THESlS

 

 

This is to certify that the
dissertation entitled
WR ITING ACH IEVEMENT:
A COGNITIVE DEVELOPMENTAL ANALYSIS " ‘
presented by

Susan Wildfong

has been accepted towards fulfillment
of the requirements for

Ph. D. mgmem Educational

 

 

Psychology

 

Dateflovember lO . 1981

MS U is an Affirmatt'w Action/Equal Opportunity Institution 0-12771

 

 

MSU

LIBRARIES

RETURNING MATERIALS:

 

 

 

Place in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped below.

 

 

 

 

‘5‘

WRITING ACHIEVEMENT:

A COGNITIVE DEVELOPMENTAL ANALYSIS

By

Susan Wildfong

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Educational Psychology

1981

ABSTRACT
WRITING ACHIEVEMENT:
A COGNITIVE DEVELOPMENTAL ANALYSIS
By
Susan Wildfong

The purposes of the dissertation were to trace the
development of children's writing over the school year using
a cognitive-development model and to determine the effect of
grade and reading ability on writing achievement.

Bereiter has proposed that writers progress from a stage
of associative writing. in which they attain fluency at idea
production and writing, to a stage of performative writing,
in which they incorporate conventions, such as mechanics for
producing correct text. The purpose of the present study was
to determine the increase in children's use of conventions.

Narrative compositions of fifty-four fourth, fifth, and
sixth grade children were analyzed for use of conventions of
story structure, cohesion, sentence formation, punctuation,
and spelling. Four multivariate analyses of variance re-
vealed a significant effect for Occasion on sentence correct-
ness and punctuation. Examinations of the means indicated
that students showed improvement over the year in punctuation.
Sentence correctness improved from Time 1 to Time 2 but
declined from Time 2 to 3.

Occasion X Grade interactions were obtained for the
other variables. Students improved over the year in spelling

accuracy, except sixth graders from winter to spring. All

Susan Wildfong .
students declined at some time in their overall use of
cohesion. Fourth graders' use of story structure increased
while fifth and sixth graders' use declined.- These data
indicated an increase in some conventions, primarily mechanics
over time. However, conventions for structuring text and
providing cohesion did not increase. Differences in growth
may have been due to the cognitive processes responsible for
the different conventions, to differences in instruction, or
to the difficulty of the conventions.

No significant main effect for Grade was found for any
variable. Grade X Occasion interactions were obtained as
reported.

A significant main effect for Reading Ability occurred
for the sentence correctness, punctuation, and spelling
variables. The better readers‘ advantage may have been due
to greater knowledge of the surface form of written prose

or to rereading skills.

ACKNOWLEDGEMENTS

I would like to thank those who helped in coding the
data: Candis Bonner, Nancy Broxholm, Cathy Emerson, Jo Meyer,
Janis Shubitowski, and Carol Weigand. I would like to express
my deep gratitude to Frank Jenkins for the effort and count-
less hours he expended on the data analysis. I also want to
thank Gabriella Belli for her role in the data analysis.

My committee has been, to a person, supportive, co-
operative, and instructive. I want to thank Helen Benedict
for inspiring my interest in language development. Bill
Schmidt aided me in design and analysis of the dissertation
and directs the project from which this study was drawn. I
want to thank Dick Prawat for his invaluable encouragement
during my program and for providing a model of scholarly and
enthusiastic scientific work. I especially want to thank my
chairman, Jere Brophy. He has been unfailingly helpful to me
during my program and this dissertation. He also is the
standard of research and teaching excellence to which I have
referred for years. I want to express my gratitude also for
the friendship of these people.

I want to thank my parents for their love over the years,
for supporting my scholarly interests and for always telling
me that my efforts were important and worthwhile. I want to
thank my daughter Becky for her love, her patience for five

years and counting, and for always making me proud of her.

ii

TABLE OF CONTENTS

LIST OF TABLES I I I I I I I I I I I I I I I I I I I

LIST OF APPENDICES I I I I I I I I I I I I I I I I I

Chapter
I.

II.

Statement of the Problem.
Need for the Study. . .
Purpose of the Study.
Design of the Study .

Models of the Writing Process . . . . . .
Bereiter's Model of Writing Development .

INTRODUCTIONI I I I I I I I I I I I I I I I I

Data Source . . . .
Research Questions.

Subjects. . . . . . .

REVIEW OF LITERATURE. . . . . . . . . . . . .

Product-Oriented Research . . . . . . . . . .
Quantitative Research on Writing Development

summary I I I I I I I I I I I I I I I I I I
Qualitative Research on Writing Development
Overall Quality . . . . . . . . . . . . . .
Organization. . . . . . . . . . . . .
Discourse Attributes. . . . . . . . .

"Process-Oriented" Research on

III.

Writing Development . . . . . . . . . . .
Evidence Related to Associative Writing .
Summary . . . . . . . . . . . . . . . . .
Evidence Relate to Performative Writing.
Discourse structure in narrative. . . .
Lower level organization in
narrative - an alternative view . .
Cohesion. . . . . . . . . . . . . . .

Mechanics . . . . . . . . . . . . .
The Pattern of Development in Writing

NIETH OD I I I I I I I I I I I I I I I I I I I I

subjeCtSI I I I I I I I I I I I I I I I I I I

Page

vi

mnouaamno\ O\ u\¢xnun»roh’ F“

Chapter . Page

Instruments . . . . . . . . . . .

Narrative Structure Instruments . . . . . . . 5l
Cohesion Instrument . . . . . . . . . . . . . 5h
Sentence Correctness. . . . . . . . . . . . . 56
Mechanics: Punctuation Maturity. . . . . . . 56
Mechanics: Spelling Maturity . . . . . . . . 57
Data Collection . . . . . . . . . . . . . . . . 58
Data Preparation. . . . . . . . . . . . . . . . 59
Data Reduction and Analysis . . . . . . . . . . 6O
DeSignI I I I I I I I I I I I I I I I I I I I I 63
Statistical Analysis. . . . . . . . . . . . . . 65

Iv I RESULTS I I I I I I I I I I I I I

O\
V

Hypothesis #1

HypOtheSis #2 : o o o o o : o o c u o o o o o : 77
HyPOtheSiS #3 o o o c o c o o o o o o o o o o o 78
HyPOtheSis #4 o o c I c o o o . o o o o o o o o o 80

V. SUMMARY AND CONCLUSIONS . . . . . . . . . . . . 90

Summary . . . .

Conclusions . . . . . . . . . . . . . . . . . . 93
TeXt StruCture I I I I I I I I I I I I I 'I I I I 93
COheSiono o o o o o o o o o o o I o o c o o I o 97
Implications I I I I I I I I I I I I I I I I I I 109

BIBLIOGRAPHY I I I I I I I I I I I I I I I I I I I I I ll 2
APPENDH A I I I I I I I I I I I I I I I I I I I I I I 115

iv

Table

A-3

A.h

LIST OF TABLES

Occasion Effect . . . . . . . . . . . . . .

Occasion X Grade Interactions . . . . . . .

Content Categories. . . . . . . .°. . . . .
Ability Effects . . . . . . . . . . . . . .
Correlations. . . . . . . . . . . . . . . .
Overall Table for the First Multivariate

Analysis of

Overall Table
Analysis of

Overall Table
Analysis of

Overall Table
Analysis of

Variance of Repeated Measures

for the Second Multivariate
Variance of Repeated Measures

for the Third Multivariate
Variance of Repeated Measures

for the Fourth Multivariate
Variance of Repeated Measures

Page

75
76

77
8h

86

115

115

116

116

Appendix
A .

LIST OF APPENDICES

vi

CHAPTER I
INTRODUCTION

Statement of the Problem

 

What characterizes development in writing? Prior to the
1970's, research on writing development was primarily re-
search on the development of features of compositions asso-
ciated with quality in writing. This approach has been
termed a "reader-based" approach to assessing writing (Tamor
and Bond, 1980). This approach provided information about
quantitative changes in lexical and syntactic features of
writing and about the development of more global features,
such as organization, content, and style (Loban, 1976: NAEP,
1975: Stahl, 1977). While this information was useful for de-
scribing writing at different ages, it offered little insight
into the cognitive processes involved in producing written
text or the skills possessed by writers of different ability
levels.

In response to the need for such research, several re-
searchers have developed models of the composing process.
Flower and Hayes (1980) have described writing as a three
phase process. The first phase is a planning phase during
which the writer defines the writing task, sets goals for
writing, and generates and organizes ideas. The second phase
is a translating phase during which the writing plan, which

1

2
may be in nonlinguistic form, is translated into written
language according to certain rules. The third phase is a
reviewing phase during which the writer reads what he has
written in order to edit it or to generate further ideas.
A similar model of the composing process has been developed
by Bereiter (1979).

Bereiter (1979) proposed a model of writing development
based on models of the composing process. According to this
model, the skills involved in producing mature prose are too
numerous to be employed simultaneously by beginning writers.
Writing development is characterized by a series of stages
each of which involves the integration of more skills to
achieve a more adaptive level of performance. Lower level
skills of content and translation are integrated first,
followed by higher order skills of adapting the text to an
audience or to the writer's personal goals.

Some research has been conducted which supports this
model. For example, Flower (1979) related the characteris-
tics of beginning and mature prose to Flower and Hayes' model
of composing. Bereiter and hiscolleagues (Bereiter,
Scardamalia, and Turkish, 1980: Bracewell, 1980) have con-
ducted a number of cross-sectional studies, and have demon-
strated an increase with age in the use of some higher level

schemes, such as those for producing certain types of cohesion.
Need for the Study

While these researchers have investigated the development

of process-related features of writing, no attempt has been

3

made to assess improvement on several features in a single
test of Bereiter's model. In addition, studies of the devel-
opment of these features have provided primarily cross-
sectional data. Such an approach has been criticized because
it allows for confounding of age and experience. This
research also has been conducted on different populations.
For example, Flower and Hayes have studied the composing
processes of college-aged adults while Bereiter and his
colleagues have focused primarily on children. Finally,

not enough data exists to allow for the characterization of
writing development as continuous or discontinuous.

This dissertation extended the above research by
examining growth on a set of features believed to reflect
development in writing. It provided longitudinal as well as
cross-sectional data on writing development. Finally, the

pattern of growth of each feature was assessed.
Purpose offithe Study

The purpose of this dissertation was to assess the
development of several features of children's writing to
determine if writing development followed stages derived
from Bereiter's cognitive developmental model. Also of
interest were the effects of age and reading ability on these

variables and the pattern of growth of each variable.

Design of the Study
Subjects
The subjects of the study were fifty—four fourth, fifth,

and sixth grade children from the area around Lansing,

L,

Michigan. The children were students of teachers who were
the subjects of a study of writing instruction conducted by
the Language Arts Project of'The Institute for Research on
Teaching. The students in each grade possessed different
social class and racial characteristics. The fourth graders
were from Okemos, a community of primarily white, upper
middle class professionals. The fifth graders were from
Dimondale and Stockbridge, two middle class, white rural
communities in central Michigan and from Lansing, a racially
diverse urban center. The sixth graders were from Lansing.

The students were given a test of reading achievement
at the beginning of the study. For the dissertation, they
were classified as above average, average, or below average
in reading ability.

Seven students in grades five and six were randomly
selected from each ability level at each grade. Only
sixteen fourth graders participated in the study: seven of
.above average, five of average, and four of below average
reading ability. To conform to a proportional design, four

fourth grade students were selected from each ability group.

Data Source

The data consisted of narrative compositions written
by these students at three times during the school year.
The compositions were produced in response to writing tasks

designed by Language Arts Project members.

Research Questions

The following research questions were investigated:

1. How does the narrative writing of fourth, fifth and
sixth grade children develop over the year on features de-
rived from a cognitive developmental study of writing devel-
opment?

2. How do these features differ for students of
different grades?

3. How do the features investigated differ in the
writing of students of above average, average, and below
average reading ability?

A. What is the relationship between features attri-
butable to a particular stage of writing development?

5. Is development on each feature continuous or
discontinuous, e.g., what is the pattern of growth for each

feature?

CHAPTER II

REVIEW OF LITERATURE

In this chapter, research investigations which form the

background for the original research to be conducted will be

reviewed.

1.

2.

3.
u.

The following areas or topics will be discussed:
"Product-oriented" research on writing development
a. Quantitative research on writing development
b. Qualitative research on writing development
Models of the writing process
Bereiter's model of writing development

”Process-oriented" research on writing development

Product-Oriented Research

Prior to the 1970's, research on the development of

writing ability focused on quantitative changes in features

of children's written products, including vocabulary, sen-

tence structure, fluency, and use of conventions of writing.

More recently, research on writing development has

investigated qualitative changes, including the overall

quality of children's writing, organization, and style.

Quantitative Research
on WfitIng Development

Vocabulary development in writing has been assessed by

the National Assessment of Educational Progress (1975).

6

7
The NAEP of writing assessed vocabulary level by counts of
the average number of letters per word. Vocabulary level was
found to increase between the ages of nine and seventeen.

Hunt (1965) developed three measures to assess syn-
tactic complexity in the oral and written language of child-
ren. These measures included number of words per T-unit,
number of clauses per T-unit, and number of words per clause.
The T-unit is defined as an independent clause and any sub-
ordinate or non-clausal constructions attached to it. Hunt
found that the mean number of words per T-unit in the oral
and written English of children in elementary and high
school increased gradually with age.

Loban (1976) conducted a longitudinal study of language
development including writing development from kindergarten
through twelfth grade. Syntactic maturity was assessed
using several measures, including number of words per com—
munication unit (a unit identical to the T-unit). Loban
also assessed the amount of subordination, transformation,
and elaboration in children's sentences. .Subordination
refers to the embedding of clausal and non-clausal units in
complex sentences. Transformations were assessed using
measures based on transformational grammar. Elaboration
refers to the expansion of a communication unit beyond a
simple subject and predicate.

The results of Loban's study showed that number of
words per communication unit increased gradually with age.
Elaboration also increased with age. One index of subordi-

nation, number of dependent clauses, increased gradually

8
until grade eight in the writing of high ability students and
then reached a plateau, increasing past this point only for
low ability students.

Friedman and Fowler (1979) assessed the development of
discrete and global variables in the writing of children in
grades two through six.- Their measure of productivity was
a count of the number of words in a composition. They found
that the number of words in children's compositions increased
from second through sixth grade. They also assessed the
development of spelling, sentence usage, and paragraph usage.
They found improvement in spelling, sentence usage, and
paragraph usage between second and sixth grade except for
fourth graders. -

In a recent study, Stewart and Grobe (1979) found that
the number of spelling errors and run-on sentencesin

children's writing decreased from fifth to eleventh grade.

Summary

The previously reported research documented the devel-
opment of aspects of children's written compositions,
assessed through quantitative measures. These aspects
include syntactic maturity, fluency, and mechanics of
writing. Syntactic maturity as measured by number of words
per T-unit shows a gradual increase with age: however,
certain aspects of sentence complexity such as number of
dependent clauses do not continue to increase in later school
years for high ability students. Older children may adopt
more advanced forms of embedding. Fluency in writing also‘

shows gradual development with age. Finally, children

9

improve with age in their use of conventions of writing such
as spelling, sentence usage, and paragraph usage.
anlitative Research on
Writing_Develqpment

While quantitative research has provided information
about the growth of features such as vocabulary and sentence
complexity, it has yielded little insight into aspects of
writing above the sentence level,such as organization and
style. However, some researchers (Lloyd-Jenes, 1974) feel
that these aspects are critical to the effectiveness of a
composition. The evaluation of more global variables, termed

holistic evaluation, has been of several types.

Overalpouality

The NAEP (1975) employed subjective ratings of the over-
all quality of compositions to assess qualitative development
in writing. This assessment revealed an improvement in qual-
ity between the ages of nine and eleven but little improve-

ment between the ages of thirteen and seventeen.

Organization

Stahl (1977) devised a scheme for rating the structural
organization of the descriptive prose of second, fifth and
eighth grade children. Nine aspects of organization were
assessed: indicated order, principle of selection, methods
of arrangement, syntax, balance, organization, use of con-
nectives, type of opening and type of closing. Indicated
order referred to the writer's use of an order of ideas,

either explicit or implicit, and to the writer's consistency

10
in following this order. The dimension of principle of
selection referred to the extent to which a writer described
critical features of an object and omitted irrelevant details.
Methods of arrangement related to the way in which a writer
presented the content.

The syntax dimension concerned sentence structure and
relations between sentences. The dimension of balance was
used to rate the amount of emphasis given to ideas. The
organization category was intended to capture the writer's
explicit use of planning and revision in composing, as
indicated the use of an outline or corrections in writing.
Finally, the introduction and closing dimensions concerned
the presence or absence of a beginning or ending which were
integrally related to the body of the composition.

Stahl rated children's compositions on a five point
scale for each of these areas. The average composition of
second grade students received a low rating on principles of
selection, arrangement, syntax, balance, connectives, and
introduction and conclusion; the average eighth grade compo-.
sition received a high rating on these categories. Fifth
grade children's compositions were bi—modally distributed
(Bereiter, 1979), receiving either a low or high rating on
most attributes. Children's compositions at all grade levels
showed little use of an indicated order of presentation or

planning and revision.

Discourse Attributes

A study of the development of different modes of dis—

course in children's writing was conducted by Britton,

ll
Burgess, Martin, McLeod and Rosen (1975). The study was
based on a model of discourse situations. The model Speci—
fied the elements of discourse situations as: the speaker or
writer, audience, subject matter, and activity. According to
Britton et. a1. there are three modes of discourse which can
be differentiated according to the elements in the writing
situation which they emphasize. Expressive writing is
characterized by an emphasis on the writer and the discovery
of writer's thoughts. In transactional discourse, the
writer and reader play the role of participants in regard
to the discourse and the focus is on the communication of
information. The final mode of discourse is poetic writing.
Poetic writing results from an emphasis on the written
product as an artistic object. Both writer and reader play
the role of spectators to the discourse.

Britton et. al. found a trend from expresSive to trans—
actional to poetic writing in children ages eleven to
eighteen. However, poetic writing declined in the upper
grades, reflecting in Britton's view.the effect of the
evaluation system in Great Britain on children's writing.

While research on written products can describe the
qualitative and quantitative changes in students' compo-
sitions, it is limited in its ability to inform about the
underlying processes or abilities in writers which result
in these changes. In the 1960's, Parke (1961) and Braddock
(1963) suggested that process-oriented research on writing
be conducted. Tamor and Bond suggest that “to be instruc-

tionally and/or theoretically useful, text analysis

12
procedures applied to compositions of primary school children
should be based on variables derived from a well-defined model

of the writing process" (Tamor and Bond, 1980, p. 7).

Models of the Writing Process

 

In order to assess the development of the integration
of skills in writing, it is necessary to specify the mental
processes that occur in the production of written text.
Models of composing derive from models of the production of
speech (Fodor, Bever, and Garrett, 197#: Clark and Clark,
1978).

Clark and Clark (1978) describe speaking as a goal-
oriented action.“ Speakers speak in order to influence
others, to obtain things, to represent knowledge, etc.
Speaking is divided into two processes: planning and exe-
cution. In the planning phase, speakers mentally represent
their discourse based on how they want to influence their
listeners. In the execution phase, speakers form the sen-
tences, phrases, and words they have planned.

The processes of planning and execution are carried out
in the following specific steps. First, the speaker formu-
lates a discourse plan by identifying the type of discourse
situation he is in and retrieving an appropriate schema.
Each discourse situation is a culturally recognized situation
which has rules specifying the sequence and content of
discourse. Speakers, through their experience with dis-
course situations have schema; or mental representations, for

different situations. After determining if he is in

13
a situation calling for an argument, a story, etc., the
speaker plans the specific utterances within a discourse
structure so that they are congruent with and contribute to
his intention.

In the next step, the speaker plans individual sen-
tences. This step requires deciding on the speech act to be
used, the thematic structure (the information that is given
or known to the listener and new) and the propositional
content. It also requires decisions about stylistic aspects
of the utterance, for example, whether it is to communicate
a message directly or indirectly.

Once the speaker has formed a general representation of
the type of sentence he will utter, he plans specific con-
stituents. A constituent is a segment of a sentence
expressing a single idea or proposition. Constituent
planning involves selecting the words or phrases the writer
wishes to put in the constituent and ordering them. After
planning the constituent, the speaker constructs the arti-
culatory program for the elements that he will utter in that
constituent. He generates and holds in working memory the
phonetic segments, stresses, and intonation he will use.
Lastly, the articulatory program is carried out according
to directions for its execution, which specify how the
muscles involved in articulation are to be used, the se-
quencing of action, and the timing of movements.

This outline specifies the process by which a speaker
constructs a discourse plan. ‘However, it does not describe

the speaking process or how execution occurs. Execution

14

seems to involve some of the same units that planning does
(Clark and Clark, 1978). In the execution phase, speakers
begin with the constituent plan. Once the meaning of the
constituent is decided, the speaker selects a syntactic
structure, to convey his meaning which specifies the major
slots. Following that, the speaker chooses the content
words (nouns and verbs etc.) which belong in each slot.
After selecting content words, the speaker plans the func-
tion words, such as articles, conjunctions, and prepositions,
prefixes and suffixes which will be part of the constituent.
The final step is the selection of phonetic segments of the
constituent syllable by syllable.

This model of speaking is similar to models of the
composing process proposed by Flower and Hayes (1980) and
Bereiter (1979). Based on their analysis of the protocols of
writers in the process of composing aloud, FloWer and Hayes
(1980) have constructed the following cognitive process
model of writing. According to this model three major ele-
ments are involved in any composing activity: the task
environment, the writer's knowledge and the composing process
itself.

The task environment includes two aspects: the rhetor-
ical problem and the task environment. The writer's knowl-
edge is the information stored in the long term memory which
is relevant to the writing task, including the writer's
knowledge of the audience, the topic, and plans for writing.
Cognitive processes of writing include planning, translating,

and reviewing.

15

The rhetorical problem is the communicative task faced
by the writer. Elements of the problem are the communica-
tive intention, the topic, and the audience's needs. Flower
and Hayes note that the rhetorical problem is "an immutable-
part of the writing process, but the way in which people
choose to represent this problem to themselves can vary
greatly from writer to writer" (Flower and Hayes, 1980,

p. 8). Less able writers may restructure writing problems,
eliminating some of their features, or altering them to re-
duce the cognitive demands they place on the writer. Able
writers on the other hand are characterized by an ability to
handle these demands and to rethink the task so that it con-
tains new features, such as the writer's own goals.

The written text is 'UHa second component of the task
environment. The topic sentence, previous sentence and word
choices act as constraints on the writer's choices of suc-
ceeding sentences and words. Again the effect of the
”growing text" varies from writer to writer. Inexperienced
writers may fail ”to consolidate new ideas with earlier
statements" (Flower and Hayes, 1980, p. 8). Able writers are
able to plan at the inter - idea and composition level.

The writer's knowledge includes knowledge about the
audience, subject area, and his representations of rhetori-
cal problems and written discourse plans. Faced with a
rhetorical problem, an experienced writer may be able to
recall relevant problems, discourse plans and content know-
ledge and to adapt this knowledge to the rhetorical problem.

On the other hand, inexperienced writers may fail to retrieve

16
relevant writing plans or to transform their knowledge to
fit the writing task.
I The final elements in a writing activity are the

composing processes. Three subprocesses comprise the act
of writing: planning, translating, and revision. The
planning process consists of several subprocesses. The
first is idea generation or the retrieval of content for a
composition from memory. The second, organization, involves
the reorganization of this content to fit the rhetorical
situation. Such reorganization may consist of establishing
new connections between elements, such as new subtopics or
"superordinate concepts". (Flower and Hayes, p. 10). Finally,
planning includes goal-setting or the creation of goals for
writing. The aims of writing are not only provided by the
rhetorical problem, but may be generated by the writer.
Goal-setting usually occurs prior to writing, but goals
also may emerge through associations to ideas produced
during the idea generation process.

Translating is the second major subprocess in composing.
It is the process of transforming the writing plan into
written English. Ideas are the concepts and relations Which
are organized in set or network form in long term memory.
These must be converted into text according to conventions
for forming letters, grapheme-phoneme correspondence, rules
of syntax and of discourse.

The third writing process is reviewing. Reviewing
processes are involved in editing text and in idea generation

and translating.

17

The writing processes are carried out according to
directions from an executive strategist or monitor. It
decides when the writer will initiate and discontinue
different writing processes, such as idea generation.

Bereiter (1979) has proposed a similar but heirarchi-
cally organized model of the writing process. Directing
the entire process is an executive scheme. This scheme
assesses the writing task and directs the writing process
according to the writer's goals and the demands of the
discourse situation.

At a lower level are genre schemes. These are the
writer's mental schemata for different types of discourse,
which are retrieved by the executive. The genre schemes
specify the following: specific intentions or speech acts
appropriate to that genre, and "a set of strategies for
carrying out these intentions" (Bereiter, 1979, p. 151).
These strategies contain content slots to be filled, search
procedures for retrieving this content from memory, and
tuning instructions for translating this content into
linguistic form. The tuning instructions specify surface
features of language to be used. e.g. style.

At the level next to the genre scheme is a "content
processor" which draws semantic material from memory and
organizes it according to instructions from the genre
scheme (Bereiter, 1979, p. 151). The product of the content
processor is the "gist", a unit like a proposition. At the

lowest level is the language processor which translates

18

propositions into written text, according to the tuning
instructions of the genre scheme.

While there are important differences in these models,
they all outline the writing process in terms of similar
subprocesses. First writing involves representation of the
task environment, including the writer's intention, audience,
discourse situation. The writer then constructs a plan,
using an appropriate discourse schema for that discourse
situation. The schema specifies a set of intentions
appropriate to the discourse situation. These intentions
specify categories of content for each sentence, and in-
struction for the language (style, tone, etc.) to be used.
The writer retrieves and organizes appropriate content and
then translates the content into written prose according to
the instructions for language use.

While these models do not describe execution, models of
the speaking process would suggest that in execution the
writer operates on the plan for discourse one constituent at
a time. The writer presumably has in mind the meaning of
the constituent and then forms a syntactic outline of the
elements to convey that meaning. The semantic features
would then be assigned to each slot in the outline. At
this point the writer would have in memory a structure with
semantic and syntactic features. Using these features the
writer would then retrieve words with these grammatical and
semantic features from his mental lexicon. Finally, a
program specifying the motor responses necessary for writing
the constituent element by element would be formed and

carried out.

19
The models agree that the final process in writing is
revision. During revision the writer reads and compares the

text to the writing plan.
Bereiter's Model of Writing Development

Bereiter (1979) has proposed a model of writing develop-
ment based on models of the composing process and information
processing views of cognitive development, particularly
Pascual-Leone's (1970) theory of cognitive development.
According to this theory, cognitive development occurs due
to an increase in the number of mental schemes that can be
kept active in working memory. Scardamalia (in press)
argues that mature writing requires attention to a number of
skills, such as "handwriting, spelling, punctuation, word
choice, syntax, textual connections, purpose, organization,
clarity, rhythm. . ." Attention to these different skills is
possible because of automatization of lower-level processes
and increasing efficiency in alternation between processes.

Bereiter categorized the skills involved in mature
writing into six skill systems. These are: writing fluency,
that is, mastery over the skills of letter and word forma-
tion: ideational fluency, (skill at retrieving and organizing
ideas): knowledge of conventions of writing, such as hand-
writing, spelling, and genre conventions: social cognition:
literary evaluation skills: and reflective thought.

In the model, development is assumed to proceed in
stages, which integrate lower level skills first followed by

higher level skills. Lower order skills are those which are

 

20

necessary for a written product of any quality to emerge.
They include idea production skills and letter and word form-
ation skills. Higher order skills are those necessary for
producing text adapted to the constraints of writing assign-
ments. They include knowledge of rules for writing, literary
evaluation skills, social cognition and reflective thinking
skills. Children integrate "what skills they can in order to
achieve a functional skill of a higher order"-(Bereiter,
1979. p- 155)-

The following are the stages in writing development:

1. Associative Writing: Associative writing is a type

 

of writing involving the integration of ideational fluency
and writing fluency. At this stage the writer can produce
and translate ideas in a process of "controlled association"
(Bereiter, 1979, p. 156). During this stage writing is under
the control of the stimulus. Bereiter describes this as
writing that is "data-driven" rather than conceptually
driven. The writer simply writes down thoughts which occur
to him as they occur and stops when he runs out of ideas. If
the topic is familiar and can be assimilated to a network in
semantic memory, a coherent, complete text can result.

Support for the existence of this phase comes from
Flower (1979) who has described the products of immature
writers and from Britton (1975) who found that expressive
writing, centered around the discovery of ideas, was the
first type of rhetoric mastered by students.

2. Performative Writing: In the associative stage, the

writer's attention is focused on the processes of idea

21
generation and producing written letters and words. In the
performative stage of writing development, the writer can
also attend to the written product. As a result, he can
begin applying mechanical and stylistic conventions for
written English to his text. These include rules for
correct spelling, grammar, sentence formation etc.

3. Communicative Writing: Communicative writing is
the product of the integration of performative writing with
social cognition. While knowledge of the recipient has been
found to shape the oral messages of children as young as
four years of age (Shatz and Gelman, 1973) in writing tasks,
the requirement to produce text of some length which con-
forms to rules for written discourse must be accomplished
first. Writing situations also are more difficult with,
respect to social cognition because communication occurs in
the absence of a reader, and a shared context.' As a result,
many writers never achieve the communicative writing stage.

A. Unified Writing: In the communicative writing
stage the writer takes account of the process, product and
reader. In the unified writing stage, consideration of the
reader's perspective is integrated with knowledge about
literary quality. The writer produces text using his per-
spective as a critical reader. The writing process now
involves a conscious effort to produce composition that is
both an aesthetically pleasing object and a document reflec-
ting the writer's personal style.

5. _Epistemic Writing; Epistemic writing involves the

integration of reflective thought with unified writing skills.

22
Written language differs from Spoken, because it allows for
the external storage, review, and alteration of information.
In combination with formal thinking skills, it is ideal for
prose which involves the development and analysis of ideas.
Theoretical writing is prose of this type.

In summary, Bereiter proposes a heirarchical model of
writing stages. This model is based on information pro-
cessing theiries of cognitive development, such as Pascual-
Leone's. It assumes that writing skills are schemes which
take up space in working memory. Writing development begins
with the integration of lower order skills to achieve at
most a functional performance and proceeds through stages
involving the integration of skills which allow for adap-
tation to constraints of the writing task, such as the
teacher, the audience, the writer's critical knowledge, and
the writer's thoughts. With practiCe, lower order skills
require less attention, which can be devoted to mastering
higher level skills. Bereiter proposes that six skills are
part of the repertoire of able writers. These include
mastery of writing, skills at generating ideas, knowledge of
conventions of written language, audience awareness, literacy
evaluation, and reflective thinking skills.

The first stage of development, the associative
stage, is oriented around the most basic demands of writing
tasks (getting content on paper). It combines the skill of
producing ideas with skill at forming letters and words.

In the next stage, performative writing, the writer

begins to adapt to the demand for correct prose. In this

23
stage, the writer can employ conventions such as penmanship,
etc. This stage involves attention to the text as well as to
the writing process.

Communicative writing occurs when the writer is able to
combine performative writing skills with awareness of audience
needs. Even though very young children can alter a message
based on a recipient's needs, writers typically do not inte—
grate this skill into their writing behavior until they have
mastered skills for content and text production.

Unified writing results when skill at considering the
view of the reader is integrated with the writer's knowledge
of his response as a reader. With this orientation, the
writer can begin producing a written product that is pleasing
to him as a reader, using his critical knowledge of litera-
ture. The formation of a feedback loop between reading and
writing allows the development of an individual style.

Research has not been conducted to test the model of
writing development. Some studies have investigated the
development of features which are related to the model. In
the following section, an attempt is made to review these
studies as well as theory and research from which measures
for testing the model can be derived.

"Process—Oriented" Research on
Writing Development

Eyidence Related to’
Associative Writing
"Writer-based" or associative prose has been investi-

gated by Flower (1979). According to her, writer-based

2h
prose has a unique structure and form._ Whereas reader-based
prose is typically organized around key concepts, a structure
suited to the reader, writer-based prose "reflects the assoc-
iated narrative path of a writer's own confrontations with the
subject" (Flower, 1979, p. 2h). Writer-based prose can fol-
low two formats. The first, the narrative format, contains
ideas or propositions linked by temporal connections. This
format reflects the structure of episodic memory, a primitive
memory for specific events. Use of this format simplifies
writing, because it allows the writer to present ideas in the
order in which they are discovered or to embed logical rela-
tions, such as causal relations, in the events which pro-
’duced them.

Writer-based prose may also follow a survey format. In
this type of writing, the writer uses the internal structure
of the writing stimulus, such as a picture or book to organ-
ize his prose. Thus, his composition can be a spatially
organized description, if the task is to describe his apart-
ment, or a chapter by chapter account of a book for a book
repdrt. Use of a survey format allows the writer to perform
"a memory dump" of information, eliminating the necessity for
reconceptualization of facts. It also removes the writer's
need to generate a communicative purpose - to adapt to the
communicative demands of the situation.

The form of writer-based prose also has unique charac-
teristics. Preoccupation with the idea production process
leads to an inability in the Writer to shift attention from
the proposition immediately under construction. The result

is prose which is similar to egocentric speech (Vygotsky,

25

1962). It consists mainly of predicates, that is, actions or
relations. The subjects of propositions, while known to the
writer, are often implicit or are referred to by nouns and
pronouns which have meaning only for the writer. A written
text as a form of representation should express meaning
autonomously. Unlike a speaker, a writer cannot rely on the
external situation, gestures, intonations or other non-
linguistic cues to convey his meaning. In order to produce
an autonomous text, he must indicate the topic, and the
relationship between successive sentences. In addition,
written prose as a system of representation, is primarily
employed to express and develop "ideational" meaning rather
than "interpersonal meaning" (Halliday and Hason, 1979).
Writers express ideational meaning by establishing a theme
and using successive ideas to develop the theme.

The encoding of topic and relationships between ideas
occurs through the use of linguistic devices which allow
for the expression of different levels of meaning in an ut—
terance (Clark and Clark, 1978). Linguists have divided the
content of sentences into three parts: thematic structure.
propositional content and communicative intention. The
thematic structure is the aspect of a sentence which indi-
cates the topic and relates it to other parts of the dis-
course. The propositional content is the predicate and
noun(s) the speaker or writer wishes to convey in that
sentence. It also is referred to as the comment. The
communicative intention is the speaker or writer's purpose

in uttering a sentence (e.g., to command, declare, propose).

26
The distinction between topic and comment is similar to a
distinction by Clark and Haviland (197A) between given and
new information. Given information is the information in a
sentence which is known to a recipient. New information
refers to content newly conveyed in a sentence. For example,
in the following sentence, The one who bought my car was
gghg, the structure indicates that the information that some—
one bought a car and that the car belonged to the speaker is
the topic of the conversation and is known to the recipient.
The sentence introduces a comment or new proposition about
who bought the car.

There are several linguistic devices for marking topic
and/or given information and comment and/or new information.
Definite articles (e.g., the) indicate that a referent is
known to the recipient, while indefinite articles (e.g., a,
an) indicate that a referent is not known or able to be
identified. Pronouns are used to indicate known informa-
tion. Topic and comment can also be conveyed through the
use of structural devices. Mbst commonly the grammatical
subject of a sentence is used to refer to the topic, while
the comment is expressed by the predicate. For example, in
the following sentence: The church has a new roof, the given

information or topic is, The church, which is fronted. That

 

it has a new roof, is the comment or new informatidn, and it
is expressed in the verb phrase.

In associative prose, principles related to the writer's
cognitive processes, rather than principles for correct usage

may dictate the writer's use of topic/comment devices. As

2?
noted, beginning writers often fail to mention the topics of
sentences or use pronouns and definite articles inappropri-
ately. Principles of cognitive processing may also determine
their use of structure rather than topic/comment consider-
ations. Factors such as vividness, motivation of the speaker,
and naturalness have been found to influence speakers' choice
of subject (Osgood and Bock, 1974). The principle of vivid-
ness determines that objects high in activity and affective
meaning tend to be mentioned first in a sentence. Components
which are important to the speaker or writer also occur
early in a sentence. Finally, the principle of naturalness
states that the order of constituents in a sentence is de-
termined by the order in which they are attended to. Usually
individuals encode the agent of an action, followed by the
action and the object of an action, if attending to an action.
In the associative stage these principles rather than topic/
comment conventions may govern word order choices.

Other types of structural errors also may occur due to
the writer's attention to content production. One such error
is the failure to combine topic and comment into a single
clause, e.g., The man with the hat, he is my boss. A second
type of error, the sentence fragment, is usually a predicate,

which takes a preceding segment of text as its subject.

Summary

According to Flower, beginning writing has a distinct
structure and form. Its structure is either a narrative or
survey structure. In the narrative the writer represents

ideas in a format which follows the idea discovery process or

28
embeds perceived relations, such as logical or causal rela-
tions, in an event script. The survey form permits the writer
to represent information by using the structure of the sub-
ject itself. The form of writer-based prose also reflects
‘the idea generating process. Topics of sentences are often
omitted. Other errors in form characterize writer-based
prose. These include ambiguous pronouns, articles and nouns,
and poor sentence constructions.

Evidence Related to
Performative Writing

 

Bereiter believes that from the time they begin to write,
children's composing processes are similar to those of mature
writers. Writing development consists of the "gradual refine—
ment and elaboration of schemes at different levels in the
process" (Bereiter, 1979, p. 152). Once lower level schemes
are developed, schemes higher in the heirarchy of composing
processes can be perfected. One way to view the level of
schemes is in terms of their salience to the writer. Most
salient are probably the writer's long term memory and stra-
tegies for retrieving and organizing content. These schemes
are developed in the associative stage. Less salient is the
task environment. The text is probably the most salient
feature of the task environment. Schemes for perfecting it
are developed in the performative stage.

A second way to view the development of abstract schemes
is in terms of information processing and writing. Like
reading, writing is an act which makes use of several know-

ledge sources. In the information processing view of reading,

29

knowledge of letters, letter-sound correspondences, and word
segments are used to identify individual words. The reader
then uses his knowledge of syntax to parse the input into
constituents. Finally, the contents of semantic memory are
applied to assign meanings to sentences and to build a mean—
ingful representation. In writing, the writer must proceed
in the reverse direction. The associative stage of writing
development involves semantic memory and enough knowledge of
text characteristics necessary to translate propositions into
workable prose. After these elements are under control,
linguistic knowledge about the surface features of the text
is employed. Linguistic knowledge can range from rules
governing Whole texts, e.g., text grammar and governing inter-
sentence relationships to rules for sentences and words.

The purpose of this study was to examine the develop-
ment of the following conventions governing different levels
of text: discourse structure, cohesion, sentence correctness,

punctuation, and spelling.

Discourse structure in narrative. In models of the com—
posing process, it is hypothesized that after constructing a
representation of the discourse problem, writers retrieve an
appropriate discourse schema. This schema is a representation
of the content categories in a type of discourse and the re-
lations between them. Discourse schemata are defined as
generalized representations of types of discourse situations,
acquired through repeated experience.

Until recently, research on discourse processing focused

on the investigation of the role of discourse schemata in the

30

comprehension and retention of prose. Bartlett (1932) was
among the first investigators to study memory for prose. He
found that in recalling a story, subjects did not remember
verbatim the structural or semantic elements of a story but
recalled versions in which transformations, deletions, and
elaborations of the original material had occurred. The
transformations were not random, but corresponded to the sub~
jects' culturally acquired knowledge of cause and effect.
Based on his findings, Bartlett argued that subjects used
schemata to interpret discourse. A schema is an "active
organization of past reactions and experiences which are
always operating in any well-developed organism" (Bartlett,
1932, p. 20). Comprehension involves an interaction between
schemata and entering information; new material is assimilated
to schemata, while, at the same time, schemata are modified
to fit new experiences. Comprehension and memory for prose
are constructive processes during which subjects attempt to
form an overall representation of a story and to remember
single elements based on their expectations about the whole.

Recently, researchers (Rumelhart, 1975: Mandler and
thnson, 1977; Stein and Glenn, 1979) have developed gram-
mars of the mental structures used in story recall. Stein
and Glenn's grammar is representative of this work. Accord-
ing to it, the basic unit in a story is the proposition.
This unit contains a predicate: an action or some relation,
and one or more arguments. The propositions in a story are
of certain types. These categories of information or nodes

are differentiated according to their function in the story.

31
Stein and Glenn assume that the categories are organized in
a heirarchical network. Categories are related to each other
within the network by relations which express the degree of
causal influence between successive categories.

The two major nodes in a story are the setting and

episode system. The setting category introduces the main
character(s) and "describes the social, physical, or temporal
context within which the remainder of the story occurs"
(Stein and Glenn, 1979, p. 59).

The second major category of a story is the episode
system. The episode system can contain one or more episodes.
Each episode describes an entire behavioral sequence. An
episode contains the following parts: "the external or
internal events which influence a character, the character's
internal response (goals, cognitions, and plans) to the
events, his external response to his goals, and the con-
sequence resulting from his next responses" (Stein and
Glenn, 1979. p. 80).

The events which bring about a response in the protag-
onist are called the initiating events. They can be actions,
internal events, or natural events. An action by the pro-
tagonist or another character can initiate a response. For
example, the main character's action of cheating on a test
might cause him to have an internal response. An internal
event such as the perception of an object might lead to a
desire for the object and some action to obtain it. A

natural event is some "change of state" in the external

32
world, such as a storm which brings about a reaction in the
protagonist.

The initiating event is causally connected to the next
higher level category: the response category.) This category
contains the character's internal response to the initiating
event and a plan sequence. The internal response category
includes statements describing the protagonist's affective,
cognitive, or goal responses to the initiating events. Each
of these types of responses can occur in the internal response
category.

Theinternal response motivates the character's plan
sequence. The plan sequence is a higher level-category con-
taining the protagonist's internal plan and plan application.
The internal plan category contains statements describing
the character's cognitions about the events and his sub-
goals for achieving his goal.

The character's internal plans motivate his plan
application. The plan application category consists of two
categories: the attempt (the character's behavioral re-
sponses to obtain his goal) and the resolution (the conse-
quences of his attempt).

The resolution category contains the direct consequence
of the character's attempt to obtain.his goals and his
reaction to the consequence. The direct consequence cate-
gory contains statements which describe the character's
success or failure at reaching his goal, describe any change
in events brought about by his attempts, or serve to bring

about a reaction in the character. The last category,

33
the reaction category, contains statements describing the
affective, cognitive, or behavioral responses of the char—
acter to the consequences of his attempt to reach a goal.

As Stein and Glenn point out, very few stories corres—
pond completely to this model. Most stories contain more
than one episode. Stories also can contain more than one
setting, internal response or other category. Some elements
such as thesetting or reaction can occur in another location
in a story than the one specified in the grammar. Finally,
some elements of a story may be omitted; the internal
response and reaction often are omitted from stories because
this information is obvious from the initiating event and
direct consequence. The internal plan is also not included
in many stories. Stein and Glenn specify that in order to
be considered an episode, the episode system must at least
-contain an initiating event or an internal response which
initiates a "goal-directed behavior sequence", an action,
and a direct consequence.

Several studies have investigated the role of dis-
course schemata in children's comprehension and recall of
narrative (Stein and Glenn, 1979: Mandler and Johnson, 1977).
Stein and Glenn (1979) found that children of all ages used
story categories in recall. Stein and Glenn also found high
consistency across age groups in the categories recalled.
The categories recalled most frequently were the setting
and direct consequence. Initiating events and attempts were

recalled with slightly less frequency. Internal responses

34
and reactions were recalled least frequently. Stein and
Glenn conjecture that the high frequency of recall for the
setting category was due to its role as a primary cue in
story recall. They hypothesize that the initiating event
and direct consequence were recalled frequently because
they are central to the "logical progression” of an epi—
sode.

Some age differences were found in these studies.
Older children tended to recall more propositions and to
recall the internal response and reaction categories better
than younger children.

Stein and Glenn (1979) examined story structure in
stories produced by children. They found that the stories
of first and second graders frequently contained only
setting information. The stories of older children con—
tained episode systems, but these systems frequently lacked
information necessary for the logical progreSsion of events.
Stein and Glenn note that logical relations between elements
of a story often have to be generated after some idea
generation takes place.

Bereiter, Scardamalia, and Turkish (1980) investigated
children's conscious knowledge and use of categories of
content in stories, arguments, and directions. Children in
fourth and sixth grade were interviewed about types of
writing and later asked to produce a composition of each
type. Children were found to name an average of 3.35 of

the categories used in Stein and Glenn's grammar. The mean

35
number of categories used by children in their stories was
5.51.

Children used more categories in their stories than
they were able to name. In addition, they most frequently
named categories which were used frequently. However, there
was a slight negative correlation between the frequency with
which children named categories and the frequency with which
they employed them in their stories.

From these results, Bereiter et.al. concluded that
"children have a substantial fund of discourse grammar
knowledge. . ." but "that the effects of discourse grammar
knowledge are not mediated through conscious thought as they
evidently are, to some degree in skilled writers" (Bereiter,
Scardamalia, and Turkish, 1980, p. 2).

In the present study, children's narratives were ana-
lyzed to determine if children appeared to use discourse
schemata. The number of categories used by children of
different ages and ability groups was determined. Develop-
ment was measured by calculating the increase over time in
the number of story grammar categories used. In addition,
the types of categories used were identified. Finally, the
correlation between number of categories used and performance
on other measures was obtained.

Stein and Glenn argue that stories are understood through
a top-down process involving the use of schemata. During
comprehension.individuals parse incoming input into cateu
gories of content. In the last step of processing individual

propositions within each category are assigned a meaning.

36

Warren, Nicholas, and Trabasso (1979) propose, in
contrast, that a bottom-up process is used in story compre-
hension. They argue that in comprehension the understander
begins at one point in a narrative, the focal event, and
using world knowledge, attempts to construct an organized
semantic representation of the linguistic input.

For example, given the following story,

David left work.

He took the bus home.

He opened the door to his apartment.

He heard shouts from within.

He'd forgotten it was his birthday.
an individual would begin with the focal event: David's
hearing shouts from the inside. He would then attempt to
construct a meaningful representation of the text using
world knowledge, e.g., by recalling what might lead to
shouts from within an apartment, what he knows about sur—
prise parties, etc. .

The statements centered around the focal event in a
narrative are referred to as the event chain. An event
chain occurs with reference to a single character. Warren,
Nicholas, and Trabasso (1979) categorized the content of
event chains along three dimensions important in social
cognition: an objective-subjective dimension, an overt-
covert dimension, which involves information about the
behavior or appearance of an individual vs. psychological
information, and a dimension concerning intentionality of

behavior.

37
They also identified inferences used in story comprehension
to relate these types of content.

Seven types of propositions can occur in an event
chain. These include: 1) states, 2) events, 3) actions,
4) cognitions, 5) displays, 6) impulses, and 7) goals.

A gjaje is an objective condition of the physical
world or of a character. An gyen£_is an objective change of
state which may be caused by the protagonict's actions,
another character's actions, or natural causes.

An action is a voluntary external behavior of the main

character. A cognition is a voluntary mental behavior, such

 

as thought, perception, judgment, image, or plan.

Displays are involuntary external behaviors of the main
character. Such acts as taking flight, or reflex movements
would be displays. Impulses are involuntary internal exper-
iences, such as feelings, intuitions, beliefs.

IA gggi is involuntary or voluntary. It can be referred
to in the text as a need, desire, or plan.

These proposition types are "non-heirarchical descrip-
tors" of the types of information in stories. They are
connected by an understander into an event strr ture by
logical inferences.

I Logical inferences serve to fill in missing slots in
the surface structure of a text, or connect basic level
events in a higher level structure.) The types of logical
inferences are: physical causation inferences, motivation
inferences, psychological causation inferences, enablement

inferences, tempora1,and additive inferences.

38

Physical causation inferences specify the mechanical
causes for objective events and states. A physical causa-
tion inference would connect the following propositions:

1. He hit the ball (Action)

2. It struck the fence (Event)

Motivation inferences specify the causes for a charac-
ter's voluntary action, thoughts, or goals. The following
propositions would be connected by a motivation inference.

l. Harry's pet canary died (Event)

2. He decided to get a new canary (Goal)

Psychological causation inferences connect involuntary
displays, impulses, or goals to their causes. A psycholo-
gical causation inference would connect the following two
propositions.

1. Sarah saw the talking bird show (Action)

2. She laughed at the cockatoo's tricks (Display)

_Enablement inferences connect events, actions, and
displays to conditions which are necessary for them to
occur but are not causal. An enablement inference would
connect the following propositions.

1. George started the car.

2. He drove off down the highway.

States, events, actions, thoughts, displays and impulses
may occur spontaneously in a story or they may be caused by
other events.

Story production, like comprehension,night occur
through bottom-up processing. The writer may first generate

a focal event and then an event chain using information

39
logically related to the event. In this study, one question
was whether children used semantic knowledge rather than
text schema in their stories. It was thought that the use
of semantic relations rather than story schemata to orga-
nize prose might be reflected in an emphasis on the event
chain in a story, which might be apparent in greater use of
the focal or initiating event category than other categories.
Focal events or initiating events are those which produce
a response in the main character. In this study the pro—
portion of propositions in the focal or initiating events
category in children's stories was compared to the proportion

of propositions in other categories.

prer level organization in narrative » an alternative
yggy. An alternative to the view that semantic knowledge
dictates story production is the argument that even begin—
ning writers use text representation in their story writing.
Such a perspective would be supported by Stein and Glenn's
finding that young children's narrative products often
contain only setting information. If story knowledge is
composed of superordinatc and subordinate nodes which are
temporally organized, it may be that beginning writers
include predominantly setting information in their stories
because of the primary effect of the setting category in
memory for text structure. When writers begin to generate
propositional content within that category, the demands of
text and idea production may make it impossible for them to

return to subsequent categories, e.g., to shift back and

4O

forth between lower level content and the higher level
categories. Emphasis on setting information in children's
first stories would seem to be evidence that children are
using a text structure to produce stories, while emphasis on
focal events would seem to indicate that semantic knowledge
is being used to guide story production. For this disserta-
tion, the difference between the proportion of propositions
in the setting and initiating event categories was of

particular interest.

Cohesion. One skill required of writers is the ability
to produce organized prose that conveys meaning independent
of the context. Bereiter and Scardamalia (1980) point
out that integrating ideas in writing is more difficult
than in speech. In speech the order of ideas is determined
by conversational rules, such as rules for turn-taking.
Coherence between ideas is provided by contextual infor-
mation and by other speakers. However, in writing, a
writer must indicate explicitly the topic and the organi-
zation of ideas. To do this, writers make use of devices
for marking topic and comment.

There are several devices to indicate topic and the
relationship between ideas in text. As indicated, gram-
matical structure is one such device.

A second type of device is the cohesive tie. Halliday
and Hasan (1976) have introduced the concept of cohesion to

refer to the unity of a text and particularly to the lexical

1:1

and syntactic devices used to integrate ideas. A disagree-
ment exists currently about the composing process respon-
sible for providing cohesion (Hidi, 1980). Flower and Hayes
(1979) maintain that the construction of cohesive prose
occurs during the translation process, while Bereiter assigns
this construction to the content processor.
The basic concept in the study of cohesion is the gig.
A tie refers to the device used to express a relation and
the elements presupposed by it. Some ties can be gygphggig,
relating an element in the external situation to an item in
the text, or endophoric. relating two or more items in a
text.
There are several types of cohesion. They differ in
the linguistic elements through which semantic relations are
conveyed. The types are:
Reference: Reference is.a type of cohesion in which
the meaning of an item can only be interpreted through
reference to another item. Types of referential ties are:
pronominal: e.g., There are several poor readers
in the school, but I don't have any of fihgm.
demonstrative: e.g., There are many career oppor-
tunities for nurses. Thggg are the ones you should
consider.
comparative: e.g., A taller man.

These types of reference can be exophoric or endophoric.

Substitution: A substitution tie is one which relates

two items with a similar grammatical function. The different

42
types of substitution are nominal, verbal, and clausal. The
following is a complete list of the items that occur as
substitutions:
nominal: one, ones, same
May so
clausal: so, not

Ellipsis: Ellipsis is, according to Halliday, "sub-
stitution by zero" (Halliday, 1979, p. 142). Ellipsis exists
when a sentence or clause contains empty slots which must
be interpreted with reference to some presupposed item in
a similar grammatical slot. An example of ellipsis is:

Do you like driving? Ndsmuch.

Lexical: Lexical cohesion involves the use of simi-
larity in word meaning to relate items. Lexical ties are of
several types:

repetition: Lexical repetition involves repetition
of a lexical item, e.g., The volcano erupted
killing several people. It is not expected to
erupt again.

synonym: A synonym is an item with a similar
meaning to the presupposed item, e.g., The average
pay for waitresses is very low. Most waitresses
need a higher wage.

general noun: A general noun is a word such as
people, place, animal, thing used to substitute
for a presupposed item, e.g., Our dog bit the
mailman again. ‘I don't know what to do with that

animal.

43
lexical superordinate: Use of a lexical super-
ordinate involves use of a category label to sub—
stitute for a member of a category, e.g., She is a
good principal. The other administrators should
follow her example.
collocation: This type of lexical cohesion is
created through use of items which are usually
associated with each other, such as boy, girl, e.g.,
The food there isn't very good: neither are the
drinks.

Conjunction: Conjunction is different from other types
of cohesion because it links clauses, sentences, or larger
sections of text. Ties which express conjunction relations
' include:
additive ties: Egg, pp, in addition, e.g.,
Yesterday I heard those students talking. And
they said the exam was too easy.

adversative ties: but, however, instead, yet, e.g.,

 

We wanted to go to the movie. Instead we ended up
at the stock car races.
causali if, consequently, pp, pgcause, therefore,
e.g., He provided her with jewelry and furs, so
she died a wealthy woman.
temporal: then, after, Eggp, e.g., We plan to see
Rome first. Then, we'll travel to Switzerland.

In summary, unity within a text is provided in part by

linguistic elements which express relations between items.

Some cohesive ties can be exophoric, linking textual elements

#h

to elements in the situation, or endophoric. linking textual
elements. The types of cohesive relations include: refer—
ence, substitution, ellipsis, lexical cohesion, and con-
junction.

Bracewell (1980) investigated the development of cohe~
sion in children's writing. Fourth, eighth, and twelfth
grade students participated in an experiment in which they
were asked to write an opinion essay. Students' essays were
analyzed using Halliday and Hasan's scheme for analyzing
cohesion. The proportion of cohesive ties used by writers
at all ages was equal: approximately .40 for all students.
Some grade level differences were found in the types of
cohesive ties used. Older students used a significantly
higher proportion of lexical superordinates and causal con-
junctions than younger students. Older students used a
higher proportion of endophoric ties than younger students,
and there was an increase with age in proportion of inter-
t-unit as opposed to intra-t—unit cohesive ties. Bracewell
hypothesized, based on Bereiter's model, that difficulty in
producing a highly coherent text was due to an inability to
attend to surface features of text independent of the
semantic content of the composition. In a second study,
Bracewell (1980) attempted to examine the effect of content
concerns on writing by manipulating the degree of subjects'
exposure to content. He exposed children to different
amounts of the same content and asked them to write about it.

He found that children's inclusion of surface features of text

#5
was inversely related to their familiarity with the content
of the composition.

A third study by Bracewell and Scardamalia (1980) exa-
mined the potentially interfering effects of the linear form
of text on children's ability to express relations between-
semantic elements. In this study students were presented
with several relatable propositions in matrix or in sentence
form. They were asked to write an integrated sentence
using these propositions. Students in the matrix condition
were able to produce more highly integrated text than
students in the sentence condition.

In the present study, the cohesion of children's
writing at different ages was assessed. Bracewell's
findings suggest that the number of ties used by writers
does not indicate attention to surface features of text:
however, the use of certain types of cohesion, a decreasing
use of exophoric ties, and the expression of inter-sentence
relations may be indicators of a deliberate effort to pro-
duce a readable text. . I.

Some cohesive devices may express more complex relations
than others and so reflect greater manipulation of ideas.
Others may demand a more conscious control over language.
Referential pronouns simply express an identity relation
between two words. Additive conjunctions, such as Egg, can
be used to express a relation between elements which were
linked together in the idea discovery process, or can express

a true co—ordination relationship between elements.

#6
Conjunctions such as adversative or causal conjunctions
integrate two or more propositions into a higher level
proposition, and so are more complex. For example, the two
propositions, "Andrew has a cold," "he is able to go on the
field trip," when connected by an adversative conjunction
"but" express a single idea, a qualified participation for
Andrew. Lexical cohesion may involve more deliberate use of
language than other types,because it occurs through vocab-
ulary selection.

The use of endophoric reference rather than text-
external reference indicates adaptation to constraints of
writing rather than the use of a conversational strategy.
The degree to which endophoric ties are used indicates the
writer's ability to respond to these constraints and produce
text which conveys meaning independent of content. Finally,
the use of inter-sentence cohesive ties may be more indi-
cative of attention to the surface text than intra—sentence
cohesion. Whereas intna-sentence cohesion can occur simply
through operations on the proposition or clause currently
being constructed, interusentence cohesion may necessitate
reading and review of the text because of the limits of
working memory.

The development of cohesion in the compositions of
fourth, fifth,and sixth grade children was investigated. The
frequency with which cohesive ties and lexical ties were
used was determined. The proportion of endophoric reference

and inter-sentence ties was calculated. Finally, the

1&7
relationship between these and other variables was deter-
mined. It was expected that increases would occur in the
use of each type of tie and that the use of cohesion would

be positively related to the use of other conventions.

Mechanics. In the performative stage, writers pre-
sumably master conventions of mechanics, such as spelling,
punctuation, and sentence usage (Bereiter, 1979). For this
study, the maturity of children's spelling, punctuation, and
sentence usage was assessed. Each composition was assigned
a rating of either poor, minimally adequate, good, or
exCellent on spelling and punctuation. The use of sentence
conventions was assessed by determining the ratio of correct
sentences to all sentence-type forms, including sentence
fragments and run-on sentences.

The Pattern of
Development in Writing

Bereiter has noted that there is insufficient data at
present to allow for the characterization of writing devel—
opment as continuous or discontinuous. Development from
one level to another could be gradual: as writers acquire
mastery over some skills, they could simply integrate new
ones into their repertoires. Alternatively, writing devel-
opment could occur in stages. Bereiter argues that this
pattern would occur if "the writing process has organi-
zation and . . . the incorporation of a new skill requires
reorganization of the process" (Bereiter, 1979, p. 154).

One question of the present study was whether growth on

each feature was linear or discontinuous.

CHAPTER III

METHOD

Subjects

Subjects of the study were fifty-four fourth, fifth,
and sixth grade students. The grade levels were selected
because important growth in writing achievement has been
found to occur during these years (NAEP, 1974: Stahl, 1977).

The subjects were students of teachers who participated
in a study of writing instruction conducted by members of
the Language Arts Project of The Institute for Research
on Teaching.

The fourth graders were drawn from a classroom in
Okemos, Michigan. The members of this community are pre—
dominately white, upper-middle class professionals. -The
fifth graders were from Dimondale, Stockbridge, and Lansing.
Dimondale and Stockbridge are rural villages near Lansing,
where residents are also primarily white but middle to
lower-middle class. Lansing is a racially mixed urban
community with a mix of social classes. tudents in the
study were from a school in a lower—middle class neighbor-
hood. The sixth graders were also from this school.

The Woodcock Reading Mastery Test was administered to

all students as part of the Language Arts Project at the

#8

#9
beginning of the study. A reading grade total score was
obtained by each student. Students were classified into
below average, average, or above average reading ability
groups using their reading grade total scores. Those
students who received a grade level score of 4.5 or below
were classified as low reading ability students. Those who
obtained grade level scores of between #.5 and 5.5 were
categorized as average, and those who obtained scores above
5.5 as above average in reading ability.

Seven students in the fifth and seven in the sixth
grade were randomly selected from each ability group. Only
one fourth grade classroom, with eighteen students partici—
pating was involved in the study. Seven of the students were
of below average ability, five of average, and four of
above average ability. The number of fourth grade students
made it impossible to use a balanced design. However, since
proportional designs can be used without violating assump-
tions of independence a proportional design was employed.
Four students from each ability group were selected from the

fourth grade.
Instruments

The Woodcock Reading Mastery Test was employed to test
reading ability. The Woodcock contains subtests of the
following reading skills: letter identification, word
identification, word attack, word comprehension, and passage
comprehension. Students' subscores on these subtests are

combined to yield a total reading grade score. Test-retest

50

and split-half reliabilities of the Woodcock have been
obtained. Test—retest reliability was .95 and split—half
reliability was .97.

The narrative compositions were produced in response
to three writing tasks developed by Dr. Laura Roehler
and other members of the Language Arts Project. In the
first task, students listened to a tape recording of the
story of Hansel and Gretel and then were asked to retell
the story from the witch's point of view. In the second
task, students were shown a photograph of a girl who had
fallen in a field and asked to tell a story about it. The
third writing assignment was to write a story about a
painting of a man sitting on a branch of a tree.

Investigators have demonstrated that the discourse
situation and content knowledge affect writing. San Jose
(1972) and Perron (1977) investigated the effect of mode
of discourse (narrative, exposition, argument, descrip-
tion) on syntactic complexity in children's writing. They
found that the degree of sentence complexity was greatest
in argument then exposition, narrative, and description.
Scardamalia et. a1. (1980) found that familiarity of
content affected the production of ideas in children's
writing.

An effort was made to design tasks which would not
require particular content knowledge. In addition, for

this dissertation, development in only one mode of discourse

51
was assessed. However, in this study the stimuli for
writing tasks were different. Confounding could have
existed between the time of year and measure. The instru—
ments used to analyze writing were the following:

Narrative Structure
Instruments

 

Stein and Glenn's story grammar was used to classify
the categories of propositions which occurred in child-
ren's stories., The categories of information in a story
according to their grammar are:

Setting: The setting provides information about

the main character and the spatial, temporal,
and physical context of the story. Setting
information about the main character is referred
to as majorsetting information. Context infor-
mation is minor setting information. '

Initiating_Event: The initiating event initiates

the episodic structure of a story. It can be

a natural event, action or internal event in the
main character. It produces a response in the
main character. The response is directed to
altering the change of state produced by the
initiating event.

Internal Response: The internal response is the

character's psychological response to the initiating

event. It can be an affective or cognitive response

52
or goal. The internal response motivates the
main character to form a plan.

Internal Plan: The internal plan consists
of the character's cognitions and sub—
goals about how to produce the desired
change motivated by the internal response.

Attempt: The attempt consists of statements

' about the character's behavioral attempts
to produce a desired state.

Direct Consequence: The direct consequence
category contains information about the
resolution of the character's attempts.
It describes some event, end state, or
action which relates to the character's
goals and functions to produce a reaction
in the character.

Reaction: The reaction category contains
statements about the character's response
to the direct consequence and his
achievement or non—achievement of his
goal. Reactions can include affective
or cognitive responses and actions
which occur as the result of the
main character's implicit or explicit

internal response.

53

The measures of structure were designed to assess both
lower level and text structures in prose.

Warren, Nicholas, and Trabasso's scheme-was used to
categorize the type of semantic elements and relations in
narrative compositions. The propositions in children's
stories were categorized as:

States: A state is an objective physical condition in
the world or in a character in the story.

Events: An event is an objective change of state which
occurs either because of natural causes or another charac-
ter's actions;

Actions: An action is a voluntary external behavior
performed by the protagonist of an event chain.

Cognitions: A cognition is a voluntary internal beha-

 

vior by the protagonist. Cognitions include thoughts, plans,
judgments, decisions.

Displays: A display is an involuntary external be-
havior performed by the protagonist.

Impulses: An impulse is an involuntary internal
behavior of the protagonist. Impulses include feelings,
intuitions, dreams, beliefs.

gpglg: .A goal is an involuntary or voluntary desire,
need, or plan of the protagonist.

These propositions may be related by motivation,
physical causation, psychological causation, er enablement
connections to form the logical structure of a story. The

types of logical connections are:

5h

Motivation connections: Motivation connections exist
between a voluntary action or cognition and the events which
cause it.

Physical causation connections: Physical causation
connections exist between an objective event or state or an
external behavior and an event which causes it.

Psychological causation connections: A psychological
causation connection is a relation between an involuntary
action or internal event and the event which causes it.

Enablement connections: An enablement connection links
an event to a state of event which is a precondition for

it, but does not cause it.

Cohesion Instrument

 

Cohesive ties were categorized according to the fol-
lowing categories defined by Halliday and Hasan.

References: Referential ties allow interpretation of
one item through reference to another item. There are three
types:

pronominal — pronouns, e.g., she, he, it
demonstrative - demonstrative pronouns, e.g., these,
these, this, that

comparative — comparative terms, e.g., better,

 

worse, best
Substitution: Substitution ties relate two items which
share a grammatical function. The types of substitution and
items that occur as substitutes are:
nominal — one, ones, same

.verbal - do (including tenses)

55
clausal - so, not

Ellipsis: Elliptical ties are emissions from the
surface structure which can be interpreted with reference
to an item in the same grammatical slot, e.g., Do you like
games? Some.

Lexical ties: Lexical ties relate elements through
similarity of meaning. The types of lexical cohesion and
the way they are achieved are:

repetition - use of repetition of a presupposed

 

item

synonym - use of a word with similar.meaning to

another word

general noun - use of a general noun, e.g., people,

place - to refer to a specific nounv

superordinate - use of a category label to refer to

a member of the category

collocation - use of words that are associated,

e.g., colors

Conjunction: Conjunctions relate clauses, sentences,

and larger units of text. The types of conjunction are:

additive Q e.g., and, er, in addition

adversative - e.g., but, however, instead, yet

causal - e.g., if, so, because, consequently,

therefore

temporal - e.g., then, after, next

56

Sentence Correctness

 

Errors in sentence formation were assessed by classi-
fying the surface structure used by students into the
following categories:

Sentence fragments: A sentence fragment was defined
as any form punctuated as a sentence which lacked a subject,
main verb or both.

Run-on sentences: A run-on sentence was defined as a
form, punctuated as a sentence, which contained more than
two independent clauses joined either by the co—ordinate
conjunction "and" or run together without a co—ordinating
conjunction.

Correct sentences: A correct sentence was any form
constructed according to rules for correct sentence forma—
tion. Sentences which contained two or fewer independent
clauses and any embedded clauses were considered correct,
even if they lacked capitalization or correct punctuation.
Mechanics:

Punctuation Maturity

Performance on mechanics was assessed using rating
scales developed by members of the Language Arts Project.
Compositions were assigned one of the four following ratings
on punctuation:

ngp: A composition with poor punctuation had many
punctuation errors making the composition difficult to read.
The errors could be omissions or inappropriately used

punctuation marks.

57

Minimally adequate: A minimally adequate rating was
assigned if a composition contained some errors but was
readable, and correct usage involved mainly simple forms,
such as capitalization.

Qpr: A composition with good punctuation had few
punctuation errors and contained a variety of types of
punctuation. However, it did not have the quality of adult
writing.

Excellent: Excellent compositions contained no punc-
tuation errors and a variety of punctuation forms like those
found in average adult writing.

A score of one was given for a poor rating and a score
of four for an excellent rating.

Mechanics:
Spelling Maturit

 

Spelling maturity was determined by using two scores:
a spelling accuracy rating and a combined spelling score,
which was the product of the spelling accuracy rating and
a rating of the difficultyof'word selection. These ratings
were developed by Language Arts Project members. The
difficulty of word choice rating was included because it
was felt that some students might achieve a high degree of
accuracy by avoiding difficult words. These dimensions were
assessed using two four point scales. The spelling accuracy
scale was as follows: . _

299;: A poor composition contained many errors making

the composition difficult to read.

58

Minimally adequate: A minimally adequate rating was
assigned to readable compositions with some errors.

Qgpg: A good rating was used for compositions with
very few misspellings.

Excellent: An excellent composition contained no
errors and resembled average adult writing.

The level of difficulty of students' vocabulary was
assessed using the following ratings:

Pppp: A poor rating was assigned to short selections
with a very simple vocabulary.

Minimally adequate: A minimally adequate composition
was one of average length with a simple vocabulary.

qug: A good composition was one of average length
which included some difficult words, but which did not
approach an average adult composition.

Excellent: An excellent rating was assigned to compo-
sitions which contained the type of vocabulary found in

average adult writing.
Data Collection

The Woodcock Reading Mastery Test was administered to
students at the beginning of the study by a member of the
project. The writing assignments were given at three times
during the year at equal intervals in November, February, and
may. The first two assignments were presented by the teacher
using a set of instructions developed by Language Arts
Projects members. A member of the project was present to

assist the teacher and to ensure that the instructions were

59
followed. The third task was presented by a project member.

Students were allowed one hour to complete each assignment.

Data Preparation

 

Copies of the students' original compositions were
used for the cohesion, Spelling, punctuation and sentence
coding. If a copy was unreadable as was the case for a few
compositions, coders were given a typed version. Only about
eight compositions were unreadable.

For the text structure and event chain coding, stu—
dents' compositions were divided into propositions and
statements by the author. Warren et. al.'s definition of
a proposition as any verbal unit plus one or more nouns
which is a basic element of a story was employed. Micro-
analysis of the text such as Kintsch (197A) has employed
was avoided. Warren et. a1. believe that this micro-
analysis is unnecessary.

Some propositions seemed to be parts of higher level
units which constituted more logical units of analysis.
Propositions in these units were grouped together for ana-
lysis, and the units were referred to as statements. Two
types of statements occurred in stories: quotations, which
contained propositions uttered by a character of the story,
as well as propoSitions referring to the speaker and the
act of speaking, and statements whiCh contained a new prop-
osition and a repetition of a previously stated proposition.

An example of a quotation in a story would be as follows:

60

1. Cynthia cried,

2. "Help me

3. I'm stuck"
In this case, the unit of meaning in the story seems to be
Cynthia's action of crying for help. For this story the
three propositions would have been grouped together to form
one statement.

An example of a statement which contained a repetition
of a previously mentioned proposition would be as follows:

1. Mary's friends came running.

2. They grabbed her and

3. she climbed out of the mud.

h. After she climbed out

5. she ran home.
In this case, the first proposition referring to mary's
climbing out of the mud would seem to be a meaningful element
of the story and the second a repetition designed to achieve
coherence, which logically belongs with the proposition which
follows it. For coding, these propositions were joined and
the higher level unit coded.

The propositions in students' stories were typed and
numbered. Propositions which were joined together in state—
ments were bracketed. Coders were then given a copy of the

typed version.
Data Reduction and Analysis

Students' compositions were coded for the features of

text structure, cohesion, sentence correctness, punctuation,

61

and spelling. The coding procedure was as follows:

The author analyzed students' compositions for sentence
correctness since sentence completeness is an objective,
easily observed dimension. The punctuation and spelling
assessment in the fall and spring was conducted by Language
Arts Project members. Punctuation and spelling were two
dimensions of a holistic assessment of students' composi-
tions carried out by the Language Arts Project. The raters
were faculty and graduate students at Michigan State
University. They were trained for a period of three
weeks. At the end of this period, rater agreement was
adequate and Project members coded the remaining composi-
tions independently.

Since the holistic assessment was conducted only on
compositions produced in the fall and spring, two additional
raters were trained by the author to rate the winter comp-
ositions. The raters were a graduate and undergraduate
student at Michigan State. They received instructions on
the rating systems and practiced on four compositions at an
initial training session. They then discussed their eval-
uations with this author and received a review of the instruc—
tions. Ten compositions were given to the coders to be
double coded. The percent of agreement was calculated for
each feature. Fer punctuation, it was 62%, for spelling
accuracy, 70%, and for difficulty of word choice, it was
61%. Because of the low inter-rater agreement and the small

number of compositions remaining, the decision was made to

62

double code the remaining compositions. Disagreements in
ratings were resolved by the author.

For the analysis of text structure, event chain rep—
resentation and cohesion, the procedures were identical and
were as follows: Two coders were assigned to code each
feature. Five of the six were undergraduate students at
Michigan State University. One graduate student was employed
to code cohesive ties. None of the coders was familiar with
the study or the hypotheses being investigated.

I The coders were given instructions at an initial ses—
sion and two compositions for coding practice. The author
then discussed their coding with them and reviewed instruc-
tions. Twenty identical compositions were given to each
coder each week for two weeks. Coders met after each week
to discuss the completed coding. The percent of coder
agreement was determined at the end of the second week.

For the cohesion coding the percent of agreement was
determined for each tie. Agreement about each type or if
there were subcategories, for each subcategory, was deter-
mined. Agreement about endophoric reference and inter—
.sentence cohesion was also determined. Failure to count an
item as a tie and conflicts about the type of tie were
counted as disagreements. For the text structure and event
chain coding, the percent of agreement with respect to each
proposition were determined. For the event chain coding the
proportion of agreements for connections was also determined.

Percentage of agreement was used to capture inter—coder

agreement because the data were nominal. Percent of

63
agreement was the ratio of the number of agreements to the
number of agreements plus disagreements.

At the end of the second week of training the percent
of agreement for the cohesion coding was 81%. For the story
structure coding it was 72%, and for the event chain coding,
75%.

At the end of the two weeks coders were told that they
would be coding different compositions but that their
agreement would be checked at random. For the next seven
weeks, coders received twenty compositions to code. Every
seventh composition was double coded. The proportions of
agreement for the double coded compositions were as follows:
For cohesion, (.80, .83, .70, .79, .89, .86, .81, .86, .88,
.90, .76, .97, .90, .81): for story structure, (.74, .83,
.79, .70, .65, .83, .80, .48, .76, .81, .70, .75, .85, .77);
and for the event chain coding (.88, .65, .75, .74, .45, .84,
.88, .88, .94, .87, .77, .76, .93, .75). The coders
continued to meet every week with the author to discuss any
'problems they were having and to review the coding instruc-

tions.

_ Design

The design was a repeated measures design with two
independent factors with three levels each. The two indepen-
dent factors and levels were: grade (fourth, fifth, and
sixth) and reading ability (above average, average, and
below average). The dependent variables were the variables

of story structure, cohesion, sentence correctness,

64

punctuation, and spelling discussed in the hypothesis ques-
tion. They include:

1. the mean number of categories from Stein and Glenn's
story grammar

2. the mean number of content categories from Warren
et. al.'s event chain representation

3. the mean proportion of cohesive ties

figof cohesive ties
# of words

 

4. the mean proportion of lexical ties

#_of lexical ties
# of words

5. the mean proportion of endophoric referential ties

fiqof endgphoric referential ties
# of total referential ties

6. the mean proportion of between-sentence ties

#qof between-sentence ties
# of between and within—sentence ties

7. the mean proportion of correct sentences
#_of correct sentences
# of sentence fragments,
run-on, and correct sentences
8. the mean punctuation rating
9. the mean spelling rating
The dependent variables included the proportion of
propositions in each category of Stein and Glenn's story
grammar.
The following were the testable hypotheses of the study:
1. H1: there will be an increase over time in the
following features of children's stories: .
a. the mean number of categories from Stein and

Glenn's story grammar

65

b. the mean proportion of lexical ties

c. the mean proportion of cohesive ties

d. the mean proportion of endophoric referential

ties

e. the mean proportion of between—sentence ties

f. the mean proportion of complete sentences

g. the mean rating on punctuation

h. the mean rating on spelling accuracy

i. the mean combined spelling score
2. H1: There will be no increase over time in the
A number of content categories in children's stories.

3a. H1: There will be a significant difference between
the fourth, fifth, and sixth graders in each feature dis—
cussed in the first hypothesis.

3b. H1: There will be a significant difference between
the fourth, fifth, and sixth graders in the pattern of growth
on each feature. '

4. H1: There will be a significant difference between
Students of above average, average, and below average reading
ability in each feature discussed in the first hypothesis.

The pattern of growth on each variable and the relation-
ship between dependent variables was determined. Finally,
the proportion of propositions in the setting category as

opposed to the initiating event category was obtained.
Statistical Analysis

Four multivariate analyses of variance of repeated

measurements were employed to test the main hypotheses and

66

to determine the pattern of growth for each variable. Pearson
Product Moment Correlations between the dependent variables

were obtained to determine their relationships.

CHAPTER IV
RESULTS

The purpose of the analysis was to trace the develop-
ment in children's narratives of features associated with
the performative stage of writing development.

Four 3x3 multivariate analyses of repeated measures
were performed on the measures of story structure, cohesion,
sentence correctness, punctuation, and spelling. The factors
in the analysis were: Grade (Four, Five, and Six) and Reading
Ability Level (Above Average, Average, and Below Average).
Each measure was taken on three occasions. Polynomial
contrasts were performed to independently assess the linear,
and quadratic trends for the set of measures. If a multi-
variate analysis indicated a significant effect for any
factor, univariate analyses of factor effects on each measure
were performed. The major analysis was conducted on the
variables of: number of story grammar categories, proportion
of cohesive ties, proportion of correct sentences, punctua-
tion rating, and spelling accuracy rating. Included in the
other analyses were other cohesion variables and the combined
spelling score. The results will be discussed for each hypo-
thesis. Interactions will be discussed first. If no interac—
tions were found in the overall analyses making it possible to

consider main effects, these effects will be discussed.

67

68

Hypothesis #1: There will be a significant increase
over time in the mean number of story grammar categories,
the mean proportions of cohesive ties, lixical ties, endo-
phoric referential and between—sentence ties, the mean
proportion of correct sentences, the mean punctuation rating,
and the mean spelling rating.

The design over measure analysis in the first multi~
variate analysis of repeated measures revealed no signifi-
cant Occasion X Grade X Reading Ability interaction.

However, a significant Occasion X Grade interaction was
obtained for the linear component of the Grade effect,
F(12,34) = 4.22, p<.0005. Univariate analyses indicated
significant Occasion X Grade interactions for the linear

term of the story structure measure, F(l,45) = 14.57, p<20005,
the linear term of the cohesion measure, Flin(l,45) = 9.45,
p(.004, and the linear and quadratic term of the spelling
accuracy measure, Flin(l,45) = 5.91, p(.02: F d(1,45) =
5.19, p<.03.

The Occasion X Grade interactions in general seemed to

qua

be due to the fourth graders' showing the greatest improve-
_ ment on the greatest number of measures and to a decline in
performance by the sixth graders at Time 3.

The Occasion X Grade interaction for the number of
story grammar categories indicated that the number of story
grammar categories in fourth graders' writing increased over
the year. The number in fifth graders' stories remained
constant between Time 1 and Time 2 and decreased between

Time 2 and Time 3, and the number in sixth graders' stories

69
decreased between Time 1 and Time 3. Cell means for this
interaction are as follows: Fourth graders (4.50, 4.75,
5.00): Fifth graders (4.90, 4.90, 4.57): Sixth graders (5.71,
”~95: 3-57)-

The Occasion X Grade interaction for the proportion of
cohesive ties indicated that the proportion of cohesive ties
in fourth graders' writing decreased from Time 1 to Time 2,
but increased from Time 2 to Time 3. Cohesive ties in fifth
and sixth graders' writing decreased from Time 1 to Time 3,
The cell means for this interaction were: Fourth graders
(.37, .34, .38); Fifth graders (.37, .35, .34): Sixth
graders (.41, .36, .32):

Univariate analyses also indicated a significant
Occasion X Grade interaction for the spelling accuracy rating.
This interaction resulted from an improvement during the
year by the fourth and fifth graders' spelling accuracy. The
sixth graders' accuracy improved from Time 1 to Time 2, but
declined from Time 2 to Time 3. Cell means for the three
times are as follows: Fourth graders (2.00, 2.17, 2.91):
Fifth graders (1.71, 2.09, 2.14): Sixth graders (1.76, 2.19,
2.05).

The univariate tests revealed no significant Occasion X
Grade interaction for the variables of punctuation maturity
and the proportion of complete sentences. Therefore, it was
possible to examine the main effect for Occasion. A signifi-
cant main effect for Occasion was found in the overall
analysis, F(12,34) = 4.99,jp(.0002. The univariate tests

revealed a significant linear trend for the punctuation

70
variable, Flin(l,45) = 6.47, p<.01 and a significant
quadratic trend for the sentence correctness variable,
Fquad(l,45) = 5.60, p<.02. The students showed a steady
improvement over the year in punctuation. The means for the
three times for this variable were: 1.74, 1.92, and 1.98.
The proportion of correct sentences increased from Time 1 to
Time 2 but decreased from Time 2 to Time 3. The mean
proportions for each time were: .68, .75, and .68.

In order to test the hypotheses that the proportion of
lexical ties, endophoric referential ties and between-
sentence ties would increase, three additional multivariate
analyses of repeated measures were performed._ One analysis
was also employed to test effects for the combined spelling
score. The first additional multivariate analysis was
performed on the variables of lexical cohesion, the combined
spelling score, the proportion of correct sentences, the
story structure and content measures. Only effects for
the lexical cohesion and combined spelling score which were
not tested in the main multivariate analysis will be
discussed. The analyses revealed a significant Occasion X
Grade X Reading Ability interaction, F(48, 133) = 1.45,
p<.05. Univariate tests indicated that an Occasion X Grade
X Reading Ability interaction existed for the combined
spelling score. 'The interaction is complex and difficult
to interpret. The fourth graders at all ability levels, the
fifth graders of average and below average ability, and the
sixth graders of below average ability showed some improve-

ment in their combined spelling scores over the year. The

71
above average fifth graders and the average sixth graders
received lower scores in the spring than in the winter.
The above average students' scores dipped in the winter.

The second additional multivariate analysis was per-
formed on the measure of endophoric reference as well as the
story structure, sentence correctness, punctuation, and
spelling accuracy measures. Only effects for the endophoric
reference variable will be discussed. The overall analysis
revealed no significant three-way interaction. It did
indicate a significant Occasion X Grade interaction for the
linear term of the Grade effect, F(12,34) = 3.36, p<.003.
The univariate analyses revealed a significant Occasion X
Grade interaction for the quadratic term for endophoric
reference, Fquad(l,45) = 5.55, p<302. The Occasion X Grade
interaction resulted from the fact that the fourth graders'
use of endophoric reference remained constant from Time 1
to Time 2 and increased from Time 2 to Time 3, while the
fifth graders' use of endophoric reference increased from
Time 1 to Time 2 but remained constant from Time 2 to Time 3.
The sixth graders showed growth in the use of these ties from
Time 1 to Time 2 but their performance decreased from Time 2
to Time 3.. Means for these groups were as follows: Fourth
graders (.67, .68, .88), Fifth graders (.64, .73, .74),
Sixth graders (.59, .74, .60).

The final multivariate analysis was conducted on the
measure of between-sentence cohesion as well as the story
structure, sentence correctness, punctuation, and spelling

measures. Again, only the variable of interest: the

72
proportion of between sentence ties will be discussed. There
was no significant Occasion X Grade X Reading Ability
interaction in the overall analysis. There was a significant
Occasion X Grade interaction, F(12,34) = 3.31, p<3003,
however, for the linear component of the Grade effect. The
univariate F test for the Occasion X Grade interaction for
the proportion of between-sentence ties was not significant.
There was a main effect for Occasion in the overall analysis.
Examination of the univariate tests for the Occasion main
effect indicated no main effect for Occasion on the proportion
of between-sentence ties.

In summary, four multivariate analyses of variance of
repeated measures were performed to test the hypothesis that
there would be significant improvement over the year in
students' use of conventions for producing written prose.

The results indicated a significant improvement in some con-
ventions. However, improvment varied as a function of the
type of convention and the grade of the student. In addition,
the use of certain features showed no increase or decrease
over the year for some students.

Students showed the greatest improvement in the use of
mechanics, such as spelling, punctuation, and sentence
conventions. Punctuation increased over the year for all
students. Spelling accuracy improved, except for sixth
graders from Time 2 to Time 3. Sentence correctness improved
from Time 1 to Time 2, but decreased from Time 2 to Time 3.
Students may have shown the most improvement in their use of

mechanics, because these conventions require attention to the

73

surface features of text, such as letter and word formation,
and sound—letter relationships rather than to the semantic
aspects of texts. Use of these conventions may require less
depth of processing during editing. In addition, these
skills traditionally receive more emphasis from teachers.
Research has demonstrated that teachers devote more time to
the instruction of mechanics than to the instruction of other
aspects of composing. Finally, the mechanical skills may
have shown greater improvement because they are skills
applied during editing which the production of cohesive text,
for example, is not. The use of sentence conventions may
have shown less improvement than the use of other surface
conventions because they apply to larger units of text than
spelling and punctuation conventions.

The use of devices to produce coherent, organized prose
showed less improvement than the use of mechanics. The only
type of cohesion to improve was endophoric reference. The
proportion of cohesive ties decreased over the three times
in fifth and sixth graders' writing and from Time 1 to Time 2
in the writing of fourth graders. The use of lexical cohesion
showed no increase over time for any grade and there also
was no increase in the use of between-sentence ties. Stue
dents did show improvement in their use of endophoric refer-
ence, indicating an adaptation with time to the demands
inherent in writing situations for explicit reference.

The use of text structure also showed less improvement

than the use of mechanics. The number of story grammar

74
categories decreased in the writing of fifth and sixth
graders but increased in the writing of fourth graders.

The absence of significant improvement over time in the
use of most types of cohesion and text structure may have
been due to the greater difficulty involved in operating on
semantic and structural aspects of text. An alternative
explanation is instructional: teachers traditionally do
not offer instruction in structural elements or devices for
improving coherence. Schemes for using cohesive ties and
story grammar may be applied during the planning process,
rather than as part of editing, and so may be independent of
conventions for improving text. A related explanation which
would clarify the decline over time in the number of story
grammar categories used by fifth and sixth graders and the
decline in the proportion of cohesive ties is that attempts
to apply mechanical conventions may interfere with skills of
idea generation and organization. Students may regress in
certain skills while attaining mastery over others.

Table 1 shows the means, standard deviations, and uni-
variate F statistics for the Occasion effect. Table 2 shows
the means, standard deviations, and univariate F statistics

for the Occasion X Grade interactions.

POQHHQ COfiMWOOO l H QQQQH

 

75

 

mm” swam ms.a mm.: as. mm.e so.a ma.m messes
*tmm m . m :ePdo usesmaw
known me pmnssz
om. mm.m ms. mH.m mm. .om.a .oo< mandamam
and
mo. ecaam. mm.m om.m ms.m mo.m mo.m mm.e mnoom weaaamam
***mmqom . m . smswpsoo
saw
*om.m snaam an. me. am. mm. mm. mm. meoCmpCmm
No. . m Pomamoo
Mo Coapmoaoam
saw
me. sagas as. mm.a as. mm.a om. as.a menses
*m:.w . m V Cowpmzpocsa
saw
mm” csfiam Hm. do. mm. mm. mm. :0. weep mosmpsmm
mm m :Cmmspmn
mo sowpaoaomm
saw .
mm.H use a mm. mm. mm. mm. mm. mm. moseamyea
0.“ .H .
*m m . ewaozmovse
Mo COHpmomoam
mad .
oou Usaam no. mo. do. mo. mo. mo. mews Hecaxma
do N . m mo Cowpaoaoma
saw .
Ha.m aside no. am. so. mm. mo. mm. mmap teammaoo
***HH.NH a m mo Cowpaoaomm
Hoo.vAH u ***
Ho.vOH u ** m epwflpm>fisb .Q.m use: .Q.m use: .Q.m sees manmwam>
mo.va u * mane m mafia a mafia

 

 

peewmm soammooo : a mqm<a

 

 

 

 

 

Hm.a sm.m om. mm.: mo.a as.m w
***N.m.d.fl .II GHQ m> W00.” 5mg: :44. 0%.: mmoH 000.: m
wmwaomepmo
memes smegma so.a oo.m mo.a ms.s mm.a om.e e weaposnpm seopm
mo menssz
em. om. mm. as. mm. an. e
. - m. 5. .sm. mm. mm. . m mmaa
#:m m n UMLO sac H : aw HmHPQoMeMmm
mo. mm. mm. me. am. so. a canonaoenm
muses awesfiq Ho soapaomonm
6 m0. WM. .30. H.3- W
7
**m#.m u GHQ Ho mo. mm. :0. mm. m
momma awesflq mo. in. we... mm. a news e>wmmnoo
mo soapnomoam
*mflomu :II. UNSO mNH> wmo WOQN m0. mHoN Adm. WK...” @
*Hnwom H HHHnH wNH> mm.- dHoN mm. mOoN mm. HBoH m 4
can-ene om. mm.m em. sa.m as. oo.m a soeesooa
weaawmam
Hoo.vAH u ***
Ho.va u ** .Q.m x .D.m x .Q.m N sumac manmwam>
mo.va u * mane N mane a mafia

m mpmflmm>HsD

 

 

waowpomaepsH seaweeoo x mumps

N mam<9

77

Hypothesis 2: There will be no increase over time in
the number of content categories in children's stories.

The four multivariate analyses were conducted with the
number of content categories as one dependent variable.
Results of the univariate tests indicated that the linear and
quadratic trends for the content variable were significant,
Flin(l,45) = 7.62, p .01: F d(1,45) = 6.62, p .01. The

qua
means for the three times for content categories were

(4.39. 4.61. 3-37).

The content categories were the types of propositions
from Warren, Nicholas, and Trabasso's scheme. It was thought
that the number of propositions a writer used would be deter-
mined by his ideational fluendy and that since ideational
7fluency is presumably mastered prior to conventions for
writing the number of categories should not increase over the
year. The results reported above indicate that the number
of content categories increased slightly from Time 1 to
Time 2 but decreased from Time 2 to Time 3. Thus, only
partial support of hypothesis was obtained. There was
development on this feature from Time 1 to Time 2, but not
from Time 2 to Time 3.

Table 3 shows the means, standard deviations, univariate

F statistics, and p values for the content category variable.

 

 

TABIEI? . . ’—
Univariate F
. . , * = (.05
Time 1 Time 2 Time 3 p
Variable x , . ""“*—‘ ** = p<201
s D x S.D. x S.D. *H = p<_001
Number of F . = 7.62**
Content 11“
Categories 4.39 1.16 4.61 1.36 3.78 1.57 = 6.62**

Fquad

78

Hypothesis 3a: There will be a significant difference
between the fourth, fifth, and sixth graders in the variables
discussed in the first hypothesis.

_Hypothesis 3b: There will be a significant difference
between children in the fourth, fifth, and sixth grades in
the pattern of growth on each feature. There was no signifi-
cant main effect for grade in the four multivariate analyses.
However, Occasion X Grade interactions were found in the
overall analyses, as reported, supporting the second hypothe-
sis. Univariate tests revealed significant Occasion X Grade
interactions for the Spelling accuracy, overall cohesion,
endophoric reference, and text structure measures. The
results of the univariate tests have been reported and will
simply be summarized here.

The fourth and fifth graders' spelling accuracy scores
improved over the year. The sixth graderS' spelling accuracy
scores improved from Time 1 to Time 2 but declined from Time
2 to Time 3. Overall cohesion showed a decline over the year,
in the fifth and sixth graders' writing. In the fourth
graders' stories, cohesion declined from Time 1 to Time 2 but
increased to its original level at Time 3. The number of
story grammar categories in the writing of fifth and sixth
graders declined over time, while the number of story grammar
categories in fourth graders' stories increased over the year.

The absence of a main effect for grade level on any
feature is surprising, since Bereiter (1979) has theorized

and researchers have discovered that writing ability improves

79
with age. The absence of age differences may be attri-
butable to social class differences between the students
in the different grades. Socioeconomic status was inversely
related to grade in this sample. Upper middle-class families
have been found to use more objective, elaborate language
with their children than lower class families, who seem to
use a semantically confined, subjective code (Bernstein,
1964). .The fourth graders may have had linguistic ability
which compensated for the greater age and experience of the
other students.

The Occasion X Grade interactions for the spelling
accuracy, cohesion, endophoric reference, and text structure
measure may be attributable to social class background or
' other features associated with grade level. The fourth
graders, who had an upper middle-class background, showed
consistent improvement on the greatest number of measures.
The fifth graders, who were from a middle-class background,
showed consistent improvement on some measures but a
decline in performance on others. The Sixth graders showed
improvement on some measures, but only from fall to winter.
Their performance on others declined. The types of measures
which showed the least improvement generally were text
structure and cohesion. Students from advantaged back-
grounds may have had more ability to master linguistic
devices for integrating content and for organizing than

students from less advantaged backgrounds.

80

Instructional factors may also have accounted for
differences in the pattern of growth. The fourth graders
could have had more instruction in writing or in text
structure and cohesion, than students in other grades. The
decline in the performance by sixth graders in the spring
may be attributable to their particular instructional
experience. They were attending graduation exercises and
other activities during the last few months of school and
so may have spent less time in academic activities.

There may also have been grade level differences in
motivation Which accounted for differences between grades
in the pattern of growth: Finally, low correlations between
students' performances at different times on the cohesion and
text structure measures were found. The absence of an
occasion effect may have been due to lack of reliability in
these measures.

Hypothesis 4: There will be a significant difference
between students of above average, average, and below average
reading ability on the variables discussed in the first
hypothesis.

The first multivariate analysis on the number of story
grammar categories, proportion of cohesive ties, proportion
of correct sentences, spelling accuracy rating and punctua-
tion rating revealed no significant Occasion X Grade X Reading
Ability interaction. There was no significant interaction
between Reading Ability and Grade or Occasion. Therefore,
the main effect for Reading Ability could be examined. A

significant main effect for Reading Ability was obtained,

81
F(12,80) = 4.39, p<30001. Univariate analyses revealed a
significant effect for reading ability on the measures of
sentence correctness, F(2.45) = 5.18, p(201, spelling,
F(2,45) = 25.09, p<.0001 and punctuation, F(2,45) = 24.52,
p(30001. The means for the three groups were as follows:

Spelling Accuracy Rating:

Above Average (2.55, 2.72, 2.89)
Average (1.50, 2.11, 2.11)
Below Average (1.33, 1.61, 1.83)

Punctuation Rating: Above Average (2.33, 2.39, 2.61)

Average (1.55, 1.94, 1.44)

Below Average (1.33, 1.44, 1.44)
Proportion of Correct Sentences: .

Above Average (.80, .81, .88)

Average (.63, .72, .62)

Below Average (.61, .73, .55)

The seccnd multivariate analysis of the measure of
lexical ties, the combined spelling score, and the other
variables revealed a significant Reading Ability X Grade X
Occasion interaction, F(48, 133) a 1.45, p<305. Univariate
tests indicated a significant Reading Ability X Grade X
Occasion interaction for the combined spelling scores.

As noted, the groups showed very complex patterns of
growth. There was an increase in the combined spelling score
over the year for the fourth graders of all ability levels,
the fifth graders of average and below average ability and
the sixth graders of below average ability. The fifth graders

of above average ability and the sixth graders of average

82
ability showed an increase in their scores from Time 1 to
Time 2 and a decrease from Time 2 to Time 3. The sixth
graders of above average ability showed a decrease from
Time 1 to Time 2.

The third multivariate analysis was conducted on the
proportion of endophoric referential ties as well as the
other measures. No significant interactions between Reading
Ability and any other factor were obtained. A significant
main effect was obtained for Reading Ability, F(12,80) =
4.39, p<.0001. However, univariate tests indicated no A
Significant effect for reading ability on the proportion of
endophoric referential ties.

The final multivariate analysis was conducted on the
fourth cohesion variable: the proportion of between-sentence
ties as well as the other measures. There was no significant
interaction between Reading Ability and any other factor in
this analysis. Again, a Significant main effect for Reading
Ability was found in the overall analysis, F(12,80) = 5.57,
pGOOOl. Univariate analysis indicated no significant effect
for reading ability on the proportion of between-sentence
ties, the variable of interest in this analysis.

In summary, the results of the four multivariate ana-
lyses revealed a significant main effect for reading ability
on three measures: the spelling accuracy, punctuation, and
sentence correctness measures. A significant Ability X Grade
X Occasion interaction was found for the combined spelling
score. Thus, higher reading ability was positively related
to the use of mechanics but not to the use of cohesion or

'text structure.

83

Greater reading ability may have had an effect on the
use of conventions because better readers have more know-
ledge of the surface features of text. Better readers also
may have greater facility at re-reading, which is a component
of the editing process. Alternatively, reading and writing
ability may both have been affected by a third factor, such
as intelligence or motivation to achieve.

The lack of a main effect for reading ability on the
proportion of cohesive ties of any type may be attributable
to the composing process responsible for cohesion. Cohesive
ties may emerge during the planning process of idea produc—
tion and organization and thus may not be related to reading
or linguistic ability.

The absence of a main effect for reading ability on
text structure usage also may be due to the composing process
responsible for providing structure to the text. Text
structure may be provided during the idea organization
process which involves operations on non-linguistic units.
Text structure may have been more related to cognitive
development than to reading ability. Formal operational
thinking may facilitate organizational processes in writing.
Finally, the lack of a main effect for reading ability on
text structure and cohesion may have been due to a lack of
reliability in these measures. Table 4 contains the means,
standard deviations, and univariate F statistics for the
reading ability effects.

Appendix A contains the overall tables for each of the

four multivariate analyses of repeated measures.

84

NH.”

nn.a
,tttmo.mw
t:¢mm.m~

¢.ma.m
.temm.:~

am.

no.

on.~

Nm.

.Hoo.vn I set
Ho.va u we NIqududHHQd

mo.vnH n 9

mm.
mm.
:m.

an.
an.
m:.

an.
on.
um.

mm.
am.
am.

me.
No.
:0.
mo.

so.
no.

.Q.m

ammue>< Soaom

mm.m
NN.¢
oo.:

mm.
mm.:
mm.:

mm.a
H0.H
mm.a

no.
ma.
mm.~

mm.
mm.
Hm.

::.H
N?

no.
no.
00.

a.
as.
a.

mo.
50.
no.

mm.
m 0
on.

cams

ae.a so.n
mm.H mm.:
mm.a sm.:
Ho.~ no.
as. oo.
sm.a am.:
so. Ha.~
on. HH.N
me. om.a
om.~ Ha.“
os.~ oo.n
so.~ -.m
mm. me.
an. we.
SN. me.
an. ss.a
an. sa.H
as. mm.fl
an. on.
am. :0.
mm. so.
mm. om.
am. me.
am. as.
so. Ha.
no. oo..
no. .mo.
mo. an.
0. mm.
0. an.
.n.m sees
mmmmqu

mn.a
mH.H

Hw.d

Ho.

ma.
ma.
am.

no.
so.
no.

we.
no.
so.

.Q.m

mm.
Hm.
om.

H0.N
mm.m
mm.m

as.
as.
oa.a

we.
we.
ms.a

mo.
mo.
mo.
em.
mm.
mm:

cams

wmwam>< o>ond

HNM HNM FINN HNM HNM HNC‘N HNM r-INM HNM HNM

egae
wane

mafia

mane
mama
mane

asaa
mafia
mane

mafia
mafia
mama

mane
mafia
mama

mama
mass
mafia

asae
wane
mane

mafia
meme
mafia

mama
mafia
mafia

mafia
mama

GEHB

mowaOMmpmo
vcmvCoo
no umnssz

mowhoMmpmo
umasmao heepm
Mo penesz

. meapmm
.oo¢ mewaaeam

muoom
mandamam
escapsoo

mmosmvam
pomaaou
Ho Cowuaanpm

mcapmm
Sowpmspocsm

new» cosmPCmm
:Cmmsumm
mo cowpaoaoum

mocmaQMmh
owaonaonso
Ho scapuoaoum

new» Hmowxma
«o soapaoaoum

mods o>wmmgoo
mo sowpuoaoam

 

magmacm>
mmmmmuuwaaapa

 

 

upoouem spaaapa - a mamas

85
One area of interest was the degree to which the
measures, which were believed to all assess skills acquired
in the performative stage of writing development, were
related. Pearson Product Moment Correlations were obtained
for the averages of the dependent variables over the three

times. Table 5 contains the correlations of the dependent

variables.

no.» . no. nu.
coauoUouou
vconuou
no eons-E
o...“ S.
ooaoooouou
Conn
no oops:
co."
ooah
oocopcoa
ocootuoa
«o .ook

Ono. 00o. ‘flo Ono. .OOoO “do “do hho. "no .Co “no

00o . who. sho duo. .0o 0». . on. . hno rho nno Oho

sh. nfloo on. “do “0.! .Qfloc duo! sh. h“. . an. an.

co.“ an.. an. o~.. on. o«.. a... . on. on. on. an.

OOGOHONOI

u. .9... .
go” go. .flol .0. “go- OOoo ”no- "‘0. n... fio-

«.5230 _ .
oeoauo .oo.‘

00o“ .0o .84. Ofloh 80o “do “no “Oo . Bflo
ooau
. 13.3
N. .thh .
8 a 2.. 3.. 8. 3.. 8.. 8. 8..
.ouon «an
ho . out
go" “Go. So go. no 80 “do

0°.u a“. un.l uu.0 un.o ~u.o

coat .hou
N. o...‘
god Odo! Odo. go. so.

:3
.Puloflou
a. £2:
o Sod Ro NOo “o

ooocoucoa
#09980
«o .z
8.." no. oo.
9:223

ooxaploo
83 no.
oodonu rho!
. no nauaonuuda
oo.n
. . . . god swoon

an.

nu.
an;
Qflo

10.:

n1:

nu...
an.

an.
2..

as.

oo.u

oouunu
:oaau hdfluohu

Down uo Sufi...

oouk ouuouccm
.cooovum
ho couuuouuna

ooCohonoK
ouuntunn-u

«o couuuonuum

ocouuucauuou
«o couounuunt

ooau
«duuuou
no .nnhh

Caouuuum
no .nuum

odouuau
:uuouaw
Mo 89».—

ooau .u.:
«o .aupa

noun
ooaoorun

ho .Aobm

ooucouuom
uoouuuu
uo .acLL.

9.42...."
e.guuu.u

3:5 an;
no 933......“

hoouaooo
9:27:

IiuvdflvOCJh

. Solo-cl

w «in

87

The Pearson Product Moment Correlations indicated that
some of the features thought to be text conventions adopted
in the performative stage of writing development were highly
related while others were not. The skills of punctuation,
spelling, and sentence formation wew~ positively related.

The proportion of cohesive ties was not related to these
measures, however, there were positive correlations between
the proportion of lexical ties, endophoric referential ties,
and between-sentence ties used and the mechanics measures.
The number of story grammar categories and content categories
employed were highly related to each other and moderately
related to the other variables except the proportion of
cohesive ties.

The correlations seem to indicate that mechanical con—
ventions are acquired by writers together and that there may
be a unitary stage of performative writing. Features such
as cohesion and text structure however are not among the
attainments of this stage either because they are not
applied during editing or because control over them is
acquired later than control over mechanics.

One question of interest concerned the pattern of growth
shown by each feature assessed. The four multivariate anal-
yses of variance of repeated measures included contrasts
of the constant, linear, and quadratic terms of the occasion
effect for each feature. The results indicated a Significant
overall linear trend for the punctuation rating, Flin(l,45) =

6.47, p<.Ol and a significant overall quadratic trend for the

88

proportion of correct sentences, Fquad(l,45) = 3.24, p<205.
There were significant Occasion x Grade interactions for
other variables. The variables which showed improvement for
any grade were: the spelling accuracy rating, the proportion
of endophoric referential ties, and the number of story
grammar categories. The means for each grade for the
spelling accuracy rating were as follows: fourth grade
(2.06, 2.16, 2.92), fifth grade (1.71, 2.09, 2.14), and
'Sixth grade (1.76, 2.19, 2.05).' The means for the proportion
of endophoric referential ties for each grade were: fourth
‘(.67, .68, .88), fifth (.64, .73, .74), and sixth (.59, .74,
.60). The use of text structure improved only for the fourth
grade. The mean number of categories used by the fourth
graders for the three occasions was: 4.50, 4.75, and 5.00.

These results demonstrate that there was a steady
improvement in the punctuation rating, overall, while there
was irregular growth in sentence correctness. Students may
have had more difficulty controlling sentence conventions
than punctuation conventions on different tasks because
sentence conventions apply to large linguistic units. The
means for the other variables reveal different patterns of
growth for each age. The means for spelling accuracy demon-
strate that spelling accuracy increased steadily in fifth
graders' writing, but in the fourth and Sixth graders'
writing it showed irregular growth. 'The growth of the
proportion of endophoric referential ties was uneven with all
grades showing improvement from one time to another rather

than over all three times. The fourth graders' showed

89
improvement only from winter to spring, while the fifth and
sixth graders showed improvement from fall to winter. The
number of story grammar categories in the fourth graders'
stories increased steadily over time, but decreased in the
writing of fifth and sixth graders.

These results do not permit any straight forward
conclusions about the pattern of growth in writing.

The final question concerned the types of story grammar
categories found in children's stories. Of particular
interest was the contrast between the proportion of propo-
sitions in the setting category as opposed to the focal or
initiating event category in children's first stories. It
was thought that an emphasis on the use of the setting cate-
gory would reflect use of text structure to construct stories,
while an emphasis on the initiating event might reflect a
strategy of using content to organize a story.

The data analysis showed that the largest proportion of
propositions in the fall occurred in the attempt category
(.36) followed by the initiating event (.17), direct
consequence (.17), setting (.10), reaction (.08), plan (.07),
and internal response categories (.05).

These results imply that the event chain or plot in a
story is emphasized more than the setting category in begin-
ning writing. Writers may construct stories in a bottom-up
fashion, beginning with the action sequence. However, the
results of this study suggest that the focal point in early
story construction is the behavior of a character. Writers
may produce stories by generating actions first and then

logical causes and effects of actions.

CHAPTER V
SUMMARY AND CONCLUSIONS

This chapter presents the results and conclusions drawn
from the data analysis. The limitations of the present
study and implications for future research are presented at

the end of the chapter.

Summary

This study was undertaken to trace the development in
children's writing of features derived from a cognitive-
developmental model of growth in writing (Bereiter, 1979).
This model incorporates concepts from current theories of
the writing process. According to one such model, developed
by Flower and Hayes, there are three elements to the writing
process: the task environment, the writer's knowledge, and
the writing processes themselves. The task environment
consists of the-writing problem and the developing compo-
sition. The writer's knowledge is comprised of the writer's
knowledge of discourse, content, and mechanics. The writing
processes are planning, translating, and reviewing. During
the planning phase, the writer generates ideas and organizes
them into some kind of structure. During the translating
phase, the writer transforms the products of the planning

phase into English words and sentences. The reviewing

90

91
phase is the phase during which the writer checks and
corrects the written text.

According to Bereiter, writing development occurs in a
series of stages, each involving the refinement of skills
employed during different phases of the composing process.
There are six different Skill systems employed during these
phases by mature writers: ideational fluency, writing
fluency, knowledge of writing conventions, social cognition,
literary criticism Skills, and formal thinking skills. Only
the first two stages were of interest in this study. During
the first stage of associative writing, skills of idea
generation and writing are developed. During the performa-
tive stage, Skill at using text conventions is developed.

It is not clear from Bereiter's description which conventions
are applied during the performative stage. Different types
of conventions apply to different levels of text: words,
sentences, segments of discourse, and discourse. Such con-
ventions include spelling, punctuation, cohesion, and dis-
course structure.

Research on process—related features of text indicates
an increase with age in the use of spelling, punctuation,
and sentence formation rules and some types of cohesion:
(NAEP, 1975, Bracewell, 1980). However, no study has
assessed the development of all these features over time.
Previous research on process-related features has been
conducted using cross-sectional rather than longitudinal
designs. It also has not been clear to what extent each

stage is a unitary one. Finally, there is not enough

92
information available on writing development to characterize
the nature of growth in writing as linear and continuous or
discontinuous.

The purpose of this dissertation was to provide longi-
tudinal as well as cross—sectional data on the development
in writing of a set of features associated with attainment
of the performative stage. These features included spelling
maturity, punctuation maturity, sentence correctness,
cohesion, and text structure. A second purpose was to
examine the effect of age and reading ability on these
.variables. In addition, the relationship between measures
was determined. Finally, the nature of growth in writing
was identified. .

The study was done using fifty-four children whose
teachers were participating in a study of writing instruction
conducted by the Language Arts Project of The Institute for
Research on Teaching. The fourth grade children were drawn
from an upper—middle class community which has only a small
percent of minority families. The fifth grade children were
from two rural, lower—middle to middle class communities
which were predominantly white, and from an urban, lower-
middle class racially mixed community. The sixth grade
children were from the urban community also.

The measuring instruments included a test of reading
ability and measures of the dependent variables. The Wood—
cock Test of Reading Mastery was used to assess reading
ability. The measures of the dependent variables were:

ratixu; scales of punctuation maturity and spelling accuracy

93
and a combined spelling score, a sentence assessment scheme,
Halliday and Hasan's scheme for analyzing cohesive ties,
Stein and Glenn's story grammar, and Warren et. al.'s event
chain scheme.

Students were assigned narrative writing tasks in the
fall, winter5and spring. Writing samples were coded for
each dependent variable. Each type of coding was performed
by two coders, with the exception of the sentence correct-
ness coding.

Four 3 x 3 multivariate multivariance analyses of
repeated measures were performed on the dependent variables.
The linear and quadratic trends for each feature were
identified. Pearson Product Moment Correlations between the

dependent variables were calculated.

Conclusions
There was improvement over the year in the use of some

'conventions for producingpcorrect written prose. These

 

results support Bereiter's hypothesis that writing develop-
ment involves an increasing ability to employ conventions
for writing. However, development varied as a function of
the type of convention and grade. The results will be

discussed separately for each feature.

Text Structure

Narrative schemata are generalized representations of
the categories of content and relations in stories. Even
young children have been found to use narrative schemata in

their comprehension and recall of stories. Researchers

94
(Stein and Glenn, 1979; Bereiter, Scardamalia, 1980) have
found that children use story schemata in their story
production. Stein and Glenn (1979) found that younger
children used fewer categories than older children. However,
Bereiter, Scardamalia, and Turkish (1980) found no difference
between fourth and sixth graders in the number of categories
used in stories. They found that fourth and sixth graders
used an average of 5.51 categories out of the 8 devised by
Stein and Glenn.

In the present study children used 5.13 categories at
Time 1, 4.89 categories at Time 2, and 4.13 categories at
Time 3. They used fewer categories than those used by
Bereiter's subjects. Bereiter and his colleagues used
subjects from Ontario, Canada. The greater number of cate-
gories in their studies may have been related to background
characteristics such as greater reading or writing experience.
The students in the present study used fewer categories but
the number they used seems to indicate that they employed
story schemata in their composing.

The number of categories used increased over the year
only in fourth graders' stories and declined in fifth and
Sixth graders' stories. However, the sixth graders used the
greatest number of categories at Times 1 and 2 followed by
the fifth graders, and then fourth graders.

According to models of the composing process, structure
in writing is provided during the organization phase of
planning. The function of this process is to select and

sequence appropriate ideas from those generated during idea

95
production. However, text structures also are one type of
literary convention, which writers can have in conscious
awareness and employ during editing. Bereiter, Scardamalia,
and Turkish compared children's conscious knowledge of story
grammar to their ability to use story categories in writing.
They found that children used more categories in their writing
than they were able to name, suggesting that children did not
employ conscious awareness of any elements in their story
production.

The decline in the number of categories used by sixth
.graders in the winter and spring, and fifth graders in the
spring suggests that while these students could employ story
categories, the structure in their stories was not the product
of conscious knowledge which could be applied regardless of
task. Structure in their narratives may have been provided
by planning rather than editing processes.

In.lris model, Bereiter proposes that writers attain
automaticity at lower level processes before refining higher
level skills. The decline in story grammar usage by the
fifth and sixth graders may indicate that these students
began to develop surface conventions before they had mastered
organizational processes. Their attention to mechanics may
have resulted in regression in the skill of organization,
because the skill still required conscious thought. Thus,
students may not follow the developmental course described

by Bereiter.

96

They may incorporate some higher level skills, especially
those such as mechanics before lower level skills are
automatic, particularly if higher level skills are empha—
sized by a teacher.

The decline in text structure usage by sixth graders
in the winter and spring may also have been due to the
cognitive processes required at the three times. As noted,
in the fall task, students were required to retell a story
from the perspective of a non-protagonist, a recall task.

In the winter and spring, children had to generate stories

in response to a photograph or painting. The decline in

text structure usage in the spring by sixth and fifth graders
could have been partially due to the stimulus. The spring
stimulus could have been a difficult one because it was very
static (The stimulus was a print of a man seated in a tree).
Students may have found it difficult to generate a plot
around this picture. .

The increase in the number of categories in the fourth
grader's stories seems to indicate that while they had less
well-established story schemata in the fall, as the year
progressed, they acquired schemata which they were then able
to employ regardless of the task. I

The absence of improvement by most students except
fourth graders in the use of text structure may be attri-
butable to other factors. Social class, teacher, and school
factors confounded with grade may have accounted for the

pattern of growth in story structure.

97

The writing samples produced by the sixth graders in the
winter and spring and the fifth graders in the spring may
have been too short to provide a representative sample of
their text structure knowledge. The mean number of words
used by children in each grade at each time was as follows:
Fourth graders (138, 92, 148): Fifth graders (168, 93, 91),
and Sixth graders (212, 124, 104).

The lack of an increase in the number of story cate-
gories may also have been due to lack of reliability of the
text structure measure. The correlations for the number of
Vcategories was (.18) between Times 1 and 2, (.11) between
Times 1 and 3, and (.02) between Times 2 and 3. The low
test-retest reliability of the measurers may have resulted
in the lack of a relationship between text structure and

other measures.

Cohesion

There was no increase over time in the proportion of
cohesive ties, or overall cohesion. The proportion of cohe-
sive ties declined in fifth and sixth graders' stories.

. Cohesive ties are lexical and syntactic devices for
relating propositions and sentences in text. They are one
means for creating unified text. Researchers (Hidi, 1980)
have disagreed about the composing process reSponSible for
cohesion. Bereiter (1979) suggests_that cohesive prose is
the result of semantic relations discovered during the idea
generation and organization processes. Hayes and Flower
(1979) suggest that the production of cohesive text is due to

the application of knowledge of the surface conventions of text.

98

Bracewell (1980) found no increase with age in the
proportion of cohesive ties. His findings and those of the
present study may suggest that the production of cohesive
text is a function of the idea production and organization
process which is mastered earlier than the transformation of
ideas into correct writing.

Hidi (1980) investigated children's ability to produce
coherent sentence sequences under two types of conditions.
The first was a Before vs. After condition in which children
were provided with a sentence and required to produce a
“semantically and logically related sentence to precede or
follow it. The second type of condition varied the type of
semantic content of the writing stimulus from concrete to
abstract. Children produced more coherent sequences in the
After condition in which they had to generate a sentence to
follow a given sentence. According to Hidi, this result
occurred because "discourse production which proceeds in a
linear, forward~going manner develops earlier than discourse
production which requires switching between forward and
backward-going production" (Hidi, 1980, p. l). Hidi also
found that the type of semantic category influenced the
ability to compose semantically and linguistically organized
text sequences. More abstract categories were more difficult
than concrete categories. I

The children in the present study may have produced
cohesive discourse early in the year because narrative
writing is a linear, forward—moving process, and coherent

text could be produced even by beginning writers. The decline

99

in cohesion over time may have been due to the requirement
in Tasks 2 and 3 for the generation of semantic frames. In
contrast, Task 1 provides students with a semantic context.

The proportion of cohesive ties used could have shown a
decline because of an increase by students in the use of
conventions, such as punctuation, for producing correct
text. Students who lacked automaticity in the idea gener—
ation and organization processes may have regressed on these
Skills when required to attend to new skills. The decline in
cohesion, like the decline in text structure suggests that
Istudents may attempt to integrate new skills into their
repertoire, especially skills which are emphasized by
teachers, before they are ready. Such a finding contradicts
a seeming assumption by Bereiter that students do not inte-
grate higher level skills until lower level skills are
automatic. An alternative to the view that cohesion is the
product of content processing is proposed by Flower and Hayes
(1979) who attribute the production of coherent text to the
translating process. The translating process is one in which
the writer transforms the output of idea generation into
linguistic form. Bracewell and Scardamalia (1980) make a
similar suggestion that cohesion is the product of the
metalinguistic knowledge of resources available for producing
coherent prose. .Some of Bracewell's research seems to
support this. He found that a writer's familiarity with
‘ content was negatively related to his ability to produce a

coherent text. The lack of development of overall cohesion

100

among the students in this study seems to indicate however

that cohesion is not the result of the deliberate applica-

tion of metalinguistic knowledge but the result of the idea
production process.

Lack of improvement in the use of cohesion may have been
due to characteristics of writing instruction. Tamer and
Bond (1980) maintain that elementary school teachers offer
instruction primarily in "transcription" skills of punctua-
tion, spelling, and penmanship rather than more complex
skills such as organization. Even when students are given
'creative writing assignments, teachers respond primarily to
the surface characteristics of text.

Students in the study, especially fifth and sixth
graders, produced Shorter samples at Time 2 and Time 3 than
at Time 1. The samples may have been too short to provide
a representative sample of their ability to integrate ideas
in text. A final factor which might account for the lack of
improvement in cohesion may have been lack of reliability
of the measure of cohesion.

There was no increase over the year in one type of
tie: the proportion of lexical ties in students' writing.
Bracewell (1980) found an increase with age in the proportion
of lexical superordinate ties. In the present study, the
use of lexical ties was positively related to the use of
conventions of punctuation, spelling, and sentence formation.
Therefore, the use of lexical cohesion seems, unlike overall
cohesion, to be related to conscious knowledge and use of
devices for producing mature text. Logically, lexical

cohesion should be related to attempts to shape language

101
because it results from the use of similar items in the
writer's lexicon, a type of linguistic knowledge. The
absence of improvement in the use of lexical ties may have
been due to their difficulty. Writing conventions apply to
outputs of different levels of the writing process. Mech-
anics, such as spelling and punctuation, apply to surface
features of text, such as sound-letter correspondences or
sentence beginnings and endings. The use of these devices
during reviewing would require the operations at the surface
.level. The use of lexical ties, however, would require the
analysis of the meaning and the selection of alternate words
to express this meaning. The use of lexical cohesion would
require deeper processing during reviewing than the use of
other conventions.

However, lexical cohesion may not have increased in
writing because of other factors, such as writing instruc-
tion.

The use of endophoric referential ties increased over
time for all students except sixth graders from Time 2 to
Time 3. Explicit reference is necessary for the interpre-
tation of most written texts. In conversations, the speaker
can rely on the listener's knowledge of the context, however,
a writer must communicate meaning independent of the situa—
tion. The increase in endophoric reference indicates that
students were adapting this convention for producing written
prose.

There was no improvement over time in the proportion

of between-sentence ties. In contrast, Bracewell (1981)

102
found an increase with age in the proportion of inter T-
unit ties. The measure employed in this study was the inter-
sentence tie because the process of dividing compositions
into T-units would have been too difficult. Since there was
a positive correlation between the proportion of between-
sentence ties and the punctuation, spelling, and sentence
formation measures, the inter-sentence tie seems to have
been a valid measure of writing ability, and the use of
inter-sentence ties also would seem to be a text convention.
The absence of growth in inter-sentence ties may have been
due to their difficulty. The use of a between-sentence tie
requires operations on the semantic content of separate
sentences. It also may require prior mastery of sentence
formation rules. The proportion of between-sentence ties
also may not have increased due to the factors associated
with grade level, the size of the writing samples in winter
and Spring, and lack of reliability of the measure.

There was no significant difference between students in
the fourth, fifth, and sixth grade in the features assessed.
The lack of grade level differences was surprising, since
previous research indicated an increase with age in writing
ability. The NAEP (1975) reported significant increases
between children ages nine and eleven in their Spelling
accuracy, punctuation maturity, and sentence formation
ability. Bracewell (1980) found an increase with age in
the proportion of lexical superordinates, endophoric
referential ties, and inter T-unit ties. Stahl (1977)

reported an increase with age in organizational features.

103

The absence of grade level differences may have been
due to confounding of social class with grade. Social class
background was inversely related to grade. Social class
background has been found to relate to linguistic ability
(Bernstein, 1964). The greater linguistic ability of
younger children may have compensated for their age.

Significant Grade x Occasion interactions were found
for the variables of spelling accuracy, proportion of cohesive
ties, endophoric reference, and number of text structure
categories. Fourth graders showed improvement over the
year in spelling accuracy, endophoric reference, and text
structure usage. The fifth graders improved from Time 1 to
Time 3 in fewer features: spelling accuracy and endophoric
reference. Their use of cohesion and text structure de-
creased. The Sixth graders showed the least improvement
over time on the fewest features: only their spelling
accuracy and endophoric reference improved from fall to
winter. Their performance on all conventions declined in
the spring. The variations in growth may have been due to
the type of convention and grade level. Important differ-
ences between conventions have been discussed. Grade level
effects also have been discussed briefly. The differences
between the grades in pattern of growth may have been due to
age or to the factors associated in this study with grade:
social class background, and educational experience.
i The fourth graders' social class background and
presumably greater linguistic ability may have enabled them

to develop over the year on a number of dimensions, including

104

text structure. If text structure is a convention, fourth
graders showed more ability to employ text categories on
any task. Fifth graders, who were from a middle class
background, showed improvement was more prolonged than that
of sixth graders. This pattern may be attributable to
social class differences between fifth and sixth graders.

Differences in educational experience may also have
accounted for the differences in writing growth. The absence
of grade level differences in the face of development over
the year points to the importance of schooling in writing
development. Britton's results suggest that schooling has
an overwhelming influence on writing development. It seems
likely then that the pattern of growth may have been related
to instruction. The decline in sixth graders' writing per-
formance in the spring may have been due to their partici—
pation in graduation activities and orientation to junior
high school in the Spring. 'Grade level differences may also
have been due to a greater frequency of writing instruction
in certain grades and/or instruction about certain features
of text such as cohesion. Finally, declining performance on
some measures may have been due to a decline in motivation.

Hgadinggability had an effect on the use of some
conventions for producing written prose. These conventions
'included the mechanical rules for spelling, punctuation, and
sentence formation. Reading ability did not relate to the
use of cohesion or text structure.

A strong relationship between reading and writing

ability has been demonstrated (Loban, 1976). Reading ability

105
may have been related to performance mechanics because it is
related to knowledge of the surface features of text.
Reading ability also may relate to facility at rereading and
therefore editing.

The absence of an effect for reading activity on the
cohesion variables may be attributable to several factors.
Bracewell (1980) proposed that reading ability might be
negatively related to the use of cohesion because better
readers are more able to abstract the propositional content
from sentences. This ability may result in better readers'
having less knowledge of, or memory for, elements of them-
atic structure than less able readers.

An alternative explanation for the lack of effect of
reading ability involves the composing process responsible
for cohesion. If devices for indicating relations between
ideas are a result of the idea generation process, there is
no reason to expect that greater linguistic knowledge (and
reading ability) Should be related to their use.

Reading ability had no effect on the number of story
grammar categories used. The absence of a relationship is
surprising because better readers would be expected to have
more well-established narrative schemata. The absence of
a relationship may have been due to the composing process
responsible for text structure. If the organization sub-
process is responsible then reading or linguistic ability
could have little influence.

Text structure usage may not have been mediated by

text structure knowledge even if it was present in readers.

106
Bereiter, Scardamalia, and Turkish demonstrated that know-
ledge does not determine the structure used by children in
their stories.

Text structure use may have been more related to
cognitive development than to reading ability. Bereiter,
Scardamalia, and Turkish suggested that because of the formal
nature of discourse grammars, use of story grammar in writing
might depend on formal operational thinking skills: however,
as stated, they found that the number of story grammar
categories used was not determined by children's conscious
'knowledge. Formal operational skills might, however, affect
ability to provide organization in prose. In the present
study, a difference between grades in text structure usage
did occur in the fall and winter. Sixth graders used the
greatest number of story grammar in the fall and winter,
followed by fifth and fourth graders.

Pearson Product Moment Correlations indicated that some
of the skills thought to be employed during_the_performative
gigge were related. Punctuation, maturity, spelling acc-
uracy, and sentence correctness were highly related. Over-
all cohesion was not correlated with these skills: however,
the proportions of lexical ties, endophoric referential
ties, and between-sentence ties were. Story grammar usage
and content coverage were moderately related to all other
variables except overall cohesion.

These results seem to indicate that skills for pro-
ducing correct text are related. They suggest that there

may be a set of skills acquired in the performative stage

107
of writing performance. These skills include punctuation,
spelling, sentence formation, lexical, between—sentence, and
endophoric referential ties, and possibly text conventions.
The use of cohesion, for reasons discussed earlier, may be
a separate skill, perhaps a function of the idea generation
process. The use of lexical, endophoric, referential, and
between-sentence ties was positively related to the use of
text conventions.‘ As stated, logically, these types of
cohesion seemed to demand more conscflnusknowledge of the
_linguistic aspects of text than other types of cohesion.
The use of story structures was moderately related to punc—
tuation, spelling, and sentence formation ability. However,
these structures were not as highly related to other skills
as they were to each other. Text structure knowledge may
not be a skill applied during the editing process. Its
moderate relationship to other features may be explained by
the fact that skillful writers are able to use both text
structure and rules for correct writing.

The absence of strong positive relationships between
the cohesion and text structure measures and other measures
may also have been due to lack of reliability of these
measures.

No clear pattern ofggrowth in writing was apparent
from the data analysis. Bereiter (1979) has pointed out
that there is not enough research evidence available to
allow for the characterization of writing growth as con-
tinuous or discontinuous. Loban (1976) found a gradual

increase from kindergarten through twelfth grade in syntactic

108
complexity and elaboration in children's writing. However,
the NAEP (1975) reported an improvement in writing quality
between the ages of nine and eleven, but no improvement
between ages thirteen and seventeen.

The question of the pattern of growth was not resolved
by the data in this study. Only two features showed Similar
patterns of growth across the three grades. One, punctua-
tion maturity, improved steadily, while the other, sentence
correctness, showed uneven growth. Development on the other
_features varied as a function of grade.

The final question of the study concerned the types of
categories used by children in story construction. AS
noted, Stein and Glenn (1979) found that the categories
used most frequently during recall by children and adults
were the setting category followed by the initiating event,
direct consequence, and attempt categories. Subjects did
not recall a high proportion of propositions from the reac-
tion or internal response categories. Their investigations
of story production demonstrated that setting information was
used most frequently by younger subjects. The purpose of
the present study was to determine the frequency with which
different types of categories were used in stories produced
at the beginning of the study. Of particular interest was
the question of whether setting or initiating event infor-
mation would be used more frequently. It was argued that
an emphasis on the setting category would be evidence that
children were using story structure in composing, whereas an

emphasis on the initiating event would result from use of

109
meaning and the most meaningful element in forming
stories.

The results indicated that the category which received
the most emphasis in the stories produced at Time 1 was
the attempt category. The category used next most frequently
was the initiating event category, followed by the direct
consequence, setting, reaction, plan, and internal response
categories. These results suggest that children use meaning
and semantic relations rather than story structure in
.organizing their first stories. However, the most meaningful
.element may not be the initiating event but the main charac-
ter's behavior. Rather than beginning with an event to which
the main character responds, students begin with his
behavior. Story production then presumably emerges through
generation of logical causes and consequences of the main
character's behavior.

The emphasis by students in the fall on semantic
relations in structuring their stories supports Bereiter's
proposal that beginning writers emphasize content production
and organization in their beginning writing. These results
may also suggest that the notion of story or text grammars,
like the notion of deep structure, may have to be revised.
Stories may not be the product of knowledge of the structure
of texts but may be representations of a writer's experien-

tial knowledge.

Implications

The results of the study suggest the following implica-

.tions:

110

1. Growth in writing seems to follow the developmental
course described by Bereiter, at least from associative to
performative writing. Writers Show an integrated increase in
the use of several conventions for producing correct text.
However, use of these conventions may vary depending on the
students' linguistic ability and instructional experience,
and the task. These results would suggest that teachers
diagnose the stage of writing development of their students
and gear their instruction accordingly. They may wish to

allow students to strengthen and consolidate Skills at one
.stage before teaching skills of a new stage. The model
also suggests a way for teachers to program writing instruc—
tion so that higher level skills such as the development of
a personal style are not taught while the student is in a
much lower stage.

2. While some features, such as mechanics, may Show
improvement, others, such as cohesion and text structure may
not, suggesting that these skills are employed at an earlier
point in composing, are more difficult, or do not receive
as much emphasis from teachers. Students may even slip back
on these skills while attempting others. The results may
indicate that teachers should wait until their students have
acquired control over idea production and organization
processes before requiring them to use all their mechanical
skills in writing. Instruction in the use of cohesive ties
and text structure which is Similar to the instruction they

receive in mechanics might give writers similar conscious

111
control over these features. Alternatively, instruction in
improving the use of text structure and cohesion may be more
useful if the writer is instructed to use them during the
planning subphases of idea generation and organization.

3. The absence of grade effects seems to indicate, as
might be expected, that writing development may depend more
on instructional experience and linguistic ability than age,
cognitive development or even prior experience, assuming
that older students are more developmentalbradvanced and

had greater writing experience. Writing may be a skill that
SIfor most children, especially lower class children, is
learned in school or not at all. AS expert writers claim,
it may be necessary to constantly practice writing to improve.

4. The relationship between reading ability and some
writing variables may indicate that reading experience can
facilitate writing achievement. However, certain aspects of
writing, such as cohesion may not relate to ability, either
because they are a function of the task or non-linguistic
processes such as idea generation, are more difficult, or
because they do not receive enough emphasis by teachers.
Even good readers may not maintain or improve these devices
in the absence of any instruction in them. Teachers may
have to help them to improve more than just the surface

features of text.

XHdVHDOI'IHIH

BIBLIOGRAPHY

Bartlett, F. C. Rememberipg. Cambridge: The University
Press, 1932.

 

Bates, E6 Language and Context. New York: Academic Press,
197 .

 

Bereiter, C. Development in Writing. In Gregg. L. W. and
Steinberg, E. (eds.) Cognitive Processes in Writipg.
Hillsdale, N. J.: Erlbaum, 1979.

' Bereiter, C., Scardamalia, M., and Turkish, L. The child
as discourse grammarian. Paper presented at the
American Educational Research Association, Boston, 1980.

Bereiter, C., Masterton{ B., and Scardamalia, M. Revision
of written text in response to new information. Paper
presented at the American Educational Research Associa-
tion, Boston, 1980.

Bracewell, R. J. Is writing retarded by reading? Paper
presented at the Conference on Writing Research,

Bracewell, R. J. and Scardamalia, M. Children's ability to
integrate information when they write. Paper presented
at the American Educational Research Association,
Boston, 1980.

Braddock, R., Lloyd-Jones, R., and Schrer, L. Research in
Written Composition. Illinois: National Council of
Teachers of English, 1963.

 

 

Britton, J., Burgess, T., Martin, N., McLeod, A., and
Rosen, H. The Development of Writing Abilities
(ll-18). London: MacMillan, 1975.

Clark, H. and Clark, E. Psychology and Language. New York:
Harcourt Brace JovanovICh, Inc., 1977.

 

Flower, L. S. Writer-based Prose: A Cognitive Basis for
Problems in Writing. College English, 1979, 41, 19-37.

Flower, L. S. and Hayes, J. R. A cognitive process theory
of writing. Paper presented at the Conference on
College Composition and Communication, March, 1980.

112

113

Fodor, J. A., Bever, T. G., and Garrett, M. The Psychology
of Language: An Introduction to Psycholinguistics and
Generative Grammar. New York: McGraw—Hill, 1974.

 

 

Friedman, M. and Fowler, E. Assessing elementary students'
writing skills. Paper presented at the American
Educational Research Association, San Francisco, 1979.

Halliday, M. A. K. and Hasan, R. Cohesion in English. New
York: Longman, Inc., 1976.

 

Hidi, S. and Hildeyard, A. The comparison of oral and
written discourse productions of two discourse types.
Paper presented at the American Educational Research
Association, Boston, 1980.

Hidi, S. How children cope with constraints in writing
coherent sentence sequences. Paper presented at the
American Educational Research Association, Boston, 1980.

.Hunt, K. W. Grammatical Structures Written at Three Grade
Levels. Champaign, Illinois: National Council of
Teachers, 1965. (Research Report No. 3),

Lloyd—Jones, R. Primary trait Scoring. In C. R. Cooper and
L. Odell (eds.) Evaluating Writing: Describingy
Measuring, Judging. Champaign, Illinois: National
Council of Teachers of English, 1974.

Loban, W. Language Develppment: Kindergarten Through
Grade Twelve. Champaign—Urbana, Illinois: National
Council of Teachers of English, 1976. (Research
Report No. 18). '

Mandler, J. M. and Johnson, N. S. Remembrance of things
parsed: Story structure and recall. Cognitive

National Assessment of Educational Progress. Writing
Mechanics 1966-1974. A capsule description of changes
in writing mechanics. Report No. 05. W-Ol. Denver,
Colorado: National Assessment of Educational Progress,

1975-

Paris, P., Scardamalia, M., and Bereiter, C. Discourse
schemata as knowledge and as regulators of text
production. Paper presented at the annual meeting
of the American Educational Research Association,
Boston, 1980.

Parke, M. B. Composition in primary grades. Children's
Writing: Research in Composition and Related Skills.
Urbana, Illinois: NCTE., 1961.

114

Rumelhart, D. E. Understanding and summarizing brief
stories. In D. LaBenge and J. Samuels (eds.)
Basic Processes in Readipg: Perception and
Comprehension. Hillsdale, N. J.: Lawrence
Erlbaum Associates, 1977. -

Scardamalia, M., Bereiter, C., Gartshore, S., and Cattain,
C. Locating the source of children's revision
difficulties. Paper presented at the American Edu-
cation Research Association, Boston, 1980.

Scardamalia, M. and Bereiter, C. Audience~re1atedness in
writing as an outcome of alternative composing
strategies. Paper presented at the Conference on
Writing Research, SUNY—Albany, 1980.

Stahl, A. The structure of children's compositions:
Developmental and ethnic differences. Research in
the Teaching of English, 1977, 11, 156—163.

.Stein, N. and Glenn, C. An analysis of story comprehension
in elementary school children. In R. Freedle (ed.)
New Directions in Discourse Processing (Volume II).
Advances in Discourse Processing. Norwood, N. J.:
Ablex Publishing Co., 1979.

Tamer, L. and Bond, T. Defining "good" writing in the
primary grades. Paper presented at the Conference
on Writing Research, SUNY-Albany, 1980.

Warren, W. H., Nicholas, D. W., and Trabasso, T. Event
chains and inferences in understanding narratives.
In R. Freedle (ed.) New Directions in Discourse
Processing_(Volume II): Advances in Discourse
Processing. Norwood, N. J.: Ablex Publishing Co.,

1979-

APPENDIX A

115

Table A.1

First Multivariate Analysis of
Variance of Repeated Measures

 

 

 

 

 

 

Source df Multivariate F p

Occasion 12,34 4.99 .0002
Grade-Linear 6,40 .58 .7464
Grade-Quadratic 6,40 1.55 .1868
Reading Ability 12,80 4.39 .0001
Occasion X Grade-Linear 12,34 4.21 .0005
Occasion X Grade—Quadratic 12,34 .66 .7748
Occasion X Reading Ability 24,68 I 1.10 .3665
'Grade X Reading Ability 24,140.75 1.42 .1053
Occasion X Grade X

Reading Ability 48,133 .96 .5445

Table A.2
Second Multivariate Analysis of
Variance of Repeated Measures
Source df Multivariate F .p

Occasion 12,34 5.30 .0001
Grade—Linear 6,40 .70 .6495
Grade—Quadratic 6,40 1.32 .2713
Reading Ability 12,80 5.26 .0001
Occasion X Grade—Linear 12,34 3.16 .0042
Occasion X Grade—Quadratic 12,34 1.10 .3891
Occasion X Reading Ability 24,68 1.1025 .3651
Grade X Reading Ability 24,140.75 1.2111 .2425
Occasion X Grade X

Reading Ability 48,133 1.4552 .0495