Handbook of psychology volume 7 educational psychology

Memory and Information Processes

RICHARD E. MAYER

47

AN INFORMATION PROCESSING VIEW OF LEARNING

AND COGNITION

47

HISTORICAL OVERVIEW

47

Associationist View

47

Gestalt View

48

TWO VIEWS OF INFORMATION PROCESSING THEORY

48

Classical View

48

Constructivist View 

49

MAJOR CONTRIBUTIONS OF INFORMATION

PROCESSING THEORY

50

Cognitive Processes: Cognitive Task Analysis

50

Mental Representations: Types of Knowledge

50

Cognitive System: Architecture of the

Cognitive System

51

INFORMATION PROCESSING AND INSTRUCTION

53

Information Processing in Reading a Passage

53

Information Processing in Writing an Essay

54

Information Processing in Solving a

Mathematics Problem

54

CONCLUSION

55

REFERENCES

56

AN INFORMATION PROCESSING VIEW OF

LEARNING AND COGNITION

How does the human mind work? What happens when some-

one learns or when someone solves a problem? According

to the information processing view, the human mind works

by forming mental representations and applying cognitive

processes to them. This definition has two elements: (a) The

content of cognition is mental representations, and (b) the ac-

tivity of cognition involves cognitive processes. In learning,

the learner takes incoming information received through the

eyes or ears and applies a series of cognitive processes to the

incoming information, resulting in the construction of a se-

ries of mental representations. For example, as you read the

words in this paragraph you form a series of mental represen-

tations by applying appropriate cognitive processes such as

mentally selecting important ideas, mentally organizing them

into a coherent cognitive structure, and mentally relating

them with prior knowledge. In this chapter I provide a brief

historical overview of the precursors to the information pro-

cessing view of learning and cognition, describe two versions

of the information processing view, examine three major

contributions of the information processing view, and then

exemplify how it contributes to theories of learning and

cognition.

HISTORICAL OVERVIEW

For more than 100 years psychologists have conducted re-

search aimed at understanding how knowledge is represented

and processed in human minds. Such issues fell under the

domain of science as psychology entered the twentieth cen-

tury, heralded by the publication of Ebbinghaus’s pioneering

memory studies in 1885 (Ebbinghaus, 1964) and Thorndike’s

pioneering learning studies in 1898 (Thorndike, 1965). Dur-

ing the first half of the twentieth century two competing

views of learning emerged—the associationist view of learn-

ing as strengthening of associations and the Gestalt view of

learning as building cognitive structures.

Associationist View

According to the associationist view, the content of cognition

consists of nodes and associations between them and the

process of cognition consists of the strengthening and weak-

ening of associations. For example, in Thorndike’s (1965)

classic study of animal learning, a hungry cat was placed in a

wooden box. The cat could escape by pulling a hanging loop

of string that opened a door allowing the cat to get out and eat

some nearby food. Thorndike noted that on the first day, the

cat engaged in many extraneous behaviors before accidentally

48

Memory and Information Processes

pulling the string, but on successive days the number of extra-

neous behaviors decreased. After many days, the cat pulled

the loop of string shortly after being placed in the box.

According to Thorndike, the cat began with a habit family

hierarchy—an ordered set of responses associated with being

placed in an enclosed box. The cat would try the most strongly

associated response first (e.g., thrusting its paw through the

slats of the box), and when it failed, the strength of the associ-

ation to that response would be weakened. Eventually, the cat

would pull the loop of string and get out, thus increasing the

association to that response. Over many days, the extraneous

responses became very weakly associated with being in the

box, and pulling the string became very strongly associated

with being in the box. Thus, Thorndike offered a clear vision

of learning as the strengthening and weakening of stimulus-

response (S-R) associations and memory as the processing of

linked nodes in a network—a vision that dominated psychol-

ogy through the 1950s and still flourishes today in revised

form.

Gestalt View

According to the Gestalt view, the content of cognition con-

sists of coherent structures, and the process of cognition con-

sists of building them. For example, Kohler (1925) placed an

ape in a pen with crates on the ground and a bunch of bananas

hanging overhead out of reach. Kohler observed that the ape

looked around and then suddenly placed the crates on top

of one another to form a ladder leading to the bananas,

allowing the ape to climb the stairs and grasp the bananas.

According to Kohler, the ape learned by insight—mentally re-

organizing the objects in the situation so they fit together in a

way that accomplished the goal. Thus, insight is a process of

structure building (Mayer, 1995). The Gestalt approach rose

to prominence in the 1930s and 1940s but is rarely mentioned

today. Nonetheless, the Gestalt theme of cognition as structure

building underlies core topics in cognitive science including

the idea of schemas, analogical reasoning, and meaningful

learning.

By the 1950s and 1960s, the associationist and Gestalt

views were reshaped into a new view of cognition, called

information processing (Lachman, Lachman, & Butterfield,

1979). The information processing view eventually became

the centerpiece of cognitive science—the interdisciplinary

study of cognition. A core premise in cognitive science is that

cognition involves computation; that is, cognition occurs

when you begin with a representation as input, apply a process,

and create a representation as output. For example, in a review

of the field of cognitive science, Johnson-Laird (1988, p. 9)

noted, “Cognitive science, sometime explicitly and sometimes

implicitly, tries to elucidate the workings of the mind by treat-

ing them as computations.” Human cognition on any task can

be described as a series of cognitive processes (i.e., a descrip-

tion of the computations that were carried out) or as a series of

transformations of mental representations (i.e., a description

of the inputs and outputs for each computation).

TWO VIEWS OF INFORMATION

PROCESSING THEORY

A central problem of the information processing approach is

to clarify the nature of mental representations and the nature

of cognitive processes. This task is made more difficult by the

fact that researchers cannot directly observe the mental rep-

resentations and cognitive processes of other people. Rather,

researchers must devise methods that allow them to infer

the mental representations and cognitive processes of others

based on their behavior (including physiological responses).

In the evolution of the information processing approach to

learning and memory, there have been two contrasting ver-

sions: the classical and constructivist view (Mayer, 1992a,

1996a).

Leary (1990) showed how progress in psychological theo-

ries can be described as a progression of metaphors, and

Mayer (1992a, 2001) described several major metaphors of

learning and memory that have emerged during the last cen-

tury, including viewing knowledge as information versus

viewing knowledge as cognitive structure. A major challenge

of the information processing view—and the field of cogni-

tive science that it serves—is to clarify the status of the

knowledge as information metaphor (which is part of the clas-

sical view) and the knowledge as cognitive structure metaphor

(which is part of the constructivist view).

Classical View

The classic view is based on a human-machine metaphor in

which the human mind is like a computer; knowledge is rep-

resented as data that can be processed by a computer, and

cognition is represented as a program that specifies how data

are processed. According to the classical view, humans are

processors of information. Information is a commodity that

can be transferred from one mind to another as a series of

symbols. Processing involves applying an algorithm to infor-

mation such that a series of symbols is manipulated accord-

ing to a step-by-step procedure. For example, when given a

problem such as “x

ϩ 2 ϭ 4, solve for x,” a learner forms a

mental representation of the problem such as “x

ϩ 2 ϭ 4”

and applies operators such as mentally subtracting 2 to both

Two Views of Information Processing Theory

49

sides in order to generate a new mental representation,

namely “x

ϭ 2.”

The classical information processing approach developed

in the 1950s, 1960s, and 1970s, although its roots predate

psychology (Lachman et al., 1979). For example, more than

250 years ago De La Mettrie (1748/1912) explored the idea

that the human mind works like a complex machine, and the

classical information processing view can be seen in Atkinson

and Shiffrin’s (1968) theory of the human memory system

and Newell and Simon’s (1972) theory of human problem

solving.

For example, Newell and Simon (1972) developed a

computer simulation designed to solve a variety of prob-

lems ranging from chess to logic to cryptarithmetic. In the

problem-solving program, information consists of “symbol

structures” (p. 23) such as a list, tree, or network, and pro-

cessing consists of “executing sequences of elementary infor-

mation process” (p. 30) on symbol structures. A problem is

represented as a problem space consisting of the initial state,

the goal state, and all possible intervening states with links

among them. The process of searching the space is accom-

plished by a problem-solving strategy called means-ends

analysis, in which the problem solver sets a goal and carries

it out if possible or determines an obstacle that must be over-

come if it is not (see Mayer, 1992b). Thus, problem solving

involves applying processes to a symbolic representation of a

problem: If the application is successful, the representation is

changed; if it is not successful, a new process is selected

based on a means-ends analysis strategy. In a complex prob-

lem, a long series of information processes may be applied,

and many successive representations of the problem state

may be created. 

Two limitations of the classical view—humans as infor-

mation processors—concern the characterization of informa-

tion as an objective commodity and the characterization of

processing as the application of algorithms. Although such

characterizations may mesh well with highly contrived labo-

ratory tasks, they appear too limited to account for the full

range of human learning in complex real-world situations.

For example, Metcalfe (1986a, 1986b; Metcalfe & Wiebe,

1987) showed that people use different cognitive processing

for insight problems (requiring a major reorganization of the

problem) and noninsight problems (requiring the step-by-step

application of a series of cognitive processes). For insight

problems people are not able to predict how close they are to

solving the problem (inconsistent with the step-by-step think-

ing posited by the classical view), but for noninsight prob-

lems they are able to gage how close they are to solution

(consistent with the step-by-step thinking posited by the

classical view). Apparently, the classical view may offer a

reasonable account of how people think about noninsight

problems but not how they think about insight problems.

Constructivist View

The constructivist view is based on the knowledge con-

struction metaphor, in which the human mind is a sort of con-

struction zone in which learners actively create their own

knowledge based on integrating what is presented and what

they already know. According to the constructivist view, learn-

ers are sense makers who construct knowledge. Knowledge is

a mental representation that exists in a human mind. Unlike in-

formation, which is an objective entity that can be moved from

one mind to another, knowledge is a personal construction that

cannot be moved directly from one mind to another. Construc-

tion involves cognitive processing aimed at sense making,

including attending to relevant portions of the presented mate-

rial, mentally organizing the material into a coherent structure,

and mentally integrating the material with relevant existing

knowledge. Unlike the view of cognitive processing as apply-

ing algorithms, cognitive processing involves orchestrating

cognitive strategies aimed at sense making. For example, as

you read this section, you may mentally select relevant ideas

such as the classical view of information and processing and

the constructivist view of knowledge and construction; you

may organize them into a matrix with classical and construc-

tivist as rows and nature of information and nature of process-

ing as columns; and you may integrate this material with your

previous knowledge about these topics.

The constructivist approach developed in the 1980s and

1990s, although its earlier proponents include Bartlett’s

(1932) theory of how people remember stories and Piaget’s

(1971) theory of how children learn. For example, Bartlett

argued that when learners are presented with a folk story, they

assimilate story elements to their existing schemas and men-

tally reorganize the story in a way that makes sense to them.

Similarly, Piaget showed how children assimilate their experi-

ences with their existing schemas in an attempt to make sense

of their environment. More recently, the constructivist view

can be seen in Ausubel’s (1968) theory of assimilative learning

and Wittrock’s (1990) theory of generative learning. In both

theories, learning involves connecting what is presented with

what the learner already knows, so the outcome of learning de-

pends both on the material presented by the instructor and the

schemas used by the learner.

Although the constructivist view addresses some of the

limitations of the classical view, major limitations of the con-

structivist view include the need to account for the social and

cultural context of cognition and the need to account for the

biological and affective bases of cognition. In particular,

50

Memory and Information Processes

the constructivist view focuses on cognitive changes within

individual learners, but this view can be expanded by consid-

ering how the learner’s cognitive processing is mediated by

the learner’s surrounding social and cultural environment.

The constructivist view focuses on what can be called cold

cognition (i.e., cognitive processing in isolation), but this

view can be expanded by also considering the role of the

learner’s emotional and motivational state. 

MAJOR CONTRIBUTIONS OF INFORMATION

PROCESSING THEORY

Three important contributions of the information process-

ing approach are techniques for analyzing cognitive process-

ing (e.g., “What are the cognitive processes involved in

carrying out a cognitive task?”), techniques for analyzing men-

tal representations (e.g., “How is knowledge represented in

memory?”), and a general description of the architecture of the

human cognitive system (e.g., “How does information flow

through the human memory system?”).

Cognitive Processes: Cognitive Task Analysis

A fundamental contribution of information processing theory

is cognitive task analysis—techniques for describing the cog-

nitive processes that a person must carry out to accomplish a

cognitive task. For example, consider the analogy problem

dog : bark :: cat : ____, which can be read as “dog is to

bark as cat is to what?” and in which the a-term is “dog,” the

b-term is “bark,” the c-term is “cat,” and the d-term is un-

known. What are the cognitive processes that a problem

solver must go through to solve this problem? Based on a

cognitive task analysis, solving an analogy problem can be

broken down into five basic steps (Mayer, 1987; Sternberg,

1977):

1. Encoding—that is, reading and forming a mental repre-

sentation of the words and accompanying punctuation,

2. Inferring—that is, determining the relation between the

a-term and the b-term (e.g., the b-term is the sound that

the a-term makes),

3. Mapping—this is, determining what the c-term is and how

it corresponds to the a-term (e.g., the a-term is a kind of

animal that makes sounds, and the c-term is another kind

of animal that makes sounds),

4. Applying—that is, generating a d-term based on applying

the relational rule to the c-term (e.g., the sound that the

c-term makes is _____), and

5. Responding—that is, physically making the response such

as writing “meow” or circling the correct answer

(“meow”) on a list. 

Cognitive task analysis has useful educational applications

because it suggests specific cognitive processes that students

need to learn. For example, the cognitive task analysis of

analogy problems suggests that students would benefit from

instruction in how to infer the relation between the a-term

and the b-term (Sternberg, 1977).

To test this idea, Sternberg and Ketron (1982) taught col-

lege students how to solve analogy problems by showing them

how to infer the change from the a-term to the b-term and how

to apply that change to the c-term. On a subsequent test of ana-

logical reasoning involving new problems, trained students

solved the problems twice as fast and committed half as many

errors as did students who had not received training.

Cognitive task analysis also offers advantages in evaluat-

ing student learning outcomes. For example, instead of mea-

suring the percentage correct on a test, it is possible to specify

more precisely the knowledge that a student possesses—

including incomplete or incorrect components. For example,

suppose a student gives the following answers on an arith-

metic test:

234

678

456

545

Ϫ156

Ϫ434

Ϫ327

Ϫ295

122

244

131

350

A traditional evaluation would reveal that the student cor-

rectly solved 25% of the problems. However, a cognitive task

analysis reveals that the student seems to be consistently ap-

plying a subtraction procedure that has one incorrect step, or

bug—namely, subtracting the smaller number from the larger

number in each column (Brown & Burton, 1978). In specify-

ing the procedure that the student is using, it becomes clear

that instruction is needed to help the student replace this

smaller-from-larger bug. 

Mental Representations: Types of Knowledge

According to the information processing approach, knowl-

edge is at the center of cognition: Learning is the construction

of knowledge; memory is the storage of knowledge; and

thinking is the logical manipulation of knowledge. Therefore,

information processing theorists have analyzed the types of

knowledge (or mental representations): factual, conceptual,

procedural, and metacognitive (Anderson et al., 2001). Fac-

tual knowledge consists of facts—that is, simple descriptions

of an object or element (e.g., “apples are red”). Conceptual

Major Contributions of Information Processing Theory

51

Figure 3.1

An information processing model of how the human mind works.

knowledge involves relations among elements within a co-

herent structure that enables them to function together, and

includes classification hierarchies, cause-and-effect models,

explanatory principles, and organizing generalizations (e.g.,

the model presented in Figure 3.1). Procedural knowledge in-

volves a procedure, method, or algorithm—that is, a step-by-

step specification of how to do something (e.g., the procedure

for how to carry out long division). Metacognitive knowl-

edge involves strategies for how to coordinate one’s cogni-

tive processing (e.g., knowing how to monitor the quality of

one’s essay-writing activity). As you can see, factual and

conceptual knowledge are knowledge of “what” (i.e., data

structures), whereas procedural and metacognitive knowl-

edge are knowledge of “how to” (i.e., processes for manipu-

lating data structures). 

Knowledge is a mental representation: It is mental because

it exists only in human minds; it is a representation because it

is intended to denote or signify something. Representations

can be classified based on the coding system used to represent

them in the cognitive system such as motoric (e.g., bodily

movement images), pictorial (e.g., mental images), verbal

(e.g., words), or symbolic (e.g., some higher level coding sys-

tem). Representations can be classified based on the input

modality including haptic/kinesthetic/vestibular (e.g., bodily

sensations), visual (e.g., imagery sensations), or auditory

(e.g., acoustic sensations).