Shock Collars: Effective Does Not Mean Ethical
When "It Works" Is Doing Rather a Lot of Work
If an electric collar stops a dog from chasing, does that make it good training? It sounds like a straightforward question, but it quietly bundles together two separate issues. Can electric punishment change behaviour, and is deliberately causing pain an acceptable way to achieve that change? Those aren't the same question, and conflating them is where a lot of the argument for shock collars falls apart.
A 2024 study by Anamarie Johnson and Clive Wynne compared electric collars with two food-based procedures intended to stop dogs chasing a fast-moving mechanical lure. The electric-collar group showed much stronger suppression of chasing during the study's tests, and that's the finding most likely to get repeated by people defending the use of shock collars. The shocks worked. An investigation should start there rather than stop there, because all dogs exposed to the electric stimulation yelped at some point during training, according to the paper and the methodological critique that followed it. Once pain has produced the desired graph, we're apparently meant to treat it as a minor administrative detail.
This piece looks closely at what the researchers actually tested, whether the comparison represented competent reward-based training, what the welfare measures could realistically detect, and what the dogs may actually have learnt. There's a more basic question underneath all of that. Should someone responsible for an animal's wellbeing deliberately cause that animal pain simply because it produces quick compliance? My own view is straightforward. I don't believe there's an ethical justification for deliberately inflicting physical pain on a dog in training, and that comes from both the welfare science and a basic duty of care. Effective is not a synonym for ethical.
Dog Shock Collar - 4500FT Dog Training Collar with Remote, IPX8 Waterproof Electric Dog Collar with 4 Training Modes, Security Lock, Rechargeable eCollar for 10-120lbs Small Medium Large Dogs
This Argument Has Been Made Before
None of this is new. Dog training has a long habit of reaching for whatever produces the fastest visible result and worrying about the rest later, and the shock collar is only the most recent version of that habit.
The electric lead itself goes back further than most people realise. A device patented in 1935 sent a shock down an electrified leash connected to a generator, with the inventor noting that the leash's length let a handler deliver the punishment from a distance. By the 1950s a version had done away with the physical connection altogether, using a radio transmitter and a receiver worn on a harness, and the patent for it claimed the new design was more humane than the models that came before, which had already been criticised for cruelty. Every generation of the technology has arrived with the same reassurance attached: this one is the humane version.
The same pattern shows up away from electricity entirely. In Britain, Barbara Woodhouse became the most recognisable dog trainer in the country through her BBC series in 1980, built around a choke chain and a sharp, well-timed jerk, summed up in her own phrase as "jerk 'em and love 'em." It worked for a great many people watching at home. It also left enough dogs with neck and eye injuries that vets began describing a specific pattern of damage as "Woodhouse neck," and her methods are now widely regarded as outdated rather than exemplary.
America had its own version a few decades earlier. Blanche Saunders spent the 1950s touring the United States promoting obedience training for ordinary pet owners, at a time when hardly anyone trained a dog beyond the basics. Her method leaned heavily on the leash jerk: dogs were corrected for getting it wrong and praised for getting it right, and food played almost no part in the process because it was seen as bribery rather than training. Saunders is credited with getting obedience training into ordinary households in the first place, which is a real legacy. The method she used to do it has not aged well.
None of these people were operating in bad faith, and none of them lacked evidence that their methods worked. Woodhouse's dogs sat, heeled and stopped pulling on camera. Saunders' students got results good enough to fill obedience classes across the country. That is exactly the problem. Effectiveness has never been the part in question. What each of these approaches shares with the modern electric collar is the same unexamined leap from "the dog stopped doing it" to "this was therefore a good way to teach it," and that leap deserves exactly as much scrutiny now as it did then.
image courtesy of https://www.amazon.co.uk/No-Bad-Dogs-Woodhouse-Way/dp/0671541854
What Did the Shock-Collar Study Actually Compare?
The experiment used dogs selected because they were willing to chase a fast-moving lure. They were split into three groups. One received electric stimulation when crossing a set point while approaching the lure. The other two underwent procedures involving food, a conditioned word (the word chosen was "banana"), and either a fast or progressively faster lure. The dogs were then tested to see whether they'd hold back from chasing in both familiar and unfamiliar conditions.
On the surface this looks like a clean comparison between electric punishment and positive reinforcement. Calling something a food-reward group, though, doesn't mean the dogs were actually given a well-designed reward-based training programme, and that distinction matters.
The researchers paired the word "banana" with food 120 times over roughly 20 minutes, presumably so the word would come to predict something highly rewarding. The trouble is that treats were delivered at regular intervals throughout, so the simple passage of time may have predicted food just as well as the word did. The study never established that every dog had actually learnt the intended association before being moved on to the far harder chasing exercise. Predicting food, in any case, isn't the same as performing a recall. A recall is an active sequence: the dog hears the cue, disengages from the distraction, turns, travels back to the handler and stays there. Teaching that behaviour properly means starting at close range in an easy environment, reinforcing the dog repeatedly for getting it right, and only then adding distance, movement and competing motivations one at a time.
In this study, the dogs were expected to hear "banana" after already travelling roughly 23 metres towards a lure they'd shown a strong desire to chase, without having practised the full recall response first in progressively harder circumstances. They sat the final exam before anyone had confirmed they understood the lesson.
Chasing is also rewarding in its own right, and every second a dog spends pursuing a fast-moving object may strengthen the very behaviour the trainer is trying to reduce. In the food conditions, dogs could carry on towards or around the lure while the cue was repeated, with food waiting whenever they eventually came back. Westlund's detailed critique points out this created the odd possibility that a dog responding promptly could receive one treat, while a dog ignoring repeated cues might end up with a bigger pile later. Call it a recall protocol designed by a committee that had never actually met a dog. The study also never established whether the food on offer was valuable enough to compete with chasing for each individual dog, and a dog selected specifically for loving the chase isn't obliged to find a snack more exciting than a lure in full flight. Competent reward-based training adjusts the reinforcer, the environment, the timing and the difficulty according to how the dog responds, and food simply being present in a study is no guarantee that positive reinforcement actually took place. A reinforcer is defined by whether it increases the desired behaviour, not by whether it was handed out.
The electric-collar dogs received up to around 60 minutes of lure-based training, and the food groups received roughly 50 minutes after the conditioning procedure. Matching the time looks tidy on paper, but different learning processes don't necessarily move at the same rate. Pain can suppress behaviour very quickly, because animals are biologically wired to learn fast about threats, whereas reward-based training usually involves building an alternative response and strengthening it across varied situations, which tends to take more repetitions. Speed alone says nothing about which process is safer, more reliable, or better understood by the dog. Picture two ways of getting someone out of a chair. In one, you teach a cue, reward the movement and practise around increasing distractions. In the other, the chair shocks them. The second method probably gets them standing faster, and faster isn't the same as better teaching.
The electric-collar group and the food groups weren't even learning the same task. A dog in the electric-collar condition could succeed simply by avoiding the lure, refusing to cross the threshold or suppressing forward movement, while a dog in the food condition was expected to disengage from a powerful attraction, turn and travel back to the handler in response to a cue that had barely been practised. Given how different those demands are, the defensible conclusion from the experiment is actually quite narrow: under this particular protocol, electric shocks suppressed movement towards the lure more effectively than the two food procedures the researchers used. Showing that is a long way from showing electric collars outperform skilled reward-based training as a whole. There was no properly developed emergency recall programme tested here, no comprehensive predation-focused intervention, and no evidence that pain was actually necessary.
When Every Shocked Dog Yelps, Is "Minimal Harm" Credible?
The most important welfare finding is also the one easiest to lose among the statistical language. Every dog in the electric-collar group yelped during the intervention. The researchers described these vocalisations as presumably pain-induced, yet still concluded they'd observed little other evidence of stress or distress, going on to suggest electric collars could have a place in addressing dangerous behaviours under expert supervision. That interpretation deserves scrutiny, because the pain gets acknowledged and then treated almost as though it sits outside the welfare assessment altogether, when pain is itself a welfare cost regardless of how carefully it's phrased.
A yelp can't tell us the exact intensity, quality or duration of what a dog experiences. Dogs vary in how readily they vocalise, some may yelp at lower intensities than others, and the absence of a yelp wouldn't prove the absence of pain either. When every single dog exposed to an intervention vocalises during it, though, that response is hard to wave away as incidental, particularly since the researchers themselves interpreted the yelps as pain-related. This wasn't an unavoidable veterinary procedure meant to diagnose or treat illness. The pain was introduced on purpose because the researchers wanted it to change behaviour, and whether it lasted seconds rather than minutes changes the scale of the harm without making the event ethically neutral.
Electric-collar advocates often reach for softer words like "stim", "tap" or "communication". Those terms might sit more comfortably with the person holding the transmitter, but what actually matters is what the dog experiences. For positive punishment to reduce behaviour, the added consequence has to be sufficiently aversive to that individual dog. The researchers discussed avoiding gradual habituation and referred to literature advocating an intense punishing stimulus that stopped short of physical damage, which is not a neutral choice of method. It's an attempt to apply something severe enough to produce rapid avoidance while staying below a threshold of visible injury, and "no tissue damage" is an extraordinarily low bar for humane care. A method can cause pain, fear or emotional conflict without leaving a burn, a wound or an abnormal veterinary finding.
The welfare measures used were also fairly narrow. The dogs were filmed and selected behaviours were coded during training and testing, but the critique of the study notes that several commonly relevant body-language indicators, such as ear position, tail carriage, flinching and crouching, weren't apparently included in the analysed welfare measures. Some behaviours occurred too rarely for statistical assessment, and behaviour between active trials wasn't examined in the same way. A dog chasing at speed is already highly aroused, so subtle signs of distress may be genuinely hard to spot mid-pursuit, and responses might only show up afterwards, as the dog recovers, re-enters the area, or encounters related cues later on. A narrow list of behaviours recorded within a narrow window can show only that the researchers didn't detect much of the specific behaviour they'd chosen to count. It says nothing about whether distress was actually absent.
Imagine assessing a medical procedure in people by recording only whether participants yawned, scratched or shook themselves afterwards. Every participant cries out when the procedure is applied, yet researchers find little else on their coding sheet. Would the sensible conclusion be that the procedure had minimal welfare impact, or would the universal cry of pain be the actual headline finding? There's a real risk in animal research that numerical measures end up carrying more authority than direct experience. Spreadsheets behave themselves; a yelp does not. Welfare assessment shouldn't work like a treasure hunt where pain only counts once it's accompanied by three approved stress behaviours and a statistically significant result.
The study also looked at faecal cortisol metabolites, though relatively few dogs supplied samples on every study day. Westlund reports that only nine of the 19 dogs initially involved produced complete daily samples, and two electric-collar dogs were later removed from the main analysis entirely because they'd exceeded the ethically permitted number of shocks. Faecal cortisol isn't an instant pain detector either. It reflects physiological arousal across a preceding period and can be affected by excitement as much as fear or discomfort, and the dogs chasing the lure without shock had an obvious source of positive arousal that muddies the picture further. Training also occupied only a small part of each day, so a faecal measure reflecting many hours could easily dilute a brief response occurring right at the moment of shock. Failing to detect a clear statistical cortisol difference is not the same as demonstrating the shocks were painless. It means this particular sample, measurement and timetable didn't pick up a reliable group difference, which is less dramatic but a good deal more accurate.
Two dogs were excluded after researchers discovered they'd received more than the permitted maximum of 20 stimulations. Removing them from the analysis may have been methodologically necessary, but it doesn't remove their experience from the experiment. They were still shocked. The incident also points to a practical problem with punishment generally, since different dogs respond differently, and a level or number of shocks that suppresses behaviour in one individual may not do so in another. When the behaviour continues, increasing the repetition or intensity becomes the obvious next move, and once the ethical reassurance rests entirely on the punishment being limited, dogs who fail to respond within that limit become inconvenient rather quickly.
Electric-collar defenders often save their criticism for poor timing, excessive settings or inexperienced owners, implying the devices become fine once experts are holding them. This study undermines that reassurance rather neatly. The intervention happened within a planned research protocol, with monitoring, formal ethical approval and trainers the authors considered competent, and every dog in the electric-collar group still yelped. Correct use didn't eliminate the painful experience, it organised it. The pain wasn't an accidental consequence of someone doing it wrong. It was part of the mechanism producing the behavioural suppression in the first place.
Stopping Behaviour Is Not the Same as Teaching
A dog receives a shock and stops moving towards the lure. From the outside, the outcome looks clear enough: the unwanted behaviour has gone. The disappearance of visible behaviour, though, doesn't prove the dog understands what we want, feels safe, or has learnt a useful alternative. All it tells us is that continuing has become costly enough to avoid, which is suppression rather than education.
In behavioural terms, adding a shock to reduce chasing is called positive punishment. Positive doesn't mean pleasant here, it just means something has been added. If the unpleasant stimulation stops when the dog changes direction, the procedure may also involve negative reinforcement, where a behaviour becomes more likely because it switches off or avoids the aversive event. Normal people can be forgiven for suspecting the terminology was designed specifically to stop normal people understanding it. In plain English, the dog learns that one action causes or predicts something unpleasant, while another action might prevent it.
The trainer can't actually control what the dog associates with the pain. The intention is for the dog to learn that chasing caused the shock, but the dog might just as easily learn that crossing this particular area is dangerous, that the moving object is dangerous, that wearing the collar here is dangerous, that moving away from this person is dangerous, or several of these at once. Nobody debriefs the dog afterwards. During later tests, the electric-collar dogs avoided progressing in the familiar arena even when the cue was absent, and in one condition, even when the handler wasn't present. Westlund argues this fits entirely with conditioned fear or avoidance tied to the location or the lure, rather than proof of a properly trained recall. The dog who calmly chooses to return and the dog who's simply afraid to cross an invisible boundary can look identical in a results table, but they are not having the same experience.
Fear and avoidance also tend to produce impressively tidy behaviour. A dog expecting something unpleasant may stop exploring, stop moving forward, stop engaging with a stimulus altogether. Stillness is quiet, and avoidance is convenient, but neither one means the dog is actually confident. Punishment can also remove useful information, given that behaviour tells us something about a dog's motivation, emotional state, skill level or how hard the situation actually is. When a dog lunges, chases, barks or refuses, that behaviour is communicating something, and suppressing the outward sign doesn't automatically change what's driving it underneath. A dog can remain powerfully interested in chasing while also predicting pain, so the visible behaviour stops while the internal conflict may still be sitting there.
Generalisation was also incomplete. In the study's first three tests, the electric-collar dogs analysed didn't chase, but the fourth test took place in a novel arena with a novel lure, and in that condition a third of the shock-trained dogs chased anyway, meaning two-thirds held back. This is the test that actually mattered for whether the response transferred beyond the setting where the punishment happened, and the method plainly wasn't reliable once the surrounding cues changed. The lesson there isn't that stronger shocks would fix it. It's that avoidance tied to a specific context has real limits.
How Effective Training Actually Gets Built
A reliable recall needs more than the absence of running off. The dog has to recognise the cue, disengage from whatever's competing for their attention, turn, return and stay with the handler, and each of those pieces gets built separately before they're stitched together.
Work starts well away from the trigger, not next to it. A dog who can't recall away from a tennis ball on the kitchen floor has no realistic chance of recalling away from a rabbit at full pace, so training begins at a distance and difficulty the dog can actually succeed at, with the cue paired to something the individual dog genuinely values rather than whatever happens to be in the trainer's pocket. Success gets reinforced heavily and often in the early stages, and only once a stage is solid does the difficulty move up: a bit more distance, a bit more movement in the distraction, a slightly higher-value temptation. Rushing that ladder is exactly what produces dogs who "know" a cue in the garden and ignore it completely in a field.
Managing the environment matters just as much as the cue itself. Every rehearsal of the unwanted chase strengthens it, so a sensible programme limits opportunities to practise chasing while the recall is still being built, whether that means a long line, a fenced area, or simply choosing walks and times where the trigger is less likely to appear unmanaged. This isn't an admission of defeat. It's the same logic used in any skill that needs protecting from being undone by its own worst-case scenario.
A dog with a strong chase drive also benefits from having that drive given somewhere legitimate to go, rather than being asked to switch it off entirely. Structured games that involve chasing a toy, tugging, or working a scent trail on cue give the dog a way to satisfy the same underlying motivation on request, which tends to make the cue to disengage from an actual squirrel considerably more achievable, because the dog isn't being asked to give up the only outlet they have.
None of this happens instantly, and it takes more patience than fitting a collar and pressing a button. But it produces a dog who understands what's being asked, chooses to comply because compliance has been consistently worth it, and can be trusted to generalise that choice across new places and new distractions, which is precisely the outcome the shock-collar study failed to reliably demonstrate.
Put the whole picture together and the case for the shock collar gets a lot thinner. The 2024 study shows that pain suppresses behaviour quickly, which nobody disputes and which the electric lead, the choke chain and the leash jerk all demonstrated decades before it. What none of them show, this study included, is that the pain was necessary. A dog can learn to leave the lure, the rabbit or the road alone without ever being hurt to get there, given training that's built at the right pace and given the time it actually needs. Once a humane method gets a dog to the same place, the argument for the collar stops being about whether it works and starts being about how much time and skill the owner is willing to put in instead. That's a far less flattering argument, which is probably why it's the one that rarely gets made out loud.
