10.3 Theories of Cognitive Categorization & Classification
Knowing What We See
The little boy runs up to the large cat, picks it up and yells ”PUPPY!” You tell him, ”No, that is a cat.” He stares you down with a snarl on his face and insists that it is, indeed, a puppy.
Why? Why does the boy believe with all his heart that the animal in his arms is a dog?
As we come into contact with the world around us, we learn to classify items into an organized manner. We create tidy little categories for each new group. Most often, these groupings or categories are based on similarities, but sometimes they can be theory based.
This lesson will walk you through the different levels of similarity-based categorization and slightly touch on theory-based categorization.
Similarity-Based View of Classification
The boy in the opening example believed the cat was a dog because he had never been exposed to a cat before, and the cat was similar to the puppy he had met before. It had fur all over it, four legs, a long tail, and was smaller than the boy. All of these similarities made it match up to characteristics seen in a puppy; thus, the new thing must be a kind of puppy.
There are three theories of categorization based on similarity: classical, probabilistic, and exemplar.
Dating back to Ancient Greece, the classical view of categorization was introduced by Plato and is a simple grouping mechanism based on similar properties. Aristotle used this approach as he categorized living things.
You’ve probably used this approach as well. Have you ever played ”20 Questions”? The first questions are normally ”Is it living?” or ”Is it a vegetable? Animal? Mineral?” These questions, and the successive questions that narrow down the mark through distinct categorization, are examples of the process of using the classical view to classify.
Structuralism is a field of study that claims that in all things, humans observe the parts of the object to identify the object. This fits in with a classical framework for categorization.
A problem with strict classical categorization is that people do not always perceive things in the same way; thus, universal classifications are difficult.
Another problem is that some items that should be classified together are different enough to be confusing. If a bird must fly and have feathers, is an ostrich a bird? It has feathers but can’t fly.
This leads us to the probabilistic view of categorization. Also known as conceptual clustering, this view builds on classical categorization by allowing for more generalized or ”fuzzy” descriptions of categories. Instead of absolute descriptions needing to be met in order to classify a new item as part of an existing category, descriptions are considered likely, leaving room for broad differences between incidences within a single category.
Probabilistic views can lead to fuzzy set categories, which are groupings of entities that share many descriptors but in differing degrees. For instance, foxes and wolves may both belong to the dog classification but are very different from each other.
You can think of the probabilistic view theory as the gray space outside of decisions that are either ”black or white.”
Naive template theory is the theory that we hold a sample template of objects in our memories. It is required that we must first come into contact with these objects, but that contact creates a prototype of the object that is stored for future use. With each object we contact, we compare it to the prototype already stored to evaluate its category.
In the beginning, the boy above had only had the puppy template (or prototype) to go by. Because the new stimulus (the cat) matched the dog template, to some degree, it was assumed this was a type of dog. When the boy accepts that the new animal is not a dog, he will create a new ”template” for cat using this object as the original prototype in his memory.
Imagine that the same child had first seen a St. Bernard dog, then the cat, and finally was introduced to a terrier dog. You might expect that the terrier would be initially slotted into the cat category as it matches that prototype much more than the St. Bernard prototype. When it is accepted that the small animal is a dog and not a cat, the boy will create another prototype assigned to the ”dog” category.
When many prototypes have been assigned to a single category to accommodate the differences between similar objects, we call this an exemplar view of classification. The exemplars are ”examples” of the category.
For some, none of the similarity-based views of classification fully account for all the intricacies of object recognition. For example, none of the similarity-based theories addresses the extreme difference between a fish and a whale or explains how to categorize a platypus. Gestalt theory states that objects are more than the sum of their parts, and our understanding of them is more than just observing their individual aspects. We see things as a whole; images and objects are assessed in context with each other.
Theory-based views of object recognition categorizes based on features, instances, and concepts viewed as a whole.
It would be theory-based categorization that would finally allow our little boy to understand that there is a canine family of many differing animals and a feline family that also holds a tremendous amount of diversity.
The majority of categorization is done through observing similarities. Indeed, earliest attempts to document formal classifications of living things by Aristotle followed the classical view of categorization. Probabilistic, or conceptual, categorization takes into account some diversity among objects belonging to the same group. Prototype and exemplar methods of categorization allow for each person to match new stimuli to existing samples in memory. When none of this works, theory-based methods of categorization allow for complexities that go beyond similarities to be used in order to form common groups.