Session Three: Learning and Cognition/Cognitive Information Processing/Schema Theory

Instructor notes written by J. David Perry, Ph.D., Indiana University.

Instructor notes: Approaches to the study of learning

Introduction to CIP

Three basic components of memory are proposed: the sensory register, short-term or working memory, and long-term memory.

The sensory register

In order for something to get into long-term memory, it must first "register" with us. The meaning we assign to sensory impressions depends on both our background knowledge and the context in which we experience something. As we've all experienced, our attention can be very selective. For example, some years ago I bought a kind of Volkwagen called a Type IV. These were only made for three years in the early '70s, so they were fairly rare. When I bought mine it was about 8 years old. I would have sworn that I had never seen a car like this before. (It looked sort of like a VW Beetle, only larger.) After I bought my car, I started seeing them all over the place. Well, not really, but I probably saw one at least three or four times a week. Did these cars just suddenly start showing up in Bloomington? I think not. They were there before, but they just didn't register with me. After I owned one, I was sensitized to them.

The role of context is obvious in language perception. For example, the word "tape" can have several meanings. But if I say, "I have to wrap a birthday present--do you know where the tape is?", most people would have no trouble perceiving that I'm looking for the roll of sticky stuff, not a video- or audiocassette. Also, you've probably had the experience of encountering someone who you recognize from one realm of your life (e.g., they work in your building) in another realm (e.g., at your daughter's soccer game). You may realize you know that person, but not be able to figure out why you know them. This is a context issue.

One of the problems that CIP researchers have wrestled with is, how do we recognize things? You should be familiar with the four different theories of pattern recognition discussed in the text.

Short-term memory

After a sensory impression has registered, it then passes into short-term memory, or working memory. The capacity of short-term memory appears to be rather limited. We can hold only about 7 "chunks" of information in short-term memory at a time. Of course, the size of a chunk is relative, not absolute. We might have trouble managing 7 single words in a language that was not known to us, but might easily be able to manage 7 sentences in our own language. The difference is that the unknown words are meaningless to us, whereas the sentences in our own language are meaningful, and therefore don't require as much working capacity.

What happens to information while it is in working memory determines whether--and how--it will get stored in long-term memory. We can hold things in working memory for a while by "rehearsing" them. An example of this is when we repeat a phone number or person's name to ourselves just long enough to dial a number or make an introduction. After we've used the information, it will probably be lost to us.

To get information stored in long-term memory requires that it be "encoded" in some way. Encoding can be accomplished in several ways. "Mnemonics" are memory tricks we can use to remember lists of names, numbers, etc. Many CIP researchers have been fascinated with what mnemonics can tell us about how memory works; however, these strategies are not in great demand in instructional settings today, since we usually don't consider memorization to be a very important learning outcome. In the longer term, the kinds of encoding strategies that work best are those that emphasize meaningfulness. One way to do this, with text for example, is to make the organizational structure of the material apparent. That's why we use outlines, headings, and other kinds of textual "cues" to indicate major and minor ideas, show relationships among concepts, etc.

Long-term memory

Some important concepts in long-term memory:

Declarative vs. procedural knowledge. Knowing "that" Bill Clinton was elected President of the U.S. is quite different from knowing "how" to conduct a successful presidential campaign. Similarly, memorizing the seven steps in the negotiating process is very different from being able to use those steps to negotiate successfully.

Declarative knowledge can be broken down in to episodic and semantic memory, or memory for events versus memory for verbal information. We have an episodic memory for the automobile accident we were in 13 years ago, but we have only a semantic memory of the "fact" that Columbus landed in the New World in 1492 (unless we were there).

Verbal and imaginal representation in memory. Words that have concrete referents (and therefore can easily be "pictured") are more likely to be remembered than abstract words. So, if I read a list of 30 words and ask you immediately afterwards to write down as many as you can remember, you will probably be more successful with words such as "skyscraper", "baboon", and "rake", than with words such as "strategy", "reference", and "nominal".

Retrieval. The difference between a "recall" task and a "recognition" task is an important one in education. As a student, I always liked multiple choice tests. Why? Because they usually require only recognition of some term or definition, which is much easier cognitively than an essay-type exam, which typically involves free "recall" tasks. However, life rarely presents us with multiple-choice options, so if we want school tasks to resemble life tasks... Well, you get the point.

Encoding specificity. The best retrieval cues are the same as the cues used for encoding. For example, remember the diagram of the theory-building process on page 7 of your text? Suppose I told you to memorize that for an exam. According to the concept of encoding specificity, it would probably be easier for you to reproduce the steps in that chart if I gave you a blank version exactly like the original chart than if I simply told you to list the seven steps. That's because the form of the chart (circular, with arrows) serves as a retrieval cue for the information.

Forgetting. Some theorists contend that we never truly "lose" anything once it is stored in long-term memory (unless the brain is damaged in some way). If that is so, then when we forget something, it must either be the case that it was never actually encoded in the first place, or that the information is still there, but we can no longer retrieve it. So the phone number that we repeated just long enough to dial is forgotten because it never really got into long-term memory. The phone number that my family had when I was a child is probably still stored in my brain, but I can no longer retrieve it. Likely, this number has suffered from retroactive interference, because of all the phone numbers I have had to learn and remember since then. (However, I could probably pick it out of a multiple-choice list!)

Some implications of CIP for instruction

Provide organized instruction. Make the structure and relations of the material evident to learners, such as through concept maps or other graphic representations.

Link new material with what is currently known. This provides a sort of mental "scaffolding" for the new material.

Recognize the limits of attention (sensory register). Help learners focus their attention through techniques such as identifying the most important points to be learned in advance of studying new material. 

Recognize the limitations of short-term memory. Use the concept of chunking: don't present 49 separate items, make them 7 groups of 7. Use elaboration and multiple contexts.

Match encoding strategies with the material to be learned. For example, don't encourage the use of mnemonic techniques unless it's really essential to memorize the material. If you want it to be processed more "deeply", then find encoding strategies that are more inherently meaningful.

Provide opportunities for both verbal and imaginal encoding. Even though it's not clear whether these are really two different systems, it does appear that imaging can help us remember. 

Arrange for a variety of practice opportunities. The goal is to help the learner generalize the concept, principle, or skill to be learned so that it can be applied outside of the original context in which it was taught.

Help learners become "self-regulated." Assist them in selecting and using appropriate learning strategies such as summarizing and questioning. 

Sidebar: Schema theory

Why do we need schema theory?

Suppose you overheard the following conversation between two college-age apartment-mates:

A: Did you order it?
B: Yeah, it will be here in about 45 minutes.
A: Oh... Well, I've got to leave before then. But save me a couple of slices, okay? And a beer or two to wash them down with?

Do you know what the roommates are talking about? Chances are, you're pretty sure they are discussing a pizza they have ordered. But how can you know this? You've never heard this exact conversation, so you're not recalling it from memory. And none of the defining qualities of pizza are represented here, except that it is usually served in slices, which is also true of many other things.

The other theories we discuss in this course would have a difficult time explaining how we can comprehend this conversation. Schema theory would suggest that we understand this because we have activated our schema for pizza (or perhaps our schema for "ordering pizza for delivery") and used that schema to comprehend this scenario. Schema theory attempts to explain how we actively make meaning of information.

What is a schema?

• Schemata are composed of generic or abstract knowledge; used to guide encoding, organization, and retrieval of information.
• Schemata reflect prototypical properties of experiences encountered by an individual, integrated over many instances.
• A schema may be formed and used without the individual's conscious awareness.
• Although schemata are assumed to reflect an individual's experience, they are also assumed to be shared across individuals [in a culture?].
• Once formed, schemata are thought to be relatively stable over time.
• We know more about how schemata are used than we do about how they are acquired.

• A play, in that it has a basic script, but each time it's performed, the details will differ.
• A theory, in that it enables us to make predictions from incomplete information, by filling in the missing details with "default values." (Of course, this can be a problem when it causes us to remember things we never actually saw...)
• A computer program, in that it enables us to actively evaluate and parse incoming information.

How are schemata created and modified?

Schemata are created through experience with people, objects, and events in the world. When we encounter something repeatedly, such as a restaurant, we begin to generalize across our restaurant experiences to develop an abstracted, generic set of expectations about what we will encounter in a restaurant. This is useful, because if someone tells you a story about eating in a restaurant, they don't have to provide all of the details about being seated, giving their order to the server, leaving a tip at the end, etc., because your schema for the restaurant experience can fill in these missing details.

Not all of the information we have about restaurants necessarily gets added to our schema. For example, there's a restaurant in Indianapolis where the seating booths are little jail cells. After you're seated, the server closes your cell doors. (Of course, you can escape any time you want, as long as you've paid your bill.) Even though I've been to this restaurant several times, I don't think my restaurant schema includes tables as miniature jail cells. This information is simply an outlier; it is too unlike my experience at other restaurants.

• Accretion: New information is remembered in the context of an existing schema, without altering that schema. For example, suppose I go to a bookstore, and everything I experience there is consistent with my expectations for a bookstore "experience." I can remember the details of my visit, but since they match my existing schema, they don't really alter that schema in any significant way. 
• Tuning: New information or experience cannot be fully accommodated under an existing schema, so the schema evolves to become more consistent with experience. For example, when I first encountered a bookstore with a coffee bar, I probably had to modify my bookstore schema to accommodate this experience. 
• Restructuring: When new information cannot be accommodated merely by tuning an existing schema, it results in the creation of new schema. For example, my experience with World Wide Web-based bookstores may be so different from my experience with conventional ones that I am forced to create a new schema. 

Why is schema theory important in teaching and learning?

• Schematic knowledge has a significant effect on organization of ambiguous or disorganized stories.
• Narrative schemata specify expected components of a story, such as the time sequence of events, and causal relations that should connect the events; during encoding or retrieval of a story, missing events may be inferred to fill in omitted information, and events may be reordered to correspond to a real-time sequence.
• Many studies have shown that the use of schematic knowledge is so powerful that listeners have little control over the retrieval strategies used during recall of narrative information; even when listeners are instructed to reproduce texts verbatim, they cannot do so when the text contains certain types of omissions or certain sequences of events.

For example, consider the following excerpt from a story:

The girl sat looking at her piggy bank. "Old friend," she thought, "this hurts me." A tear rolled down her cheek. She hesitated, then picked up her tap shoe by the toe and raised her arm. Crash! Pieces of Piggo--that was its name--rained in all directions. She closed her eyes for a moment to block out the sight. Then she began to do what she had to do.

If you have a well-developed schema for "piggy banks", this story should be readily comprehensible. You would understand that traditional piggy banks were usually made of some fragile, brittle material, that they contained a slot for inserting and saving coins, and that the money could only be removed by breaking them.

On the other hand, if you have no schema for piggy bank, the story probably makes little sense.

What are some implications of schema theory for instruction?

• Provide unifying themes for content, since information that lacks a theme can be difficult to comprehend, or, worse, the learner may "accrete" the information to the wrong schema.
• Choose texts with "standard" arrangement so that they conform to student expectations.
• Encourage students to read titles and headings.
• Point out the structure of particular kinds of texts; e.g., what are the common features of published research articles?
• Ask questions to determine what students' current schemata might be.
• Pay attention to student answers and remarks that may give clues about how they are organizing information; i.e., what schemata are they using?

Problem solving and critical thinking

 We often think of problem solving in a narrow way, such as getting the right answer on a math or physics question. However, when problem solving is defined in a broader way, much of human behavior can be framed as problem solving. What clothes should I wear today? What route should I take to work? Where should I eat lunch? How should I approach this new project or class assignment? How can I get my daughter to go to bed at 8:30? Each of these fits the general definition of a problem: a desired state of affairs that differs from the current state of affairs. (Example: The current state is that it is lunchtime and I am hungry; the desired state is that in an hour I want to be back at work having consumed a satisfying, nutritious meal at a reasonable price.)

Note that we often solve such problems without conscious awareness. Which raises a question: Could we spend more of our time in ³desired² states if we approached problems with greater awareness?

Some useful concepts in problem solving

Well-defined vs. ill-defined problems. Well-defined problems are the kind we often encounter in math or science classes, as noted above. There is usually one correct answer and an algorithm, or set of rules, that virtually guarantees finding the right answer when correctly applied. Unfortunately, nearly all of the important problems in the world are of the second kind‹ill-defined. These problems have multiple possible solutions and instead of a well-defined algorithm for solving them we have, at best, some heuristics or guidelines that do not guarantee a correct solution.

Functional fixedness. Our ability to solve some problems may be limited by our inability to see objects in a new light or consider solutions that fall outside of a set of unnecessary constraints we have imposed. For example, if I am fixed on finding a quick, pleasant route to drive to work each day, I may never consider alternatives such as taking the bus or working at home.

Identifying the problem. This is more than just the first step in the general problem solving model. It is crucial, because if we cannot identify that a problem exists, then obviously we cannot solve it. Note that in school learning it is often the teacher or textbook that identifies problems; the studentıs only task is to solve a problem once identified. In the real world, though, problems are often not easy to identify. For example, floods create misery for thousands of people each year in the U.S. and many other countries. But what, precisely, is the problem? Has global warming led to changing weather patterns? Do we need more dams, or fewer? Or do we need laws that prohibit building houses in flood plains? Unless we can frame such problems appropriately, we have little chance of solving them satisfactorily. Donald Schon, in his work on reflective practice, contends that the ability to frame (identify) problems is one of the most important factors that distinguishes experts from novices.

Domain-specific knowledge and the problem of transfer. In addition to their greater ability to identify problems, another important factor that characterizes the problem-solving ability of experts is the amount of domain-specific knowledge they can bring to bear on a task. No matter how much generalized problem-solving ability I have, I am more likely to solve any particular problem successfully if I have relevant domain-specific knowledge. For example, if the problem I have set for myself is to buy a new car that meets my familyıs needs and budget, I am more likely to succeed the more knowledge I have about vehicle types, makes, reliability ratings, gas mileage, availability of local service, etc.

Critical thinking

Critical thinking is more difficult to define than problem solving. It is a broader term, and it means different things in different disciplines. It may help to think of it as more process-oriented, while problem solving is goal-oriented. The best way to understand what is meant by critical thinking is probably to consider the kinds of abilities that are thought to comprise it: analyzing arguments; using deductive and inductive reasoning; judging the reasoning of others; making value judgments; etc.

Learning to read and reading to learn

When you think about the act of reading in light of the basic information processing model we discussed in the previous unit, it seems like a truly remarkable feat. For example, each word we read must be attended to and must stay in working memory long enough for us to comprehend the overall sentence or paragraph. Otherwise, we would understand individual words, but have no idea of what the text as a whole meant! There are several aspects of information processing that likely aid us in this seemingly difficult task:

Propositional encoding. Fortunately, when we read it is usually not necessary to remember words and sentences in their exact written form. It appears that we convert reading material into sets of propositions that carry the essential meaning while reducing the burden on our memories. For example, consider this passage:

Josh leaned back and tossed the football in a high, lazy arc. Maria looked up to see it spiraling toward her and caught it easily in her outstretched hands.

We would likely encode such a passage as two simple propositions: Josh threw the football. Maria caught the football. Or perhaps as just one: Josh threw the football to Maria.

Chunking. Beginning readers are still struggling with the meaning of words and may have difficulty holding more than a few words at a time in working memory. Experienced readers, on the other hand, can manage larger chunks, or units of information, and can likely hold the meaning of several sentences in working memory.

Bottom-up and top-down processing. We appear to use both of these processes in reading. We use bottom-up processing when we read words or sentences and encode them in memory. We use top-down processing when our initial understanding of a passage activates relevant schemata which, in turn, shape our expectations of what is to come and aid in our comprehension of the material.

Automaticity. As with any cognitive activity, our skill improves when some aspects of the task become automatic. So, for example, when we can decode the meaning of most words without conscious effort, more of our time and attention can be directed to comprehending the overall meaning of what we are reading.

Meaningfulness. One of the fundamental principles of cognition is that new information is meaningful to the extent to which it can be related to what we already know. The same is true for reading. That is why beginning readers, in particular, should be provided with material that contains mostly familiar settings and concepts.

Metalinguistic awareness. Think of this as a subset of metacognition that has particular relevance to reading. In order to be a successful reader we must be aware of a number of things that most of us take for granted: That printed words correspond to spoken language; that sentences generally consist of one or more propositions; that propositions usually have a subject and a verb, and often an object; etc. Most children seem to develop sufficient metalinguistic awareness without these things being taught explicitly.

Some other issues related to reading and cognition:

Definitional versus contextual word knowledge. This closely parallels the distinction in concept learning between knowing the definition of a concept and being able to use it to correctly classify things we encounter in the world. Merely having the textbook definition of the concept mammal, for example, does not necessarily mean that we could correctly classify bats and whales as mammals if we encountered them. Similarly, with word knowledge, knowing the definition of irascible is not the same as being able to use it appropriately in speech or writing.

Schema theory and text comprehension. Schema theory holds that, when appropriate schemata are activated as we read, they facilitate comprehension and shape our expectations about what is to come. (Of course, when inappropriate schemata are activated the likely result will be confusion rather than comprehension.) This explains why well-written titles and headings can be useful. Advance organizers might be thought of as a way to activate an appropriate schema, or even to provide one when none exists in the learner.

Cognitive approaches to mathematics and science: Implications for instruction

Build on studentsı informal knowledge. As in any domain, new material is learned more meaningfully when it can be related to what the learner already knows.

Identify students' current "theories" or algorithms.

Use student errors as a source of information about their mental models. These ³buggy algorithms² and naïve misconceptions are typically based on uncontrolled observations and need to be brought to the studentıs awareness before they can be challenged.

Use "think aloud" activities, since these help to uncover current models.

Model real problem-solving for students. Students need to see that solving problems is not just a matter of plugging numbers into an algorithm; rather it is a matter of determining the kind of problem so that an algorithm can be successfully applied.

Explicitly teach problem-solving strategies. Donıt expect that students will acquire appropriate strategies merely by seeing the teacher use them. Students will need guided, hands-on experience in using these.

Focus on processes, structures, and decisions, not answers. If students have a broad, conceptual understanding they will more likely be able to solve other kinds of problems in the future, not just the limited set they encounter in school.

Provide a mix of problem types, rather than grouping problems of one type; otherwise, students won't develop skill at determining the problem type and choosing an appropriate solution strategy.

Learning activities

Learning Activities for Session Three