Punishment, Fallout and Recovery in canine training.
Updated: Aug 27
Is coercion, as a training fundamental, really so wrong?
If animals learn to avoid objects/behaviors through aversive consequences, then how can utilising aversives in training be wrong or ineffective?
Debunking the “myths” and the “science”.
Introduction
Advocates for “balanced canine training” – commonly based on the misinterpretation (or over-interpretation) of Skinner’s Behaviourist Model of learning[1] - go to extraordinary lengths to demonstrate that punishment is not detrimental in the long term, as the animal recovers and, further, that punishment is not just effective, but essential, as a canine training tool. In summary, this line of argument making the case in favor of punishment with no lasting fallout – which we will expand upon – overlooks several fundamental issues, importantly:
· The very definition of coercion through aversives.
· The impact of breed, personality and past experience on canine reaction to a stimulus.
· The concept of teaching right from wrong – can punishment teach “constructive adaption” – that is an alternative action which would be rewarding?
· The fallacy of extrapolating the behavioural reactions evidenced in caged rat experiments, especially employing electric shocks to generate behaviors, as “scientific proof” of canine behavior, just as human behaviour and reactions – thought processes – are commonly postulated to be reflected by dogs.
· The fallout – or long-term effects – of punishment are only poorly demonstrated by short term experiments on any animal, not only because of interspecies differences, but because of the inability to measure levels of sensitivity, reactivity and the canine-human relationship following punishment episodes.
· The canine genetically engineered desire to be a companion – to befriend, to remain close to and to be responsive to human guardians – guides a dog’s desire to please, willingness to respond to training and resilience in the face of applied aversives.
To begin with the fundamentals, we need to discuss what constitutes “coercion and punishment with aversives” in the canine training arena. This is not a simple issue, as the actual aversive, particularly its amplitude/severity and timing, as well as the canine’s breed, nature and past experience, all contribute to the impact and the outcome.
The science and the myths
1. The aversives definition, right from wrong and animal experience
a. Definition of aversives in the coercion paradigm
As discussed in companion articles on this website (specifically Aligning "positive reinforcement" with "use of aversives" dog trainers and Reinforcement vs punishment: A dog trainer's perspective) the definition is not as simple as “aversive means aversive”[2]. AVSAB (American Veterinary Behaviourist Society)[3] do provide a fairly comprehensive definition of aversive in the dog behaviour context as “anything unpleasant (emotionally or physically) that is used to decrease an unwanted behaviour”[4]. This is a good general guide, but not perfect, nor religeously adhered to here, as the tendency of the force free/positive reinforcement only fraternity is to look at extremes.
We need to recognise that aversives can range from a tight leash momentarily because the dog pulls ahead of the trainer, the trainer cuing a verbal “No”, through to some explicit forceful action by the trainer, such as a “leash pop/jerk” (in the extreme using a slip leash , choke chain or prong collar and jerking a dog until its airway is constricted or it is pulled off its feet with the leash). For simplicity and to get to the main lines of argument in this paper, we will go with the rationale expressed in the afore-referenced articles of “acceptable aversive” and what aversives are absolutely “unacceptable”?[5]
Whatever definition is preferred, aversives can generally be regarded as part of a “coercive training regime”, although note that in the realm of playing semantics, the fundamental coercion technique of “escape and avoidance” training has elsewhere been labelled “relief training” as a nicer sounding description[6], despite much of the “theory and science” being based on rather callous experiments on “traumatic avoidance learning”[7]. The question this raises is “Why use aversive pressure to utilise escape-avoidance, when canines read body language first and foremost, so surely hand and body gestures (including the familiar techniques of luring and shaping) can offer a successful (and preferred) training path?” The insight, unfortunately, is that humans favour coercion as the go-to technique to impose their will. The answer offered by those favoring the need for punishment (i.e., an insistence on using “all four Skinner Quadrants”[8]) is that following the 1980’s “Inducive Training Revolution” and the replacement of coercion as the core dog training technique, the “failure” was to move from the fluency stage of training[9] to “aversive control”, postulating that, for example, being lured into a “Down” position is completely different to a commanded “Down” under pressure – believing that although topographically the same skill set, they are psychologically completely different[10]. This very thin argument completely overlooks the elements of fear, escape and avoidance, Pavlovian responses, Operant Conditioning and muscle memory, thereby offering a poor excuse to have to overlay force and punishment on a positively reinforced training program to achieve results – and may lead to anxiety, reflexive stiffness and even aggression in the dog – something MWD (Military Working Dog) trainers figured out many years ago[11].
b. Right from wrong
An argument favored by adherents to an over-interpretation of Skinner’s Behaviourist Model advocate that if a trainer does not punish a dog, how will it know what it shouldn’t do? This has been addressed in the companion article on this website The Operant Conditioning Model Fallacy, so to keep it brief here, the answer to this lies in replacing the word “punishment” with “redirection” steering the dog away from what not to do and showing it what to do instead, without a role for punishment in that equation.
Redirection is not simply a “correction - which is another word for punishment, or at least a correction which grades into a punishment”; “redirection is about teaching for success in developing a skill”, adding the concept of “restructure of the training” and replacing the incorrect or undesirable action/behavior with the desirable/favored action/behavior – showing the dog what success looks like[12]. Only the desired action/behavior is rewarded, increasing the likelihood of that action being repeated – the unfavorable/undesirable one is not rewarded – it is ignored or possibly addressed with a calm “No” followed by the structuring of the desired behavioural outcome, reducing the animal’s propensity to repeat the unwanted action. It is important to appreciate that the cued response of “No” (apart from those religiously “force free” who classify “No” as an aversive) can be delivered as an aversive, a forewarning of the dire consequences of punishment, or it can preferably be delivered non-coercively as a simple marker, a signal that positive reinforcement will follow a different, preferred, action. Now, some may regard this "redirection protocol" as simply negative punishment - and that would be a fair conclusion on the basis that something is taken away ("negatived" - that is, no reward) to reduce the likelihood of that undesirable action reoccuring (the definition of "punishment"). However, the important point is that the desired action is shown/demonstrated/taught and learned - and therefore rewarded (positive reinforcement).
Note that in more urgent situations, some may refer to the trainer administering a “positive interrupter”- rapidly halting the undesirable action so as to implement the redirection - with no need to follow this with positive punishment – very far removed from the aversive auditory hiss, squirt of water or finger poke.
So, put simply, “No” does not need to mean “stop - because I’m going to choke you or whack you”, it can be taught to mean “pause - try something else – a different action which will bring a reward”[13].
c. Aversives in “life experience”
Canine handlers in favor of applying aversives in the training regime commonly make the argument that aversives constitute an essential part of life’s experiences to learn the difference between “safety and danger”, or “what actions will cause pain and which will give pleasure”. This sounds a lot like “life wasn’t meant to be easy – it’s tough out there – so we need to include punishment as a key ingredient in any training regime to ensure normality”. This is a classic example of twisted logic – the fallacy (or "logical fallacy"- sometimes referred to as “junk cognition”) that two things which, on the surface, seem to be related and parallel – the one demonstrating the other – are in reality separable parts of learning, whether one leans on the foundation of Behaviorism or of Experiential Learning.
Canine training generally relies on the “motivational control of goal-directed action”[14], also referred to as instrumental learning, which builds the knowledge of the “instrumental contingency” between action and outcome, that is, with an outcome deliberately associated with an action, so the outcome (usually a primary incentive) becomes the goal, at least during the acquisition phase of training.
Now, is there any possible common thread - any consilience or concordance - of the two examples of “aversives in the life experience learning” (Experiential Learning) and the “trainer applied use of aversives in coercive learning” (Instrumental Learning)? Certainly, we can envisage commonalities, but therein lies the “proof of logic trap”. We first need to distinguish between "random happenstance" - that is an accidental pain infliction, such as tripping over - and a "deliberately engineered act" - such as poking a cactus bush or attacking a spiny anteater, porcupine or echidna. In either case, because an animal experiences an aversive by performing a certain action – such as a torn muscle after jumping from a height, walking on sharp rocks, eating a foul tasting, noxious/toxic plant or carrion, or attacking a porcupine, so that in future it avoids such actions (which it probably won't) does not prove that applying aversives in a canine training regime is essential, nor even effective. Not to belabor the point, but because this argument raises it head so often, the “logic link” breaks down for two reasons.
The first is that, in the “life situation", the canine has created the event delivering the aversive outcome through its natural actions related to its movements – not as part of coercion to supposedly learn what a trainer desires. A dog which turns it head suddenly and bangs into a tree or a wall, will no more avoid those actions in future any more than stubbing a toe prevents a human ever doing that again (a "random happenstance" - recognizing the aforementioned danger of drawing the human-canine parallel with “disjointed logic”, just to illustrate the point).
The second is that the often referenced “hot pot/stove burner” anthropomorphic analogy (“people learn not to touch a hot pot from a single painful burn experience” - a "deliberately engineered act") - although seemingly simple and logical for humans – is severely flawed. The canine having such an experience will most likely become wary of all pots/stove burners - the canine will not just avoid such hot objects, but all look-alike objects, as they will not be able to envisage the distinction between the two. This latter was exemplified in the behavior training Hearne[15] indulged in, wherein, following the coercive canine training methods of Koehler[16], to stop her dog from digging holes in her garden, she had the dog watch her gleefully further dig out the hole, then fill it with water, then hold the dog’s head in the hole under the water. The result? The dog became scared of all holes anywhere – on a walk the dog would skirt suspiciously/ fearfully around any depression in the ground. Hearne was delighted with her thorough success and, yes, the dog stopped digging holes in the garden – but ask yourself, is this the relationship and training regime you want with your dog?
A variation in this "deliberately engineered act" category is the canine chasing a prey animal and being hurt by running into a fence or being damaged by the animal it catches - numerous illustrations of this not deterring that behavior in future (see discussion in the companion article on this website Why use electric shock collars? A plea for enlightenment.)
So, the argument is not that a dog should never experience stress or hurt of any kind in its life. The argument is, this does not mean that adding stress through aversive actions in a coercive canine training regime is justified by “the natural stresses of life experiences”, nor that adding aversives is a better teaching method than aiming for positive reinforcement outcomes, teaching the dog the desired behaviors and actions.
2. The caged rat “scientific proof” of canine psychology and behavior
Canine trainers who attempt to educate dog owners about the use of punishment in dog behavior modification or obedience outcomes, favour references to caged rat or pigeon experiments[17]. The fundamentals of the caged rat experiments (also commonly using pigeons) to demonstrate and statistically record reactions and behavior modification, rely on a certain action, such as pressing a lever (or pecking a button) delivering a reward (such as a food pellet) or causing a punishment, commonly a shock delivered through the electrified floor of the cage. Before we look at the deductions drawn from such experimental data and extrapolated to the canine training world, we need to revisit the applicability of rat – or pigeon - behaviors to any other species, particularly dogs.
In what might seem like a seminal paper, recording animal experiments using Operant Conditioning (Instrumental Conditioning)[18] that did not go the way the designers hoped with rats, pigeons, chickens and other species, such as raccoons and marine life (porpoises and whales) as the subjects, Breland and Breland[19] recognised that not only did different species they experimented with have unexpected reactions or behaviors in response to the instrumental conditioning/training applied, but a significant proportion of subjects (20% in the case of chickens) failed to respond to the experimental design imperatives. This was explained principally with reference to “instinctive behavioral drift”, that is reverting to the natural behaviors of the subject species, or (quote) “learned behavior drifts towards instinctive behavior”. Paraphrasing the most significant conclusions of the Breland and Breland observations: (1) no animal subject starts out as a “blank slate”, (2) species differences are not insignificant and (3) not all behavioral responses are equally “conditionable” to all specified stimuli. Hence, the behaviors of any species cannot adequately be understood, predicted or controlled without knowledge of its instinctive patterns, evolutionary history and ecological niche.
Now, these conclusions might seem either revolutionary or obvious (depending on the reader’s knowledge) but they were previously observed and highlighted a decade earlier by such ethologists as Lorenz[20] and Tinbergen[21]. The latter acknowledged gene related behavioral “taxonomic characteristics” and Lorenz[22] particularly observed that stimuli (which might be an operant conditioning attempt to generate or reinforce a particular behavior) can in reality serve to "unlock" or “release” an innate/instinctive reaction and decided that behavior is a taxonomic character and therefore a diagnostic trait of a particular species, not synchronous nor transferrable across species. At a much later time, Pellis and Pellis[23] observed that lumping observed behaviors into a single category may seem to be useful, but runs the risk of “pigeonholing behaviors that only really make sense when species are compared within a clade of related species”.
Additional to these observations about the predictability of behavioral responses in one species based on “scientific observations” of a different species – given the various innate/instinctive behavioral reactions shown by each different species, one also needs to consider Pavlovian Interference and Pavlovian Instrumental Transfer, which will impact how an animal responds to a given stimulus (to appreciate these elements, refer to the companion article on this website Are you confusing your dog? Canine learning interference as well as Holland [23A] and Mason [23B]).
In summary, the lesson here is not simply that “sometimes operant conditioning doesn’t work so well”, but rather, although acknowledging that certain continuities between species do exist, that “the influence of species and individual factors within a species (breed differences, intellectual differences, experiential learning variations, Pavlovian conditioning, etc.) can interfere with and even override instrumental instruction and expected behavioral outcomes”. The caution is that laboratory based experimental results from one species may well provide an indication of behavioral types (and we will explore results from some of these experiments further below) but not “norms”. Beware of the inter-species, one-to-one, direct correlation of what appears to be “empirical scientific behavioral evidence” – it may not actually prove anything relevant, especially not to canine training, so bear that in mind as we delve into some of the experiments pro-punishment canine trainers like to quote.
The bottom line? “Scientific evidence” from experiments on caged rats or pigeons does not tell us everything we need to understand about canine psychology, natural reactions and training - and might even be leading us to false conclusions.
The long term, or even permanent, effects of aversive experiences causing stresses building emotional and mental distress (lasting or recurring anxiety/sorrow/pain/upset/ unhappiness) can generally be well understood across the animal (canines, humans and other species) fields of behavioral psychology. The analysis of “coercion”[24] has added the descriptor of “fallout”, somewhat analogous to “nuclear fallout”, under the particular thesis of Sidman[25], notably in regard to human behavior. Again, being wary of the “inter-species analogy trap”, testing, analyzing and describing punishment as a regular part of a canine training regime or behavior modification raises similar questions surrounding our understanding of “fallout”. What deleterious effects can punishment, as an applied training tool have, under what level/severity, timing and regularity of aversive treatment will the fallout be real, noticeable and lasting and then, just how long will the effects actually last? All of these questions can be scrutinized by reference to empirical evidence from species other than canines (including humans) but nonetheless is certainly evidenced in canines, particularly through body language and undesirable behaviors under certain circumstances and/or reactions to various stimuli.
b. Punishment research
Estes and Skinner[26] used electric shocks, preceded and, therefore, over several repetitions, predicted, by a “conditioning” sound. The shock - and subsequently the sound pulse - become a “disturbing stimulus”, disrupting the rats previously learned positive lever pressing behavior, demonstrating the establishment of a condition of anxiety in the subjects. The deduction was that anxiety affects the rate of responding, and hence had a detrimental and lasting impact on previously conditioned responses. Now, one might conclude, therefore, that positive punishment can supress behaviors. The two flaws in that conclusion on a stand-alone basis are that (a) the punishment can also suppress desirable (conditioned/trained) behaviors (that is in fact precisely what the experiments did demonstrate) and (b) that the punishment alone does not offer an alternative behavior, but rather a tendency to “do nothing”, an outcome also demonstrated in these rat experiments.
However, there is an important caveat. That is, in these experiments the “warning sound” heralding a shock was independent of any action the rats had taken, so this is referred to (arguably semantically misleading) as “Conditioned Suppression” – non-contingent upon action/behavior – also called a “Conditioned Emotional Response” (“CER” – for those who favour making the description appear more scientific or technical through the use of acronyms). So, the aversive shock is response/action/behavior independent – just “coming out of the blue”[27] for no apparent reason. Estes[28] found that both categories of aversives – dependant and independent of the subject’s prior action – did suppress the targeted future response/behavior/action, but suggested that punishment is effective only if it elicits responses that are incompatible with the punished target action[29]. This might all seem a little confusing in terms of exactly what punishment does and how effective it can be, but under any interpretation, the warning to the canine training world is clear - the grumpy, impatient and frustrated handler who appears (to the dog anyway) to be issuing random punishments (in the world in which punishment is part of the tool kit for incorrect actions) may delude themselves that the canine’s behavior “improves” in alignment with this Estes’ theory. But the more effective – and humane - training protocol is to offer the dog an alternate desired behavior which is rewarded, rather than just letting the dog choose its own response, which may be to bite the trainer or to shut down and do nothing[30].
Azrin and Holz[31] decided that response suppression, as a result of punishment, is actually more complex than a simple and direct behavior-punishment correlation and that multiple interdependencies arise. Focussing this broad conclusion on canine training, the questions remain whether this response suppression or fallout of punishment (a deliberately applied aversive) is (a) generalised – for example to the environment in which it occurs/is applied and/or to the entity administering the punishment (the trainer) and (b) will this effect of suppression fade over time – is it transient? The answers relate to two elements, viz, (a) the resilience (“bounce back”) of the species/individual and (b) the availability/application of countering reinforcement/rewards for the preferred/replacement behavior.
c. The canine punishment justification
Referring back to the “evidence” from caged rat experiments, those favouring punishment in canine training claim that these results:
1. Clearly demonstrate that punishment works to suppress designated actions[32]
2. That although “fallout” impacts the training regime in some way (commonly related to a Pavlovian association) these effects are transient and will fade or resolve in time – a process facilitated/sped up by offering alternative actions to replace the undesirable behavior.
To amplify this latter point of lessening fallout, these same pundits will also commonly argue that positive punishment is “most effective” when:
· the target behavior is novel and not entrenched - the behavior has not developed into such a habit that it becomes insensitive to “instrumental reprogramming”. This is referred to as “Pavlovian Instrumental Transfer” – because the undesirable reaction/behavior becomes impulsive and “uncontrolled”,
· the aversive chosen is new and introduced suddenly,
· the timing of punishment application is timely (immediately following the undesirable action/behavior),
· when alternative behavior is reinforced,
· when the undesirable behavior is not produced out of fear or in self-defence i.e. not acquired traumatically[33],
· (to which we might add) is not a primal behavior, such as the predatory sequence.
Taken together, these constraints on the effective use of aversives suggest that;
· punishment won’t work well on entrenched undesirable behaviors (the “big” issues),
· that the use of extreme “training tools”- such as prong and shock collars – won’t work universally and will not be effective in teaching a dog the preferred behavior,
· the use of – or combination with – positive reinforcement is the preferred path to effective canine training.
These conclusions are so self-evident that one wonders why so many dog trainers waste so much time justifying positive punishment as a key training tool, following the pathways of caged rat and pigeon experiments to prove what to many are self-evident outcomes. However, those researchers ever seeking more answers and understanding of escape, avoidance and the apparent persistence or transience of fear in animals responding to electric shocks, naturally took this one step further, as explained below.
d. Are fallout effects transient and extinguishable?
Various experiments on dogs were conducted in the 1950’s (e.g., Solomon, et al and Solomon and Wynne)[34] which purportedly demonstrated that, in experiments involving canine escape and avoidance learning through the application of caging and severe electric shocks[35] with an escape route available, dogs gradually “acclimatized” to the punishment and showed signs of less fear. Now, fortunately for the subjects, these types of harsh and inhumane experiments are no longer conducted in formal research (although arguably implemented by certain punishment-oriented dog trainers) however these seventy-year-old experimental results are still quoted as some form of proof that:
(a) punishment is just fine and
(b) that “fallout” is non-persistent and therefore not a big concern
Neither conclusion is consistent with what Seligman and others covered in dissecting “learned helplessness”[36] and are inconsistent with numerous research findings that demonstrate that although circumstances leading to extinction as a response-weakening process that dampens the original association, it does not eliminate it (see discussion in Mason: Footnote 23B). To emphasise this point, the fallout transience conclusion - that apparent suppression of fear expression - ignores the research which suggests that behavioral, autonomic and endocrine responses to fear and stress does not erase the original memory of the fear inducing stimulus. Recovery and reinstatement of the fear state can occur soon after the supposed extinction.[38]
Furthermore, although punishing bad behavior or undesirable actions may seem to terminate those undesirable behaviors/actions, this does not teach replacement good/desirable behavior/actions. Sidman[39] describes experiments in which punishment suppresses a learned behavior which had previously delivered reinforcement through a primary reward. However, if no alternative behavior will deliver that primary reward, that suppression is overcome and the “undesirable” behavior is repeated despite predictable and ongoing punishment, because there is no alternative means of gaining access to that desirable reinforcer.
Even a moment’s thought will suggest that, on the one hand, it is not altogether surprising that dogs in the aforementioned experiments showed lesser anxiety signs over time and experimental repetitions, figuring that this was their lot and in certain tests they could learn avoidance behaviors. Then, on the other hand, what modern canine trainer is going to argue that they can continue to practice escape and avoidance training coupled with sometimes severe punishment because the dog will “get over it”, showing no visible emotional response[37] and that this treatment will not have a lasting effect on other canine behaviors or reactions, or indeed, the human-canine bond?
e. Fear Learning and the Warning and Safety Signals theories
In a further attempt to demonstrate that a canine training regime of punishment and induced fear is not just OK - due to good outcomes and no lasting effects (disputed here on both fronts in canine training applications) - a body of research has attempted to elucidate the interplay and outcomes of “warning signals” and “safety signals”. The warning signal predicts eminent onset of punishment (in experiments usually electric shock) and the safety signal predicts that the danger of that punishment either will pass momentarily (a forward predictor) or has now passed (theorized as a reinforcer by backward signaling of safety)[40]. The breathtaking conclusion some draw from these studies is that both types of signals can be viewed as a positive reinforcement effect - BUT – this can also be interpreted simply as the knowledge and certainty of what happens next via the warning signal (the element of control coming from predictability[41]), or the euphoria of relief that follows escape and avoidance or cessation of the punishment, or risk of punishment. That is, any reinforcing element may be related to the adrenaline rush following survival of natural or self-imposed risk (the “fear relief” of Konorski as well as Domjan and Burkhard[42]). This is arguably the basis of escape/avoidance training and negative reinforcement training techniques.
Now, here is a rather dramatic inference for the canine training and companionship arena. If a canine guardian/handler is predisposed to using punishment as a key ingredient in their regime, then not only might they be attempting to impose or alter a mental state and behavior which is evolutionary in perspective and automatic in response (innate/instinctive) – in the case of fear deeply seated in the amygdala (or lizard brain) – any “reprogramming” attempt will be unresponsive to cognitive learning[43]. Furthermore, the very approach of the guardian may act as a warning signal activating defensive behavior, just as their departure may signal safety.
Incidentally, in experiments on caged animals using punishment if a certain learned action is not taken, the animal may do little else apart from remaining at the required location ready to perform the specified act to avoid further punishment. Surely such forcefully acquired behavior is not just a block to thinking, exploring and new learnings, but inhumane into the bargain.
An interesting aspect of “fear learning” is the research – on animals and human children – into “vicarious learning” (and here one may draw parallels with “warning signals” discussed above). In simple terms, vicarious learning is the acquisition of fear via observation of the fearful responses of others, even beyond the primary stimulus to a second order stimulus[44]. This is a complex research area relying on neurology, Pavlovian Conditioning, innate danger stimuli linked to the neural circuitry of the amygdala – likely of an evolutionary origin[45] - however there is a relevance to canine training. Firstly, using punishment as a part of a training regime in the presence of other dogs in the training session may well impact how the canine reacts in turn to the cues or actions they are being tasked with. Secondly, fear freezing, escape or attack may result – the very hallmarks of escape and avoidance training. Further, such fear learning, whether primary or vicarious, may be difficult to reverse as it is emotional and disassociated from cognitive appreciation of stimuli context.
4. The canine desire to be a companion
Canine handlers may argue that using punishment simply teaches the dog what not to do and, in any case, the punished dog “bounces back” and is either restrained in their behavior (apparently obedient) or even friendly towards the handler. There are at least three challenges to this false logic.
Firstly, a punished dog will look to prevent, or escape, the aversive situation by flight. If no escape is possible, then the best defense is offense – i.e., attack whatever appears to be causing the pain – the handler, a nearby dog or even the apparatus through which the punishment or constraint on flight is administered.
Secondly, if the dog collapses into a state of no action whatsoever, taking on a “cannot escape this punishment so will just not move”, this is the very definition of learned helplessness. Unfortunately, uninformed dog trainers will think this is successful retraining and behavior modification.
Thirdly, dogs have been bred for thousands of years to be the companions of human beings, so there is an innate and compelling desire to cooperate and “remain friends” - and so to gain access to the primary needs of safety, shelter and food. It is not surprising that dogs “bounce back” from punishment.
5. Does punishment work in canine training?
Modern dog training pundit’s efforts to prove that an essential ingredient of canine training is applying a coercive/punishment/escape-avoidance regime are really redundant, the proponents being guilty of adhering to an ideology, doctrine, belief, or predisposition - without considering modern advances in canine teaching – and making abundant (and redundant?) social media claims about the complex variety of situations and animal behaviors wherein aversives are the very best tool. The fact is that coercive canine training and behavior change techniques, using punishment as a key tool, have been aptly demonstrated decades ago by Most and Koehler[46] and illustrated in the reference to Hearne above. We need to appreciate that dog training understanding and philosophy have migrated from those past times, through many intermediate steps (such as the well-meaning Woodhouse[47] moving towards positive reinforcement, through to the Cesar Milan[48] methods retaining force, escape and avoidance based on outdated alpha/pack leadership theory of dominance) to the present confusion between coercive, balanced, reinforcement and force free trainer ideologies[49]. Fortunately, for the canine species, the preference has moved from coercion to a more enlightened preponderance of positive reinforcement – although the questions surrounding “is there such a thing as positive reinforcement only” and “is there really such a thing as purely force free” are the subjects of an extended debate (touched on in the companion article of this website The Operant Conditioning Model Fallacy).
So, does punishment work in canine training? Yes, of course, it does - dogs are not stupid – why keep acting out a behavior or action if they get punished for it – provided they (a) are given a path to an alternate acceptable and rewarded behavior and (b) that the “root cause” of their undesirable behavior (such as a past trauma) is understood and can be reprogrammed in the canine brain[50]. In fact, in the face of punishment as a standalone tool, the easiest action is “do nothing”- so often evidenced in aversive dog training social media videos.
For the modern canine trainer, the questions become;
(a) What training regime and human-canine relationship does a guardian want?
(b) What level of “aversive” (remembering the acceptable/unacceptable debate) is actually helpful or required to achieve the training objective?
(c) Will a system of “redirection and reward” (as discussed above) do the job?
(d) Is a firm “No – leave it” justifiable in certain circumstances and is it going to work once properly trained starting with positive reinforcement?
(e) Is leash restraint justified and appropriate?
(f) Do you really need an electric collar, choke chain or physical punishment and, if so, Why? (see companion article on this website Why use electric shock collars? A plea for enlightenment.).
In answering these questions, trainer frustration, some outdated “pack theory of alpha dominance/leadership” or ideological devotion to a misunderstood Behaviorist Model being adhered to, can all come into play. Perhaps it sheets home to an overload of experiences with tough spirited, recalcitrant dogs lacking proper puppy socialisation and training which require a constant “firm hand”, overlooking breed and personality differences and then expanding those experiences to a general training paradigm, whilst ignoring the various outcomes sought in the myriad of training programs and objectives[51]? Perhaps this reflects one of the greatest oversights of certain punishment training advocates, overlooking the spectrum of training needs and outcomes. For example, consider the training of psychiatric service dogs, for which enlightened trainers advocate building the human-canine “trust through patience, consistency and positive interactions throughout training”[52] – these trainers would not consider punishment to be a part of their toolkit.
In summary, answers to these types of questions will tell you more about the personality and proclivities of the dog trainer than about the best training methods and whether a single, universal, method actually exists.
Conclusion
Advocates for “balanced canine training” go to extraordinary lengths to demonstrate that not only is punishment not detrimental in the long term, but indeed is effective and essential, as a canine training ingredient. This line of argument making the case in favor of punishment draws on some questionable data and overlooks several fundamental challenges, described in this paper.
In the words of Sidman[53], “…. common sense tells us that we have to use whatever effective means are at hand. Mistakes, a temporary lack of relevant information, or an occasional emergency may justify punishment as a treatment of last resort, but never as the treatment of choice. To use punishment occasionally as an act of desperation is not the same as advocating the use of punishment as a principle of behavior management.
If there is a catch phrase which dog trainers could have in mind in deciding their preferred training methods, it might be “Consider options, boundaries and consequences – then behave as the dog’s best friend”.
References
[1] See companion article on this website The Operant Conditioning Model Fallacy.
[2] Oxford Dictionary/Oxford Languages: causing strong dislike or disinclination…..relating to or denoting aversion therapy, a type of behaviour therapy designed to make patients give up an undesirable habit by causing them to associate it with an unpleasant effect. "A program of aversive treatment for criminal offenders".
[3] American Veterinary Society of Animal Behavior (2021) Glossary of Terms used in AVSAB Position Statements: www.AVSAB.org
[4] Examples of an aversive may include verbal reprimands, pushing an animal into a position (alpha rolls, dominance downs) threatening body language, shaker cans, spray bottles, citronella collars, leash corrections, choke chains, prong collars, or shock collars. This list really covers the field at the extreme end and does not leave room for “yes, but….”. However, the “milder” aversives, over which ardent force-free protagonists also hold sway, is left undefined (such “aversives” as a momentary tight leash, a leash “pop”, a “Hey, stop that”, or a “Nope/Oops/Wrong….” if the wrong action is displayed in response to a cue).
[5] We note that an ardent force-free reader (if such a person really exists and religiously applies their mantra) maintains that no aversives are acceptable under any circumstances (even in safety emergency situations?). See especially the definitions of these categories offered in Aligning "positive reinforcement" with "use of aversives" dog trainers.
[6] Hilliard, S. (2026) Kynology 4; Michael Ellis School for Dog Trainers; California [by enrolment/subscription only: April 14-18 2026, Santa Rosa, CA].
[7] Solomon, R. L., Kamin, L. J. and Wynne, L. C. (1953) Traumatic Avoidance Learning: The Outcomes of Several Extinction Procedures with Dogs; J. Abnormal and Social Psychology; 48 (2) pp.291-302: Solomon, R. L. and Wynne, L. C. (1953) Traumatic Avoidance Learning: Acquisition in Normal Dogs; Psych. Monographs: General and Applied; 67 (4); Whole No. 354, pp. 1-19.
[8] See companion article on this website The Operant Conditioning Model Fallacy
[9] Canine training is commonly referred to as occurring in 4 stages, viz., Acquisition, Fluency, Generalisation and Maintenance.
[10] Hilliard, S. (op.cit.) apparently feeling the need to instruct canine trainers on the need and value of “aversive outcome stimuli” (Hilliard terminology).
[11] U.S. Military’s Dog Training Handbook: Official Guide for Training Military Working Dogs (2019) Dept. Defense; Pepper Press; 183pp: see p.59 {quote} “….. hasten the Sit! by applying physical force…. The handler intends to demonstrate to the dog the consequences of sitting slowly, but unwittingly he/she actually constructs a very effective classical conditioning trial – Sit! develops the power to trigger responses…… from pain-elicited aggression and biting to avoidance and cowering….. or anxiety and reflexive stiffening ….to defend against the correction”; and at p.65 which recommends {quote} “…..the proper use of inducive training is to teach the dog skills, while the proper use of compulsive training is to enforce the performance of these skills (if necessary). The optimal way for the handler to build rapport and a good working relationship ….. is to perform inducive training …. And avoid the use of physical compulsion”; also p.99 on Clear Signals Training method (CST).
[12] This concept of redirection is also found in the training of service dogs – see for example Hurd, M. (2025) Psychiatric Service Dog Training Handbook: Amazon, Sydney, Australia. General theme and especially at p.55.
[13] Sidman, M. (1989; revised paperback Ed. 2000) Coercion and its Fallout; Authors Coop., Inc., Publ., USA; 300 pp; esp. at p. 243 in Yr. 2000 Ed. (this concept is built throughout the book).
[14] Dickinson, A. and Balleine, B (1994) Motivational control of goal-directed learning; Animal Learning and Behavior, 22 (1) pp 1-18.
[15] Hearne, V (1986) Adam’s Task: Calling the Animals by Name; NY; A. A. Knopf (reprint 2007; Skyhorse Publ.; 288pp).
[16] Koehler, W. R. (1962) The Koehler Method of Dog Training: Howell; republ. 1996 Hall & Co. USA: 378pp
[17] Particularly those described by W. K. Estes, B. F. Skinner, J. A. Dinsmoor, R. L. Solomon, et al. and N. H. Azrin & W. C. Holz, all referenced herein.
[18] Again, refer to the companion article on this website The Operant Conditioning Model Fallacy.
[19] Breland, K. and Breland, M (1961) The Misbehavior of Organisms; Am. Psych. 16, pp 681-684.
[20] Lorenz, K. (1950) Innate Behavior Patterns Symp. Soc, Experimental Biology, 4; in Physiological Mechanisms in Animal Behavior; New York Academic Press.
[21] Tinbergen, N. (1951) The study of instinct; Oxford; Clarendon: and (1963) On aims and methods of ethology; Zeitschrift fur Tierpsychologie 20; pp 410-433. A possibly more accessible account of this work (English language) is summarised by Miklosi, A (2015) Dog Behaviour, Evolution and Cognition; Oxford University Press; Ch 2 pp 16 – 38.
[22] Lorenz, K. (1937) The Companion in the Bird’s World: The Auk, Volume 54, Issue 3, 1 July 1937, Pages 245–273, https://doi.org/10.2307/4078077: Pellis, S. and Pellis, V. (op.cit.) entire Ch 6; but especially p.110.
[23] Pellis, S. and Pellis, V. (2009; reprint 2017) The Playful Brain – Venturing to the Limits of Neuroscience; Oneworld Publ.; 258pp (reprint); entire Ch 6; but especially p.110.
[23A] Holland, P. C. (2004). Relations between Pavlovianinstrumental transfer and reinforcer devaluation. Journal of Experimental Psychology: Animal Behavior Processes, 30, 104–117.
[23B] APA Handbook of Behavior Analysis [Vol.1] Methods and Principles (2012)
Madden, G. (Ed in Chief): American Psychological Association (Publ.) 607pp: esp. pp298-299 in Lattal, K. M. ; Pt III The Experimental Analysis of Behavior: see particularly Ch. 13 Pavlovian Conditioning and Ch. 14 The Allocation of Operant Behavior (refers to "Pavlovian-to-instrumental transfer").
[24] Coercion definition – persuasion to do something (an act or behavior) by using force or threats.
[25] Sidman, M. (op.cit.).
[26] Estes, W. K. and Skinner, B. F. (1941) Some Quantitative Properties of Anxiety; Journal of Experimental Psychology 29 (5) pp. 390-400.
[27] For the canine, this idiomatic saying can mean that although the trainer is frustrated by the dog’s incorrect/undesirable action/behavior and so administers a punishment directed at that behavior, the dog itself actually has no idea that it did something wrong or unwanted, because of the environmental distractions or simply the lack of appropriate training - hence the punishment appears to it as unrelated to a wrong action so constitutes “Conditioned Suppression” – non-contingent upon action/behavior – generating a “Conditioned Emotional Response”.
[28] Estes, W. K. (1944) An experimental study of punishment; Psychological Monographs, 57 pp. 1-40.
[29] Holthe, T. H. (2005) Two Definitions of Punishment; The Behavior Analyst Today; V 6, N. 1, pp. 43-47.
[30] Note that studies indicate that “traumatically acquired habits” (referred to as “traumatic avoidance learning”) maintain a strong resistance to extinction, even without positive reinforcement of that behavior. Solomon, R. L., Kamin, L. J. and Wynne, L. C. (1953) Traumatic Avoidance Learning: The Outcome of Several Extinction Procedures with Dogs; Jour. Abnormal and Social Psychology, 48 (2) pp. 291-302.
[31] Azrin, N. H. and Holz, W. C. (1966). Punishment; pp. 380-447 in W.K. Honig (Ed.), Operant behavior:
Areas of research and application. New York: Appleton-Century- Crofts.
[32] Bolles R. C. Holtz R. Dunn T. Hill W. (1980). Comparisons of stimulus learning and response learning in a punishment situation. Learning and Motivation 11, 78–96., is a research example commonly quoted.
[33] Referred to as “traumatic avoidance learning” by Solomon et al (op. cit.).
[34] Solomon, R. L., et al (op. cit.): Solomon, R. L. and Wynne, L. C. (1953) Traumatic Avoidance Learning: Acquisition in Normal Dogs; Psychological Monographs: General and Applied, 67 (4) 19 pp.
[35] The subject dogs reacting with avoidance of entering the cage, defecation, urination, yelping, trembling and attacking the “responsible” apparatus.
[36] Seligman, M. E. P. (1975) Helplessness: on Depression, Development, and Death; Publ. W. H. Freeman & Co., 250pp: Petersen, C. Maier, S. F. and Seligman M. E. P. (1995) LEARNED HELPLESSNESS: A Theory for the Age of Personal Control; Reprint; Oxford Univ. press; 373pp.
[37] Words of Solomon et al. (op. cit.) reporting on experiment observations.
[38] Schiller, D. et al. (2008) Evidence for recovery of fear following immediate extinction in rats and humans; Learning and Memory, 15, pp. 394-402. Note that this research appears to contradict Myers, K. M., Ressler, K. J. and Davis, M. (2006) Different mechanisms of fear extinction dependent on length of time since fear acquisition; Learning and Memory, 1, pp. 216-223; however, the difference may be “apparent extinction” of the immediate high level of fear, but the retention over the longer term of the memory of the fear producing stimulus – and therefore recurrent wariness or even fear.
[39] Sidman (op. cit.) at pp. 71-77.
[40] Rescorla, R. A., & Lolordo, V. M. (1965). Inhibition of avoidance behavior. Journal of Comparative and Physiological Psychology, 59(3) pp. 406–412 [Sidman experimental concept shuttle box experiments on dogs concluding that Pavlovian fear conditioning and avoidance reaction could be enhanced or depressed according to warning signals and contrasting signals]: Dinsmoor, J.A. (2001) Stimuli inevitably generated by behavior that avoids electric shock are inherently reinforcing J. Experimental Analysis of Behavior, 75: pp. 311-333 [attempts to demonstrate that warning and termination of punishment via electric shock on rats has a positive reinforcing effect]: Schiller, D. Cain, C. K., Curley, N.G, et al (2008) Evidence for recovery of fear following immediate extinction in rats and humans; Learn. Mem. 15 pp. 394-402 [Extinction of fear response by warning stimuli not followed by the fear inducing aversive – i.e., fear recovery]: Konorski, J (1967) Integrative Activity of the Brain; U. Chicago Press; 543pp. [motivational processes in the brain based on studying Classical and Operant (instrumental) conditioning interplay; described “fear relief” and “transient memory” as well as “prospective” and “retrospective” memory, both required for organised behavior].
[41] Baumeister, R. F. and Vohs, K. D. (2007) Encyclopedia of Social Psychology; Sage Publications; 1248pp: describes the concept of “forewarned is forearmed,” i.e., forewarning often leads to resistance, but can also temporarily lead to acquiescence: so, the forewarning interpretation is really not simple.
[42] ]: Konorski, J (1967) Integrative Activity of the Brain; U. Chicago Press; 543pp: Domjan, M. (1993/1998). Domjan and Burkhard's "The principles of learning and behavior" (3rd ed.). Thomson Brooks/Cole Publishing Co., 435pp; especially Ch. 9 Aversive Control: Avoidance and Punishment and Ch. 10 Classical-Instrumental Interactions and the Associative: Structure of Instrumental Conditioning and also The Concepts of Habituation and Sensitization in Ch.2.
[43] Ohman, A. and Mineka, S. (2001) Fears, Phobias and Preparedness: Toward an Evolved Module of Fear and Fear Learning; Psych. Rev. 108 (3) pp. 483-522.
[44] Reynolds, G. Field, A. P. and Askew, C. (2017) Learning to fear a second-order stimulus following vicarious learning; Cognition and Emotion; 31: 3, pp. 572-579.
[45] Ohman, A. and Mineka, S. (op.cit.).
[46] Most, K (Colonel) (reprint 2001) Training Dogs: A Manual; original 1910, translated to English from the German 1954; Dogwise Publ.; 214pp. Koehler, W. R. (1962) The Koehler Method of Dog Training: Howell; republ. 1996 Hall & Co. USA: 378pp
[47] Woodhouse, B (1984) No Bad Dogs: The Woodhouse Way; Touchstone publ.; 128pp.
[48] Numerous YouTube videos in which coercive training methods are cloaked in simple three step pseudo- psychology employing dominance theory, demonstrating aversive techniques such as forcing the dog to take a feared action and using flooding to counter an unwanted behavior based on fear.
[49] Burch, M R (2022) The Evolution of Modern-Day Dog Training: Today’s dog trainers owe much to their predecessors; Dog Savvy Los Angeles; URL http://www.dogsavvylosangeles.com/blog/2022/8/13: provides an excellent historical evolution summary and commentary.
[50] This opens a discussion of a longer topic, dealt with, for example, in the companion articles on this website:
[51] The military canine trainers – in some cases still referencing the dominant/submissive “alpha dog” concept, do recognise that {quote} “attempts to physically punish a dominant dog into cooperative behavior normally only results in handler aggression and the dog and handler becoming suspicious of one another”: on the other hand, a subdominant or “submissive dog will often exhibit what is called willingness or eagerness to please. This behavior can greatly facilitate the dog’s training if the handler has established a positive rapport with the animal”. U.S. Military’s Dog Training Handbook: Official Guide for Training Military Working Dogs; Dept. Defense; Pepper Press; 183pp; quote from p. 55. Although the “dominant/subdominant” characters might be explained and classified in other ways than pack leadership and willingness/eagerness can be explained and developed in different ways, the point regarding excessive corrections and positive punishment across the spectrum is well made.
[52] Hurd, M. (op.cit.) General theme and at p.49.
[53] Sidman (op. cit.) at p. 6
