Have you ever watched a quiz show and wondered how the winner knew so much? The famous American on the hundred dollar bill, the names of the Three Musketeers, the director of the 2014 Oscar winner for Best Picture. … But it’s not just quiz show winners who possess a huge amount of knowledge. You do, too. You easily can recall zillions of facts: the name of your high school, whether Mexico is south or north of the United States, the approximate price of a gallon of milk. You also can remember how to do a zillion things: scramble an egg, drive a car, ride a bike, speak in grammatically correct sentences. Now that we’ve seen how information makes its way from sensory to short-term to long-term memory, our next question is how your long-term memory can retain such vast amounts of information.

Psychologists address this question by studying knowledge representation, which refers to the form in which we retain knowledge in long-term memory and connect different pieces of knowledge to one another. Let’s introduce the topic with an analogy.

Instead of asking how knowledge is represented in the human mind, suppose you asked, “How is knowledge represented in libraries?” (Let’s put aside digital storage and Internet access and imagine a traditional library with hard-copy books.) The answer would have a few parts: (1) Knowledge is stored using symbols (e.g., words) printed on paper; (2) large numbers of symbols, usually representing knowledge of a particular topic, are gathered together into books; and (3) the books are organized by a cataloguing system (e.g., the Library of Congress system) that enables people to retrieve any book relatively quickly. This answer says nothing about which specific books are in any particular library. Instead, it addresses a broader question: How, in general, is knowledge represented?

Psychologists ask this question about the mind. To answer it, they have proposed three main theoretical models, each of which contributes to our scientific understanding of memory and mind: semantic network, parallel distributed processing, and embodied cognition models.

To introduce the idea of semantic networks, let’s first look at the idea of a network, which is any collection of interconnected elements. The Internet, for example, is a communications network in which the elements that are interconnected are computers. In any network, elements in the network vary in how closely they are connected to one another. Two computers in a science laboratory might be closely connected, through a local network that serves the lab, and distantly connected, through the Internet, to a computer on the other side of the world.

In many networks, if two elements have a close connection, activating one of them will strongly influence the other. As an example, imagine a physical network, such as a fishing net. The knots represent the elements of the network. They are connected by cords running from one knot to another. If you shake one knot, the knots closer to it will move a lot, but knots farther away may move only a little.

Now, let’s use these ideas to understand the mind. A semantic network model is a conceptual model of long-term memory in which memory consists of a large set of individual concepts that are connected to one another. Any one concept (e.g., “spring break”) is linked to another (“vacation”) and another (“get a tan”). The interconnections create a network of concepts.

233

What determines how closely any two concepts are related in the mind? In principle, it could be anything: size (maybe concepts of whales and passenger planes will be closely related), color (licorice and bowling balls), or sound (actors and tractors). However, what usually connects two concepts is semantics. Semantics refers to the meaning of words, sentences, or other communications. “Cold” and “snow” are semantically similar because they are both related in meaning to the idea of winter weather. “Vacation” and “tan” are related in meaning to “spring break.”

In the semantic network model shown in Figure 6.9, the mind processes one concept at a time (Collins & Loftus, 1975). In this network, you may be thinking about fire engines at one moment, and at the next, ambulances. The distance between two concepts—represented, in the model, by the length of the line between them—determines the likelihood that thinking of one will lead you to think of the other. Notice that the distance between some concepts, such as “car” and “truck,” is shorter than between others, such as “red” and “sunrises.” So, according to this network model, you would be more likely to think of “truck” when you think of “car,” than you would to think of “sunrises” when you think of “red.”

234

RESEARCH TOOLKIT

235

Concepts can become active in memory without your even knowing it. People often are not aware of how concepts are connected in their memory or how activation spreads throughout their semantic network. They sometimes aren’t even aware of the events that prime concepts; in the priming study described in the Research Toolkit, photos were presented so briefly that the participants weren’t even aware of their content. Nonetheless, priming affected their thinking.

This lack of awareness has important implications for everyday behavior. People’s behavior can be affected by social factors of which they aren’t even aware. Briefly presented information—for example, words or pictures in an advertisement—can spread so rapidly through semantic networks that people are not even aware of their effects. As a result, social information can influence your thoughts and feelings without your even knowing it (Wilson & Dunn, 2004; see Chapter 12).

In sum, semantic network models have proven valuable for understanding how knowledge is represented in the mind. By modeling knowledge as a network of linked concepts, psychologists can explain how ideas trigger one another, and how social influences affect thinking. Yet, as the study of mind advanced in the latter decades of the twentieth century, psychologists identified limitations to the semantic network approach and sought alternatives.

An alternative to semantic network models is parallel distributed processing models of knowledge representation, or PDP models, for short (McClelland & Rogers, 2003; Rogers & McClelland, 2004; Rumelhart, McClelland, & the PDP Research Group, 1986). The PDP approach differs from semantic network models in the following manner.

In semantic network models, the simplest elements were individual concepts; concepts such as cars, trucks, and vehicles (Figure 6.9) were interconnected in the network. In PDP models, the basic elements are much simpler: processing units that do nothing other than turn on or off. No individual processing unit represents a concept. Rather, concepts are represented by patterns of activation in large numbers of units.

An analogy helps explain this. Imagine an electronic sign made up of numerous individual light bulbs; let’s say it’s a sign that displays scrolling news headlines, and the headline at the moment is “John Kerry visits Russia.” Now answer two questions: (1) Which light bulb represents the concept “John Kerry”? (2) Which light bulb is scrolling? Obviously, no individual light bulb represents the concept and no light bulbs are scrolling. The information represented in the sign is contained in the sign as a whole. PDP models are like this. Many individual units take part in representing any piece of knowledge; the knowledge thus is “distributed” across the units.

236

An appealing feature of the simple units in PDP models is that they resemble, in their simplicity, the units contained in the brain: neurons, or brain cells (Chapter 3). Neurons turn on and off; that is, they fire (“on”) or are in a resting state (“off”). Similarly, the units of a PDP model either turn on or off. Inputs from other neurons determine whether a given neuron fires. Similarly, inputs from other processing units determine whether a given PDP processing unit turns on. The resemblance between PDP models and brain processes is an advantage in pursuing the long-term goal of relating processes of the mind to activities in the brain.

Our first two models of knowledge representation, semantic networks and parallel distributed processing, were the primary theories informing research on memory and mind in the last quarter of the twentieth century. Here in the twenty-first century, scientific understanding has been expanded by a new theoretical model: embodied cognition.

The central question embodied cognition theorists address is, “When you think, what part of your mind do you use?” The question might seem silly; it’s the “thinking part”—some part of mind devoted to memory, reasoning, and other types of complex thought—isn’t it? You might expect so; but consider these research findings:

237

From the perspective of semantic network and PDP models, these findings are weird. Why would bodily movements affect concentration or physical warmth affect thoughts about other people? Why would thinking about leaves speed thinking about blenders—particularly odd since leaves and blenders are not semantically linked? The answer, according to embodied cognition, is that perceptual and motor systems take part in thinking processes (Anderson, 2003; Lakoff, 2012; Wilson, 2002). In embodied cognition, parts of the mind that previously were thought to participate only in relatively simple perceptual and motor-movement tasks also contribute to complex thought. This happens even when the thoughts are not, in any literal way, related to perception or physical movement.

Here’s how embodied cognition explains the three findings above:

In each case, a part of the mind that normally carries out perception or bodily movement—an “embodied” part of mind—also contributes to thinking.

Figure 6.10 contrasts semantic network and embodied cognition approaches to memory (Barsalou et al., 2003). In a semantic network model, perceptual systems play no role in remembering; they are active only when people first experience an event. In embodied cognition, by contrast, the visual system is active both when people experience an event and when they remember it. People remember information by reactivating their original perceptual experience, essentially “replaying” events in their mind.

Embodied cognition is consistent with evolutionary principles (Glenberg, 2010). In the distant past, survival required keen perception (e.g., to spot prey) and good motor skills (to evade predators and create tools; Mithen, 1996). Human perceptual and motor systems evolved to meet these needs. Once they evolved, these same systems were used for thinking; the evolved brain systems got “recycled” (Dehaene, 2004). Parts of the mind and brain that evolved to perform one function (e.g., perception, movement) were reused for others (e.g., thinking, memory).

238

An aspect of everyday thinking that embodied cognition can explain is people’s frequent use of metaphors. Metaphors are figures of speech in which one thing is described in terms of something else to which it is not directly related. If you say a friend is in “the depths” of depression or an athlete is at “the height” of her career, you are using physical metaphors (depth, height) to describe emotion and achievements. Why are such physical metaphors so common? Semantic network and PDP models don’t tell us, but embodied cognition does. Our visual system, which evolved to detect the presence and location of physical objects, takes part in thinking. When it does, it brings its physical framework with it, which fosters physical metaphors when thinking (Lakoff & Johnson, 1999).

239