Improving Learning Disabled Students' Skills at Composing
Essays: Self-Instructional Strategy Training
ABSTRACT: This study was conducted to determine if
self-instructional strategy training would
improve learning disabled students' writing. Students were
taught a strategy designed to
facilitate the generation, framing, and planning of argumentative
essays. Training effects were
investigated using a multiple-baseline across-subjects design, with
multiple probes in baseline.
Strategy instruction had a positive effect on students'
writing performance and self-efficacy.
Effects were maintained over time and transferred to a new setting
and teacher. Evidence for
generalization to a second genre, story writing, was also obtained.
The students and their special
education teacher recommended the use of the strategy with other
students. * Learning disabled students have been characterized as having
severe and persistent writing problems (Graham & MacArthur, 1987).
Graham and Harris (1989) noted that these students have considerable
difficulty executing and monitoring many of the basic cognitive
processes central to effective writing. Of particular interest are
learning disabled students' difficulties in generating, framing,
and planning text.
In terms of generating content, students with learning disabilities
produce written texts that are inordinately short (cf. Deno, Marston,
& Mirkin, 1982; Nodine, Barenbaum, & Newcomer, 1985). Though it
is tempting to attribute their content-generation problems to a lack of
knowledge about the topics they are assigned, recent evidence suggests
that learning disabled students' problems are, in part, due to
difficulties in expressing the knowledge they have. Graham (1989) found
that simply encouraging students with learning disabilities to
"write more" doubled to tripled their output. MacArthur and
Graham (1987) reported that these students may possess more knowledge
than is reflected in their written products; learning disabled
students' dictated compositions were three to four times longer
than handwritten or word-processed papers. Englert and her colleagues
(Englert & Raphael, 1988; Englert, Raphael, & Anderson, 1986;
Thomas, Englert, & Gregg, 1987) have further argued that students
with learning disabilities have content-generation problems because they
are not particularly successful in employing strategies for
self-directed memory search.
In addition to their having difficulties in generating text, the
content generated by students with learning disabilities often falls
short of adequately meeting the purpose, conventions, and features of
the genre under consideration (Graham & Harris, 1989). To meet the
demands inherent in a particular writing task, good writers commonly
develop an ongoing frame for their text by using their knowledge of
genre patterns or other discourse schema to decide what information to
include in their papers and to regulate what will go where (Englert
& Raphael, 1988; Scardamalia & Bereiter, 1986). Students with
learning disabilities appear to be somewhat successful in using
knowledge of genre patterns to develop and frame text, as evidenced by
their ability to generate compositions that contain at least some of the
elements common to the type of writing they are assigned (cf. Graham,
1989; Graham & Harris, in press; Nodine et al., 1985; Thomas et al.,
1987). However, their knowledge is either incomplete or they are not
able to gain conscious access to all the structural knowledge they do
possess. Examinations of learning disabled students' compositions
reveal that they frequently fail to include critical elements such as
how a story ends (Graham & Harris, in press; Nodine et al., 1985) or
the premise and conclusion of their essays (Graham, 1989). Furthermore,
a considerable amount of irrelevant or nonfunctional information is
often generated (Graham, 1989). Thus, use of genre-specific knowledge to
retrieve and organize relevant information appears to be limited.
Whereas little is known about how students with learning
disabilities plan their compositions, it does not appear that much
planning is done in advance of writing. MacArthur and Graham (1987)
found that students with learning disabilities averaged less than 1
minute of time between the examiner's instructions to "plan
their papers in advance of writing" and the physical start of
writing, typing, or dictation. Similar results have been reported with
normal students (Scardamalia & Bereiter, 1986). In addition, Thomas
et al. (1987) reported that the students with learning disabilities in
their study approached writing by converting the assigned writing task
into a question-answering task, telling whatever came to mind and then
terminating their responses or answering in short choppy phrases. A
student using a planning strategy of this nature will direct little
attention to the needs of the reader, the constraints imposed by the
topic, or the whole-text organization of the written composition.
In the present study, students with learning disabilities were
taught a strategy designed to facilitate the generation, framing, and
planning of text; these are areas demonstrated by previous research to
be critical to effective writing (see Hayes & Flower, 1986) and also
to be areas of difficulty for students with learning disabilities.
Specifically, students were taught a strategy including a series of
self-directed prompts that required them to (a) consider their audience
and reasons for writing, (b) develop a plan for what they intended to
say using knowledge-of-discourse schemas or frames to generate and
organize writing notes, (c) evaluate possible content by considering its
impact on the reader, and (d) continue the process of content generation
and planning during the actual act of writing. Students were taught to
use the strategy when writing argumentative essays.
The strategy was taught using the self-instructional strategy
training procedures recommended by Graham and Harris (Graham &
Harris, 1987, 1989; Graham, Harris & Sawyer, 1987; Harris &
Graham, 1985, 1988). Self-instructional strategy training involves
interactive learning between teacher and students; the teacher initially
provides strong external support to students, with ultimate
responsibility for recruiting, applying, and monitoring the strategy
gradually being transferred to the student. Explicit explanations
regarding the rationale and the potential benefits of the strategy are
discussed, and procedures for promoting maintenance and generalization
are embedded throughout training. Students also learn self-statements
designed to support and regulate the smooth execution of the strategy.
To date, self-instructional strategy training has been successfully used
to teach students strategies for writing stories (Graham & Harris,
in press; Harris & Graham, 1985), revising essays (Graham &
MacArthur, 1988), and mathematical problem solving (Case & Harris,
Following recommendations made by Harris (1985), multiple methods
were used to evaluate the impact of the strategy and the corresponding
training procedures. First, changes in writing performance were measured
by analyzing students' essays in terms of functional and
nonfunctional elements, number of words written, cohesiveness, and
overall quality. Second, students' judgments of their ability to
write and revise a "good essay," a measure of self-efficacy,
were investigated before and after training. Third, generalization to
both a new setting and a different writing genre (story production), as
well as maintenance of effects over time, was appraised. Fourth,
validation of instructional manipulations and confirmation of mediating
responses (i.e., treatment validity) were obtained by collecting
evidence of students' use of the strategy. Fifth, the social
validity of the strategy and training procedures was assessed.
Subjects were three sixth-grade learning disabled (LD) students
(Elaine, Morgane, and Arthur) receiving resource room services in a
suburban elementary school in the northeastern United States. These
students were selected on the basis of the following stepwise criteria:
identification as LD by the school district, IQ scores between 85 and
115 on an individually administered intelligence test (Wechsler
Intelligence Scale for Children-Revised [WISC-R], or Slosson
Intelligence Test), achievement at least 2 years below age-grade level
in one or more academic areas, absence of any other handicapping
condition, and interviews with teachers indicating that significant
composition problems were evident.
In addition, the Vocabulary and Thematic Maturity subtests from the
Test of Written Language (TOWL) were administered to each student. The
standard scores for these subtests have a mean of 10 and a standard
deviation of 3 (Hammill & Larsen, 1983).
The first student, Elaine, was 12 years and 3 months of age at the
start of the study and had previously been retained in the fifth grade.
Her full-scale score on the WISC-R was 101; and her math, writing, and
knowledge achievement scores on the Woodcock-Johnson Psychoeducational
Battery (WJ) were at least 3 years below her age level. Elaine's
standard scores on the Vocabulary and Thematic Maturity subtests of the
TOWL were 9 and 8 respectively.
Morgane, the second student, was 12 years and 3 months of age and
had been retained in the first grade. Her score on the Slosson
Intelligence Test was 89, and her reading achievement score on the WJ
was 2 years below her age level. Morgane's standard scores on the
Vocabulary and Thematic Maturity subtest of the TOWL were 9 and 6,
respectively. The third student, Arthur, was 12 years and 7 months of
age and had been retained in the third grade. His full-scale score on
the WISC-R was 99, and his reading and math achievement scores on the WJ
were 2 years below his age level. His standard scores on the Vocabulary
and Thematic Maturity subtests of the TOWL were 9 and 6, respectively.
A graduate student majoring in special education served as the
instructor. Before the study began, the instructor received extensive
training and practice implementing the experimental procedures. A
notebook with detailed lesson plans containing a step-by-step
description of the instructional procedures was provided, and the
instructor was told to check each step of the daily lesson plan as it
was completed. The instructor was also directed to keep a daily log
recording (a) student and teacher comments and (b) any pertinent
observations on the learning and use of the strategy. The instructor was
introduced to each student as a special writing teacher and met with
each in a quiet room at the student's school.
Before the study began, 19 essay topics were generated; and the
students' resource room teacher was asked to evaluate each topic in
terms of its appropriateness. On the basis of the teachers'
evaluations, two topics were dropped. An example of the essay topics is:
"Do you think children should be allowed to have their own
pets?" The essay topics were randomly ordered and preassigned for
each of the planned essay probes.
Similarly, seven black-and-white pictures were selected and
evaluated. Pictures were selected so that they would be of interest to
elementary-age children and fairly easy to write about. None of the
selected pictures was evaluated negatively by the resource room teacher.
An example of the pictures is: a boy walking across a field. Pictures
were randomly ordered and preassigned for each of the story writing
On days in which the student's assignment was to write either
an essay or a story in response to a preassigned topic or picture, the
following instruction was given: "Look carefully at (the sentence
on this card/this picture) and make up a good (essay/story) to go with
it." For essay writing probes, the examiner read the topic sentence
to the student. In addition, at the start of each writing session, the
student was given two pieces of paper and two pencils. The examiner did
not provide assistance or feedback to the student on the quality of the
essay or story. However, if the student was clearly off task, the
examiner prompted her or him to keep working.
The essays and stories produced by each student were shared with a
classmate assigned by the student's teacher. This provided a real
audience and purpose for writing.
Training emphasized the student's role as an active collaborator
and incorporated principles of interactional scaffolding and Socratic
dialogue (Graham et al., 1987; Meichenbaum, 1977). Strategies were
explicitly and overtly modeled in context; the goal and significance of
the strategies were also made clear. Training was criterion based, and a
student did not progress to the next level of training until the
previous skills had been mastered. Total number of training sessions
required for Elaine, Morgane, and Arthur were seven, five, and eight,
respectively; each session was approximately 40 minutes long. The
instructional steps and generalization and maintenance components were
Pretraining. Pretraining focused on defining and generating the
components of a good essay. This process was facilitated by the use of a
small chart which provided a mnemonic (TREE) for four prompts: Note
Topic sentence, note Reasons, Examine reasons (Will my reader buy
this?), and note Ending. The mnemonic and its prompts correspond to the
general framework for arguments, which consists of premise, reasons and
data to support premise, and conclusion. After discussing the meaning of
each of the prompts, the student practiced until the mnemonic and the
meaning of each of the steps could be recited from memory. The student
also assisted the instructor in generating a premise, reasons, and a
conclusion for a particular topic.
Review of Current Performance Level. The student and instructor first
discussed the student's baseline performance on writing essays.
This included an examination of the inclusion of the basic components as
well as the quality of each component. The instructor and student then
discussed the goal of training (to write better essays), why this is
important, and how the inclusion of the basic components could improve
an essay. Each student was also asked to write and sign a goal
indicating her or his commitment to learning a strategy for writing
Description of the Composition Strategy. A small chart was used to
introduce and discuss a three-step strategy for writing good essays. The
three steps were: 1) Think, who will read this, and why am I writing
this? 2) Plan what to say using TREE (note Topic sentence, note Reasons,
Examine reasons, note Ending). 3) Write and say more. The strategy
required the writers to consider their reasons for writing the
composition, evaluate writing content by considering the potential
reader(s) of their text, generate notes for their compositions that
corresponded to the basic components of a good essay, and continue the
generation of content and planning during the actual writing of the
text. The instructor described why and how each step of the strategy was
used in planning and writing an essay.
Modeling the Strategy. The three-step strategy chart was placed in
front of the student and the instructor modeled the use of the strategy
by writing an essay while "thinking out loud." The instructor
modeled four additional types of self-instruction during the development
and writing of the essay: problem definition, planning, self-evaluation,
and self-reinforcement. Following modeling, the instructor and student
discussed the importance of what we say to ourselves while we work.
Students then generated and recorded on a small chart their own examples
of the four types of self-instruction.
Mastery of Strategy Steps. Students were required to practice the
three-step strategy and TREE until memorized. The student was allowed to
paraphrase as long as meaning remained intact. Additionally, students
practiced the self-instructions previously generated and recorded.
Controlled Practice. The instructor and student conjointly composed
an essay using the three-step strategy and self-instructional
statements. The strategy chart, as well as the student-generated
self-instruction list, were available as prompts. Although the
instructor directed and monitored the process, she did not write the
Independent Performance. The student independently composed two or
three essays using the three-step strategy and self-instructional
statements. Positive and corrective feedback was provided as needed and
transition to covert self-instruction was encouraged. The strategy chart
and self-statements were initially available as prompts, if necessary,
but then faded. Independent practice was continued until the student
demonstrated proficiency in the use of the strategy.
Generalization and Maintenance Components. Following mastery of the
strategy, the student and instructor discussed how the three-step
procedures might be adapted (e.g., eliminate TREE as a prompting device)
for use on other writing assignments. In addition, other procedures for
promoting generalization and maintenance were incorporated throughout
training: Students were asked to share what they were learning with
their teachers; the resource room teacher was asked to initial all
instructional materials and compositions; students were directed to be
prepared to use what they were learning in other settings and discussed
opportunities for, and instances of, generalization with their
instructor; and each student kept a writing folder with copies of all
training materials and compositions written.
Before the experiment began, the instructor received training in
administering writing probes, the social validation questionnaires, and
the self-efficacy measure. Practice was also provided in scoring
prewriting time, number of words, and self-efficacy. Essay elements and
coherence, as well as story grammar scores, were independently
calculated by a trained examiner. Subjective quality ratings of essays
and stories were obtained at the conclusion of the study.
Essay Elements. All essays were divided into the following minimal
parsable units: premise, reasons, conclusions, elaborations, and
non-functional (see Scardamelia, Bereiter, & Goelman, 1982).
Premise, reasons, conclusions, and elaborations were classified as
functional text; that is, they directly supported the development of the
argument. Premise represented the writer's statement of belief. To
be scored as a premise, the writer's stated belief had to be clear
without having to reference the original essay prompt. Reasons were
explanations as to why the writer believed a particular premise.
Conclusion was defined as a closing statement (e.g., a statement that
brings everything together). In addition, text could be scored as an
elaboration on a premise, reasons (e.g., an example), or a conclusion.
Finally, nonfunctional text was any unit that was repeated but had no
discernible rhetorical purpose, or any unit that was unrelated to the
argument under consideration.
Coherence. Based on the scoring procedures developed by Scardamalia
et al. (1982), all essays received a score for the longest number of
functional essay elements that were included in one of the following
coherent ordering of units: 1) premise followed by reasons with
elaborations following the initial statement to which they were related
(e.g., premise, elaboration on premise, reason 1, reason 2, elaboration
on reason 2, etc.) or 2) premise followed by reasons with elaborations
in parallel with the initial statements to which they were related
(e.g., reason 1, reason 2, elaboration on reason 1, elaboration on
reason 2, etc.). Any departure from these patterns was treated as a
break in the coherent string. Contrary to the scoring procedures for
essay elements, a premise that could not stand alone (simply stating
"No" or "Yes" to the essay question) was counted as
a functional unit.
Number of Words Written. All essays and stories were scored in terms
of number of words written. Total number of words included all written
words, regardless of spelling, that represented a spoken word.
Prewriting Time. Prewriting time was determined for all essays
(except for generalization probes). This was the amount of time between
the end of the instructor's directions to write and the actual
start of writing by the student.
Story Grammar Elements. A scale for assessing the schematic structure
of stories developed by Graham and Harris (in press) was used to score
all stories. The scale assessed the inclusion and quality of eight story
grammar elements identified by Stein and Glenn (1979): main character,
locale, time, starter event, goal, action, ending, and reaction. For
each element, a score of 0 was assigned if the element was not present,
and a score of 1 was assigned if the element was included. Highly
developed elements received a score of 2. For the story element goal, a
third point was awarded if two or more goals were present. Similarly,
for the story element, action, a third or fourth point was assigned if
the actions or events happened in a logical manner or if there was more
than one well-defined episode. By totaling the scores for the individual
elements, a total story grammar element score was calculated (19
Quality ratings. Essays and stories were scored separately using a
holistic rating scale following procedures recommended by Graham (1982).
Prior to scoring, all identifying information was removed (e.g., name,
condition, etc.), spelling and punctuation errors were corrected, and
the composition was typed.
For both essays and stories, two undergraduate students who were
unfamiliar with the students, design, or purpose of the study
independently scored the compositions. They were told to read each paper
attentively but not laboriously to obtain a general impression of
quality. Papers were scored on a 1- to 8-point scale, with 1
representing the lowest quality of writing and 8 representing the
highest quality. Examiners were provided with a representative sample of
a low-, medium-, and high-scoring essay and story to use as guides or
anchor points. For each paper, the scores of the two examiners were
Self-Efficacy. A scale for assessing students' self-efficacy
concerning their ability to write an essay was administered before and
following training. According to Schunk (1985), self-efficacy refers to
personal judgment regarding one's performance capabilities in a
specific situation. The scale used in this study was developed in
accordance with guidelines described by Bandura and Schunk (1981) and
Schunk (1983). The scale ranged from 10 to 100 in 10-unit intervals from
"not sure" (10), to "maybe" (40), to "pretty
sure" (70), to "real sure" (100). Students were
familiarized with the different numerical values on the scale by judging
their certainty of successfully jumping progressively longer distances.
Five questions were used as stimulus items. The questions measured the
students' perceived ability to write an essay that had a
"good" beginning, that gave three reasons to support the
premise, and had a "good" ending. In addition, students were
questioned concerning their perceived ability to locate errors in their
writing and revise their essays so that they were "better."
The scale was administered by reading each question aloud; the student
was advised to be honest, and then privately circled an efficacy value
on the 10- to 100-point scale.
Strategy Usage. Because the use of the three-step strategy required
that the student generate notes in response to the second step, it was
possible to obtain evidence on strategy usage by examining the papers
students wrote on during their writing probes.
Social Validation. At the conclusion of the study, the resource room
teacher and the three students were interviewed separately by the
instructor to obtain information on perceived effectiveness of the
intervention, recommendations, and other feedback. In addition, the
resource room teacher was queried regarding evidence of strategy
The effects of strategy training on essay writing skills were
assessed through the use of a multiple-baseline across-subjects design,
with multiple probes in baseline. The following conditions were
Baseline. During baseline, pretreatment response rates on essay
writing behaviors were established. Throughout baseline, general
procedures were in effect.
Training. Training was initiated for the first student, Elaine, after
a stable baseline was established in terms of total number of essay
elements. Training was continued until the student was able to write
essays that contained all the basic parts and demonstrated independent
mastery of the strategy. Training was not initiated for the second
student, Morgane, until the first student's posttreatment essay
performance reached a criterion level of at least one and a half times
the mean number of elements produced during baseline. Identical
procedures were followed when introducing and terminating treatment with
the third student, Arthur.
Posttreatment Essay Probes. Three to four posttreatment essay probes
were administered during the first 2 weeks immediately following the
termination of treatment. General procedures were in effect.
Maintenance Probes. For the first student, Elaine, three maintenance
writing probes were collected. These occurred at 6, 11, and 12 weeks
following the termination of training. For the second student, Morgane,
one follow-up maintenance probe was collected 3 weeks following
training. General procedures were in effect. No follow-up probes were
administered to the third student, Arthur, because the end of the school
year precluded the collection of these probes.
Setting Generalization Probes. Before training and during the
posttreatment phase, a generalization writing essay probe was collected
from each student by the resource room teacher. Procedures for
collecting these probes were identical to baseline conditions.
Task Generalization Probes. To determine if the strategy training
procedures resulted in transfer to a different genre, multiple story
writing probes were collected during baseline and following the
administration of posttreatment essay probes. General procedures were in
effect for all generalization story probes.
For two students, Elaine and Morgane, an additional instructional
session was initiated following their first and second task
generalization probes, respectively. The instructional session was
initiated when it became clear that these two students did not
demonstrate adequate transfer from essay writing to story writing.
During the instructional session, each student received practice in
using the mnemonic SPACE (note: setting, purpose, action, conclusion,
and emotion) during the second step of the strategy (Plan).
Reliability of Measures
At the conclusion of the study, 25% of the essays were randomly
selected for rescoring. All identifying information was removed and an
examiner unfamiliar with the purpose and design of the study was trained
to score the essays. To determine interrater reliability, Pearson
product-moment reliability coefficients were calculated between the
scores assigned by the instructors and independent examiner. Interrater
reliabilities were as follows: total number of elements (.89), premise
(.77), reasons (.79), conclusion (1.00), elaborations (.83),
nonfunctional (.97), coherence (.83), and number of words (.99).
Similarly, 29% of stories were randomly selected for rescoring at
the end of the study. Identifying information was removed, and stories
were scored by the examiner who scored the reliability essays.
Interrater reliabilities for story grammar elements and number of words
were .81 and 1.00, respectively. The Pearson product-moment correlation
coefficients between the quality-rating scores assigned by the two
undergraduate students were .77 for essays and .83 for stories.
Elements. Figure 1 presents the total number of functional elements
(premise, reasons, conclusions, and elaborations) for each essay written
by each student during all phases of the study. All three students
evidenced low levels of performance throughout baseline. Elaine,
Morgane, and Arthur's mean baseline response rates in terms of
total number of functional elements were 2.3, 3.5, and 3.8,
respectively. Following training, Elaine's mean performance on the
posttreatment essay probes was 7.0, clearly exceeding the previously
established baseline level. Average performance for Elaine on the
maintenance probes (5.7) continued to be considerably higher than
The second student's (Morgane's) mean performance on the
posttreatment essay probes was 6.25, and she had a score of 7.0 on the
maintenance essay administered 3 weeks following training. The third
student, Arthur, evidenced the largest gains following training; his
mean performance on the posttreatment essays was 8.7. Posttreatment
performance of the LD students in terms of number of functional elements
is equivalent to or exceeds that of normal students as reported by
Scardamalia et al. (1982). Thus, stable baselines were established, and
the sequential application of the instructional procedures resulted in
substantial increases in average number of functional essay elements;
these gains were maintained over time for at least two of the three
For all three students, treatment effects in terms of total number
of functional essay elements generalized from the training setting to
the students' resource room. Before training, the number of
functional elements written by Elaine, Morgane, and Arthur when
composing in the resource room was 3, 3, and 4, respectively. Following
training, all students at least doubled their production of functional
essay elements on the generalization essays administered in the resource
room. Elaine, Morgane, and Arthur's scores on this essay were 7, 6,
and 8, respectively.
It is also important to examine the distribution of specific essay
elements before and after treatment. During baseline, 45% of all
elements were classified as nonfunctional, i.e., material unrelated to
the writer's argument or the repetition of textual material.
Following training, the proportion of nonfunctional elements dropped to
15%. In contrast, training resulted in increases for three of the four
functional elements. During baseline, 29% of all elements were reasons,
and premise and conclusions each accounted for only 3% of the elements.
For essays written after training, 41% of elements were reasons, and
premise and conclusions accounted for 11% and 12% of all elements,
respectively. No changes were noted in the proportion of elaborations;
20% of essay elements prior to and following training were elaborations.
The impact of training can also be gauged by examining the
proportion of essays written before and after training that contained
all the basic components of an essay: premise, reasons, and conclusion.
Before treatment only 7% of essays contained all of these elements,
whereas following training 82% met this requirement.
Coherence. The mean coherence scores for Elaine, Morgane, and Arthur
on the baseline essays were 2.7, 4.0, and 3.2, respectively. For two of
the students, coherence scores on the posttreatment essays increased
substantially; Elaine's average score was 4.7 and Arthur's was
7.0. Furthermore, Elaine's average coherence score on the
maintenance probes (6.0) exceeded both her baseline and posttreatment
performance. Nonetheless, Morgane did not evidence a sizable increase
(over baseline levels) in terms of average posttreatment coherence
scores (4.5). She received a score of 7.0, however, on the one
maintenance probe she completed. Finally, all three of the
students' generalization essays evidenced higher coherence scores
following training. Before training, Elaine, Morgane, and Arthur's
coherence scores on essays written in the resource room were 1, 3, and
4, respectively. On the essay written in the resource room following
training, Elaine's coherence score increased to 7, Morgane's
to 6, and Arthur's to 8.
Number of Words. The effects of training on the mean length of essays
were not consistent across subjects. For Elaine, the mean length of
baseline essays was 40.3. Her posttreatment essays, however, averaged 80
words in length, whereas her maintenance essays averaged 93.7 words.
Similarly, Arthur's baseline essays averaged 71.6 words, and his
posttreatment essays increased to an average of 88.8 words. In contrast,
Morgane averaged 52 words on her baseline probes, but her posttreatment
performance dropped to 46 words per essay; she wrote only 43 words on
her maintenance probe. All three students evidenced an identical pattern
on the generalization essay probes collected in the resource room. It
should be noted that number of words written was not a target of
Prewriting Time. All students averaged less than 12 s of prewriting
time on their baseline probes. Following training, average prewriting
time increased for all three students. Elaine and Morgane's average
prewriting times on posttreatment essays were approximately 8 and 9 min,
respectively; both spent approximately 6 min on their maintenance
probes. Arthur had approximately 8 min of prewriting time on his first
posttreatment essay, but prewriting time on his final two essays was
only 5 and 10 s, respectively.
Quality Ratings. The mean holistic quality ratings for Elaine,
Morgane, and Arthur's baseline essays were 2.3, 2.5, and 3.3,
respectively. Following the completion of training, mean quality ratings
for the posttreatment essays rose to 6.2 for Elaine, 4.0 for Morgane,
and 6.0 for Arthur. Improvement in essay quality was maintained over
time; Elaine averaged a mean quality score of 5.5 on the maintenance
probes, and Morgane received a score of 5.0 on the maintenance probe she
completed. Furthermore, in all cases, the generalization probe taken
after training received a quality rating 3 to 4 1/2 points higher than
that received by the probe taken during baseline.
Generalization to Stories
Story Grammar Elements. Figure 2 presents the students' scores
on the story grammar scale for each generalization story written during
baseline and after the administration of posttreatment essays. Elaine,
Morgane, and Arthur's mean performance scores on baseline stories
in terms of the story grammar scale were 7.0, 4.0, and 7.3,
respectively. With the exception of Arthur's first baseline story,
all subjects' baseline stories were missing three or more of the
basic story grammar elements identified by Stein and Glenn (1979).
Following the completion of training and the posttreatment essays,
two students improved in their performance on the story grammar scale.
Morgane's average score (based on the first two stories) increased
to 9.5, whereas Arthur's mean score rose to 12.0. Though
Arthur's posttraining stories contained either all or all but one
of the basic story grammar elements, this was not the case for Morgane.
Her stories consistently lacked three elements: starter event, ending,
Elaine's first story following the completion of training and
the posttreatment essays evidenced a decline in overall performance
(story grammar score of 5). In addition, four basic elements were
notably missing from her story; and it appeared that she attempted, at
least in part, to use the mnemonic TREE as an aid in generating story
content. Consequently, an additional instructional session was initiated
with both Elaine and Morgane; each student received practice using a
mnemonic (SPACE) for generating content related to common story
Following this session, Elaine's stories improved
dramatically; her average story grammar score was 13.0, and her stories
contained all the common elements. Although Morgane's performance
dropped slightly (8.0), her story contained all but one of the basic
Number of Words. Elaine, Morgane, and Arthur's mean length of
baseline stories was 40, 61.5, and 97.7 words, respectively. Following
the completion of training and the posttreatment essays, all
students' stories increased in length; Elaine's first story
was 59 words long, and Morgane's (based on her first two stories)
and Arthur's stories increased to 92.5 and 118 words, respectively.
Following the additional instructional session, Elaine evidenced a
dramatic increase in story length ([M.sub.x] = 151). Morgane, however,
produced a story (36 words) that was shorter than the other stories that
Quality Ratings. Elaine, Morgane, and Arthur's mean quality
ratings for baseline stories were 1.0, 1.3, and 4.0 respectively.
Following the completion of training and posttreatment essays,
Arthur's stories evidenced a modest increase to 5.5 in story
quality. Elaine's first story, in contrast, did not improve, and
Morgane's stories improved only slightly to 1.8. Following the
additional instructional session, the quality of Elaine's stories
([M.sub.x] = 5.3) were judged to be much improved, whereas
Morgane's story quality increased only slightly to 2.0.
Following training, Elaine and Morgane used the mnemonic TREE to help
generate notes on all essay probes, including posttreatment,
maintenance, and generalization. In contrast, Arthur made notes only for
the first posttreatment essay following training. When asked why he had
stopped making notes in advance of writing, he replied "I
don't need to make a planning sheet because I have the steps in my
Although his prewriting time was much shorter when the mnemonic was
not written out in advance of writing, measures of content and quality
were only slightly affected. Two students, Elaine and Morgane, were
taught to use SPACE in the second step of the strategy. Both used the
mnemonic each time they were directed to write a story following
During baseline, Elaine, Morgane, and Arthur's scores on the
self-efficacy measure were 64, 52, and 88, respectively. Following
training, Elaine evidenced a small, insignificant 4-point increase on
the self-efficacy measure, whereas Morgane and Arthur's scores
increased by 18 and 12 points, respectively.
Social Validity Interviews
All three students indicated that they believed the strategy helped
them to write better. Comments suggested that the strategy made the task
of writing essays easier and provided a helpful device for organizing
content. Similarly, all thought that the strategy should be taught to
their friends because it would improve their writing. The only negative
comments made were that the training procedures were "boring
sometimes" (one student) and having to make notes and write the
essay was too much (one student).
The resource room teacher reported that she felt that the
instructional procedures had a positive impact on the writing of all
three students and that she would recommend the use of the strategy with
other LD students. She further indicated that two students now had a
better understanding that "writing is a process" and were
asking about using the strategy with specific classroom assignments.
In the present study, LD students were taught a strategy for planning
and writing essays. On the essay probes administered during the 2 weeks
immediately following training, all three students participating in the
study evidenced substantial gains over baseline levels in several
distinct areas of writing performance.
First, training resulted in a positive change in the number of
functional elements students included in their essays. Essays written
after training contained, on the average, approximately two to three
times as many functional elements as those produced during baseline.
Second, and equally important, changes occurred in terms of the types of
essay elements students included in their compositions.
Following training, the proportion of nonfunctional essay elements
in students' papers dropped considerably, and the proportion of
elements classified as premise, reason, and conclusion increased. The
magnitude of change in the construction of students' essays can be
gauged by the finding that four out of every five of the papers written
after training contained all the basic parts of an essay; less than 10%
of baseline essays met this same criteria. Third, essays written
following training were judged to be qualitatively superior to those
written during baseline.
Gains in two other areas of writing performance were also
demonstrated by Elaine and Arthur. These students' posttreatment
essays were, on the average, longer than the papers they composed during
baseline. Similarly, these same two students evidenced a substantial
increase, following training, in the average length of the longest
number of functional essay elements that were coherently ordered. Thus,
training in the use of the target strategy allowed these students not
only to produce more content (both words and functional elements) but
also to arrange the generated content in a meaningful and coherent
In addition to improved essay writing performance, Morgane and
Arthur demonstrated a concomitant change in self-efficacy following
training. They became more confident in their ability to write and
revise a "good essay." Equally notable was the relatively high
score among all students on this self-efficacy measure during the
pretest ([M.sub.x] = 68). The students clearly overestimated their
composition abilities. This finding adds to a growing body of literature
that indicates that learning disabled students have difficulty
accurately assessing or predicting their performance capabilities
(Brown, Bransford, Ferrara, & Campione, 1983; Graham & Harris,
in press; Harris, Graham & Freeman, 1988).
Self-instructional strategy training has been recommended as a
particularly promising means for achieving maintenance and
generalization (Harris, 1982). Maintenance data was collected for Elaine
and Morgane. Both students were able to maintain the gains they had
demonstrated immediately following training; there was little difference
between their maintenance and posttreatment essays in terms of total
number of functional elements and overall quality. Furthermore, their
maintenance probes generally had longer coherent strings of functional
elements than their posttreatment essays, and one of the students
generated more content when writing the maintenance essays.
Two methods were used to assess generalization. First,
generalization across settings and teachers was assessed by having the
students' special education teacher administer in the resource room
an essay probe during baseline and immediately following training. For
all three students, the generalization essay written after training was
superior to the baseline probe in terms of overall quality, total number
of functional elements, and coherence. Elaine and Arthur also evidenced
an increase in the amount of content that they generated.
Generalization across tasks involved having students write stories
both before and following the completion of treatment. Though students
briefly discussed during the course of treatment how the strategy might
be used with other writing assignments, direct instruction in adapting
the strategy for this purpose was not provided. Thus, students'
performance on the story probes provided a procedure for assessing
generalization to a different writing genre, narrative stories. Evidence
on the transfer effects of training to story writing were mixed.
Consequently, Elaine and Morgane received one training session where
they were taught a mnemonic for helping them generate content for each
of the elements of a story and received practice in using the mnemonic
in concert with the other basic steps of the strategy. Following this
one training session, both students' stories contained all or all
but one of the basic story elements. This training session also had a
powerful effect on the length and quality of Elaine's stories.
Thus, although limited generalization across writing tasks
occurred, a single, brief generalization training session was
successful. Such results in terms of generalization across tasks, which
is more difficult to establish than generalization across settings, are
encouraging. However, additional modifications and some instruction may
be necessary if students are to realize the potential effects of this
strategy across different types of writing tasks. For example, the basic
steps of having students consider their purpose and audience, generate
notes in advance of writing, and continue content generation and
planning during text production can be used with a variety of classroom
writing tasks. Teachers may find the procedure to be more effective,
however, if they provide students with aids or procedures that will help
them generate content relevant to the genre under consideration, as was
done with the TREE and SPACE aids in the present experiment.
Treatment validity was assessed by examining any planning notes
written by students during the process of writing. Evidence of strategy
usage provides additional confirmation that the instructional
manipulations were, in fact, responsible for changes in students'
writing behavior. Following training, Elaine and Morgane demonstrated
evidence of strategy usage on every essay they wrote. Arthur used the
essay mnemonic to generate notes just once following training, but
indicated that the strategy had been internalized and, as a result, it
was not necessary to make notes in advance of writing. His decision not
to overtly use the strategy did not have much of an effect on the
content or quality of writing. It should also be noted that when
students overtly used the strategy there was a corresponding increase in
the amount of time they spent planning before actually writing their
paper. Finally, evidence of social validity was established, as each of
the students and their special education teacher indicated that use of
the strategy improved writing performance and that the strategy should
be taught to other students.
The present study adds to a growing body of literature supporting
the use of self-instructional strategy training procedures as an
effective means for improving LD students' writing performance and
other academic skills (cf. Graham & Harris, 1989; Graham et al.,
1987; Case & Harris, 1988). Future research needs to document the
contributions of specific components of self-instructional strategy
training and to continue to address issues related to maintenance and
generalization (Harris, Wong, & Keogh, 1985). [Figures 1 to 2
STEVE GRAHAM and KAREN R. HARRIS are Associate Professors, Department
of Special Education, University of Maryland, College Park.