The Trouble with Shock Collars: A Real World Example


Photo credit: Schill via Foter.com / CC BY-NC

A family in my neighborhood welcomed a Great Dane puppy into their lives last fall. The once-gangly, now-100+-lb grey dog regularly greets me and my dog while we walk by his yard with booming barks. This situation has apparently become unacceptable for his owners: this week we have not been regaled with the hound’s vocalizations and he is sporting a shiny shock collar high on his neck.

The training method that shock collars, and all punitive training methods, rely on the operant conditioning strategy of positive punishment. The principal is simple: the dog does something the owner does not like and receives the shock – thereby decreasing the likelihood of the dog to repeat that undesired behavior in order to avoid the aversive consequence. For operant conditioning to be successful, two criteria must be met: timing and persuasiveness. These two criteria are precisely why punitive training methods like shock collars are fraught with trouble – and I’ll tell you why:

Shock collars are not smart.  When using operant conditioning, the trainer’s timing must be impeccable in order to link the correct behavior to the correct consequence, because animals do not possess the mental capacity to link consequences of behaviors that are more than a few seconds apart. Is my neighbor’s dog linking the shock with his barking behavior – or the appearance of another dog? Or a squirrel that just happened to be running down the street at the same time? Or the truck that just drove by?

All types of operant conditioning always involves a little trial and error on the animal’s part. Using aversive consequences for training causes stress in dogs and can likely contribute to the development anxiety disorders, aggression, and a host of other behavioral problems.

Shock collars are not safe. The other vital criteria for successful operant condition is that the consequence of the animal’s behavior must be compelling enough to actually change the animal’s behavior. This is why many dog trainers instruct owners to turn up their dog’s shock collar, despite contrary instructions from the manufacturer.

So what if my neighbor’s dog is connecting the significant shocks to the appearance of me and/or my dog? Well, if human +/- dog = pain, it will behoove the dog to prevent humans and/or dogs from getting near to him – so he will become aggressive towards walkers and their dogs.

Shock collars are not necessary. While I’m having to re-route my walking path due to my safety concerns about my neighbor’s dog (which is actually quite the ordeal for my OCD-anxiety dog), I must admit that I am frustrated by my neighbor’s numerous decisions leading up to this point:

  1. They purchased a dog too large to be adequately contained in their yard
  2. They leave him outside unattended
  3. They are upset that a dog bred specifically for its guarding skills barks at strangers
  4. They chose to install a shock collar on him rather than safely train him to tolerate walkers

Any one of these decisions could have been made differently and our whole neighborhood would be safer! Of course, the welfare of the dog is also a significant concern on mine so I would prefer that my neighbors would train their dog not to bark in a manner that will not cause pain and anxiety.

Journal Article: Where’s the Stress From?

Title: Comparison of behavioral and physiological responses of dogs wearing two different types of collar.

Authors: Philip Ogburn, Stephanie Crouse, Frank Martin, Katherine Houpt

Published: February 1998 in Applied Animal Behavior Science

General overview: This research compared behavioral (head position, ear position, tail position and posture) and physiological (blood pressure, pulse rate, respiratory rate and pupil diameter) responses of dogs wearing a head collar (like a Gentle Leader collar) and a buckle collar during brief obedience trials (walking, sitting and turning). The authors found that while there was no significant difference between the physiological responses of dogs to the different collars, the buckle collar resulted in more unruly behaviors (i.e. leash pulling) and the head collar resulted in more pawing at the head and ears and less eye contact with the handler.

My comments:

I noticed a major flaw with this paper right at the beginning of the Methods section: their study population was poorly defined. I believe these dogs are unowned dogs (perhaps from an animal control facility?) because the authors mention that each dog experienced identical housing, feeding, animal care personnel and exercising routines. Furthermore, the authors say that they have no history on each dogs’ previous exposure to head collars but to their knowledge, none of the dogs have worn a head collar before. So…with no history, it’s safe to assume the none of the dogs had ever worn a head collar? That seems like an unscientific conclusion to me.

The researchers’ methodology doesn’t seem awful (each dog was used as its own control: all dogs were tested with a head collar and with a buckle collar, randomly assigned to be test with either collar first). Some of the terminology in the paper seems outdated: the researchers deem a dog “subordinate” or “dominate” based on body posture at the end of testing (to my knowledge, it’s not appropriate to label a dog “subordinate” or “dominate” because those terms describe reactions to situations, not overall temperaments). The authors also reference a few ecological studies that have since been highly refuted to legitimize the way head collars work. But this paper is sixteen years old!

The biggest red flag in this paper is the combination of the findings that no physiological difference was detected between collars but an overall trend of diminishing physiological responses during the test. Since all dogs experienced the same care routines, I believe it is safe to say these dogs were being housed in a kennel situation. Kennels can be extremely stressful housing environment for dogs because they are unfamiliar (to many dogs), don’t allow for much positive human contact and involve many foreign sounds and smells. The fact that blood pressure and respiratory rate and etc. went down during the test regardless of treatment or control might be the result of the dog being removed from their kennel and having a calm interaction with a human.

I was looking forward to reading this study because while I have read papers that looked worse alternatives to buckle collars, this is the first that I’ve seen evaluate a potentially better substitute. Because of the substantially under-described study population, however, I fail to see how this paper contributes all that much besides evidence that dogs wearing head collars do indeed paw their head/nose more than those wearing buckle collars, and that buckle collar-wearing dogs pull more.

Journal Article: Shock Collars Cause Stress and Unintended Consequences

Title:  Training dogs with help of the shock collar: short and long term behavioural effects

Authors: Matthijs B.H. Schilder, Joanne A.M. van der Borg

Published: March 25, 2004 in Applied Animal Behavior Science

General overview: The authors studied video recordings of dogs in training for police or guard use to compare behavioral responses of dogs trained with remote electronic collars (e-collars) and dogs that were trained without e-collars. The authors noted that the training protocols used for all dogs included punitive techniques such as choke or prong collars, kicks and beatings. The authors found statistically significant increases in behaviors associated with stress (tongue flicking, tucked tail posture and averted ear position) in dogs trained with the e-collars than the dogs trained without the e-collars. The authors also found that dogs trained with e-collars may associate training, the training area and their trainer with stressful stimuli, which is important because it meant that the dogs were not necessarily associating the aversive stimuli of the e-collar with the behavior that the trainer was trying to prevent.

My comments:

Especially when compared to this UK study of e-collar training, this study was not as rigorous as I might have appreciated. Although the authors had a very detailed ethogram (an inventory of behaviors or actions exhibited by an animal) to catalog the study dogs’ responses, the researchers themselves reviewed the recordings. This method is clearly susceptible to bias, because the researchers would have been to tell which dogs were being trained with a bulky e-collar and without one. I wish the researchers had an independent observer score the dog training videos according to their ethogram and performed the analysis on that data. I can’t put much stock in their comparative findings because, whether intentional or not, the researchers may have noticed stress-related behaviors in the dogs they knew had e-collars on if they presumed that exposure to an e-collar might cause increased stress in the dog.

However, I do find the conclusions that dogs trained with e-collars appeared to associate their trainer, the training area and being given commands with stressful stimuli very interesting. From my understanding, e-collars work on the principal of avoidance conditioning: a subject learns a behavior to avoid a painful stimuli.

Here’s why I think this is interesting: the author mentions that most shocks were given to dogs when they did not obey the “let go” command. Using this command as an example, the point of shocking the dog when he (most study dogs were male) did not obey the command “let go” is that the dog learns not obeying the command results in a shock. Instead, it appears that the dogs may have learned that being near the trainer, being the training yard or being given a command results in a shock. If this is true, the e-collars were not doing what they are intended to do.

I found this article to be a fascinating read also because of its commentary on training techniques and breeding standards of guard and police dogs. The description of the “harsh” training protocols is rather horrifying: dogs are taught to bite well before they are taught to “let go”, subject to beatings and kicks and show visible signs of distress just because their trainer is present. The authors comment on the dogs bred for such jobs as purposefully selected to be highly excitable, temperamental and with low biting thresholds. It made me question the ethicalness of using dogs for police or guard work, for both the quality of life of the dog and the safety of communities within which these dogs work and reside.

This article was discussed, in great detail, on Dr. Sophia Yin’s blog.

Journal Article: Shock Collar Trainers Don’t Follow Device Instructions with Questionable Welfare Implications

Title: The Welfare Consequences and Efficacy of Training Pet Dogs with Remote Electronic Training Collars in Comparison to Reward Based Training

Authors: Jonathan J. Cooper, Nina Cracknell, Jessica Hardiman, Hannah Wright, Daniel Mills

Published: September 3, 2014 in PLOS ONE 

General overview: This article studied the welfare impacts and efficacy of remote electronic training collars (e-collars) in comparison to positive-reinforcement based training. After a preliminary study of 9 dogs to determine method feasibility, the authors compared three groups of 21 dogs each (63 dogs total): Group A (trained using e-collars), Group B (trained by the same trainers as Group A but without e-collars) and Group C (trained without e-collars by trainers who used positive reinforcement-based training protocols). Data collected from the three groups included video recordings of dog behavior, which were evaluated by an independent team of trained observers and salivary and urinary cortisol* levels. Behavioral and cortisol differences were marked between the groups in the preliminary study while behavioral differences only were significant in the larger study. Cortisol levels were elevated in the preliminary study in dogs trained with the e-collars. Stress and displacement behaviors were seen more frequently in dogs trained with the e-collars in both the preliminary and larger study.
*Cortisol is a hormone that is released in association with stress.
My comments: 
I first heard of this study on Science Daily, and it was also reported on the DogingtonPost and DogTime. The revelatory lead sentence to the Science Daily story was, “…the immediate effects of training pet dogs with an electronic collar cause behavioural signs of distress, particularly when used at high settings.” Who would have guessed that giving a dog an electric zap would cause distress? I would have thought that was common sense but I decided to read the study understand the details.
The purpose of this study was to provide evidence to support or detract from the argument that, when used according to the manufacturer’s instructions, e-collars are effective tools for avoidance conditioning (learning a behavior to avoid an aversive stimulus). Furthermore, the authors point out that there have been studies on the immediate impacts of e-collars in dog training but they have been restricted to specific populations of dogs, like police dogs or laboratory dogs, and the generalizability of those findings to the general population of dogs is questionable.
From my perspective, the methodology of this study is sound. The authors did a great job defining the study group and control groups as well as the way data was collected and evaluated. The use of the three groups controlled for differences between trainers with diverse opinions about positive vs. aversive training techniques. The use of independent observers for the recorded behavior of the test dogs avoided bias for behavioral responses. I can’t think of any potential confounding aspects of this study that weren’t addressed by the authors.
I understand that this study was novel in its study population (“regular” dogs) but I don’t think it’s particularly interesting that the authors found use of e-collars during training was associated with stress and displacement behaviors in dogs. I’ve read a few other studies about aversive training techniques and behaviors like yelping, moving away from the trainer and lowered ears all seem to be common and expected behaviors with that type of training.
I do think it was interesting that all but one trainer who used e-collars regularly did not follow the manufacturer’s instruction to use the lowest setting that the dog responded to – the majority of these trainers set the devices on the highest or nearly the highest setting right away. The oddest part about this is that the trainers were nominated by the Electronic Collar Manufacturers Association. Why did these trainers ignore their own professional association’s directions?
I’m not sure why there were differences in cortisol levels during the preliminary study between e-collar exposure groups and not in the larger study. The authors noted that it may have been due to the time frame differences between the short preliminary study and the several months that the larger study took, but I don’t know enough about cortisol levels in dogs to say whether or not this is a valid argument.
The major finding of this study is that there was no difference in training efficacy between the three control groups: positive reinforcement based training was just as effective as more aversive-based training. The authors note that positive reinforcement training protocols are completely void of the questionable welfare implications of aversive training techniques like e-collars.