Behavior Analysis Analyzed
Gary Wilkes
Arizona State University (Presented as a course handout when I was an associate professor at the Morrison School of Agri-Business at ASU, teaching pre-vet students.)

Within the study of psychology are several sub-disciplines that focus on animal behavior. One of these fields is called behavior analysis. Unlike ethology, the study of how animals behave in their natural habitat, behavior analysis deals primarily with the way behavior is changed by the environment. This field is also called the experimental analysis of behavior. Behavior analysis began in the early 1900’s through the work of scientists such as Edward Thorndike, John Watson and B.F. Skinner. As a foundational cadre they developed terms and concepts to create a discipline that would form a science of behavior. Their original terminology and perspective of behavior analysis has remained the dominant jargon of this and other related fields. When we use terms such as “reinforcement” to describe the strengthening of a behavior, we are using terms coined by those original researchers.

It should not be a surprise that current investigations into behavior use this existing framework. What may be surprising is that this seminal resource for modern animal behaviorists is only tangentially connected to animal behavior. Though most of the research in behavior analysis is conducted with animals, the choice of experimental subjects was done for reasons other than facilitating the care or understanding of those species. Skinner, et al, chose pigeons and rats as their primary research animals as models for human behavior. Paradoxically, these species were not selected because their behavior closely resembles humans. They were selected because they were small, easy to care for and inexpensive.1 Skinner believed that the behavior of any animal is basically analogous to that of other species. Within 20 years of the acceptance of rats and pigeons as experimental standards, that vast majority of behavioral experimentation was/is confined to these two species. When peer review literature grew to a substantial size, behavior analysts confined themselves almost exclusively to these species. Logically it made sense to foster uniformity and gain apple to apple comparisons between scientific studies. The problem is that rats and pigeons are not good models for the behavior of cats and dogs.

Peckers vs. predators
Obviously, rats and pigeons are not predators. Both species are browsing animals whose survival requires methodical, stereotypical behaviors such as pecking and foraging for food. Modern companion animals are primarily predators or the descendents of predators. Predators have a completely different way of surviving that makes direct comparison between the two types of animals problematic. EG: Pigeons peck things. Dogs don’t. If your research depends on an animal pecking a lighted disk to quantify behavioral data, the dog must learn the behavior from scratch. The pigeon performs the behavior instinctively, therefore automatically. The dog’s nose-bumping must be reinforced to create a baseline, the pigeon pecks in the absence of reinforcement. Unless comparison studies can demonstrate a predictable and calculable skew between the species, these two studies are not logically connected. Comparison studies of this sort are fantastically rare. In essence, while attempting to match apples to apples, behavior analysts have excluded oranges altogether. Sticking with this metaphor, veterinarians are primarily concerned with oranges and apple data is of little use.

Beyond instinctive or species specific behaviors, the physiology of predators and prey also compounds the problem of comparison. The primary experimental procedure for behavior analysis is based on reinforcing a particular behavior. Simply stated, the behavior “causes” the presentation of food or water which increases the behavior’s strength and likelihood that it will be repeated. This new reinforced behavior is called an “operant” because it is determined purely by its consequences. Once an operant is formed, the reinforcement procedure is then contrasted to the removal of reinforcers until the behavior stops entirely or goes back to a previously established baseline. This process is termed extinction. The reinforcement and extinction paradigm works best with animals that are in constant need of food and water. Experimental animals are kept at less than normal body weight or dehydrated to facilitate reinforcement. Both rats and pigeons require relatively constant access to food and water. Much of their day is devoted to this process. Virtually all the available research on the use of positive reinforcement is based on these parameters.

In contrast to browsing, grazing animals, predators tend to gorge and starve depending on availability of prey and successful hunting. Rather than being constantly active, wild canids and felines spend most of their time sleeping and conserving energy. Additionally, most wild canids are crepuscular creatures that are active primarily at dawn and dusk and tend to “lay up” during the day, even if they have not eaten in several days. Their biological clocks may suppress normally occurring behavior at various times of the day. This means that dogs and cats don’t react to food like rats and pigeons. For instance, dogs may stop reacting to food reinforcers even if they have not eaten for more than 24 hours. Cats are even more finicky and may simply stop eating because a particular food source isn’t properly palatable. In both of these species the nutritional quality of the food is a great predictor of the animal’s enthusiasm for acquiring it. For top-end predators, low protein, low fat foods are invariably less interesting and less reinforcing than high protein, high fat foods. Pet owners are usually resistant to keeping their dog or cat at 15% below its normal body weight merely to facilitate training. The practice of keeping animals under weight to facilitate experiments is called “establishing operations” to fit the experimental time-table, not to yield great insight into animal behavior. Ultimately, the differences between the typical Skinnerian experiment and the reality of training dogs and cats in pet homes make it difficult to draw analogies. i.e. Comparing positive reinforcement/extinction findings gleaned from rats and pigeons may have little or nothing to do with similar analysis of dogs or cats.

Aversive control — Another Failed Analog
If you are starting to guess that behavior analytic knowledge about reinforcement isn’t an analog for real dog and cat behavior, the other shoe is just about to drop. A more serious problem lies in the most important tool available for effective behavioral control — punishment. Loosely defined, punishment causes animals to decrease or stop a behavior because of an unpleasant event. If a dog avoids charging through a sliding glass door because he “bonked” himself on his first attempt, the smack on the head punished the behavior of charging through the doorway. This is a perfectly normal reaction that rarely leads to phobias, aberrant escape behaviors or personal injury. In its scientific sense, punishment does not imply abuse. It merely describes an influence on behavior caused by an aversive stimulus. It should have the same neutral connotation as surgery or anesthesia — but it doesn’t. Virtually no one uses the term punishment without an implied antipathy, especially behavior analysts. Modern dog trainers and most psychologists have an overt bias against this behavioral effect. This bias stands in the way of progress in the world of veterinary behavioral therapy. To understand the issue requires a look at a further limitation of observing prey animals and extrapolating the results to cover predators.

To kill game, predators risk death and dismemberment as the prey animals resist being eaten or flee over potentially dangerous terrain. i.e. Dogs/wolves/cats are willing to “take the hit” to get what they need. When predators are aggressively aroused they become nearly insensate. They may sustain terrible battle damage that may only be sensed later.2 By contrast, pigeons and rats avoid any circumstance that would cause physical damage. They are not prepared to “take the hit” and use flight as their primary means of survival. Studies of aversive control on rats and pigeons have little to do with speculation about what the same process does to dogs. Pavlov routinely conditioned dogs to accept relatively high levels of electric shock as a conditioned predictor of food presentation. The dogs made perfect conditioned associations between the two stimuli and would wag their tails and salivate after receiving a substantial jolt. In essence, a properly conditioned dog considered the electric shock as a secondary reinforcer — not a secondary punisher as most behavior analytic research would suggest. Note: Anyone who has watched a Labrador receive a heavy jolt from a shock collar and then enthusiastically retrieve a duck understands that the dynamics of punishment cannot possibly be summed up as a simple, linear process.

Despite the overwhelming evidence that dogs are resilient in their ability to absorb aversive training, much current discussion about dog behavior revolves around grave concerns about aversive control. The problem is that the studies linked to these concerns were based on rat and pigeon studies — not cat or dog studies. The “proof” that punishment is a bad thing is derived from scientific observation of animals that are inordinately sensitive to any form of aversive control.

Instinct, we don’t need no stinkin’ instinct!
Perhaps the weakest aspect of behavior analysis is its focus on ontogeny without regard to phylogeny. Worldwide, dogs are represented by about 400 distinct breeds. Though modern kennel clubs focus on cosmetic features, the majority of these breeds were established based on behavioral criteria. Many breeds even get their names from their behavioral peculiarities: Pointers point, Springers spring, Setters set, Terriers dig and Heelers bite the heels of livestock. These are behaviors that need no education and exist regardless of whether you ignore them or reward them. This sensitivity to particular stimuli and resistance to operant control is widespread among dogs. Some of these innate behaviors are modifiable, some are stoppable and some are beyond operant control. Since many of the behaviors that we consider objectionable have genetic components, no single rule or guideline is going to automatically apply. Additionally, the physical attributes of the breed may attenuate standard training methods. EG: Labrador and Chesapeake Bay Retrievers are often insensitive to physical correction. Their heritage includes leaping into freezing cold water for hours on end. Without a built-in numbness they would be unwilling to do their job. A traditional choke-chain correction that would injure an Old English Sheepdog may have no effect on a Labrador of equal size. Yorkshire Terriers may have a hyper-sensitivity to cold and may be difficult to housetrain because they are unwilling to get their feet cold or wet — they have little or no protective tissue to insulate their feet from cold.

Conclusion: Behavior analysis has limited its study of behavior to a small sample of species-specific stereotypic behavior, namely pecking keys and pressing levers. The species most commonly studied have little in common with the primary species treated by veterinarians. To create a practical form of veterinary behavior analysis requires a restructuring of the current methodology, research standards and practices. Drawing conclusions based on current behavior analytic research may be a matter of hanging a very large hat on a very small peg.

Endnotes: 1. Personal Communication: Dr. Ogden Lindsley, B.F. Skinner’s first teacher’s assistant at Harvard University. Confirmation of Lindsley’s comments was given to me by Marian Breland Bailey, Skinner’s second graduate student at Harvard and lab assistant throughout the development of the tools and experimentation of behavior analysis.