Aversive Training Methods in Dogs: Neurological Effects, Stress Responses and Long-Term Welfare Risks

1. Introduction

1.1 What "Aversive" Means

In learning theory, punishment is the application of an unpleasant stimulus intended to reduce a behavior. In practice this spans physical discomfort, startling stimuli, and intimidation. The defining feature for this article is not the specific tool but the mechanism it relies on: changing behavior by making the animal's experience worse, rather than by making a desired behavior pay.

1.2 Why This Question Is Contested at All

Few areas of dog training generate more heat, and the reason is not primarily evidential. The methods in question work in the short term, they are widely taught, and a substantial industry rests on them.

What has changed is that the comparison has been made under more controlled conditions, by more than one group, with results that mostly point the same way – and one recent result that does not. This article sets out what those studies found, what mechanism is proposed, and — importantly — which parts of the mechanism have been measured in dogs and which are imported from other species.

1.3 How to Read the Evidence

Two cautions frame everything below. First, the neural mechanisms invoked – amygdala-based fear learning, cortisol effects on the hippocampus, HPA-axis activation – are established mainly in rodents and humans; they are applied to dogs because the systems are deeply conserved, but most have not been imaged in the dog brain directly. Second, the dog studies themselves vary in strength: some are owner surveys comparing training styles, which are informative but open to confounds; others are field experiments with controlled assignment and physiological measurement, which carry far more weight. This article flags which is which rather than treating all "studies show" claims as equal.

2. Learning Mechanisms and Punishment

2.1 What Punishment Does – and Does Not – Teach

Punishment can reduce a behavior in the short term, but it carries a structural limitation: it tells the dog what not to do without teaching what to do instead, and it tends to create associations between the aversive event and whatever happened to be present when it occurred. Suppression is not the same as learning an alternative, and a suppressed behavior often re-emerges when the threat of punishment is absent (inhibited behavior is not erased behavior).

2.2 Two Different Neural Routes

Reward-based and punishment-based learning are not mirror images running through the same machinery. Reward learning engages dopaminergic motivation and reinforcement circuitry (the neurochemistry of learning through reward), building behavior the dog is motivated to repeat and tying it to a signal of "good things predicted" (the prediction-error basis of reward learning). Punishment engages threat and stress circuitry instead. Because these routes are thought to be neurologically distinct, their downstream consequences are expected to differ in kind rather than only in degree – a plausible reason, drawn from other species, for why the two approaches diverge in their side effects.

2.3 The Four Quadrants and Where Aversives Sit

Two of the four operant arrangements involve aversive stimuli, and they are frequently confused.

Positive punishment adds something unpleasant after a behavior, reducing its future frequency. Negative reinforcement removes something unpleasant when the desired behavior occurs, increasing that behavior's frequency.

Both require the aversive stimulus. In negative reinforcement the pressure must already be present for its removal to reinforce anything — which is why leash pressure, collar pressure and sustained electronic stimulation belong in this discussion even where nothing is described as punishment.

2.4 Why the Distinction Matters Less Than It Sounds

Practitioners sometimes defend negative reinforcement as gentler than punishment on the grounds that it teaches what to do rather than what not to do. The behavioral logic is sound. What follows for welfare is more conditional: the consequences depend on intensity, predictability, controllability and the individual animal, but both approaches rely on aversive stimulation, and the welfare measures in the studies discussed below do not separate the two.

2.5 What Timing Requires

For punishment to suppress a specific behavior, it must follow that behavior closely, consistently and at an intensity sufficient on the first application. Those conditions are difficult to meet outside a laboratory.

Where they are not met, the aversive event becomes associated with whatever else is present — the handler, the location, the other dog — rather than with the behavior it was aimed at (which is ordinary classical conditioning operating alongside the intended operant effect).

---

2.6 Why "It Worked for My Dog" Is Not Evidence

The claim is usually true and rarely informative, for three reasons that operate together.

Behavior problems fluctuate, and interventions are typically started when things are at their worst — so improvement is expected regardless of what was done. The person assessing the outcome is the person who chose the method. And the costs described in this article are not visible in the behavior that was targeted; they appear elsewhere and later.

None of this means individual experience is worthless. It means it cannot distinguish between an intervention that worked and one that coincided with improvement, which is what controlled comparison exists to do.

---

3. What Is Actually Being Used

3.1 The Range

The category covers a wide span, and treating it as one thing obscures differences that matter for both mechanism and welfare.

Physical corrections: leash jerks, collar pressure, pinning, physically forcing a position or forcing an item from the mouth.

Equipment-delivered aversives: choke and prong collars, remote electronic collars, spray and citronella collars.

Social and vocal: shouting, sustained staring, looming, the "alpha roll".

Environmental: water sprays, thrown objects, noise devices.

3.2 Why the Category Holds Together Anyway

What unites them is functional rather than technological. Each works by presenting or maintaining something the dog will act to avoid, and the physiological response depends on how the dog experiences it rather than on what generated it.

A dog that finds a raised voice genuinely threatening is in the same physiological state as one responding to collar pressure, and the difference in equipment does not appear in the cortisol measurement.

3.3 The Intensity Problem

Aversives are frequently applied at an intensity insufficient to suppress the behavior on first application, which produces the worst combination available: the stress response occurs, the behavior does not stop, and the dog habituates to that intensity.

The usual response is escalation, and escalation to an intensity that would have worked initially is now required against a partially habituated animal. The pattern is described by practitioners and follows from learning theory; it has not been measured systematically for most of the tools in the list.

3.4 What "Balanced Training" Describes

The term is now the most common self-description among trainers using aversives, and it denotes the combination of reward-based and aversive methods rather than a distinct third approach.

Section 10.2 gives the relevant finding: in one owner survey, dogs whose owners combined positive reinforcement with positive punishment had the highest mean aggression score of the method groups (Blackwell et al., 2008). The label describes the combination that fared worst on that measure in that survey, which is worth knowing when it is offered as a moderate position.

3.5 Where the Owner Sits in This

One consequence is easy to overlook. If the aversive is delivered by a person, that person is present when it happens — which is precisely the association the operant framework does not intend and the classical one produces anyway (with the relationship consequences documented separately).

---

4. The Amygdala: Fear and Threat Processing

4.1 Fear Memory Formation

The amygdala is the brain's hub for emotionally significant, especially threatening, stimuli. On the model from other species, an aversive event during training strongly engages the amygdala, which facilitates rapid, durable fear learning (how conditioned fear is formed and stored). Evolutionarily this is adaptive – fast learning about danger keeps animals alive – but in a training context that speed and durability work against the trainer's intent.

4.2 Misplaced Associations

The critical problem is what the fear attaches to. Because the amygdala binds threat to whatever is salient at the moment of the aversive, punishment delivered in the presence of other dogs, unfamiliar people, or a particular place can teach the dog that those cues predict danger. The intended lesson ("don't do that behavior") is not necessarily what gets learned; what can be learned instead is that a situation is unsafe. Over time this can produce fearfulness, avoidance, and defensive aggression (a nervous system tuned toward reactivity) – outcomes precisely opposite to a calm, trained dog.

5. The Hippocampus and Contextual Memory

The hippocampus encodes where and under what circumstances significant events happened, and it is highly sensitive to stress hormones. In rodents and humans, elevated cortisol alters hippocampal memory processing and plasticity, and chronic exposure can impair the structure's function (the full cortisol–hippocampus story, and its limits in dogs). Applied to training, the expectation is that repeated aversive sessions lay down strong contextual memories linking an environment to fear or discomfort, which can generalize into anxiety in any setting that resembles it – even after the aversive itself is gone. This is a reasonable, mechanistically grounded prediction; it should be stated as cross-species inference rather than as something measured in the dog hippocampus, because the latter has not been done.

5.1 Why Context Encoding Matters Practically

The hippocampus binds an emotional event to where and when it occurred, which is why fear acquired in one place can be absent in another and why a dog that was fine on the training field is not fine at the vet.

It also explains a common clinical picture: a dog punished during an encounter with another dog on a particular street becomes reactive on that street, then on similar streets, then on lead generally. Each step is ordinary generalization (as the conditioning literature describes).

5.2 What Stress Does to the Structure Itself

The hippocampus is also one of the structures most affected by sustained glucocorticoid exposure, which produces a circular problem: the system that encodes context is degraded by the stress it is encoding.

That relationship is established in rodent and human research and has not been demonstrated directly in dogs — an extrapolation, stated as one.

---

6. The HPA Axis and the Stress Response

The hypothalamic-pituitary-adrenal axis is the body's central stress system: a perceived threat drives the hypothalamus and pituitary to signal the adrenal glands to release cortisol, mobilizing energy and raising alertness. This is adaptive acutely and costly chronically. Here the dog-specific evidence is direct, and more differentiated than it is often presented. In an electronic collar field study, high-intensity stimulation without warning cues in a preliminary phase produced negative behavioral changes and elevated cortisol; in the main study with lower settings, e-collar dogs were more often tense and yawned more than dogs trained by reward-based trainers, while cortisol did not differ significantly (Cooper et al., 2014). In laboratory dogs, salivary cortisol rose significantly when shocks were poorly predictable but not when they were tied directly to touching a prey dummy (Schalke et al., 2007). Across training schools, dogs trained with aversive methods showed larger post-training cortisol increases than reward-trained dogs (Vieira de Castro et al., 2020). Repeated activation of this system is associated with the longer-term shifts in emotional regulation and stress sensitivity described across the chronic-stress literature (how the arousal system becomes miscalibrated).

7. Behavioral Consequences

The neural picture translates into observable outcomes. Punishment-based training has been associated in several studies with more stress-related behavior, fear and aggression (Guilherme Fernandes et al., 2017), with dogs less likely to interact with a stranger and less playful (Rooney & Cowan, 2011), and with more problem behaviors overall (Hiby et al., 2004) – patterns consistent with anxiety, avoidance and behavior that is suppressed rather than understood (anxiety has its own measurable neurobiology). A dog that stops engaging is often read as "calm" or "respectful," when the underlying state is avoidance and inhibition – and inferring the inner state from the quiet exterior is exactly the mistake to avoid (behavior does not equal emotion).

7.1 Why Suppression Looks Like Success

The practical difficulty is that punishment often works in the short term and by a measure that flatters it. The behavior stops, the handler observes the improvement, and the improvement is real as far as the observation goes.

What is not visible is whether the motivation producing the behavior has changed, whether the warning signals preceding it have been suppressed along with the behavior, and what has been associated with the aversive event in the process (with the inference problem set out separately).

7.2 The Warning Signal Problem

Where the punished behavior is a growl or a threat display, the consequence is specific and dangerous. Those signals exist to prevent escalation, and removing them does not remove the state that produced them.

The result is the dog described as biting without warning — which is a description of what the observer saw rather than of what the dog did (as the guarding literature documents in detail).

---

7.3 The Redirection Problem

One consequence appears frequently in behavior caseloads and rarely in training discussions. A dog prevented from acting on a motivation does not lose the motivation, and where the original outlet is blocked the behavior may appear elsewhere.

A dog corrected for lunging at other dogs may begin lunging at cyclists; one corrected for barking at the window may begin barking at sounds. The suppression is specific to what was punished; the state driving it is not. The pattern is reported from practice rather than measured in studies.

7.4 Effects on the Handler Relationship

Where the aversive is delivered by the owner, the owner becomes a predictor of it. That does not usually produce a dog that avoids its owner — the attachment relationship tends to be robust — and it can produce a dog that is less willing to offer behavior, slower to approach, and more inhibited in training.

Reduced offering is the practical cost, and it is invisible unless someone is looking for it: a dog that does nothing is easily read as a dog that is calm (which is the distinction the suppression literature turns on).

---

8. Learned Helplessness: Suppression Mistaken for Success

The most concerning outcome deserves its own treatment. Learned helplessness – first demonstrated, as it happens, in dogs given inescapable, uncontrollable aversives – occurs when an animal cannot predict or control the unpleasant events it experiences and eventually stops trying to respond at all (the neurobiology of learned helplessness). In training, a dog repeatedly punished with no clear, controllable way to avoid it may cease offering behavior, exploring, or engaging. From the outside this looks like obedience: the dog is quiet, still, compliant. The proposed reading is that this is closer to a suppression of behavioral output than to successful learning.

That inference has to be made carefully. Similar outward behavior can have several causes — a trained alternative, inhibition, avoidance, low motivation, or a stress response — and calm stillness should not automatically be interpreted as learned helplessness (which is the general limit on reading states from behavior). What distinguishes them is what else the dog does: whether it offers behavior, approaches readily, and recovers normally when the situation changes. The distinction is practical, not academic: a dog that has genuinely learned stays curious, engaged, and willing to offer behavior, whereas a helpless dog is subdued, passive, and low in initiative (the prefrontal machinery for engaged, flexible responding goes quiet). Reading collapse as calm is one of the most consequential errors in training.

8.1 What the Original Experiments Showed

The phenomenon was identified in dogs. Dogs exposed to inescapable shock subsequently failed to learn to escape when escape became available (Overmier & Seligman, 1967). In a follow-up with a yoked design, dogs that could end the shock by pressing a panel later learned to escape normally, whereas dogs receiving identical but inescapable shocks did not (Seligman & Maier, 1967).

The variable was controllability rather than the shock itself, which is the finding that carries forward.

8.2 The Modern Reformulation

The interpretation has since been revised substantially: passivity under uncontrollable stress is now understood as the default mammalian response, with the detection of control being what is learned (Maier & Seligman, 2016) (with the reformulation set out separately).

That revision does not weaken the practical point. It relocates it: what training either provides or withholds is the experience of controllability, and a dog whose behavior has no effect on what happens to it is in the condition the original experiments created.

---

9. The Evidence – and Its Quality

9.1 Welfare Findings

A growing body of research converges on elevated stress and poorer welfare with aversive methods. Dogs trained with aversive techniques showed more stress-related behaviors, larger cortisol rises, and a more "pessimistic" cognitive bias outside training than reward-trained dogs (Vieira de Castro et al., 2020); the e-collar field study found more tense behavior and yawning, though not higher cortisol, in its main study (Cooper et al., 2014); a review of 17 studies concluded in favor of reward-based approaches (Ziv, 2017), and a second review agreed on the direction but judged the evidence too limited for strong conclusions (Guilherme Fernandes et al., 2017). Common stress signals – lip-licking, yawning, lowered posture, avoidance – recur across these studies.

9.2 The Efficacy Question

The strongest counter-argument for aversives is that they are simply more effective, and this is where recent evidence matters most. Comparing e-collar training against reward-based training for exactly the problem e-collars are marketed to fix – unreliable recall – a controlled study found reward-focused training to achieve better recall and sit responses, with no deterrent advantage for the aversive method (China et al., 2020). For recall training, then, if aversives are not more effective and carry welfare costs, the central justification for using them largely collapses. The relevant question is not whether aversives can suppress behavior — they can — but whether they perform better than the alternative and at what cost. On current evidence for recall and general obedience the answer is: not better on the measures compared, and at documented cost; section 9.6 describes the one task on which a recent study found otherwise (reinforcement schedules explain how reward-based behavior is built to last).

9.3 Methodological Caveats

Intellectual honesty requires noting the limits. Some welfare evidence comes from owner surveys comparing training styles, which cannot fully rule out that dogs sent for aversive training differed at the outset, or that owners who choose harsh methods differ in other ways – correlation is not cleanly causation (the general difficulty of measuring behavior and its causes). The most weight therefore falls on the controlled field studies with physiological measurement and assigned groups (Cooper et al., 2014; China et al., 2020) and on cognitive-bias experiments (Vieira de Castro et al., 2020), which are harder to explain away. The overall pattern is consistent and points one direction; the confidence attached to it should still track the quality of each study rather than the vividness of its conclusion.

9.4 What the Electronic Collar Trials Found

The controlled comparisons are the strongest design available here. Sixty-three pet dogs referred for recall problems were assigned to three groups of 21: industry-approved trainers using electronic collars, the same trainers without electronic collars, and trainers from a pet dog trainers' association working reward-based, with behavior and physiological measures recorded throughout (Cooper, Cracknell, Hardiman, Wright & Mills, 2014).

Dogs trained with electronic collars spent more time tense, yawned more and explored less than dogs in the reward-based group, and the e-collar brought no consistent benefit. A subsequent analysis with the same design and sample size found better recall and sit responses in the reward-focused group (China, Mills & Cooper, 2020). The two analyses should not be counted as independent trials, and the efficacy analysis has been criticized for semi-random group assignment and for training the reward group at a different time and place (Sargisson & McLean, 2021).

9.5 Why That Combination Is Decisive

An intervention carrying welfare costs can be justified if it delivers something the alternative does not. For recall training, these trials remove that argument: the welfare costs were present and the efficacy advantage was not.

That is a narrower finding than "electronic collars do not work" — they do suppress behavior — and a more useful one, because it addresses the actual claim made for them.

9.6 The Finding That Runs the Other Way

One recent study reached a different conclusion and belongs here rather than in a footnote. Comparing electronic collars against food-reward methods in randomly assigned groups for stopping dogs chasing a lure, dogs receiving shocks stopped chasing within one or two ten-minute sessions and did not chase in the first three tests, although 67 percent chased again in the final test in a novel arena, while two reward-trained groups did not stop across five training sessions and failed all four tests (Johnson & Wynne, 2024).

The authors report that apart from yelps at the moment of shock, no group showed signs of stress or distress, and they conclude that e-collars may be appropriate for expert trainers addressing behaviors with serious welfare consequences — while calling for work on the expertise required, the problems best suited, and longer-term implications.

9.7 How to Weigh It

Three things at once. It is a genuine efficacy result on a specific task, obtained where earlier studies found no advantage. Its welfare assessment is less extensive than the measures used by Cooper and colleagues, and absence of observed distress is not equivalent to demonstrated absence of cost. And the population — expert application, a defined task, short duration — is the favorable case rather than the typical one.

The honest summary is that the efficacy question is less closed than the welfare question, and that the predatory context specifically may differ from the general training case (where the intervention evidence is thin in every direction).

9.8 Predictability in the Laboratory

A laboratory study reconstructed everyday training situations, such as stopping hunting behavior or failing to respond to a recall, in laboratory dogs. Salivary cortisol rose significantly in the groups for which shocks were poorly predictable, but not in the group for which the shock was tied directly to touching a prey dummy (Schalke, Stichnoth, Ott & Jones-Baade, 2007).

The finding matters for the correct-use argument: even under controlled conditions, predictability determined the stress response, and predictability is exactly what is hardest to guarantee outside a laboratory.

---

10. What the Owner-Survey Data Show

10.1 The Population-Level Picture

In a questionnaire survey of dog owners, 88 percent of whose dogs had received some form of training, 16 percent of owners reported using only positive reinforcement, and 72 percent used some form of positive punishment. The authors report associations between the methods used and several categories of undesirable behavior, including aggression (Blackwell, Twells, Seawright & Casey, 2008).

10.2 The Combination Finding

The result that complicates the simple framing concerns owners who combined rewards and punishment. Owners using positive reinforcement together with positive punishment had dogs with the highest mean aggression score of the method groups; no group in the survey used punishment alone (Blackwell et al., 2008).

One reading is that inconsistency is its own problem: a dog for whom the same behavior sometimes produces reward and sometimes punishment cannot predict the outcome, and unpredictability is a documented stressor in its own right (with the physiology described separately).

The design cannot establish that reading, and the finding argues against the common compromise position that mixing approaches is a moderate middle path.

10.3 The Earlier Survey

The pattern was visible two decades ago. Questioning 364 dog owners about how they trained seven basic tasks, researchers found 66 percent using vocal punishment, 12 percent physical punishment, 60 percent praise, 51 percent food rewards and 11 percent play (Hiby, Rooney & Bradshaw, 2004).

Owner-rated obedience in eight tasks correlated with the number of tasks trained using rewards but not with the number trained using punishment. The number of 16 common problem behaviors reported correlated with the number of tasks trained using punishment but not with rewards (Hiby et al., 2004).

10.4 Confrontational Methods Specifically

One survey addressed the confrontational end directly, asking owners which interventions they had used and what the dog did in response. Physically forcing an item from the mouth, staring the dog down, hitting or kicking, and the alpha roll each produced aggressive responses in at least a quarter of the dogs on which they were used (Herron, Shofer & Reisner, 2009).

Non-confrontational interventions produced aggressive responses far less often. This is owner-reported and retrospective, and it is the most direct data available on the specific techniques still taught as leadership exercises.

10.5 Why Surveys Alone Would Not Settle It

Every study in the preceding sections relies on owner report, which raises an obvious objection: owners who use punishment may perceive or report their dogs differently, and the association could be an artifact of who answers how (with the general measurement problem set out separately).

10.6 The Study That Addressed It

One study reduced the problem by observing outcomes directly. Fifty-three owners reported how they trained seven common tasks and were filmed with their dog in standardized scenarios at home, including a training session on a novel task, with behavior scored from the footage (Rooney & Cowan, 2011).

Dogs whose owners reported using more rewards tended to perform better in the novel training task, dogs whose owners reported more punishment were less likely to interact with a stranger, and dogs whose owners used physical punishment tended to be less playful (Rooney & Cowan, 2011). The outcomes were observed rather than reported, which removes part of the survey weakness; the training history itself was still reported by the owners.

10.7 What the Convergence Means

Three designs with different weaknesses — owner questionnaires, direct observation of learning, and controlled comparison of training with and without electronic collars — produce results in the same direction, with the chasing study as the notable exception on efficacy (Johnson & Wynne, 2024).

None of them is a randomized controlled trial assigning dogs to methods and following them for years, and the agreement between designs with different weaknesses is the strongest form of evidence this field currently offers.

---

11. Welfare and Practical Implications

From a welfare standpoint, methods should minimize stress and avoid unnecessary suffering, and because aversives measurably increase stress indicators in dogs, their use raises a real ethical burden – one not offset by a clear efficacy advantage in the comparisons made for recall training. The constructive alternative is well defined, and reward-based methods allow learning without the risks that come with aversive consequences. Reinforcement-based training builds desired behavior through motivation rather than suppressing unwanted behavior through threat; it fosters engagement, confidence, and emotional stability; and it avoids straining the dog–human relationship (the bond that underpins cooperative training). It also aligns with how dogs actually learn (the neurology of learning and behavior), which is why it tends to produce a dog that participates rather than one that merely stops.

11.1 Why Professional Bodies Have Moved

The convergence in the evidence has been followed by convergence in position statements. Veterinary behavior organisations and welfare bodies across several countries now recommend against aversive methods in routine training, and a number of European jurisdictions restrict or prohibit electronic collars specifically.

Regulatory status differs between countries and changes over time; the current position in any given jurisdiction is a question for the relevant authority rather than something this article can settle.

11.2 What This Does Not Establish

Three limits worth stating plainly. The welfare findings concern group comparisons and do not predict how any individual dog responds. No study has followed dogs for years after training with either approach. And the efficacy comparisons were made on specific tasks with trainers experienced in the method, which is not the condition under which most equipment is actually used.

The last point cuts against aversive methods rather than for them: the trials represent competent application, and the everyday version is less controlled.

---

11.3 What to Ask a Trainer

Four questions separate approaches more reliably than the labels used to describe them, because almost everyone now describes their methods as positive.

What happens when the dog gets it wrong? The answer is the method. "Nothing" and "we make it easier" are different from "we correct it".

What equipment is used, and what does it do when the dog pulls? Equipment that applies pressure, tightens or delivers a stimulus is delivering an aversive whatever it is called.

How is progress measured? Suppression of an unwanted behavior and acquisition of a wanted one look different in the answer.

What happens if it does not work? Escalation of intensity is the pattern described in section 3.3.

---

12. What the Alternative Requires

12.1 Why "Just Use Rewards" Is Insufficient Advice

The evidence supports reward-based methods and does not make them automatic. A poorly timed reward teaches the wrong thing as reliably as a poorly timed aversive, and the reinforcement literature is specific about what makes the difference (with schedules affecting persistence directly).

12.2 What Aversive Methods Are Usually Reaching For

Two things, and both have alternatives. Interruption of a behavior already in progress, and reliability under distraction.

Interruption can be achieved by managing the situation so the behavior does not start, which is less satisfying and more effective. Reliability is a training problem — sufficient repetitions, graduated distraction, an adequate reinforcement history — and punishment is not a shortcut around it.

12.3 Where the Emotional Layer Is the Problem

Much of what owners want suppressed is fear-driven, and punishment applied to a fear response adds an aversive event to a situation the dog already finds aversive. Not every dog becomes fearful through punishment, but in a fearful or insecure dog punishment often intensifies exactly the emotional state that drives the problem.

The supported approach targets the emotional response rather than the behavior expressing it (with the method and its limits set out separately), and it is slower, which is a substantial part of why the alternative persists.

12.4 The Medical Differential

A dog whose behavior has changed suddenly, or which reacts to handling in a way it did not before, warrants physical examination before any training decision. Pain features heavily in behavior caseloads, and an aversive applied to a pain response compounds the original problem (with the evidence set out separately).

12.5 What Cannot Be Promised

Reward-based methods are not faster, not effortless, and not universally successful. What the evidence supports is that they perform at least as well on the effectiveness measures compared for recall and obedience, with the chasing study as an exception, while avoiding the documented welfare costs — which is a comparative claim rather than a guarantee.

---

13. Why the Methods Persist

13.1 They Work in the Moment

The behavior stops, immediately and visibly. That is the strongest possible reinforcement for the handler, delivered on the shortest possible schedule — which means the person applying the aversive is being conditioned at least as effectively as the dog.

This is rarely stated and it explains a great deal. A method that reliably produces an immediate result will be repeated regardless of what it produces over months.

13.2 The Costs Are Delayed and Attributed Elsewhere

Fear generalization, suppressed warning signals and altered associations with the handler can appear weeks or months later, in situations that look unrelated to the training session.

By the time a dog reacts on lead or snaps without warning, the connection to a correction administered in another context is not visible — and the usual explanation available is the dog's temperament.

13.3 The Framing Is Available

Aversive approaches come with an explanatory vocabulary that makes them sound principled: leadership, boundaries, respect, structure. Several of these rest on the dominance framework, which does not survive scrutiny (as documented separately).

The vocabulary persists after the framework it came from has gone, which is a pattern this library documents in several places.

13.4 The Alternative Is Slower

Reward-based work on an established problem takes weeks to months and requires consistency the household may not have. An owner choosing between a solution that appears to work tonight and one that requires three months is not making an evidential judgment.

That is the honest competitive position, and pretending otherwise does not help anyone assessing the options.

---

14. Summary at a Glance

Punishment suppresses without teaching — It reduces the frequency of a behavior and supplies no information about what to do instead.

Two neural routes operate at once — The intended operant effect and an unintended classical association form together, and the second is not under the trainer's control.

Timing conditions are rarely met — Close, consistent and sufficiently intense on first application; where they fail, the aversive attaches to whatever else is present.

Punishment can stop behavior — The decisive question is what was learned and at what cost; a stopped behavior is not automatically a resolved one.

Electronic collars produced welfare costs without efficacy gains in recall training — In controlled comparison, training with remote electronic collars showed no advantage over reward-based training while producing behavioral indicators of poorer welfare; the two analyses share one design and sample (Cooper et al., 2014; China, Mills & Cooper, 2020).

One study found the opposite on efficacy — Electronic collars stopped dogs chasing a lure where food-reward methods did not, although most shocked dogs chased again in a novel arena (Johnson & Wynne, 2024).

Method was associated with problem behavior — In an owner survey, 72 percent of owners used some form of positive punishment, and the authors report associations between methods and several categories of undesirable behavior (Blackwell et al., 2008).

Combining rewards and punishment fared worst on aggression — In one owner survey, positive reinforcement combined with positive punishment had the highest mean aggression score among the method groups (Blackwell et al., 2008).

Direct observation points the same way — Dogs whose owners reported using more rewards performed better in a novel task under videotaped assessment (Rooney & Cowan, 2011).

Controllability was the variable in the original experiments — Dogs that could end the same shocks did not develop the escape deficit, whereas yoked dogs did (Seligman & Maier, 1967).

---

14.1 Research Gaps

The physiology is largely extrapolated. Amygdala, hippocampal and HPA mechanisms come from rodent and human research. What has been measured in dogs is the observable output — cortisol, stress behaviors, learning performance, welfare indicators — rather than the neural pathway.

No long-term follow-up exists. No study has tracked dogs for years after training with either approach, which means the durability of both the welfare costs and the training outcomes is unknown.

Owner report dominates the population data. The direct-observation study (Rooney & Cowan, 2011) is the exception rather than the rule, and the questionnaire studies share a common weakness.

Individual variation is not modelled. Group comparisons cannot say which dogs are most affected, and there is reason to expect the effects differ substantially between individuals (with coping style varying between animals).

Intensity and frequency are rarely quantified. "Aversive methods" covers a single leash correction and a daily regime of electronic stimulation, and few studies separate them.

---

14.2 Where the Field Stands

The convergence described in section 10.7 is unusual for canine behavior research, where designs with different weaknesses more often produce results pointing in different directions.

What is missing is not more evidence of the same kind but a different kind: longitudinal follow-up, individual-level prediction, and quantification of intensity and frequency. Those are the studies that would move the question from "which approach is associated with better outcomes" to "what happens to this dog".

---

15. Conclusion

Aversive training methods affect more than behavior. 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 proposed mechanism involves the neurobiology of fear and stress – amygdala-based fear learning, stress effects on hippocampal memory, and repeated HPA-axis activation – with the honest caveat that it is largely established in other mammals and carried over to dogs, while the direct canine evidence is strongest at the level of measured stress and welfare. For recall training, the best current evidence indicates no efficacy advantage for aversive methods, while one study on lure chasing found one. Understanding this does not require accepting every claim uncritically; it requires weighing a coherent mechanism against a converging, if imperfect, body of canine research. That weighing points clearly toward reinforcement- and cooperation-based methods as more humane and, on most comparisons, no less effective – which is what makes them the better-supported choice.

Key Insights (Takeaways)

• Punishment can stop a behavior; the decisive question is what was learned and at what cost. On the model from other species, aversive methods engage threat and stress systems, and reward-based and punishment-based learning are thought to run through different routes, so their side effects are expected to differ.

• Fear can attach to whatever is present when the aversive occurs, so punishment near other dogs, people, or places can teach that those things are dangerous. In a fearful or insecure dog, punishment often intensifies exactly the state that drives the problem. The cortisol–hippocampus and HPA mechanisms are established mainly in rodents and humans.

• The dog-specific evidence is strongest on welfare: aversive-trained dogs showed more stress behaviors, larger cortisol increases and more pessimistic cognitive bias (Vieira de Castro et al., 2020), e-collar dogs were more tense (Cooper et al., 2014), and stress responses depended on predictability (Schalke et al., 2007). Reviews agree on the direction and differ in firmness (Ziv, 2017; Guilherme Fernandes et al., 2017).

• On efficacy, reward-focused training achieved better recall 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). One study found e-collars more effective at stopping lure chasing (Johnson & Wynne, 2024). For recall training, aversives are not more effective and carry welfare costs.

• Controllability was the decisive variable in the original helplessness experiments with dogs (Seligman & Maier, 1967). Owner surveys carry confounds, while observed outcomes and controlled comparisons carry more weight – and they point the same way on welfare. Reward-based methods allow learning without the risks of aversive consequences.

References

Blackwell, E. J., Twells, C., Seawright, A., & Casey, R. A. (2008). The relationship between training methods and the occurrence of behavior problems, as reported by owners, in a population of domestic dogs. Journal of Veterinary Behavior, 3(5), 207–217. https://doi.org/10.1016/j.jveb.2007.10.008

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

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

Hiby, E. F., Rooney, N. J., & Bradshaw, J. W. S. (2004). Dog training methods: Their use, effectiveness and interaction with behaviour and welfare. Animal Welfare, 13(1), 63–69. https://doi.org/10.1017/S0962728600026683

Johnson, A. C., & Wynne, C. D. L. (2024). Comparison of the efficacy and welfare of different training methods in stopping chasing behavior in dogs. Animals, 14(18), 2632. https://doi.org/10.3390/ani14182632

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

Overmier, J. B., & Seligman, M. E. P. (1967). Effects of inescapable shock upon subsequent escape and avoidance responding. Journal of Comparative and Physiological Psychology, 63(1), 28–33. https://doi.org/10.1037/h0024166

Rooney, N. J., & Cowan, S. (2011). Training methods and owner–dog interactions: Links with dog behaviour and learning ability. Applied Animal Behaviour Science, 132(3–4), 169–177. https://doi.org/10.1016/j.applanim.2011.03.007

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

Seligman, M. E. P., & Maier, S. F. (1967). Failure to escape traumatic shock. Journal of Experimental Psychology, 74(1), 1–9. https://doi.org/10.1037/h0024514

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