as Confounds
If a dog is sampled after a car journey to a new place, the result reflects journey, place and session together. The within-dog design tried to separate these components, and even there the higher cortisol in the unfamiliar room cannot be attributed to novelty alone, because the settings differed in other ways (Ng et al., 2014). Transport data from the 2021 review and the shelter-dog study suggest the journey alone can raise arousal (Glenk & Foltin, 2021; d'Angelo et al., 2021).
7.3 Missing Control Conditions
Many studies compare pre- and post-session values on the same day (Clark, Martin, et al., 2020; Shih et al., 2025) or treat dog-only periods within the session as control (Melco et al., 2020). A non-working day under otherwise identical conditions, including transport and the same room, is rarely used.
7.4 Small Samples and Selected Dogs
Sample sizes are small: one dog (Palestrini et al., 2017; Hill et al., 2023), two (Uccheddu et al., 2019), four (Clark et al., 2019; Pirrone et al., 2017; Giuliano et al., 2025; Dowling-Guyer et al., 2024), five (d'Angelo et al., 2021), six (Erichsmeier et al., 2026). Even the larger studies have 15 to 26 dogs (Ng et al., 2014; Clark, Martin, et al., 2020; McCullough et al., 2018). The only dataset with 63 dogs relies on handler ratings (Haven-Pross et al., 2026). The dogs studied are usually certified, experienced and selected, and belong to motivated volunteers; dogs that dropped out or were never certified are not represented, so the evidence describes those that cope.
7.5 Handler-Rated Welfare
Handler ratings agree with physiology in some studies and not in others. Handler-perceived dog stress correlated with cortisol in the visit pilot (Clark, Smidt, et al., 2020), whereas handler-attributed emotions did not match cortisol in the Austrian work (Haubenhofer & Kirchengast, 2007). In the prison program, dogs were rated as stressed in 11% of cases without a significant change in cortisol (Koda et al., 2015). At the outdoor events, every dog showed avoidance, but most owners rated stress low (Erichsmeier et al., 2026). The French survey suggested a positive bias in how handlers see welfare (Mignot et al., 2022). Where handlers are the only behavioral source (Giuliano et al., 2025) or score their own dogs (Haven-Pross et al., 2026), findings must be read with this in mind.
7.6 Reporting, Ethics and Funding
A review of 139 AAI research articles found that only about 10% reported approval by an animal care and use committee and about 3% an adverse outcome for at least one animal; training, certification and veterinary care were inconsistently reported (Ng et al., 2019). In the Dutch field study, the dogs were not classed as research animals (Haven-Pross et al., 2026). A One Health framework has called for integrated designs that measure benefits and costs for humans and animals together (Hediger et al., 2019), and an overview of the field highlighted animal welfare research and guidelines as central issues (Fine et al., 2019).
8. Guidelines and Standards
The IAHAIO guidelines, which also require regular veterinary checks, are an organizational white paper, not a peer-reviewed study (IAHAIO, 2018). The US survey found little consensus on strictly reward-based training (Serpell et al., 2020). A narrative review of European and US frameworks described governance as fragmented and proposed moving from dog-level pass-or-fail certification toward assessment of the team in its context, with reward-based, consent-oriented preparation, multimodal monitoring, workload regulation and retirement planning (Shajid Pyari et al., 2026). Winkle et al. (2020) likewise tie suitability to the fit between dog, task and handler, not to a test passed once.
9. Practical Consequences for Program Design
9.1 What the Studies Support
Within the limits of the evidence, several design principles are consistent with the data. Allowing the dog to move freely and to approach or avoid contact is consistent with the non-randomized on-lead and off-lead comparison and the free-choice experiment (Glenk et al., 2013; Sarrafchi et al., 2025). Reducing novelty, for example by visiting familiar rooms, is consistent with the higher cortisol in an unfamiliar room, although novelty was not isolated as the cause (Ng et al., 2014). Keeping transport short or allowing time to recover after arrival is supported by the transport findings (Glenk & Foltin, 2021; d'Angelo et al., 2021). Preventing recipients from acting on the dog in ways it cannot avoid is consistent with the pediatric data and, indirectly, the assistance-dog data (Glenk & Foltin, 2021; Burrows et al., 2008).
9.2 Monitoring Instead of Assumptions
Because handler ratings can be biased, programs benefit from structured observation, such as a short ethogram or video review by a second person. The visit pilot showed that handlers can perceive their dog's stress (Clark, Smidt, et al., 2020); the outdoor study and the survey showed that they can also underestimate it (Erichsmeier et al., 2026; Mignot et al., 2022) (why owner and handler ratings can miss the mark). Training handlers to read subtle signals is a recurring recommendation (Townsend & Gee, 2021; Mignot et al., 2022) (how to read canine body language).
9.3 Workload and Career
No study establishes a safe maximum session length or frequency (Serpell et al., 2020). The guideline limit of 30 to 45 minutes (IAHAIO, 2018) or the survey values of up to 60 minutes and three sessions per week reported by Glenk (2017) are reasonable starting points, not thresholds (what research shows about session length and breaks in training). Criteria for reducing workload and retiring dogs currently rest on handler judgment (Ng & Fine, 2019).
In practice: Many programs keep a simple log per dog with date, setting, duration, transport time, number of clients and notable behavior. Gradual changes, such as more frequent avoidance or slower recovery, then show up earlier than from memory.
10. Summary at a Glance
Most physiological studies found no statistically significant increase in the stress markers examined during sessions, and some found lower HR, higher HRV or reduced cortisol by the end (Ng et al., 2014; Clark et al., 2019; Clark, Martin, et al., 2020; Shih et al., 2025); early Austrian data showed higher cortisol on therapy days, with unclear valence (Haubenhofer & Kirchengast, 2006, 2007).
Behavioral studies found more play and affiliation during sessions in one sample (Corsetti et al., 2019), but frequent appeasement and avoidance at outdoor events (Erichsmeier et al., 2026), and few positive behaviors in the handler-scored Dutch dataset (Visser et al., 2026). Factors associated with stress include unfamiliar settings, transport, recipients acting on the dog, younger clients, lack of choice and handler stress (Ng et al., 2014; Glenk & Foltin, 2021; Haven-Pross et al., 2026; Sarrafchi et al., 2025; Silas et al., 2019).
11. Limitations and Research Gaps
11.1 Short Time Frames
Almost all data cover single sessions or a few weeks; the longest observation periods are three months (Haubenhofer & Kirchengast, 2006, 2007), 90 days (Giuliano et al., 2025) and four months (McCullough et al., 2018). There are no longitudinal data across a dog's working career and no evidence-based criteria for retirement (Ng & Fine, 2019).
11.2 No Evidence-Based Session Limits
Recommended limits are expert-based, organizations differ widely, and study results on frequency and duration conflict (IAHAIO, 2018; Serpell et al., 2020; Haubenhofer & Kirchengast, 2006; Clark et al., 2019).
11.3 Valence Remains Unresolved
Cortisol and HR cannot separate distress from positive arousal (Handlin et al., 2011; van Houtert et al., 2021). Validated positive-welfare indicators are missing. The gap identified in 2022 (Miller et al., 2022) was partly addressed in 2026 (Haven-Pross et al., 2026), but validation of positive affect remains limited, and expert thresholds are not independently validated (Visser et al., 2026). HRV during sessions is still at the feasibility stage (Suttkus et al., 2026).
11.4 Design Weaknesses
Matched non-working control days with the same transport and setting are rare, randomization concerns patients, not dogs, and most samples are below ten dogs (Glenk & Foltin, 2021; Shih et al., 2025).
11.5 Recipients, Choice and Handlers
Group size and recipient behavior have been described, not varied experimentally. Choice has been tested in one experiment with three-minute exposures (Sarrafchi et al., 2025), and the lead comparison rests on one non-randomized study of established handling groups (Glenk et al., 2013). The link between handler and dog stress is correlational (Silas et al., 2019; Koda et al., 2015; Sundman et al., 2019).
11.6 School and Reading Dogs, Reporting
Dog-side data for reading programs come from two dogs (Uccheddu et al., 2019) and one in-session activity (Melco et al., 2020); the main review of reading programs does not address dog welfare (Hall et al., 2016). Reporting of animal welfare outcomes and ethical approval in AAI research remains poor (Ng et al., 2019).
12. Conclusion
The available studies do not show that well-run animal-assisted interventions are generally stressful for the dogs involved. Several suggest that dogs can be calm or even more relaxed at the end of a session (Ng et al., 2014; Clark, Martin, et al., 2020; Shih et al., 2025). But the evidence comes from few, mostly experienced and selected dogs, short observation periods and measures that cannot tell positive from negative arousal. Behavioral data show that avoidance and appeasement occur and that positive behaviors may be less frequent than hoped, though the latter comes from one handler-scored dataset (Erichsmeier et al., 2026; Visser et al., 2026).
For program design, the most defensible conclusions concern conditions rather than numbers: give the dog choice and a way to withdraw, reduce novelty and transport load, control how recipients interact with the dog, monitor behavior systematically, and make the handler responsible for the dog first. Whether a dog enjoys the work has to be shown again for every dog, not assumed for the role.
Key Insights (Takeaways)
Most physiological studies found no statistically significant rise in examined stress markers in therapy dogs during sessions, but samples were mostly below ten dogs (Glenk & Foltin, 2021).
In a within-dog comparison, an unfamiliar room (handler present, no strangers) was associated with higher cortisol than a 60-minute visiting activity; novelty was not isolated as the cause (Ng et al., 2014).
Lower heart rate and higher HRV after sessions in hospital programs suggest relaxation, but these markers cannot prove enjoyment (Clark, Martin, et al., 2020; Shih et al., 2025).
Free choice of contact changed dogs' behavior compared with forced touch on leash (Sarrafchi et al., 2025).
Handler stress is associated with dog stress, and handler ratings can underestimate the dog's state (Silas et al., 2019; Koda et al., 2015; Erichsmeier et al., 2026).
In the largest dataset, 63 dogs scored largely by their own handlers and analyzed in two papers that are not independent replications, overt stress was largely managed, but positive behaviors were infrequent (Haven-Pross et al., 2026; Visser et al., 2026).
Recommended session limits are guideline or expert values; reliable evidence on safe time limits is sparse (IAHAIO, 2018; Serpell et al., 2020).
References
Burrows, K. E., Adams, C. L., & Millman, S. T. (2008). Factors affecting behavior and welfare of service dogs for children with autism spectrum disorder. Journal of Applied Animal Welfare Science, 11(1), 42–62. https://doi.org/10.1080/10888700701555550
Clark, S. D., Martin, F., McGowan, R. T. S., Smidt, J. M., Anderson, R., Wang, L., Turpin, T., Langenfeld-McCoy, N., Bauer, B. A., & Mohabbat, A. B. (2020). Physiological state of therapy dogs during animal-assisted activities in an outpatient setting. Animals, 10(5), 819. https://doi.org/10.3390/ani10050819
Clark, S. D., Smidt, J. M., & Bauer, B. A. (2019). Welfare considerations: Salivary cortisol concentrations on frequency of therapy dog visits in an outpatient hospital setting: A pilot study. Journal of Veterinary Behavior, 30, 88–91. https://doi.org/10.1016/j.jveb.2018.12.002
Clark, S. D., Smidt, J. M., & Bauer, B. A. (2020). Therapy dogs' and handlers' behavior and salivary cortisol during initial visits in a complex medical institution: A pilot study. Frontiers in Veterinary Science, 7, 564201. https://doi.org/10.3389/fvets.2020.564201
Cobb, M. L., Iskandarani, K., Chinchilli, V. M., & Dreschel, N. A. (2016). A systematic review and meta-analysis of salivary cortisol measurement in domestic canines. Domestic Animal Endocrinology, 57, 31–42. https://doi.org/10.1016/j.domaniend.2016.04.003
Corsetti, S., Ferrara, M., & Natoli, E. (2019). Evaluating stress in dogs involved in animal-assisted interventions. Animals, 9(10), 833. https://doi.org/10.3390/ani9100833
d'Angelo, D., d'Ingeo, S., Ciani, F., Visone, M., Sacchettino, L., Avallone, L., & Quaranta, A. (2021). Cortisol levels of shelter dogs in animal assisted interventions in a prison: An exploratory study. Animals, 11(2), 345. https://doi.org/10.3390/ani11020345
Dowling-Guyer, S., Dabney, K., Robertson, E. A. R., & Mueller, M. K. (2024). Behavior coding of adolescent and therapy dog interactions during a social stress task. Veterinary Sciences, 11(12), 644. https://doi.org/10.3390/vetsci11120644
Dreschel, N. A., & Granger, D. A. (2009). Methods of collection for salivary cortisol measurement in dogs. Hormones and Behavior, 55(1), 163–168. https://doi.org/10.1016/j.yhbeh.2008.09.010
Erichsmeier, R., Arney, D., & Soonberg, M. (2026). Behavioral observations of dogs during animal-assisted interventions and their handlers' perceptions of their experienced level of stress. Journal of Applied Animal Welfare Science, 29(3), 612–623. https://doi.org/10.1080/10888705.2025.2555266
Fine, A. H., Beck, A. M., & Ng, Z. (2019). The state of animal-assisted interventions: Addressing the contemporary issues that will shape the future. International Journal of Environmental Research and Public Health, 16(20), 3997. https://doi.org/10.3390/ijerph16203997
Giuliano, V. O., Sacchettino, L., Rusu, A. S., Ciccarelli, D., Gazzano, V., de Cesare, M., Visone, M., Mizzoni, V., Napolitano, F., & d'Angelo, D. (2025). Well-being indicators in autistic children and therapy dogs during a group intervention: A pilot study. Animals, 15(14), 2032. https://doi.org/10.3390/ani15142032
Glenk, L. M. (2017). Current perspectives on therapy dog welfare in animal-assisted interventions. Animals, 7(2), 7. https://doi.org/10.3390/ani7020007
Glenk, L. M., & Foltin, S. (2021). Therapy dog welfare revisited: A review of the literature. Veterinary Sciences, 8(10), 226. https://doi.org/10.3390/vetsci8100226
Glenk, L. M., Kothgassner, O. D., Stetina, B. U., Palme, R., Kepplinger, B., & Baran, H. (2013). Therapy dogs' salivary cortisol levels vary during animal-assisted interventions. Animal Welfare, 22(3), 369–378. https://doi.org/10.7120/09627286.22.3.369
Hall, S. S., Gee, N. R., & Mills, D. S. (2016). Children reading to dogs: A systematic review of the literature. PLOS ONE, 11(2), e0149759. https://doi.org/10.1371/journal.pone.0149759
Handlin, L., Hydbring-Sandberg, E., Nilsson, A., Ejdebäck, M., Jansson, A., & Uvnäs-Moberg, K. (2011). Short-term interaction between dogs and their owners: Effects on oxytocin, cortisol, insulin and heart rate – An exploratory study. Anthrozoös, 24(3), 301–315. https://doi.org/10.2752/175303711X13045914865385
Haubenhofer, D. K., & Kirchengast, S. (2006). Physiological arousal for companion dogs working with their owners in animal-assisted activities and animal-assisted therapy. Journal of Applied Animal Welfare Science, 9(2), 165–172. https://doi.org/10.1207/s15327604jaws0902_5
Haubenhofer, D. K., & Kirchengast, S. (2007). Dog handlers' and dogs' emotional and cortisol secretion responses associated with animal-assisted therapy sessions. Society & Animals, 15(2), 127–150. https://doi.org/10.1163/156853007X187090
Haven-Pross, S. C., Jens, A. L., Maarleveld, K. N., van Honk, P., de Kort, M., & Visser, E. K. (2026). Dogs' behavioural responses to dog-assisted interventions: A field study. Animals, 16(7), 1063. https://doi.org/10.3390/ani16071063
Hediger, K., Meisser, A., & Zinsstag, J. (2019). A One Health research framework for animal-assisted interventions. International Journal of Environmental Research and Public Health, 16(4), 640. https://doi.org/10.3390/ijerph16040640
Hill, J., Driscoll, C., Cawdell-Smith, J., Anderson, S., & Ziviani, J. (2023). Investigating dog welfare when interacting with autistic children within canine-assisted occupational therapy sessions: A single case study. Animals, 13(12), 1965. https://doi.org/10.3390/ani13121965
International Association of Human-Animal Interaction Organizations (IAHAIO). (2018). IAHAIO white paper 2014, updated for 2018: The IAHAIO definitions for animal assisted intervention and guidelines for wellness of animals involved in AAI [Organizational document, not peer-reviewed]. https://iahaio.org/wp/wp-content/uploads/2018/04/iahaio_wp_updated-2018-final.pdf
Koda, N., Watanabe, G., Miyaji, Y., Ishida, A., & Miyaji, C. (2015). Stress levels in dogs, and its recognition by their handlers, during animal-assisted therapy in a prison. Animal Welfare, 24(2), 203–209. https://doi.org/10.7120/09627286.24.2.203
Kolevská, J., Brunclík, V., & Svoboda, M. (2003). Circadian rhythm of cortisol secretion in dogs of different daily activities. Acta Veterinaria Brno, 72(4), 599–605. https://doi.org/10.2754/avb200372040599
McCullough, A., Jenkins, M. A., Ruehrdanz, A., Gilmer, M. J., Olson, J., Pawar, A., Holley, L., Sierra-Rivera, S., Linder, D. E., Pichette, D., Grossman, N. J., Hellman, C., Guérin, N. A., & O'Haire, M. E. (2018). Physiological and behavioral effects of animal-assisted interventions on therapy dogs in pediatric oncology settings. Applied Animal Behaviour Science, 200, 86–95. https://doi.org/10.1016/j.applanim.2017.11.014
Melco, A. L., Goldman, L., Fine, A. H., & Peralta, J. M. (2020). Investigation of physiological and behavioral responses in dogs participating in animal-assisted therapy with children diagnosed with attention-deficit hyperactivity disorder. Journal of Applied Animal Welfare Science, 23(1), 10–28. https://doi.org/10.1080/10888705.2018.1536979
Mignot, A., de Luca, K., Servais, V., & Leboucher, G. (2022). Handlers' representations on therapy dogs' welfare. Animals, 12(5), 580. https://doi.org/10.3390/ani12050580
Miller, S. L., Serpell, J. A., Dalton, K. R., Waite, K. B., Morris, D. O., Redding, L. E., Dreschel, N. A., & Davis, M. F. (2022). The importance of evaluating positive welfare characteristics and temperament in working therapy dogs. Frontiers in Veterinary Science, 9, 844252. https://doi.org/10.3389/fvets.2022.844252
Ng, Z., Morse, L., Albright, J., Viera, A., & Souza, M. (2019). Describing the use of animals in animal-assisted intervention research. Journal of Applied Animal Welfare Science, 22(4), 364–376. https://doi.org/10.1080/10888705.2018.1524765
Ng, Z. Y., & Fine, A. H. (2019). Considerations for the retirement of therapy animals. Animals, 9(12), 1100. https://doi.org/10.3390/ani9121100
Ng, Z. Y., Pierce, B. J., Otto, C. M., Buechner-Maxwell, V. A., Siracusa, C., & Werre, S. R. (2014). The effect of dog–human interaction on cortisol and behavior in registered animal-assisted activity dogs. Applied Animal Behaviour Science, 159, 69–81. https://doi.org/10.1016/j.applanim.2014.07.009
Odendaal, J. S. J., & Meintjes, R. A. (2003). Neurophysiological correlates of affiliative behaviour between humans and dogs. The Veterinary Journal, 165(3), 296–301. https://doi.org/10.1016/S1090-0233(02)00237-X
Palestrini, C., Calcaterra, V., Cannas, S., Talamonti, Z., Papotti, F., Buttram, D., & Pelizzo, G. (2017). Stress level evaluation in a dog during animal-assisted therapy in pediatric surgery. Journal of Veterinary Behavior, 17, 44–49. https://doi.org/10.1016/j.jveb.2016.09.003
Pirrone, F., Ripamonti, A., Garoni, E. C., Stradiotti, S., & Albertini, M. (2017). Measuring social synchrony and stress in the handler-dog dyad during animal-assisted activities: A pilot study. Journal of Veterinary Behavior, 21, 45–52. https://doi.org/10.1016/j.jveb.2017.07.004
Sarrafchi, A., de Zwaan, N., Tucker, M., & Merkies, K. (2025). Can I touch you? A pilot study comparing consensual and non-consensual human-dog touch interactions. Applied Animal Behaviour Science, 285, 106560. https://doi.org/10.1016/j.applanim.2025.106560
Serpell, J. A., Kruger, K. A., Freeman, L. M., Griffin, J. A., & Ng, Z. Y. (2020). Current standards and practices within the therapy dog industry: Results of a representative survey of United States therapy dog organizations. Frontiers in Veterinary Science, 7, 35. https://doi.org/10.3389/fvets.2020.00035
Shajid Pyari, M., Guo, K., & Kargas, N. (2026). Preparing therapy dogs for animal-assisted services: Governance, welfare-centred preparation, and a lifecycle research agenda across selected European and United States frameworks. Animals, 16(16), 2557. https://doi.org/10.3390/ani16162557
Shih, H.-Y., Martin, F., Ness, D., Romine, W., Peck, T. L., Turpin, T., Horoschak, R., Steeby, C., Phillips, H., Claypool, M., Theuer, A., Herbeck, G. M., Sexton, J., Pittman, E., Bellamkonda, E., Mohabbat, N. M. L., Lyn, S. A., Bauer, B. A., & Mohabbat, A. B. (2025). Behavioral and physiological responses of therapy dogs to animal-assisted treatment in an inpatient stroke rehabilitation program. Animals, 15(2), 121. https://doi.org/10.3390/ani15020121
Silas, H. J., Binfet, J.-T., & Ford, A. T. (2019). Therapeutic for all? Observational assessments of therapy canine stress in an on-campus stress-reduction program. Journal of Veterinary Behavior, 32, 6–13. https://doi.org/10.1016/j.jveb.2019.03.009
Sundman, A.-S., Van Poucke, E., Svensson Holm, A.-C., Faresjö, Å., Theodorsson, E., Jensen, P., & Roth, L. S. V. (2019). Long-term stress levels are synchronized in dogs and their owners. Scientific Reports, 9, 7391. https://doi.org/10.1038/s41598-019-43851-x
Suttkus, S., Schumann, A., de la Cruz, F., Geisler, M., Gupta, Y., Rieger, K., Ganser, G., & Bär, K.-J. (2026). Feasibility of extracting reliable canine heart rate variability from ambulatory electrocardiogram recordings during animal-assisted psychotherapy [Preprint, not peer-reviewed]. Preprints.org. https://doi.org/10.20944/preprints202607.2168.v1
Townsend, L., & Gee, N. R. (2021). Recognizing and mitigating canine stress during animal assisted interventions. Veterinary Sciences, 8(11), 254. https://doi.org/10.3390/vetsci8110254
Uccheddu, S., Albertini, M., Pierantoni, L., Fantino, S., & Pirrone, F. (2019). Assessing behaviour and stress in two dogs during sessions of a reading-to-a-dog program for children with pervasive developmental disorders. Dog Behavior, 4(3) [issue dated 2018]. https://doi.org/10.4454/db.v4i3.83
van Houtert, E. A. E., Endenburg, N., Rodenburg, T. B., & Vermetten, E. (2021). Do service dogs for veterans with PTSD mount a cortisol response in response to training? Animals, 11(3), 650. https://doi.org/10.3390/ani11030650
Visser, E. K., Jens, A. L., van Honk, P., van Asselt, M., & Haven-Pross, S. C. (2026). Establishing behavioural thresholds for dogs in animal-assisted services: Expert-derived thresholds and field study comparison. Animals, 16(7), 1078. https://doi.org/10.3390/ani16071078
Winkle, M., Johnson, A., & Mills, D. (2020). Dog welfare, well-being and behavior: Considerations for selection, evaluation and suitability for animal-assisted therapy. Animals, 10(11), 2188. https://doi.org/10.3390/ani10112188
Zenker, F., Thodberg, K., & Sandøe, P. (2026). Partners or tools? Ethical framings shape welfare research on dogs and horses in animal-assisted services. Frontiers in Veterinary Science, 13, 1960144. https://doi.org/10.3389/fvets.2026.1960144