When we think in simple terms, it is easy to understand how people could equate positive to good things and the words punishment and negative as bad things. This is not exactly how all these terms are applied in training dogs and sometimes it is necessary to consider how the dog perceives the application.
Positive reinforcement as it is technically applied is a “procedure in which the instrumental response turns on or produces an appetitive stimulus” and contrary to this, punishment of an instrumental response “produces or turns on an unpleasant or aversive stimulus” (Domjan, 2003).
Negative reinforcement is a little trickier but technically is explained as “the response turns off or prevents the presentation of the environmental event” (Domjan, 2003). Negative reinforcement produces two types of responses escape and avoidance. With escape, a dog can terminate an aversive stimulus by moving away and with avoidance; a dog can avoid the presentation of an aversive stimulus by providing the correct response prior to the presentation of the aversive stimulus.
Positive reinforcing events serve to satisfy either physiological or psychological needs. Certain learning events form contingencies between behavioral responses and consequences and identified by presence of cues or discriminative stimulus. The most important lesson learned is the dog realizes they control their environment making the entire experience intrinsically reinforcing.
Attending Behavior – Why is this important and how does it relate to positive reinforcement?
Attention is considered the most basic form of behavior and “both classical and instrumental elements closely cooperate” mediating effective “perception and action” (Lindsay, 2000). In a broader view, “attentional activities specify a dog’s intentions, reveal a dog’s motivational state” and sometimes define what he is prepared to learn, thus “attentional activities” are said to “reflect a dog’s overall disposition to learn” (Lindsay, 2000).
How we stimulate and control dog’s attentional behavior can have profound effect on training and behavior modification. Lindsay (2000) says “dogs pay attention to occurrences that are significant to them and learn to ignore occurrences that are irrelevant” and stimuli associated with pleasurable events or those associated with fearful events gain the most attention than other irrelevant stimuli.
To emphasize why attending behavior is important Lindsay (2000) says, “…attention is highly correlated with reinforcement (both positive and negative).” Animals become more attentive through experience and “since attending behavior is present in most successful learning situations” it could be considered a “dominant class of higher-order behavior” and may affect all other classes of instrumental behavior in accordance with its frequency and probability and “the least carefully studied.”
Training should always begin by gaining cooperative attending behavior and how this behavior is trained will ultimately have profound effects on the subsequent dog-human relationship.
How Positive and Negative Events Affect Dog’s Behavior
When we condition stimuli using classical conditioning we are setting the subject up to understand a specific expected outcome as a result, the same holds true in instrumental conditioning when we use signals to gain responses, that in turn give the subject information regarding whether reinforcement or punishment will be delivered. In instrumental conditioning, we use acquisition, reinforcement schedules, and extinction to provide necessary information regarding expected outcomes or contingencies. It is with this information the subject can draw conclusions regarding their own behavior and consequences. One may use continuous reinforcement, intermittent reinforcement or differential reinforcement for other desirable behavior.
However, it is fundamentally important for dog trainers and owners to realize how predictability for both rewarding and punitive consequences may affect the learning process. It is necessary in training that we provide clear links with any proceeding antecedents and subsequent behavior otherwise; the subject may be unable to associate their behavior with the rewarding or punitive consequences. This would create a very unstable relationship, which can lead the subject to either learned laziness or even worse learned helplessness.
According to Lindsay (2000), the “lack or loss of controllability of positive outcomes affects not only subsequent appetitive training but also the animal’s ability to learn aversive contingencies’ and additionally one may inadvertently reward undesirable behavior and superstitious behavior.
Lindsay (2000) says, “…unpredictable and uncontrollable aversive stimulation” and its effects can even be more “pervasive and debilitating, when a subject is not given the opportunity to learn avoidance cues pertaining to negative reinforcement and noncontingent punishment. In addition, “…the loss of control over significant events via the noncontingent presentation of appetitive or aversive stimuli results in reduced operant initiative and retards associative learning processes.”
The devastating effects on dogs can include becoming “overly cautious, nervous, and insular” since they are unable to predict outcomes concerning their behavior. Additional observed behavior might include punishment passivity, pain insensitive, stubborn, failing, and resistant to learning and appearing to struggle with training often resorting to withdrawal (Lindsay, 2000).
Lindsay’s Alternative Theory of Reinforcement
According to Lindsay (2000), “sharp lines of distinction between instrumental and classical phenomena do not exist except under the artificial conditions of the laboratory and not really there either.” He proposes, “…successful control depends on adequate prediction and adequate prediction depends on successful control” and that when “significant events are adequately predicted and controlled; the consequence is adaptive success-an enhanced state of well-being, confidence, and power.”
Within this proposed framework, the dog can meet both his “biological and motivational inclinations” that drive his behavior coupled with his previous reinforcement history form his “disposition to learn.” The need to “predict and control” ones environment will be affected by necessary “biological, emotional and psychological homeostasis and security” and ones overall goals are “survival, adaptive success, enhanced power [and]…reproduction.”
Lindsay (2000) asks, what is the relationship between reinforcement and punishment. Behavior analysts have defined punishment in terms of its effect on behavior through presentation of positive or negative punishment and its ability to either suppress or lower the frequency or probability the behavior occurs in the future. He says, punishment defined “as a suppressive event” only describes it using the most ‘superficial and general attributes.”
Using an alternative definition, he suggests, “punishment be stated in terms of prediction and control.” In addition, “punishment is defined as occurring whenever a behavior fails to anticipate or control a significant event.” Punishment is not something done to a behavior or animal but rather something the behavior itself does or fails to do. Essentially punishment then “fails to appropriate an important resource or escape or avoid an aversive or dangerous situation.”
To use an easily understandable analogy, one could use the common cue sit, an instrumental trained response. If a dog is hungry and fails to obtain a food reward for sitting because it missed a cue/signal…the dog is punished, not indirectly because the food reward was withdrawn, an appetitive opportunity, but directly because they failed to control an opportunity to obtain food. This occurs using the same analogy when a dog misses a signal/cue when they fail to sit and is punished because the dog fails to control the presentation of an aversive event. It is an aversive event when a hungry dog is deprived of a food reward for non-compliance. On the other hand, if the dog is not hungry, has other motivational need, taking a food reward for compliance with a requested behavior can be a punishment!
Lindsay (2000) cautions punishment is associated with emotional states and punishment failing to predict reinforcing events results in fear and anxiety. Failure to control the occurrence of a reinforcing event results in frustration. In some instances, “emotional reactions facilitate adaptation in cases where prediction and control are compromised,” fear and anxiety heighten vigilance improving future anticipation for stimuli associated with reinforcement, and frustration invigorates or amplifies behavioral effects directed toward the restoration of instrumental control over available reinforcers.
To some degree, anxiety and frustration contribute to ones overall learning experience, but in cases where high levels of fear and anxiety are present learning will be adversely affected. High levels of anxiety produce unpredictability and frustration producing uncontrollability and have causal effects toward potential learning dysfunctions. These types of learning environments tend to produce emotional states similar to PTSD and abnormal behavior patterns i.e. learned helplessness. An alternative environment providing a necessary amount of predictability and control over one’s resources produces a more successful adaptation and sense of well-being (Lindsay, 2000).
There exist differences in how punishment is viewed, “whether one views punishment from the perspective of an event produced by behavior (animal’s perspective) or as an event done to behavior (trainer’s perspective).”
How does prediction control expectancies help dogs adapt to their environment?
Prediction control expectancies and cost-benefit analysis are controlling factors for dogs making decisions regarding approach and avoidance behavior. According to Lindsay (2000) “prediction and control expectancies share a common cognitive axis mediating reinforcement and punishment” combined these two forces guide “purposive behavior” including appetitive and escape-avoidance either confirming (verified) or disconfirming when the expectancy is either an attractive outcome (reinforcing) or less (punishing) than the expected outcome. The two predictors’ help dogs control significant events taking into account antecedents that verify outcomes or disconfirm expectancies allowing the animal to adjust their behavior accordingly. When predictive disconfirmation is attractive, it can arouse the dog in either direction of surprise or disappointment. When the dog encounters an aversive outcome the arousal is “startle or relief” depending on the nature of the aversive. In the case of disappointment, the disconfirmation provides feelings associated with “need-anxiety” if the animal is startled, the disconfirmation may be “threat-anxiety.” In either case, a “preparatory adjustment” of increased vigilance and autonomic arousal is in effect until the dog has reappraised the situation and “forms new expectancies and behavioral strategies based on this new information (Lindsay, 2000).
The process of disconfirmation coupled with control expectancies is related to behavioral change and motivation. In a dog’s effort to control “attractive resources,” their “inquisitiveness” could be enhanced or “satiated.” When their efforts fail “frustrative-lose” occurs resulting in “invigoration-persistence or despair” depending on motivation, past history and context. Depending on an individual’s past history in coping during “frustrative non-reward” situations and attractive outcomes the individual will either persist and try harder or with “repeated control disconfirmations” result in a motivational change in the direction of “hope or loss-anger” (Lindsay, 2000).
When dogs confront aversive outcomes they react with courage, if less effort is required, their behavior may shift in the opposite direction toward “threat-anger” because the aversive situation requires more effort than expected (Lindsay, 2000). Aversive situations are capable of causing more problems for dogs including “disorganized efforts involving anger-anxiety loops of aggression.” Contrary to this dogs provided a “high degree of predictability and control” over their environment have feelings of “security and safety” until an event or situation is disconfirmed. Finally, “pathological helplessness and behavioral disorganization” ensues when dogs find attractive and aversive events both uncontrollable and unpredictable (Lindsay, 2000).
What do dogs learn from expectancy disconfirmation?
Behavior is modified by producing attractive or aversive outcomes disconfirming previously established prediction-control expectancies and relative to frequency, size, quantity, or quality of those outcomes.
What this means is dogs having the opportunity to operate in stable environments and prediction control expectancies are confirmed there is no need for dogs to adjust their behavior. This type of predictive environment provides dogs a “sense of enhancing efficacy beliefs and feeling of well-being” and does affect dogs continuing adaptation, further behavioral acquisition or extinction (Lindsay, 2000).
Reinforcement and punishment do the same thing by regulating instrumental behavior and providing dogs with control over the environment they consider significant. They receive reinforcement when they succeed in controlling both attractive and aversive events. Punishment for behavior results in less control over attractive and aversive events. Both reinforcement and punishment when disconfirmed by increasing or decreasing control over relevant attractive and/or aversive events changes an animal’s behavior. An animal’s predictive control expectancies are modified to agree with “cumulative behavioral successes or failures” based on the dog’s behavioral efforts to access available attractive opportunities and to escape-avoid aversive threats.
The necessities for control expectancies help avoid “distressful emotional arousal” leading to anxiety and frustration, learned helplessness or dysfunctional behavior associated with impulsivity or compulsive excesses. A small amount of frustration and anxiety can be helpful in learning, but excessive amounts have harmful effects on learning.
It is my opinion; shelter dogs, rescue dogs and dogs continuing to live in dysfunctional home environments are probably most affected by lack of clear prediction control expectancies. This may ring true even for those dogs who have come through shelters and rescue organizations and readopted to new homes where proper education is still insufficient, especially related to proper management, training, and clear understanding of normal dog’s behavior.
Understanding how these concepts relate in both training and the behavioral environment enhances one’s ability to address and problem solve by offering alternatives to “reactive force and punishment.” It is my opinion; a dog trained using positive reinforcement learns more actively and exhibits confidence and optimism compared to those trained with force alone. A trainer or behavior consultant’s preferential outcome for behavior modification should provide a “system of communication between owner and dog” providing clear expectancies regarding their behavior. The establishment of a “mutually cooperative” relationship based on “constructive mediational behaviors” meets both owner and dogs common needs. This type of approach coupled with appropriate and proper training methods forms a foundation for “interactive harmony based on realistic boundaries and cooperative exchange.”
A word about punishment & aversive control
“Punishment is an inescapable fact of life” (Lindsay, 2000) and the desire for rewarding outcomes and the ability to escape or avoid aversive outcomes provides the “yin and yang” for one’s behavior.
A brief review of the applicable theories
Mowrer’s two-factor theory of avoidance learning
Mowrer (1960) argued the simple association Thorndike proposed explaining behavior associated with punishment was far more complicated. Mowrer proposed two distinct features are added to the training environment when using punishment. He proposed punishment did not suppress an undesirable behavior; 1.) it strengthens behavior directly associated with its termination and 2.) any antecedent stimuli/cues occurring prior to the onset of punishment become emotionally conditioned with fear” (Lindsay, 2000).
Mowrer was particularly interested in why avoidance learning is so resistant to extinction and how it is maintained even when reinforcement for an avoidance response would rarely be presented. To explain this, Mowrer proposed a “two-factor theory of avoidance learning” including the Pavlovian side, explaining the conditioned emotional responses and the “Thorndikian” side explaining “habit formation” (Lindsay, 2000).
By classically conditioning a tone, shock and after repeated trials the tone acquired the same characteristics associated with the US (unconditioned stimulus) normally responsible for eliciting fearful and avoidance behavior. The classically conditioned CS (conditioned stimulus) is able to elicit the same physiological responses previously associated with the US. He theorized, that animals’ find emotional reactions aversive learning to escape precisely the same way they learn to escape direct aversive stimulation-negative reinforcement (Lindsay, 2000).
Kamin (1956) “found if the CS continued beyond the emission of an avoidance response, avoidance learning would be disrupted. The extended CS punished the avoidance response.” This research supported Mowrer’s theory (Lindsay, 2000). Further studies conducted by Rescoral and LoLordo (1965) used a pre-conditioning phase training dogs to avoid shock jumping over a barrier without using a CS (conditioned stimulus) predicting an occurrence of shock. The two groups were later trained using two different conditioned sets of cues one included a tone regularly followed by shock (CS1) and the variable delay group received the same tone (CS1) and an additional tone (CS2) but without shock. The results found dogs receiving the CS1 stimulus gave more responses than the conditioned dogs using the CS1-CS2 combined stimulus. This suggested the “dogs were less worried about the occurrence of shock in the presence of the CS1-CS2 arrangement” creating what they termed a “safety signal” concluding “some variable emotional factor alleviates or potentiates avoidance responding” (Lindsay, 2000).
Mowrer (1960) refined his theory along the same lines as Tolman’s “cognitive learning theory.” He said, “we discard the notion that behavior itself is learned…as habit or as conditioned reflex.” We retain the concept of conditioning using this to explain how certain internal events are connected to new (extrinsic or intrinsic) stimuli. We view them as “hopes and fears” that generally help animals by guiding, selecting, and controlling their behavior in ways that are more adaptively driven (Lindsay, 2000).
The cognitive theory of avoidance learning
Seligman and Johnston (1973) posited the cognitive theory of avoidance learning. It suggests “avoidance signaling results from both emotional conditioning and cognitive information.” Avoidance training i.e. using e-stimulus devices is dependent on dogs acquiring an expectation their behavior controls the occurrence of aversive events. The use of well-timed avoidance cues during acquisition gives dogs the opportunity to avoid aversive stimulation provided appropriate responses are given. With continued learning, the expectancy based on behavioral responses provides predictive control over the occurrence of aversive events. This is maintained through continuing confirmation of successful avoidance of aversive events (Lindsay, 2000).
As the learning phase continues, reinforcing confirmations increase dog’s confidence in the presence of a previous fear eliciting stimuli and unless a disconfirmation is presented as an aversive (punishment), behavior is maintained. According to Siligman and Johnston’s cognitive theory a discriminative stimulus, using positive reinforcement training can also be used as a signal to avoid aversive events. The cognitive differences are using positive reinforcement “learning is based on the acquisition of a promised or hoped-for-outcome in the form of reward.” Avoidance learning is based on behavior avoiding an aversive stimulus providing emotional relief or relaxation by removing, postponing, or avoiding an aversive event (Lindsay, 2000).
Lindsay (2000) concludes, “positive and negative reinforcement paradigms depend on learned expectancies based on history of confirmatory outcomes.” Rather than viewing them as two separate ways for learning to take place, he suggests, “viewing them as two sides of a single process within a broader context of expectancy and confirmation.” This helps clarify the nature of learning and respective role each reinforcement paradigm plays in the learning process.”
The Safety Signal Hypothesis
The earlier experiment conducted by Rescoral and LoLordo (1965) suggests the use of the conditioned CS1-CS2 arrangement depressed the dogs response to shock. The experiment concluded dogs tested using this arrangement learned avoidance preparation and responding was unnecessary in the presence of this combined conditioned stimulus suggesting the “dogs appeared to feel more relaxed or safe even though the signal had no real relevance to the actual arrangement of the avoidance contingencies” (Lindsay, 2000).
According to Lindsay (2000), “dogs experience stimuli associated with relief from aversive stimulation” suggesting they view this as positive reinforcement. He further suggests “praise represents a safety signal” and during training its associative value coupled with appetitive events and presented on a regular basis may “gradually [become] highly desirable in itself and may be treated as a kind of conditioned positive reinforcer.”
As a leading proponent for using safety-relaxation theory and avoidance learning, M. Ray Denny (1971) proposed “avoidance responding is acquired through the antagonistic dynamics of fear and relief-relaxation” (Lindsay, 2000). He proposed the negative emotional state during a fearful event coupled with the successful avoidance or escape turn off and replaced by relief or relaxation. In turn, this stimulates approach behavior and “successive relief and relaxation responses serve to reinforce avoidance behavior” (Lindsay, 2000).
According to Denny (1976), “relief and relaxation are differentiated along two primary dimensions briefly outlined in the following and for further review see Lindsay (2000, pg 295-296).
Relief – onset is 3-5 seconds after aversive stimuli is removed continuing 10-15 seconds – affects muscle & motor functions.
Relaxation – onset is 2.5 minutes following removal of aversive stimuli and before full benefit takes effect.
The use of avoidance training – using conditioned safety or relief cues – should be presented 2-5 seconds after removal of aversive stimuli and continued for several seconds there after.
Inter-trail exposure – the between trial time and presentation of the aversive stimuli should be at least 2.5 minutes
In addition, ‘the effects of safety appear to double when both relief and relaxation…are associated with a particular stimulus’ according to Denny (1983), “safety signals take on conditioned positive-reinforcing properties” and further supported by Weisman and Litner (1969) (Lindsay, 2000).
These implications may provide possibilities in changing previously learned behavior based on avoidance learning. According to Lindsay (2000) “not only does relaxation positively support avoidance learning, it simultaneously results in its gradual extinction” occurring through “back-chaining and counter-conditioning effects originating in the safe…relaxed situation” and through generalization gradually back to the original aversive event.
Tortora (1983) tested the use of safety signals in the treatment of avoidance-motivated aggression commonly diagnosed as dominance. This type of aggression is often caused from dysfunctional avoidance responding (Lindsay, 2000). The important point Tortora makes is his reference to the behavior as “dysfunctional avoidance responding” and its relationship to predictive control expectancies discussed earlier.
Often encountered are cases of aggressive behavior that a dog tried to employ avoidance or escape behavior during a relational context. If the dog is reinforced by the owners’ reaction and/or prevented from escape or avoidance from a perceived aversive event, his behavior changes from a simple stimulus reflexive response to a learned behavior. For example, if a dog is eating, a household member approaches, dog shows signs of aggressive behavior, the household member moves away, the dog is reinforced by avoiding what he perceives as a threat to a resource (food). This same scenario is played out with dogs appearing to guard the couch or bed, and allowed to prevent a person’s access, the dog is reinforced as a positive outcome by preventing the owner from accessing the couch and the owner is negatively reinforced because they are prevented from gaining access.
More support for the use of safety signals is derived using the opponent process theory and studies conducted by Solomon and Corbit (1974). According to the opponent process theory “the offset of any hedonically significant stimulus results in a recoil of opposing emotional reactions” (Lindsay, 2000). Simply stated, when an aversive stimulus is terminated, the opposing pleasurable relief provides a source of covert reinforcement, either strengthening desirable behavior or inadvertently reinforcing undesirable behavior (Lindsay, 2000).
This is generally, what trainers refer to as ending on a positive, allowing the animal an opportunity to perform a task resulting in positive reinforcement after the removal of an aversive stimulus. By not providing the opportunity the animal is left with choices such as “running away or avoiding the owner” that would be counterproductive to establishing what is expected or preferred behavior. The most effective aversive events are those that simultaneously suppress unwanted behavior while teaching desirable or incompatible alternatives for undesirable behavior.
What are SSDR’s Species-Specific Defensive Reactions?
In 1960 Bolles (1960) said, “[t]here is something fundamentally wrong with our traditional interpretations of avoidance learning” and those experiments currently conducted were “basically wrong.” His reasoning based on other researchers’ assumptions concerning animal learning principles concentrated on matters unrelated to “the choice of the response” as being a serious matter of consideration. He said, “…we have become inclined to believe that we may choose any animal, choose any response in its repertoire, and strengthen that response as much as we please by the appropriate application of a suitable reinforcer.”
Bolles (1960) in support of his argument says “the learning of an avoidance response Ra is greatly facilitated if it is chosen to be one of the S’s [subjects] innate defense reactions” and avoidance responses are learned only if they are species-specific defense reactions (SSDR).
According to Bolles (1960) responses associated with these SSDR’s are the easiest to learn during aversive stimulation, within specific contexts and resistant to extinction.
Bolles suggests when subjects are trained in avoidance responding using simple ‘one-way’ avoidance situations the subject learns quickly to avoid the shock by running to the other side of the box, shock is avoided and the response is easily learned providing a “no escape contingency is necessary.” In addition studies using a “running wheel” option as an escape mechanism provided “faster acquisition and better performance levels” allowing rats to alter their location while providing them the use of a normal species-specific defense reaction preferring to leave aversive situations than to fight.
There is a clear alternative to the prevailing view that Ra can be any response in the S’s repertoire. Bolles hypothesis states, “…the Ra is either a defense reaction or some very slight topographic modification of a defense reaction” suggesting the rat has two types of reactions to aversive stimuli “fleeing and freezing.”
Many animals, including the rat, have a particularly interesting defense reaction that is not defensive but rather offensive. Bolles says, “…when a rat is subjected to an aversive event (shock) they…exhibit certain unconditioned reactions…such as jumping, running, or flinching, depending on the intensity” and these behaviors appear to be “largely under the control of the prevailing shock stimulation” and the subject is largely bound to these types of responses. An interesting observation occurred when rats introduced to another subject and presented with a shock stimulus, the subject often attacked the other animal. Bolles (1960) suggests “the precise nature of the animal’s repertoire when being aversively stimulated is more a function of other environmental stimuli than might be thought.” He concluded, “the defensive repertoire is not inflexible by any means” rather it is “highly adaptable to specific environmental constraints” and animals more often choose flight when available, and when not available the animal will freeze, if provocation continues, the animal may attack another individual or object.
Regarding dogs, Bolles suggests, the trick in avoidance situations is punishing all wrong responses so that the right response occurs (Lindsay, 2000). Dogs trained using forceful methods react systematically using various defensive posturing and reactions typical to the dog as a species. These defensive reactions include bolting, jumping, dropping down and freezing, balking and struggling to pull away, and biting the leash. Some dogs exhibit a wide array of passive submissive displays or the extreme opposite with threats and snapping at handlers. Lindsay (2000) says “early stages of avoidance training (punishment training) involves systematically suppressing these innate defensive reactions” replacing them with forcefully prompted alternatives, once all defensive reactions are punitively suppressed or reduced to the obedient target response does systematic and formal avoidance training begin.
How negative reinforcement impacts avoidance learning
According to Lindsay (2000), negative reinforcement occurs when the probability of a behavior’s future emission is increased by 1.) escape from ongoing aversive stimulation and 2.) avoidance of an anticipated aversive outcome. This definition does not include for a dog responding to any predictive stimuli foreshadowing an aversive event. This definition only implies an “escape response” terminating an aversive situation. Reinforcement is both “response correlated and response contingent” meaning “one response turns on the aversive event while another one turns it off.”
When using avoidance training (negative reinforcement) it is necessary to provide a sufficient amount of aversive stimulation to counter the motivation of the animal to avoid its presentation. In some instances, the competing reinforcing qualities may be worth the dog engaging in them despite the aversive event that may occur. The “effectiveness of punishment and negative reinforcement depends not so much on its pain eliciting characteristics as on the elicitation of a startle response.” Fear is the central motivational substrate regulating avoidance learning, dogs learn to fear the presentation of the aversive stimulus or correction, and therefore dogs are learning how to avoid an aversive stimulus. Since eliciting fear is incompatible with positive reinforcement, distractions aversively counter-conditioned are something avoided rather than pursued (Lindsay, 2000).
Avoidance Learning, Fear and Pain
The relationship between pain and fear requires understanding this relationship in avoidance learning. According to Lindsay (2000), “conditioning accounts seem to presume that fear and pain are coextensive events,” however, this does not appear to be true. According to Panksepp (1998), “current neurobiological research disputes the widely held belief that fear is a conditioned response to cues associated with pain.” Pain is capable of eliciting fear and fear of pain is a strong behavioral motivation and plays a role in adaptation. Fear is also a source of “maladaptive aversive arousal” influencing development and expression for many behavior problems.
Punishment and negative reinforcement are often confused, punishment always offers a positive contingency between the aversive stimulus and the instrumental response, compared to negative reinforcement, “there is a negative response-outcome contingency” meaning the expected response will either terminate or prevent the delivery of the aversive stimulus (Domjan, 2003).
According to Lindsay (2000), “punishment and negative reinforcement often occur together” by definition they are both functionally opposite. When a subject successfully avoids or terminates an aversive stimulus the behavior is negatively reinforced and is more likely to be elicited in the future. Contrary to this punishment makes behavior less likely to occur in the future through the presentation of an aversive stimulus positive punishment (+P) or through the withdrawal of a desirable outcome providing negative punishment (-P). Concluding “positive and negative reinforcement function to strengthen behavior” and punishment is used to weaken undesirable behavior.
Is Punishment Bad?
Contrary to Thorndike and Skinner’s previous positions regarding the use of punishment, behavior analysts have recently criticized its lack of effectiveness and potential fallout. Hineline says, “…reinforcement procedures are similarly temporary when reinforcement procedures are discontinued” (Lindsay, 2000).
Lindsay (2000) concedes aversive events do cause side effects, caused by specific situations, abusive treatment, not punishment. Sidman (1989) has written in length about various side effects and problematic features associated with coercive methods of control. He says most behavior is modifiable “without resorting to aversive methods.” Lindsay (2000) has little disagreement with selecting positive training methods but suggests excluding punishment arbitrarily is both counterproductive and artificial.
In light of social reform in the dog training industry with “all due respect for the accomplishments of both Thorndike and Skinner, some of their more extreme views about punishment must be questioned in light of scientific advances and empirical finding derived from practical experience (Lindsay, 2000). The “stubborn reliance on punishment and negative reinforcement to an unequally extreme in which the use of punishment and negative reinforcement is shunned to embrace a so-called ‘positive’ approach to training and behavioral control” are “extreme positions.” Whether they are based on good intentions or not, irrational beliefs and assumptions, not scientific knowledge and experience, is taking a naïve approach.
According to Lindsay (2000) adopting, a one or the other option or method “reflects a core of misunderstanding about how dog behavior is most efficiently modified” and these contrary views are based on “distortion of the subject matter and basic facts.” Despite this debate, “the majority of dog trainers and behaviorists remain pragmatic opportunists about the use of reward and punishment” suggesting “do what works within the context of practical considerations and ethical standards.”
There are occasions when punishment in unavoidable. Those cases are usually defined when “a dog’s unwanted behavior endangers either the dog itself or others with whom the dog comes in contact. In most cases, humane trainers select the least intrusive punishment necessary to achieve their behavioral objectives striving to minimize its use whenever possible. They keep in mind the use of punishment is to avoid using punishment in the future. Rather than adopting extreme positions, that sometimes include “accusatory innuendo, moralizing, and half-truths” a balanced and informed attitude regarding its practical use, misuse and abuse would be more acceptable (Lindsay, 2000).
Does Punishment Work?
The effectiveness of punishment is not really in doubt when science is accepted as a final arbiter of the debate. Punishment applied properly works, in some instances quickly. In many cases the “suppressive effects of punishment are permanent” with several hundred studies suggesting its effectiveness. According to Azrin and Holz (1966), “one of the most dramatic characteristics of punishment is the virtual irreversibility or permanence of the response reduction once the behavior has become completely suppressed.”
Misrepresentation and confusion of facts coupled with “excessive moralizing” may have negative effects on the dog owning public. This often occurs when we ostracize well-meaning owners using “aversive prerogatives” attempting to establish constructive limits and boundaries. Dogs benefit using directive training, combined with a balanced application of behavior modification. A dog’s welfare is better served when we teach owners when punishment is necessary and how to use it effectively and humanely (Lindsay, 2000).
The role of punishment and neurosis in dogs
Numerous studies confirm the dangerous effects from aversive stimulation. This includes but not limited to “puppies exposed to excessive physical punishment…abusive treatment and stressful rearing practices.” These welfare issues cause developmental problems for dogs including hyper-vigilance, irrational fear, heightened irritability, impulsive-explosive behavior, hyperactivity, aggression evoked with minimal provocation, withdrawal, social avoidance, loss of sensitivity to pleasure and pain and depressed mood. These cases require special applications of predictability and control, factors important for behavior modifications effectiveness and eliminating side effects of punishment (Lindsay, 2000).
Using non-contingent punishment has four key factors contributing to maladaptive behavior according to Solomon (1964) outlined as follows.
- the stimulation generates vigorous and sustained emotional arousal
- the stimulation is unpredictable
- the stimulation is uncontrollable
- the stimulation is inescapable
The non-contingent use of punishment often used by unknowledgeable owners and/or trainers occurs when punishment is delivered after an offending behavior has occurred. This results in confusing the dog and affects trust and affection toward the owner. This type of “treatment is not punishment” but rather “simple irrational and ineffectual abuse that should be strictly abstained from by professional dog trainers and behaviorists” (Lindsay, 2000).
Are there any positive benefits?
According to Kazdin (1989) “some side effects of punishment…may actually be beneficial” and “punitive events often help set and enforce social boundaries, promote impulse control, reinforce social status, and provide various other generalized effects” ensuring optimal adaptation in both social and physical environments.
Coercion and Conflict
Frequently when dog trainers use coercive methods inducing dogs to perform certain behaviors conflicting motivations may occur between the choices the dog wishes to make and those the trainer wants the dog to make. These conflicting motivational desires most often result in abusive treatment. When we use compulsive training to counter conflicting motivations, potential for blocking or interfering with natural functioning behavior and satisfying a targeted behavioral system could potentially generate some degree of internal disruption (stress) and homeostatic imbalance. This is why it is necessary to provide outlets for normal “drive satisfaction” to promote “healthy emotional development and equilibrium.”
Before trying punishment, these seven alternatives for changing undesirable behavior should be used first.
- Modify the unwanted behavior into an acceptable form
- Modify the environment so the unwanted behavior cannot be performed
- Redirect the unwanted behavior into a more acceptable outlet
- Bring the behavior under stimulus control, then signal it only under acceptable conditions
- Modify reinforcement contingencies maintaining behavior (extinction)
- Select and reinforce alternative behavior incompatible with undesirable behavior
- For intrinsically reinforced behavior, bring the behavior under the control of an extrinsic reinforcer, then extinguish it
Intrinsic – part of the task itself – internally motivated having both positive and negative behavioral results.
Extrinsic– external to the task – derived from sources other than the behavior itself and having both appetitive and aversive consequences – usually controlled or manipulated by a trainer.
If punishment is considered the following rules should be applied.
- Punishment should result in the dog emitting a behavior incompatible with the one punished.
- The emission of the alternative behavior should occur with the onset of relief from punishment.
The use of punishment requires great knowledge, practical experience, compassion, refined and expert skills and most importantly self-mastery (Lindsay, 2000).
Breaking the Breaker by C.B. Whitford
Do as little breaking as possible; try to encourage the dog to do the proper things and develop him as much as possible with the least amount of control. As final word to the breaker, it may be said that he should so educate himself that he will know that it is always wise, when in doubt, to give the dog the benefit of the doubt. Not only should he know this, but also he must have such complete control of his feelings as to give his knowledge effect. The breaker who spends much time in considering his own weaknesses will profit by this effort (Lindsay, 2000, pg. 305).
The use of excessive punishment
The use of excessive punishment in dog training should be avoided at all costs. In the case of aggression, “excessive punishment may suppress vital threat displays, making future attacks more difficult to anticipate and avoid safely.” These “misguided training efforts may produce a more difficult and dangerous situation to control.”
Punishment is an important tool for the control of dog behavior. Its use “tempered by informed judgment, ethical restraint, and compassion.” Dog trainers and behaviorists should follow the spirit of the Hippocratic Oath ‘do no harm’ avoid methods that obviously ‘do harm’ dogs and the human-dog relationship (Lindsay, 2000).
Bolles, Robert C. (1960). Species-Specific Defense Reactions.
Obtained from http://users.ipfw.edu/abbott/314/Bolles.htm
Domjan, Michael. The Principles of Learning and Behavior (Fifth Ed.).
CA: Wadsworth/Thomson. (2003).
Hineline, Philip. Aversive Control: A Separate Domain?
JEAB: 1984, 42, 495-509.
Lindsay, Steven R. Handbook of applied dog behavior and training. 2 Vols.
Iowa: Iowa SP. (2000). Vol. 1.
Responsible Dog & Cat
Training and Behavior Solutions, LLC
Joyce Kesling, CDBC
Certified Dog and Cat Behavior Consultant
Professional Dog Trainer
The greatness of a nation and its moral progress can be judged by the way its animals are treated. Mahatma Gandhi 1869 – 1948
Copyright Joyce Kesling 2005-2020