Introduction

In the previous chapter, I asked you to imagine a mouse, a cat, and an elephant. You can only do this if you know what these animals are in the first place.

Well, do you? What is a cat? Do you remember learning that? What about animal? Do you remember learning the concept of animal? And how do you know that a cat is an animal?

You may or may not remember being explicitly taught these things by someone else. It will vary, depending on what kind of early education you had, and whether or not you owned a cat. However it got there, though, you know what a cat is, you know what an animal is, and you know that a cat is an animal. These things that you know are what we’re going to call concepts, and in this chapter, we’ll explore how they are organized in your brain, and how they come to be.

Hierarchical representation

Hierarchical organization comes naturally to humans. Files go in folders, which go in bigger folders. An employee works for a boss, who works for a bigger boss. Someone governs a town, which is part of a larger region that someone else governs, which is part of an even larger region that someone else governs. We can easily apply this sort of structure to concepts. You can see an example of this in Figure 4.1.

In Figure 4.1, you can see how Liz’s concepts are organized. Liz knows that a robin is an animal because robin is stored under bird, which is in turn stored under animal.

Let’s get some hierarchy terminology down while we’re here. A cat typically falls under what we’d call a basic category: a term we use naturally on a daily basis. For example, if I ask you what the thing on the left is in Figure 3.2, you’d probably say “cat.” You probably wouldn’t say “mammal” or “animal,” or even “pet” or “thing,” even though all those things are true. That’s what makes a basic category. It’s a bit of a subjective definition, and it does depend on context. For example, at a cat shelter, if someone points and says “what’s that one,” it would be supremely unhelpful to answer “cat.” In that context, the specific breed or name of the cat might be a more natural basic category.

The level above a basic category is called the superordinate, and the level below it is called the subordinate. So, we can say that animal is a superordinate category for bird, and robin is a subordinate category for bird.

This is all very nice, but how do you test if the concepts in our brain are actually organized this way? Surprise: you can use mental travel time! The thing we already talked about! We just have to take the idea of a “map” and make it a little more abstract.

Consider again Liz’s concept hierarchy in Figure 4.1. Now suppose you ask Liz a few questions and measure how long it takes her to answer those questions. You can see this in Figure 4.2, which is a fictional example of the same kind of data a 1969 study¹ uncovered.

Let’s look first at the questions whose answer is “yes.” Liz is quickest to respond to the question “is a robin a bird?” and the slowest to respond to the question “is a robin a thing?” Despite knowing that an object belongs to a big, broad category, it often takes us longer to name it as such than as part of a smaller, narrower category. Hence, it’s quicker to call a robin a bird than it is to call it a thing. Just like with mental travel time across imagined maps, our brains have to travel across a longer distance to get from robin to thing than from robin to bird.

There is also a question whose answer is “no”: “is a robin a place?” Liz takes the longest to answer this question. Why is that? Here’s one possible explanation that uses the hierarchical model: suppose that upon seeing these questions your mind begins at the robin node and travels upwards, checking to see if the question has been answered. At the first node, you arrive at bird. At the next node, you arrive at animal But, following this tree, when will you ever hit place? You won’t! So, you essentially run all the way up your hierarchical tree until you reach the end, and then your brain reports to you “Error! No answer found!” Searching through the whole tree takes the maximum amount of time, so a negative answer to a category question takes the longest,

The hierarchical model makes intuitive sense, and data supports a lot of it. However, there are some things that don’t fit well into this model. We’re now going to take a look at those things.

Problems with Hierarchical Representation

Let’s take another look at Liz’s hierarchy in Figure 4.1. Is it correct? That is, is it scientifically correct? It’s hard to define what an idea is in scientific terms, but there is one thing that’s definitely not in the correct spot: a person is an animal. Why does Liz place person as its own category?

In fact, categories are subjective, and don’t have to obey the laws of science at all. Consider the tomato. You probably lump it in your head with vegetables, but in terms of plant science, it is technically a berry. If you order a slice of berry pie and get a pizza, will you be pleased? I wouldn’t! Tomato, in my head, sits under the vegetable heading, despite the fact that it is technically not a vegetable. But that does not mean it must look like that in your own head – you may place tomatoes with the fruits if you wish. We all have a mental hierarchy that is unique to us.

Individual variation is not the only source of ambiguity among hierarchies, though. The location of a category can vary between people…but it can also vary within one person. Consider Liz’s hierarchy again below, with a few ambiguous terms placed within:

Sydney is the name of a city in Australia, but it’s also the name of Liz’s friend. Does the term Sydney appear in her hierarchy twice? Liz knows that whales are mammals, but they also share many traits with fish. Where should she store whale? If Liz has a category called plant and a category called animal, where does food go? Food can be either one (if you’re not vegetarian) …so should it go above plant and animal? But not all plants and animals are foods, so…should there be three categories? Non-food plants, non-food animals, and food, which contains food plants and food animals? This is getting complicated, fast.

Spreading Activation

Sometimes, when a model is making things too complicated, it’s because the model itself is too complicated. What if we took some of the restrictions away from the hierarchical model? Behold Figure 4.4: the spreading activation model of concept representation.

In the spreading activation model², Liz’s concepts are not organized in a hierarchy. Instead, they are connected by association alone. The birds that Liz knows are no longer underneath the concept of bird; rather, they are connected to it in a web. When Liz thinks of the idea of a bird, she often thinks of a robin. Other categories connect to bird as well; sometimes, thinking about bird makes Liz think of animal, and sometimes, it makes her think of food. An insect is not a bird, but they can be associated in your mind through concepts like flying and the fact that many birds eat insects, so insect can be connected to bird as well. This solves the problem of wondering where to place some ambiguous terms in the hierarchy. If you associate robin with both bird and Batman, you can simply connect it to both concepts.

Another thing to note about Figure 4.4 is the distance between certain concepts. These distances are intended to represent the strength of Liz’s association between the concepts. While the connection between bird and robin might be very strong for Liz, the connection between bird and animal is not as strong. You can also represent this idea by changing the thickness of the lines, so that thicker lines represent stronger connections, like in Figure 4.5.

The theorists of spreading activation theory supported their arguments with some experiments. We can demonstrate these results using Liz’s associations. First, we need to flesh out her tree a little.

Feel free to peruse Liz’s tree to see what her associations are. Your own may be the same, or they may be completely different. In any case, now that we have a lot more items in her associative web, we can explain the experimental procedure. To begin, please direct your attention towards the flower and color sections of the web (you’ll see we got there by associating bird with nest, nest with tree, tree with plant, plant with flower, flower with rose, rose with red, and red with color). For Liz, flower is associated with several types of flowers, including rose, daisy, and tulip. Color is associated with several colors, including red, yellow, green, and blue.

To do the experiment, Liz will need to complete some word pairs. For example:

COLOR: Y _______

What might she fill that space in with? Likely, yellow. Next, up, we have another word pair:

FLOWER: R_______

Perhaps, she’ll fill it in with rose.

You might be thinking, “Hey! This test doesn’t seem very hard.” And you’d be right! We’re not trying to trip Liz up here. Instead, we’re measuring how long it takes her to come up with these words. Let’s say her examination results look like this:

Word Pair	Liz’s Answer	Reaction Time
COLOR: Y_____	yellow	1.56 s
FLOWER: R_____	rose	1.55 s
COLOR: B_____	blue	1.34 s

Now, wait a minute! Why was Liz faster to come up with blue than she was with yellow? Indeed, Liz’s reaction time to name a color will change depending on how much time has passed since she last thought of a color.

Number of Distractors	Looks like…	Reaction Time to say “Blue”
0	COLOR: Y_____ COLOR: B_____	1.22 s
1	COLOR: Y_____ FLOWER: R_____ COLOR: B_____	1.34 s
2	COLOR: Y_____ FLOWER: R_____ BIRD: H____ COLOR: B_____	1.48 s
3	COLOR: Y_____ FLOWER: R_____ BIRD: H____ ANIMAL: W_____ COLOR: B_____	1.56 s

If Liz answers two color questions in a row, she’s very fast at the second color question. If she answers two color questions with one flower question in between, she’s a little slower. If there are two questions in between, she is slower still. The spreading activation model can explain this! This is where the “spreading activation” term comes into play. The idea of this model is not just that your concepts are stored in a web, but also that when you think of one concept, it makes you think of all the other things you associate with that concept. In other words, it “activates” the concepts around it, and that activation “spreads.” So, thinking of bird will make Liz think of things she associates with birds, and thinking of color will make her think of colors. By the time the second color question rolls around, Liz is already thinking about colors, so she is faster at coming up with a color. You can see a visual representation of this in Figure 4.7.

This model—that our concept knowledge is just a giant network of association—is fantastic at explaining a wide range of things, so much so that association is one of the core themes of this textbook. More than anything else, your mind runs on association. Why? One answer is neurons!

Concepts as Networks of Neurons

A neuron is a cell of the nervous system whose primary job is to communicate. As a communicator, it has a receiving end, a central processor, and a transmitting end. You can see a diagram of some neurons communicating in Figure 4.8.

The receiving end of a neuron is called the dendrite. Here, other neurons (or other stimuli) send signals to the neuron. These signals can take many forms: neurotransmitters, other chemicals, electrical changes, or even physical touch. Often, information that stimulation has occurred will travel down the neuron, arriving at the cell body. The cell body, or soma, is the central processor of the neuron. In other words, it’s the regular “animal cell” part of the neuron. If you have learned about cells before, you may be familiar with terms like mitochondria, nucleus, and endoplasmic reticulum. These are the sorts of things you will find inside the cell body.

When a neuron is firing more action potentials than average, we say that it is excited. When a neuron is firing fewer action potentials than average, we say that it is inhibited. You can examine a common way we represent this visually in Figure 4.9.

One thing neurons are very good at is associating. When two things happen together, the connection grows stronger between the neurons that represent those things. When you’re born, you don’t know any concepts (or at least, you can’t articulate any concepts through language). As you learn words, however, concepts form in your mind – first haphazardly, perhaps with errors (look at all those chickens), but they firm up over time. When you first see an apple, you might hear someone say the phrase, “Hey, can I have that apple?” Your brain then logs that the sound apple is associated with a round, red object. When you next see an apple, someone might say, “This is way better than an orange!” Your brain could then mistakenly associate that round, red object with the sound orange.

Eventually, however, you will hear the sound “apple” while looking at an apple way more often than anything else, and your brain will learn the concept apple, because the same neurons are firing every time you think about an apple. You can think of a concept as a pattern of neurons firing. If we add back in the idea of spreading activation, we can also conclude that when you think apple, you’re also activating the neural patterns associated with red, sweet, and so on. Related concepts will activate overlapping populations of neurons, like in Figure 4.10.

We’ve covered what your understanding of concepts looks like. In the next chapter, we’ll talk about what it is.