Researchers watched 837 therapy sessions to ask whether the dogs were actually okay
Doges Editorial · 2026-05-07 · 6 min read
A new study coded 19 behavioral indicators across 63 therapy dogs and 837 campus sessions — and found that while most dogs show engagement and comfort, young clients and heavy workloads generate real stress signals that programs have been too quick to overlook.
Therapy dog programs have expanded dramatically across university campuses over the past decade, driven by a logic that is both compassionate and commercially legible: students experiencing anxiety or stress report feeling better after time with a dog, the evidence behind those self-reports is reasonably strong, and the programs are popular. What has received less systematic attention is the other participant in the interaction. The dog is present at every session, often for extended periods, and often with multiple clients in succession. The dog did not consent to the job in any sense we can access. And the dog's emotional state during those sessions — whether it is experiencing something like genuine engagement or something more like chronic low-level stress — is not something the student on the other side of the interaction is typically equipped to notice or evaluate.
Eight hundred and thirty-seven sessions under observation
Researcher John-Tyler Binfet and colleagues at the University of British Columbia Okanagan set out to take that question seriously in the study's design rather than as an afterthought. Their paper, published in the journal Animals in 2026, video-recorded 837 therapy dog sessions involving 63 dogs working in on-campus programs. Two independent coders assessed each recording against a 19-item behavioral checklist covering indicators of positive affect, stress, arousal, and uncertainty. The indicators ranged from obvious signals — tail wagging as positive, lip licking and yawning as stress markers — to more nuanced cues like backing away from contact, decreased engagement with clients over the course of a session, or active play solicitation behavior that signals an actively positive experience rather than passive tolerance.
The overall picture was mixed in a way that is useful precisely because it does not produce the flattering conclusion that program administrators might prefer. Most sessions showed dogs displaying the behavioral profile of comfortable, engaged animals: tail wagging was common, active engagement with clients was frequent, and overt stress signals were not the modal experience. But stress signals were present in a meaningful proportion of sessions, and they were unevenly distributed in ways that the data could identify and describe with precision.
Where the stress signals concentrated
Two patterns emerged with particular clarity in the analysis. First, sessions with younger clients — children rather than young adults in the typical university context — generated heightened uncertainty and tension indicators in the therapy dogs. Children's movements tend to be less predictable, their vocalizations more abrupt and high-pitched, and their physical interactions with animals less consistent with the handling patterns a trained dog has been conditioned to expect from adult handlers. Even dogs with extensive therapy certification showed elevated stress markers during sessions with young clients. This finding does not suggest children cannot benefit from therapy dog programs; it suggests the dogs assigned to those programs need to be selected and monitored specifically for their comfort in high-stimulation, unpredictable-handling environments.
Second, older dogs showed less playfulness and more settled behavior across sessions — which sounds benign and is largely benign, but the data also showed that older dogs in high-volume programs accumulated stress indicators across extended sessions in ways that suggested fatigue effects. Female dogs showed more uncertainty and arousal markers than males across the full dataset. These granular findings give program administrators specific, actionable information: not broad policy mandates but targeted insights about matching individual dogs to appropriate client populations and managing per-dog workloads to prevent cumulative stress buildup.
Dogs are widely recognized as sentient beings capable of experiencing a full range of affective states, including fear, joy, frustration, comfort, and pain.
— Haven-Pross et al., Animals journal, 2026
The consent problem in therapy animal work
The philosophical dimension of this research is one that the authors approach carefully and practically rather than as an abstract exercise. A therapy dog cannot inform a program coordinator whether it would prefer to be doing something else. The closest available proxy is behavioral observation — watching for the signals that animal behaviorists associate with negative affect, monitoring for stress accumulation across sessions, and taking those signals seriously rather than assuming that a dog trained to be calm in a given setting is therefore flourishing in that setting. Training can suppress the behavioral expression of stress without eliminating the underlying physiological experience. That distinction matters if animal welfare is treated as a genuine design constraint for therapy programs rather than a compliance checkbox.
Optimizing dog welfare requires matching dogs to suitable roles, attentive session planning, and managing workload — not simply selecting well-trained dogs and deploying them at scale.
— John-Tyler Binfet, University of British Columbia Okanagan
What good program design looks like going forward
The study's practical recommendations are specific. Programs should monitor individual dogs across sessions using structured behavioral observation rather than relying on handler impressions alone, since handlers who have deep bonds with their dogs are not always the most objective assessors of stress signals in animals they care for. Session length limits should be calibrated to observed behavioral data from individual dogs rather than generic guidelines applied uniformly. Dogs should receive adequate rest between sessions. Young-client populations should be matched with dogs who have been specifically assessed for high-stimulation tolerance — not assumed to be appropriate for any dog that has completed standard therapy certification.
The broader implication is that the rapid expansion of therapy dog programs — which has accelerated in lockstep with the documented mental health crisis on university campuses over the past decade — needs to be matched by commensurate investment in the welfare infrastructure that makes those programs sustainable over the long term. A program that burns through therapy dogs by running them at high volume without adequate monitoring, rotation, and individual welfare assessment is not serving students well, is not serving handlers well, and is certainly not serving the dogs well. The evidence base for what therapy dogs offer to human mental health is now substantial and growing. Building the parallel evidence base for what we owe therapy dogs in return is not a secondary consideration — it is overdue. The animal welfare community has spent the last decade building the scientific case that dogs experience complex emotional states and that those states carry moral weight. The therapy dog context is one of the most direct and actionable arenas where that argument has practical implications — not as abstract ethics but as a design requirement for programs that intend to operate sustainably. Programs that take dog welfare seriously will monitor behavioral indicators systematically, match dogs to client types by individual assessment, and build rest and rotation into their operational structure as a baseline expectation rather than an optional enhancement. The Binfet research gives program administrators the evidence and the vocabulary to make that choice. Whether they use it is a different question — but the research has made the choice harder to avoid.