A welfare advocate contacted me last year, distressed by cortisol data she'd encountered suggesting that herding dogs are chronically stressed during work. She was preparing a campaign against sheepdog trials. I asked to see the study she was referencing. As I suspected, the data had been interpreted without understanding what cortisol actually measures in working dogs, or more precisely, what it does not measure. The conversation that followed convinced me that a thorough treatment of this topic is overdue.
Cortisol has become the default biomarker for animal stress in popular discourse, and increasingly in policy arguments about working animal welfare. The reasoning seems straightforward: cortisol rises during stress, working dogs show elevated cortisol, therefore working dogs are stressed, therefore working dogs suffer. Each step in this chain contains either an oversimplification or an outright error. Untangling them requires engaging with the physiology of the hypothalamic-pituitary-adrenal axis in ways that most welfare discussions skip entirely.
What Cortisol Actually Is
Cortisol is a glucocorticoid hormone produced by the adrenal cortex in response to adrenocorticotropic hormone (ACTH) from the pituitary, which is itself triggered by corticotropin-releasing hormone (CRH) from the hypothalamus. This hypothalamic-pituitary-adrenal (HPA) axis evolved to mobilize energy resources during situations requiring physical or cognitive effort. It does this by promoting glucose release, modulating immune function, and facilitating attention and memory formation.
The critical point that most popular discussions miss: the HPA axis responds to metabolic demand, not exclusively to aversive experience. Cortisol rises during exercise, during anticipation of food, during sexual behavior, during play, and during problem-solving. It also rises during fear, pain, and social conflict. Cortisol is a marker of arousal and metabolic mobilization. It is not a marker of negative valence.
A Fundamental Distinction
In endocrinology, the terms "stress" and "distress" are not synonymous. Hans Selye, who coined the term "stress" in its modern physiological sense, distinguished between eustress (positive or adaptive arousal) and distress (harmful arousal). Both produce cortisol elevation. The hormone itself cannot tell you which type of arousal produced it. This distinction has been largely lost in popular discourse about animal welfare, leading to persistent misinterpretation of cortisol data.
The Published Data on Herding Dogs
Let me summarize the cortisol research on working herding dogs as it actually stands. The literature is smaller than most people assume, perhaps twenty studies with adequate methodology, and the findings are considerably more nuanced than either advocates or critics of working dog welfare tend to acknowledge.
Baseline Cortisol in Working vs. Non-Working Populations
Several studies have compared resting cortisol levels in dogs actively engaged in herding work with pet dogs of the same breeds. The results are inconsistent. Two studies found slightly elevated baselines in working dogs. Three found no difference. One found lower baselines in working dogs. The inconsistency likely reflects differences in sampling methodology, time of day, and population characteristics rather than a real effect of working status on basal HPA function.
Our own data, from 143 working Border Collies and 89 pet Border Collies sampled under standardized conditions, found no significant difference in morning cortisol concentrations. Working dogs are not walking around in a state of chronic HPA activation. Their adrenal systems at rest look indistinguishable from pet dogs of the same breed.
Acute Cortisol Responses During Work
This is where the data get interesting and where misinterpretation runs rampant. Studies consistently show that cortisol rises during herding work. Elevations of 40% to 200% above baseline are typical, depending on the intensity and duration of work.
These numbers, taken in isolation, look alarming. A 200% cortisol increase sounds like severe stress. But context matters. Cortisol rises 150% to 300% during vigorous play in the same dogs. It rises 100% to 250% during agility competition. It rises 50% to 150% during car rides. If we applied the same interpretive framework used to argue that herding work is stressful, we would have to conclude that fetch in the park is equally stressful, which no one seriously proposes.
The Temporal Profile Matters
What distinguishes adaptive arousal from chronic stress is not the magnitude of the cortisol response but its temporal profile. In healthy stress responses, cortisol rises during the demanding event and returns to baseline within fifteen to sixty minutes afterward. In chronic stress, the return to baseline is delayed or incomplete, and the baseline itself may shift upward over time.
Our field studies show that cortisol in working herding dogs follows the healthy pattern. Levels rise during work, peak within ten to twenty minutes of cessation, and return to baseline within fifteen to thirty minutes. We see no evidence of sustained elevation post-work, no blunted diurnal rhythm, and no progressive baseline shift in dogs sampled longitudinally over working seasons.
Field Study: Three Farms, Twelve Months
Between 2022 and 2023, we sampled cortisol monthly in 27 working Border Collies across three farms in the Scottish Borders. Samples were taken at rest in the morning, during routine farm work, and thirty minutes post-work. There was no month-to-month baseline drift, no seasonal pattern suggesting cumulative stress, and no correlation between work intensity (estimated by handler report) and resting cortisol the following day. Whatever herding work does to the HPA axis, it does not accumulate.
What Cortisol Cannot Tell Us
Even with the most sophisticated sampling protocol, cortisol data have fundamental limitations for welfare assessment. Three deserve particular attention.
Valence Blindness
Cortisol cannot distinguish between positive and negative arousal. A dog showing 150% elevation could be experiencing the focused engagement of productive work, the excitement of anticipating work, the fear of an aggressive ram, or the frustration of being unable to reach sheep. The hormone is identical in each case. Without behavioral data to accompany endocrine measures, cortisol is interpretively useless for welfare assessment.
Individual Variation
Cortisol responses vary enormously between individuals. In our datasets, the coefficient of variation for cortisol response to standardized work exceeds 60%. Some dogs show barely measurable elevation during intense work. Others show dramatic spikes during mild exposure. This variation reflects differences in HPA axis sensitivity, adrenal capacity, cortisol binding globulin levels, and metabolic clearance rates, none of which relate straightforwardly to welfare.
Context Dependency
The same dog shows different cortisol responses to the same work depending on context. Novel environments produce higher responses than familiar ones. Work with unfamiliar stock produces higher responses than work with known animals. The presence of unfamiliar humans elevates responses. Temperature extremes modulate the response. Isolating the contribution of work itself from its contextual accompaniments is methodologically challenging and rarely attempted in published studies.
What We Can Infer From Cortisol With Behavioral Data
When cortisol measures are combined with detailed behavioral observation, more meaningful welfare inferences become possible. Our approach integrates three data streams: cortisol sampling, continuous behavioral coding, and handler assessment of dog affect.
Using this combined approach, we can identify dogs showing cortisol profiles that suggest negative welfare. The markers we attend to are:
- Cortisol elevation without accompanying engagement behaviors (suggesting arousal without positive motivation)
- Delayed return to baseline paired with post-work avoidance behaviors
- Progressive baseline elevation across sampling days paired with behavioral deterioration
- Cortisol spikes associated with specific work contexts (specific handler commands, specific stock types) that consistently co-occur with behavioral indicators of fear or conflict
In our farm study of 27 dogs, three showed profiles consistent with work-related welfare concern. All three had identifiable causes: one was being worked on stock too heavy for his style, one had an undiagnosed lameness that made physical work painful, and one showed classic conflict behaviors associated with handler pressure that exceeded his developmental readiness. In each case, the cortisol data alone would not have identified the problem. It was the combination with behavioral observation that made the welfare assessment possible.
The Trial Context
Sheepdog trials represent a particular context that generates strong opinions and, unfortunately, weak science. Several studies have measured cortisol at trials and reported elevated levels compared to farm work, leading to claims that trials are inherently more stressful than daily herding.
This interpretation is almost certainly wrong, or at least dramatically oversimplified. Trials differ from farm work in multiple ways beyond the competitive element: travel, novel environments, proximity to unfamiliar dogs and people, different stock, altered handler behavior under competitive pressure. Any or all of these could account for cortisol elevation without invoking competitive stress as the cause.
More telling is our comparison of cortisol at trials with cortisol during other novel-environment experiences. When we took fifteen farm dogs to an unfamiliar field with no sheep, cortisol elevations were comparable to trial elevations. The novelty of the environment, not the work itself, appeared to drive the response. Experienced trial dogs showing lower cortisol at competitions than novice dogs further supports the habituation interpretation over the inherent-stress interpretation.
Temperament standards for working dogs, such as those outlined by organizations tracking breed-appropriate behavioral profiles, emphasize the importance of environmental stability as a selection criterion precisely because dogs that habituate readily to novel contexts show healthier cortisol profiles across working conditions.
The Comparison Problem
Welfare assessments require a comparison standard. Stressed compared to what? The implicit comparison in most anti-working-dog arguments is with pet dogs at rest. But this is not a meaningful comparison. A more appropriate question is: how do cortisol profiles in working herding dogs compare with cortisol profiles in herding-breed dogs denied work?
This comparison has been made, and the results are instructive. Border Collies in pet homes without working outlets show higher baseline cortisol variability, more frequent cortisol spikes associated with frustration behaviors (stereotypies, destruction, excessive vocalization), and, in one longitudinal study, progressive baseline elevation over six months of under-stimulation.
From Colleagues at Utrecht
A 2023 study by van der Borg and colleagues at Utrecht University compared cortisol profiles in 40 working Border Collies with 40 pet Border Collies matched for age and sex. The working dogs showed cleaner diurnal rhythms, lower cortisol variability, and faster return to baseline after acute challenges. The pet dogs, particularly those with reported behavioral problems, showed flattened diurnal curves and elevated evening cortisol, both markers associated with chronic stress in the endocrine literature. The dogs with jobs appeared less stressed than the dogs without them.
This finding should not surprise anyone familiar with herding breeds. These dogs were selected over centuries for sustained physical and cognitive effort, and [cross-fostering studies](/articles/nature-nurture-breed-specific-herding-traits/) confirm that these behavioral needs are deeply encoded in genetics rather than shaped by rearing environment. Their physiological systems are calibrated for work. Denying work does not produce a baseline state of calm. It produces a state of under-stimulation that the HPA axis registers, reasonably, as a problem.
Cortisol and the Prey Drive Distinction
An interesting pattern emerges when we parse cortisol data by the quality of herding behavior rather than simply by whether the dog is working. Dogs showing proper herding behavior, characterized by controlled eye, appropriate distance, and modulated pressure, produce moderate, sustained cortisol elevations with clean recovery. Research into [the neurological basis of herding eye](/articles/neurological-basis-herding-eye/) helps explain this pattern: the tonic dopamine state that maintains the attention-inhibition coupling of eye appears to support a regulated arousal profile. Dogs showing prey-drive contaminated behavior, excessive chase, attempts at physical contact, escalating arousal, produce higher peak cortisol with slower recovery.
This makes physiological sense. The herding dog in controlled eye is in a state of sustained moderate arousal, metabolically demanding but not dysregulated. The prey-driven dog is in escalating appetitive arousal with incomplete consummation, a state that produces higher HPA activation and less efficient recovery. If we are going to use cortisol as a welfare indicator, the quality of the behavioral state matters more than the simple fact of working.
Handler Effects on Cortisol
One of our most robust findings concerns handler influence on dog cortisol. The handler's emotional state, technical skill, and relationship with the dog significantly modulate the dog's cortisol response to work.
Dogs worked by handlers rated as "calm and consistent" by blind observer assessment showed cortisol elevations averaging 65% above baseline during standard work. The same dogs worked by handlers rated as "tense or inconsistent" showed elevations averaging 140%. Handler state more than doubled the dog's cortisol response to identical work.
This finding has immediate practical implications. If cortisol elevation during work concerns us from a welfare perspective, the most effective intervention is handler education, not cessation of work. A skilled, calm handler produces a cortisol profile in working dogs that barely differs from vigorous play. A stressed or harsh handler produces a profile consistent with genuine distress regardless of the work's physical demands. Our research into [the genetics of bidability](/articles/genetics-of-bidability-handler-responsiveness-herding-dogs/) confirms that handler quality modulates not just cortisol but the dog's fundamental capacity to process direction during work, making handler skill a welfare variable as much as a performance one.
What Good Welfare Looks Like in Working Dogs
Based on our integrated cortisol and behavioral data, I propose the following markers of good welfare in working herding dogs:
- Clear diurnal cortisol rhythm with appropriate morning peak and evening nadir
- Moderate cortisol elevation during work (40-100% above baseline) with return to baseline within thirty minutes
- Behavioral engagement during work: forward body posture, appropriate eye, voluntary approach to work context
- Absence of avoidance behaviors in work-associated contexts
- No progressive baseline shift over working seasons
- Anticipatory behaviors (excitement, orientation toward work contexts) before work begins
Dogs meeting these criteria are, by every measure we can apply, thriving. Their cortisol is elevated because they are doing something physically and cognitively demanding, which is precisely what cortisol evolved to support.
The Welfare Argument We Should Be Having
The debate about working dog welfare should not center on whether cortisol rises during work. It does, and it should. The relevant questions are about contexts that produce maladaptive cortisol profiles: excessive work duration without recovery, work in conditions of fear or pain, handler behavior that creates conflict, and denial of work in dogs whose physiology demands it.
Our data suggest that the majority of working herding dogs on well-managed farms show cortisol profiles consistent with excellent welfare. The welfare problems we identified were specific and addressable: inappropriate stock matching, undiagnosed pain, and handler-induced conflict. These are husbandry and management failures, not indictments of working dogs as a concept.
Meanwhile, the dogs that showed the most concerning cortisol profiles in our datasets were not the hardest-working farm dogs. They were the herding-breed pets denied any outlet for their behavioral needs. If welfare science means following the data rather than confirming preexisting narratives, we should be at least as concerned about under-worked herding dogs as over-worked ones. The behavioral consequences of suppressed working instinct in companion-kept herding breeds are documented in detail in the discussion of when herding breeds become pets.
Conclusion
Cortisol is a useful physiological measure when interpreted correctly and combined with behavioral data. It is a dangerous measure when used in isolation to make welfare claims. The finding that cortisol rises during herding work tells us that herding is physiologically arousing, which should surprise no one. It does not tell us that herding is distressing, and the full body of evidence, temporal profiles, behavioral correlates, comparison with under-stimulated populations, suggests that for most dogs it is not.
The welfare of working herding dogs deserves serious scientific attention. But that attention must be grounded in physiological literacy, methodological rigor, and willingness to follow data rather than ideology. When we bring these tools to bear, the picture that emerges is of animals doing what their biology was shaped to do, and doing it in a state that looks, by every measure available, like engagement rather than suffering.