The Fallout of Aversives: Neurobiological Consequences of Punishment in Dogs

1. Introduction

1.1 Punishment as a Physiological Event

The central reframe is simple: what looks like "correcting" a dog is, biologically, an aversive event, and aversive events are handled by the systems that detect threat and organize avoidance. How strongly those systems are engaged depends on intensity, predictability and the individual dog (see 5.5), but the dog's brain does not process an aversive event according to the trainer's intended meaning; it processes the consequences through the existing threat, learning and motivation systems. Once the threat systems are engaged, the animal does not stop learning — but, on the model from other species, what it learns can be less flexible, more strongly emotional and less tied to the intended lesson (behavior is the output of an emotional brain, not a moral choice).

1.2 How to Read the Evidence

Two layers must be kept apart. The neurobiological cascade below — amygdala threat detection, cortisol effects on the hippocampus, prefrontal suppression, motivational shutdown — rests on decades of rodent and human research and is extended to dogs because these systems are evolutionarily conserved. The canine-specific evidence, by contrast, sits mostly at the level of physiology and behavior: measured cortisol, stress signals, cognitive bias, and welfare outcomes. This article names which is which, so that the strength of "aversives harm dog welfare" (well evidenced) is not confused with the precision of any particular claim about the dog hippocampus (inferred).

Its focus is what aversive experience is expected to do in the organism and which of those effects have actually been measured in dogs. The training studies, including the efficacy comparisons and the owner surveys, are treated in more depth in the companion article on aversive training methods; the canine findings summarized here are the ones the physiological account has to explain.

1.3 What Counts as an Aversive Here

The term covers a wide range: physical corrections, electronic collars, choke and prong collars, startling noises, spray devices, verbal intimidation, and physical manipulation such as pinning or rolling. What unites them is not intensity but function — something the animal will work to avoid.

That functional definition matters because it is the dog rather than the handler who determines membership. A tone of voice one dog ignores is aversive to another, and a device marketed as gentle is aversive if the dog is avoiding it.

One distinction belongs here as well. In learning theory, punishment also includes negative punishment: removing something the dog wants, such as ending a game or withholding an expected reward. It works by taking something away rather than by adding something the dog avoids, and although the loss can be frustrating, it is not what the studies in this article examined. Everything below concerns aversive stimulation — positive punishment and the other methods listed above.

1.4 Why "It Depends on How You Use It" Is Not an Answer

The most common defense of these methods is that the problem lies in misapplication rather than in the method. That claim is testable, and the controlled trials described later are what testing it looks like: trained operators, calibrated equipment, supervised protocols.

Where effects persist under those conditions, the misapplication argument is weakened on its own terms, although that does not settle the question for every method and setting.

It is worth conceding one point plainly: punishment can change behavior, and in the short term it can reliably stop a behavior. The decisive question is not whether it works but what was learned and at what cost. A behavior that has stopped is not automatically a behavior that has been resolved.

2. What Happens When Owners Try It

2.1 A Different Kind of Evidence

The controlled trials discussed elsewhere in this article measure welfare indicators under supervised conditions. A separate line of work asks something blunter: what happened when ordinary owners applied these methods to their own dogs before seeking professional help.

A thirty-item survey was distributed to all owners making appointments at a university referral behavior service over one year, and 140 surveys were completed, covering which interventions they had used, where the recommendation came from, and how the dog responded (Herron, Shofer & Reisner, 2009).

2.2 The Numbers

Several confrontational methods elicited an aggressive response from at least a quarter of the dogs on which they were attempted: hitting or kicking the dog for undesirable behavior (43%), growling at the dog (41%), physically forcing the release of an item from the dog's mouth (39%), the alpha roll (31%), staring the dog down (30%), the dominance down (29%), and grabbing the dog by the jowls and shaking (26%) (Herron et al., 2009).

Reward-based training elicited aggression in very few dogs, regardless of the presenting complaint (Herron et al., 2009).

The comparison is what gives the numbers their weight. Both sets of methods were applied by the same population of owners to the same population of dogs, so the difference between them is not explained by the households or by the animals.

2.3 Where the Risk Concentrates

Dogs presenting for aggression toward familiar people were more likely than dogs with other complaints to respond aggressively to the alpha roll and to being yelled at (Herron et al., 2009).

That is the group these techniques are most often recommended for, which makes the finding the most consequential in the paper. In this sample, the dogs already presenting for aggression toward familiar people were the ones most likely to respond aggressively to these two techniques (the broader literature on aggression toward familiar people). Advice to confront an aggressive dog is therefore likely to be particularly risky in exactly the group it is aimed at.

2.4 Where the Advice Came From

The most frequently listed sources of the recommendations were the owners themselves and dog trainers (Herron et al., 2009). Neither is surprising and both are worth stating, because it locates the problem outside the veterinary consultation where the survey was collected. By the time a dog reaches a behavior referral, several methods have usually been tried on the advice of people with no training in behavior, and the dog's history includes all of them.

2.5 What This Design Cannot Establish

This is a referral population: owners who sought specialist help for a behavior problem, which is not a sample of dogs generally. The reports are retrospective and owner-recalled, the interventions were not standardized, and nobody observed them.

Nor does the design establish direction. Owners of dogs that already behave aggressively may reach for confrontation more readily, so an association between the two is compatible with the technique causing the response and with the response prompting the technique (what a correlational design can and cannot establish).

2.6 Why It Matters Anyway

Set alongside the controlled trials, the survey answers a question they cannot. The trials show what happens under supervision with trained handlers and calibrated equipment. This shows what happens in households.

The percentages are not effect sizes, and they come from a referral population, but they give an indication of how often these methods are associated with an aggressive response in the hands of the people most likely to use them. For a consultation, that is more useful than any mechanism: a technique associated with an aggressive response in roughly a third of the dogs it was tried on is a safety issue before it is a welfare one.

3. The Immediate Response: Amygdala and Midbrain

3.1 The Amygdala as a Rapid Threat Detector

The amygdala is the brain's hub for emotionally salient, especially threatening, stimuli, receiving sensory input by both a slow cortical route and a fast subcortical one that enables reflexive responses before conscious evaluation (LeDoux, 2000). This is superb design for surviving predators and useless for distinguishing a genuine danger from a trainer's imposed aversive — the threat system does not sort them by the handler's intention. On activation it drives the sympathetic release of adrenaline and noradrenaline, stimulates the HPA axis toward cortisol secretion, and enhances the consolidation of fear-based memories (the durable machinery of conditioned fear).

3.2 What the Dog Actually Learns

On the model from other species, none of this is reflective learning. Threat systems encode predictive relationships between stimuli and danger, so a dog is unlikely to process punishment as "correction" and more likely to form associations such as human proximity predicts danger, this place predicts danger, or my behavior produces unpredictable bad outcomes. What looks from the outside like learning can instead be the formation of threat-related associations — and the difference matters enormously for what the dog becomes (how threat processing differs from cognitive learning).

3.3 The Periaqueductal Gray and Defensive States

The periaqueductal gray (PAG) integrates amygdala signals and organizes species-typical defense: freezing, flight, or fight. In the model from other species, its activation is not a choice but a largely hardwired output, and once strong aversive stimulation triggers it, the animal is operating in a defensive state rather than a learning one. That may help explain why punished dogs can suddenly show aggression, bolt, or shut down — responses better understood as defensive than as defiance or confusion.

4. The Hippocampus: Contextual Learning Under Stress

4.1 Cortisol and Hippocampal Function

The hippocampus contextualizes experience — encoding where, when, and under what conditions events occurred — which is what lets an animal tell safe from unsafe. It is also dense with glucocorticoid receptors, and chronic cortisol elevation is associated, in rodents and humans, with reduced neurogenesis, dendritic retraction, and impaired plasticity (McEwen, 2007). Two honest caveats: the strongest of these claims (frank neuronal loss) is less certain even in those species, and none of it has been demonstrated in the dog hippocampus — it is a well-grounded expectation, not a canine measurement (the fuller cortisol–hippocampus story and its limits in dogs).

4.2 Loss of Contextual Precision

On this model, one behavioral consequence of stress-affected hippocampal function would be a loss of specificity: fear generalizes more widely. A dog punished in one situation may not learn "this behavior is unwanted here" but instead acquire a broad fear spanning similar environments, related stimuli, and social interactions. This spreading of fear beyond its original context is a signature of stress-affected contextual processing.

4.3 Impaired Extinction

Extinction — the updating and suppression of a previously learned fear response — depends on coordinated activity in brain systems including the hippocampus and prefrontal cortex (Maren, 2011), and in other species stress has been reported to weaken extinction. That may contribute to why some dogs exposed to aversive training do not simply "recover" once the punishment stops: the systems that would revise the threat associations may themselves be affected (and suppressed responses tend to return), leaving behavioral flexibility reduced.

4.4 Why Poor Contextual Encoding Shows Up as Unpredictability

If stress degrades the precision with which context is encoded, then a fear learned in one setting should generalize more widely than it otherwise would. That is what owners describe when a dog that had one bad experience becomes wary of a whole category of situations.

The account is mechanistic and drawn from other species. What can be observed in dogs is the spread itself: police dogs with recent shock-collar experience showed lower ear posture and more stress-related behaviors than control dogs even in training sessions without shocks and during walks in a park (Schilder & van der Borg, 2004), which is the reason the term fallout is used at all. A household that describes a dog as having become unpredictable is frequently describing a fear that has generalized rather than one that has intensified.

5. Sensitization, HPA Dysregulation, and the Prefrontal Cortex

5.1 Amygdala Sensitization

In other species, repeated activation of fear circuitry can make it more efficient, not less: thresholds drop, responses speed up, and fear generalizes more readily. Over time the animal becomes hyper-responsive, reacting to stimuli that once seemed neutral — and this is elaborated in the companion account of the neurological effects of aversive methods.

5.2 HPA Dysregulation and Hypervigilance

In other species, persistent threat signaling can keep the HPA axis chronically engaged, which is associated with elevated baseline cortisol, prolonged stress responses, and slower physiological recovery — a state expressed behaviorally as hypervigilance, a dog that never fully stands down (arousal regulation as a core casualty).

5.3 Prefrontal Cortex Impairment

The prefrontal cortex supports impulse control, decision-making, and emotional regulation, and acute stress signaling degrades its function (Arnsten, 2009). In those species the result is described as a functional imbalance — an overactive, bottom-up amygdala and an under-functioning, top-down prefrontal cortex — in which behavior is governed more by emotional reactivity than by deliberation; in dogs this has not been measured (see 10.5) (the prefrontal basis of canine self-control). A dog in a strongly stressed state may not be "choosing" to ignore known cues; its capacity to comply may be reduced.

5.4 Timing Is Where Most of It Goes Wrong

For an aversive to suppress a specific behavior, it has to arrive close enough in time for the animal to associate the two. In practice it frequently arrives after the behavior, after the dog has turned toward the handler, or after something else has intervened.

What gets associated is then whatever was salient at that moment, which is often the handler, the location or the other dog rather than the behavior anyone intended to punish. This is the mechanism behind the most common complaint about aversive methods in practice — that the dog became worse around the very thing the handler was trying to make it comfortable with.

5.5 The Intensity Problem

An aversive strong enough to suppress reliably is often strong enough to produce the effects described in this article, and one mild enough to avoid them may fail to suppress, which can lead to escalation. The electronic collar data illustrate the gradient: high-intensity stimulation without warning cues produced marked behavioral responses and elevated cortisol in a preliminary study, whereas lower settings with a pre-warning function produced less marked responses that were still more tense than in reward-based training (Cooper et al., 2014).

This is not an incidental difficulty of application. It is structural, and it is why the question of correct intensity has no stable answer. The window between too mild to work and strong enough to cause the effects described here is narrow, it differs between dogs, and it moves within the same dog depending on arousal, pain and what has already happened that day.

6. Motivation and Learned Helplessness

6.1 When the Animal Stops Trying

Uncontrollable aversive experience also reaches the motivational systems, and the classic result is learned helplessness — originally described in dogs — in which an animal exposed to inescapable, uncontrollable aversives eventually stops trying to respond at all (the neurobiology of learned helplessness). The modern reformulation is important and often missed: reviewing fifty years of neuroscience, the theory's own authors concluded that passivity and shutdown are the default mammalian response to prolonged uncontrollable stress, and that what is actually learned — when the animal detects its actions have effects — is control (Maier & Seligman, 2016). Either way, the outward picture is a quiet, compliant, "calm" dog; what can be observed is reduced engagement and initiative, which is not successful learning, and 6.2 sets out why the helplessness label itself remains an inference (and the outward stillness does not reveal the inner state). Reading that shutdown as obedience is one of the most consequential errors in training.

6.2 Helplessness Is a Strong Term

Learned helplessness names a specific experimental phenomenon produced under conditions of inescapable aversive stimulation, and its modern interpretation has been substantially revised (Maier & Seligman, 2016). Applying the label to a dog that has gone quiet in training is an inference, not an observation.

What is observable is a reduction in offered behavior, in exploration and in engagement. That is worth taking seriously in its own right without importing a laboratory construct to explain it (how affective state is measured in dogs). A dog that has stopped offering behavior is harder to train by any method, which is a practical cost quite apart from what it says about the animal's state.

6.3 Why a Quiet Dog Is Not Evidence of Success

The most consequential misreading in this whole area is that a dog which has stopped doing something has learned not to do it. Suppression and learning are different outcomes, and they look identical from the outside on the day.

The distinction becomes visible later — in a new context, under arousal, or when the aversive is absent — which is exactly when a household concludes that the training stopped working. It rarely stopped working. Often it was not doing what the household believed it was doing in the first place.

7. Possible Long-Term Consequences of Chronic Stress

7.1 Cellular Aging

Chronic stress promotes oxidative stress and inflammation, which are associated with accelerated telomere shortening; in a study of 250 dogs, those living in stressful or less enriching conditions, such as laboratory settings, showed shorter telomeres (Dutra et al., 2025), raising the possibility that aversive-driven chronic stress contributes to faster biological aging (the telomere–stress link in dogs, held cautiously). This remains an association, not a demonstrated causal chain from punishment to cellular aging.

7.2 Immune and Gastrointestinal Effects

Chronic stress dysregulates multiple systems, with plausible effects on immune function (suppression or overactivation), disease susceptibility, and gastrointestinal stability — consequences documented broadly in stress physiology and expected, by extension, in chronically stressed dogs.

7.3 Gut–Brain Axis

The gut microbiome participates in emotional regulation, and chronic stress can shift microbial composition in ways that feed back on neurotransmission and behavior (the gut–brain axis in dogs) — another route by which sustained aversive stress may extend beyond the moment.

7.4 Social Bonding

Threat-related associations formed in training may alter how the dog responds to the handler — more avoidance, more hesitation to approach, less willingness to offer behavior. Whether that amounts to damage to the attachment bond itself is considerably harder to measure than the behavioral change, and no canine study has demonstrated it.

7.5 Emotional Contagion

Finally, the handler's own state is part of the loop. Dogs respond to human emotional signals, and a stressed or angry handler may amplify the dog's fear (stress transmits between human and dog), so aversive handling can create a self-reinforcing cycle of mutual dysregulation.

7.6 How Firm the Systemic Claims Are

The chapter above sets out effects on cellular aging, immune function, gastrointestinal state, social bonding and emotional contagion. These vary considerably in how well established they are, and running them together as a list makes them look equivalent.

Chronic stress physiology in dogs is documented (Beerda et al., 1999). The telomere connection is a small and internally contested canine literature. The gut–brain and immune links are largely extrapolated. None of these is the reason to avoid aversive methods, and presenting them as though they were weakens an argument that does not need them. A reader who checks one of the weaker links and finds it thin has reason to discount the whole chapter, including the parts that are solid.

8. What the Dog Evidence Actually Shows

8.1 The Canine Trials

The mechanism above is a model. What grounds it empirically is a body of canine research measuring the fallout directly — and this is where confidence should be highest. In a comparison of companion dogs from training schools using different methods, dogs trained with aversive methods showed more stress-related behaviors, larger post-training cortisol rises, and a more "pessimistic" cognitive bias outside training than reward-trained dogs (Vieira de Castro et al., 2020). In an electronic-collar field study, a preliminary phase with high-intensity stimulation produced negative behavioral changes and elevated cortisol after stimulation; in the main study of 63 dogs with lower settings and a pre-warning function, e-collar dogs spent more time tense, yawned more and explored less than dogs trained by reward-based trainers, while urinary and salivary cortisol did not differ significantly (Cooper et al., 2014). Laboratory work found that the physiological response depended on predictability: dogs shocked when touching a prey dummy showed no significant rise in salivary cortisol, whereas dogs shocked for not responding to a recall command or at random did (Schalke et al., 2007). Six weeks of social and spatial restriction in beagles produced behavioral signs of chronic stress, including low posture, autogrooming, paw lifting, repetitive behavior and coprophagy (Beerda et al., 1999). And crucially for the "but it works" defense, in 63 dogs with poor recall, reward-focused training achieved better responses to recall and sit than e-collar training, and the e-collar did not create a greater deterrent for disobedience (China et al., 2020). That study has been criticized for semi-random group assignment and for training the reward group at a different time and place (Sargisson & McLean, 2021), and its design and sample size match the Cooper study, so the two analyses should not be counted as independent trials. The observable fallout in dogs, in other words, is not hypothetical: it is measured, and with those caveats it points consistently in one direction.

8.2 What the Controlled Trials Can and Cannot Settle

The electronic collar trials are the strongest evidence in this article and they were run under conditions that favor the method: trained operators, calibrated devices, manufacturer-nominated or industry-approved trainers, supervised protocols throughout (Cooper et al., 2014; China et al., 2020).

That matters for interpretation. Effects observed under best-case application are plausibly a lower bound on what happens in ordinary use, not an average. The survey data described earlier are the other end of the same range: the same category of method, applied by owners without supervision, with outcomes recorded after the fact.

8.3 Efficacy and Welfare Are Separate Questions

A method can work and still cost more than it is worth, and the two questions get merged constantly in this debate. The canine trials addressed both, which is what makes them useful.

Reading them only for the welfare result, or only for the efficacy result, misses the point of running them together. A method that is neither more effective nor welfare-neutral has no remaining argument in its favor, and that is the shape of what the electronic collar analyses found for recall training — with the qualification that the welfare and efficacy results come from one design and sample, and that the efficacy analysis has been criticized for its assignment procedure (see 8.1).

8.4 Effects Beyond the Moment of Stimulation

The most direct canine evidence that aversive training affects a dog beyond the moment of stimulation comes from guard dog training of German shepherds. The immediate reactions of 32 dogs to 107 shocks included lowered body posture, high-pitched yelps, avoidance, redirected aggression and tongue flicking, most lasting only a fraction of a second (Schilder & van der Borg, 2004).

The authors then compared 16 dogs that had received shocks in the recent past with 15 control dogs that had received similar training without shocks, using only sessions in which no shocks were delivered. The shocked dogs showed a lower ear posture and more stress-related behaviors during free walking on the training grounds, during obedience training and during bite work, and similar differences appeared during free walking and obedience exercises in a park, even though the control dogs had also been trained in a fairly harsh way (Schilder & van der Borg, 2004). The groups were not randomly assigned, so the comparison is observational.

8.5 What the Reviews Conclude

A review of 17 studies, including surveys, observational studies and interventions, concluded that aversive training methods can jeopardize both the physical and mental health of dogs, and that although positive punishment can be effective, there is no evidence that it is more effective than training based on positive reinforcement (Ziv, 2017). The same review named small samples, missing effect sizes and possible bias in behavioral coding as methodological concerns.

A second review reached a more cautious conclusion: aversive methods are correlated with indicators of compromised welfare, such as stress-related behaviors during training, elevated cortisol and fear or aggression, but limitations, including heavy reliance on surveys rather than objective measures, prevent strong conclusions (Guilherme Fernandes et al., 2017). Both readings point in the same direction and differ in how firmly they state it.

9. Implications for Training and Behavioral Practice

9.1 Reward-Based Learning

On the model from other species, positive reinforcement engages dopaminergic reward and motivation circuitry (the neurochemistry of reward-based learning), building behavior without recruiting the threat systems described above. Reward-based methods allow learning without the risks that come with aversive consequences, and they work with the dog's cognition instead of against it (what dogs are genuinely good at).

9.2 Control and Predictability

In animal research, control over outcomes reduces the impact of stressors (Maier & Seligman, 2016), and the canine laboratory data point the same way for predictability (Schalke et al., 2007); methods that give a dog choice and predictability are therefore likely to reduce stress (a question tied to how dogs monitor and act on information).

9.3 Recognizing Stress Signals

Displacement behaviors, avoidance, muscle tension, and shutdown all mark a compromised learning state and should redirect the trainer, not be pushed through. Visible stress means the session has already left the zone where learning happens.

9.4 The Pain Connection

Undiagnosed pain, such as from osteoarthritis, lowers a dog's threshold for defensive aggression, so a painful dog is more likely to perceive threat and react — and punishing that reaction is especially damaging, adding an aversive to an animal already in distress (the hidden link between pain and aggression).

9.5 What to Say to Someone Who Reports It Worked

Owners who have used these methods often report success, and they are usually reporting accurately. The behavior did stop. Disputing that observation is both unnecessary and counterproductive.

The useful response addresses what was not observed: whether the behavior returns in a new setting, what the dog now does when the handler approaches, and what has happened to the behaviors the dog used to offer freely. Those three questions are answerable by the owner and they change the conversation from a disagreement about method into a description of the dog.

9.6 The Escalation Trap

Where an aversive stops working, the available responses are to increase it or to change approach, and increasing it is the path of least resistance because it worked before. That trajectory is visible in most case histories where a household has arrived at a serious incident. The first application is rarely what causes the harm; the eighth one, at an intensity nobody would have started with, usually is.

Naming it early — that a method requiring steady escalation is telling you something — is more useful than arguing about the first application.

9.7 What Replaces It

The alternative is not permissiveness. It is management that removes the opportunity, reinforcement of an incompatible behavior, and graduated exposure below the threshold at which the dog stops coping.

That is more work in the first weeks and less work in the second year, which is a hard argument to make to a household in the middle of a problem and the honest one.

9.8 Where the Handler's Own State Enters

Confrontational methods are most often reached for at the moment a handler is frightened, embarrassed or out of patience, which is also the moment least suited to calibrating anything. The survey data do not test this directly, although the interventions with the highest aggressive-response rates are ones that require little deliberation (Herron et al., 2009).

A plan that depends on the handler being calm at the worst moment of the week is not a plan. Management that removes the situation is more reliable than a technique that has to be applied correctly under pressure. This is also why advice given in a calm consultation frequently fails in the household: the plan assumed a version of the handler that does not exist at the moment it is needed.

9.9 Fearful and Insecure Dogs

Not every dog becomes fearful through punishment. In a fearful or insecure dog, however, and especially where the problem involves fear or aggression, punishment often intensifies exactly the emotional state that drives the problem. That is a practical judgment rather than a measured frequency, but it follows the direction of the canine findings in this article: dogs presenting for aggression toward familiar people were more likely to respond aggressively to confrontation (Herron et al., 2009), dogs with shock-collar experience showed more stress behavior even without stimulation (Schilder & van der Borg, 2004), and aversively trained dogs judged ambiguous cues more pessimistically (Vieira de Castro et al., 2020).

A dog whose problem arises from fear needs safety in order to learn differently. A consequence that adds threat works against exactly that (how anxiety in dogs develops and is maintained).

10. Which Findings Come From Which Species

10.1 An Unusually Favorable Split

This article is better placed than most in this series, because its central practical claim rests on canine evidence rather than on extrapolation. That is worth stating explicitly, since the neurobiology chapters can make it look otherwise.

10.2 What Was Measured in Dogs

Owner-reported aggressive responses to confrontational methods have been surveyed (Herron et al., 2009). Welfare consequences and efficacy of electronic collar training have been tested with assigned groups (Cooper et al., 2014), and remote electronic collars have been compared against reward-based approaches for recall and general obedience (China, Mills & Cooper, 2020). Physiological responses to electric collars have been documented under laboratory conditions (Schalke et al., 2007), and longer-lasting behavioral differences have been observed in shock-collar-trained police dogs (Schilder & van der Borg, 2004). Training method has been related to welfare outcomes in companion dogs (Vieira de Castro et al., 2020), and chronic stress indicators have been characterized in this species (Beerda et al., 1999).

That is seven canine sources, with two reviews summarizing the field (Ziv, 2017; Guilherme Fernandes et al., 2017), although two of the electronic collar analyses appear to share one design and sample. Few topics in canine behavior have this much species-specific evidence.

10.3 What Was Established Elsewhere

The amygdala account of threat processing is human and rodent (LeDoux, 2000). Allostatic load and the brain's role in stress adaptation is likewise (McEwen, 2007), as is the inverted-U relationship between stress and performance and individual variability in it (Sapolsky, 2015), the prefrontal impairment account (Arnsten, 2009), the extinction and fear-memory work (Maren, 2011) and learned helplessness (Maier & Seligman, 2016).

Six sources, none of them canine. They explain why the canine findings look the way they do; they are not themselves canine findings.

That is not a criticism of their inclusion. A mechanism is what turns a set of associations into an account, and no canine equivalent exists for any of them. The requirement is only that a reader can tell which column a given statement came from.

10.4 Why the Argument Should Lead With the Dogs

An argument that aversive methods are harmful because of what they do to the amygdala borrows authority from work in other species. An argument that they are associated with poorer welfare indicators, higher stress signaling and aggressive responses rests on measurements taken in dogs.

The second is both stronger and harder to dismiss, and it does not depend on any neural claim surviving revision. It is also the argument that survives a hostile audience, which the first one does not: a mechanism from another species can always be waved away, a controlled study in dogs much less easily.

10.5 What Is Still Missing

No canine study has imaged amygdala activity during or after aversive training, measured hippocampal change, or demonstrated impaired extinction in dogs trained aversively. The mechanism chapters describe a coherent account of the canine observations rather than a set of canine measurements.

That gap is unlikely to close soon, and it does not weaken the practical conclusion, which was never resting on it. It does mean that confident neural language in training material — claims about what an aversive does to a dog's amygdala — is running ahead of what anyone has measured in this species, in a direction that happens to be right.

11. Summary at a Glance

Confrontational methods were frequently followed by aggression — Hitting or kicking (43%), growling at the dog (41%), forcing an item from the mouth (39%) and the alpha roll (31%) each elicited an aggressive response from at least a quarter of dogs on which they were attempted (Herron et al., 2009).

Reward-based methods rarely were — In the same survey, reward-based training elicited aggression in very few dogs regardless of the presenting complaint (Herron et al., 2009).

The risk is highest where the methods are most recommended — Dogs presenting for aggression toward familiar people were more likely to respond aggressively to the alpha roll and to being yelled at (Herron et al., 2009).

Controlled canine trials exist — Welfare consequences and efficacy of electronic collar training have been tested directly in dogs (Cooper et al., 2014; China et al., 2020; Schalke et al., 2007), although two of these analyses appear to share one design and sample.

Effects reach beyond the moment of stimulation — Police dogs with recent shock-collar experience showed lower ear posture and more stress-related behaviors than controls even in sessions without shocks and in a park (Schilder & van der Borg, 2004).

Training method is associated with welfare outcomes — Companion dogs trained with aversive methods showed poorer welfare indicators than those trained with reward-based methods (Vieira de Castro et al., 2020).

Suppression is not resolution — Punishment can stop a behavior, but a behavior that stops under punishment has not automatically been unlearned, and the underlying emotional state may be unchanged or worse.

The neurobiology is borrowed — Amygdala, hippocampal, prefrontal and learned-helplessness accounts come from human and rodent work, not from dogs.

The practical conclusion does not depend on it — The canine welfare and behavioral evidence stands on its own.

12. Research Gaps and Critical Appraisal

The strength of this account varies by claim and should be read accordingly.

Mechanism is cross-species. The amygdala, hippocampal, prefrontal, and motivational mechanisms are established in rodents and humans and applied to dogs by extension; they have not been imaged in the dog brain.

Dog evidence is behavioral and physiological. The solid canine findings concern cortisol, stress behavior, cognitive bias, welfare, and efficacy (Vieira de Castro et al., 2020; Cooper et al., 2014; Schalke et al., 2007; China et al., 2020) — not neural structure. The neurobiology explains these findings; it is not itself measured in dogs.

Some mechanistic claims are too strong. "Neuronal loss," a total "collapse" of motivation, or a straight line from punishment to accelerated aging overstate what is established; the better-supported statements are dendritic and neurogenic changes, motivational suppression, and stress–telomere associations.

Confounds in field studies. Some welfare comparisons are observational: the police-dog comparison did not assign groups (Schilder & van der Borg, 2004), and the companion-dog study compared dogs from training schools that already used different methods rather than assigning dogs to methods (Vieira de Castro et al., 2020). The strongest weight therefore falls on studies with assigned groups and physiological measurement, and those carry limits of their own, set out below.

The electronic collar trials are not independent. The efficacy analysis and the welfare analysis appear to share one design and sample of 63 dogs, and the efficacy analysis has been criticized for semi-random assignment and differing training conditions (Cooper et al., 2014; China et al., 2020; Sargisson & McLean, 2021).

Reviews differ in firmness. One review concludes that aversive methods can jeopardize dogs' health and are not more effective (Ziv, 2017); another finds the same direction but judges the evidence too limited for strong conclusions (Guilherme Fernandes et al., 2017).

Stated honestly, none of this removes the practical conclusion; it sharpens what the conclusion rests on. The mechanism is a coherent model from other species, and the measured canine findings point in one direction, with the two reviews differing in how firmly they state it.

The survey evidence is retrospective and from a referral population. Owner-recalled reports of unstandardized, unobserved interventions in dogs already presenting for behavior problems (Herron et al., 2009) cannot establish causal direction or a general base rate.

The controlled trials were run under favorable conditions. Trained operators and calibrated equipment (Cooper et al., 2014; China et al., 2020) plausibly place a lower bound on the effects of ordinary use rather than describing it.

No canine imaging supports the mechanism. Amygdala sensitization, hippocampal change and impaired extinction after aversive training are inferred from other species and have not been measured in dogs.

Long-term follow-up is largely absent. The claim that suppressed behavior returns, or returns worse, follows from learning theory and is thinly documented in longitudinal canine data.

13. Conclusion

Aversive methods can have consequences that reach well beyond the behavior they target. Yes, punishment can stop a behavior. But a stopped behavior is not automatically a resolved one, and if the price is fear, conflict, avoidance or a deteriorating relationship between dog and human, that is not a good training solution, especially when welfare-friendly alternatives exist. The most direct evidence comes from dogs. In a referral population, confrontational techniques applied by owners were frequently followed by an aggressive response, most of all in dogs already presenting for aggression toward familiar people (Herron et al., 2009). Police dogs with recent shock-collar experience showed more stress-related behavior even in sessions without shocks (Schilder & van der Borg, 2004). Companion dogs from schools using aversive methods showed more stress behavior, larger cortisol rises and a more pessimistic cognitive bias (Vieira de Castro et al., 2020), and in assigned groups electronic collar training brought more tense behavior and no efficacy advantage for recall (Cooper et al., 2014; China et al., 2020) — two analyses of one design, the second criticized for its assignment procedure. Two reviews agree on the direction and differ in how firmly they state it (Ziv, 2017; Guilherme Fernandes et al., 2017). Behind these findings sits a mechanistic account drawn from rodent and human neuroscience: repeated threat and stress are expected to sensitize fear circuits, blur contextual learning, weaken executive control and suppress motivation. None of that has been measured in the dog brain, and the practical conclusion does not depend on it. What appears as compliance can mask a dog that has stopped offering behavior, and a still, "obedient" dog may be a shut-down one. A scientifically grounded approach prioritizes methods that allow learning without the risks of aversive consequences over those that buy short-term suppression at a cost that is often delayed and easy to miss.

Key Insights (Takeaways)

• An aversive is, biologically, handled by threat and avoidance systems: in the model from other species it engages the amygdala and midbrain defensive circuits (LeDoux, 2000), not a "correction" module, with effects that scale with intensity and predictability. What looks like learning can instead be a threat-related association — the dog may encode this person / place / situation predicts danger rather than the intended lesson.

• The neurobiological cascade — cortisol effects on the hippocampus (McEwen, 2007), extinction (Maren, 2011), prefrontal impairment (Arnsten, 2009), and learned helplessness (Maier & Seligman, 2016) — is drawn from rodents and humans and applied to dogs by extension. It has not been measured in the dog brain.

• What is measured in dogs is the fallout the mechanism predicts: more stress behaviors and a more pessimistic cognitive bias with aversive methods (Vieira de Castro et al., 2020), more tense behavior with e-collars (Cooper et al., 2014), stress responses that depend on predictability (Schalke et al., 2007), and stress signs in shock-collar-trained police dogs even outside the moment of stimulation (Schilder & van der Borg, 2004). A shutdown can be mistaken for calm obedience.

• The efficacy defense does not hold up for recall training on current evidence: reward-focused training achieved better responses than e-collar training (China et al., 2020), although that analysis shares its design with the welfare study and has been criticized (Sargisson & McLean, 2021), and a review found no evidence that punishment is more effective than reward-based training (Ziv, 2017). A later study found electronic collars more effective at stopping lure chasing. In owners' hands, confrontational methods were frequently followed by an aggressive response in a referral population, a design that cannot establish direction (Herron et al., 2009).

• Punishment can stop behavior; the decisive question is what was learned and at what cost. In fearful or insecure dogs, punishment often intensifies the very state that drives the problem. Practically, favor reward-based learning, build in control and predictability, read stress signals as stop signs, and rule out pain before treating "defiance" as a training problem. Where fallout accumulates, it is easy to miss until it is severe; the long-term canine follow-up that would show how often that happens is largely absent.

References

Arnsten, A. F. T. (2009). Stress signalling pathways that impair prefrontal cortex structure and function. Nature Reviews Neuroscience, 10(6), 410–422. https://doi.org/10.1038/nrn2648

Beerda, B., Schilder, M. B. H., van Hooff, J. A. R. A. M., de Vries, H. W., & Mol, J. A. (1999). Chronic stress in dogs subjected to social and spatial restriction. I. Behavioral responses. Physiology & Behavior, 66(2), 233–242. https://doi.org/10.1016/S0031-9384(98)00289-3

China, L., Mills, D. S., & Cooper, J. J. (2020). Efficacy of dog training with and without remote electronic collars vs. a focus on positive reinforcement. Frontiers in Veterinary Science, 7, 508. https://doi.org/10.3389/fvets.2020.00508

Cooper, J. J., Cracknell, N., Hardiman, J., Wright, H., & Mills, D. (2014). The welfare consequences and efficacy of training pet dogs with remote electronic training collars in comparison to reward based training. PLoS ONE, 9(9), e102722. https://doi.org/10.1371/journal.pone.0102722

Dutra, L. M. L., Souza, F. S., Vasconcellos, A. S., Young, R. J., & Schork, I. G. (2025). Telomere Tales: Exploring the impact of stress, sociality, and exercise on dogs' cellular aging. Veterinary Sciences, 12(5), 491. https://doi.org/10.3390/vetsci12050491

Guilherme Fernandes, J., Olsson, I. A. S., & Vieira de Castro, A. C. (2017). Do aversive-based training methods actually compromise dog welfare?: A literature review. Applied Animal Behaviour Science, 196, 1–12. https://doi.org/10.1016/j.applanim.2017.07.001

Herron, M. E., Shofer, F. S., & Reisner, I. R. (2009). Survey of the use and outcome of confrontational and non-confrontational training methods in client-owned dogs showing undesired behaviors. Applied Animal Behaviour Science, 117(1–2), 47–54. https://doi.org/10.1016/j.applanim.2008.12.011

LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, 155–184. https://doi.org/10.1146/annurev.neuro.23.1.155

Maier, S. F., & Seligman, M. E. P. (2016). Learned helplessness at fifty: Insights from neuroscience. Psychological Review, 123(4), 349–367. https://doi.org/10.1037/rev0000033

Maren, S. (2011). Seeking a spotless mind: Extinction, deconsolidation, and erasure of fear memory. Neuron, 70(5), 830–845. https://doi.org/10.1016/j.neuron.2011.04.023

McEwen, B. S. (2007). Physiology and neurobiology of stress and adaptation: Central role of the brain. Physiological Reviews, 87(3), 873–904. https://doi.org/10.1152/physrev.00041.2006

Sapolsky, R. M. (2015). Stress and the brain: Individual variability and the inverted-U. Nature Neuroscience, 18(10), 1344–1346. https://doi.org/10.1038/nn.4109

Sargisson, R. J., & McLean, I. G. (2021). Commentary: Efficacy of dog training with and without remote electronic collars vs. a focus on positive reinforcement. Frontiers in Veterinary Science, 8, 629746. https://doi.org/10.3389/fvets.2021.629746

Schalke, E., Stichnoth, J., Ott, S., & Jones-Baade, R. (2007). Clinical signs caused by the use of electric training collars on dogs in everyday life situations. Applied Animal Behaviour Science, 105(4), 369–380. https://doi.org/10.1016/j.applanim.2006.11.002

Schilder, M. B. H., & van der Borg, J. A. M. (2004). Training dogs with help of the shock collar: Short and long term behavioural effects. Applied Animal Behaviour Science, 85(3–4), 319–334. https://doi.org/10.1016/j.applanim.2003.10.004

Vieira de Castro, A. C., Fuchs, D., Morello, G. M., Pastur, S., de Sousa, L., & Olsson, I. A. S. (2020). Does training method matter? Evidence for the negative impact of aversive-based methods on companion dog welfare. PLoS ONE, 15(12), e0225023. https://doi.org/10.1371/journal.pone.0225023

Ziv, G. (2017). The effects of using aversive training methods in dogs – A review. Journal of Veterinary Behavior, 19, 50–60. https://doi.org/10.1016/j.jveb.2017.02.004