Canine Behavior

---

The purpose of this page is to break down the psychology behind basic animal behavior and explain some of the psychological jargon. The most important part is the positive reinforcement quadrant of operant conditioning, but it is important to see the whole picture in order to understand how positive reinforcement works and why it is important.

Classical Conditioning

A stimulus is anything that can be perceived by an animal, whether it is an object or experience, and whether it is positive, negative, or neutral. Food, praise, and play are positive stimuli for most dogs; pain, fear, and discomfort are negative stimuli; and most other things are neutral stimuli.

Classical conditioning is when an animal makes a natural association between something neutral and something meaningful. The classic example is the experiment that precipitated the term: Ivan Pavlov rang a bell every time he fed his dogs dinner, the dogs naturally associated the bell with the food, and the sound of the bell induced salivation even if there was no actual food present. The bell has no natural significance to a dog, but it became significant when it was associated with food (a positive stimulus). 

In modern dog training, one of the most important uses for classical conditioning is in clicker training. The 'click' sound is, for most dogs, a neutral stimulus, but it is far quicker and more consistent than verbal praise, and when paired with something motivating (usually some kind of edible treat) it becomes an effective training tool. The edible treat is a primary reinforcer because it is innately rewarding; the click from the clicker is a secondary reinforcer because it only has a rewarding effect on the dog after it has been paired with a primary reinforcer like food.

Another extremely useful function for classical conditioning is during systematic desensitization training, during which an aversive stimulus (usually something the dog is afraid of or reactive to) is associated with a positive stimulus (usually a click that has been conditioned to be positive AND a primary reinforcer) strongly and repeatedly enough that the aversive stimulus gradually becomes less aversive and eventually rewarding. Throughout this kind of training, it is crucial that the aversive stimulus is presented at a low enough level that the dog is not fearful or reactive. For example, if the dog is reactive to strangers, you may have to start with the stranger across the street or even down the block rather than approaching the dog.

Operant Conditioning

Operant conditioning is all about contingencies, or if-then scenarios, and it is used to shape animal behavior by replacing undesired behaviors with desired behaviors. Operant conditioning consists of two axes and four quadrants.

The horizontal axis describes whether the behavior in question increases or decreases as a result of the contingency. If the behavior increases, it is an example of reinforcement (the right side of the matrix), because the animal's motivation to conduct that behavior is reinforced by the contingency. If the behavior decreases, it is an example of punishment (the left side of the matrix). If the frequency of the behavior does not change, it is neither reinforcement nor punishment, regardless of the intention of introducing the stimulus in question.

The vertical axis describes whether a stimulus is added or removed in order to influence the frequency of the behavior. If something is added, it is called positive (top half of the matrix), but if something is removed, it is called negative. 

Note that positive and negative do not mean "good" and "bad"—they mean "additive" and "subtractive"; this is important to remember because negative reinforcement (bottom right quadrant) consists of removing a stimulus and consequently increasing a behavior (negative reinforcement is not punishment). Negative reinforcement generally comes in one of two forms: escape (e.g., de-electrifying a rat's cage when it presses a lever, alleviating pressure on a dog's neck when it stops pulling when walking with a prong collar) and active avoidance (e.g., a dog hiding in another room to avoid the scary vacuum cleaner). Notice how negative reinforcement—despite rewarding the behavior—typically involves some kind of aversive stimulus; because of this, it is generally not considered an ethical way to shape animal behavior. The other three quadrants are more straight-forward: positive reinforcement (top right quadrant) is increasing a desired behavior by adding something rewarding; positive punishment (top left quadrant) is decreasing a behavior by adding something aversive (e.g., shaking a can of coins, smacking a dog on the nose); and negative punishment (bottom left quadrant) is decreasing a behavior by removing something positive (e.g., putting a child in time-out).

There are different types of contingencies in operant conditioning. Fixed ratio is when the reward or punishment is presented after a fixed number of instances of the behavior (e.g., every time your dog sits on command, after every third press of a lever in a rat's operant chamber). Variable ratio is when the reward or punisher is presented after a variable number or repetitions (for example, a VR3 schedule might involve rewarding a rat for pressing a lever 1-5 times with an average requirement of 3 lever presses). Fixed interval is when the reward or punisher is presented upon the first instance of the behavior after a set time period (e.g., a rat's first lever press after a period of 30 seconds yields a pellet). Variable interval is when the reward or punisher is presented after a variable time period. Extinction is when an animal gives up on a behavior (extinguishing the behavior) after it repeatedly fails to yield the desired result.

What works best for dog training?

Positive reinforcement is the only quadrant that does not involve aversive stimuli or consequences. Aversive methods typically yield confusion and distrust, and trust and respect are essential to building a mutually beneficial relationship with your dog, so positive reinforcement is best for any scenario. As far as contingencies go, fixed ratio is generally regarded as the most ethical because the animal can count on receiving the reward it worked for, but it is also the most susceptible to extinction because the dog might give up on the behavior after the first time you fail to provide a reward. Variable ratio results in the greatest increase in the target behavior, and it is the most resilient to extinction, but it can also lead to learned helplessness if the dog repeatedly tries a learned behavior but never receives a reward because it has not yet reached the criteria. When pairing a strong reinforcer (treats) with a weak reinforcer (verbal praise and/or click paired with the treat via classical conditioning), it is possible to combine fixed ratio (using the clicker and/or verbal praise every time) with variable ratio (giving treats intermittently) in order to accelerate training without causing undue stress and confusion. When extinction is the goal of training (e.g., your dog jumps on you every time you walk in the door), ignoring the behavior is always preferable to punishing it—you will receive the desired result just as quickly, and without instilling fear or distrust in your dog. The most important thing is to always ensure that your behaviors do not sacrifice the relationship of trust and respect you have built with your dog.

Why is punishment dangerous?

Dogs are naturally not aggressive except in situations where prey drive comes into play, but they will defend themselves when threatened. The vast majority of the time, when a dog exhibits "aggressive" behavior, it is really only trying to defend itself from a perceived threat. When training dogs to be less aggressive, it is essential to train the dog not to perceive harmless situations as a threat rather than to train them not to growl or snarl. When you punish the reactive behavior, you are not making the dog any less afraid of whatever it views as a threat, and you are ultimately making the dog more unstable. For example, if a dog feels threatened by strangers and growls at them, and you punish the growl, the dog will learn to stay silent at that stress level, but it will still be just as stressed; if a stranger corners the dog, they will skip the growl and go straight to a bite. Every time somebody's "well-behaved" dog "snaps" and bites someone, it is because the dog has been trained that apprehensive warning behaviors like growls and snarls will be punished but the dog is still afraid of the same trigger. Using positive reinforcement and/or systematic desensitization to create positive associations with the trigger is the only safe way to train aggression.