Emotional Contagion in Dogs: Can Human Stress Influence Canine Stress Responses?
Michael Sauerwein
Dogs are extraordinary readers of human emotion. Thousands of years of living alongside us have tuned them to our gestures, faces, voices, and moods, and a large literature confirms they can tell a happy human from an angry one and adjust accordingly. But a more striking possibility has come into focus recently: that dogs do not merely read our emotional states but partly catch them — that a stressed owner can, through channels the dog never chose, shift the dog's own physiology and behavior. This is emotional contagion, and if it is real in the dog–human relationship, it means the human at the end of the leash is part of the dog's emotional environment, for better and for worse.
This article reviews what is known and holds it to a clear standard. It keeps two distinctions in view throughout. The first is between recognizing an emotion and sharing it: contagion is an automatic, low-level state-matching, not evidence that the dog understands or feels for the person. The second is between what the evidence shows and what it can prove: much of this research is correlational, so a genuine synchronization between owner and dog does not, by itself, establish that the human's state caused the dog's. Read with those caveats, the findings are relevant: behavioral responses, physiological associations and experiments with human odor samples show that dogs can respond to cues of human stress. What this means for welfare and training in everyday life has not been studied directly for every situation.
1. Introduction
1.1 From Reading Emotions to Catching Them
There is a meaningful gap between a dog that behaves appropriately around an angry person and a dog whose own arousal rises to match that person's. The first is recognition — a learned or perceived cue guiding behavior. The second is contagion — an actual shift in the dog's internal state toward the human's (dogs are unusually skilled at reading human signals to begin with). This article is mostly about the second, harder claim, and about the evidence that human stress specifically is one of the states that transfers. That evidence does not show that every human emotion transfers to every dog: the studies concern a small set of states, mostly distress and stress, in particular test situations, and whether a given owner's state reaches a given dog in everyday life depends on the channel, the situation and the individual animal.
1.2 How to Read the Evidence
Two cautions matter. Emotional contagion is a basic mechanism — automatic emotional synchronization, not the higher-order, perspective-taking empathy that would require the dog to model the human's mind (what dogs can and cannot be said to understand). And much of the evidence is correlational: owner and dog states are found to move together, which is compatible with human-to-dog transfer but does not prove its direction on its own. Where a study is correlational, this article says so.
1.3 Why the Question Is Harder Than It Looks
Three claims are routinely bundled together and require different evidence.
That dogs detect human emotional states — well supported, through several sensory channels. That dogs take on those states — the contagion claim, supported by physiological and behavioral convergence. And that dogs respond in order to help — which is where the evidence runs out and the confident statements continue.
This article keeps them separate throughout, because almost every popular account of canine empathy slides from the first to the third without pausing at the second.
2. Emotional Contagion as a Basic Form of Empathy
2.1 What It Is
Emotional contagion is the automatic synchronization of emotional states between individuals, and it is considered one of the most basic building blocks of empathy (Preston & de Waal, 2002). Unlike cognitive empathy, it requires no reasoning about another's perspective; it runs on fast, low-level processes. In social species it is adaptive — the rapid spread of fear or vigilance through a group lets everyone respond to a threat before each individual has independently detected it. Framed this way, catching a companion's alarm is not a sophisticated feat but an ancient survival shortcut.
2.2 Why Dogs Are So Sensitive to Us
Dogs bring unusual sensitivity to human signals into this system. Using a cross-modal task pairing faces with voices, researchers showed that dogs integrate visual and auditory cues to categorize emotional states in both dogs and humans — matching an angry face to an angry voice, a happy face to a happy voice — without prior training on this particular matching task, although an influence of earlier experience with such signals cannot be fully excluded (Albuquerque et al., 2016). The finding supports the perception and integration of emotional cues; whether this leads to taking on the emotional state is a separate question.
2.3 The Proposed Mechanism
The standard account describes a perception–action link: perceiving another's state activates, in the observer, some of the same representations that would accompany that state directly (Preston & de Waal, 2002).
On this model, no conscious inference about the other's inner state is required. It remains a theoretical account that the dog studies discussed here do not directly demonstrate as a neural mechanism.
2.4 Which Channels Carry It
For three sensory channels, studies show that dogs perceive or respond to emotional cues; that does not yet show emotional contagion for each channel. A fourth is plausible in everyday life. Visual — facial expression and posture, with evidence that dogs process human emotional expressions as such. Auditory — vocalizations, including recordings with no visual component at all. Olfactory — human stress chemosignals, covered in section 6. Tactile and mechanical — leash tension, muscle tone through contact, handling; plausible, but not isolated experimentally in the studies reviewed here.
Isolating one channel is what the experimental designs do and is not what happens on a walk, which is worth remembering when reading effect sizes from single-channel studies.
2.5 Why Dogs Rather Than Other Domesticates
Dogs are among the earliest domesticated species and have lived closely with humans for thousands of years. Selection, rearing and individual experience may have shaped their responsiveness to human signals. The ability to read human gestures and expressions is present across breeds, although breeds also differ in it.
Whether that history produced heightened contagion specifically, or whether dogs simply have more opportunity to display it, has not been tested comparatively against other domesticated species living in similar proximity.
3. Physiological Synchronization Between Dogs and Humans
3.1 Heart Rate and Autonomic Coupling
Contagion shows up not only in behavior but in the body. Measuring heart rate variability, a fine-grained index of autonomic (sympathetic/parasympathetic) balance, in owner–dog pairs under a stress condition imposed on the owner, researchers found that the dog–owner correlation in one heart rate measure only tended to be higher under stress than in the control condition. The clearer result was that the dog–owner correlations in heart rate variability measures increased with how long the dog had lived with the owner; a closer bond as such was not measured, and only 14 of 34 pairs could be analyzed (Katayama et al., 2019). This is compatible with shared experience influencing the measured coupling but does not prove it in the small cross-sectional sample (arousal regulation is a shared, coupled process).
3.2 Long-Term Cortisol Synchronization
The most compelling endocrine evidence comes from hair cortisol, which reflects stress over weeks to months rather than minutes. In a study of 58 pairs of Shetland sheepdogs or border collies and their female owners, dogs' long-term cortisol at two sampling points, summer and winter, correlated with their owners'. Strikingly, the dogs' physical activity and the number of training sessions did not affect their hair cortisol, whereas owner personality traits did (Sundman et al., 2019). A shared cause is the alternative that has to be named: owner and dog inhabit the same household, the same routine and the same disruptions, and a stressful period affects both without either transmitting anything to the other. One possible reading is that the caregiver's living conditions or stress are related to the dog's stress physiology (the machinery of chronic cortisol and the HPA axis). The honest caveat is that this is a correlation: it is compatible with human-to-dog influence but cannot exclude a shared environment, influence in the other direction or further confounds.
3.3 What Heart Rate Coupling Does and Does Not Show
Autonomic coupling between dog and owner is measurable and is the least interpretable of the findings. Two individuals in the same room, responding to the same events, will show correlated heart rates without either one catching anything from the other.
The designs that address this manipulate one party's state and observe the other, rather than simply recording both. Where that has been done, as in the heart-rate study with a stressor imposed on the owner (Katayama et al., 2019), the difference between the stress and the control condition was only a statistical trend in a small analyzed sample, so this remains the weakest of the three lines of evidence discussed here.
3.4 Why the Hair Cortisol Study Is Different
Long-term cortisol measured from hair integrates over months rather than minutes, which removes the shared-moment problem entirely. Two individuals reacting to the same doorbell do not thereby have matching hair cortisol across a season.
That is what makes the synchronization finding substantive: it describes a sustained relationship between two organisms' stress physiology rather than a shared response to a shared event (Sundman et al., 2019).
3.5 The Direction Finding
The same study provided indirect evidence on the possible direction of the association, and reported that owner personality traits (neuroticism, conscientiousness and openness) related to the dog's long-term cortisol (owner effects on dog behavior more broadly), whereas the dog's own personality had little effect on it. From this the authors concluded that dogs mirror their owners' stress rather than the reverse (Sundman et al., 2019).
That is one dataset and a correlational design. The results justify a testable hypothesis about the direction of influence but do not replace an experimental or longitudinal design. It also did not hold for every kind of dog: in ancient and solitary hunting breeds, no synchronization of long-term cortisol with the owner was found, and the dogs' values were related mainly to the owner–dog relationship (Höglin et al., 2021).
3.6 Why Owners Notice This Before Researchers Did
The phenomenon was described by dog owners long before it was measured, and the descriptions were largely accurate about the observation while wrong about the mechanism.
"He knows when I'm upset" is a report of covariation between two states, which is what the synchronization data document. "He's trying to cheer me up" adds an intention the data do not reach. Separating the two is most of what this article does.
4. The Behavioral Evidence
4.1 The Crying Study
The most-cited behavioral demonstration placed a dog in a room with its owner and a stranger. Each person alternately talked, hummed or pretended to cry, and the dog's behavior was recorded (Custance & Mayer, 2012).
Dogs oriented toward the person more often when that person was crying than when talking or humming, and observers blind to the condition mostly classified their approaches during crying as submissive. Critically, they did so whether the crying person was the owner or the stranger.
4.2 Why the Stranger Result Matters
If approaching were simply a bid for reassurance, the dog would be expected to seek its owner regardless of who was crying. That dogs also approached a crying stranger argues against the simple explanation that they only seek comfort from the owner. Whether this approach was meant to help the other person or to regulate the dog's own response remains open.
The authors describe the responding as empathic-like and are careful about the label, noting that a non-insightful explanation is available: dogs may have been rewarded for approaching distressed humans throughout their lives.
4.3 The Physiological Test
A separate study with 75 dogs removed the visual component entirely, playing recordings of an infant crying, an infant babbling, and white noise (Yong & Ruffman, 2014).
Cortisol rose only after crying — not after babbling, and not after white noise. The behavioral response to crying was a combination of submissiveness and alertness that appeared in neither control condition.
4.4 Why the Control Conditions Carry the Argument
White noise is an acoustic control condition, babbling a human vocalization; whether a particular sound is experienced as unpleasant cannot be inferred from its category alone. Only crying produced a cortisol response, together with a combination of submissiveness and alertness that the authors describe as a distinct behavioral response to crying.
That pattern is difficult to explain as a generic response to noise or to human voices, and the authors read it as the first clear physiological evidence of a primitive cross-species emotional contagion.
4.5 The Alternative Reading
One interpretation runs through this whole literature and is rarely resolved. A dog that approaches a distressed person may be responding to that person's state, or may be seeking social contact to reduce its own distress — which the contagion has just produced.
Both accounts can predict approach. Measuring the dog's own arousal before and after contact would help; in the studies discussed here this was not tested directly (which is the general limit on reading motivation from behavior).
4.6 What Would Separate Them
Three designs. Comparing approach to a distressed person against approach to a neutral person offering equivalent social contact. Measuring the dog's physiological recovery with and without contact. And testing whether dogs approach a distressed person they cannot obtain contact from.
These contrasts have not been examined systematically in the studies reviewed here. It therefore remains open what motive lies behind an approach in a given case.
5. Oxytocin and the Social Bond
How strongly contagion operates may depend partly on shared history, since in one small study the dog–owner coupling increased with the duration of ownership (Katayama et al., 2019), and oxytocin has been proposed as one reason why. Oxytocin, central to social attachment, is released during positive contact such as mutual gaze and touch; in a well-known study, longer dog–owner mutual gaze was associated with rises in urinary oxytocin in both species, which the authors proposed as a positive feedback loop of the kind described between human parents and infants (Nagasawa et al., 2015). This system is thought to support attachment and cooperation (how the oxytocin bond between dog and human develops) — but the same sensitivity that makes a bonded dog responsive to a caregiver's warmth may also make it more responsive to that caregiver's distress. A closer bond would then be not only protective but possibly also a wider channel (which is why attachment quality shapes how a dog copes).
5.1 Why Oxytocin Alone Does Not Explain Contagion
The oxytocin loop describes a mechanism for bonding rather than for state transmission. A dog and owner whose oxytocin rises during mutual gaze are engaged in an affiliative exchange, which is not the same phenomenon as one party's stress appearing in the other.
The two are related and are frequently conflated. Responses to a stranger's stress odor and to recorded infant crying were observed without any bond to the source of the stimulus. That shows a response to emotional cues, not necessarily the adoption of the same emotional state.
6. Olfactory Detection of Human Stress
6.1 Chemosignals of Fear
Dogs may also pick up human emotion through smell, via chemosignals in sweat (a channel suited to the dog's extraordinary olfactory system). Labrador and golden retrievers, in the presence of their owner and a stranger, were exposed to armpit sweat collected from men in a fear or a happiness condition, or to a control condition with no sweat. The fear odor produced measurable effects: dogs showed more stress-related behaviors and higher heart rates than in the happiness and control conditions. With happiness odor, by contrast, they directed fewer behaviors to the owner and more to the stranger (d'Aniello et al., 2018). The donors were not present, so the dogs had no visual or auditory information about their state, and the odor condition was associated with changes in their behavior and physiological measures. Whether they actually took on the donors' emotional state remains open.
6.2 Stress Odor and Judgment Bias
More recent work sharpens the point. Testing whether the smell of a stressed stranger changes how a dog interprets ambiguity, researchers tested 18 dogs: with the stress odor of unfamiliar people, dogs in the third test session were less likely to approach a food bowl at one of three ambiguous locations, the one near the unrewarded location, than with no odor. The authors read this as possible risk-reduction behavior (Parr-Cortes et al., 2024). This matters because it suggests that human stress odor may not just raise arousal in the moment but may also affect how a dog evaluates an ambiguous situation. The finding concerns one location in one session and has not yet been replicated (the kind of negative-expectation state associated with anxiety).
6.3 What a Dog Smells Is Not a Diagnosis
The finding is not that a dog smells sadness. What is detectable is a change in body chemistry associated with a state, and the association between that change and anything the person is going through is supplied by context and by the dog's history rather than by the odor.
The compounds have not been identified, so nobody can say what the dog is responding to. "Dogs smell fear" compresses a measurable chemical effect into a claim about content, and the second part is the invented one.
6.4 What the Chemosignal Studies Add
Odor experiments are an important test of the effect of human stress cues, for a methodological reason: odor can be delivered without the dog seeing, hearing or touching anything, and without the source person being present at all.
Sweat samples collected in a fear or a happiness condition produced different behavioral and physiological responses in dogs, compared with each other and with a no-sweat control (d'Aniello, Semin, Alterisio, Aria & Scandurra, 2018), and the odor of an unfamiliar stressed person shifted responses in a cognitive bias test (Parr-Cortes et al., 2024).
6.5 Why the Cognitive Bias Result Matters Most
A judgment bias test measures how an animal interprets an ambiguous cue, and it is one of the few available windows onto affective state rather than arousal (with the method and its limits set out separately).
Finding a shift in that measure after exposure to a stranger's stress odor would be a stronger claim than finding raised heart rate: it would indicate a change in how the dog was evaluating its situation, not merely that something registered. So far it rests on one location in one session with 18 dogs.
6.6 What Is Still Missing
The compounds involved have not been identified, the dose–response relationship is unknown, and the studies reviewed here do not establish how long the effect persists after exposure ends. The phenomenon is documented and the mechanism is a black box.
7. What Contagion Is Not
7.1 Not Empathy in the Full Sense
Emotional contagion is the basal layer of a proposed hierarchy: catching another's state, without necessarily understanding that the state belongs to someone else (Preston & de Waal, 2002).
Above it sit sympathetic concern — acting to relieve the other's state — and perspective-taking, which requires representing what another individual knows or feels as distinct from one's own position. The canine evidence sits at the first level, with the second contested and the third unsupported.
7.2 Not Understanding
A dog whose arousal rises alongside its owner's is not thereby aware that the owner is worried, or why. Contagion requires no model of the other's mind, which is precisely what makes it evolutionarily widespread and available to species that show no other evidence of mentalizing.
7.3 Not Evidence of Comfort-Giving
The step from "the dog approached" to "the dog was comforting" is the one this literature does not currently support, for the reasons set out in section 4.5.
That matters practically: owners describe dogs as knowing when they are upset and trying to help, and the first half is defensible while the second is an interpretation of behavior whose motivation has not been established.
7.4 Not a One-Way Street
The synchronization findings run in both directions by design. Long-term cortisol was synchronized between dogs and owners, owner personality related to the dogs' long-term cortisol while the dogs' own personality hardly did, and the authors read this as dogs mirroring their owners' stress (Sundman et al., 2019), and in ancient and solitary hunting breeds the synchronization itself was absent (Höglin et al., 2021) (behavioral synchronization between dogs and humans).
That asymmetry is worth stating precisely: it is a finding about direction of influence in one dataset, not a general law about who affects whom.
8. Dog to Dog
8.1 The Same Mechanism Within the Species
Almost everything in this article concerns transmission from humans to dogs, which reflects where the research attention has gone rather than where the phenomenon operates.
Contagion is a basal mechanism found across social species, and there is no reason to expect it to work between a dog and a person while failing between two dogs. Playback work points the same way: dogs hearing recorded whines of a familiar or an unfamiliar dog showed more alertness and stress-related behavior than with acoustically similar control sounds, offered more comfort behavior to their familiar partner afterwards, and familiar whines tended to keep cortisol higher (Quervel-Chaumette et al., 2016).
8.2 Why the Within-Species Case Is Practically Important
Multi-dog households, daycare groups, training classes and kennels place dogs in continuous proximity to other dogs' states. If contagion operates between them, then one anxious animal is an input to every other animal present.
That has obvious implications for group settings and it has barely been studied — most of the group-housing literature addresses space, noise and routine rather than affective transmission.
8.3 What This Means for Assessment
A dog assessed in a room with other stressed dogs is not being assessed under neutral conditions, which is one more reason single-occasion behavioral tests carry less information than the decisions made on them assume (illustrated by behavior tests for resource guarding).
8.4 What Would Need Testing
The within-species question has a clean design available: measure one dog's physiology while a familiar conspecific is exposed to a stressor out of sight, with a control for the stressor itself being detectable.
Something close to this has been done with playback of distress vocalizations (Quervel-Chaumette et al., 2016). What has not been done is the version using a real conspecific in real time, which is the situation every multi-dog household presents daily.
9. How It Is Measured
9.1 Four Instruments, Four Limitations
The evidence in this article comes from four kinds of measurement, and each answers a slightly different question.
Heart rate and variability capture autonomic arousal within seconds, and cannot distinguish positive from negative states. A dog whose heart rate rises during play and during fear produces the same reading.
Salivary cortisol responds with a delay to changes in the HPA axis and depends, among other things, on time of day, sampling and prior activity. It is not an unambiguous marker of a particular emotional state (with the neurochemistry set out separately).
Hair cortisol reflects a longer period, depending on growth and the hair segment sampled. Momentary reactions recede into the background; longer-lasting shared environmental influences remain possible confounds.
Behavioral coding captures what the animal did and requires an inference about why, which is the limitation running through this entire literature.
9.2 Why Convergence Matters Here
No single measure establishes contagion. What the case rests on is that different instruments with different weaknesses point the same way: cortisol responses to crying, hair cortisol synchronization over months, behavioral approach to distress, and judgment shifts after odor exposure.
Agreement between different kinds of findings increases the plausibility of an association but replaces neither replication nor a direct measurement of shared emotional experience. Alternative explanations remain, and 9.3 sets out the most important one.
9.3 The Question the Instruments Cannot Answer
None of them distinguishes catching a state from responding to a signal. A dog whose cortisol rises when it hears crying may be experiencing something like the human's distress, or may simply find the sound aversive.
The distinction is the whole content of the word "contagion", and no available canine measurement addresses it directly.
10. Research Gaps and Methodological Challenges
The findings are real but should be held with calibrated confidence.
Correlation versus causation. Cornerstone results such as the cortisol synchronization are correlational; they are compatible with such an influence but cannot alone prove its direction, and influence in both directions has not been excluded (the general difficulty of pinning down cause in behavior research).
Contagion versus recognition versus learning. A dog that acts stressed near a stressed owner might be catching the emotion, reading a learned cue, or responding to changed handling. Disentangling true state-matching from these alternatives is methodologically hard.
Individual differences. In one small study, the strength of coupling varied with ownership duration (Katayama et al., 2019); an effect of the bond itself and of temperament is plausible but has not been measured, and population findings do not map neatly onto every dyad (coping style shapes emotional reactivity), and breed and early experience add further variation (breed is a weak but non-zero factor).
What it is not. None of this shows that dogs experience human-style empathy or grasp why their person is upset. It shows responses and physiological associations that are compatible with emotional contagion, without clearly separating it from other explanations in every study.
10.1 Additional Gaps
The direction question is largely unaddressed. Most designs measure whether states co-occur rather than which one moves first, and the cortisol synchronization study is the exception rather than the rule.
Individual variation is barely modeled. Group means dominate, and there is reason to expect dogs differ substantially in how readily they take on a human state.
Positive states are under-studied. Nearly all the work concerns distress, fear and stress. Whether calm and positive affect transmit as readily has been examined far less, which leaves the practically useful half of the question open.
Sample sizes are small. The behavioral studies work with tens of dogs, and several key findings rest on a single experiment that has not been replicated independently.
11. Why the Ownership Duration Finding Is Interesting
11.1 The Result
In one small study, the dog–owner coupling in heart rate variability increased with the duration of ownership: dogs that had lived with their owners longer showed it more strongly (Katayama et al., 2019).
11.2 Three Readings
Learning. The dog has accumulated experience of what a particular person's states predict, which makes the signals more informative rather than more contagious in any basal sense.
Attachment. A stronger bond produces stronger state transmission, which would fit the affiliative framing and predict that contagion should be weaker with strangers.
Selection. Owner–dog pairs that synchronize poorly may be less likely to stay together, which would produce the correlation without any within-pair change at all.
11.3 Why the Third Reading Is Rarely Considered
Cross-sectional data comparing long-standing pairs with recent ones cannot distinguish a process that develops over time from a filter that removes the pairs where it does not.
The stranger findings complicate the attachment reading too: dogs responded to the odor of unfamiliar stressed people (Parr-Cortes et al., 2024) and approached crying strangers (Custance & Mayer, 2012), which shows that dogs can also respond to emotional cues from strangers. Whether they take on the same emotional state remains open, and so does whether a bond strengthens such responses.
12. Implications for Behavior and Training
12.1 The Handler's Emotional State Is Part of the Training Environment
If dogs partly synchronize with human stress, then the handler's emotional condition is not a neutral backdrop to training — it is an input. A tense, anxious, or frustrated human may, without intending to, send cues of possible danger. Whether this affects the dog's arousal or learning depends on the situation, prior experience and other factors. It is one possible contributor to why a dog's behavior differs between people and days; direct evidence for all everyday situations is lacking.
12.2 Regulation as a Two-Body Problem
The practical corollary is that emotional regulation in the human is a training skill, not a soft extra. Calm, predictable handling can make orientation easier; that it measurably lowers the dog's arousal in every situation is not shown by the studies cited here. This reframes difficult training moments: before asking why the dog is escalating, it is worth asking what the dog is catching — and a heightened state in a dog whose person is stressed may be less defiance than resonance (fear and stress can be transmitted as readily as calm).
12.3 What Handlers Can and Cannot Control
Advice to "stay calm" is common and close to useless as an instruction, because affective states are not directly volitional. What is controllable is narrower and more useful.
In practice: Breathing and posture are under voluntary control and are among the channels the dog reads. Leash handling is mechanical and trainable independently of how the handler feels. Situation selection — choosing not to walk past the thing at the time the handler is least able to manage it — removes the problem rather than requiring composure.
12.4 Why This Is Not a Blame Framework
The evidence supports adding handler state to the list of contributing variables. It does not support the inference that a dog's problem behavior originates in its owner's emotions, and that inference is made frequently in training contexts where it does real harm.
An owner told that their anxiety caused their dog's anxiety has been given something unfalsifiable to feel guilty about, and the guilt is itself a stressor in a system the evidence says is coupled.
12.5 What to Tell an Owner
Three statements the evidence supports, stated in the form a consultation can use.
Your dog can perceive cues about your state through several sensory channels (an owner's guide to how dogs read our emotions). In one studied group of owner–dog pairs, hair cortisol values were associated over longer periods; that is not a fixed pattern for every pair. And what your dog does when you are upset is a response to your state — whether it is an attempt to help you or to settle itself is not something anyone can currently tell you.
The last sentence is the one usually omitted, and leaving it out is what turns a finding into a story.
13. Where This Shows Up in Practice
13.1 The Veterinary Visit
An owner who is anxious about the appointment arrives with a dog that has been in a car with an anxious person for twenty minutes. The chemosignal and behavioral evidence both suggest this is not neutral for the dog, although neither was collected in that situation.
In practice: The practical consequence is that preparing the owner is part of preparing the dog, and it is the part usually left out (what your dog's reactions reveal about you).
13.2 The Reactive Dog on Leash
A handler who tenses at the sight of another dog produces a change the dog can detect through several channels at once — leash tension, posture, breathing, and possibly odor. Whether the dog's reaction follows from the handler's state or the handler's state follows from anticipating the dog's reaction is usually impossible to determine in the moment (with the reactivity picture set out separately).
Both directions are plausible and the loop is self-sustaining either way, which is why handler-focused work has a place alongside dog-focused work.
13.3 Households Under Strain
The long-term cortisol finding has an uncomfortable implication for behavior consultations. A household under sustained stress can be a changed environment for a dog as well. Whether its physiological values or behavior are related to that has to be checked in the individual case.
Naming it is more useful than ignoring it, and it may be one reason behavior problems sometimes resolve when circumstances change without any intervention.
13.4 What Not to Conclude
That the dog's behavior is the owner's fault. Contagion is a mechanism, not an attribution, and a dog with a genuine anxiety disorder does not have it because someone was worried.
The finding supports adding the handler's state to the list of variables worth considering. It does not support relocating the problem into the owner.
14. Summary at a Glance
Cortisol rose only to crying — Not to infant babbling and not to white noise, with a behavioral combination of submissiveness and alertness unique to the crying condition (Yong & Ruffman, 2014).
Dogs approached the crying stranger, not only the owner — Which argues against approach being a simple bid for reassurance (Custance & Mayer, 2012).
The authors call it empathic-like, not empathy — And name the available alternative: dogs may have been rewarded for approaching distressed humans (Custance & Mayer, 2012).
Long-term stress levels were synchronized — In herding breeds, with owner personality traits related to the dog's hair cortisol and the dog's own personality hardly at all (Sundman et al., 2019); not found in ancient and solitary hunting breeds (Höglin et al., 2021).
Stress odor may affect judgment — With the odor of unfamiliar stressed people, dogs in a cognitive bias test approached one ambiguous location less often (Parr-Cortes et al., 2024).
Duration of ownership matters — In one small study, the dog–owner coupling in heart rate variability increased with how long the dog had lived with the owner (Katayama et al., 2019).
Contagion is the basal layer, not the whole of empathy — Catching a state does not require understanding whose state it is (Preston & de Waal, 2002).
Whether approach is comforting or self-soothing is unresolved — Both accounts predict the same behavior, and the study separating them has not been run.
14.1 What Would Move the Field
Three studies would resolve most of what is currently assumed. A design measuring whether the dog's own arousal falls after contact with a distressed person, which would separate comfort-giving from self-soothing. A longitudinal design following new owner–dog pairs, which would separate development from selection in the duration finding. And work identifying the chemosignal compounds, which would turn a documented phenomenon into a mechanism.
None requires new technology. All three require follow-up over time, which is the scarcest resource in this literature.
14.2 A Note on the Popular Version
The claim that dogs sense emotions circulates in a stronger form than any of this supports — that they detect illness, grief or intention, and act on it deliberately.
What the evidence establishes is narrower and still remarkable: in one studied group, and not in all breeds, the long-term hair cortisol of dogs and their owners correlated, a stranger's stress odor may affect how a dog evaluates an ambiguous situation, and recorded crying produces a cortisol response that recorded babbling does not.
14.3 What This Does Not License
Two inferences the evidence does not support, both common.
That a dog showing concern-like behavior understands what its owner is going through. Contagion requires no such understanding, and the studies that document it explicitly do not claim it.
That a dog's emotional problems can be traced to its owner's emotional state. Synchronization describes a coupling between two systems, not a cause located in one of them.
15. Conclusion
Emotional contagion is one possible mechanism in the close relationship between humans and dogs; its share cannot be determined reliably from the evidence available so far. Across behavioral, autonomic, hormonal and olfactory channels, dogs respond to human emotional states in ways that go beyond recognizing them: cortisol rose to recorded infant crying but not to babbling or white noise (Yong & Ruffman, 2014), with the odor of stressed strangers, dogs approached one ambiguous location less often in a cognitive bias test (Parr-Cortes et al., 2024), and in herding breeds long-term cortisol was synchronized with the owner's (Sundman et al., 2019), although not in ancient and solitary hunting breeds (Höglin et al., 2021). Several of these effects did not require a bond: dogs responded to a stranger's stress odor and approached a crying stranger as well as a crying owner (Custance & Mayer, 2012). Three honest qualifications keep this in proportion: an observed association is not automatically evidence of shared experience or human-like empathy; much of the evidence is correlational; and whether a dog that approaches a distressed person is comforting that person or settling itself has not been resolved. Even so held, the conclusion is important and actionable: a dog's behavior and certain physiological measures can be related to human stress cues and living conditions. Recognizing that the human is inside the system — without turning that coupling into blame — is the foundation of a more honest and more welfare-oriented approach to training and to living with dogs.
Key Insights (Takeaways)
Emotional contagion is automatic emotional state-matching, a basic building block of empathy (Preston & de Waal, 2002) — not the higher-order, perspective-taking kind. Dogs "catching" a human's stress does not mean they understand or feel for the person.
In a test not specifically trained for this matching, dogs linked congruent emotional cues from faces and voices (Albuquerque et al., 2016). That shows perceptual ability, not yet a demonstrated adoption of the emotional state.
Synchronization is measurable in the body: owner–dog heart-rate coupling that, in one small study, increased with the duration of ownership (Katayama et al., 2019), and long-term hair-cortisol levels that correlated with the owners' while the dogs' own activity had no effect (Sundman et al., 2019), although not in ancient and solitary hunting breeds (Höglin et al., 2021). The cortisol finding is correlational and does not by itself establish the direction of influence.
Human stress reaches dogs even through smell: fear-odor chemosignals raise dogs' stress behaviors and heart rate (d'Aniello et al., 2018), and with the odor of stressed strangers, dogs approached one ambiguous location less often in a cognitive-bias test (Parr-Cortes et al., 2024), which may point to an effect beyond arousal.
Practically, the handler's emotional state is part of the training environment, because it is among the signals the dog responds to. A dog escalating alongside a stressed person may be resonating rather than disobeying — which makes the human's own emotional regulation a genuine training skill.
References
Albuquerque, N., Guo, K., Wilkinson, A., Savalli, C., Otta, E., & Mills, D. (2016). Dogs recognize dog and human emotions. Biology Letters, 12(1), 20150883. https://doi.org/10.1098/rsbl.2015.0883
Custance, D., & Mayer, J. (2012). Empathic-like responding by domestic dogs (Canis familiaris) to distress in humans: An exploratory study. Animal Cognition, 15(5), 851–859. https://doi.org/10.1007/s10071-012-0510-1
Höglin, A., Van Poucke, E., Katajamaa, R., Jensen, P., Theodorsson, E., & Roth, L. S. V. (2021). Long-term stress in dogs is related to the human–dog relationship and personality traits. Scientific Reports, 11, 8612. https://doi.org/10.1038/s41598-021-88201-y
Katayama, M., Kubo, T., Yamakawa, T., Fujiwara, K., Nomoto, K., Ikeda, K., Mogi, K., Nagasawa, M., & Kikusui, T. (2019). Emotional contagion from humans to dogs is facilitated by duration of ownership. Frontiers in Psychology, 10, 1678. https://doi.org/10.3389/fpsyg.2019.01678
Nagasawa, M., Mitsui, S., En, S., Ohtani, N., Ohta, M., Sakuma, Y., Onaka, T., Mogi, K., & Kikusui, T. (2015). Oxytocin-gaze positive loop and the coevolution of human–dog bonds. Science, 348(6232), 333–336. https://doi.org/10.1126/science.1261022
Parr-Cortes, Z., Müller, C. T., Talas, L., Mendl, M., Guest, C., & Rooney, N. J. (2024). The odour of an unfamiliar stressed or relaxed person affects dogs' responses to a cognitive bias test. Scientific Reports, 14, 15843. https://doi.org/10.1038/s41598-024-66147-1
Preston, S. D., & de Waal, F. B. M. (2002). Empathy: Its ultimate and proximate bases. Behavioral and Brain Sciences, 25(1), 1–20. https://doi.org/10.1017/S0140525X02000018
Quervel-Chaumette, M., Faerber, V., Faragó, T., Marshall-Pescini, S., & Range, F. (2016). Investigating empathy-like responding to conspecifics' distress in pet dogs. PLoS ONE, 11(4), e0152920. https://doi.org/10.1371/journal.pone.0152920
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
Yong, M. H., & Ruffman, T. (2014). Emotional contagion: Dogs and humans show a similar physiological response to human infant crying. Behavioural Processes, 108, 155–165. https://doi.org/10.1016/j.beproc.2014.10.006
d'Aniello, B., Semin, G. R., Alterisio, A., Aria, M., & Scandurra, A. (2018). Interspecies transmission of emotional information via chemosignals: From humans to dogs (Canis lupus familiaris). Animal Cognition, 21(1), 67–78. https://doi.org/10.1007/s10071-017-1139-x