Assistance dogs (including those for PTSD) undergo several major life transitions that disrupt their physical and social environments (54, 55, 62, 63) (Figure 1). Selectively bred assistance dogs or those sourced as puppies are typically placed with foster families at 8–10 weeks for 12–18 months (51, 54–56, 59, 62, 63, 87). With different degrees of agency oversight, the foster family is responsible for socialising the puppy, exposing them to a wide range of public places/situations, and teaching them basic obedience (55, 62, 78, 87). Upon reaching adolescence, dogs return to agency kennels, where they are housed in stalls (often individually) over the period of advanced training, culminating in pair-training (36, 54–56, 59, 62, 63). The length of this kennel-life period varies from 4–8 months (54, 56, 62). Dogs sourced outside the breeding programs as adolescents/adults (e.g., rescued/donated dogs) likely spend their full training time at the kennel.

Some evidence suggests 17 months is the optimal age for transitioning back to kennels (55), allowing time for certain genetic or behavioural issues to manifest (54) and agencies to decide on further training investment. Although using puppy-raisers and the subsequent advanced training at kennels is common among ADI-accredited PTSD assistance dog schools, specific practices vary. For instance, WCC’s dogs live with puppy-raisers until 21 months, followed by 3–4 months of kennel-based advanced training and a week of personalised training with the recipient (WCC, personal communication, April 25–26, 2024).

Dogs are usually placed with recipients at 18–24 months of age (54, 55, 62, 78), unless enrolled in the program later. Being matched does not necessarily mean the dog has found a long-term home. Assistance dogs have been relinquished due to temperament mismatch, behavioural issues, or handler-related circumstances (34, 35, 62). Unfortunately, there is also evidence for an instrumental view of assistance dogs—among recipients who treat them merely as tools for health-related goals. Williamson et al. (35) reported a case where a Veteran, despite bonding with their PTSD assistance dog, relinquished the dog after a year, feeling the dog’s support was no longer needed. If legal ownership remains with the agency, such dogs may be retrained and/or reassigned (35, 62). Dogs deemed unsuitable for assistance work are retired prematurely (62).

Retired assistance dogs may be adopted by their handler or handler’s family/friends, return to the puppy-raiser, or go back to the agency to be released for adoption (34, 51, 62, 63). Dogs who drop out of training follow similar paths (88, 89). However, when a dog was privately trained, or agencies transfer legal ownership completely to handlers, and no family/friends can adopt the dog, the retiring assistance dog might be relinquished to an animal shelter (34). While their training and temperament typically make assistance dogs desirable for adoption (51), euthanasia is possible for those not placed in a no-kill shelter.

Such a succession of families and uprooting from familiar environments can negatively affect a dog’s welfare (34, 54) and potentially foster behaviour problems that hinder future performance (63). These could include separation anxiety, excessive barking/vocalising, hypervigilance, destructive behaviour, inappropriate elimination, and difficulty focusing during training/working (90). Such disruptions can be especially stressful for assistance dogs, who are typically selected for their tendency to form strong attachment bonds with humans—and may be particularly vulnerable to separation anxiety (91). Although the exact impact of repeatedly breaking these bonds is unclear, many experts argue it can be deeply distressing (32, 54, 63).

Mengoli et al. (56) found that assistance dogs had significantly lower free and total oxytocin levels than companion dogs, potentially due to repeated separations from foster families during kennel-training period (characteristic to their setting). It was suggested that such social disruptions alter dogs’ bonding-related neuromodulatory system, negatively affecting future attachment processes (56). Conversely, Hall et al. (55) encouraged foster families and trainers to bond with dogs, suggesting that secure and high-quality attachment bonds in earlier life facilitate developing bonds later and thus positively impact the dogs’ well-being and performance after placement. Attachment to foster families is also encouraged in the European standards (78).

Varying personalities, expectations, and training/interacting styles among foster families and handlers further challenge the dogs, especially when placed with inexperienced people (34, 35, 54, 63, 92). Poor training and handling skills can confuse dogs and reduce their performance, affecting the handler’s health benefits and satisfaction (63), and potentially lowering their commitment to the dog’s well-being (80). Although pair-training includes developing recipients’ dog-training/handling skills, the varying lengths of pair-training across agencies (from a week to 13 months as reviewed above) raises serious welfare concerns for PTSD assistance dogs.

Major life changes, like moving to a new home, can be stressful for a dog even when remaining in the same family (93). Drastic changes to the dog’s physical and social environment—like moving from family home to kennel—can be even more challenging, particularly for dogs who did not experience kennel life during early puppyhood (54, 94). Life in kennels varies across organisations. While some agencies incorporate training sessions, outings, playgroup-time, and enrichment activities into the dogs’ schedule (WCC, personal communication, April 25, 2024), others may take the dog out of the stall for only 20–60 min each day (54) and provide little to no enrichment (63).

Kennel stress is not an uncommon reason for assistance dog candidates to drop out of programs, just like it is not uncommon to develop behavioural issues during kennel stays (54). Although ADI standards emphasise the importance of enrichment during kennel housing (77), kennels remain impoverished environments—even at top-tier agencies (54, 94, 95). Loud, crowded, unfamiliar, and understimulating kennel environments can elicit stress, (separation) anxiety, stereotypic behaviours (e.g., pacing, excessive barking), as well as hyperactivity, fearfulness, and dog-directed aggression (63, 94–97). It goes without saying that similar welfare concerns apply to relinquished/retired assistance dogs in animal shelters.

Rescuing a dog from a shelter offers obvious welfare benefits (admittedly, depending on the contrast between the shelter and the new home). However, such dogs have been subjected to kennel conditions and may have developed above-described issues that could compromise their future performance and welfare as assistance dogs. Shelter-sourced dogs are carefully screened, but the lack of validated screening procedures and the latent nature of many behaviour problems (63) remain major challenges. Their unknown background—particularly during early puppyhood—is another concern. Hence, critics argue that organisations using shelter dogs risk providing animals unsuitable for assistance work (35, 36, 43, 63).

Early experiences during the sensitive period of socialisation—typically between 4 and 16 weeks in dogs—are critical for behavioural development (54, 55, 63, 90, 98–101). During this time, puppies are prone to bond easily and accept novel social and environmental stimuli (98), building lasting associations that are difficult to unlearn later in life (63, 98, 101). Prospective assistance dog puppies must be introduced to a wide range of potentially fear-provoking stimuli related to their future occupation (including various people, animals, locations, sounds) in a safe and positive way to shape their perception of what is “normal” in their environment and prevent future undesirable responses (54, 55, 63, 64, 78, 98, 101).

Shelter dogs’ unknown background entails the possibility of negative or insufficient exposure during the socialisation period or growing up in an environment not conducive to learning to cope with novelty. This may lead to fearful or aggressive responses to novel or negatively associated situations/stimuli (63, 98, 101)—common reasons for dogs failing assistance dog training. In contrast, breeding programs can provide controlled environments and structured socialisation to ensure the dogs learn to handle relevant stimuli and novelty with calm and confidence (55, 60, 63, 77). Moreover, some suggest that positive associations formed during proper socialisation may boost the dog’s motivation to work (55) and make the work more enjoyable, as the dog could develop cognitive abilities to better understand the tasks’ purpose and rely less on operant conditioning (63).

WCC’s Puppy Enrichment Center supports early socialisation (87). Additionally, they involve Veterans with PTSD in socialising and training future assistance dogs for fellow Veterans—both to prepare the puppies and provide occupational therapy for the Veterans (25, 45, 49). Nevertheless, not all assistance dog breeding kennels offer adequate socialisation opportunities (54, 63), setting dogs up to struggle with future work-related demands. After all, the “flunk-out” rate is high among agency-bred dogs alike and impoverished early environments are a contributing factor (54, 63). In such cases, transferring puppies to foster homes earlier can help. However, if agencies provide sufficient enrichment and outside exposure, keeping puppies with their mother until 12 weeks supports social learning (by observing her performance), with a to-be-determined benefit for the pups’ training outcomes (55).

Further factors that contribute to the high drop-out among agency-bred dogs and lifelong welfare concerns are genetic health issues (e.g., progressive retinal atrophy, osteoarthritis), inbreeding, susceptibility to infectious diseases, and other problems associated with a narrow gene pool (36, 54, 63). Along with uncertainties around trait heritability and breed preferences, this raises doubts about whether breeding programmes of solely pure-bred lineages are actually desirable for this type of work (36, 63).

Thus, while some evidence links selectively bred assistance dogs to better training/working success (36, 68, 70), these benefits can be undermined by organisational practices and conditions. More research is needed to lower failure rates and improve dog welfare—such as comparing welfare and return rates of shelter-sourced vs. purpose-bred PTSD assistance dogs and examining recipient health outcomes. Recommendations have included exploring out-crossing and crossbreeding to boost the health of selectively bred stocks, while also highlighting the need for better behavioural screening methods, particularly for identifying suitable shelter dogs (63, 102). For instance, Mengoli et al. (56) suggest that neurochemical profiling (e.g., prolactin, oxytocin, serotonin) could enhance assistance dog selection, breeding, and welfare monitoring.

Despite ADI’s recommendation to use humane, evidence-based, minimally aversive training methods (77), as discussed earlier, aversive and dominance-based methods are relatively widespread among PTSD assistance dog schools—even some ADI-accredited ones. The use of dominance-based methods contradicts ADI standards, as dominance theory within the context of human-dog relationships is outdated and lacks scientific support [see (103–106) for explanations for dismissing the dominance-based approach]. Insurmountable evidence shows that dominance-based human-dog interactions (e.g., alpha-rolls, stare-downs) induce fear in dogs, increase handler-directed aggression, and damage dog-handler relationships by undermining trust, emotional bonds, communication, and cooperation (90, 99, 100, 104, 107).

The use of aversive stimuli within positive punishment and negative reinforcement can be considered evidence-based when understood as integral principles of operant conditioning. However, it is important to consider all the evidence, including the growing research highlighting their harm to dogs’ physical and mental health (36, 54, 55, 63, 90, 100, 107, 108). Aversive methods can also elicit unwanted behavioural side-effects, hinder performance and motivation, and compromise dog-handler relationships (36, 55, 90, 99, 100, 107–109).

Ziv’s (107) review of scientific studies on dog training methods shows positive punishment and negative reinforcement often increase dogs’ fear (including fear of novelty and punishment-related stimuli), aggression (toward people/dogs), stress (e.g., stress-indicative body language, restlessness, cortisol levels), distractibility, escape/avoidance behaviours, hesitance to explore, apathy/passiveness, and hypervigilance. These outcomes—especially heightened fearfulness, aggression, stress, and anxiety—are consistently documented by scholars and practitioners (55, 63, 90, 98–100, 104, 110).

Aversive techniques linked to negative behavioural and psychological effects include verbal (e.g., yelling “no,” scolding), physical (e.g., kicking, hitting, shaking, scruff-grabbing, alpha-roll), and collar corrections (e.g., choke, tone, shock, prong, spray collars, or yanking by a flat collar) (107). Many also pose physical health risks. These involve injuries from kicking or forced restraint (e.g., alpha-roll), and damage from punitive collars—including increased intraocular pressure, spinal cervical injury, nerve damage, esophageal/tracheal injuries, and impaired respiration (100, 107). Stress-induced concerns, like gastrointestinal problems and weakened immunity add further risk (107).

Supporters of aversive training argue that, when used professionally, such tools do not harm dogs. However, Ziv (107) found negative welfare impacts even when applied by experienced trainers. Risks increase—potentially reaching abuse—when used by unskilled handlers lacking timing and consistency (107). With no trainer qualification standards, recipients’ general inexperience, and brief pair-training, aversive techniques pose significant welfare concerns for assistance dogs. In contrast, positive reinforcement-based training carries far lower risks of behaviour problems, aggression, and fear—even when applied inexpertly (100, 107).

Considering that fearfulness and aggression are highly undesirable in PTSD assistance dogs and top reasons for training failure, it is regrettable that some agencies’ methods contribute to these problems. Even when dogs complete such programs, the harm and suffering caused remain. Moreover, research shows punitive techniques often reduce dogs’ willingness to approach strangers and interact/play with handlers (107). The latter reflects damage to the dog-handler relationship due to dogs associating handlers with aversive stimuli (99, 100, 104, 107).

There is a major research gap on how different training methods affect assistance dogs specifically (36). However, LaFollette et al. (79) found similar outcomes: PTSD assistance dogs of Veterans prone to use positive punishment were more fearful, less active, and less playful than dogs of Veterans using positive reinforcement-based techniques. Moreover, frequent positive punishment use was linked to weaker perceived Veteran-dog bonds and less eye contact, while positive reinforcement and bond-based interactions were associated with stronger bonds (79).

As discussed, sociability with strangers is highly desirable in PTSD assistance dogs and the recipient-dog bond is crucial for recipient health benefits (29). Thus, aversive methods should be impermissible when training/handling PTSD assistance dogs—both because they harm dog welfare and undermine training goals. Moreover, positive reinforcement-based methods promote handler patience, attention, emotional regulation, and expression of positive emotions (e.g., when praising the dog), mitigating PTSD symptoms like emotional numbing, avoidance, and hypervigilance (49). Hence, they benefit both the dog and the person with PTSD.

Certain situations may require assistance dogs to take initiative or disobey commands (“intelligent disobedience”) (54, 55, 64). Encouraging this is incompatible with associating disobedience with aversive stimuli (55). Exposure to aversive stimuli perceived by the animal as inescapable and/or unpredictable (e.g., due to inconsistent or poorly timed punishment) can cause the animal’s learned helplessness (100, 104, 110), manifesting as passivity, and impaired decision-making and initiative-taking (7, 111, 112).

Some tasks expected of assistance dogs (including some PTSD assistance dogs)—e.g., wheelchair-pulling—lack discrete beginning/ending, might be physically distressing, are not intrinsically motivating, and the dog may not fully understand their purpose (54, 63). In such cases, it is sometimes argued that positive reinforcement-based techniques are impractical for meeting training goals and maintaining performance levels after placement (63, 92). Although none of the studies reviewed by Ziv (107) analysed wheelchair-pulling per se, no evidence indicated that aversive methods are more effective. In fact, reward-based training is linked to increased motivation, improved performance (e.g., in obedience, protection, scent-detection) (107, 108), and higher trainability (79). Incorporating punishment in training complex tasks can make them even more aversive, further reducing the dog’s willingness to work, and triggering avoidance and other unwanted side-effects. Tasks requiring the most stamina and motivation are precisely those expert trainers avoid associating with punishment (63).

Hopefully, the prevalence of aversive methods will decline substantially, at least in Europe, now that the new EU-wide standards explicitly prohibit their use (64, 78). Non-aversive alternatives should be explored where positive reinforcement falls short. Hall et al. (55) suggest investigating Pavlovian-conditioned reinforcers (e.g., clickers) for sustaining performance when primary reinforcers (e.g., treats) are impractical. If applied appropriately, gradually introducing intermittent reinforcement schedules can also increase resistance to extinction and reduce frustration from inconsistent reinforcement in real-world conditions (55, 113–115). However, research focused on such non-aversive solutions is still lacking.

Bremhorst and colleagues (36) noted that agencies still prioritise assistance dogs’ physical health, while psychological/emotional welfare often receives scant attention. Several researchers as well as the European standards urge a broader assistance dog welfare approach aligned with the Five Domains model. One general concern is that—due to limited knowledge or financial restraints (34, 116)—recipients may not always provide adequate canine care despite education during pair-training (35, 42, 47). Further concerns stem from the nature of assistance dog work: insufficient rest, constant work, unpredictable routines, limited play/recreation opportunities, unintentional maltreatment, zoonotic disease risk, and even obesity (34, 36, 51, 62, 63, 117). Another issue is assistance dogs’ inability to control their physical and social environment or escape stressful, potentially dangerous situations (63, 117). Examples include busy streets and crowded public spaces where avoiding unwelcome interactions is difficult, and home settings where contact with non-recipient family members may be restricted (62).

While PTSD assistance dog recipients have reported frustration from the dog’s around-the-clock company (35), it is unclear whether dogs feel similarly. A qualitative study of 11 assistance dogs for children with autism reported several dogs resisting sleeping in the child’s bedroom—whimpering and pawing at the door—suggesting a stronger bond with a parent (the primary handler) (62). There is little evidence on whether dogs whose strongest bond is with the recipient also occasionally prefer solitude or alternative company—and how often they can act on these preferences. Encouraging positive relationships with other family members might be important, as their difficulties accepting the dog may raise relinquishment risk. For instance, partners of people with PTSD have reported struggling to adjust their caregiving role and mixed emotions about the dog (35, 47, 118, 119).

Meeting assistance dogs’ age-dependent needs and managing retirement raises further welfare concerns (36, 51, 63, 102, 117)—particularly because best practices remain unclear and much depends on whether the agency offers lifelong support beyond the assistance role. Retirement adjustment is stressful for dogs due to sudden routine changes (36, 51, 63). Discouraging or no longer rewarding assistive tasks may also elicit confusion, frustration, anxiety, and behaviour problems (51), although intermittent reinforcement during training may buffer this effect.

Retirement plans where dogs remain with the handler pose challenges including dogs’ inability to accompany the handler everywhere and adjusting to a replacement dog—potentially eliciting separation anxiety or inter-dog aggression, respectively (51). However, PTSD assistance dogs have an advantage over dogs with permanently impaired handlers—PTSD symptoms might improve enough to make a replacement dog unnecessary. Moreover, gradually reducing dog-accompanied outings and phasing out maintenance training may benefit both parties. It has been suggested that PTSD assistance dogs could hinder handlers’ habituation to trauma reminders and independent coping (11, 24, 39). Progressively fostering independence may thus facilitate the dog’s retirement transition while supporting the handler’s self-mastery during recovery. This aligns with the European standards, which recommend a progressive, phased approach to assistance dog retirement to support both the dog and handler in adjusting to the transition (78). Returning retired dogs to puppy-raisers or third-party adopters may alleviate confusion from described handler-related issues (51), but does not offset the distress of separation from their bonded person. Adapting to a new home, people, and possibly animals further amplify the stress (51).

The chronic stressors outlined—or their analogues—apply across assistance dog types (9, 25, 35, 43, 117). Specific concerns for PTSD assistance dogs include uncertainty about coping with all PTSD symptoms (9, 25) and ethical considerations when using animals in trauma response. For example, PTSD is robustly associated with dysregulated anger and elevated aggression/violence in civilian and military populations alike (1, 35, 120–122). The anger may manifest as irritability, hostility, or self-directed aggression (including recklessness, suicidality), increasing risk of interpersonal conflicts and violence (122). A review by MacManus et al. (121) found combat-related PTSD significantly associated with criminal and domestic violence among military personnel and Veterans. Notably, irritability and anger often persist post-treatment (122). Although some speculate PTSD assistance dogs may respond to violent outbursts with fear or defensive aggression (9), no studies have actually investigated how dogs cope with these symptoms or whether they may be victims of violence. The study on autism assistance dogs found dogs were often targets of child aggression and learned to distract/comfort or avoid the child depending on the distinct state of anger (62). Predicting the outcomes of their handler’s emotional states is likely a daily challenge for PTSD assistance dogs trained to calm an agitated person (35).

This highlights how assistance dogs face ambivalent situations, having to choose between trained and self-preserving behaviours. Such situations require great emotional control, as dogs are expected to remain calm and never retaliate or exhibit aggression, even under high stress (56, 62, 123, 124). For instance, similar concerns about dogs having to suppress their emotional response in challenging situations have also been raised in guide dog research (124). Research is urgently needed to assess how effectively PTSD assistance dogs maintain emotional balance while upholding training standards—and whether those bred for attentiveness to human emotions and low arousal outperform shelter-sourced dogs. Burrows et al. (62) emphasised teaching dogs that leaving the handler in potentially abusive situations is acceptable—a principle also relevant for PTSD assistance dogs. Still, as with intimate partner violence, dogs may not be in a position to escape.

Handlers’ reckless or angry behaviour not directed at the dog can also endanger them. For instance, Veterans in Williamson et al. (35) reported their PTSD assistance dog’s presence helped suppress road rage. While positive, this underscores risks for dogs paired with impulsive/reckless handlers.

Another welfare concern is that—through synchronisation, emotional contagion, or association with unpleasant human behaviour—assistance dogs may experience stress or anxiety themselves when detecting handler distress related to PTSD intrusion/hyperarousal symptoms. Several studies show dogs can detect human stress (125–128) and might respond with emotional contagion (129–132). PTSD assistance dogs are routinely trained to detect and interrupt handler’s distress episodes by comforting or distracting them. The mechanism behind alerting/disrupting tasks may even involve negative reinforcement, with dogs seeking to alleviate their own distress induced by the handler’s emotional state. These concerns are echoed in the newly published European standards, which emphasise the importance of PTSD assistance dogs’ ability to cope with high levels of handler emotional arousal while performing tasks without experiencing anxiety (64). However, specific strategies to support the dogs in these situations have not been proposed, and the impact on their emotional state remains unstudied.

Two studies on hair cortisol demonstrated that herding breeds bred for human cooperation synchronise long-term stress levels with their guardians (133), while ancient and solitary hunting breeds do not (134). Retrievers—also bred for human cooperation (134)—may similarly synchronise stress. Parr-Cortes et al. (132) found dogs exposed to human stress odors made more pessimistic choices in cognitive bias tests, indicating adverse emotional impact. Future studies should explore whether stress-synchronisation/emotional contagion presents greater welfare risks for PTSD assistance dogs of retriever breeds—and whether the selective breeding for stress tolerance sufficiently mitigates this (acknowledging individual differences may outweigh breed differences). Their simultaneous selection for sensitivity to human emotions could also factor into the equation.

Positive reinforcement-training may help PTSD assistance dogs develop positive associations with human distress (135). A meta-analysis of judgment-bias studies showed animals in favourable conditions—e.g., those trained with reward-based methods—exhibit more optimistic responses to ambiguous stimuli and greater positive affect (136). Positive associations are less likely in dogs trained with aversives or without specific training. However, in the absence of research, no data prove that even exclusively positive reinforcement-based training fully neutralises negative effects. Potential shortcomings might be leveraged by strengthening the dog’s association of human stress stimuli with reinforcers through Pavlovian conditioning, including counterconditioning (55).

Common PTSD comorbidities—like depression and substance use disorder—also pose welfare risks. Continuing a pattern in other areas we highlighted earlier, research on human-animal interactions and depression largely focuses on human benefits, with little attention to animal impact. However, some evidence indicates harm: Hunt et al. (137) found handler depression and PTSD symptoms predicted problem behaviours (e.g., excessive attention-seeking, separation anxiety) in search and rescue dogs, while Dodman et al. (138) found men with moderate depression more often practiced aversive training on companion dogs. Severe depression could also prevent the handler from offering affection, attention, or adequate care (including maintenance training, exercise/play, veterinary care). Moreover, depression and PTSD increase suicidal impulses (1, 122, 139), heightening risks of neglect, relinquishment, or abandonment. Substantial evidence shows PTSD assistance dogs significantly reduce recipient depression symptoms and suicidal ideation (6, 24, 25, 35, 43, 46, 47, 140). Nevertheless, no studies have examined whether handler depression/suicidality endangers dogs—highlighting a critical research gap.

Current understanding of dog-guardianship and substance use disorder mirrors that of depression. Research has focused on how dogs aid recovery, with preliminary evidence suggesting they provide structure, purpose, and responsibility (141). PTSD assistance dogs have also been associated with reduced substance use (35, 140, 142, 143). While these findings are promising, having an assistance dog does not necessarily preclude problematic substance use—underscoring the need for research into impacts on dog welfare.

One reassuring aspect is that many providers rigorously screen PTSD assistance dog recipients. Reported exclusion criteria include insufficient functionality to care for the dog (e.g., difficulty getting out of bed or leaving home) (47), elevated suicide/homicide risk (50, 144), history of violence/animal cruelty (42), severe substance use disorder, and certain psychiatric conditions like bipolar disorder or schizophrenia (50). Still, without standardised guidelines, screening practices vary, and optimal approaches for safeguarding dog welfare remain unstudied.

Further concerns stem from PTSD assistance dogs’ trained skills. Beyond diffusing distress and anger, interrupting nightmares may also entail long-term health risks. Burrows et al. (62) found autism assistance dogs frequently lost sleep during nights when following unsettled children, leading to exhaustion and poor performance the day after. Similar issues may affect PTSD dogs paired with handlers experiencing significant sleep perturbations and nightmares. Research is also needed on dogs’ physical safety during nightmares/flashbacks where handlers fighting imagined assailants may inadvertently hurt the dog who is trying to disrupt the episode.

Many other tasks—like “block,” “cover,” room checks, or retrieving items—seem low-risk, but observational studies are needed to confirm this. For instance, in busy areas, tasks like creating personal space or guiding a disoriented handler (who might be uncooperative) may elicit stress or pose safety risks for dogs.

PTSD assistance dog welfare can further vary depending on the tasks tailored to the handler’s needs. Mobility assistance tasks like bracing, door-opening, or wheelchair-pulling may cause discomfort, pain, and long-term musculoskeletal issues. Wheelchair-pulling particularly, is linked to serious physical concerns, including body misalignment and joint, back, and shoulder strain [see Bremhorst et al. (36), Coppinger et al. (54), Serpell et al. (63), and Coppinger et al. (92) for details]. Similarly, door-opening can strain dogs’ bodies and teeth [see Bremhorst et al. (36), Serpell et al. (63), and Coppinger et al. (92) for details]. These effects are exacerbated when the dog is overweight or wears poorly fitted/designed gear—both common among assistance dogs (36, 54, 92, 145). In guide dogs, musculoskeletal problems—often linked to suboptimal harnesses/handles—are a leading cause of early retirement (51, 102, 145). Thus, such problems not only harm dogs but can also shorten their working life, affecting the handler. As a promising development, the EU standards safeguard dogs providing mobility assistance, for example, by prohibiting their use as the primary source of wheelchair propulsion and restricting handlers from resting their weight on the dog (78).

The welfare concerns discussed above certainly do not apply to all PTSD assistance dogs. Welfare likely varies widely depending on the handler’s PTSD symptoms and comorbidities (and their severity), the dog’s background and training, their unique relationship, and the education/support provided by the agency. Some studies report handlers aware of possible shortcomings regarding meeting their dog’s needs (35) and prioritising play and rest (26). However, as no studies have systematically investigated PTSD assistance dog welfare, the severity and prevalence of the discussed welfare issues remain unknown.

An important advance in animal well-being is the shift toward respecting animal agency (31, 84, 146, 154). All sentient beings are agents with desires/preferences and the capacity to make choices based on those desires/preferences (146, 153). Work conditions promoting well-being should offer animals opportunities to exercise agency by having some control over their circumstances (31, 84, 146, 152, 153). This could be achieved by allowing animals to make choices, provided these choices are not self-destructive (31, 83, 84, 148, 153, 155).

A myriad of options exist to give PTSD assistance dogs opportunities to exercise choice in their daily lives. Providing options among essential resources, like treats, toys, or bedding supports well-being even within limited selection (83). During down-time, dogs should be free to choose whose company or which room to stay in, whether to be indoors or outdoors, or seek privacy. Research shows animals considerably benefit from having such choices even if not utilising them (148). Down-time should also offer ample opportunities for species-specific behaviours (83), like running free, exploring surroundings, or interacting with other animals (31, 38). Walks can support agency if dogs can choose directions and freely engage in sniffing when feasible/safe. It is encouraging that the EU standards recognise the latter need by mandating daily leisure walks and a minimum of weekly off-leash free-running sessions, during which sniffing, playing, and interacting with other dogs is encouraged, and by requiring sufficient daily mental and physical rest (64, 78).

Ensuring sufficient leisure time for PTSD assistance dogs considerably enhances their well-being by offering more control over their lives (31, 83, 148). However, respecting animal agency should also extend to their work. Some ethicists argue animals should choose their work (148, 153)⁷. Assistance dog screening partly allows this by selecting dogs’ career paths based on their individual traits, personality, and skills, excluding those reluctant or unsuited to the tasks and interactions (153, 155) (WCC, personal communication, April 25, 2024).

Cochrane (153) further argued that good work allows animals some control over their work tasks. The array of PTSD assistance dog tasks is diverse enough to enable dogs to choose which they are willing to perform. Many agencies likely consider a dog’s enthusiasm for specific tasks when assembling their repertoire. However, since tasks are tailored to the handler’s needs, the dog’s input is unclear. Ideally, the dog’s preferences should match the recipient’s needs.

Another important step in respecting PTSD assistance dogs’ agency is institutionalising a dissent model (148, 154). In human bioethics, the gold standard is informed consent—which requires being as fully informed as possible (relative to development and level of understanding) of all the benefits/risks associated with the intervention and a freedom to withdraw at any time (148). When individuals are incapable of providing informed consent, assent—willing participation with sufficient understanding—is acceptable if accompanied by a legal guardian’s (or other legally appropriate third party’s) informed consent (148, 154). For very young children and people with significant cognitive impairments, sustained dissent—unwillingness to participate, expressed by repeated attempts to remove themselves from the situation—is increasingly favoured (148, 151). Unlike consent or assent, sustained dissent does not require understanding the intervention’s goals, risks or benefits; expressing a sustained unwillingness to experience distress, pain, or suffering is enough to warrant exclusion from the relevant activity (148, 154). This helps prevent harm from overriding an individual’s expressed preferences given the aversiveness of what they are (expecting) to experience (151).

Some scholars seek to adapt these human bioethics concepts for animal bioethics. Given uncertainty about animals’ ability to provide informed consent or meet the full requirements of assent, several scholars propose sustained dissent as the gold standard in animal bioethics (148, 151, 155). Others, like Jones (156), contend that the current human-centric definition of consent is inapplicable to animals like dogs but advocate for a species-specific definition. Jones (156) defines consent as, “the voluntary agreement for something to occur [while being connected to agency] in that it may afford dogs the freedom to choose when and how to interact and when to opt out” (p. 10). She notes that, through learning, dogs can make informed choices in some contexts and effectively communicate their willingness to participate (e.g., cooperative care); other contexts—like medical interventions—may require guardian (or other appropriate third party) decisions based on the dog’s best interest, aligning with assent (156). Similarly, Blattner (148) proposes an embodied consent model which she thinks resembles informed consent and suggests that animals’ consent to participate could be inferred by interpreting their communication (including body language and vocalisations)—although she acknowledges animals may not fully grasp the risks and benefits of the activity. Blattner (148) therefore recommends prioritising respect for dissent, while working toward broader consent and assent models.

To avoid misaligning the terms, so that the constraints on how these terms are used in human bioethics are not sufficiently captured in animal bioethics, this paper focuses on sustained dissent as a starting point in working with assistance dogs. Until more is known about dogs’ capacity for consent or assent, respecting sustained dissent is appropriately responsive to dogs’ expressed preferences and reflects a commitment to their well-being. Dogs can express sustained dissent in various ways (148, 153, 155). They regularly do so by disobeying commands, decelerating work, exhibiting avoidance behaviours, taking unauthorised breaks, objecting vocally, damaging equipment, or refusing to work under inadequate external (e.g., heat) or internal (e.g., illness) conditions (148).

Work environments aligned with interspecies justice should allow PTSD assistance dogs to stop or refuse work by honouring their sustained dissent (148), similar to how human workers can quit or refuse tasks. ADI-accredited providers implement this in part through annual recertification. However, it remains unclear whether yearly interval is optimal, and how effectively the varying follow-up care of different agencies compensates for this. The newly published EU standards also emphasise respecting sustained dissent by recommending the withdrawal of dogs exhibiting persistent stress or anxiety during training/work; however, they do not outline how this principle should be implemented (64, 78). Ng and Fine (51) recommend biannual assessments, particularly for older dogs. Some non-accredited organisations, however, lack regular recertifications altogether or conduct them infrequently, like every three years.

Blattner (148) and Coulter (155) argue that working animals should have the right to dissent at any time, enabling them to promptly cease work that causes discomfort or reluctance—just as human sustained dissent should be respected. Establishing respect for sustained dissent as a standard work condition for PTSD assistance dogs requires knowledge of and responsiveness to canine communication (148, 155). While recipients are normally educated on canine care/communication during pair-training, it is unknown how often dogs’ daily expressions of dissent—like stress signals or avoidance behaviours—are acknowledged. Burrows et al. (62) reported that one autism assistance dog’s reluctance to wear his service jacket was dismissed as laziness, though the jacket was later found to have become too small, causing discomfort. Similar reports of mobility assistance dogs in obvious distress while pulling wheelchairs or opening heavy doors (54), cast doubt on how well assistance dogs’ dissent is respected in practice.

Additional safeguards against assistance dog exploitation involve promoting the dogs’ well-being by ensuring they enjoy and benefit from their work (153, 155). Like good human jobs, good work for animals should minimise risks, foster pleasure, and offer meaning (146, 153). Cochrane (153) suggests that pleasurable animal work requires sufficient task variability to engage a range of enjoyable skills. Unlike, say, scent-detection or sled dogs focused on one task, PTSD assistance dogs use a broad skill set: olfactory and visual acuity to detect intrusion/hyperarousal symptoms (125), social skills to calm their handler and facilitate social interactions, fetching objects/medication, cognitive skills for tasks like “block”/“cover” or dialling emergency contacts, and, sometimes, physical abilities for mobility support.

However, a distinct question is whether these tasks are enjoyable or meaningful for the dogs. This matters both for their well-being and assistance effectiveness, as enjoyment affects motivation. Reflecting this, the European standards emphasise that training records shall document whether dogs enjoy particular assistive tasks and display natural aptitude for them (64). While screening potentially selects dogs who enjoy these tasks, many assistance tasks fall outside canine species-specific behavioural repertoire, offering little internal motivation (55, 92). Unlike sled or sheep dogs driven by innate motor patterns that make performance inherently rewarding (54, 63, 92), assistance tasks often lack relevance or intrinsic benefit to dogs (e.g., switching on lights, retrieving medication). Some, like wheelchair-pulling or door-opening, may even induce physical distress (54, 63, 92).

Due to the lack of internal motivation, biological relevance, and explicit reward, the motivation to perform such tasks must be generated externally (55, 92), relying heavily on operant conditioning. This parallels human labour, where difficult or unpleasant tasks are driven by external incentives (e.g., salary, academic degrees). As noted, aversive techniques are often used to “motivate” dogs, especially for ongoing (i.e., without clear beginning/ending) or physically uncomfortable tasks. However, aversive training is unequivocally unethical, contradicting principles of animal agency, respect for sustained dissent, and enjoyable work. Although even continuous positive reinforcement might not prevent a trained skill from diminishing if it diverges from intrinsic behaviours (63, 92), positive reinforcement-based training is more likely to increase motivation and make work more enjoyable.

Positive reinforcement-based training also aligns with other ethical principles of providing animals good work. Based on voluntary cooperation, it lets the animal choose whether to perform the task and receive the reward, or not perform and forgo it (90, 99, 100, 154, 156). Yet even positive reinforcement can sometimes become coercive (151), especially under high external pressure (such as the high demand and low supply of PTSD assistance dogs), where the principle of voluntary cooperation is overlooked and animals are pushed to work despite frustration or diminished motivation. In such circumstances, institutionalising a work ethic that respects sustained dissent is crucial. Furthermore, positive reinforcement-training can address the primary weakness of the dissent model—while dissent allows refusal, it does not confirm willingness (148). Truly voluntary, non-coercive, positively reinforced work can reflect assent (or acquiescence), as the animal chooses to work for a reward, indicating some motivation (83).

One non-coercive approach to boost dogs’ motivation for assistive tasks is to explore whether selectively bred dogs are more intrinsically motivated for PTSD assistance work, as proposed by WCC (60). Other strategies include broadening breed selection (including cross-breeding) to identify dogs with greater motivation for specific tasks or, as Gadbois and Reeve (157) suggest, considering breed differences in dopaminergic systems to select more work-motivated dogs. If no efforts succeed to make these tasks enjoyable and non-coercive methods fail, using dogs for such purposes should cease. That dogs can be trained for these tasks does not mean that they should be—especially if it presents a morally unacceptable risk of harm. For example, motorised/electric wheelchairs can eliminate the need for dogs to pull them (78). While occasional frustration and discomfort are natural parts of learning—much like in human skill acquisition—PTSD assistance dog training and work must be enjoyable overall to be ethically acceptable (153, 155). Tasks that remain consistently unpleasant and distressing even in fully trained dogs should be excluded.

Another way to promote benefits for PTSD assistance dogs involves deliberately nurturing their bond with the recipient. An increasing body of research shows the human-dog bond offers substantial advantages for people with PTSD (9, 24–26, 29, 47). Strong recipient-dog bond also improves assistance dog training outcomes, work performance, and the overall success of the partnership (26, 29, 35, 47, 55). Accordingly, most PTSD assistance dog agencies already prioritise this bond, dedicating varying amounts of time to its development (26, 35, 47, 50).

A strong recipient-dog bond should also guarantee benefits for the dog, as a human-animal bond is, by definition, a mutually beneficial positive relationship that promotes both partners’ well-being and health (35, 55, 158). However, few studies have investigated how bond quality affects assistance dogs and whether such bonds are consistently present. Lane et al. (80) found no conclusive evidence that mobility assistance dogs’ welfare was influenced by their relationship with recipients. A review by Hall et al. (55), however, reported that human-dog bonds based on secure attachments are linked with dogs’ lower stress and neuroticism, and greater persistence in novel contexts. It also remains unclear whether instrumentalising dogs is detrimental to their well-being in contexts where bonding is encouraged. Overall, more research is urgently needed on how attachment styles and recipient-dog relationship dynamics can advance ethical, sustainable assistance dog practices (55).

Finally, Coulter (155) argues that enjoyable work includes supporting the animal’s well-being outside working hours—both daily and across their lifetime. While we have addressed daily needs like quality downtime, ethical work standards require a holistic approach that also prioritises well-being before and after their formal careers. Coulter (155) illustrates this by pointing to human parallels like childcare, education, pensions, and elder care. If the welfare issues outlined in subsection 3.1 are adequately mitigated, the upbringing (by dedicated foster families) and training of PTSD assistance dogs can significantly enhance their lifelong well-being by preparing them for future roles. This rationale implies that dogs entering the program as puppies may enjoy better overall well-being throughout their working lives than those recruited later.

Additionally, much can be done to promote an enjoyable retirement for former assistance dogs, regardless of living arrangements. Alongside facilitating retirement adjustment, providing ample enrichment, stimulation, and activities that match their interests is essential (51). This priority is acknowledged in the new European standards, which require a dedicated health and well-being plan for retired assistance dogs (78). A major unresolved question—reflected in the lack of standards—is when to retire an assistance dog. While illness impairing work (e.g., musculoskeletal disease, cognitive dysfunction) clearly warrants retirement (51, 102), truly enjoyable retirement requires retiring dogs while still healthy and free of debilitating conditions. This implies enforcing retirement before signs of illness appear (51), creating a potential conflict of interest, as retiring a fully functional dog hardly aligns with the recipient’s interests.

To fulfil our ethical duty to provide good retirement, data on the age at which assistance dogs retire and the reasons for retirement should be systematically collected and analysed. Such information is essential for understanding how best to support healthy ageing and to determine the optimal retirement age for these dogs (102). Ng and Fine (51) recommend retiring dogs during the final 25% of their life expectancy—typically their senior life stage. This accounts for breed-specific lifespans and roughly compares to human retirement. In addition to monitoring behavioural changes or performance decline, they suggest measuring hair cortisol to detect rising overall stress and conducting surveys to recognize declines in quality of life (51). These tools could help determine when to retire otherwise healthy dogs.