The Effects of Peer-Assisted Sentence-Combining Instruction on the Writing Performance of More and Less Skilled Young Writers

Abstract

Mastering sentence-construction skills is essential to learning to write. Limited sentence-construction skills may hinder a writer's ability to translate ideas into text. It may also inhibit or interfere with other composing processes, as developing writers must devote considerable cognitive effort to sentence construction. The authors examined whether instruction designed to improve sentence-construction skills was beneficial for more and less skilled 4th-grade writers. In comparison with peers receiving grammar instruction, students in the experimental treatment condition became more adept at combining simpler sentences into more complex sentences. For the experimental students, the sentence-combining skills produced improved story writing as well as the use of these skills when revising. (PsycINFO Database Record (c) 2012 APA, all rights reserved)

Full text

The Effects of Peer-Assisted Sentence-Combining Instruction on the
Writing Performance of More and Less Skilled Young Writers
Bruce Saddler
University at Albany, State University of New York
Steve Graham
Vanderbilt University
Mastering sentence-construction skills is essential to learning to write. Limited sentence-construction
skills may hinder a writer’s ability to translate ideas into text. It may also inhibit or interfere with other
composing processes, as developing writers must devote considerable cognitive effort to sentence
construction. The authors examined whether instruction designed to improve sentence-construction skills
was beneficial for more and less skilled 4th-grade writers. In comparison with peers receiving grammar
instruction, students in the experimental treatment condition became more adept at combining simpler
sentences into more complex sentences. For the experimental students, the sentence-combining skills
produced improved story writing as well as the use of these skills when revising.
Current educational reform, as reflected in the No Child Left
Behind Act of 2001 (NCLB), places considerable emphasis on
improving children’s reading and mathematics performance. Un-
fortunately, other important skills such as writing are not empha-
sized, even though these skills are critical to school success.
Writing, for instance, is the primary means by which students
demonstrate their knowledge in school and the major instrument
that teachers use to evaluate academic performance (Graham &
Harris, in press). It provides a flexible tool for gathering, remem-
bering, and sharing subject matter knowledge (Durst & Newell,
1989) as well as an instrument for helping children explore,
organize, and refine their ideas about a specific subject (Applebee,
1984). It also offers a medium where students can explore their
interests, feelings, and experiences as well as the artistic and
creative aspects of self-expression.
One possible reason for not emphasizing a skill such as writing
in current educational reforms such as NCLB is that schools
already do a good job of teaching it. That is not the case, however.
In both the 1998 and 2002 National Assessment of Educational
Progress, the majority of students in 4th, 8th, and 12th grade
demonstrated only partial mastery of the writing skills and knowl-
edge needed at their respective grade level (Greenwald, Persky,
Campbell, & Mazzeo, 1999; Persky, Daane, & Jin, 2003). As a
result, many students are starting college with poorly developed
writing skills. One fifth of college freshmen are required to take a
remedial writing class, and more than one half of new college
students are unable to write a paper relatively free of errors
(Intersegmental Committee of the Academic Senates, 2002).
In an effort to focus national attention on this crises in writing,
the College Board, an organization of more than 4,300 colleges,
established the National Commission on Writing in America’s
Schools and Colleges in 2002. Their report, The Neglected “R,”
highlighted the need to place writing squarely in the center of the
school-reform agenda, noting that the writing of students in the
United States “is not what it should be” (National Commission on
Writing, 2003, p. 7). Although this report provided suggestions for
improving writing instruction, these recommendations were
sketchy and incomplete. The commission’s decision to focus its
efforts on establishing the need for action is understandable, as a
transformation in writing practices is unlikely to occur if the public
and policy-making communities are not aware of the scope and
depth of the problem. Nevertheless, an essential element in devel-
oping a comprehensive writing policy is the identification of
effective instructional procedures, not just at the secondary level
(where the report concentrates most of its attention) but with
younger students as well. This is important for two reasons. First,
providing effective writing instruction to children during the ele-
mentary grades should maximize the writing development of chil-
dren in general. Second, it should minimize the number of children
who experience difficulties learning to write (Graham & Harris,
2002).
In the present study, we examined the effectiveness of an
intervention for improving a basic foundational writing skill, sen-
tence construction. The study was conducted with children in
fourth grade who were more and less skilled writers, both in terms
of their abilities to construct sentences as well as their overall
ability to write. We decided to focus our efforts on sentence-
construction skills for two reasons. First, in the seminal work
conducted by Hayes and Flower (1986), sentence generation was
one of the three major processes (in addition to planning and
revising) that skilled writers used as they composed. This involved
transforming ideas and intentions into written sentences that make
sense and conform to the permissible syntax of the language.
Limited knowledge about effective writing formats at the sentence
level may hinder a writer’s ability to translate his or her thoughts
Editor’s Note. Michael Pressley served as the action editor for this
article.—KRH
Bruce Saddler, Department of Educational and Counseling Psychology,
University at Albany, State University of New York; Steve Graham,
Department of Special Education, Vanderbilt University.
Correspondence concerning this article should be addressed to Bruce
Saddler, Department of Educational and Counseling Psychology, Univer-
sity at Albany, State University of New York, Education 226, 1400
Washington Avenue, Albany, NY 12222-0001. E-mail: bsaddler@
uamail.albany.edu
Journal of Educational Psychology Copyright 2005 by the American Psychological Association
2005, Vol. 97, No. 1, 43–54 0022-0663/05/$12.00 DOI: 10.1037/0022-0663.97.1.43
43
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

into text, constraining the richness and quality of the writing.
Difficulties constructing well-designed, grammatically correct sen-
tences may also make the text more difficult for others to read.
Teachers’ evaluations of the quality of students’ text are influ-
enced by sentence structure (Freedman, 1979).
Second, the process of constructing formal sentences is quite
complex, as the writer must deal with a number of demands,
including word choice, syntax, textual connections, clarity,
rhythm, and so forth. This process of translating ideas into written
text requires considerable effort (as measured by the time it takes
to react to a secondary task), even for college-age students
(Kellogg, 1987). Presumably, the mental load imposed by sentence
construction is even higher for children, as they have less volun-
tary control than adults over their syntactic choices and the process
of translation is also less fluent (Scardamalia & Bereiter, 1986).
Thus, sentence-production skills may inhibit or even interfere with
other composing processes during the act of writing (Graham, in
press; Scardamalia & Bereiter, 1986; Strong, 1986). For example,
if a writer has to devote considerable attention to the process of
sentence construction, this depletes the cognitive resources avail-
able for other important writing processes such as planning or
revising. Furthermore, interference may occur as the writer
switches attention from a writing process, such as planning, to
crafting a sentence, as plans or ideas being held in working
memory may be lost or forgotten.
We anticipated that efforts to improve sentence-construction
skills would be beneficial for both less and more skilled young
writers. Less skilled writers often have considerable difficulty with
sentence construction, as they typically produce relatively short
and syntactically simple sentences that often contain errors involv-
ing grammar, usage, and vocabulary (Houck & Billingsley, 1989;
Morris & Crump, 1982; Myklebust, 1973; Newcomer & Baren-
baum, 1991). Although more skilled writers produce sentences that
are longer, more syntactically correct, contain fewer errors, and are
qualitatively better than those produced by their less skilled coun-
terparts (Graham & Harris, 1988; Houck & Billingsley, 1989;
Newcomer & Barenbaum, 1991), their sentence-construction skills
are still developing (Hunt, 1965; Scardamalia & Bereiter, 1986).
Consequently, efforts designed to improve sentence-writing skills
of these children should be beneficial too (cf. Combs, 1975;
O’Hare, 1973).
The intervention used in the present study directly taught stu-
dents how to construct more complex and sophisticated sentences
by combining two or more basic (i.e., kernel) sentences into a
single sentence (Ney, 1981; Strong, 1976). This instructional
method, referred to as sentence combining, is not only designed to
teach students how to craft more syntactically complex sentences
but to produce better sentences, ones that more closely convey the
writer’s message. The assumed benefits of sentence combining rest
on three principles. One, students need to be taught the concept of
sentence formulation and the syntactic options that are available to
them when producing a sentence (Neuleib & Fortune, 1985;
Rhodes & Dudley-Marling, 1996). Two, once the process of sen-
tence formation and reformation becomes more habituated or
automatized as a result of sentence-combining instruction, the
overall cognitive strain of writing should be reduced (Graham,
1982), allowing writers to devote additional cognitive resources to
other writing processes. Three, the process of combining sentences
is primarily a revising skill (Zimmerman & Kitsantas, 2002),
providing writers with another tool for recrafting already written
text.
Although sentence-combining instruction is not currently in-
cluded as a component in popular approaches to writing instruc-
tion, such as Writers’ Workshop (Calkins, 1986; Pritchard, 1987),
it can be used to improve students’ sentence-construction skills
(Hillocks, 1986; O’Hare, 1973). Most of this evidence, however,
has been collected with students in middle school, high school, and
college, and the impact of sentence-combining instruction on im-
proving the overall quality of students’ writing has been mixed
(see Hillocks, 1986). In addition, few investigations have exam-
ined whether the skills that students learn when combining sen-
tences are used when they revise their text (see Horstman, 1989,
for an exception). Examining the effects of sentence combining on
writing quality and revising with young and less skilled writers is
especially important, as these students often have difficulty gen-
eralizing new skills (Graham, Harris, & Mason, in press).
The present study extended previous intervention research on
sentence combining in several important ways. First, we integrated
sentence combining with another effective instructional method,
peer-assisted learning strategies (PALS). This approach has been
used successfully to improve young students’ reading, writing, and
math skills (D. Fuchs, Fuchs, Mathes, & Simmons, 1997; D.
Fuchs, Fuchs, & Thompson, 2001; Utay & Utay, 1997). With
PALS, a stronger and weaker student are paired together to prac-
tice applying a target skill, and each student alternatively acts as
the coach or tutor as the other child applies the procedure. Thus,
the less competent as well as the more competent student acts as
both tutor and tutee. PALS support learning through active aca-
demic responding, collaborative practice, and immediate feedback
and assistance from a peer (D. Fuchs et al., 2001; Greenwood,
Carta, & Kamps, 1990).
In this investigation, the instructor first explained and modeled
how to apply a specific sentence-combining strategy. The student
pair then practiced applying the procedure under instructor guid-
ance. This was followed by the student pair alternatively acting as
tutor and tutee (i.e., PALS), as they independently practiced the
sentence-combining strategy. Although PALS has been an effec-
tive instructional tool for teaching reading and math skills (see,
e.g., D. Fuchs et al., 1997; L. Fuchs, Fuchs, Phillips, Hamlett, &
Karns, 1995; L. Fuchs, Fuchs, & Kazdan, 1999), it has not been
applied to the teaching of sentence-combining and other writing
skills, with the exception of spelling (see, e.g., Maheady, Harper,
Mallette, & Winstanley, 1991).
In contrast to previous studies (with the exception of Horstman,
1989, with high school students), this investigation also provided
children in the experimental condition with opportunities to prac-
tice applying the sentence-combining strategies they were taught
as they revised material they previously wrote. They were specif-
ically directed to apply the sentence-combining procedures as they
revised a conjointly written paper with their partner and revised a
paper of their own independently. As Kendall (1989) noted, good
instruction does not ensure that learned skills are applied to other
pertinent tasks. This is especially true for children who experience
difficulty learning academic tasks such as writing (Wong, 1994).
First, we hypothesized that students in the experimental group
would be more adept at combining sentences following instruction
44 SADDLER AND GRAHAM
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

than children in a comparison group who received grammar in-
struction (an approach commonly used to improve the syntactic
complexity of students’ writing). Sentence-combining skills were
assessed with a series of progress-monitoring tests administered
during instruction as well as a norm-referenced test administered
after treatment ended. Second, we predicted that sentence-
combining instruction would have a positive impact on the quality
of students’ writing. A traditional holistic measure (Cooper, 1977)
was used to assess the writing quality of stories students wrote
(i.e., first draft) and revised (i.e., second draft) before and after
treatment. Although the effects of sentence-combining instruction
on the quality of students’ writing have been mixed in previous
research (Hillocks, 1986), we anticipated that experimental stu-
dents’ improvements in their sentence-combining skills would
enhance writing quality. Learning how to better construct sen-
tences should reduce cognitive load during writing, freeing up
resources that can be applied to other important writing processes
(Graham, 1982), and sentence structure influences perceptions of
writing quality (Freedman, 1979). Third, no prediction was made
about changes in the length of students’ writing, as sentence-
combining instruction might reduce length (e.g., text such as “The
big dog ran” and “The dog was black” might be combined into the
shorter sentence “The big black dog ran”) or increase it (e.g.,
attentional resources that were previously devoted to sentence
construction could alternatively be devoted to content generation).
Fourth, we predicted that when students in the experimental con-
dition revised the first draft of their stories, they would be more
likely than their counterparts in the comparison condition to make
revisions involving the sentence-construction skills they were
taught. The only exception to this prediction was for revisions that
involved the conjunction and. Fourth-grade students already use
this conjunction quite often when they write, and we did not expect
that sentence-combining instruction would increase the number of
revisions involving this skill.
Method
Participants
Screening. A two-step process was used to select participants. First, all
fourth-grade students from nine classrooms across three schools located in
the Washington, DC, metropolitan area were administered Form A of the
Sentence Combining Subtest from the Test of Written Language—3
(TOWL–3; Hammill & Larsen, 1996). This measure assesses a child’s
ability to integrate the meaning of several short sentences that address a
related topic into a single grammatically correct sentence (coefficient
alphas ranged from .83 to .84 for 9- to 11-year-olds). The measure was
used to identify students who were skilled and less skilled at sentence
combining (referred to as more skilled and less skilled writers, respec-
tively). Less skilled writers scored one standard deviation or more below
the mean for the normative sample, whereas skilled writers’ scores were at
or above the mean of this subtest (see Table 1 for means and standard
deviations). Second, teachers were asked to confirm that students who
scored at or above the mean on this subtest were not experiencing writing
difficulties and that students who scored one or more standard deviations
below the mean were experiencing difficulty with writing. Using these
procedures, we identified 44 students (22 more skilled and 22 less skilled
writers). A one-way analysis of variance (ANOVA) confirmed that more
skilled writers had higher sentence-combining scores than did less skilled
writers, F(1, 42) ⫽227.4, MSE ⫽153.5, p⬍.01.
Groups. The 44 students were randomly assigned to two conditions,
sentence combining or grammar instruction, so that there was an equal
number of more and less skilled writers in each treatment at each school.
This allowed us to pair each less skilled writer with a more skilled writer
in each instructional condition. Within a school, a more and less skilled
writer within a treatment condition were paired together on the basis of
teachers’ judgments about their compatibility and scheduling issues. Be-
fore the study was completed, a less skilled writer in the sentence-
combining condition moved. This child was replaced by another student for
the remaining instructional sessions, but data for the replacement student
were not included in any of the analyses. Thus, each group of more and less
skilled writers for the two conditions contained 11 participants, with the
Table 1
Student Characteristics by Treatment Condition and Type of Reader
Variable
Sentence combining Grammar
More skilled Less skilled More skilled Less skilled
MSDMSDMSDMSD
TOWL–3
Sentence Combining 10.4 0.5 6.7 0.7 10.9 1.4 7.0 0.3
Composite score 9.9 1.2 7.6 0.9 10.0 1.9 7.8 1.1
WJ-III
Letter–Word Identification 104.5 16.6 100.9 19.0 99.2 10.3 105.5 13.9
Reading Fluency 103.6 14.5 101.7 7.2 98.6 5.6 100.6 13.8
CELF–R: Formulated Sentences 9.1 2.4 9.0 2.2 11.1 3.4 10.4 3.1
Chronological age 9.2 0.3 9.5 0.7 9.3 0.4 9.3 0.4
Gender
Male 7 4 3 7
Female 4 6 8 4
Note. For all students, the mean for the Test of Written Language—3 (TOWL–3) Sentence Combining subtest
was 10 (SD ⫽3); the mean for the TOWL–3 composite score, which is based on the average score of the
Language Conventions and Story Construction subtests, was 10 (SD ⫽3); the mean for the Woodcock-Johnson
III Tests of Achievement (WJ–III) was 1,000 (SD ⫽15); and the mean for the Clinical Evaluation of Language
Fundamentals—Revised (CELF–R) was 10 (SD ⫽3).
45
SENTENCE COMBINING AND WRITING PERFORMANCE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

exception of less skilled writers receiving sentence-combining instruction
(N⫽10).
Additional evidence that these more and less skilled writers differed in
their writing abilities before the start of instruction was obtained by
administering two other subtests, Contextual Language and Story Con-
struction, from the TOWL–3 (Hammill & Larsen, 1996). The first subtest
measures a student’s ability to construct grammatically correct sentences
and use appropriate vocabulary (coefficient alphas ranged from .77 to .78
for 9- to 11-year-olds), whereas the second measure assesses a child’s skills
in writing a story (coefficient alphas ranged from .88 to .90 for 9- to
11-year-olds). As recommended in the TOWL–3 manual, we obtained a
composite score for these two tests by computing the average of the
standard score for these two tests (see Table 1 for means and standard
deviations). A two-way ANOVA, with treatment and student type as the
independent variables, showed that more skilled writers obtained higher
scores than less skilled writers, F(1, 39) ⫽5.1, MSE ⫽22.6, p⫽.03, but
that there was no statistically significant difference between treatment
conditions, nor was the interaction statistically significant.
At the start of the study, the mean age of the participating students was
9 years 3 months (range ⫽9 years 0 months to 11 years 2 months). There
was an almost equal number of boys (n⫽21) and girls (n⫽22). The racial
demographics of students were diverse (and consistent with the makeup of
the participating schools), as 25 students were African American (58%), 12
were White (28%), 4 were Hispanic (9%), and 2 were Asian (5%). English
was the primary language for all students. Although the schools would not
provide information on the socioeconomic status of the participating stu-
dents, the percentage of students receiving free or reduced meals in each
school was available. This ranged from 1% to 49%. It should be noted that
there were no statistically significant differences between students in the
two treatment conditions of more and less skilled writers for either age or
gender (see Table 1).
Because the effectiveness of the experimental treatment condition, sen-
tence combining, depended in part on students’ reading and oral language
skills, we assessed pretreatment performance in each of these areas. Read-
ing was assessed with the Letter–Word Identification and Reading Fluency
subtests from the Woodcock–Johnson III Test of Achievement (WJ-III;
Woodcock, McGrew, & Mather, 2001). The first subtest measures stu-
dents’ ability to read printed letters and words (coefficient alphas ranged
from .90 to .94 for 9- to 11-year-olds), whereas the second subtest mea-
sures children’s ability to rapidly read and respond to a statement (coeffi-
cient alphas ranged from .87 to .90 for 9- to 11-year-olds). Oral language
was assessed by administering the Formulated Sentences subtest from the
Clinical Evaluation of Language Fundamentals—Revised (Semel, Wiig,
Secord, & Sabers, 1987). This subtest measures students’ ability to produce
an oral sentence in response to a picture of an everyday activity (coefficient
alphas ranged from .70 to .78 for 9- to 11-year-olds). Means and standard
deviations by treatment condition and student type for each of these
measures are presented in Table 1. There were no statistically significant
differences between treatment conditions or student type for either of the
reading measures or the oral language test.
Instructional environment. Students’ regular writing instruction was
primarily patterned after Writers’ Workshop (Calkins, 1986; Graves,
1983). This is the most popular approach to writing instruction in elemen-
tary schools throughout the United States (Pritchard, 1987). It involves
setting up a classroom routine where students are expected to plan, draft,
revise, edit, and publish their work. Students share their progress and
completed work with peers, conference with the teacher about their writ-
ing, and often choose their writing topics. Instruction is typically provided
through minilessons. There was variation among teachers, however, in how
much time they devoted to writing instruction (30 min to 240 min a week)
as well as how often conferencing, sharing, revising, the teaching of
mechanics, and student choice occurred. For example, most of the teachers
taught writing mechanics (i.e., capitalization and punctuation) several
times a week, but others provided such instruction only on a monthly basis.
Likewise, some of the teachers conferenced with students each day,
whereas two teachers only did this several times per year.
1
General Procedures
Each student pair in each treatment condition received 30 lessons, 25
min in duration, three times a week for 10 weeks (12.5 hr of instruction).
Thus, each child received 750 min of instruction. Instruction took place in
a quiet place in the school, and all students were told that they were
receiving a special program to help them with their writing.
Instruction was delivered to students in the two treatment conditions by
six college students majoring in education. To control for possible instruc-
tor effects, each instructor taught an equivalent number of pairs of children
in both treatment conditions. Before the start of the study, instructors were
taught how to implement both treatment conditions; instruction continued
until each instructor could implement each treatment without error. For
each treatment condition, instructors were provided with a notebook that
contained detailed directions for implementing each activity and lesson
(this included a space to check off each step as it was completed). The
value of each treatment was stressed so that instructors would not be
predisposed to one treatment over the other. Instructors were also blind to
study hypotheses.
Experimental Treatment: Sentence Combining
The sentence-combining treatment was based in part on a previous
program developed by Strong (1986). It was designed to improve students’
sentence-construction skills and promote use of these skills when revising.
Instruction was broken down into five units, consisting of six lessons each.
The first unit focused on combining smaller related sentences into a
compound sentence using the connectors and,but, and because (e.g., “The
worm was squishy” and “The worm did not taste bad” combined to “The
worm was squishy, but it did not taste bad”). Starting with the conjunction
and allowed us to begin instruction with a fairly easy skill, as more and less
skilled fourth-grade writers typically use this conjunction to form com-
pound sentences. The next unit involved embedding an adjective or adverb
from one sentence into the other (e.g., “They run to a cave” and “They run
quickly” combined to “They run quickly to a cave”). The third and fourth
units concentrated on creating complex sentences by embedding an adver-
bial and adjectival clause, respectively, from one sentence into the other
(e.g., “The students all cheered” and “The movie stopped” combined to
“They all cheered when the movie stopped”). The final unit extended the
embedding skills taught in Units 2– 4 by teaching students to make multiple
embeddings involving adjectives, adverbs, adverbial clauses, and adjecti-
val clauses (e.g., “Ralph stuck his head out,” “Ralph was in Ryan’s
pocket,” “Ralph looked around,” and “Ralph did not know where he was”
combined to “Ralph, who was in Ryan’s pocket, did not know where he
was, but stuck his head out and looked around”).
Instruction was scaffolded so that the instructor first explained and
modeled how to use the new sentence-combining procedure. The student
pair then practiced applying the procedure to combine sentences orally,
with the instructor writing the pair’s responses on paper. The student pair
next practiced combining sentences, writing out their own responses. The
instructor provided assistance as needed. This was followed by indepen-
dent practice applying the sentence-combining procedure using a peer-
assisted format (each student alternatively acted as the coach as the other
child applied the procedure). In each of these steps, students discussed and
evaluated their solutions. All of these steps were included in Lessons 1 and
2, and the oral practice activity continued in Lesson 3.
1
Although there were some differences among schools in their writing
programs, the school attended by experimental students was not related to
posttreatment performance on any of the measures used to assess the
effects of sentence-combining instruction (all ps⬎.06).
46 SADDLER AND GRAHAM
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Instruction was further scaffolded so that students moved from combin-
ing sentences that contained cues for generating a successful solution to
open exercises where they were asked to combine multiple sentences
without the aid of cues to more natural practice where they used these skills
to write and revise their own work. There were two types of cues embedded
in all of the practice exercises in Lessons 1 and 2. The first provided a cue
that was positioned in parentheses at the end of one of the sentences to be
combined. For example, in Unit 1 the students were provided with this
kernel sentence pair to combine:
Joe dipped a worm in cornmeal.
Joe put the worm in a frying pan. (and)
With the second cue, some words or phrases in one of the sentences to be
combined were underlined. The underlined words or phrases were modi-
fiers that were inserted into a base sentence either before or after the words
modified. The use of both of these cues is illustrated below:
They tried to put the worm in their bag.
The worm did not fit in their bag. (but)
Students were told that both the underlined and the parenthetical clues must
be included verbatim in the new sentence combination and that parenthet-
ical content must precede underlined content when the sentence was
rewritten. Thus, the two sentences above might be rewritten as “They tried
to put the worm in their bag, but the worm did not fit.”
The underlining cue continued to be used for oral sentence-combining
practice during Lesson 3, but the cue in parentheses was eliminated at this
point. The only exception involved Units 1 and 5, where the sentences in
the oral practice activity continued to contain the cue in parentheses. This
was continued at these two points because we believed that students would
benefit from the extra scaffolding as they were first learning to use these
procedures (Unit 1) and as they practiced applying all that they had learned
with multiple sentences (Unit 5).
Lesson 3 also saw the introduction of a more open-ended exercise where
students were asked to combine multiple sentences without cues. This
exercise contained anywhere from 10 to 20 sentences that told a story.
Students were informed that they could add connecting words, take out
unneeded words, move words around, and change word endings. Finally, in
Lessons 4 – 6, students applied the sentence-combining skills they were
learning as they wrote and revised their own compositions.
All of the sentence-combining exercises in Lessons 1–3 were created
using story ideas from popular works of children’s fiction. The selected
books were from a list of books for fourth graders compiled by Scholastic
book publishers. Selected passages from the books were decombined into
kernel sentences and were presented in the same sequence as the original
story, thus in effect retelling it in a simplified manner. Breaking the
material up into smaller kernel sentences resulted in a readability level of
first and second grade for the decombined passages (according to the Fry,
1977, readability formula). This was well within the participating chil-
dren’s reading capabilities, as the average reading performance of both
skilled and less skilled writers was at about the 50th percentile of fourth-
grade students nationwide (see Table 1).
Lesson 1. This lesson began with the instructor introducing and ex-
plaining that good writers often play with their sentences to make them
sound better and that sometimes good writers may change words, move
words around, add words, or take words out. The instructor then told the
student pair that they were going to learn a new method or trick for doing
this, and it would make writing more interesting for them and their readers.
The instructor modeled, while thinking aloud, how to apply the target
sentence-combining procedure for that unit. This involved showing the
students a set of kernel sentences and reading them aloud. The instructor
modeled how to combine the first sentence combination, writing the new
sentence so that the students could see it. The students were then asked in
turn to provide a solution for how succeeding sentences could be com-
bined. Solutions were written down by the teacher as they were dictated by
the students. Each solution was discussed for rhetorical effectiveness, as
the instructor asked which sounded better and why. If neither student
thought of a solution, the instructor provided one as an example.
Next, students practiced combining sets of kernel sentences by creating
a written solution instead of a verbal one. After reading a set of kernel
sentences on their practice sheet, students wrote down their suggestion for
a combined sentence. If a problem was encountered with punctuation, the
students discussed possible solutions with the instructor. Before proceeding
to the next set of kernel sentences, the rhetorical effectiveness of each
student’s solution was discussed, and if any sentence was not grammati-
cally correct, the instructor helped the student modify it. Only when an
ungrammatical sentence was produced was the student’s writing corrected
and used as a basis for learning and problem solving. Otherwise, any
pertinent response was acceptable.
The teacher-guided written practice was followed by peer-directed in-
dependent practice. The peer-assistance format was based on the PALS
approach designed by Fuchs and Fuchs (cf. D. Fuchs et al., 1997). The
instructor provided the students with a worksheet containing sets of kernel
sentences and explained that they would work as a team to figure out
different ways to combine them. Each student served alternately as a coach,
providing directions for the task and corrective–supportive feedback as
needed, or as a player who accomplished the task.
To guide the process, the coach was provided with instruction cards.
These cards included directions for how the sentences should be combined
as well as directions for the coach and player’s actions. This included
telling the player to (a) first, read the sentence pair out loud, (b) second,
decide the best way to combine the sentences, (c) third, write the answer on
the sheet, and (d) fourth, read the new sentence. If the sentence was
grammatically correct, the coach was to reinforce the player by saying
“good job.” If the sentence was not grammatically correct, then the coach
provided suggestions on what changes to make in the sentence. If the coach
could not fix the sentence, the instructor provided assistance as needed.
After the sentence was written correctly, the roles were reversed and the
sequence repeated until all sentence pairs were completed. Once all kernel
pairs were completed, each student read a sentence to the instructor. The
instructor praised the students’ effort and asked if there was any other way
that the sentence could have been combined.
Lesson 1 ended with the instructor reviewing the new sentence-
combining procedure and prompting transfer. This involved revisiting the
purpose of the target sentence-combining procedures and how it worked
(this occurred in subsequent lessons). In addition, students were asked to
apply the sentence-combining procedure in their regular classroom. They
were further told that at the beginning of the next lesson, the instructor
would ask them to name each instance where they had applied this skill and
they would record the total number of instances on a chart.
Lesson 2. The second lesson was identical to the first lesson, with one
exception. At the start of the lesson, the instructor asked the students
whether they used the sentence-combining procedure for any sentences
they wrote in their regular classes. They put a sticker on a chart to record
each instance they identified, providing a daily record of how often they
used the procedures outside of the instructional environment. The process
of reminding students to use what they were learning in their regular
classes and recording how often this happened continued in all of the
subsequent lessons.
Lesson 3. The third lesson was the same as Lesson 2, except that
students did not do the teacher-guided written practice or the peer-assisted
independent practice activities. Instead, they worked together to produce a
revised paragraph from a series of related kernel sentences. No clues were
given so that many combinations were possible. The students were
prompted to use everything they had learned to help write the new para-
graph. Once they were done, 1 of the students read it aloud, and the
47
SENTENCE COMBINING AND WRITING PERFORMANCE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

instructors and students discussed the rhetorical effect of their revisions.
The instructor then prompted the student pair to revise their new paragraph
to make it different from their first version.
Lesson 4. In Lesson 4, students worked together to write and revise a
short story using the sentence-combining skills taught in that unit and
preceding units. The writing activity was structured so that the students
were provided with a planning facilitator that contained three columns. The
first column included two different characters, the second column de-
scribed two different settings, and the third column provided two possible
topics to write about. In addition, there were five kernel sentences students
were to use for their ending.
Students were first asked to combine three or more of the kernel
sentences to construct their ending. Keeping their ending in mind, they then
spent 5 min planning the rest of their story, using the planning facilitator
and their own imagination to flesh out the elements and detail of their
composition. Before writing their story, the instructor reminded them to
look for ways to make their sentences different and interesting by using the
sentence-combining procedures they were learning. After 15 min of writ-
ing, they were asked to read what they had written. They then looked for
ways to make their sentences better and revised them using their combining
procedures. The instructor provided the spelling of any word the students
needed.
Lesson 5. In Lesson 5, students completed a five-item sentence-
combining test. This test served as a progress monitoring probe (see the
Measures for Assessing the Effects of Treatment section), providing a
measure of students’ mastery of the sentence-combining skills taught in
that unit. Students also independently wrote a story in response to a picture.
This offered an additional opportunity for them to apply their sentence-
combining skills as they wrote.
Lesson 6. Students were asked to revise the story they wrote during
Lesson 5. They were specifically directed to revise at least three sentences
using the sentence combinations they had learned in the current or prior
units. Assistance with the first revision was provided as needed.
Comparison Treatment: Grammar Instruction
Instruction in the comparison treatment involved teaching specific gram-
mar skills; namely, parts of speech. It was designed to enhance the
precision of students’ vocabulary in their writing through practice in using
more descriptive nouns, verbs, adjectives, and adverbs. As with sentence-
combining instruction, there were five units, consisting of six lessons each.
Unit 1 focused on verbs, Unit 2 on nouns, Unit 3 on adjectives, Unit 4 on
adverbs, and Unit 5 on subjects and predicates. The experimental and
comparison treatments were similar in that both used the same type of
instructional activities in each lesson.
Lesson 1. This lesson began with the instructor introducing and ex-
plaining that good writers know the parts of speech and how to use these
parts to make their sentences and writing better. The instructor then told the
student pair that they were going to learn how to do this, and it would make
writing more interesting for them and their readers. The instructor mod-
eled, while thinking aloud, how to apply the target part of speech for that
unit. This involved showing the students a sentence with the target part of
speech missing and reading the sentence aloud. The instructor modeled
how to complete the sentence, defining the part of speech needed and
supplying a correct word as the completed sentence was written on the
practice sheet. The students were then asked in turn to provide a solution
for succeeding sentences by adding an appropriate word for the target part
of speech. Solutions were written down by the teacher as they were dictated
by the students. Each solution was discussed for rhetorical effectiveness, as
the instructor asked which one sounds better and why. If neither student
thought of a solution, the instructor provided one as an example.
Next, students practiced completing sentences by supplying the missing
correct part of speech. After reading an incomplete sentence on their
practice sheet, students wrote down their suggestion for a completed
sentence. If a problem was encountered with providing the correct part of
speech, students discussed possible solutions with the instructor. Only
when an incorrect part of speech was supplied was the student’s writing
corrected and used as a basis for learning and problem solving. Otherwise,
any pertinent response was acceptable.
The teacher-guided written practice was followed by peer-directed in-
dependent practice. The format for this practice was identical to the one
used with sentence-combining treatment and consequently will not be
reiterated here. The only difference involved the task (i.e., sentence com-
pletion) and the procedure students used to complete it. At the start of this
task, the instructor provided students with a worksheet containing sen-
tences where the target part of speech was missing. Students were directed
to work as a team using the PALS format (D. Fuchs et al., 1997) to figure
out how to complete the sentences with the target part of speech.
Lesson 1 ended with the instructor reviewing the part of speech and
prompting transfer. This involved revisiting the part of speech and how it
helps the writer (e.g., “Verbs help show action”). This review occurred in
all subsequent lessons. In addition, students were asked to apply what they
were learning outside the instructional setting. They were told that if they
brought sentences with the target part of speech underlined to the next
lesson, they would record this on a chart.
Lesson 2. The second lesson was identical to the first lesson, with one
exception. At the start of this lesson, the instructor asked the students to
show how many sentences with the target part of speech underlined they
had produced since the last lesson. They put a sticker on a chart to record
each sentence, providing a daily record of how often they used the
procedures outside of the instructional environment. The process of chart-
ing the number of sentences produced by students continued in all subse-
quent lessons.
Lesson 3. The third lesson was the same as Lesson 2, except that
students did not do the teacher-guided written practice or the peer-assisted
independent practice activities. Instead, they worked together to complete
a paragraph where the target part of speech was missing in each sentence.
Once they were done, 1 of the students read it aloud, and the instructors and
students discussed the rhetorical effect of their revisions.
Lesson 4. In Lesson 4, students worked together to write and revise a
short story using the parts of speech taught in that unit and preceding units.
The writing activity was identical to the one used in the experimental
treatment condition. It included a planning facilitator with three columns:
one for characters, another for setting, and the third for topics. In each
column, two options were provided. In addition, there were five sentences
students could use for their ending.
Students were first asked to use three or more of the five sentences for
their ending. Keeping their ending in mind, they then spent 5 min planning
the rest of their story, using the planning facilitator. Before students began
writing their stories, the instructor reminded them to use the parts of speech
they had learned. After 15 min of writing, they were asked to read what
they had written. They then looked for ways to make their story better
using the parts of speech. The instructor provided the spelling of any word
the students needed.
Lesson 5. In Lesson 5, students completed the five-item sentence-
combining test taken by students in the sentence-combining treatment.
Students also independently wrote a story in response to a picture. This
offered an additional opportunity for them to use the parts of speech they
were learning as they wrote.
Lesson 6. Students were asked to revise the story they wrote during
Lesson 5. They were specifically directed to revise at least three sentences
using the parts of speech they had learned in the current or prior units.
Assistance with the first revision was provided as needed.
Fidelity of Treatment Implementation
To ensure that the experimental and comparison treatment conditions
were delivered as planned, the following safeguards were implemented.
48 SADDLER AND GRAHAM
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

First, instructors received intensive practice in applying the instructional
procedures for both treatments. Second, instructors met with Bruce Saddler
weekly to discuss any glitches that occurred in implementing procedures.
Reported glitches or deviations from instructional plans occurred rarely
and usually involved an inadvertent mistake by an instructor. Third, in-
structors were provided with a checklist that contained step-by-step direc-
tions for each lesson. As they completed a step, they were asked to check
it off. Examination of these checklists by Bruce Saddler showed that
instructors completed 98.3% of the steps in the sentence-combining treat-
ment and 99.7% of the steps in the grammar treatment. The reliability of
Bruce Saddler’s scores were checked by having a trained assistant inde-
pendently rescore 30% of the checklist (reliability between Bruce Saddler
and the assistant was .99). Fourth, 30% of all lessons were tape recorded
and checked by the assistant to determine whether each step of a lesson was
executed as intended by the instructor. There was 100% agreement be-
tween what the instructors checked as completed and what the assistant
checked. To ensure that the assistant’s scores were reliable, Bruce Saddler
rescored 30% of the tapes (reliability between Bruce Saddler and the
assistant was .98).
The assistant also rated the overall quality of the lessons that were tape
recorded. This involved assessing the following six areas: (a) level of
student engagement, (b) students’ responses to instructor’s questions and
participation in discussion, (c) quality of instructor’s responses to students’
questions, (d) completion of each lesson step, (e) efficiency of instruction,
and (f) pacing of instruction. Each area was rated using a 5-point Likert-
type scale (scores ranged from 0 to 4, with 0 representing a score of not
evident and 4 a score of excellent). A score of 24 was the maximum number
of points that could be assigned to a lesson.
To make the scoring system more concrete, the assistant was provided a
verbal description for each Likert-type scale point in each of the six areas.
The assistant practiced applying the scoring system until she reached 70%
reliability with Bruce Saddler. She then independently scored all of the
tapes. Once this was completed, Bruce Saddler independently rescored one
third of them (reliability between Bruce Saddler and the assistant was .80).
Overall, the six instructors maintained a high quality of instruction for both
treatment conditions. The mean quality rating for sentence combining was
22.9 (SD ⫽1.39). It was 23.23 (SD ⫽1.09) for grammar instruction. The
six instructors’ ratings ranged from a low average of 20.4 to a perfect 24.0
high average.
Measures for Assessing the Effects of Treatment
The measures described below were administered to assess the effects of
treatment; each measure was administered to students individually. Three
of the measures described below (writing quality, length, and revisions)
were obtained from stories that students were asked to write and revise
immediately before and after treatment. Students wrote their stories in
response to a picture prompt. To increase motivation, students were pro-
vided with a choice of two prompts at both pretest and posttest. The picture
prompts involved line drawings depicting one or more characters involved
in an activity (i.e., a rabbit lecturing other rabbits about carrots, a child
swinging on a vine through the jungle, an alien using a phone booth to
make a telephone call, and a girl showing two boys a snake). Before the
start of the study, fourth-grade students in two classes were asked to
evaluate 14 line drawings and select the one they thought would be most
interesting to write about. The four prompts used in this study and de-
scribed above were the most popular. Two of the writing prompts were
randomly assigned to pretest and the other two were assigned to posttest.
During testing, the administration of writing prompt sets was counterbal-
anced across pretest and posttest.
When writing a first draft, students were given the two pictures and
directed to write a story about one of them. They were given 15 min to
write their story and 15 min during the next session to revise it. During the
drafting and revising sessions, instructors provided no assistance, nor were
students prompted to use any of the skills they were taught. Prior to
scoring, compositions were typed (errors were not corrected) and identi-
fying information removed.
With the exception of the writing quality measure, the assessments
below were scored by Bruce Saddler. A second scorer who was blind to the
purpose and design of the study independently rescored one third of the
protocols (randomly selected). For writing quality, two former teachers
(who were blind to the purpose and design of the study) independently
scored all compositions. To determine interrater reliability between the
scores assigned by the two raters, we calculated a Pearson product–moment
correlation coefficient for each measure.
Sentence-combining measures. To assess changes in students’
sentence-combining skills, we administered two measures. First, in the fifth
lesson of each unit, all students completed a five-item progress-monitoring
test designed to measure their mastery of the sentence-combining skills
taught in that unit to students in the experimental treatment condition. In
Units 1– 4, each item included two sentences that students were asked to
combine into a written single sentence (e.g., “Combine ‘The students sat
still’ and ‘The principal finished talking’ into one better sentence”),
whereas Unit 5 involved combining four or five sentences into a written
single sentence (e.g., “Combine ‘Ralph is a mouse,’ ‘Ralph is gray,’ ‘Ralph
has many brothers,’ and ‘Ralph has many sisters’ into one better sen-
tence”). This was consistent with the practice activities completed by the
students in the sentence-combining condition.
Items on the five progress-monitoring tests were scored as correct if the
student produced a single written sentence that incorporated the important
elements from the stimulus sentences and if this sentence was grammati-
cally correct. Interrater reliability for the two scorers was .92. A student’s
score for the progress-monitoring tests was the average of the five tests.
Alpha coefficient for this average score was .85.
Second, students were administered Form B of the Sentence Combining
subtest from the TOWL⫺3 (Hammill & Larsen, 1996) once instruction
was completed (Form A of this subtest was administered as a screening
measure before the start of treatment). Form B contains 20 items that
require students to produce increasingly complex and grammatically cor-
rect written single sentences by combining and integrating the meaning of
two or more sentences together (this test measures a wider range of
sentence-combining skills than does the progress-monitoring tests). Coef-
ficient alpha ranges from .83 to .86 for 9- to 11-year-olds for Form B of this
subtest (correlations with Form A are high as well, as alternate form
reliability coefficients range from .87 to .89).
Items on the Sentence Combining subtest were scored as correct if the
student produced a single written sentence that incorporated the important
elements from the stimulus sentences and if this sentence was grammati-
cally correct. Spelling, punctuation, and capitalization errors were ignored
in scoring the produced sentence as correct or incorrect. Interrater reliabil-
ity between the two scorers was 1.00. For the purpose of data analysis, raw
scores were converted to standard scores (M⫽10; SD ⫽3), using the
normative tables in the test manual.
Writing quality. To examine treatment effects on quality of students’
story writing, we scored first and second drafts of each story using a
traditional holistic quality rating scale (Cooper, 1977). Examiners were
asked to read the paper attentively to obtain a general impression of overall
writing quality. Stories were then scored on an 8-point scale, with 1
representing the lowest quality of writing and 8 representing the highest
quality. Examiners were told that ideation, organization, grammar, sen-
tence structure, aptness of word choice, and mechanics should all be taken
into account in forming a judgment about overall quality and that no one
factor should receive undue weight.
Examiners were provided with a representative paper (or anchor point)
for a low-, middle-, and high-quality score. These stories were collected
from two fourth-grade classes that did not participate in the study. Two
former elementary school teachers selected the best, average, and poorest
49
SENTENCE COMBINING AND WRITING PERFORMANCE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

quality stories on the basis of the scoring criteria described above. These
papers served as the anchor points for their respective genre.
To train examiners to use the quality rating scale, Steve Graham first had
them discuss the distinguishing features of each of the anchor points. Next,
they practiced applying the scale to a series of compositions that varied
widely in quality. After independently scoring each practice story, exam-
iners compared their scores and resolved any differences through discus-
sion. Training continued until examiners obtained scores that differed by
no more than 1 point on 10 consecutive compositions. Interrater reliability
was .72 between the two examiners when they independently scored all
first and second drafts from pretest and posttest. The quality score for a
student’s first or second draft was the average score for the two raters.
Length. To assess the impact of the two treatments on story length, we
scored all first and second drafts of papers for number of words written.
Total number of words included all written words, regardless of spelling,
which represented a spoken word. Interrater reliability between the two
examiners was 1.00.
Sentence-combining revisions. To determine whether students com-
bined sentences when revising, we laid a students’ first draft and revised
draft side by side. The examiner then identified each revision that involved
combining two or more sentences into a single sentence. The score for this
measure was the total number of separate revisions involving sentence
combining. Interrater reliability between the two scorers was .95.
Results
We first examined the impact of the two treatments on students’
sentence-combining skills. The impact of the treatments on story
writing quality, length, and sentence-combining revisions were
examined next. Effect sizes were computed for each statistically
significant difference by dividing the mean difference between the
two treatments conditions by the pooled standard deviation.
Sentence Combining
Means and standard deviations for the sentence-combining
progress-monitoring measure are presented in Table 2. The means
and standard deviations for the Sentence Combining subtests
(Forms A and B; Hammill & Larsen, 1996) are presented in
Table 3.
Progress-monitoring tests. We predicted that students who
received sentence-combining instruction would be more adept at
applying the sentence-combining skills that were taught than stu-
dents who received grammar instruction. To test this hypothesis,
we compared the performance of the two treatment groups on the
five progress-monitoring tests administered during instruction.
Students’ average score for these five tests were analyzed using a
2 (treatment) ⫻2 (student type) ANOVA. There was a statistically
significant main effect for treatment, F(1, 39) ⫽31.3, MSE ⫽
37.7, p⫽.00, effect size ⫽1.31. Students in the sentence-
combining condition were twice as likely as comparison students
to produce a correct written sentence (i.e., grammatically correct
and containing all critical ideas) on these unit tests (see Table 2).
Thus, sentence-combining instruction was effective in improving
the sentence-combining skills that were taught.
Sentence Combining (TOWL–3). We also expected that
sentence-combining instruction would have a positive impact on a
norm-referenced measure of students’ sentence-combining skills.
This norm-referenced measure assesses a broader array of
sentence-combining skills than the progress-monitoring measures.
To assess this hypothesis, we analyzed students’ posttest scores
from Form B of the Sentence Combining subtest of the TOWL–3
(Hammill & Larsen, 1996) using a 2 (treatment) ⫻2 (student type)
analysis of covariance (ANCOVA). Students’ scores on the
screening measure (Form A of the Sentence Combining subtest of
the TOWL–3) served as a covariate (the ANCOVA assumption
that the regression slopes were homogeneous was met). We de-
cided not to use testing time (preintervention vs. postintervention)
as an independent variable in this analysis, because Form A (the
screening measure) was group administered and Form B (admin-
istered following treatment) was administered individually.
As predicted, there was a statistically significant main effect for
treatment, F(1, 38) ⫽8.8, MSE ⫽184.5, p⬍.01. Neither the main
effect for student type (more or less skilled writers) or any of the
interactions were statistically significant. Thus, the effects of
sentence-combining instruction were evident not only on the
researcher-designed progress-monitoring tests that just assessed
the skills that were taught but on a norm-referenced measure of
sentence combining that assessed a broader set of skills.
Writing Measure
Means and standard deviations for writing quality and length are
presented in Table 4. Means and standard deviations for revising
are presented in Table 5.
Table 2
Means and Standard Deviations on the Sentence-Combining
Progress-Monitoring Measure
Student type Sentence combining Grammar
More skilled writers
M3.8 1.7
SD 0.8 1.1
Less skilled writers
M3.7 2.1
SD 1.1 1.3
All skilled writers
M3.8 1.9
SD 0.9 1.3
Note. The progress-monitoring test is the average score for the five unit
tests administered during treatment (scores ranged from 0 to 5).
Table 3
Means and Standard Deviations for the Sentence Combining
Subtest From the Test of Written Language—3 (TOWL–3)
Time of testing
Sentence
combining Grammar
More
skilled
Less
skilled
More
skilled
Less
skilled
Pretest (Form A)
M10.4 6.7 10.9 7.0
SD 0.5 0.7 1.4 0.3
Posttest (Form B)
M12.9 12.5 10.3 10.3
SD 2.2 2.8 2.8 2.5
Adjusted M
a
10.5 10.4 5.8 6.4
a
Posttest mean on Form B of the Sentence Combining subtest of the
TOWL–3 adjusted for pretest score on Form A of the Sentence Combining
subtest of the TOWL–3.
50 SADDLER AND GRAHAM
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Writing quality. It was predicted that sentence-combining in-
struction would have a positive impact on the quality of students’
story writing. To test this hypothesis, we analyzed holistic quality
scores for students’ story writing using a 2 (treatment) ⫻2
(student type) ⫻2 (draft of paper) ⫻2 (time of testing) ANOVA,
with repeated measures on the last two variables. There were
statistically significant effects for the Time ⫻Draft interaction,
F(1, 39) ⫽4.8, MSE ⫽1.8, p⫽.04, the Time ⫻Draft ⫻
Treatment interaction, F(1, 39) ⫽7.1, MSE ⫽2.7, p⫽.01, and the
Draft ⫻Treatment ⫻Student Type interaction, F(1, 39) ⫽5.9,
MSE ⫽2.2, p⫽.02. None of the main effects nor the other
interactions were statistically significant.
Follow-up analysis focused only on the Time ⫻Draft ⫻Treat-
ment interaction. We did not analyze the Time ⫻Draft interaction,
as it was mediated by treatment, nor did we analyze the Draft ⫻
Treatment ⫻Student Type interaction because the dependent
measure for such an analysis was the average of students’ pretest
and posttest quality score.
For the Time ⫻Draft ⫻Treatment interaction, we ran a 2
(treatment) ⫻2 (draft of paper) ANOVA, with a repeated measure
on the last variable for both the pretest and posttest data. At pretest,
the main effect for treatment and draft were not statistically sig-
nificant (both ps⬎.69), nor was the interaction between these two
variables ( p⫽15). At posttest, however, there was a statistically
significant effect for draft, F(1, 41) ⫽6.3, MSE ⫽2.7, p⫽.02,
and a Draft ⫻Treatment interaction, F(1, 41) ⫽4.8, MSE ⫽2.1,
p⫽.04. Follow-up analysis for the Draft ⫻Treatment interaction
showed that for students in the sentence-combining condition,
revising improved the quality of their posttest stories, F(1, 20) ⫽
14.7, MSE ⫽4.7, p⬍.01, effect size ⫽0.64. In contrast, revising
had virtually no impact on the average quality of stories produced
by students in the comparison condition at posttest (see Table 4).
For students in the sentence-combining condition, posttest story
quality improved from 3.4 (SD ⫽0.1) on the first draft to 4.1
(SD ⫽1.3) on the revised draft. For students in the comparison
condition, average posttest story quality remained the same on
both first and revised drafts (M⫽3.5).
Length. No prediction was made for the effects of sentence
combining on the length of students’ stories, as such instruction
might theoretically result in longer or shorter text. Length of
students’ story writing was analyzed using a 2 (treatment) ⫻2
(student type) ⫻2 (draft of paper) ⫻2 (time of testing) ANOVA,
with repeated measures on the last two variables. There were
statistically significant effects for draft, F(1, 39) ⫽22.1, MSE ⫽
2,234.8, p⫽.00, and the Time ⫻Draft interaction, F(1, 39) ⫽7.2,
MSE ⫽656.0, p⫽.01. None of the other main effects nor
interactions were statistically significant. Revised drafts were
longer than first drafts at pretest, F(1, 42) ⫽239.2, MSE ⫽
832,031.1, p⬍.01, and at posttest, F(1, 42) ⫽21.8, MSE ⫽265.1,
p⬍.01. At pretest, revised papers were approximately 11.0 words
longer, whereas at posttest revised papers were only 3.5 words
longer. At pretest, the mean for Draft 1 and Draft 2 papers were
92.7 (SD ⫽36.6) and 104.0 (SD ⫽48.1), respectively. At posttest,
Table 4
Means and Standard Deviations for Story Quality and Length
Variable
Sentence combining Grammar
More skilled Less skilled More skilled Less skilled
MSD M SDMSD M SD
Pretest story quality
Draft 1 3.8 1.3 3.7 1.9 3.4 1.6 3.7 1.9
Revised draft 3.5 1.4 3.5 1.6 3.7 1.6 3.6 1.8
Posttest story quality
Draft 1 3.5 0.8 3.4 1.1 3.0 1.0 4.0 2.0
Revised draft 3.8 1.1 4.4 1.4 3.4 1.1 3.7 1.6
Pretest story length
Draft 1 98.7 27.6 93.6 44.4 83.1 42.4 95.6 35.7
Revised draft 108.0 38.6 105.1 61.4 94.5 56.3 108.6 38.9
Posttest story length
Draft 1 82.8 24.1 103.0 33.7 88.6 32.5 108.9 49.3
Revised draft 86.0 24.6 104.5 33.6 92.6 34.1 114.1 55.1
Note. Story quality scores ranged from 1 (low)to8(high).
Table 5
Means and Standard Deviations for Number of Sentence-
Combining Revisions (SCR)
Type of SCR
Sentence
combining Grammar
More
skilled
Less
skilled
More
skilled
Less
skilled
Conjunction and
Pretest
M0.0 0.3 0.2 0.5
SD 0.0 0.7 0.4 1.1
Posttest
M0.0 0.2 0.2 0.4
SD 0.0 0.4 0.6 0.9
All other SCR
Pretest
M0.3 0.0 0.1 0.4
SD 0.6 0.0 0.3 0.7
Posttest
M0.5 0.6 0.0 0.0
SD 0.9 1.3 0.0 0.0
51
SENTENCE COMBINING AND WRITING PERFORMANCE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the mean for the first draft was 95.7 (SD ⫽36.4), whereas the
mean for the revised draft was 99.2 (SD ⫽38.8).
Sentence-combining revisions. We predicted that students in
the sentence-combining treatment would be more likely than their
peers in the comparison condition to make revisions in their first
draft that involved the sentence-combining skills that were taught.
To assess this hypothesis, we used two 2 (treatment) ⫻2 (student
type) ⫻2 (time of testing) ANOVAs, with a repeated measure on
the last variable, to analyze revisions involving sentence combin-
ing. The first ANOVA examined whether there was a statistically
significant difference between the two treatment conditions in
terms of the number of sentence-combining revisions involving the
conjunction and. We analyzed this sentence-combining revision
separately, because fourth-grade students already typically use and
to form compound sentences. As predicted, none of the main
effects nor interactions were statistically significant for this vari-
able (see Table 5 for means and standard deviations).
The second ANOVA examined whether there was a statistically
significant difference between the two treatments in total number
of all other sentence-combining revisions made by participants.
There was a statistically significant Treatment ⫻Time interaction,
F(1, 39) ⫽6.0, MSE ⫽1.0, p⫽.02. None of the main effects nor
any of the other interactions were statistically significant. There
was no statistically significant difference in the total number of all
other sentence-combining revisions made by students in the two
treatment conditions prior to instruction ( p⫽.59), but students
who received sentence-combining instruction made more of these
kinds of revisions following instruction, F(1, 42) ⫽5.7, MSE ⫽
3.5, p⫽.02, effect size ⫽0.69. It must be noted that the mean
number of sentence-combining revisions made by students in
either condition was small (see Table 5). At posttest, students who
received grammar instruction made no sentence-combining revi-
sions, other than those involving the conjunction and, whereas
students in the experimental treatment condition averaged less than
one of these types of revisions when revising a paper (M⫽0.6;
SD ⫽1.1).
Discussion
A writer’s proficiency in translating intentions and ideas into
well-crafted sentences can impact both the writer and the reader. It
may handicap young writers because they have to devote so much
attention to it that this minimizes or even interferes with their use
of other important writing processes (Scardamalia & Bereiter,
1986). It can enhance or hinder reading by making text more or
less readable. Thus, an important goal in learning to write is to help
children become more facile at sentence construction.
In this study, we examined whether a peer-assisted approach to
sentence-combining instruction would enhance the writing of
fourth-grade students. As expected, such instruction had a positive
impact on the sentence-combining skills of both more and less
skilled writers at this grade level. When students were tested on the
specific sentence-combining skills they were taught during instruc-
tion (progress-monitoring measure), they were twice as likely to
combine two or more sentences into a semantically and syntacti-
cally correct single sentence than their counterparts who received
grammar instruction. Likewise, on a norm-referenced test
(TOWL–3) that assesses a broader range of sentence-combining
skills, students in the experimental condition again outscored their
peers in the comparison condition. For both of these measures, the
effects of treatment were strong, as the effect sizes for the
progress-monitoring measure and the norm-referenced test were
1.31 and 0.81, respectively. These findings demonstrate that
sentence-combining instruction can be effective in improving the
sentence-construction skills of young developing writers in the
elementary grades. Most of the previous research in contrast has
examined the effectiveness of this procedure with older students
(Hillocks, 1986; O’Hare, 1973).
It is also instructive to examine the relative standing of exper-
imental students before and after instruction on the norm-
referenced sentence-combining test. On the basis of the normative
data for the TOWL–3, weaker writers in the sentence-combining
condition evidenced an increase of almost two standard deviations
from pretest to posttest, whereas more skilled writers in this
condition increased by five sixths of a standard deviation. The
average posttest performance of the more and less skilled writers
in the experimental group placed them at the 90th and 87th
percentiles, respectively, of the normative sample for the
TOWL–3.
In addition, our predictions that sentence-combining instruction
would have a positive impact on the writing performance of
students was partially confirmed. Following instruction, students
in the experimental condition were more likely to revise their
papers by combining sentences than their peers who received
grammar instruction. Although an effect size of 0.69 indicated that
the experimental treatment had a moderate impact on promoting
sentence-combining revisions, this must be tempered by the fact
that students in the sentence-combining condition averaged less
than one of these revisions per paper. The sentence-combining
revisions they did make were almost equally divided between
using but and because to create compound sentences and combin-
ing two simpler sentences to create a more complex one (contain-
ing at least one independent and dependent clause). Their
sentence-combining revisions did not involve embedding an ad-
jective or adverb from one sentence into another (e.g., turning
“The cat is quick” and “The cat is hungry” into “The cat is quick
and hungry”). It should be noted that there were few opportunities
to do so, because students did not typically write pairs of sentences
like the ones in the above parentheses when composing their first
drafts at pre- or posttest.
In terms of writing quality, students in the experimental condi-
tion evidenced a single advantage over their counterparts in the
comparison condition. When students in the sentence-combining
condition revised their posttest papers, the overall quality of their
writing improved. This did not occur for children who received
grammar instruction. The impact of sentence-combining instruc-
tion on quality change following revising was moderate, as the
effect size was 0.64. Although it is tempting to attribute improve-
ments in writing quality from first to second draft to the sentence-
combining revisions experimental students made, the correlation
between quality change and number of sentence-combining revi-
sions was only .34 at posttest for experimental students. Thus,
other aspects of the sentence-combining treatment must have con-
tributed to this improvement as well.
Why did sentence-combining instruction not have a stronger
impact on writing quality? An assumption underlying sentence-
combining instruction is that this approach makes the process of
sentence construction more habituated and less effortful, freeing
52 SADDLER AND GRAHAM
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

up cognitive resources that can be devoted to other writing pro-
cesses (i.e., the cognitive load reduction hypothesis). It is further
assumed that these resources will be used wisely, resulting in a
better written product. For the young writers in this investigation,
the approximately 12 hr of instruction may not have reduced the
cognitive strain of sentence construction enough to obtain the
hypothesized results. Even if the experimental treatment did make
sentence construction less effortful, however, the participating
students may not have taken full advantage of this situation, failing
to apply all of their available resources or applying them
ineffectively.
Despite the importance of the cognitive load reduction hypoth-
esis, investigators have not examined its viability for sentence-
combining instruction. If this hypothesis is valid, there should be
a reduction following sentence-combining instruction in the
amount of time students take to react to a secondary task (e.g., an
audible signal) that occurs while they construct sentences (see
Kellogg, 1987, for an example of how a secondary-task reaction
time technique can be used to measure cognitive effort during
writing). Likewise, instructed students should become more fluent
(as well as adept) at combining sentences. There should also be a
proportional increase in the amount of time that these students
devote to other writing processes (see Rijlaarsdam & van den
Bergh, 1996). Future research needs to examine whether sentence-
combining instruction yields such effects.
In summary, findings from the current study replicate and ex-
tend previous research by showing that a peer-assisted sentence-
combining treatment can improve the sentence-construction skills
of more and less skilled young writers. Such instruction can also
promote young students’ use of sentence-combining skills as they
revise. Finally, sentence-combining instruction can have a positive
effect on the quality of young students’ writing, specifically in
terms of revising the first drafts of their papers. Additional re-
search is needed to replicate these findings. Perhaps just as im-
portant, researchers need to examine other approaches for improv-
ing children’s sentence-construction skills. There are few
scientifically validated practices for improving children’s sentence
skills outside of sentence-combining instruction.
References
Applebee, A. (1984). Writing and reasoning. Review of Educational Re-
search, 54, 577–596.
Calkins, L. (1986). The art of teaching writing. Portsmouth, NH:
Heinemann.
Combs, W. E. (1975). Some further effects and implications of sentence
combining exercises for the secondary language arts curriculum (Doc-
toral dissertation, University of Minnesota, 1975). Dissertation Ab-
stracts International, 36, A1266.
Cooper, C. (1977). Holistic evaluation of writing. In C. Cooper & L. Odell
(Eds.), Evaluating writing (pp. 3–31). Urbana, IL: National Council of
Teachers of English.
Durst, R., & Newell, G. (1989). The uses of function: James Britton’s
category system and research on writing. Review of Educational Re-
search, 59, 375–394.
Freedman, S. (1979). How characteristics of student essays influence
teachers’ evaluations. Journal of Educational Psychology, 71, 328 –338.
Fry, E. (1977). Fry’s readability graph: Clarifications, validity, and exten-
sion to Level 17. Journal of Reading, 17, 242–252.
Fuchs, D., Fuchs, L., Mathes, P., & Simmons, D. (1997). Peer-assisted
learning strategies: Making classrooms more responsive to diversity.
American Educational Research Association, 34, 174 –206.
Fuchs, D., Fuchs, L., & Thompson, A. (2001). Peer-assisted learning
strategies in reading: Extensions for kindergarten, first grade, and high
school. Remedial and Special Education, 22, 15–21.
Fuchs, L., Fuchs, D., & Kazdan, S. (1999). Effects of peer-assisted learning
strategies on high-school students with serious reading problems. Reme-
dial and Special Education, 20, 309 –318.
Fuchs, L., Fuchs, D., Phillips, N., Hamlett, C., & Karns, K. (1995).
Acquisition and transfer effects of classwide peer-assisted learning strat-
egies in mathematics for students with varying learning histories. School
Psychology Review, 24, 604 – 620.
Graham, S. (1982). Composition research and practiced: A unified ap-
proach. Focus on Exceptional Children, 14, 1–16.
Graham, S. (in press). Writing. In P. Alexander & P. Winne (Eds.),
Handbook of educational psychology. Mahwah, NJ: Erlbaum.
Graham, S., & Harris, K. R. (1988). Instructional recommendations for
teaching writing to exceptional students. Exceptional Children, 54, 506 –
512.
Graham, S., & Harris, K. R. (2002). Prevention and intervention for
struggling writers. In M. Shinn, H. Walker, & G. Stone (Eds.), Inter-
ventions for academic and behavior problems: II. Preventive and reme-
dial approaches (pp. 589 – 610).Bethesda, MD: National Association of
School Psychologists.
Graham, S., & Harris, K. R. (in press). Writing instruction. In B. Wong
(Ed.), Learning about learning disabilities (3rd ed.). San Diego, CA:
Elsevier Science.
Graham, S., Harris, K. R., & Mason, L. (in press). Improving the writing
performance, knowledge, and self-efficacy of struggling young writers:
The effects of self-regulated strategy development. Contemporary Ed-
ucational Psychology.
Graves, D. (1983). Writing: Teachers and children at work. Exeter, NH:
Heinemann.
Greenwald, E., Persky, H., Campbell, J., & Mazzeo, J. (1999). National
assessment of educational progress: 1998 report card for the nation and
the states. Washington, DC: U.S. Department of Education.
Greenwood, C., Carta, J., & Kamps, D. (1990). Teacher versus peer-
mediated instruction: A review of educational advantages and disadvan-
tages. In M. Foot, F. Morgan, & R. Shute (Eds.), Children helping
children (pp. 177–205). Sussex, England: Wiley.
Hammill, D., & Larsen, S. (1996). Test of Written Language—3. Austin,
TX: Pro-Ed.
Hayes, J. R., & Flower, L. S. (1986). Writing research and the writer.
American Psychologist, 41, 106 –113.
Hillocks, G. (1986). Research on written composition. Urbana, IL: ERIC
Clearinghouse on Reading and Communication Skills. (ERIC Document
Reproduction Service No. ED265552)
Horstman, F. R. (1989). The effects of instruction in sentence combining
and revision on ninth and tenth graders’ explanatory writing (Doctoral
dissertation, University of Maryland, 1989). Digital Dissertations, AAT
9012476.
Houck, C. K., & Billingsley, B. S. (1989). Written expression of students
with and without learning disabilities: Differences across the grades.
Journal of Learning Disabilities, 22, 561–575.
Hunt, K. W. (1965). Grammatical structures written at three grade levels.
(Research Rep. No. 3). Champaign, IL: National Council of Teachers of
English.
Intersegmental Committee of the Academic Senates. (2002). A statement of
competencies of students entering California’s public colleges and uni-
versities. Sacramento, CA: Author.
Kellogg, R. (1987). Effects of topic knowledge on the allocation of
processing time and cognitive effort to writing processes. Memory &
Cognition, 15, 256 –266.
Kendall, P. (1989). The generalization and maintenance of behavior
53
SENTENCE COMBINING AND WRITING PERFORMANCE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

change: Comments, considerations, and the “no cure” criticism. Behav-
ior Therapy, 20, 357–364.
Maheady, L., Harper, G., Mallette, B., & Winstanley, N. (1991). Training
and implementing requirements associated with the use of classwide
peer tutoring system. Education and Treatment of Children, 14, 177–
198.
Morris, N. T., & Crump, W. D. (1982). Syntactic and vocabulary devel-
opment in the written language of learning disabled and non-learning
disabled students at four age levels. Learning Disabilities Quarterly, 5,
163–172.
Myklebust, H. R. (1973). Development and disorders of written language.
Studies of normal and exceptional children (Vol. 2).New York: Grune
& Stratton.
National Commission on Writing. (2003). The neglected “R.” Washington,
DC: College Entrance Examination Board.
Neuleib, J., & Fortune, R. (1985). The use of sentence combining in an
articulated curriculum. In D. A. Daiker, A. Kerek, & M. Morenberg
(Eds.), Sentence combining: A rhetorical perspective (pp. 127–137).
Carbondale: Southern Illinois University Press.
Newcomer, P. L., & Barenbaum, E. M. (1991). The written composing
ability of children with learning disabilities: A review of the literature
from 1980 to 1990. Journal of Learning Disabilities, 24, 578 –593.
Ney, J. W. (1981, March). Sentence combining: Everything for everybody
or something for somebody. Paper presented at the annual meeting of the
Conference on English Education, Anaheim, CA. (ERIC Document
Reproduction Service No. ED199753)
No Child Left Behind Act of 2001, Pub. L. No. 107–110, 115 Stat 1425.
O’Hare, F. (1973). Sentence combining. Champaign, IL: National Council
of Teachers of English.
Persky, H., Daane, M., & Jin, Y. (2003). The nation’s report card: Writing.
Washington, DC: U.S. Department of Education.
Pritchard, R. (1987). Effects on student writing of teacher training in the
National Writing Project Model. Written Communication, 4, 51– 67.
Rhodes, L. K., & Dudley-Marling, C. (1996). Readers and writers with a
difference: A holistic approach to teaching struggling readers and
writers. Portsmouth, NH: Heinemann.
Rijlaarsdam, G., & van den Bergh, H. (1996). The dynamics of compos-
ing—An agenda for research into an interactive compensatory model of
writing: Many questions, some answers. In M. Levy & S. Ransdell
(Eds.), The science of writing: Theories, methods, individual differences,
and applications (pp. 48 –75).Mahwah, NJ: Erlbaum.
Scardamalia, M., & Bereiter, C. (1986). Research on written composition.
In M. C. Wittrock (Ed.), Handbook of research on teaching (3rd ed., pp.
778 – 803). New York: Macmillan.
Semel, W., Wiig, E. H., Secord, W., & Sabers, D. (1987). Clinical
Evaluation of Language Fundamentals—Revised. San Antonio, TX:
Psychological Corporation.
Strong, W. (1976). Close-up: Sentence combining. English Journal, 24,
56 – 65.
Strong, W. (1986). Creative approaches to sentence combining. Urbana,
IL: ERIC Clearinghouse on Reading and Communication Skills and the
National Council of Teachers of English. (ERIC Document Reproduc-
tion Service No. ED274985)
Utay, C., & Utay, J. (1997). Peer-assisted learning: The effects of coop-
erative and cross-age peer tutoring with word processing on writing
skills of students with learning disabilities. Journal of Computing in
Childhood Education, 8, 165–185.
Wong, B. (1994). Instructional parameters promoting transfer of learned
strategies in students with learning disabilities. Learning Disability
Quarterly, 17, 110 –119.
Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). Woodcock–
Johnson III Tests of Achievement. Itasca, IL: Riverside.
Zimmerman, B., & Kitsantas, A. (2002). Acquiring writing revision and
self-regulatory skill through observation and emulation. Journal of Ed-
ucational Psychology, 94, 660 – 668.
Received March 1, 2004
Revision received September 8, 2004
Accepted September 15, 2004 䡲
Wanted: Old APA Journals!
APA is continuing its efforts to digitize older journal issues for the PsycARTICLES database.
Thanks to many generous donors, we have made great strides, but we still need many issues,
particularly those published in the 1950s and earlier.
If you have a collection of older journals and are interested in making a donation, please e-mail
journals@apa.org or visit http://www.apa.org/journals/donations.html for an up-to-date list of the
issues we are seeking.
54 SADDLER AND GRAHAM
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.