Is it all in the telling?


Material Information

Is it all in the telling? a study of the role of text schemas and schematic text structures in the recall and comprehension of printed news stories
Physical Description:
xiii, 223 leaves : ill. ; 29 cm.
Durham, Meenakshi Gigi, 1961-
Publication Date:


Subjects / Keywords:
Reporters and reporting   ( lcsh )
Reading comprehension   ( lcsh )
Schemas (Psychology)   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1990.
Includes bibliographical references (leaves 211-222).
Statement of Responsibility:
by Meenakshi Gigi Durham.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001687943
notis - AHZ9973
oclc - 25116707
System ID:

Full Text








Copyright 1990


Meenakshi Gigi Durham

For Frank

and for the rest of my wonderful family.


Producing one of the first doctoral dissertations from

the University of Florida's College of Journalism and

Communications has involved enormous struggles and,

ultimately, enormous rewards.

I could not have seen my way through it without the

encouragement and guidance of my committee members--Dr.

Leonard Tipton, Dr. Kurt Kent, Dr. Mickie Edwardson, Dr.

Nora Hoover, and Dr. Tom Fillmer. I thank them individually

and collectively for giving me so much of their time and

erudition during the research and writing of this

dissertation. Special thanks to Dr. Edwardson and Dr.

Fillmer for allowing me to conduct part of my research in

their undergraduate classes.

Dr. Thomas Abbott and fellow-Ph.D.-student Polly Moore

Shipp rallied to the cause when I cried for help with

finding experimental subjects. Heartfelt thanks to both of

them for letting me into their classrooms to badger their


Many thanks to my friends and coworkers at the Graduate

School for their understanding and sympathy, especially when

I was in the last throes of dissertation dementia. Dan

Hanson and Sherman Martin's help with my modem and with the


final printing of this work will not be soon forgotten.

Julie Shih was the kindest and most generous of supervisors,

allowing me time off whenever I needed it to research and

write this dissertation. Special thanks to Rose Barnett for

her sympathy, encouragement, and free counseling.

Many thanks also to my parents (without whom I would

not be in college at UF), Debbie Burke, Claire Walsh, Renee

and Hamilton Williams, Vishu Kulkarni, Mona Durham, and

finally to Dr. Roger Keroack for prodding me at just the

right time to dislodge me from a quicksand of inertia.

This dissertation is above all the product of years of

unyielding love, patience, and support from my husband

Frank. With all my heart, I thank him. Now he can be a

doctor's husband.



LIST OF TABLES .. . .. vii

LIST OF FIGURES .. .. . .. xi

ABSTRACT .. . .... ... .... xii

Research Question and Justification for This Study 1
Text Schemas and Text Structures: An
Overview . . 2
Goal of This Research . 3
The Concept of Schema . 4
Definition of Schema . 5
Functioning of Schemas . 6
Schema-Theoretic Information Processing and Its
Application to the Mass Communication Model 9

The Foundations of Schema Theory . 14
Bartlett's Experiments On Remembering 15
Minsky's Frameworks for the Representation of
Knowledge . 19
Schema-Theoretic Approaches to Reading
Comprehension .... 20
Text Structures and the Reading Process .. 22
Schemas for the Structure of News Stories 28
Limitations of the Schema Concept .. 39
Predicted Associations. . ... 42
Nominal Definitions . 43

METHOD . . .
Hypotheses . .
Operational Definitions .
The Independent Variables .
The Dependent Variables .
Reliability and Validity of
Instruments . .
Reliability . .
Validity . .
The Control of Possibly
Variables .
The Experimental Design .
Subjects . .
Method . ..


. a .
. .
. .
. .

Some Considerations: Threats to External
and Internal Validity . .





General Results . 80
Effects of Altering the Story Structure on
Short-Term Recall ........ 84
Effects of Altering the Story Structure on
Long-Term Recall . 88
Effects of Altering the Story Structure on
Comprehension . 93
Summary . . 97
Short-Term Recall of Text Content:
A Closer Look .. . 98
Summary of Analysis of Short-Term Recall
Protocols . 109
Some Theoretical Implications:
Correspondence of Strength of
Text Schema With Recall and
Comprehension . 111
Comprehension: A Closer Look . .130
Summary of In-Depth Analysis of Comprehension
Scores . . 137
General Summary . . 138

DISCUSSION . . .. 141
Theoretical Implications . 141
Problems Encountered and Possible Solutions 147
Suggestions For Further Research . 150
Implications for the Field . 152


SCORING SHEETS .. .... . 185






Table 3-1. Descriptive statistics on the stimulus
used to measure recall and comprehension ... 54

Table 3-2. Descriptive statistics on the stimulus
passages used to measure strength of text schema 57

Table 3-3. Reliability coefficients of measurement
instruments . . .. 61

Table 3-4. Summary table of descriptive statistics of
sample subjects . .. 70

Table 3-5. Conceptual and operational definitions 78

Table 4-1. Analysis of variance of short-term recall
across text structure . .. 82

Table 4-2. Analysis of variance of long-term recall
across text structure . ... 83

Table 4-3. Analysis of variance of comprehension across
text structure ... . 84

Table 4-4. Comparison of mean short-term recall scores
of narrative story group with news story group 86

Table 4-5. Comparison of mean short-term recall scores
of expository text group with news story group 88

Table 4-6. Comparison of mean short-term recall scores
of narrative story group with expository text
group .. . . ... 89

Table 4-7. Comparison of mean long-term recall scores
of narrative story group with news story group 90

Table 4-8. Comparison of mean long-term recall scores
of expository text group with news story group 91

Table 4-9. Comparison of mean long-term recall scores
of narrative story group with expository text
group . ... .. 92

Table 4-10. Comparison of mean comprehension scores of
narrative story group with news story group 94

Table 4-11. Comparison of mean comprehension scores of
expository text group with news story group 94


Table 4-12. Comparison of mean comprehension scores of
narrative story group with expository text group 96

Table 4-13. Percentage of recall of the propositions
representing each terminal node for the expository
version of the panther passage .. 100

Table 4-14. Percentage of recall of the propositions
representing each terminal node for the narrative
version of the panther passage . 102

Table 4-15. Percentage of recall of the propositions
representing each terminal node for the news
version of the panther passage . 104

Table 4-16. Percentage of recall of the propositions
representing each terminal node for the expository
version of the political uprising passage .. 106

Table 4-17. Percentage of recall of the propositions
representing each terminal node for the narrative
version of the political uprising passage 108

Table 4-18. Percentage of recall of the propositions
representing each terminal node for the news
version of the political uprising passage .. 110

Table 4-19. Comparison of mean strength of schema scores
of narrative story group with news story group 112

Table 4-20. Comparison of mean strength of schema
scores of expository text group with news story
group . . 113

Table 4-21. Comparison of mean strength of news schema
scores between infrequent newspaper readers and
frequent newspaper readers . 114

Table 4-22. Correlations between strength of text
schema and short-term recall . 115

Table 4-23. Comparison of mean comprehension scores of
subjects with high prior knowledge versus subjects
with low prior knowledge . 124

Table 4-24. Comparison of mean comprehension scores of
subjects with high interest in the passage topic
versus subjects with low interest in the passage
topic ...... ...... ......... .. 126

Table 4-25. Crosstabulation of frequency of news media
use with interest in topic of passage .... .127

Table 4-26. Comparison of mean comprehension scores on
textually explicit, textually implicit, and
scriptally implicit questions ... .132


Figure 2-1. Fedler's inverted pyramid . 30

Figure 2-2. Diagrammtic representation of van Dijk's
news macrostructure . 33

Figure 2-3. Newsom and Wollert's traditional inverted
pyramid ........ ........ ..... 35

Figure 2-4. Newsom and Wollert's modified inverted
pyramid . . 36

Figure 4-1. Plot of short-term recall scores against
strength of text schema for panther passage, news
version . . 117

Figure 4-2. Plot of short-term recall scores against
strength of text schema for political uprising
passage, news version . 118

Figure 4-3. Plot of short-term recall scores against
strength of text schema for panther passage,
narrative version . ...... 119

Figure 4-4. Plot of short-term recall scores against
strength of text schema for political uprising
passage, narrative version . 120

Figure 4-5. Plot of short-term recall scores against
strength of text schema for panther passage,
expository version . 121

Figure 4-6. Plot of short-term recall scores against
strength of text schema for political uprising
passage, expository version . .. 122

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Meenakshi Gigi Durham

May 1990

Chairperson: Dr. Leonard P. Tipton
Major Department: Journalism and Communications

On the basis of schema-theoretic approaches to the

reading process, this study focused on the extent to which

the underlying structure of news stories and readers'

schemas for text structures affect the comprehension and

recall of the stories' content.

Because readers' awareness of text structures has been

shown to influence their cognitive processing of text

content, it was hypothesized that altering a news story so

that the text conformed to a more familiar structure might

increase comprehension and recall.

Two stimulus news passages were rewritten twice--once

to conform to a story grammar and once to conform to an

expository/attribution text structure. Subjects (n=104)

read a stimulus passage, performed distractor tasks


including a measure of the strength of the subject's schema

for the stimulus passage's text structure, and then

responded to measures of free short-term and long-term

recall and comprehension.

Readers were found to have significantly higher short-

term recall of the narrative and expository versions of both

text passages than the news stories, E (2,50) = 2.881, p =

0.065 for Passage I and F (2,45) = 3.529, p = 0.038 for

Passage II. However, subjects' comprehension of the two

passages did not appear to be significantly affected by the

experimental manipulation overall, E (2,50) = 1.985, p =

0.148 in the case of Passage I, and (2,45) = 0.179, p =

0.836 in the case of Passage II. Long-term recall was

significantly affected by alteration of text structure for

Passage I, E (2,35) = 3.102, p = .057, but not for Passage

II, E (2,25) = 0.907, p = 0.416.

Subjects' interest in the topic of the passage was

found to influence comprehension in the case of Passage I (t

= 2.36, p = 0.011) but not for Passage II (t = 1.21, E =




News writing has traditionally aimed at presenting

information in a clear, concise manner designed to render

the news content of a story accessible to as large a number

of readers as possible (McCombs & Becker, 1979, p. 18).

Reporters are exhorted to write simply and clearly, using

short sentences and familiar words (cf. Berner, 1984;

Hutchison, 1986; Jones, 1978). Mencher (1984) notes, "News

is information people need in order to make rational

decisions about their lives" (p. 77). If the information is

meant to be useful, it follows that an ideal news story

would be one that a reader could easily comprehend and also

could easily recall when the information in it needs to be


Research Question and Justification for This Study

The goal of this research was to investigate a specific

domain within the broad topic of schematic cognition and its

bearing on the comprehension and recall of news. In

particular, this study focused on the structure of print

news messages and the ways in which readers' schemas for

such structure influence their interaction with the content

of the messages.

Text Schemas and Text Structures: An Overview

Numerous studies indicate that a reader's familiarity

with the way in which information is organized in a text has

a significant impact on how well he/she comprehends and

remembers it (Adams & Collins, 1977; Anderson, 1977;

Marshall & Glock, 1978-79; Taylor, 1980; Pearson &

Camperell, 1985; McGee, 1982; Whaley, 1981a, 1981b; Gourley,

1984; Rumelhart, 1985; Bobrow, Black, & Turner, 1985;

Ruddell & Speaker, 1985). Research also demonstrates that

different types of text have underlying "grammars" or

linguistic structures that depend on the goal of the text--

narrative structures have been identified for simple fiction

stories (Rumelhart, 1975; Mandler & Johnson, 1977; Stein &

Glenn, 1979), and a variety of structures have been found

for various types of nonfiction text (chronological,

taxonomical, persuasive, directive, expository, etc.; see

Gillet & Temple, 1986, pp. 247-254). Van Dijk (1983, 1988a)

has identified an underlying structure for hard news


Story grammars are usually the first text structures

acquired by readers, through early exposure to stories in

childhood (Applebee, 1978; Hoover, 1981). Nonfiction


structures are learned later, generally from the third grade

on (Gillet & Temple, 1986, p. 49). Richards (1978) has

observed that many students who are reading fluently in the

beginning grades start to experience reading problems in the

fourth and fifth grades, when a greater amount of nonfiction

material is introduced into the curriculum. She speculates

that this phenomenon arises from children's lack of

familiarity with text structures other that of the fictional


Green (1979) claims that the unfamiliar and often

illogical organization of the typical hard news story can

actually impede comprehension of the information contained

in it. However, she provides no experimental data to

support this hypothesis. How, in fact, does the traditional

"inverted pyramid" structure of a news story influence

comprehension and recall of news? And how might such

effects be related to the receiver's prior knowledge of text

structures? These questions are considered herein.

Goal of This Research

This study investigates some factors influencing the

information gained from news stories, focusing principally

upon the ways the structural organization of hard news

stories affects their comprehension and recall.

Numerous studies have explored the comprehension and

recall of news stories. Most of the research to date has


measured audience members' memory for the factual content of

news stories (Edwardson, Kent, & McConnell, 1985; Katz,

Adoni, & Parness, 1977; Gunter, 1981); little attention,

however, has been paid to the cognitive processes involved

in the assimilation, storage, and later retrieval of news

content, or to the relationship between these processes and

the news message itself. Recent research indicates that

comprehension and recall are interactive operations in which

a reader brings his/her prior knowledge and beliefs into

play while encoding new information and later activates that

knowledge for retrieval of the information (e.g. Bobrow &

Norman, 1975; Schank & Abelson, 1977; Pearson, Hansen, &

Gordon, 1979; Graesser & Nakamura, 1982; Freebody &

Anderson, 1983; Stahl & Jacobson, 1986). The theoretical

position on which these findings are predicated is known as

schema theory.1

The Concept of Schema

The psychological concept of the schema emerged

initially as a reaction to the traditional associationist

models of memory and learning (e.g. Ebbinghaus, 1964), in

which recall occurred simply as a response to a stimulus.

The associationist model gradually gave way to the trace

theory of mental representation, which evolved into schema

'In this document, the word "schema" will be pluralized as
schemass" rather than as schemataa," as per the style used by
Mandler (1984, p.2, note 1)

theory as we know it today (for a more complete account of

the history of schema theory, see Hastie, 1981). The notion

of a schema was first used in studies of memory and

remembering and was later applied in the study of reasoning,

learning, language processing, problem solving, reading, and

countless other cognitive and psychosocial processes.

Definition of Schema

In general terms, a schema may be defined as a dynamic,

generic mental framework for the hierarchical representation

of knowledge. Anderson (1977) asserts, "A schema represents

generic knowledge; that is, it represents what is believed

to be generally true of a class of things, events, or

situations" (p. 2). Each schema contains slots for its

various components: for example, a rudimentary schema for a

human face would contain slots for eyes, a nose, a mouth,

and ears. Each new face encountered by an individual with

this schema would present new information that could easily

be fitted into these slots. Schemas are generally created,

or instantiated, through experience; once in place, they are

key to innumerable cognitive processes.

Graesser and Nakamura (1982), in an extensive

exposition on the role of schemas in comprehension and

memory, define schemas as "generic knowledge structures that

guide the comprehender's interpretations, inferences,

expectations, and attention. A schema is generic in that it

is a summary of the components, attributes, and

relationships that typically occur in specific exemplars"

(pp. 60-61).

Fiske and Taylor (1984) refer to a schema as "a

cognitive structure that represents organized knowledge

about a given concept or type of stimulus" (p. 140). In

their view, a stored schema represents preexisting knowledge

about a given topic and also guides the assimilation,

interpretation, and subsequent recall of incoming

information. They note that "cognitive research on the role

of generic prior knowledge has demonstrated the importance

of schemata in basic processes of understanding and memory"

(p. 145).

Functioning of Schemas

An essential aspect of the concept of schema concerns

the notion that schemas are active rather than immutable or

static; schemas perform a variety of different operations

and are constantly modified, altered, and elaborated as

cognitive processes occur. Piaget (1952) writes of the

processes of assimilation and accommodation of schemas that

transpire whenever schemas are used in interpreting

information. New information is assimilated via an existing

schema; an individual with a more elaborate schema for the

incoming information will be in a better position to absorb

more details of the incoming data. Simultaneously, the


schema becomes enriched by accommodating or assimilating the

new information. These phenomena occur constantly as

information is continually processed, encoded and modified.

Rumelhart (1980) posits that schematic learning involves

three processes: the learning of facts, or accretion, in

which information is encoded into schemas; elaboration and

refinement of schemas through continued experience, or

tuning; and the creation of new schemas, known as

restructuring. Rumelhart's model illustrates the flexible

nature of schemas.

Another key characteristic of schemas is their

hierarchical organization. Rumelhart and Ortony (1977) note


each schema is characterized in terms of lower level
constituents, or subschemata. Presumably, the
dependence that schemata have on lower level schemata
must ultimately stop, that is to say, some schemata
must be atomic in the sense that they are not
characterized by reference to any other constituent
schemata. Thus, our entire knowledge system
would appear to ultimately rest on a set of atomic
schemata. (p. 106)

Rumelhart (1980) describes schemas as "the fundamental

elements upon which all information processing depends" (p.

33). Schemas can be used to interpret information, to

retrieve information from memory, to organize actions, to

set goals, to allocate resources, and to organize all

information processing, but above all they are useful in the

process of constructing meaning, or comprehending.

Rumelhart defines a schema as "a data structure for

representing the generic concepts stored in memory" (p. 34).

Rumelhart and Ortony (1977) have identified four

essential properties of schemas, to wit (a) schemas possess

variables, (b) schemas can embed in one another, (c) schemas

represent generic concepts of varying levels of abstraction,

and (d) schemas are representations of knowledge, not

definitions. To these four characteristics, Rumelhart

(1980) added two more--(e) schemas are active processes, and

(f) schemas are recognition devices whose processing is

aimed at the evaluation of their goodness of fit to the data

being processed.

Hastie (1981) recognizes three distinct types of

schemas. The simplest are the central tendency schemas,

also termed prototype schemas. Such a schema refers to "a

member of a stimulus set that is located at the statistical

center of the distribution of items in the set" (Hastie,

1981, p. 40), or to "the member of a category with the most

attributes in common with other members of the category and

the fewest attributes in common with members of other

contrasting categories" (Hastie, 1981, p. 40). In other

words, such a schema would represent the archetype of a

given concept.

The second type of cognitive schema Hastie identifies

is termed a template schema. These schemas classify, store,

and coordinate incoming information. They are more active

in nature than prototype schemas in that they can add

generic information to the schematic structure when

anticipated information is not supplied during assimilation,

they can modify schema boundaries depending on the nature of

the incoming information, and they can perform tests on new

information to determine its proper classification.

The most complex type of schema is labeled a procedural

schema. Such a schema directs the exploration and

information-seeking that make new information more readily

available to the individual. It specifies the criteria by

which new information is encoded into particular schemas or

subschemas. It is elaborated and enriched when new

information is assimilated. This type of schema "is a

pattern of action as well as a pattern for action" (Neisser,

1976, p. 54).

Schema-Theoretic Information Processing and Its Application
to the Mass Communication Model

In the schematic view of cognition, incoming

information is encoded and stored via an appropriate schema

or pre-existing mental knowledge structure. Graesser and

Nakamura (1982) recognize two stages in the functioning of

schemas during learning: schema identification and schema

application. During schema identification, the learner

selects a schema which matches some aspects of the input

data. Here, the incoming information "matches the

components, attributes, and relationships of a particular

schema better than alternative schemas" (Graesser &

Nakamura, 1982, p. 62). Once an appropriate schema has been

identified, schema application, the second stage, ensues.

In this stage of information processing, the schema directs

the perception and interpretation of new information and

provides the prior information necessary to comprehend the

new information. The schema also determines the amount of

attention the learner gives to the elements in the incoming

data and aids the learner in formulating expectations about

relevant events or information that may follow. As Graesser

and Nakamura point out, "[S]chemas are very powerful and

intelligent knowledge structures" (p. 63). During recall,

information is retrieved via the schema through which it was


Variations on this model of information processing have

been developed (cf. Norman & Bobrow, 1976; Rumelhart, 1980),

but the fundamental mechanism remains essentially the same

in all of them. Schemas play a crucial and multifold role

in the processes of comprehension and memory. Schemas

generate a wealth of prior knowledge essential for

understanding, they organize and format incoming

information, and they provide an efficient mechanism for the

retrieval of stored information.

These operations are all necessary in the comprehension

and recall of news and other messages generated via the mass

media. In Lasswell's model of the communication process,

the transmission of information via mass media is expressed

in the question, "Who says what to whom through what

channels of communication?" (Smith, Lasswell, & Casey, 1946,

p. 121). Another widely accepted model of communication,

the Shannon-Weaver model, also incorporates the sender-

message-medium-receiver chain (Shannon, 1949). The

processing of media-generated messages once they have

reached their destination--i.e., the individual receiver

(newspaper or magazine reader, broadcast listener or

viewer)--is crucial to the success of any mass

communication. Woodall, Davis, and Sahin (1983) point out


the process of understanding the news is a cumulative
process both for the individual and for society. The
ability or inability to understand and remember the
news presented to viewers on any given day will leave
viewers more prepared or less prepared to understand
the news tomorrow. As a society, we make decisions
about collective actions based on our understanding of
the world around us which we derive in part from news
stories. If there is widespread and increased
misunderstanding of certain news stories, we may all
make poorer decisions. (p. 194)

The receiver in the mass communication model has been

studied extensively from a variety of perspectives: A

substantial body of research addresses the individual's use

of the mass media and the gratifications derived therefrom

(e.g. Blumler, 1979; Windahl, 1981) and the effects of mass

media messages on the individual's behavior (e.g. Becker &

Whitney, 1980; McLeod & Reeves, 1981; Weaver, 1982; Hawkins

& Pingree, 1983; Bryant & Zillman, 1986). Scant attention

has been paid, however, to the cognitive processing of the

mass media news message and the ways in which it is mentally

stored and later retrieved (or not) by the receiver.

Research of this type is still relatively in its infancy in

the annals of mass communication scholarship.

Much of the existing research on the processing of

mass media news messages focuses on memory for news,

although a few studies do address the issue of comprehension

of news stories. Very little empirical evidence is

available to support existing hypotheses on these topics,

and that which exists focuses primarily on the

quantification of news recall (Booth, 1970; Neuman, 1976;

Gunter, 1980, 1981; Findahl & Hoijer, 1975, 1981, 1985;

Edwardson, Grooms & Pringle, 1976; Edwardson, Grooms, &

Proudlove, 1981; Edwardson, Kent, & McConnell, 1985).

Studies of recall and comprehension of news grounded in

theory of cognition are virtually nonexistent.

The Swedish research team of Olle Findahl and Birgitta

Hoijer has long been interested in the study of recall and

comprehension of news messages. Several of their studies

have shown that prior knowledge is vital to comprehension

and recall of news (cf. Findahl & Hoijer, 1981; Findahl &

Hoijer, 1985). In other words, the existence of a schema

for a news topic or for some other aspect of a news story


will improve comprehension and recall of the story. As they

point out:

Schema theory stresses the organization of earlier
knowledge in memory in general or prototypical
schemata, representing standard situations, events, or
structures. Two different kinds of schemata have been
proposed: one deals with knowledge about recurrent
events and situations ; the other deals with
knowledge about the typical structure of stories .
In news comprehension both kinds of schemata (about
recurrent events and about the structure of news items)
are probably activated. (Findahl & Hoijer, 1985, p.

In a departure from the strictly numbers-oriented

tradition of this line of research in mass communication,

Woodall, Davis, and Sahin (1983) proposed a theoretical

framework for memory and understanding of news based on

principles of episodic memory and on the trace theory of

memory and understanding. Another pioneer in this domain,

Doris Graber (1988), conducted in-depth interviews with 21

subjects in Evanston, Illinois, to study their schema-based

strategies for processing information gained mainly from the

news media. These theoretical perspectives are strongly

tied to the schematic model of information processing, the

origins and ramificatications of which are discussed in the

next chapter.


Schema-based models of information processing have led

to increased understanding of the recall and comprehension

of written text. Central to the development of such models

is the evolution of the construct of the schema as a

paradigm for cognition.

In this chapter, the development of the schema theory of

cognition is traced from its early applications in studies

of memory processes to its current role in research on

reading comprehension and recall of text information.

The Foundations of Schema Theory

Two seminal pieces of research form the keystones of

schema theory as it is applied in the study of reading

cognition today. The first is the work done by Frederic

Bartlett of Cambridge University, England, as set forth in

his book Remembering (1932). The second is a paper in the

realm of artificial intelligence written by Marvin Minsky


Bartlett's Experiments On Remembering

Near the beginning of this century, Bartlett conducted

a series of experiments with a view to studying

systematically the specific social factors that influence

visual perception, imaging, and memory, arguing that

perception and imagery are both significantly affected by

recalled prior information, and that they in turn influence


To study perception, Bartlett exposed experimental

subjects to shapes and patterns of varying complexity, and

to pictures representing easily-identifiable objects and

situations, for intervals ranging from 1/15 to 1/4 of a

second. Subjects were asked to reproduce the illustration

they had seen or simply to describe it. Bartlett observed

that the subjects' perceptions of the stimuli were

determined by their attitudes, interests, and dispositions.

He observed that "in perceiving, the data presented have to

be connected with something else before they can be

assimilated" (Bartlett, 1932, p. 46).

Imaging was studied by means of a similar experiment in

which Bartlett showed subjects various inkblots and

encouraged them to describe them. The results of this

experiment indicated that, again, subjects drew on prior

information and attitudes when describing the inkblots, and

that their reactions to the inkblots were part of a search

for meaning in the stimuli. Bartlett noted that

the subject, confronted by his task and having to
use the same instruments of subjective tendencies,
bias, interests, and temperamental factors, casts about
for analogies with which to subdue the intractability
of the perceptual data. (1932, p. 45)

Bartlett's work on remembering is extensive,

incorporating a series of experiments in which five

different methodologies were used: the method of

description, the method of repeated reproduction, the method

of picture writing, and two methods of serial reproduction.

In the first experiment, Bartlett showed subjects a

series of five picture postcards, each of which bore the

face of a different naval or military officer. Subjects

were briefly exposed to each postcard separately and thirty

minutes after the exposure were instructed to describe the

various cards in the order in which they thought them to

have been presented. Recall was tested several times

thereafter, following intervals of one week and longer, and

questions on the stimuli were posed.

In the second experiment, Bartlett asked subjects to

read the short story "The War of the Ghosts" by Franz Boas.

Each subject read the story twice, silently, and after a

lapse of 20 hours orally retold the tale. Bartlett studied

the accuracy of the retellings and the various types of

deviations from the original story stimulus.

The third experiment utilized the method of picture

writing, where subjects memorized various symbols that

represented words and were then asked to use the symbols in


writing down a dictated account of a short story containing

words for which symbols had been taught. Bartlett studied

methods of learning and conventions of representation,

concluding that a few broad patterns of recall and

transmission of symbols exist that are largely social in

origin and character. He also found a negative correlation

between determination to remember and actual forgetting.

Bartlett's final experiments were based on two methods

of serial reproduction and were designed to study the

effects of the transformations brought about by many

individuals in the transmission of a stimulus. The

experimental method used in the first of these experiments

was similar to the method of repeated reproduction. Again,

a subject was asked to read and retell a story, but the

retelling was then reproduced by a second subject, whose

retelling was in turn reproduced by a third subject, and so

on. Bartlett's focus was the main trends of change in

series of reproductions gleaned from a number of different

subjects. A variety of stimuli were used, including folk

tales and nonfictional passages. The experiment was

conducted using a variety of different social and ethnic

groups of subjects.

The second experiment using the method of serial

reproduction was based on pictorial material--decorative

patterns, illustrations, art forms, and other items. These

picture representations were submitted to a course of


repeated and serial reproductions in the same manner as were

the stories in the first experiment.

In both experiments, Bartlett found that

transformations occurred which resulted in an overall loss

of the individualizing features of each stimulus, so that in

the final reproduction the stimulus had been changed into

some more "conventional" representation of the original.

The theory of remembering that evolved from Bartlett's

experiments laid the foundation for much of the schema

theory on which this research is based. Bartlett (1932)

noted that

when any specific event occurs some trace, or some
group of traces, is made and stored up in the organism
or in the mind. Later, an immediate stimulus re-
excites the trace, or group of traces, and, provided a
further assumption is made to the effect that the trace
somehow carries with it a temporal sign, the re-
excitement appears to be equivalent to recall. (p. 197)

Bartlett postulated that these memory traces are linked

in organized frameworks that support the process of learning

and remembering. Each of these frameworks, or "postural

models", he termed a "schema." In Bartlett's words,

"'Schema' refers to an active organisation of past

reactions, or of past experiences, which must always be

supposed to be operating in any well-adapted organic

response" (p. 201). In his view, schemas both influence the

acquisition of new information and are continually being

elaborated by the addition of new information.

Minsky's Frameworks for the Representation of Knowledge

By synthesizing a number of classical and modern

paradigms from psychology, artificial intelligence, and

linguistics, Marvin Minsky of the Massachusetts Institute of

Technology developed a theory of memory based on the

existence of "frames" that are similar in function to

Bartlett's schemas. As Minsky defines it, "A frame is a

data-structure for representing.a stereotyped situation"

(Minsky, 1974, p. 211); it is also "a network of nodes and

relations" (p. 211). In his model, a frame has top levels

which represent unvarying knowledge about the situations in

question, and lower levels containing terminals or slots

that are filled by specific instances or data. Each

terminal or slot can specify conditions for the incoming

data. Collections of frames and sub-frames are linked into

complex frame systems, and frames within each frame system

share terminals, which makes it possible to coordinate

information gathered from different viewpoints. In the

learning or data-acquisition process, information is matched

with a frame and then assimilated. If no frame exactly

matches the new information, the best available one is

modified and stored, in much the same way that Bartlett's

schemas are elaborated when new information is acquired.

Memory, according to Minsky's hypothesis, is a function of

frame retrieval--a process corresponding to schema

activation during remembering.

The work of Bartlett and Minsky has had significant

impact on the study of cognition and forms the underpinning

of much of the recent research in reading cognition.

Schema-Theoretic Approaches to Reading Comprehension

As Durkin (1984) has pointed out, reading instruction

has long taken into account a reader's prior knowledge as a

crucial factor in the comprehension of printed text. Schema

theory is predicated upon the notion that this prior

knowledge is organized into dynamic knowledge structures in

the brain that are activated during the reading process, as

well as during other types of cognitive processing. As

Adams and Collins (1977) have observed:

The goal of schema theory is to specify the interface
between the reader and the text--to specify how the
reader's knowledge interacts with and shapes the
information on the page and to specify how that
knowledge must be organized to support the interaction.
(p. 5)

Anderson (1985) has pointed out that from a schema-

based perspective, reading involves the analysis of text at

many different levels simultaneously. In his definition,

processes that stem from the actual print on the page are

termed "bottom-up" or "data-driven," while processes that

originate with the reader's prior information about the text

content are "top-down" or "hypothesis-driven" (p. 376).

Adams and Collins (1977) provide a model of reading

comprehension that describes the act of comprehension as one

in which "top-down" and "bottom-up" schematic processing

occur simultaneously. Using Aesop's fable "Stone Soup," the

authors analyze the reading process on four levels: letter

and word, syntactic, semantic, and interpretative. Letter

and word recognition are fundamentally bottom-up processes,

although words are recognized by letter and holistically--

letter schemas activate word schemas, and, as a word schema

becomes active, "it proportionally and reciprocally

facilitates the letter schemata corresponding to its

component letters" (p. 21). Syntactic, semantic, and

interpretative processing are "top-down" operations, wherein

prior knowledge or existing schemas about fables, word

meanings, and problem-solving are brought into play so that

the new information in the stimulus story can be efficiently

encoded and understood. Schema theory has been applied to

explain the comprehension of written text in numerous

studies in the last decade (e.g. Anderson, 1977; Pace, 1978;

Adams & Bruce, 1980; Pearson, Hansen, & Gordon, 1979;

Anderson, 1985; Pearson & Camperell, 1985; Stahl & Jacobson,

1986). This recent research stems from the work of


Schank and Abelson (1977) defined a schema as "the

large repertoire of knowledge structures brought to the

reading task by the reader which enable him to understand

information which is not directly contained in the text of a

given action" (pp. 9-10). In other words, a schema-


theoretic approach to reading would suggest that reading is

a process requiring a significant amount of prior knowledge

on the part of the reader. Anderson (1977) proposes that

language comprehension involves rapidly sorting information

into slots in a schema. Slots are "placeholders" in schemas

into which elements of incoming information can be fitted.

To comprehend written material, a reader is required to fit

textual information into slots in his/her schema for that


The notion that a schema must have been instantiated

for reading comprehension to occur may be applied at several

levels of the reading process. Some schema for the content

of the text passage must exist for comprehension to take

place (e.g. Stahl & Jacobson, 1986; Pearson, Hansen, &

Gordon, 1979). Adams and Collins (1977) suggest that

readers use schemas both to decode written symbols and to

extract meaning from written words. And a growing body of

literature indicates that some knowledge of the discourse

structure or "text grammar" of a passage, i.e. a "text

schema," is essential for reading comprehension.

Text Structures and the Reading Process

In his book Remembering, Bartlett proposed that recall

of written material depends on a reader's schema for the

structure of a written passage (Bartlett, 1932). Such

structures, also known as text grammars, have been theorized

to be fundamental in the organization of all written text

(Rumelhart, 1975; Rumelhart, 1977; Mandler & Johnson, 1977;

Meyer, 1977a, 1977b; Stein & Glenn, 1979; Marshall & Glock,

1979; Taylor, 1980; Taylor & Samuels, 1983). Underlying

structures have been isolated for narrative texts (story

grammars), as well as for descriptive, expository, and

argumentative texts. A rudimentary story grammar for a

fable, for example, would comprise a story and a moral.

Research indicates that a knowledge of text grammars is

essential to the reading process; thus, a reader would have

to have a schema for the "fable grammar" to recognize and

assimilate a fable in its entirety.

Handler and Johnson (1977) refer to a story schema as

"an idealized representation of the parts of a typical story

and the relationships among those parts" (p. 112).

The basic organizational pattern of the Handler &

Johnson (1977) story grammar comprises the following


1. The setting of the story: The protagonist, and

possibly other characters, are introduced; the

temporal and physical locations of the story are

indicated; any other information is provided which

may assist the reader to follow subsequent events.

2. An event: An action or idea that precipitates

further story developments.

3. An internal reaction, which may be either simple

or complex: When the character's internal

reaction leads to a single action, it is termed a

simple reaction; when the internal reaction

results in the setting of a goal, which is

followed by an attempt to reach that goal, the

internal reaction is a complex reaction.

4. An outcome: This is the result of the

protagonist's attempt to arrive at the goal set in

the internal reaction. If the outcome is not

successful, the protagonist may try again, with a

different outcome. The pattern may thus be

repeated within a story.

5. An ending: A state of affairs in which the story

is wrapped up "with a dramatic flourish" (p.


Numerous studies have demonstrated that a knowledge of

story grammars, i.e. the presence of story schemas in

readers, facilitates reading comprehension and recall of

stories. Mandler (1978) investigated the effects of

scrambling two-episode stories generated on the basis of the

Mandler and Johnson (1977) story grammar when testing for

recall using second- fourth- and sixth-grade children as

subjects, as well as a group of adult subjects. She found

that the structure of the stimulus stories had a significant

impact on the quantity, quality, and temporal sequencing of


recall for children and adults; when the basic story grammar

was violated, subjects showed a tendency to recall stories

according to the correct grammatical structure rather than

in the form in which they were presented. It also appeared

that children relied more on story schemas than did adults.

In a similar study using only adult subjects (n=64),

Stein and Nezworski (1978) examined recall of stories of

four types: (a) Exact Order (stories that followed the Stein

& Glenn story grammar), (b) Slightly Disordered (two

elements of the grammatical story transposed), (c) Randomly

Ordered (story elements presented in random order), and (d)

Unrelated Statements (twelve statements from which no causal

order could be inferred). Subjects in the Exact Order

treatment demonstrated nearly perfect recall of the stimulus

material; recall dropped in each experimental condition as

more story conventions were violated. Additionally,

subjects who were instructed to recall the stories "in the

form of a good, coherent story" rather than in the

presentation format tended to recall the text in an order

which corresponded almost exactly to story grammar order.

The data indicated that the underlying structure of a story

had a significant influence on its retention and recall.

In a more recent experiment, twenty fourth-grade

children considered average or below-average readers, and

who lacked a sense of story structure, were randomly

assigned to one of two treatment groups: instruction in

narrative structure or instruction in dictionary usage and

vocabulary (Fitzgerald & Spiegel, 1983). It was found that

the children provided with instruction in story structure

showed significant gains in reading comprehension compared

with the control group. Spiegel and Fitzgerald (1986)

provide details of the instruction given to the treatment


Fitzgerald (1984) found that in fourth- and sixth-grade

children, there was a significant positive correlation

between reading achievement and ability to anticipate

narrative structure. The correlation was consistent across

grade level.

Van Dijk and Kintsch (1985) hypothesized that (a) in

order to understand narratives, subjects must have a general

knowledge of a conventional narrative structure, (b) a

subject's construction of a story structure comprises a

necessary component of story comprehension, and (c) the

information stored in memory corresponds to the structure of

a text. All three hypotheses were experimentally supported.

Similarly, it has been shown that knowledge of

nonfictional text structures helps readers better recall and

comprehend nonfictional text (Meyer, 1975; Meyer, 1977a,

1977b; Meyer, Brandt, & Bluth, 1980; Meyer & Freedle, 1979;

Mulcahy & Samuels, 1987). The expository text structure is

written with the purpose of describing or explaining

something. Meyer (1977b) notes that recall of expository


prose is primarily a function of recognizing superordinate

(or top-level) and subordinate (lower-level) structural

propositions in the text:

The top level information in the content structure is
similar to what educators have identified as the main
ideas of a passage and the interrelationships among
those ideas. The top levels of the structure appear to
carry the central message of a passage. In contrast,
the low-level information in the content structure is
not part of the central message of a passage although
it often supports various aspects of the message;
instead, the low levels of the structure appear to
contain information peripheral to the central message
of a passage. (Meyer, 1977b, pp. 330-331)

Meyer has identified several types of organizational

structures and relationships characteristics of nonfictional

texts. These include taxonomical, chronological, cause and

effect, directive, comparison and contrast, and enumerative

or attributional. These organizational schemas are acquired

gradually through long-term exposure to different discourse


Bartlett (1978) found the enumerative or attributional

structure to be the most common type of textbook

organization. An outstanding feature of the attributional

expository text structure is the theme paragraph, where each

paragraph begins with a topic sentence synopsizing the

paragraph's content, followed by elaborative sentences.

Taylor (1980) found that children who are "good"

readers (scoring higher on standardized reading tests than

other children in the same grade) use prose text structure

to organize recall. Taylor and Samuels (1983) found that

superior recall for expository text could be attributed to

the use of text structure as a retrieval aid. More

recently, in an examination of the complementary roles of

text schemas and content schemas in reading, Ohlhausen and

Roller (1988) determined that both types of schema are used

in reading and comprehending expository text.

Schemas for the Structure of News Stories

Research which integrates cognitive science, schematic

views of reading, and the processing of news information is

still in its infancy. News stories would appear to combine

the characteristics of narrative and expository text and

thus possess a structure uniquely their own. That news

stories follow a set structural pattern is a notion which

has been intuitively acknowledged as true for many years.

In the textbook Writing for Mass Communication, Hutchison

(1986) points out that all hard news stories2 should have a

formal structure, beginning with a lead:

In a good lead, the important things come first.
They provide the umbrella under which all details of
the story will fit comfortably. The details
usually flow from the lead in order of descending
importance into the succeeding paragraphs. A simple

2The inverted pyramid story structure is characteristic of
that of hard news stories, i.e. stories that are factual accounts
of events, usually with a time element. Soft news stories (news
stories with a human interest focus, written in a lighter vein) or
feature stories often do not follow the inverted pyramid structure.
The present research is therefore confined to hard news stories.
For a more detailed classification of news story types, see
McCombs, Shaw, and Grey (1979), Handbook of Reporting Methods, p.

news story about a minor traffic accident or a minor
house fire will look like an inverted pyramid. (p. 125)

One form of the inverted pyramid structure is that

described by Fedler (1989), shown in Figure 2-1 and

discussed in detail later in this section.

The structuring of hard news stories is in fact so

entrenched in the newswriting process that Tuchman (1978)

claims that most news stories consist of prestructured

patterns of words into which reporters insert "factoids."

The traditional "inverted pyramid" structure of news

stories corresponds to the concept of a schematic structure

or text grammar. Van Dijk (1983, 1988a, 1988b) observed

that a news story can be viewed in terms of schematically

structured discourse. Using cognitive models, Van Dijk

examined media discourse and its representation in memory,

and his analysis of newspaper stories led him to postulate

an underlying macrostructure for news stories

The overall organization of news discourse reflects the
importance of macrostructures. These will typically be
expressed by titles or headlines, by initial or final
summaries, or by leads. .The lead, often printed in
bold type will express, in a first few sentences
(which are, by definition, "thematical sentences"), the
full macrostructure of the news discourse. Following
sentences will then progressively specify further
details of the events, with the less important ones at
the end (with the practical consequences that these
can, if necessary, be cut by the editor). Unlike
argumentatively structured discourse, such as the
scholarly paper, where the important conclusion comes
at the end, news in the daily press is organized by the
principle of relevance or importance, along a dimension
of decreasing prominence with respect to the macro-
structure. (Van Dijk, 1983, p. 35)

Summary lead

Most important facts

Next most important facts




Related facts


Figure 2-1. Fedler's inverted pyramid.

Van Dijk's grammar for news discourse is outlined


1. Summary/introduction
1.1 Headlines
1.2 Lead
2. Episode(s)
2.1 Events
2.1.1 Previous information
2.1.2 Antecedents
2.1.3 Actual events
2.1.4 Explanation Context Background
2.2 Consequences/reactions
2.2.1 Events
2.2.2 Speeches
3.1 Expectations
3.2 Evaluation
A diagrammatic representation of this structure is

shown in Figure 2-1.

Van Dijk has noted that while the elements of this news

macrostructure are present in almost all hard news stories,

their sequence may vary depending on such factors as the

semantic content of the story, the news values of the

writer, the complexity of the story, and so on.

In his study, van Dijk analyzed news stories from a

number of international newspapers; however, the

macrostructure he proposed does not correspond very well

with the average hard news story found in American

newspapers, especially with regard to the last category in

the structure--the "Comments" category. In general, hard

news stories produced in the United States do not include

evaluations or predictions of possible consequences of the

actual events detailed in the stories.

In a popular journalism textbook, Fedler (1989)

describes the structure of a hard news story thus:

The lead in an inverted pyramid story summarizes the
topic, and each of the following paragraphs presents
some additional information about it: names,
descriptions, quotations, conflicting viewpoints,
explanations, background data and so forth. Most
paragraphs are self-contained units that require no
further explanation, and only the summary of the entire
story appears in the lead. News stories end with their
least important details--rarely with any type of
conclusion. (pp. 135-136)

Fedler's version of the inverted pyramid story may be

graphically represented as shown earlier in this chapter in

Figure 2-1.

Newsom and Wollert (1988, p. 120) determined that most

news stories have the following elements:

1. The lead (the main point)

2. Secondary points in a tie-in transition

3. Elaboration on the main point

4. Support for the lead

5. Background

6. Development of the main idea

7. Details



















Figure 2-2. Diagrammatic representation of van Diik's
news macrostructure

They offer two diagrammatic representations of the

inverted pyramid: the "traditional" inverted pyramid (Figure

2-3) and the "modified" pyramid (Figure 2-4).

For the purposes of this study, Newsom and Wollert's

modified inverted pyramid structure will be adopted as the

typical structure for a breaking hard news story in an

American newspaper. Unlike the van Dijk structure, the

Newsom and Wollert modified pyramid does not include any

subjective elements such as comments on or analyses of the

events contained in the story; the absence of these elements

is far more typical of American news stories. Second, the

Newsom and Wollert modified pyramid includes quotes and the

possibility of a secondary theme in the story, which more

complex news stories often contain; the traditional pyramid

does not accommodate these elements. In this respect, the

Newsom and Wollert pyramid is a more useful descriptor than

the Fedler pyramid, which makes a provision for quotes but

not for a secondary theme within a story.

Green (1979) suggested that a "typical" news story

organization can, in fact, be detected, but that the

organization hampers comprehension instead of facilitating

it. The typical news story, in Green's view, is

"disorganized and undirected, unconnected and jumbled up,

with the result that it is difficult to follow" (p. 5).

Who What When Where Why How 16 to 25 words

One sentence connecting one element of the lead to the body

Development of the most important WWWWWH elements of the lead

Second most important element of WWWWWH

Further development of most important element

Other elements.

The least important facts of the story-nothing new introduced

Lead 5W&H


Most Important

2nd Most Important



Figure 2-3.




Newsom & Wollert's traditional inverted



1st Graph:

2nd Graph:

3rd Graph:

4th Graph:

Major theme, could be significance of event, rather than fact
May be two sentences
May not include 5W&H

The left-overs of the 5W&H not mentioned in the lead

Explication of the lead incident, quote, meaning, or
background of event-how something came to be

Additional information about most important fact of lead
Something to give credibility or significance to lead

Secondary theme or supporting documentation for the lead

Any other details, in order of significance to lead


Documentation or Explication=Background or History

Elaboration of Lead

Secondary Theme


Supporting Facts, Quotes

Least Significant


Figure 2-4. Newsom and Wolert's modified inverted pyramid

She recommends a complete overhaul of the guidelines

reporters follow to shape their stories in favor of a

structure that would ease comprehension.

Although the above analyses have revealed evidence of

an underlying news story structure, Thorndyke (1977) found

that altering the structure of a news story did not

significantly influence readers' recall or comprehension of

the content, the implication being that the structuring of a

news story does not consistently affect the processing of

information contained therein, contrary to the similar

studies of this nature that have been conducted using

narrative and expository text. Thorndyke, however, did not

identify a standardized news or narrative structure in his

experiment, and thus the scrambling of stories in his study

was not systematic, which may render his results less than


House (1984) posited that the linguistic complexity of

a news story influenced readers' recall and comprehension

far more than did the structure of the story. He found that

in the case of television news stories, linguistic

complexity was indeed a better predictor of recall and

comprehension than was story structure. However, the

inverted pyramid structuring of a story in fact lends itself

to a lower level of linguistic complexity than would be

found in a narrative or expository story due to the absence

of cohesive devices, connectives, and transitions in

inverted pyramid stories.

Nolan (1989) found that when a news story was rewritten

to follow a chronological order, the gist of the story was

better recalled by subjects who read it in the inverted

pyramid form.

More research into this aspect of the nature of news

stories and the processing of news content is needed at this

juncture. Conventional perspectives and some empirical

evidence support the concept of an underlying structural

grammar for news stories. But to what extent does this

structuring influence readers' interaction with printed

news? Are stories which conform to the grammar more easily

comprehended and recalled than stories which do not? Would

familiarity with the structure of news stories increase

readers' comprehension and recall of news? Concomitantly,

if news stories were written to conform to structures with

which readers were already very familiar--e.g. a narrative

or expository structure--would recall and comprehension


The implications of research of this nature would be

multifold. Identification of a text structure which

significantly influenced comprehension and recall of news

might revolutionize the way newswriting is taught in

journalism curricula. Several studies have been conducted

to assess the influence of newspaper reading on reading

skills (cf. Stetson, 1977; Heitzman, 1979; Cheyney, 1984),

but the schematic structures of news stories and their use

by readers to facilitate comprehension and recall have not

been incorporated into educational programs such as

Newspapers in Education. Thus, instruction on news grammars

in reading classrooms and literacy education programs--

particularly those in which newspapers are used as teaching

tools--could enhance readers' use of newspapers in the

context of their daily lives as a direct result of increased

comprehension and recall of news stories' content.

Limitations of the Schema Concept

Although cognitive models, particularly those utilizing

the concept of schema, provide a comprehensive and useful

framework within which to study information processing

(Lachman, Lachman, & Butterfield, 1979), various critics of

schema theory have pointed out its limitations as a research

paradigm. The research conducted in this dissertation could

not be fully and critically evaluated without some

discussion of these limitations and their possible effects

on the results of this study.

Schema theory is perhaps best characterized as a form

of causal process theory, defined as "a set of descriptions

of causal processes" (Reynolds, 1971, p. 11). This form of

theory should incorporate (a) a set of definitions,

including definitions of theoretical concepts, using nominal

and operational definitions; (b) a set of existence

statements describing the situations in which the causal

processes are expected to occur; and (c) a set of causal

statements that describe the effect of one or more

independent variables on one or more dependent variables

(Reynolds, 1971, p. 97).

These three conditions have been met in the case of

schema theory. (1) The term "schema" has been defined in a

variety of ways, using both primitive and derived terms; some

of these nominal definitions have been presented earlier in

this work (see Chapter 1). (2) Schemas are thought to guide

information processing in virtually any situation requiring

cognitive activity. (3) Schema theory describes the way

cognitive structures cause information to be assimilated,

stored, and later retrieved. These cognitive structures are

posited to affect such dependent variables as thinking,

perception, comprehension, memory, information gain, and

concept formation.

Fiske and Linville (1980) evaluate schema theory

according to attributes of a good theory such as predictive

capability, link to observables, and heuristic value. While

schemas have good predictive power and tremendous heuristic

provocativeness, as is exemplified by the numerous studies

predicated on the schema concept, the link between schemas

and observable entities remains tenuous. As Fiske and

Linville observe, "There is no manipulation check for a

schema" (p. 547). This fact remains the most significant

limitation of schema theory: because no method of measuring

a schema is consensually accepted at present, "schema" is

still considered an abstract and vague concept.

A second criticism hinges on the idea that the schema

concept cannot be falsified because a schema can be used to

explain virtually any experimental result, even results that

contradict one another. In theory testing, as described by

Popper (1963), the theory needs to be weighed in relation to

empirical findings which constitute attempts to falsify the

theory. Reynolds (1971) observes that a theory can best be

refined by testing axioms or statements in the theory which

are most likely to be false; when empirical findings refute

a theoretical statement, the theory is modified and

improved. However, the failure of empirical findings to

refute the causal process statements implicit in schema

theory should not constitute an automatic indictment of the

theory. Rather, embeddeddd in a well-specified theory of

process, the schema construct becomes more clear,

consistent, powerful, and thus more falsifiable" (Fiske &

Linville, 1980, p. 546).

According to Fiske and Linville (1980), some critics

claim that schema theory explains phenomena already

adequately explained by attitude and attribution theory--

i.e., that an individual's schema for something is simply

another word for his/her attitude toward it or a way of

describing attributions made about it. However, as these

authors point out, the schema can be thought of as a

metaconstruct which explains both attitude formation and

attribution, thus providing a framework within which to

study the internal processes guiding attitude change and

behavior. Rather than reiterating old findings, schema

theory actually presents new conceptualizations of

previously-observed phenomena.

In short, it would appear that schema theory provides a

distinctively new orientation to the study of cognitive

processes. New research questions and strategies have been

generated from the study of schematic information

processing, and explanations for previously unexplained

phenomena have been established. On the basis of these

criteria, schema theory can be accepted as a paradigm to be

used to understand and explain human cognition.

Predicted Associations

The theory and research findings summarized in this

chapter point to a number of empirical relationships that

were explored in some depth in this study. The schematic

structure of text has been shown to significantly affect

readers' recall and comprehension of the information

contained therein; similarly, readers' familiarity with text

structures--i.e., the strength of their text schemas--has

been demonstrated to affect their comprehension and recall

of the text information. The extent to which these

relationships hold true in the case of printed news messages

was the focus of the present research.

Nominal Definitions

To clarify the method and hypotheses employed in this

study, explicit definitions of the concepts under

investigation should first be provided.


The concept of schema forms the crux of this study;

while definitions and descriptions of this concept abound,

as illustrated in Chapter I, above, these definitions differ

only slightly; researchers seem generally to share an

agreement regarding the notion that the term "schema" refers

to a malleable cognitive structure representing generic

knowledge. The definition of schema used by Graesser and

Nakamura (1982) seems to express explicitly and concisely

the nature and function of schemas and thus will be the

definition adopted for this study. Thus, a schema will

hereinafter be defined as a dynamic, generic knowledge

structure that guides interpretation, inferences,

expectations, and attention (see Graesser & Nakamura, 1982,

pp. 60-61).

Text schemas and text structures

The specific type of schema known as a text schema

bears particular relevance to this study, along with the

related concept of a text structure or text grammar. The

term "text structure" refers to the underlying

organizational pattern of a given text; the representation

of this pattern in the reader's mind is the text schema.

Mandler and Johnson (1977) define a story schema as "an

idealized internal representation of the parts of a typical

story and the relationships among those parts" (p. 111).

Stein and Glenn (1979) describe a story schema as "the

underlying structure used to comprehend the informational

units in a story and the relations that occur between the

units" (p. 53). This notion of a cognitive model

representing the structure of a story can be applied to

other types of written text also. Therefore, hereinafter a

text schema will be defined as an idealized mental

representation of the informational units in a typical

example of a particular type of text and the relationships

among those units. Narrative, or story, schemas, expository

text schemas, and news story schemas will be studied in this


Stein and Glenn (1978) noted, "Stories can be described

in terms of a hierarchical network of categories and the

logical relations that exist between these categories" (p.



This underlying organizational structure of a text will

be identified in this work as the text structure or the text

grammar (the two terms will be used interchangeably).

Recall and comprehension

Recall and comprehension are the dependent variables of

interest in this study. Both terms encompass a wide variety

of cognitive phenomena. However, the two constructs are

closely linked. Van Dijk (1987) observes that "one result

of understanding a text is a representation of the meaning

of the text in (episodic) memory" (p. 165), the direct

implication being that text comprehension always results in

the storage of information in long-term memory for later

retrieval, i.e. text comprehension always precedes long-term

recall of text information. Voss (1984) corroborates this

notion. He writes

While reading, the individual is assumed to interpret
the text contents in terms of his or her own knowledge,
interests, and attitude. During the interpretive
process the individual develops a representation of the
contents of the text. Learning is thus presumed to
involve the storage of information via the development
of the representation. (p. 197) (emphasis added)

Irwin (1986) describes comprehension as comprising

several processes that proceed simultaneously:

microprocesses, involving "[t]he initial chunking and

selective recall of individual idea units within individual

sentences" (p. 3) as well as integrative processing, in

which the relationships between clauses and sentences are

inferred; and macroprocesses, involving elaborative and

metacognitive processes. The ultimate result of these

comprehension processes is a good representation of the

text's ideas in memory.

Thus, separating comprehension from memory as two

distinct constructs gives rise to some difficulties.

Nevertheless, the inferential and elaborative aspects of the

comprehension process do distinguish it from the phenomenon

of storing literal information drawn from a text.

Recall. As Belli (1986) observes, rival psychological

theories have resulted in very different interpretations of

the memory process. The mechanistic model of memory, for

example, views it as a passive process, whereas schema-based

models regard memory as an active, adaptive operation. The

latter position will be adopted for the purposes of this


Memory is generally measured as recall, a term which

also possesses different meanings in different contexts. In

the mass communication literature, recall is further

classified into aided and unaided recall. Katz, Adoni, and

Parness (1977) refer to unaided recall as "spontaneous

recall" (p. 232). British psychologist Martin Le Voi, on

the other hand, terms it "free recall" (1986, p. 105) and

describes the process as happening in a situation where "the

subject is free to recall any items and create and use

helpful cues in any way he or she wishes" (p. 105).

Generally, unaided or free recall means the unprompted

remembrance of information; aided recall, on the contrary,

refers to the process of remembering information in a

situation where prompts or cues are provided. Le Voi calls

this latter phenomenon cuedd recall" (p. 106). He notes

that models of memory based on the encoding specificity

principle (ESP) make no distinction between cued recall and


To avoid any confusion, hereinafter unaided or free

recall will simply be termed "recall" while aided or cued

recall will be termed "recognition".

In this study, the research question was framed in

terms of long-term benefits to the reader from the

assimilation of information contained in printed news

stories. Thus, recognition is not of as great interest as

recall. In addition, long-term recall is of greater

significance in this investigation than short-term recall.

Nevertheless, both short-term and long-term recall were


Comprehension. Irwin (1986) defines comprehension as

the process of using one's own prior experiences
(reader context) and the writer's cues (text context)
to infer the author's intended meaning. (p. 9)

The assessment of text comprehension has traditionally

been effected via the use of text-based questions (see

Trabasso, van den Broek, & Liu, 1988). Comprehension in this

experiment was assessed using a questionnaire that measured

subjects' literal, inferential, and evaluative processing of

the stimulus texts. Literal comprehension refers to the

retention of facts from a text; inferential comprehension

requires the reader to use his or her prior knowledge in

conjunction with information in the text to construct

meaning from the text; and evaluative comprehension occurs

when the reader is able to formulate some judgment about

information contained in text.

Literal comprehension may be assessed using textually

explicit questions, defined by Pearson and Johnson (1978) as

questions having "obvious answers right there on the page"

(p. 157). Inferential comprehension is measured through the

use of textually implicit questions. "Comprehension is

regarded as textually implicit if there is at least one step

of logical or pragmatic inferring necessary to get from the

question to the response and both question and response are

derived from the text" (Pearson & Johnson, 1978, p. 161).

Evaluative comprehension is assessed by means of scriptally

implicit questions:

Scriptal comprehension .. occurs when a reader gives
an answer that had to come from prior knowledge (it is
not there in the text) to a question that is at least
related to the text (that is, there would be no reason
to ask the question if the text were not there). It is
similar to textually implicit comprehension in that an
inference is involved; however, it is different in that
the data base for the inference is in the reader's
head, not on the page. (Pearson & Johnson, 1978, p.


To explore the connection between the schematic

organization of news stories and the receiver's cognitions

regarding the news message, an experimental study of the

relationships between these variables was undertaken.

Green's (1979) hypothesis that the organization of news

stories hinders comprehension and recall of the message was

empirically tested; in addition, the significance of the

existence of a reader's schema for news story structure was

investigated under experimental conditions.

The most commonly encountered types of text

organization are the narrative or story grammar and the

expository text structure. The story grammar forms the

basic framework of all simple stories (see Handler &

Johnson, 1977; Stein & Glenn, 1978). While many types of

nonfiction text structures have been identified, Bartlett

(1978) found that the expository structure, in which facts

are organized in terms of main points and supporting

details, is used most frequently in nonfiction text. Thus,

in schema theoretic terms, it would appear that these two

organizational patterns would be the most familiar to


The simple story grammar and the expository text

structure were thus selected as the text grammars with which

to compare the inverted pyramid news story structure.

Readers' schemas for these patterns were also measured to

gauge the extent to which strength of text schema affected

information processing from a printed news story.


This study investigated the influence of text grammars,

the underlying organizational structures of text passages,

on readers' recall and comprehension of passage content.

Generally, it was hypothesized that narrative and expository

text grammars would be more conducive to high recall and

comprehension than the news story grammar.

The following hypotheses were tested:

HI: A text passage organized according to a narrative story

grammar will be better remembered in the short term

than a passage organized according to an expository

text grammar, which in turn will be better remembered

in the short term than a passage organized according to

a news story grammar.

H2: A text passage organized according to a narrative story

grammar will be better remembered in the long term than

a passage organized according to an expository text

grammar, which in turn will be better remembered in the


long term than a passage organized according to a news

story grammar.

H3: A text passage organized according to a narrative story

grammar will be better comprehended than a passage

organized according to an expository text grammar,

which in turn will be better comprehended than a

passage organized according to a news story grammar.

The following hypotheses were postulated to find a

theoretical basis for explaining the results of the above


H4: Readers with highly developed schemas for a particular

text structure will have higher recall of that type of

text than readers with weak schemas for that structure.

H5: Readers with highly developed schemas for a particular

text structure will have better comprehension of that

type of text than readers with weak schemas for that


Operational Definitions

The Independent Variables

Short-term and long-term recall of text and text

comprehension were hypothesized to vary with the underlying

structure of a text. Thus, the principal independent

variable in this study was text structure, which was

manipulated to assess its effects on the cognitions under

examination. Readers' text schemas were also hypothesized

to affect their recall and comprehension of text in the

various structural conditions.

Schematic structure of text

The variable manipulated to predict readers' recall of

a news message was the schematic structure of a stimulus

hard news story. This structure was systematically varied

so as to measure fluctuations in recall that might result

from changes in the news story structure.

In general, news writers in the United States follow

Associated Press guidelines for the organization of news

story material. Stimulus news stories were generated by

selecting two front-page hard news AP stories from a

national newspaper according to how well they typified a

hard news story as it is defined by McCombs, Shaw and Grey

(1979) (see Footnote 2 in the preceding chapter).

Stimulus news stories were rewritten twice: once to

follow the Mandler & Johnson (1977) narrative grammar and

again to follow the expository/attribution text structure

outlined by Meyer (1975), both described in an earlier

chapter. The purpose of the rewriting was to provide

stimulus materials to test which, if any, structural pattern

contributes most to increasing levels of comprehension and

recall of the passage's content.

The news stories chosen as experimental stimuli were

rewritten to follow the patterns for narrative and

expository text as closely as possible without significantly

altering the content, length or readability level of the

passages (see Table 3-1). The content of the stories was

such that they could not be rewritten to follow exactly the

story or expository text grammars, but the narrative and

expository versions produced for use in this experiment were

reasonably close facsimiles of the ideal structures.

Text schemas

A second independent variable hypothesized to predict

variations in readers' comprehension and recall of text was

the strength of their schemas for the particular pattern of

organization inherent in the text. This factor is not,

strictly speaking, an independent variable in that it was

not experimentally manipulated. However, since it was

hypothesized to be a predictor of changes in recall and

comprehension, it was analyzed as an independent variable in

this study. It has been shown that recognition of text

patterns increases recall and comprehension (Mandler, 1978;

Fitzgerald & Spiegel, 1983). Wicks (1986) points out,

"Schema theory suggests that individuals possessing a well

defined schema in [a] domain will have more success at

recall of related information (p. 7)."

Table 3-1

Descriptive statistics on the stimulus passages used to

measure recall and comprehension

Wicks notes that generally schemas have been measured

either by means of survey questionnaires or through

experimentation, adding:

Most of the studies aimed at demonstrating the
presence of a schema rely on measurement approaches
that test recall, evaluate inferential capabilities and
assess the tendency of an individual to cluster related
concepts. (p. 4)
In this study, the method of measurement used by Bower,

Black, and Turner (1979) and Kinney (1984) was adopted: The

subjects' familiarity with different text structures was

assessed by their responses to a task in which they were

instructed to reconstruct scrambled passages to form well-

ordered, typical news, narrative, or expository texts,

depending on which type of scrambled passage they were

given. The scrambling of the stimulus passages was

systematic in that each text passage was broken down into

its nodes according to the text structure on which it was

based. Each subject was randomly assigned to one of three

structure conditions (news, narrative, or expository).

Again, the passages used were similar in terms of

length and readability level, although the content differed

(see Table 3-2).

The Dependent Variables

The response variables hypothesized to change with the

structural organization of the stimulus text and with

readers' schemas for that structural organization were

recall and comprehension of the content of the text. As

indicated in the previous chapter, a substantial body of ex-

perimental research indicates that variations in text

organization result in significant changes in recall of both

fiction and nonfiction passages; these findings provided the

basis for the present experiment, in which a news story was

used as the primary stimulus for further investigations

along these lines


Typically, memory for measured as either recall

or recognition. Recognition measures include multiple-

choice questionnaires and retelling tasks in which subjects

are prompted to remember specific pieces of information.

Recall is frequently measured more informally, usually by

means of a request to "write down brief descriptions" of

what is recalled (Gunter, 1980) or requests for verbal

descriptions of the stimulus passages (Edwardson, Kent, &

McConnell, 1985).

In the present experiment, recall of the stimulus

passages was measured according to the procedure developed

by Meyer (1975), adapted by Taylor (1980) and Taylor and

Samuels (1983),and later used by McGee (1982) for scoring

recall of expository text--a method similar to the scoring

procedure followed by Mandler and Johnson (1977) for

measuring recall of narrative text. After being given an

interference task in which they provided the researcher with

Table 3-2

Descriptive statistics on the stimulus passages used to

measure strength of text schema


demographic information and answered a test evaluating their

reading level (the West Informal Reading Evaluation; West,

1978), subjects were asked to write down an account of the

stimulus passage they had read, keeping as close to the

original version as possible. The recalled texts were

scored by comparing them to the originals on the basis of

the proportion of elements of the initial passage recalled

per structural node. The sequencing of the recalled

propositions was not analyzed in this experiment, since

sensitivity to text organization was assessed using the text

schema instrument. A simple score based on the overall

proportion of the stimulus passage recalled was judged to be

sufficient. The overall proportion of terminal nodes

recalled was computed as a percentage score.


Comprehension was measured on the basis of nine

questions. Questioning is a standard method of gauging

comprehension (see Anderson & Biddle, 1975; Johnson, 1983;

Wilten, 1987; Trabasso, et al., 1988). The types of

questions used were loosely based on Pearson and Johnson's

description of textually explicit and scriptally implicit

questions (Pearson & Johnson, 1978) as well as on the

comprehension questions used in Johns' (1988) Basic Reading

Inventory, a standardized informal reading evaluation

instrument. Five of the nine questions were literal,

eliciting information based on main points or details that


could be found directly in the text. Of the remaining four

questions, two were evaluative, requiring the reader to

assess a situation; one was inferential, calling on the

reader's prior knowledge of the situation; and one was a

global question asking for a summarization of the main point

of the stimulus passage.

A tenth question was included which asked, "How well do

you feel you understood the story?" This general self-

evaluative measure was based on the concept of comprehension

used by Housel (1984) and Thorndyke (1977).

Reliability and Validity of Measurement Instruments

As Selltiz, et al. (1976) point out, "The quality of

research depends not only on the adequacy of the research

design but also on the quality of the measurement procedures

employed" (p. 160). Good measurement instruments must be

dependable measures of the target concept--i.e., they must

be as free as possible from random error caused by testing

conditions or inconsistencies among scorers or observers.

In addition, they must accurately identify and measure the

concept in question. The two principal properties of an

instrument that affect its usefulnesss as a measurement tool

are its reliability and its validity.


Reliability of a measurement instrument refers to

the steadiness of scores on the instrument. Reliability may

be measured in terms of stability, or consistency of scores

over time; internal consistency, sometimes called

homogeneity--the similarity of items within a test or other

instrument; and equivalence, or consistency across different

forms of the same instrument. This third aspect of

reliability becomes important only when different forms of

an instrument are being used to measure the same construct.

The reliability of the instruments used in this study

was assessed in terms of stability and equivalence.

Reliability of the instruments was measured using the

alternate forms method. Twenty-four undergraduate students

at the University of Florida were asked to participate in

the reliability study. The students responded to the

various measures in the experimental sequence described

later in this chapter. Two days later, the experiment was

repeated with the same class; however, while students

remained within the same experimental condition (news,

narrative, or expository), they were given different

stimulus passages on the second day. Thus, they were

effectively given alternative forms of a single test.

A coefficient of stability and equivalence was computed

according to the formula

A5B = (A-Ma)(B-Ms)


where A represents a subject's score on the first test

(Test A), B represents the subject's score on the second

test (Test B), MA represents the mean score on Test A, Me

represents the mean score on Test B, SDA represents the

standard deviation of scores on Test A, and SDB represents

the standard deviation of scores on Test B. (For a more

detailed explanation of the alternative forms method of

reliability assessment, see Walsh & Betz, 1985, pp. 50-51,

and Horvath, 1985, pp. 71-85).

Reliability coefficients for the instruments used in

this experiment are given in Table 3-3. In some cases the

reliability coefficients computed were slightly below 0.50.

Table 3-3

Reliability coefficients of measurement instruments


Schema measure

News 0.57

Narrative 1.00

Expository 0.44

Recall measure 0.44

Comprehension measure 0.46

These low coefficients could be attributed to small

size of the sample used in the reliability tests; the

instruments weregenerally considered acceptably reliable.

Selltiz, et al. (1976) point out that low reliability

coefficients are not necessarily indicators of low validity

of measurement instruments (pp. 194-197). They argue that

in some cases, fluctuations in scores on measurement

instruments from one test administration to another or even

within a test are desirable in that tests that produce

extremely homogenous results are not as useful for making

fine discrimination among responses and may in fact reflect

a high degree of constant error. They note that the

assessment of reliability and validity occurs along a

continuum from convergence of scores to divergence,

depending on the correlations being computed, and that "if a

measure can be shown to be reasonably valid it must

ipso facto be reasonably reliable, since a measure with a

large error component could not show such consistent

relationships" (p. 197).


"The validity of a measuring instrument may be defined

as the extent to which differences in scores on it reflect

true differences among individuals on the characteristics

that we seek to measure" (Selltiz, Wrightsman, & Cook, 1976,

p. 169). In other words, the validity of a measure refers

to the extent to which it is a true gauge of the construct

it is supposed to measure. Here, the crucial questions

would be whether the recall instruments were accurate

measures of subjects' memory for the stimulus passages and

whether the comprehension instruments were accurate measures

of subjects' understanding of the passages' content.

Face validity

As a very superficial test of an instrument's validity,

the relevance of the instrument to the construct under

investigation should be apparent "on the face of it"

(Selltiz, et al., 1976, p. 178). Because the measurement

instruments used were derived from the stimulus passages

themselves and were constructed following the methods used

by earlier investigators of similar phenomena, the

instruments exhibited significant face validity.

Content validity

Content validity is an estimate of the extent to which

the measurement instrument is an adequate sample of the

domain or process being measured (Selltiz, et al., 1976, p.

179). Generally, content validity is assessed by submitting

the measurement instrument to the scrutiny of experts, who

verify that all facets of the construct or domain under

investigation are represented in the instrument. The

instruments used in this experiment possessed considerable

content validity because they were derived exclusively from

the stimulus passages read by the subjects, measuring recall

of each structural proposition within each passage and

comprehension of ideas contained within the passage. The

measures thus represented an adequate sample of the

processes under examination.

Construct validity

In this experiment, the dependent variables "recall"

and "comprehension" are constructs or abstractions that

describe traits possessed by the subjects--i.e., the ability

to remember and to understand text. Construct validation

refers to the process of estimating to what extent the

measurement instruments measure these latent traits.

Construct validation may be accomplished by means of

examining patterns of correlation of a measure with other

validated measures of the same trait (convergent validity)

and by showing that the trait as measured by the instrument

in question can be differentiated from other traits or

constructs (Selltiz, et al., 1979, p. 174).

Curtis and Jackson (1962) have suggested that high

correlations between measures intended to measure different

but theoretically related constructs provide evidence of

convergent validity. In this study, comprehension and long-

term recall were expected to vary together; they are

theoretically related but conceptually distinct constructs.

Their construct validity was estimated by measuring the

degree of correlation between subjects' scores on the

measure of comprehension and the measure of long-term recall

based on the same stimulus passage. The Pearson correlation

coefficient was found to be 0.33 (N=69); this correlation

was statistically significant, p = .003. Short-term recall

was measured by means of the same instrument used to assess


long-term recall; thus, convergent validity was established

for the comprehension and recall measures.

The instruments used to assess recall of the passages

were thus generally judged to be reliable and valid.

The Control of Possibly Confounding Variables

The dependent variables, recall and comprehension of

the informational content of text, are influenced by

numerous factors. To date, most of the research on this

process has been conducted using children or adults with

reading problems as subjects; knowledge remains scarce

regarding influences on the reading recall and comprehension

of normal adult readers.

Age has been determined to be a major factor affecting

the reading ability of children (Chall, 1983).

Developmental stages of reading ability have been found

among adults with reading difficulties (Norman & Malicky,

1987), but these are not associated with age. Thus, while

subjects' ages were noted during data collection for this

experiment, they were not taken into account when the data

were analyzed.

Factors influencing the reading and learning abilities

of adults with impaired reading facility include cultural

background, physiological influences, and educational level

(Newman, 1980). Bowen and Zintz (1977) list sociological,

physical, environmental, and psychological factors as


significant influences on the reading abilities of adults in

literacy programs; they also include learning ability or IQ

as an important determinant of reading skill.

In this experiment, the final subject group chosen

consisted of undergraduate students at the University of

Florida. Their ages ranged from an 18-to-24-year-old group

to a 50-to-64-year-old group. The modal age range was 18 to

24 years, with 87.5 percent of the subjects falling into

this group. Most of the subjects used in the experiment

happened to be women (89 out of 104 subjects were female).

Ninety-four of the subjects were white, six were black, two

were Asian, one was Hispanic, and one did not indicate a

racial designation. In a general sense, the group could be

seen as homogenous: little variation in age, SES,

educational background, race, or cultural background was

evident. The selection of subjects alone resulted in the

effective control of sociological, environmental, and

educational variables which may have interfered with the

experimental manipulation.

Norman, Malicky, and Fagan (1988) have pointed out that

the reading level of the adult reader could affect text

comprehension and recall. Although it could fairly safely

be assumed that because the subjects were all college

undergraduates at the same institution, their reading levels

were on a par, reading level was nonetheless measured using

the West Informal Reading Evaluation (West, 1978). Data


collected from the eight students reading below the college

level, as estimated by the WIRE, were not included in the

data analysis. Reading level was thus eliminated as a

confounding variable.

Schema theoretic views of reading suggest that prior

knowledge of text content strongly influences reading

comprehension and recall (Pearson, Hansen, & Gordon, 1979;

Stahl & Jacobson, 1986; Perin, 1988). Thus, two text

passages were used in the experiment as a verification

tactic, to determine whether the observed effects occurred

regardless of text content. Passage I was drawn from a wire

story published in The Independent Florida Alligator. The

original article referred to incidents in Beijing, China, in

May 1989, but the setting was changed to Paramaraibo,

capital of the South American country of Surinam. All names

in the passage were changed. Passage II was a brief wire

article about a captive-breeding program to protect the

endangered Florida panther.

Within each text passage condition, the different

structural versions were kept as similar as possible in

terms of their length, average sentence length, and

readability (calculated according to the Fry formula

conceived by Edward Fry, 1977), so that any influences on

recall and comprehension arising from these factors might be

minimized. Only the passages' structures were varied.

Descriptive statistics pertaining to these characteristics

of the stimulus passages are provided in Table 3-1.

The content of the two passages precluded rewriting to

exactly follow the narrative or expository text grammars,

but these were approximated as well as possible. The

narrative version of the political uprising story fell short

of representing a true story in that it lacked a single

protagonist--the political demonstrators filled that role en

masse. In the panther story, a goal was implied in the

setting in that the protagonist's desire to combat a problem

was tacitly expressed in the first sentence. However,

generally speaking, the passages adhered to the Handler and

Johnson (1977) story grammar in their narrative forms and to

the Meyer (1975) expository/attribution grammar in their

expository forms.

The Experimental Design

The overall experimental design comprised a single-

factor posttest-only procedure. The manipulated variable

was text structure; the response variables were short-term

recall, long-term recall, and comprehension. The strength

of subjects' text schemas for each of the structures used

was also measured as a quasi-independent variable. The

experiment consisted of six tasks.


The power of an experiment is the probability of

rejecting the null hypothesis if the null is false. This

probability is increased by minimizing the chances of making

a Type II error by finding support for a false null

hypothesis. The power of an experiment is determined by the

size of the sample of the subject population (see Winer, p.

104). Fifteen subjects per cell were computed to be

required to minimize Type II error at a 0.05 level of

significance. Because there were six cells in the design,

90 subjects were needed for this experiment.

Subjects were undergraduate students at the University

of Florida, drawn from senior-level classes in journalism,

education, and speech communication. However, not all

subjects were majoring in these disciplines. After the

elimination of the poor readers, one hundred and four

students participated in this study. All students were

juniors or seniors. Subjects' ages ranged from 18 to 64.

Table 3-4

Male 15
Female 89



Some high school
High school diploma
Some college
Bachelor's degree

INCOME: Less than $6,000
$ 6,000-$11,999

$60,000 or more

(Missing = 3)






C"Vnm +-=hl do of descriptive statistics on sa- ILLL m aYUhjm-*=

rlrmms~rv +~h~er

~f ~Ps~ri~+iv~

~fnficfi~c ~1~ c;rmrrlc~ ~rrki~~+c


Two stimulus news stories were rewritten to follow

three schematic structures: (a) the prototypical Newsom and

Wollert (1988) "modified inverted pyramid" news structure,

(b) the Mandler and Johnson (1977) story grammar, and (c)

the expository/attribution text structure described by Meyer


Subjects were randomly assigned to one of the six

passage conditions--Passage I news, Passage I narrative,

Passage I expository, Passage II news, Passage II narrative,

or Passage II expository.

Each subject received a packet containing the

measurement instruments. All measurement instruments and

stimulus passages are included in Appendices A and B.

Subjects were first asked to complete the Media Use


Subjects were then presented with one of the two

stimulus passages in one of the three organizational

patterns. They were given sufficient time to read the

passages; the stimuli were then removed.

Subjects were next asked to complete a multiple-choice

questionnaire requesting demographic information, and then

they were given the WIRE test.

Next, subjects were tested for strength of text schema

by means of an unscrambling task based on three passages

constructed according to prototypical narrative, news, and

expository structures. These passages were written by the

author; the narrative story was based on the fable "The Owl

Who Was God" by James Thurber (Thurber, 1941), the

expository text was based on a passage used in the 1978

Florida Literacy Test (Morrison, 1978), and the news story

was an Associated Press story. Each subject was given a

labeled envelope containing a number of slips of paper; each

scrap of paper was printed with a node from the scrambled

stimulus passage. Subjects were instructed to organize the

pieces of paper so that they formed a typical text passage

of the type printed on the front of the envelope--either

news, narrative ("story"), or expository ("textbook"). The

structure of the passage used for the unscrambling task

corresponded to the structure of the initial stimulus

passage that each subject read. Once the subjects had

arranged the pieces of the passage to their satisfaction,

they were asked to number the scraps in sequence. Subjects'

sequencing of each passage was compared to the original,

parsed version of the passage, and a difference score was

computed to determine the correlation between subjects'

organization of the nodes and the actual, prototypical

organization. Subjects' reorganizations of the passage were

scored by comparing the relative order of propositions in

the subject-generated passages to the original, correctly

organized passages. Scores were computed by calculating

rank-order correlation coefficients (Spearman's rho)

comparing subjects' ranking of the propositions to the

rankings in the original passages (a full account of the

scoring procedure is given in Appendix C; also see Hays,

1963, p. 645, for a more complete description of the

calculation of difference scores and their use in computing

rank-order correlation coefficients).

Various methods may be applied to the calculation of

rank-order correlations. Perhaps the simplest way to

calculate the proportion of pairs of items having the same

relative position in two rankings is a graphic method. In

this method, the objects ranked are listed, once in the

order of the first ranking, and once in the order of the

second. Then straight lines are drawn connecting the same

objects in the two parallel rankings. This method, however,

is time-consuming, and was therefore rejected in favor of

the computation of difference scores based on calculating

the numerical difference between two rankings xi and y; ,

such that the difference score D; = x; y1 and squaring

the result. The sum of the squared differences was then

used in the following formula to compute Spearman's rho, the

rank order correlation coefficient:

r = 1- 6 Di

[N(Nz 1)

After these three distractor tasks, subjects were asked

to recall the stimulus passages and write them down using

language as close to the originals as possible. Their


instructions included the caveat that if they were not able

to remember the original wording of the passage,

approximations were permissible.

Finally, subjects were asked to answer the

comprehension questionnaire.

The recall instrument was administered one week later

to the same subjects. Subjects were asked to recall the

passage they had read and write it down using language as

close to the original as possible.

Recalls were scored by parsing the subject-generated

passages and recording the number of text elements recalled

from the original for each node in the appropriate text

grammar. These numbers were converted to percentages. A

full explanation and example of the scoring procedure is

given in Appendix C.


To determine the overall effects of altering the

structure of the stimulus news stories, an analysis of

variance using the fixed-effects model was conducted to

compare the mean recall and comprehension scores across the

three text structure conditions for each of the two

passages. Although the hypotheses, as stated above, focused

on comparisons between pairs of means, the analyses of

variance were efficient methods of summarizing the results.

Each hypothesis was then tested by means of a t-test

comparing mean scores on the dependent variables as

specified in the hypothesis.

All statistical analyses were performed using SPSS-X.

The level of significance generally used in this study

was = .05, although if R lay between .05 and .10, the

relationship was judged to be statistically significant.

Rejecting the null hypothesis at the .05 level indicates

that if the null were true, the probability would be less

than .05 of obtaining a test statistic value as favorable to

the alternative hypothesis as the one observed. This

research was in a sense exploratory in that the method and

the measurement instruments were developed for this study

and had not been refined through years of experimentation;

thus, the significance levels used were to some extent


Some Considerations: Threats to External
and Internal Validity

As is the problem with most experiments conducted in

controlled conditions, any results obtained cannot easily be

generalized beyond the laboratory situation, which presents

a serious, and inevitable, threat to the external validity

of the proposed study.

However, to compensate for this inherent flaw in the

proposed method, threats to internal validity were minimized

by randomizing subjects' assignment to the six experimental

conditions, standardizing stimuli and experimental

conditions, and controlling for variables which may have

interfered with the relationship being studied. These steps

helped ensure that the manipulation of the independent

variable was the cause of any observed variation in the

dependent variables.

Internal validity in many experiments can be reduced

because of maturation of subjects, mortality or attrition of

the subject pool, history (i.e. some unexpected factor which

may coincide with the independent variable and produce a

similar effect), or reaction to a pretest. Since the

proposed experiment was conducted within a fairly short

time, it can be assumed that maturation, mortality, and

history did not threaten this particular experiment. No

stories similar to the stimulus passages appeared in local

newspapers during that period. A pretest that may have

interfered with performance on the measures of the dependent

variables was not administered, so this threat to internal

validity may also be ruled out; the absence of a pretest

also eliminates any possibility of a regression toward the

mean in scores on the dependent variable measures.

Any threat to internal validity that may be due to

inadequacy of the measurement instruments was minimized by

testing all measures for reliability and validity before

they were used. Some attrition occurred between the

administration of the first set of instruments and the test

of long-term recall, which may have affected the validity of

these results to some extent.

4-4 ac U)
0 C 4 4J
a -4H i

0 >m
-H 4-I C
4J 0 40
o0 -4 -I 4)

LI --4
0I >i.C A
Li4..4 0

4l -H
C0 0 4-
CO 04
40 OU0
co .
'-M 4J CP 4J

a)X C
S -i 0 o: -H
U0 4- 4-
04 M r-
C --i 0 C 0
-. -i 3 0 o

C -*H


0 -4 0

0 G
00 04-H
. 0'T3
0 0
4 )4

4 00
10 0 U


0a o 0
0 E-H 1
0 CO 4J 0
C -H >

0 O
0 0

40J (0 4-4
X 01
00o (fl-
4J 0 0-4
M4-1 ao -

0 I
1 0

4- O 04 0 o

I C -H -4
004J 4J 400 4J
S0 -H 0 (

) .--I (04-
0, -- 0 C4

0) -H U) 4J 0
4-1 E
dr-I 44 M
O(O 0)
(40 0
4J O-Hl Cu a
SC 4-4 0Q. Q.-H
) C > >i
e -H 4J 4r

(D ) 0

0n o 0 0 0

0- C C -4 OU -
j-1 (
I- o 0 0 .C 04'
0-A a) c rio arl 4
3 ) 4J () I -rH 4J -L,
04 0 C H 04J 0

4. 040 0 0 2.0
H a o a

CO) -l-HaOr O)
- a d 0 ) 0) 0) to
-, W 4)i 0 m V

S4-H p 0 L) 0) -

4 4-J U O Q) 04 4

4U 4)

0 (0

N -H



4 -H

N 0


, 0 --

0 0

:c 0


Y 6
O- I



4-4 -r

4. -4





E 14
0 oI

(a -ri 0
0a) a>
) 1 r-4 C4
1 4-)
C aU 10 -H 0
aO 4 4i

04) 4 ->-1

rC 4) M) Ca
- -1-1 4 4 -1 4-

0 m

0 0)
4 P
a) .

U 0 I

>i 0(0
a) > O
-! 04a) a)
- ^ s-I

a) E (a -H
p ()
4-4 4-) U)
-H 4- a

U) 0
) r-d 4-4
-r-i '-I C 04
a aO )

4 0 U)
4j *) 0
4- 4 *

4 *



>1 0

C -H

I U 43
0 )i ( 4-1

O U CO a
-rl a) U *a
0 U cr--4
4-H 4 -) 4-a (0

0 43 0
4-I C a)4-
O -H1 a) p J
I C C 4-
040(0 -H
-H U) a) 4-)
a) 4) U) r- -rH
4c -H i0 0 X M
E4 ) 4 (4 :

0 a)
10 10

S-H 4 -1

0(0 a)
>U C
CO M 4)
C (3

n0 -4 0D U)
03 a) O a)
(0 0 (0
5- HU-4
E-i PQ (0

( 4
I -

0 (d 0 E-

S0 >i -iHO
QM 4-
4-4 rl U) a) C
0-H 04H ) -H,
'0 O' :
C 01 UG
0 Ur Q) r.C 4-)
-rH r 0 H4 0 i
4-) -H- -- O () -H- 0
0 '0 4-) P P -4-
E C co 4j 04 U) O
so M 4- () o0 -H
04 U) 01 U 0 14

> *
C r-A -1 (d (0

-H 0"--0
OC ( c141QUw

U) 4 4-1 3 0) -H 1)
O0 a) Xa T 0 U)1

0, C3 4>UCU
a) -il +J 4J 4J 01
4J 4-) r4 -HI -H -H
U -HI -H U 0 U a)
a) I O -H.- -r-4 -H
,4 4 4J r- r-- -4 4-)
s- aC) Q 4H
O -i 0 x ee
U rC 0 a)-H-H 0

Ia -4

4O 40
e- HGe -

4 (d U)
4 a) -r04-

*0 0 0
CO a.4
aE .Ma -a
r-1 a )
n) g, *o 4
U M ( 0) 4


) 3 U) rH
3 >1 -H-
:3 (0 Ia)

44 CO 0-H1
S0 ) -r-4
a) (0 >

Ca 4I *aC
0 '3 -H 0
G) r -H


9 >


0 O

U a


-H 0l

U) -H
P 0
-rl *O
0 -ri
M o


An investigation was conducted to determine whether

altering the underlying organizational structure of a

printed news passage would affect readers' recall and

comprehension of the information contained in it. The

independent variable in this experiment was the text

structure of the passage, which was altered to follow a

typical news, narrative, or expository structure. The

dependent variables were short-term recall, long-term

recall, and comprehension of the passage's informational


Two passages were used as stimuli in the experiment.

Subjects were randomly assigned to read either the passage

about a captive-breeding program to protect the Florida

panther or the passage about a political uprising in


General Results

The data were analyzed to test the hypotheses proposed

in Chapter 3. Initially, separate one-way analyses of

variance were conducted for each passage to compare the mean

recall and comprehension scores of subjects in each

experimental condition and summarize the observed results.

Detailed descriptions of these analyses as well as the

statistics used to test each hypothesis are given below.

Manipulating the structure of the stimulus stories had

somewhat different effects on each of the dependent


Overall, it appeared that the subjects' short-term

recall of the passages was significantly influenced by

manipulation of the text structure, F (2,50) = 2.881, R =

.065 for the panther passage and F (2,45) = 3.529, E = .038

for the political uprising passage (Table 4-1).

However, subjects' comprehension of the two passages,

as measured by the ten-item questionnaire described in

Chapter 3, did not appear to be significantly affected by

the experimental manipulation overall, F (2,50) = 1.985, E =

.148 for the panther passage and F (2,45) = 0.179, E = .836

for the political uprising passage (Table 4-2).

Long-term recall appeared to be significantly affected

only in the case of the panther passage, E (2,35) = 3.102, E

= .057; no significant differences were found for the

political uprising passage, F (2, 25) = 0.907, E = 0.416

(Table 4-3).

Table 4-1

Analysis of variance of short-term recall across text


Panther Passage







712.14 2.881 .065



Political Uprising Passage







449.44 3.529 .038



Table 4-2

Analysis of variance of comprehension across text structure


Panther passage







2471.57 2 1235.78 1.985 .148

32380.53 52 622.70

34852.10 54 645.40

Political Uprising Passage




2 103.84

46 579.17

48 559.36

.179 .836

Table 4-3

Analysis of variance


of lanar-term recall acrnss te~t


Panther Passage








1007.14 2 503.57 3.102 .057

6006.75 37 162.34

7013.90 39 179.84

Political Uprising Passage







.907 .416

These broad trends were investigated in terms of the

experimental hypotheses expressed in Chapter 3 by conducting

a series of t-tests to study the differences in mean scores

on the dependent variables between each pair of experimental

groups. The results of these tests are discussed in the

following sections.

Effects of Altering the Text Structure on Short-Term Recall

HI: A text passage written to follow a narrative story

grammar will be better remembered in the short term

than the same passage organized according to an

expository text grammar or a news story grammar;

of -~~~... recal ac-ros... text

similarly, a text passage organized according to an

expository text grammar will be better remembered than

the same passage organized according to a news story


The hypothesis predicting associations between the text

structure of a passage and short-term recall was, overall,


This hypothesis was tested by comparing the mean short-

term recall scores of subjects who read the narrative

version of each of the two stimulus text passages to the

mean short-term recall scores of subjects who read the

expository and news versions of each of the two stimulus


Results of the t-tests for each text condition are

described below and shown in Tables 4-4 through 4-6.

Narrative Versus News Grammars

Panther passage

The mean short-term recall score of subjects who read

the narrative version of the panther passage was

significantly higher (M = 31.7) than the mean score of

subjects who read the news version of this passage (M =

22.3), t (35) = 2.06, p = 0.024.

Political uprising passage

The data for the political uprising passage paralleled

the findings for subjects who read the panther passage.

Those subjects who read the narrative version of the

political uprising passage also displayed better short-term

recall of the passage's content than did subjects who read

the news version of this passage. The mean short-term

recall score of subjects who read the narrative version was

25.3, while the mean recall score of the group that read the

news version was 14.9. The t-test indicated a significant

difference between these two groups, t (33) = 2.59, R =


Table 4-4

Comparison of mean short-term recall scores of narrative
story group with news story group


Panther Passage

Narrative 17 31.7 12.83 3.11 2.06 33 .
News 18 22.3 13.83 3.26

Political Uprising Passage

Narrative 18 25.3 14.34 3.40 2.59 31 .
News 15 14.9 6.50 1.68



Thus, in the case of both text passage conditions, the

narrative structure appeared to result in better short-term

recall of the passages' content than the news version.

Hypothesis 1 was supported.

Expository Versus News Grammars

Again, separate t-tests were performed for subjects in

each text passage condition to compare the mean short-term


recall scores of subjects who read the expository version of

each passage with those of subjects who read the news

version of each passage. T-test results are reported below

and given in Table 4-5.

Panther passage

The analysis showed that subjects had better short-

term recall of the passage written according to an

expository text structure (M = 34.2) than of the passage

written according to the news story structure (M = 22.3), t

(37) = 2.14, p = 0.020.

Political uprising passage

Again, in the case of the political uprising passage,

the short-term recall of the expository version of the

passage was higher (M = 21.1) than short-term recall of the

news version of the passage (M = 14.9), t (31) = 1.93, R


Thus, the second hypothesis was also supported by the

experimental data.

Narrative Versus Expository Grammars

A third set of t-tests was conducted to compare

subjects' short-term recall of the narrative version of each

of the two passage to their short-term recall of the

expository version of the each of the two passages. These

results are given below and shown in Table 4-6.