We often hear conflicting information about the word “generalization” when it applies to dogs. We are often told that “dogs don’t generalize well”, but then, there are instances when generalization takes place quite quickly as many owners of fearful dogs can attest. Today, we’ll therefore be taking a closer look into dog generalization, the meaning of the word, how it happens and the process it entails. We will also be taking a peak at some examples of generalization taking place in dog training and dog behavior as it can happen in specific contexts.
Generalization is often defined as “taking something specific and applying it more broadly.”We often use generalization in our linguistic world to cover make broad statements. For example, if we say “customers are always right” we are making a broad statementRichard Nordquist, a Grammar & Composition Expert, the English word dog, comes from the earlier word dogge, which was originally used to depict a powerful breed of dog that originated in England. Today, the word dog has evolved though and generalized as an umbrella term used to include all of our domesticated canine companions.
To give a more technical explanation we can say that generalization occurs when a new stimulus/setting/situation that has similar characteristics to a previous stimulus/setting/situation comes to trigger the same response. So if Mary listens to the radio one day and falls in love with a specific heavy metal song made by a specific band, she may then later on find out that she likes other heavy metal bands so starts listening to more and more heavy metal songs and even starts attending concerts. In this case, her passion has generalized from one song, to other songs made by different bands that share similar features with the original band/song she first fell in love with.
Stimulus generalization occurs when a response is “provoked not only by the object or event that originally provoked it, but also by objects or events that are similar to the original stimulus.” Karen Overall
When applied to dog training, generalization can be defined as the process where, the settings in which a dog initially responds to, start to increase so that the dog starts responding to more and more different settings. We often talk about generalization when a dog who was initially trained to sit in a quiet room, starts then sitting in rooms with more distractions, then starts sitting in the yard, and then on walks, at the park etc. Since we have (hopefully!) helped the dog progress by training in more and more distracting environments through gradual exposure, and have rewarded each time the dog responded, the dog succeeds despite the increase in challenge.
We can therefore pat ourselves on the back and claim that the behavior of sitting has successfully generalized to other settings other than the one in which the dog was originally trained. We have therefore taken something that was initially trained in a specific setting (sit in a quiet room) and expanded it to encompass various settings (sit in the yard, sit in the park, sit on walks). When people say “dog’s don’t generalize well” they often refer to the fact that with potty training, it’s often hard to train a puppy to use a pee pad at home and then go potty outdoors on a totally different surface or the fact that it’s hard to train a dog to sit in a quiet room and then ask the dog to sit on walks. However, this is not the dog’s fault! This tends to occur when people fail to use high-value treats (the more distracting the environment, the higher value the treats) and don’t take time to make sure the sitting behavior in quiet settings is fluent enough before moving to training in more and more distracting areas. Slow and steady wins the race!
Did you know? Generalization, strictly speaking is not really something somebody does (the dog doesn’t really “generalize”) but it rather refers to a process. So it would be more correct to say “the dog’s behavior of sitting in the park is sign that generalization has occurred.”
Another example of generalization is seen when we are in the process of increasing criteria and we start training our dog so that he learns to respond to more and more subtle forms of our prompts that share similar features with the original version. For example, if we have trained a dog to spin in a circle by bending down and moving our arm and pointed finger in an imaginary circle that the dog must follow, at some point we might want to make the signal less significant. So we work on making the hand gesture gradually less and less evident by bending down less, then making the circle less and less wide, up until we reach a point where the dog responds just to a mere small imaginary circle drawn with a finger.
Since we have gradually made the hand gesture less and less significant and we have rewarded the dog for responding to these more subtle gestures, we have helped the dog succeed. We can therefore pat ourselves on the back and claim that the behavior of responding to our initial prompt has successfully generalized to other prompts. We have therefore taken our initial pronounced prompt (bending down and moving our arm and pointed finger in an imaginary circle) gradually morphed it so that dog responds to incrementally more and more subtle prompts (circle carried out while bending less, less wide circle.) and at the end, we may have finally decided to make the small imaginary circle drawn with a finger the new permanent cue.
Tip: the more similar the new prompt is to its previous version, the higher the chances for the dog to succeed.
Generalization can also occur in dog behavior and we often see examples of this with fears. When people say “dogs’ don’t generalize well” owners of fearful dogs may disagree when they notice how quickly a dog’s fear can generalize and spread like a wild fire! We can see an example of this phenomenon in people. Let’s say a young boy is attacked by a black dog. At some point, the child may acquire fear of black dogs, and then later, fear of all dogs (despite coat color) even though the initial negative encounter encompassed a single black dog. Generalization in this case occurs because of shared features (having four legs and a tail) with the original dog who attacked him.
In dogs, we can see something similar occur. A dog may one day get traumatized by an unusually loud crack of thunder. Soon, the dog comes to react fearfully to the noise of gunshots, then firecrackers, and then other similar loud noises such as a person closing a car door or person clapping hands. Or in another example, a dog who becomes reactive one day towards a man in uniform coming to read the meter, may in the future start becoming reactive towards other people wearing uniforms such as the Fedex and UPS guys. It’s therefore important tackling fearful behaviors at their early onset before generalization occurs and things get more complicated to treat.
“The more similar the original and subsequent stimuli, the more similar and intense the response.”~ Karen Overall
Did you know? According to James O’ Heare, President of The Companion Animal Sciences Institute, sometimes a dog’s fear generalizes so much that at a certain point you may have a hard time identifying the original stimulus that caused the fear to occur in the first place!
One of the most popular examples of stimulus generalization affecting fear comes from the famous Little Albert experiment conducted by behaviorist John B. Watson and Rosalie Raynor. What happened in this study? A child around 9 months known as Albert was exposed to various stimuli including a white rat, a monkey, a rabbit and masks and his reactions to these stimuli was observed. Albert showed no signs of fear towards these stimuli whatsoever.
At some point though, Watson decided to make a loud noise by hitting a metal pipe with a hammer the moment Little Albert was shown a white rat. This caused Little Albert to cry. After several repetitions of pairing the rat with the sound, Albert starts to cry at the mere sight of the rat. This was a classical example of associative learning where Albert learned to associate the sight of the rat with the loud noise so much so that just seeing the rat resulted in a crying spell. The experiment though didn’t end here… After further experiments, Watson noticed how Albert’s fear wasn’t just limited to the white rat, but soon began generalizing to a wide variety of similar white objects such as Raynor’s fur coat and Watson’s Santa Claus beard! This phenomenon was therefore called stimulus generalization.
“Generalization is often an adaptive function that allows an organism to rapidly respond to novel stimuli that are related in some way to a previously learned stimuli.”Joseph E. Dunsmoor, Stephen R. Mitroff and, Kevin S. LaBar
- Manual of Clinical Behavioral Medicine for Dogs and Cats, By Karen Overall, Mosby; 1 Pap/DVD edition (July 9, 2013)
- Watson JB, Rayner R (1920) Conditioned emotional reactions. J Exp Psychol 3:1–14
- Generalization of conditioned fear along a dimension of increasing fear intensity, Joseph E. Dunsmoor, Stephen R. Mitroff, Kevin S. LaBar, Learn. Mem. 2009. 16: 460-469
- James O’ Heare, The Dog Aggression Handbook, Dogpsych Publishing; 0003- edition (December 1, 2007)