The topic for this post is Conceptual Knowledge, and more specifically, why and how our brain classifies and stores it. The question of how we go about defining categories, has been a key area of study for Cognitive Psychology, for the past 50 years or so.  

To begin we need a few definitions to ensure we have a shared vocabulary:

  •  What is “Conceptual Knowledge”? — We define it as knowledge that allows us to recognize the items, events, and concepts in the world around us and deduce conclusions about their characteristics.
  •  For our purposes we will define a “Concept” as a mental representation of a case or type. We could say that our concept for a Thai-Chinese Restaurant is our answer to the question: “What is a Thai-Chinese Restaurant?

Using our sparkling new definition for Concepts, we can now go ahead and say that our category for Thai-Chinese Restaurants would define all the possible examples of these restaurants we can imagine.  

Why Are Categories Important?  

Before we go further it makes sense to ask what is it about categorical thinking that is so important and useful? We touched on this earlier — categories are highly useful shortcuts on our route to understanding how the world around us works. For example, from the moment I share with that Chandrphen (https://www.tripadvisor.com/Restaurant_Review-g293916-d3709901-Reviews-Chandrphen_Restaurant-Bangkok.html) is a Thai-Chinese Restaurant, you’re able to deduce a substantial amount of information and expectations about its menu, decor, and pricing, without ever having been there, or hearing about it.

You derived all this information from the fact that your category for Thai-Chinese restaurants stores all this information for you, and once it is applied to any new concept you encounter, it bequeaths all this data to it.

Categorical thinking is the result of an evolutionary process designed to help us accomplish the incredible feat of being able to process and act effectively in a nearly limitless universe, with a computer that weighs about as much a laptop, uses less power than a light-bulb, and is made largely of fatty goop.

The History Of Categorization — Definitions

Categorical thinking is a tenement of Western Philosophy. From the dawn of philosophical thinking, until very recently, the process of categorizing was understood as one where a definition is formed, and concepts are then examined to see how well they match it, in order to test whether they should be included in the defined category.  This process is the one we used earlier while laying down our shared vocabulary and establishing what may be considered as Conceptual Knowledge .

According to this line of thought, similarly, once I have a definition for a Thai-Chinese restaurant, I can then categorize the world into things that are, and aren’t Thai-Chinese restaurants.  

Where Definitions Fail

It was the linguistic philosopher Ludwig Wittgenstein who in the 1950s came to the realization that, despite their exalted place in Western thinking, definitions aren’t as useful as we think for categorizing things. Wittgenstein used games as an example for demonstrating how definitions can be ineffective.  

While we all intuitively know what a game is, and have no trouble classifying Chess, Basketball, and Angry Birds as games, writing a definition that would easily allow us to categorize all three, is nearly impossible. Don’t believe me? Go ahead and try…

Family Resemblance

Wittgenstein wasn’t satisfied with merely identifying this challenge, he wanted to solve it as well, and true to his Teutonic origin, his proposed solution was wonderfully pragmatic — Wittgenstein suggested that coping with definitions was unnecessary, instead categories could be established based on family resemblance. A category, to his thinking, was a grouping of all the concepts that have a filial resemblance to each other. Using this approach, it’s easy to see how the question of categorizing games is easily resolved, since all the exemplary concepts listed earlier share a fundamental similarity, insofar as they’re amusing pastimes.

Archetypes As Definitions

In the 1970s Dr Eleanor Rosch, a pioneering Cognitive Psychologist, took Wittgenstein’s ideas a step further by suggesting his approach to categorization wasn’t merely a philosophical construct, but also the foundation for how our brain processes data. Rosch and her colleagues hypothesized that in order to categorize a new concept, our brain compares it to a categorical archetype, and based on how well the concept conforms to the archetype, we either do, or do not, assign that archetype’s category to the concept.  

In other words, when trying to establish whether Chandrphen is a Thai-Chinese restaurant we run a quick analysis of whether it matches the archetypical Thai-Chinese restaurant we already have in our mind.

The obvious question then becomes what is an archetype, and how do we go about identifying one to begin with?

How Much Is A Chair Furniture?  

Rosch and her collaborators were highly intrigued by this question and ran experiments to test precisely that. They started by examining how typical, on a scale of 1 to 7, people rated examples of items to the categories they belong to. We can run a thought experiment to replicate their process now. Here’s a list of concepts you may find in a restaurant:

  • Chair
  • Table
  • Mirror
  • Ceiling fan
  • Kitchen cabinet.

Take a moment to rate each of these from 1 to 7 for how typical it is of the category Furniture.  

Now, following Rosch’s experiment, write for each of the concepts listed what are the characteristics you believe establish it as belonging in the furniture category.  

Archetypes Are The Examples With Most Shared Characteristics

When Rosch and her colleagues conducted their experiments in the 70s, they found that archetypical concepts were those who had the largest pool of shared characteristics with all the other items in the category. In our example it’s likely that distinction belongs to chairs. They’re probably going to be the item that has the widest list of shared characteristics with the all the others.

 Archetypes Are The Fastest Examples  

Rosch’s study sparked interest in the budding Cognitive Psychology community and her work was followed up in a series of studies run by  Dr. Edward Smith. In his experiments Smith presented subjects with sentences establishing a concept as belonging to a category. He then measured how long it took them to respond affirmatively. His subjects heard sentences like “A kitchen-cabinet is furniture” and had to press a button as quickly as possible to confirm they agree.  Smith predicted and confirmed that archetypical items would be those for which the response time was the shortest.

What Is A Bird?  

Rosch’s archetypal approach works well for some categories, but not all.  

Take a moment and think about your archetype for a bird.  

It’s likely the picture in your mind is something that resembles a pigeon or starling. Now take a moment to consider penguins and ostriches. Neither of these has the remotest resemblance to a pigeon. In fact, we could argue that bats are more similar to pigeons than either, yet we easily categorize both the penguin and the ostrich as birds, but not the bat. The question is why?

It’s examples like the one we just examined that led cognitive psychologists to suggest that rather than having a single archetype per category, our brain actually use a variety of examples, and is flexible in expanding this list as needed. This approach works much better than the singular

archetype hypothesis, especially for categories like birds, that have a large variety of characteristics which aren’t necessarily shared by all the categories concepts.  

More recent studies suggest we use both the examples and archetype models when categorizing.

Categorization In The Brain

The study of categorization is still very much unresolved and ongoing however, the availability in more recent years, of functional MRIs and other technologies that allow us to peer into the brain at work is providing scientists with new insights and sparking new theories. 

We now know, for example, that particular aspects of the physical world trigger specific neural networks to fire in a set fashion. For example, looking at your hand, as you move it horizontally from right to left in front of your face, triggers neural networks specialized in identifying that particular horizontal movement, as well as networks specialized in identifying the tone of your skin, etc. One of the more popular current hypotheses suggests that our brain understands our hand as a concept represented by the sum total of all the possible patterns of neural network firings ever triggered by stimuli deriving from our hand. Our hand isn’t stored in our brain in any particular neuron or neural network, nor is it simplistically understood as a single object. Instead, all possible representations and experiences of our hands are stored and reflected in all the potential constellations of firing neurons that may respond to stimuli our hand generates. 

These theories provide a good basis for understanding how we can expand our understanding of categories based on new stimuli we receive from our environment. Given that cognitive psychology is heavily influenced by computer science theory, it might be unsurprising to learn this hypothesis also bears a great degree of similarity to how we understand the training of AI agents through machine learning.

Neural Networks Are Efficient and Resilient 

One might ask what’s the point of all this staggeringly complex architecture? Why are our brains built this way? The answer is that this arrangement is both efficient and resilient — Breaking down concepts into elemental perceptions allows the neural networks responsible for these perceptions to be reused across the totality of our experiences, which is of course far more economical than having to store unique data for each experience. Furthermore, this dispersed architecture allows for the brain to suffer tremendous damage, and still be able to function effectively, because even with part of any individual neural network damages, there’s enough retained in overlapping routes, to minimize the likelihood of a total wipe out.

Conclusion 

The study of the processes by which our brain absorbs, organizes, and stores data, is still very much ongoing, and while these processes aren’t always yet entirely clear to us, it appears they’re largely the result of evolution’s attempt to provide us with the best overall hardware package for understanding a nearly infinite universe, within the limitations of a human body. For what it’s worth, the 1.4 kg of fatty goop we each carry between our ears, are still capable of accomplishments far surpassing anything we’ve been able to engineer ourselves, and even the most optimistic projections for advances in computer science put our ability to build a comparable machine many years in the future. 

The universe has provided us all with an amazing brain — What we each choose to do with it is up to us…