The Effects of Sentence-Combining Instruction on the Writing of Fourth-Grade Students With Writing Difficulties
Bruce SaddlerBita BehforoozKristie Asaro
University of Albany
The Journal of Special Education
Volume 42 Number 2August 2008 79-90
? 2008 Hammill Institute on Disabilities
hosted at http://online.sagepub.com
One area of writing that may be particularly problematic for less skilled writers and writers with learning disabilitiesis constructing well-formed sentences. In this single-subject design study,sentence-combining practice with a peer-assistance component was used to improve the writing ability of 6 fourth-grade students with and without learningdisabilities. The results support the use of sentence-combining practice to increase sentence construction ability.Furthermore,sentence-combining instruction led to gains in story quality and writing complexity.Keywords:sentence combining; learning disabilities; writing disabilities; peer assistance
riting is an important tool for learning. In fact,academic progress in school depends on anadequate degree of writing fluency (Martlew,1983).Although many students struggle occasionally withwriting,for less skilled writers and writers with a learn-ing disability,writing is especially difficult (Graham& Harris,1989). One area of writing that may be par-ticularly problematic for a less skilled writer and awriter with a learning disability is constructing well-formed sentences. In comparison to their more skilledcounterparts,these students may produce sentencesthat are generally less syntactically complex,andtheir sentences typically contain more grammaticalerrors (Myklebust,1973). These writers also producesentences that are shorter; have higher percentages ofcapitalization,punctuation,and spelling errors; andare lower in overall quality than thoseof their moreskilled peers (Graham &Harris,1989; Houck &Billingsley,1989; Newcomer & Barenbaum,1991).Overcoming problems in sentence construction isimportant to young writers for several reasons. First,problems with sentence production skills may inter-fere with other processes such as planning,contentgeneration,and revising because attention devoted tolower level skills depletes available cognitive resourcesthat could be applied to higher level processes (Graham,
1997; Scardamalia & Bereiter,1986; Strong,1986).Second,lack of knowledge of effective writing for-mats at the sentence level hinders a writer’s abilityto translate his or her thoughts into text (Hayes &Flower,1986),affecting the complexity and coher-ence of the communication. Finally,difficulties con-structing well-designed,grammatically correct sentencesmay make the material the student writes more diffi-cult for others to read.
To mitigate such problems,Graham,Harris,MacArthur,and Schwartz (1998) suggest that lessskilled writers need to develop proficiency in framingtext within a variety of sentence formats,for example,by expressing their thoughts in the context of a com-plex sentence instead of a series of simple ones. Toaccomplish this,Andolina (1980) recommends thatteachers provide direct,stimulating language experi-ences to accelerate the development of syntacticalpatterns throughout the school years. One instructionalmethod that provides direct practice with sentenceconstruction skills is sentence combining. Sentence-combining practice requires the manipulation of
Authors’Note:Address correspondence to Bruce Saddler,University at Albany,Educational and Counseling Psychology,ED226,Albany,New York 12222; e-mail:email@example.com.
80The Journal of Special Education
ideas by a writer through rewriting and transformingbasic,or kernel,sentences (Strong,1976). For example,if a student characteristically composes simple kernelsentences such as “My dog is short. My dog is brown,”the student can learn through sentence-combiningpractice to change these sentences into more syntac-tically complex and mature sentences such as “Mydog is short and brown”or “The short brown dog ismine,”depending on what facet of the sentence thewriter wishes to emphasize.
Many studies involving sentence-combining researchhave been conducted,and with few exceptions,sen-tence combining has been shown to be an effectivemethod for helping students produce more syntacti-cally mature sentences (e.g.,Gale,1968; Hunt,1965;Mellon,1969; O’Hare,1973). In addition,sentencecombining has positively influenced the quality ofstudent writing across various modes of discourse(e.g.,Combs,1975; O’Hare,1973; Saddler & Graham,2005). There is also some evidence suggesting thatsentence combining has a positive effect on regularlyachieving students’revising skills (Horstman,1989).Despite the evidence that sentence combining can beused to improve students’sentence construction skills(Hillocks,1986; O’Hare,1973),sentence-combininginstruction is not currently included as a componentin popular approaches to writing instruction,such asWriters’Workshop (Pritchard,1987).
A recent study by Saddler and Graham (2005) is par-ticularly relevant to this investigation. These researchersassessed the effects of a sentence-combining procedureinvolving peer-assisted practice with more and lessskilled young writers. Forty-two students in the fourthgrade received either sentence-combining instruction orgrammar instruction. Students were paired for instruc-tion and received 30 lessons,25 minutes in duration,three times a week for 10 weeks outside of their regularclassrooms. The results indicated that in comparison topeers who received grammar instruction,students in theexperimental treatment condition became more adept atcombining simpler sentences to create more complexsentences. In addition,for the experimental students,thesentence-combining skills they were taught transferredto a story-writing task,resulting in improvements inwriting quality and revising ability.
The Present Investigation
This study replicates and extends the Saddler andGraham (2005) investigation. In this study,sentence-combining practice with a peer-assistance component
was used to improve the abilities of young studentswith weak writing skills to construct sentences andcompose stories higher in overall quality than storieswritten at pretest. However,we extended the Saddlerand Graham study in two ways. First,we wanted toimprove the generalization of the sentence-combiningskills to story writing. Campbell,Brady,and Linehan(1991) suggested that if generalization is to occur,training for generalization should be incorporatedas an integral part of the instructional program andshould be implemented either during instruction ordirectly after the objective criterion has been reached.In addition,Strong (1986) recommended the use ofparallel writing tasks,using the application of targetskills to create connections in the writer’s mind thatmay increase generalization. Therefore,we includedtwo additional sentence-combining practice exer-cises recommended by Strong that were not used bySaddler and Graham. Both exercises focused on para-graph production. The first prompted the students tocombine a series of kernel sentences into a paragraphfrom a series of kernel sentences. The second requiredthe students to combine phrases into sentences andthe sentences into a paragraph (these components arefully explained in the General Procedure section).We believed these additional practice componentswould better facilitate generalization of the sentence-combining skills to story writing.
Second,we modified the instructional pairings.Whereas in the Saddler and Graham (2005) study amore skilled writer in terms of sentence constructionability was paired with a less skilled writer duringinstruction in a peer-assisted framework (e.g.,Fuchs,Fuchs,& Thompson,2001),in this study,all of thewriters were considered less skilled and demonstratedsimilar writing abilities. This pairing was used to ascer-tain how less skilled writers might complete sentence-combining tasks without the benefit of immediatesupport from a more skilled student.
Teachers in the fourth grade at an urban elementaryschool in the Northeast were asked to nominate studentsin their classrooms who met the following criteria:(a)identified by the school as having a learning disabilityand (b) showed evidence of weak writing skills. Only 3children met both requirements; therefore,3 childrenwho did not have an identified learning disability but
who had weak writing skills nonetheless were includedin the study. The 6 students were randomly assigned topairs for instruction. Names have been changed to pro-tect the students’identities.
Pair 1 consisted of Bob and Carol. Bob was 10years 6 months and Caucasian; he had been identifiedin the second grade as having a learning disability. He had a Wechsler Intelligence Scale for Children–Revised (WISC-R; Wechsler,1974) performancestandard intelligence quotient (IQ) score of 98 andthe equivalent of a second-grade reading level accord-ing to the Wide Range Achievement Test(WRAT-3;Wilkinson,1993). His individualized education pro-gram (IEP) included several goals in written expres-sion. Carol was 9 years 1 month and African American;she had also been identified in the second grade ashaving a learning disability. Her WISC-R perfor-mance standard score was 95,and her WRAT-3 read-ing level was second grade. Her IEP also includedwritten expression goals.
Pair 2 consisted of Pete and Joni. Pete was 9 years 8months and African American. He had been identified 2years prior as having a learning disability. His WISC-Rperformance standard score was 84,and he had asecond-grade reading level according to the WRAT-3.He had goals on his IEP in written expression,and histeacher reported that his stories were very short anduninteresting. Joni was 9 years 4 months and Caucasian.Because she was not identified as needing any specialservices,no intelligence information was available. Herteacher nominated her for this study because she haddemonstrated weak writing abilities on in-class assign-ments and on a statewide assessment of writing. Her in-class stories were very short and her sentences verysimple,and her score on a writing sample from the statewriting examination indicated that she was below aver-age in her writing ability.
The third pair was Tom and Sam. Tom was 9 years9 months and African American. Sam was 9 years 5months and African American. Neither student wascurrently being served under any special educationcategory but had been nominated to participate in thisproject by their teachers because of weak writing abil-ity. The teachers reported that neither student enjoyedwriting nor invested themselves in planning or revis-ing. In addition,the stories they wrote were short andlacking in details. Finally,both had achieved scoreson a writing sample from the state writing examina-tion that indicated they were below average in theirwriting abilities.
To provide additional evidence of the participants’writing abilities,each student was administered the
Saddler et al. / Sentence Combining81
story-writing probe from the Test of Written Language–3(TOWL-3; Hammill & Larsen,1996). The writing sam-ple was analyzed using the TOWL-3 Story Constructionsubtest criteria. This subtest measures story-writingquality as revealed by plot,prose,development ofcharacters,interest to the reader,and several othercompositional aspects (coefficient alphas rangedfrom .88 to .90 for 9- to 11-year-olds). The stories werescored by two graduate students in educational psy-chology who were unaware of the research questions.To determine interrater reliability between the scoresassigned by the two raters,a Pearson product momentcorrelation coefficient was calculated. Interrater reli-ability between the two scorers was .95. The overallscores for each examiner were then averaged. For thepurpose of data analysis,raw scores were convertedto standard scores (M=10,SD=3) using the normativetables in the test manual. Three students,Bob,Pete,and Joni,achieved a standard score of 6 points. Carolachieved a standard score of 7 points,and Tom andSam achieved a standard score of 8 points.
The elementary school had a total population of 685students in Grades K through 5. The population wasdiverse,with 51.3% of the students being White;36.2% African American; 6.2% Hispanic; and 6.2%American Indian,Alaskan,or Pacific Islander. In addi-tion,4.0% of students received services in English as asecond language. Approximately 37.0% of the studentsreceived free lunch,and 12.0% received reduced-pricelunch. The three classes participating in this study hada general education teacher and a special educationcoteacher available during the school day. The generaleducation teachers,who were primarily responsible fordelivering instruction,had an average of 13 years ofteaching experience. The students received between 3and 4 hours of writing instruction per week,arrangedin a writing workshop approach.
Each student pair received 18 lessons,25 minutes induration,separated into three units of instruction,threetimes a week for 6 weeks,for a total of 450 minutes ofinstruction and practice. Instruction took place in thehallway near their classrooms,and students were toldthat they were receiving special instruction by a writ-ing coach to help them with their writing. Instructionwas delivered by the second author. Before the startof the study,the second author was trained by the firstauthor to implement instruction. The second author was
82The Journal of Special Education
Kernel Sentence Cluster Examples by Unit of
provided a notebook that contained detailed directionsfor implementing each activity and lesson (this includeda space to check off each step as it was completed).The sentence-combining treatment was adaptedfrom the curriculum used in the Saddler and Graham(2005) study. Instruction was segmented into threeunits consisting of six lessons each. Each unit taught aparticular type of sentence construction. In the firstunit,students combined kernel sentence clusters ofthree or more sentences by using adjectives. In thesecond unit,students combined sentences by insertingphrases,and in the third unit they used the connectorsbutand becauseto combine sentence kernels. SeeFigure 1 for examples from each unit of instruction.Lessons 1 and 2 procedures.The first lesson ofeach unit began with the instructor introducing andexplaining the sentence-combining strategy practicedin that unit. The instructor introduced sentence com-bining to the students as a trick good writers use tomake their sentences and stories easier to understandand more interesting. Following the introduction,there was a brief oral warm-up portion during whichthe instructor read a set of kernel sentences to thestudents. The instructor then modeled combining thefirst sentence pair. The students were asked to taketurns completing the remaining sentence pairs. If nei-ther student suggested a solution,the instructor pro-vided an answer and then moved to the next set ofsentences. After the warm-up sentences were com-pleted,the instructor asked the students to explainhow they knew to put the sentences together. The
instructor then stated what operations the studentsneeded to carry out in order to combine the sentences,namely,using a connecting word (if appropriate),get-ting rid of words they did not need,moving wordsaround,changing words if needed,or adding words.A written guided practice portion followed thewarm-ups. This practice consisted of the students com-bining sets of kernel sentences into a single sentence.The students were asked to combine the first set of sen-tences individually and then write out their answers onthe worksheets. They were instructed to stop after fin-ishing each set of sentences and read their answers toeach other out loud. While one student was reading hisor her response,the other student used hand signals torate the sentence. Ratings for each sentence wereestablished through discussion about the sound of thesentence. If the sentence sounded great the studentwould give a thumbs up,if it sounded okay the studentplaced his or her thumb parallel to the ground,and ifthe sentence did not sound right the student gave it athumbs down. If a sentence was a thumbs-down sen-tence,then the partner and the instructor discussed howit could be improved.
Lesson 3 procedures.The purpose of the third les-son in each unit was to transfer the skills the studentswere practicing via the sentence-combining exercisesto a revision task. The lesson started with an oralwarm-up time identical to that in Lessons 1 and 2.After the warm-up,the students assisted each other inproducing a revised paragraph from a series of kernelsentences that did not contain any combination cues.This exercise was not included in the Saddler andGraham (2005) study.
Lesson 4 procedures.The fourth lesson in each unitwas not included in the Saddler and Graham (2005)study. This lesson was designed to promote storyplanning and to provide additional practice with thesentence combination skills. The lesson began withthe same oral warm-up activity as in the previouslessons with one minor change; this time,both studentsattempted each of the warm-up problems to ascertainif each student could create a different solution foreach problem.
Next,each student was given a copy of a practicesheet that included random facts about a given topic.The facts were not written as whole sentences butinstead were short phrases. The students were instructedto pick out different sentence parts that made sensetogether and then combine the parts into a sentence.
Next,they were instructed to write out the combinedsentences into a paragraph about that topic.Lesson 5 procedures. Lesson 5 in each unit con-sisted of a sentence-combining progress-monitoringprobe created by the first author to measure the acqui-sition of the skills presented during the unit. The testincluded 10 kernel sentence clusters ranging from 3to 6 kernels long. Students were given 20 minutes tocomplete the test.
Lesson 6 procedures.During Lesson 6 of each unit,the students each wrote a story from pictures similarto the pictures used at pre- and posttest to further facil-itate transfer of the sentence-combining skills to con-nected writing.
Fidelity of Treatment Implementation
To ensure that the treatment was delivered withdocumented fidelity (R. H. Horner et al.,2005),thefollowing safeguards were implemented. First,theinstructor received intensive practice with the firstauthor in applying the instructional procedures. Second,the instructor met with the first author weekly to dis-cuss any problems that occurred in implementingprocedures. Third,the instructor was provided with achecklist that contained step-by-step directions foreach lesson. Each step was checked off when com-pleted. Examination of the checklists by the first authorshowed that the instructor completed 100% of thesteps in the sentence-combining treatment. Fourth,toreduce the possibility of instructional contamination,each classroom teacher was asked to not discuss theintervention with their students who were participatingin the study until the study ended. Furthermore,theteachers were not explained the particular objectivesof the study or the details of the intervention until thestudy concluded. Finally,each teacher was observedtwice by the first author during the time allocatedfor writing. No evidence of any sentence-combininginstruction was evident during these observations.
Four measures were used to document progress:sentence-combining ability,story quality,writing com-plexity,and instances of taught sentence-combiningconstructions in connected text. Students were testedin pairs. All of the measures were rated by two gradu-ate students in educational psychology who did not
Saddler et al. / Sentence Combining83
know the purpose of the study. To determine interraterreliability between the scores assigned by the tworaters,a Pearson product moment correlation coeffi-cient was calculated for each measure. The overallscores for each examiner were then averaged.Sentence-combiningmeasure.To assess changes instudents’sentence-combining skills,each student com-pleted Form A of the Sentence Combining subtestfrom the TOWL-3 (Hammill & Larsen,1996) beforethe start of the study. After instruction was completed,students were administered the alternate form (FormB) of this subtest. Forms A and B both contain 20 itemsthat require students to produce increasingly complexand grammatically correct written single sentences bycombining and integrating the meaning of two or moresentences together. Interrater reliability between thetwo scorers was .93. 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 connected text.Before instruction,immedi-ately following instruction,and after 3 weeks,studentswere asked to write stories. These compositions allowedus to examine how sentence-combining instructioninfluenced quality of writing,T-unit length (a writingcomplexity measure),and the instances of sentence-combining usage in connected text. Students wrote theirstories in response to picture prompts. To increase moti-vation,students were provided with a choice of twoprompts for each story written during baseline andposttest. The picture prompts were line drawings depict-ing one or more characters involved in an activity (i.e.,arabbit lecturing other rabbits about carrots and a girlpulling a dog in a wagon).
When writing a first draft,students were given thetwo pictures and directed to write a story about one ofthem. They were given 20 minutes to write their stories,which was consistent with the amount of time providedduring the Saddler and Graham (2005) study. Instructorsprovided no assistance,nor were students prompted touse any of the skills they were taught. Prior to scoring,compositions were typed (errors were not corrected),and identifying information was removed.
To examine treatment effects on the quality ofstudents’writing,each story was scored using a holis-tic quality-rating scale originally developed by Grahamand Harris (1989). Examiners were asked to read thepaper attentively to obtain a general impression ofoverall writing quality. Stories were then scored on an 8-point scale,with 1 representing the lowest quality
84The Journal of Special Education
of writing and 8 representing the highest quality.Examiners were told that ideation,organization,gram-mar,sentence structure,aptness of word choice,andmechanics should all be taken into account in forminga judgment about overall quality and that no one factorshould receive undue weight. Examiners were pro-vided with a representative paper (or anchor point) fora low-,middle-,and high-quality score. These paperswere the same used in the Saddler and Graham (2005)study. By discussing the distinguishing features ofeach of the anchor points,the first author trained theexaminers to use the quality-rating scale. Next,theexaminers practiced applying the scale to a series ofcompositions that varied widely in quality. After inde-pendently scoring each practice story,examiners com-pared their scores and resolved any differences throughdiscussion. Training continued until the examiners’scores differed by no more than 1 point on 10 consec-utive compositions. Interrater reliability was .85 betweenthe two examiners.
To quantify writing complexity,a ratio based onseminal work by Hunt (1965)—mean words per T-units—was used. Hunt defined each main clausealong with its subordinate and nonclausal elementsas a T-unit. Therefore,any simple sentence would beconsidered a T-unit,but a compound or complexsentence would consist of two or possibly more T-units. Interrater reliability was .75 between thetwo examiners.
To evaluate the extent the writers used the sentence-combining skills they were practicing while compos-ing their stories,the examiners counted the numberand type of sentence-combining constructions in eachstory. The author trained the examiners by practicingscoring papers that included examples of each type ofsentence-combining skill found in the treatment.Training continued until the examiners’scores differedby no more than 1 point on five consecutive composi-tions and the examiners agreed completely on the typeof sentence-combining skill demonstrated (i.e. adjec-tive,phrase insertion,connection). Interrater reliabilitywas .85 between the two examiners when they inde-pendently scored drafts from pretest and posttest.
To demonstrate experimental control,the interven-tion effects on the dependent measures were assessedvia a multiple baseline across-subjects design withmultiple probes during baseline (R. D. Horner &Baer,1978). Prior to the introduction of treatment,each student’s writing performance was measured
over time to establish a baseline of typical ability. Thefollowing conditions existed during the study:base-line (pretest),during which each student wrote a storyto establish pretreatment skill level; treatment,duringwhich instruction began for each pair of students aftereach child established a stable baseline for story qual-ity followed by 18 lessons; posttreatment (posttest),when students wrote three stories immediately fol-lowing instruction under the same conditions as dur-ing baseline; and maintenance,during which eachstudent wrote a story 3 weeks following the posttestunder the same conditions as during baseline.
Examples of a pretest and a posttest story from aparticipant are provided in Figure 2. Figure 3 summa-rizes story quality,and Figure 4 shows average lengthof the T-units in the stories written at pretest,post-treatment,and maintenance. Finally,Figure 5 revealsthe amount of constructions (phrase embeddings,adjectives,connectors) included. In addition to thevisual analysis of level and trend,the data were alsoanalyzed using the percentage of nonoverlapping data(PND) procedure outlined in Scruggs,Mastropieri,and Casto (1987). The following guidelines recom-mended by Scruggs et al. were used:90% of the post-treatment and maintenance points exceeding theextreme baseline value indicated a very effective treat-ment,70% to 90% an effective treatment,50% to70% a questionable treatment,and less than 50% anineffective treatment.
All of the students improved their ability to com-bine sentences. On average,the group achieved astandard score of 6.3 at baseline and improved to an11.5 average at posttest. Pete’s improvement was themost robust,as he climbed from a baseline score of 7to a posttest tally of 13; however,all of the studentsimproved by at least 5 points. The PND for this vari-able was 100%; therefore,the intervention was veryeffective in increasing sentence-combining ability,aswas expected. Interrater reliability was .93.
As can be seen in Figure 3,all of the studentsimproved the quality of their stories. Carol’s storieswere particularly improved,as she increased from alow of 2 quality points at baseline to a high of 6 during
Saddler et al. / Sentence Combining85
Pre- and Posttest Story Examples
Bob Pretest Story
One summer it was dogs they could talk. No one ever know that. They went for a boat ride
they was friends. People thought that they was gods but they wasn’t. They took a people in thereboat said that was awesome. Some started to laugh about it. They keep on paddling out. Theyhad a lonely boat ride then it turn dark. They still was paddling then they got scared and startedto tell stories. Then they were scared no more. Then at 5:00 they got across the land. They ownersaid they not done. Bob Posttest Story
One day there was a boy Jesse and a girl named Jessica. They was practicing hitting a ball because they had baseball practice. Jesse hit the ball and it almost hit a window. He it the ball again and it went over the roof and landed in the street. The car hit the ball. It rolled and he caught it. Then Jessica said that she could do better than Jesse then it was his turn again and thistime he hit the window. It broke and they was scared they didn’t know what to do so they hid the bat and glove behind there back but no one came out. There parents said what is that? He tried to take off his hat and put it on his face. Then the police came because of the alarm and they were scared. Then the parents come outside they seen the ball and the bat. Then for that they was punished for 3 weeks.
posttest. The PND for this variable was 87.5%,indi-cating that the intervention was effective in increasingstory quality. Interrater reliability was .85.
Figure 4 reveals that the T-unit length per sen-tence did increase for all of the writers. This indi-cates that in terms of T-unit length,the sentencesthey wrote became more complex following instruc-tion. The PND for this variable was 91.6%,indicat-ing that the intervention was effective in increasingthe complexity of the sentences. Interrater reliabilitywas .75.
during the posttest for several of the writers. Forexample,in one pretest story,Tom included a stringof adjectives to describe a dog falling from the sky(the big,hairy,mean,dumb-looking,fat dog).According to the scorers,this style did nothing toimprove the quality of his stories. At posttest,Tom’soverall use of adjectives declined,but the adjectivespresent were used to enhance the descriptions of thecharacters in a way that added value to the story interms of overall quality.
Identifying if sentence combining is a useful areato develop more automated and fluent composers hasgreat importance for our field. Ironically,unlike read-ing,where we have outstanding early interventionsfor phonological and phonemic skills,few empiricallyvalidated interventions on the sentence level exist inwriting. This research helps advance an evidenced-based approach for a potentially important writingsubskill and represents a balanced approach to writinginstruction that combines bottom-up and top-downapproaches.
In this research,two questions were asked. First,wouldsentence-combining practice with a peer-assistancecomponent improve the ability of young students withweak writing skills to construct sentences and compose
As reported in Figure 5,several of the studentsincreased their use of the constructions included inthe sentence-combining curriculum. Interrater relia-bility for number and type of sentence-combiningconstruction occurring in each story was .85.
The PND for this variable was 71%,indicating theintervention was effective in increasing the numberof taught constructions in the stories. Although theintervention did not have as strong of an effect in thisarea as hoped,the results were interesting. Of thethree constructions,the use of adjectives was themost prevalent during pretest and posttest. However,the effectiveness of adjective use seemed to improve
86The Journal of Special Education
Figure 3Story Quality
stories higher in overall quality and longer than storieswritten at pretest? Second,when working in a peer-assistance format,could less skilled writers effectivelycomplete sentence-combining tasks without the bene-fit of immediate support from a more skilled other?Several outcomes were consistent with the Saddlerand Graham (2005) study. As expected,supplementarypractice in sentence combining led to increased abilityto combine sentences for each of the students and morecomplex sentences in terms of T-unit length. Sentence-combining instruction has proven in this and other stud-ies to be an effective method of increasing skill in onecritical component of writing:sentence construction. Inaddition,in this study there were some improvementsin the overall quality of all of the stories written.
Unlike the Saddler and Graham (2005) study,in thisstudy the taught sentence-combining constructions didappear in the posttest writing task to a greater degreethan at pretest. However,although this result wasencouraging,there was actually little gain in the numberof constructions from pretest to posttest. There may beseveral reasons for this. First,the participants may nothave had enough practice opportunities during the inter-vention to internalize the constructions to the pointwhere they could be fluently recalled during the actualwriting process. Second,the generalization activities
may not have emphasized the transference of these skillseffectively. During Lesson 3,the participants revisedkernel sentences into a paragraph by using the taughtconstructions,and in Lesson 4,the students combinedphrases into whole sentences. Although both of theseactivities represented extensions of the exercises pre-sented in Lessons 1 and 2,they may have still been toofar removed from the actual writing process to be valu-able during connected writing.
Furthermore,even if the constructions had increasedto a greater degree,it is less clear what actual impacttheir presence may have on the overall holistic rating ofthe stories. Although it is tempting,it would be unwiseto suggest that the inclusion of more adjectives,embeddings,or connecting words by themselves wouldaccount for an improvement in overall story quality. Inaddition,the measure of use of sentence-combining skillwhile writing may in itself only be an indirect measureof linguistic features that might reasonably be expectedto change after sentence-combining instruction.
If taught constructions did not markedly increase,yet story quality improved,some other aspects of thesentence-combining treatment must have contributedto these improvements. Several explanations are possi-ble. First,perhaps while practicing the constructionand revision of sentences,the writers began to attend to
Saddler et al. / Sentence Combining87
Figure 4T-Unit Length
the options found in our language to a greaterdegree. Second,it may be that by practicing makingdecisions about the various ways sentences could becombined and the relative merit of particular combi-nations during the exercises the students becamemore aware of the decision-making process writersengage in when creating and revising prose. Finally,researchers (e.g.,Mellon,1969; Saddler & Graham,2005) have suggested that sentence-combininginstruction may make the process of sentence con-struction more automated and less effortful,therebyfreeing up cognitive resources for use in other writ-ing processes such as planning and composing (thehypothesis of cognitive load reduction). Furthermore,wise use of these cognitive resources will result in abetter written product. Because of the limited timeof the intervention,the research design did not allowexploration of this theory to a great degree.We also investigated the effect of pairing two writ-ers with similar abilities within a peer-assisted learningformat. In prior research (Saddler & Graham,2005),amore skilled writer was partnered with a less skilledwriter. This type of pairing allowed the more skilledwriter to provide support and grammatically correctsolutions to the less skilled partner when needed.Because much of the current intervention consisted ofthe partners crafting solutions together,this designallowed an examination of how two writers with lim-ited skill in writing would approach these tasks.
It is interesting that in each pairing one of thestudents acquired the target skill in each unit before theother and then proceeded to offer verbal and writtensupport to the partner. The nature of the supportchanged depending on the need. In one instance,forexample,a student read difficult words to his partner.In another case,a student spelled words for her partner.
88The Journal of Special Education
In a third case,a student offered words his partner coulduse as adjectives. Often during the warm-up exercises,one partner would offer alternatives to the other partnerif the partner was struggling with a particularly difficultcombination. Thus,in this study,the pairings led to sup-portive relationships and successful outcomes.
instructional activities on children in other gradelevels without learning disabilities or with disabili-ties who may present different areas of weakness isunknown. Third,only one writing genre was used.Fourth,fidelity of treatment was not collected by anindependent person. Fifth,one student’s (Bob’s)baseline story quality was trending upward during aphase change from baseline to instruction.
There are several limitations to this study. First,no peer discourse during instruction was collected;therefore,no formal analysis of the peer-to-peer interac-tions between students during instructional settings,which may help to explain the results,is possible.Second,the sample size was small and only repre-sented one grade level; therefore,the effects of these
Research in sentence combining presents a uniqueopportunity to explore the potentiality that what maybe needed in effective writing instruction research areinterventions that connect several different theore-tical approaches. In other words,writing research that
employs principles of direct and explicit instructionin the skilled aspects of writing (such as sentencecombining),while embedding this instruction withinstrategy frameworks that are built on cognitive andsociocultural theories employing a cognitive appren-ticeship approach to teaching,may be efficacious. Toexplore this possibility,extended research in sentencecombining is needed in several areas.
First,although we know that sentence-combininginstruction can improve the ability to write more var-ied and syntactically complex sentences,we do notknow yet how best to transfer this skill to connectedwriting. Therefore,future studies need to explore gen-eralization as a primary goal. Second,in exploring thegeneralization issue,researchers need to investigatethe metacognitive aspects of this instruction. We knowvery little about how a student approaches a sentence-combining task cognitively or how such instructionmay help reorganize existing ideas about sentenceconstruction. This information is crucial in designingactivities that may foster generalization. Third,futurestudies should record and distill the interactions betweenpairs of writers as they approach sentence-combiningtasks. Such information is vital in understanding thecognitive impact of sentence-combining skill as it is being acquired and also the effects of peer inter-action on this process. Fourth,curriculums taughtover a longer time period or longitudinal studies may strengthen the results of this intervention. Fifth,because this instruction was not taught by the class-room teacher,nor was it part of the normal classroominstruction,contextual studies of this instruction undernormal classroom conditions is warranted. Sixth,mul-tiple grade levels should be used to discern when suchinstruction might be implemented most effectively.Seventh,the context for developing sentence-combiningskill should be explored. Specifically,future studiescould examine if sentence-combining activities taughtin a less decontextualized,more interactive manner,wherein students are co-constructing texts and utiliz-ing sentence-combining skills within these texts withmore knowledgeable others,would be more effica-cious than more direct instruction approaches. Eighth,future research should include participants with vary-ing manifestations of learning disabilities. Ninth,interventions could include vocabulary instruction anda measure of vocabulary knowledge,as vocabulary isan important component of writing and could affectsentence-writing ability. Finally,because writing ispartially a social activity,the role of audience and pur-pose in a created story should be examined.
Saddler et al. / Sentence Combining89
Although there has been much attention given byresearchers to cognitive strategy instruction,text struc-tures,and process writing over the past several decades,little research has focused on one of the fundamentalbuilding blocks of good writing,namely,the ability towrite effective and complete sentences. This study rep-resents an important contribution in this area. Combinedwith recent work on the role of handwriting and spellingin developing writing fluency and more effective composition skills,sentence combining and other fundamental interventions are in some ways parallelconcepts to the role of efficient decoding and fluency oncomprehension processes in reading.
Furthermore,this study has potential to add to ourunderstanding of what counts as effective writing inter-vention. The need for evidenced-based interventions forimproved sentence construction is warranted,especiallyas part of a balanced writing program. The findingsfrom this study replicate and extend previous researchby showing that a peer-assisted sentence-combiningtreatment can improve the sentence construction skillsof young writers with weak writing skills. It also extendsthis research by showing that such instruction can pro-mote young students’use of sentence-combining skillsas they write and can positively affect the quality ofyoung students’writing. Finally,these results suggestthat less skilled writers can support one another effec-tively during peer-assisted practice.
Although sentence-combining exercises haveproven effective in increasing the syntactical fluency ofwriters,they only represent one component in a writ-ing program. These exercises cannot be a panacea forevery challenge writers will face when composing;therefore,teachers should not rely on them exclusively.They cannot replace other validated writing instructionpractices,nor are they a quick fix. However,sentence-combining exercises are valuable as a component of awell-rounded writing program that includes ample timefor writing,conferencing between peers and teachers,minilessons to increase skills,teacher modeling,andchoice in assignments.
Andolina,C. (1980). Syntactic maturity and vocabulary richnessof learning disabled children at four age levels. Journal ofLearning Disabilities,13,27–32.
Campbell,B. J.,Brady,M. P.,& Linehan,S. (1991). Effects ofpeer-mediated instruction on the acquisition and generaliza-tion of written capitalization skills. Journal of LearningDisabilities,24,6–14.
90The Journal of Special Education
Combs,W. E. (1975). Some further effects and implications ofsentence combining exercises for the secondary language artscurriculum. Dissertation Abstracts International,36,A1266.Fuchs,D.,Fuchs,L. S.,& Thompson,A. (2001). Peer-assistedlearning strategies in reading:Extensions for kindergarten,first grade,and high school. Remedial and Special Education,22,15–21.
Gale,I. F. (1968). An experimental study of two fifth grade lan-guage arts programs Dissertation Abstracts International,28,A4156.
Graham,S. (1997). Executive control in the revising of studentswith learning and writing difficulties. Journal of EducationalPsychology,89,223–234.
Graham,S.,& Harris,K. R. (1989). Improving learning disabledstudents’skills at composing essays:Self-instructional strat-egy training. Exceptional Children,56,201–214.
Graham,S.,Harris,K. R.,MacArthur,C.,& Schwartz,S. (1998).Writing instruction. In B. Wong (Ed.),Learning about learn-ing disabilities(2nd ed.). New York :Academic Press.
Hammill,D.,& Larsen,S. (1996). Test of Written Language–3.Austin,TX:Pro-Ed.
Hayes,J. R.,& Flower,L. S. (1986) Writing research and thewriter. American Psychologist,41,106–113.
Hillocks,G. (1986). Research on written composition.Urbana,IL:ERIC Clearinghouse on Reading and Communication Skills.Horner,R. D.,& Baer,D. M. (1978). Multiple-probe techniques:A variation of the multiple baseline. Journal of AppliedBehavior Analysis,11,189–196.
Horner,R. H.,Carr,E. G.,Halle,J.,McGee,G.,Odom,S.,&Wolery,M. (2005). The use of single-subject research to iden-tify evidence-based practice in special education. ExceptionalChildren,71,165–179.
Horstman,F. R. (1989). The effects of instruction in sentencecombining and revision on ninth and tenth graders explana-tory writing. Digital Dissertations,AAT 9012476.
Houck,C. K.,& Billingsley,B. S. (1989). Written expression ofstudents with and without learning disabilities:Differencesacross the grades. Journal of Learning Disabilities,22,561–575.Hunt,K. W. (1965). Grammatical structures written at threegrade levels(Research Report No. 3) Champaign,IL:National Council of Teachers of English.
Martlew,M. (Ed.). (1983). The psychology of written language:Developmental and educational perspectives.New York:JohnWiley & Sons.
Mellon,J. C. (1969). Transformational sentence combining:Amethod for enhancing the development of syntactic fluency inEnglish composition.Champaign,IL:National Council ofTeachers of English.
Myklebust,H. R. (1973). Development and disorders of writtenlanguage. In Studies of normal and exceptional children(Vol.2,p. 79).New York:Grune & Stratton.
Newcomer,P. L.,& Barenbaum,E. M. (1991). The written com-posing ability of children with learning disabilities:A reviewof the literature from 1980 to 1990. Journal of LearningDisabilities,24,578–593.
O’Hare,F. (1973). Sentence combining.Champaign,IL:NationalCouncil of Teachers of English.
Pritchard,R. (1987). Effects on student writing of teacher train-ing in the National Writing Project Model. WrittenCommunication,4,51–67.
Saddler,B.,& Graham,S. (2005). The effects of peer-assistedsentence-combining instruction on the writing performance ofmore and less skilled young writers. Journal of EducationalPsychology,97,43-54.
Scardamalia,M.,& Bereiter,C. (1986). Research on writtencomposition. In M. C. Wittrock (Ed.),Handbook of researchon teaching(3rd ed.,pp. 778–803). New York:Macmillan.Scruggs,T.,Mastropieri,M.,& Casto,G. (1987). The quantita-tive synthesis of single subject research. Remedial and SpecialEducation,8,24–33.
Strong,W. (1976). Close-up:Sentence combining. EnglishJournal,24,56–65.
Strong,W. (1986). Creative approaches to sentence combining.Urbana,IL:ERIC Clearinghouse on Reading andCommunication Skills and the National Council of Teachersof English.
Wechsler,D. (1974). WISC-R:Wechsler Intelligence Scale forChildren–Revised. New York:Psychological Corporation.Wilkinson,G. (1993). WRAT-3:Wide Range Achievement Test,administration manual. Wilmington,DE:Wide Range.Bruce Saddleris an assistant professor in the Department ofEducational and Counseling Psychology,Division of SpecialEducation,at the University at Albany. His current research inter-ests include writing disabilities,self-regulation,and inclusion.Bita Behforoozis a doctoral student in the Department ofEducational and Counseling Psychology at the University atAlbany. Her current research interests include children with writ-ing disabilities and second-language acquisition.
Kristie Asarois a doctoral candidate in the Department ofEducational and Counseling Psychology,Division of SpecialEducation,at the University at Albany. Her current research inter-ests include children with autism spectrum disorders,writing dis-abilities,and self-regulation.