Unless otherwise noted, the publisher, which is the American Speech-Language-Hearing Association
(ASHA), holds the copyright on all materials published in Perspectives on Language Learning and
Education, both as a compilation and as individual articles. Please see Rights and Permissions for
terms and conditions of use of Perspectives content: http://journals.asha.org/perspectives/terms.dtl
Sentence Combining: Assessment and Intervention
Applications
Cheryl M. Scott
Rush University Medical Center
Chicago, IL
Nickola Wolf Nelson
Western Michigan University
Kalamazoo, MI
Abstract
Clinicians can use sentence combining tasks to assess students’ written language needs
in the area of syntax and also as a context for language intervention. In this brief article,
we describe current efforts to develop a formal standardized measure of written syntactic
development and offer suggestions for using sentence combining as a context for
intervention targeting higher level syntactic complexity.
An essential component of literate language competence is the ability to comprehend
and formulate complex sentences to communicate about relationships among propositions or
chunks of content. A quick look at any newspaper or magazine shows many multi-clausal
sentences, coding relationships among propositions. For example, the following sentence
contains 32 words organized into 5 clauses and 1 T-unit:
The bad news touched discount and luxury stores alike, casting fresh doubt on the
survival of some chains and signaling what is likely to be the weakest holiday shopping
season in decades.
As school-age students and young adults grow in the ability to understand subtle
relationships between syntactically encoded ideas and events, the sentences they formulate
“grow” in length and complexity as well. Higher level literate syntactic formulation is marked by
the ability to combine, flexibly and fluently, several varieties of subordinate and coordinate
clauses within a single sentence (Berman, 2007; Nippold, Hesketh, Duthie, & Mansfield, 2005).
Skillful written language composition presents particular challenges to students with
language-learning disability (LLD). Students with typical language (TL) gradually transition
from producing greater syntactic complexity while speaking to producing more complex
sentences in writing (Perera, 1986; Scott & Windsor, 2000). By contrast, students with LLD
persist in showing greater complexity in their spoken than written language (Gillam &
Johnston, 1992), while producing written syntax that is less complex (Lewis, O’Donnell,
Freebairn, & Taylor, 1998; Nelson & Van Meter, 2007) and includes more errors (Gillam &
Johnston, 1992; Scott & Windsor, 2000) than that of their peers. In addition, students with
LLD produce a higher frequency of grammatical errors in written texts when compared to
spoken versions of the same content (Windsor, Scott, & Street, 2000). Microstructure
difficulties in using cohesive devices and complex grammatical structures, in fact, differentiate
stories produced by students with LLD better than macrostructure difficulties involving
omission of story grammar components (Liles, Duffy, Merritt, & Purcell, 1995). Students with
LLD also have difficulty adapting sentence forms to meet varied pragmatic demands, in which
14
“the specific grammatical form is partly determined by discourse needs and presuppositional
inferences” (Eisenberg, 2006, p. 153). Of particular relevance to the current discussion, the
spoken and written language produced by students with LLD can be distinguished by
limitations in the ability to combine different types of subordinate clauses within one sentence
(Gillam & Johnston, 1992; Marinellie, 2004).
The purpose of this article is to consider how a particular type of writing, which Hunt
(1977) called “rewriting” but others call “sentence combining” (Eisenberg, 2006; Strong, 1986),
can be used by clinicians to understand students’ written language needs, particularly in the
area of syntax. We describe current efforts to develop a formal standardized measure of written
syntactic development. We also offer suggestions for using sentence combining (hereafter SC)
as a context for intervention on higher level syntactic complexity.
Sentence Combining and Written Syntax
Assessment of written syntax requires a writing task that can be accomplished in a
short time frame and yields a written product that is relatively easy to evaluate. For this to
occur, clinicians need measurement techniques that can provide a valid and reliable reflection
of a child’s level of syntactic ability (see Scott, 2008, for a review of reliability and validity
issues associated with common syntax measures).
Story writing tasks are well suited to the purpose of probing students’ abilities to
generate ideas and use writing processes of planning, drafting, and revision to produce original
written stories. Original story probes can be gathered by asking students in a class to write a
story (real or imaginary) that tells about a problem and what happened (Nelson, Bahr, & Van
Meter, 2004; Nelson & Van Meter, 2002). The child’s writing processes are observed during the
sampling activity, and the resulting product is analyzed using a set of multi-level procedures
for characterizing writing processes and elements of discourse-, sentence-, and word-level
abilities. Such methods can provide rich information for planning intervention and measuring
progress, but they also are demanding in terms of the time and expertise required to gather
and analyze the samples. In addition, purposeful limitations on external control make it
difficult to generate normative data from such samples for comparing the performance of
students with special needs to data for peers with TL (see Nelson & Van Meter, 2007, for a
summary of grade-level data).
SC tasks offer an alternative to original composition as a means of assessing a student’s
written syntax proficiency. Hunt’s early work (Hunt, 1965, 1970) was based on analysis of
1000-word “free writing” samples produced during the normal course of the school day, but he
later pointed to advantages of using “rewriting,” rather than “free writing” (Hunt, 1977). When
researchers (or examiners) control language input, it becomes easier to compare output
produced by writers at different points of maturity. As Hunt described the task, “A student is
given a passage written in extremely short sentences and is asked to rewrite it in a better way”
(pp. 91-92). Controlled rewriting tasks make it possible to compare samples across writers who
begin with the same input. As Hunt wrote
When studying free writing, a researcher sees only the output. The input lies hidden in
the writer’s head. Its presence is conjectural and can only be inferred. But in rewriting,
one sees both input and output equally well. Neither is conjectural. (Hunt, 1977, p. 97)
As an example, Hunt (1977) described a rewriting task starting with six short sentences
as input: (1) Aluminum is a metal, (2) It is abundant, (3) It has many uses, (4) It comes from
bauxite, (5) Bauxite is an ore, and (6) Bauxite looks like clay. Hunt’s solution resonated with
the transformational grammar model predominant at the time (Chomsky, 1965), when
educators were accustomed to thinking about one-clause, unembellished “kernel” sentences
and the rules for combining these into longer sentences. Hunt provided an example of how a
typical fourth grade student might convey the controlled content:
15
Aluminum is a metal and it is abundant. It has many uses and it comes from bauxite.
Bauxite is an ore and looks like clay. (p. 95)
He contrasted this with an example of how a typical eighth grader might convey the
same content:
Aluminum is an abundant metal, has many uses, and comes from bauxite. Bauxite is
an ore that looks like clay. (p. 95)
A challenge facing researchers and clinicians, then and now, is how to capture the
differences between these two samples quantitatively. It is evident that simple word counts will
not work because the less mature fourth-grader conveys the information in 25 words, whereas
the more mature eighth-grader communicates the same content using only 20 words.
Computation of mean length of T-units (MLTU) using a measure of average sentence length in
words incorporated into “minimal terminable” units (T-units; first defined by Hunt in 1965),
does capture the difference. Hunt (1977) defined a T-unit as a single independent clause “plus
whatever other subordinate clauses or nonclauses are attached to, or embedded within, that
one main clause.” He defined an independent clause as “as a subject (or coordinated subjects)
with a finite verb (or coordinated finite verbs)” (pp. 92-93).
In the prior example, the fourth-grader produced 3 T-units with an MLTU of 8.3
words; whereas the eighth-grader encoded the same content using 2 T-units with an MLTU of
10 words.
SC tasks show promise as a means of quantifying students’ written syntax in a
standardized measure of language and literacy abilities. The technique is currently under
investigation in a set of coordinated subtests called “Reading and Writing the News” that are
part of a beta research edition of a Test of Integrated Language and Literacy Skills (TILLS;
Nelson, Helm-Estabrooks, Hotz, & Plante, 2007). “Reading the News” is designed to measure
reading fluency by asking the examinee to read the words of simple content units (similar to a
kernel, one-clause sentence) representing notes for a news story. The examinee is then asked
to “put these notes together to make a story that sounds better.” The SC process is modeled
with an example story. Students within a defined range of grade levels receive the same input.
The figure below provides examples of output for pairs of second grade and sixth grade
students.
Figure 1. Examples of rewritten stories produced by student with typical language (TL) or language-learning
disability (LLD)
Second Grade
Student with TL
Student with LLD
The class has a pet hamster. / It got out
because the cage door was open. / The
children looked for him and found him and put
him back in his cage. / So he found a corner
and went to sleep.
The class has a pet. / It is a hamster. / It got out. /
The cag[sp] was open. / The door was open. /
The children looked. / The children found him. /
They put him back. / They put him in the cage. /
He found a corner. / He went to sleep.
11 Input units – 0 omitted = 11 Output units
Output units/T-units = 11/4 = 2.75 SC ratio
Total words/T-units = 39/4 = 9.75 MLTU
11 Input units – 0 omitted = 11 Output units
Output units/T-units = 11/11 = 1.0 SC ratio
Total words/T-units = 46/11 = 4.18 MLTU
16
Sixth Grade
Our school was closed last Wednesday, on a
school day. / It was six o’clock am when the
janitor opened the school. / He smelled
something strong, / and it was a skunk. / It
almost knocked him over. / He opened the
doors, left them open, and searched. / He
looked in the cafeteria and found two hungry
skunks eating cookies. / He called animal
control right away. / The workers came, took
the skunks, and let them go in the woods. /
A school was closed by Wednesday / and it was
closed all day. / A Janitor came inside a
classroom that he smelled something strong
[awk]/ he thought there was a skunk. / the Janitor
wint[sp] inside the school cafeteria / he saw two
skunks eating food. / The Janitor called the
animal control department[sp] to come get the
skunks. / The animal workers came and toulk[sp]
the skunks and left[sp] them go in the woods./
24 Input units – 2 omitted = 22 Output units
Output units/T-units = 22/9 = 2.44 SC ratio
Total words/T-units = 84/9 = 9.33 MLTU
24 Input units – 11 omitted = 13 Output units
Output units/T-units = 13/8 = 1.65 SC ratio
Total words/T-units = 69/8 = 8.63 MLTU
The best techniques for analyzing these written language samples are still under
investigation, but Figure 1 illustrates techniques that hold promise. The first steps are to count
the number of input content units that are maintained in the student’s output. The student’s
written language output also is divided into T-units. The resulting totals are used to compute a
sentence combining (SC) ratio by dividing the number of output content units by the number of
T-units. Mean length of T-unit (MLTU) is computed as well by dividing the total number of
words by the number of T-units. Both the SC ratio and MLTU can be compared across
students. At the second-grade level in this example, both students produced all 11 input
content units. In fact, except for one spelling error, the student with LLD copied the content
units exactly, yielding exactly the same number of output content units and T-units, with an
SC ratio of 1.0 and MLTU of 4.18. The second grade student with typical language (TL)
conveyed all 11 content units in 39 words and 4 T-units, yielding an SC ratio of 2.44 and an
MLTU of 9.75 words. The data for the sixth-grade dyad showed that the student with LLD
omitted almost half of the content units, combining the remaining content into 8 T-units, with
an SC ratio of 1.65 and MLTU of 8.63; whereas the student with TL retained almost all of the
original content, combined into 9 T-units, with an SC ratio of 2.44 and MLTU of 9.33. These
two examples show how the measures might work, even if a student with less mature syntax is
skilled at copying all the words. If this pattern continues across the data set, it is likely that SC
ratios and MLTU can be used to establish normative expectations for students at each
age/grade level, making it possible to evaluate written syntax ability relative to standardized
data. Underscoring the potential of this task to differentiate levels of maturity, we found
evidence in a pilot study of a robust developmental course for SC for students with TL ranging
from 7 to 15 years. Importantly, the syntactic measures based on SC samples correlated
significantly with syntactic complexity measures in the same students’ original written story
samples (Scott, Nelson, Andersen, & Zielinski, 2006)
Intervention Strategies
In our clinical experience, the ability to turn ideas into well crafted written sentences
with flexibility and fluency is one of the last “bastions” of language struggle seen in students
with LLD. Several reasons no doubt account for this, some of which are non-linguistic (e.g.,
inadequate information base, lack of conceptual understanding, compromised motivation).
Nevertheless, we have encountered many students who can explain content orally using
complex syntax, but who struggle to express the same thoughts in writing. One explanation is
that these students have difficulty in the metalinguistic processes required for the deliberate,
thoughtful, and decontextualized act of writing. For these students, SC offers an intervention
context that is uniquely suited to develop complex sentence fluency. Because SC, by nature,
requires students to manipulate clause combinations consciously via rules of deletion,
17
substitution, insertion, and rearrangement, sentence construction is facilitated at the
metalinguistic level required for writing.
In recognition of these benefits, SLPs in recent years have made increasing use of SC
techniques. Nelson and colleagues have included SC as part of their regimen of sentence-level
mini-drills and online scaffolding under the umbrella of the writing process framework that
forms the basis of the writing lab approach and related interventions (Nelson et al., 2004;
Nelson & Van Meter, 2006; Nelson, Roth, & Van Meter, 2008). Practice in controlled contexts
can provide a prelude to scaffolding students to combine their original ideas syntactically to
convey higher level logical, temporal, and causal relationships. Eisenberg (2006) recommended
using SC as one of several formats to encourage complex syntax. Scott and Balthazar (2008)
currently are investigating SC as one of several techniques in a feasibility study concerned with
teaching complex syntax fluency.
Several characteristics of SC broaden their appeal for SLPs. SC exercises can be
designed to vary in difficulty level and address a variety of syntactic targets so that an SLP can
match the exercise to a student’s individualized needs. To illustrate, the first example below
targets clause connectivity via adverbial clauses. It is a straightforward path to get from the two
kernels to the combined version by adding the second clause at the end of the first,
substituting an appropriate pronoun in the second clause to avoid redundancy, and choosing a
subordinate conjunction that makes sense semantically.
The student found something else to do the day of the debate team try-outs.
The student did not have much confidence in his public speaking ability.
The student found something else to do on the day of the debate team try-outs because
he did not have much confidence in his public speaking ability. (combined version)
The second example targets clause connectivity via center-embedded object relative
clauses and is more difficult. The student would have to determine where to insert the second
clause and appreciate that the noun being relativized (i.e., the SUV) is the grammatical object of
the second clause rather than the subject.
The new SUV was nearly totaled in the accident.
My brother bought the SUV last Saturday.
The new SUV that my brother bought last Saturday was nearly totaled in the accident.
(combined version)
Several systematic reviews and meta-analyses have been conducted regarding the
evidence base for using SC techniques in general and special education. Andrews et al. (2006)
located 18 studies on SC that met criteria for review, justifying the conclusion that, for
students between the ages of 5 and 16, SC provides an effective means for improving syntactic
maturity in writing (usually measured in terms of sentence length). They cited an early study
by O’Hare (1973) in which seventh graders were assigned randomly to experimental and control
groups, identifying it as the best study to date. Students in the experimental group who
received SC training obtained significantly higher scores on six measures of syntactic maturity
following training and performed similarly to 12th grade students on several measures. In a
recent analysis of writing instruction techniques for adolescents, Graham and Perin (2007)
found a weighted mean effect size of .50 (moderate) for SC (across 5 studies). Saddler and
Graham (2005) randomly assigned 4th-grade writers to either SC or traditional grammar
instruction (teaching parts of speech) groups. Although no treatment effect was found for
quality ratings of first drafts, students in the SC group performed better on post-treatment SC
tasks and also on their ability to improve writing when they revised their papers. In this study
as well as several previous projects, SC had its greatest impact on students with the lowest
pre-treatment written language scores.
18
Summary
Although sentence combining is an old technique in the writing instruction arsenal of
teachers and special educators, it is a technique that continues to deliver. For SLPs in
particular, there is promising new evidence that SC tasks may have value as part of an
assessment protocol for students with language learning disabilities. In addition SC tasks hold
promise as intervention activities that target written sentence level complexity in a
metalinguistic format needed by many students.
References
Andrews, R., Torgerson, C., Beverton, S., Freeman, A., Locke, T., Law, G., et al. (2006). The effect of grammar teaching on
writing development. British Educational Research Journal, 32, 39-55.
Berman, R.A. (2007). Developing language knowledge and language use across adolescence. In E. Hoff & M. Shatz (Eds.),
Handbook of language development (pp. 346-367). London: Blackwell.
Chomsky, N. (1965). Aspects of a theory of syntax. Cambridge, MA: M.I.T Press.
Eisenberg, S. L. (2006). Grammar: How can I say that better? In T. A. Ukrainetz (Ed.), Contextualized language intervention:
Scaffolding PreK-12 literacy achievement (pp. 145-194). Eau Claire, WI: Thinking Publications.
Gillam, R. B., & Johnston, J. (1992). Spoken and written relationships in language/learning impaired and normally achieving
school-age children. Journal of Speech and Hearing Research, 35, 1303-1315.
Graham, S., & Perin, D. (2007). A meta-analysis of writing instruction for adolescent students. Journal of Educational
Psychology, 99, 445-476.
Hunt, K. W. (1965). Grammatical structures written at three grade levels. Urbana, IL: National Council of Teachers of English.
Hunt, K. W. (1970). Syntactic maturity in school children and adults. Monographs of the Society for Research in Child
Development, No. 134.
Hunt, K. W. (1977). Early blooming and late blooming syntactic structures. In C. R. Cooper & L. O’Dell (Eds.), Evaluating
writing: Describing, measuring, judging (pp. 91-106). Urbana, IL: National Council of Teachers of English.
Lewis, B. A., O’Donnell, B., Freebairn, L. A., & Taylor, H. G. (1998). Spoken language and written expression—Interplay of
delays. American Journal of Speech-Language Pathology, 7(3), 77-84.
Liles, B. Z., Duffy, R. J., Merritt, D. D., & Purcell, S. L. (1995). Measurement of narrative discourse ability in children with
language disorders. Journal of Speech and Hearing Research, 38, 415-425.
Marinellie, S. A. (2004). Complex syntax used by school-age children with specific language impairment (SLI) in child-adult
conversation. Journal of Communication Disorders, 37, 517-533.
Nelson, N. W., Bahr, C. M., & Van Meter, A. M. (2004). The writing lab approach to language instruction and intervention.
Baltimore: Paul H. Brookes.
Nelson, N. W., Helm-Estabrooks, N., Hotz, G., & Plante, E. (2007). Test of Integrated Language and Literacy Skills (beta
research edition; not for distribution). Baltimore: Paul H. Brookes.
Nelson, N. W., Roth, F. P., & Van Meter, A. M. (2008). Written composition instruction and intervention for students with
language impairment. In G.A. Troia (Ed.), Instruction and assessment for struggling writers (pp. 187-212). New York: Guilford.
Nelson, N. W., & Van Meter, A. M. (2002). Assessing curriculum-based reading and writing samples. Topics in Language
Disorders, 22(2), 35-59.
Nelson, N. W., & Van Meter, A. M. (2006). The writing lab approach for building language, literacy, and communication
abilities. In R. McCauley & M. Fey (Eds.). Treatment of language disorders in children: Conventional and controversial
interventions (pp. 383-421). Baltimore: Paul H Brookes.
Nelson, N. W., & Van Meter, A. M. (2007). Measuring written language ability in narrative samples. Reading and Writing
Quarterly, 23, 287-309.
Nippold, M. A., Hesketh, L. J., Duthie, J. K., & Mansfield, T. C. (2005). Conversational versus expository discourse: A study of
syntactic development in children, adolescents, and adults. Journal of Speech, Language, and Hearing Research, 47, 1048-1064.
O’Hare, F. (1973). Sentence combining: Improving student writing without formal grammar instruction. Research Report No. 15.
Urbana, IL: National Council of Teachers of English.
Perera, K. (1986). Language acquisition and writing. In P. Fletcher & M. Garman (Eds.), Language acquisition (2nd ed., pp. 494518). Cambridge, UK: Cambridge University Press.
19
Saddler, B., & Graham, S. (2005). The effects of peer-assisted sentence-combining instruction on the writing performance of
more and less skilled young writers. Journal of Educational Psychology, 97, 43-54.
Scott, C. (2008). Language-based assessment of written language. In G. A. Troia (Ed.), Instruction and assessment for struggling
writers (pp. 358-385). New York: Guilford.
Scott, C., & Balthazar, C. (2008, November). Building sentence complexity: Evidence-based approaches with school-age
children and adolescents. A paper presented at the annual meeting of the American Speech-Language-Hearing Association,
Chicago, IL.
Scott, C., Nelson, N., Andersen, S., Zielinski, K. (2006, November). Development of written sentence combining skills in schoolage children. Poster presented at the annual meeting of the American Speech-Language-Hearing Association, Miami, FL.
Scott, C., & Windsor, J. (2000). General language performance measures in spoken and written narrative and expository
discourse of school-age children with language learning disabilities. Journal of Speech, Language, and Hearing Research, 43,
324-399.
Strong, W. (1986). Creative approaches to sentence combining. Urbana, IL: ERIC Clearinghouse on Reading and Composition
Skills and the National Conference on Research in English. (ERIC Document Reproduction Service No. ED274985)
Windsor, J. Scott, C., & Street. C. (2000). Verb and noun morphology in the spoken and written language of children with
language learning disabilities. Journal of Speech, Language, and Hearing Research, 43, 1322-1336.
20