Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/221735881

The Narrative Language Performance of Three Types of At-Risk First-Grade


Readers

Article  in  Language Speech and Hearing Services in Schools · January 2012


DOI: 10.1044/0161-1461(2011/11-0024) · Source: PubMed

CITATIONS READS

16 588

3 authors, including:

Teresa Ukrainetz
Utah State University
42 PUBLICATIONS   1,408 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Phonological Awareness View project

All content following this page was uploaded by Teresa Ukrainetz on 07 March 2019.

The user has requested enhancement of the downloaded file.


LSHSS

Article

The Narrative Language Performance


of Three Types of At-Risk
First-Grade Readers
Melissa M. Allen, a Teresa A. Ukrainetz, a and Alisa L. Carswell b

Purpose: This study investigated the narrative language per- of different words, and number of communication units. The
formance of 3 types of readers who had been identified as being narratives of early resolvers and good responders differed sig-
at risk through code-based response-to-intervention (RTI) nificantly on the productivity index, number of coordinating
procedures. conjunctions, and number of episodic elements. There were no
Method: In a retrospective group comparison, 32 at-risk other significant differences.
1st-grade readers were identified: children who resolved with- Conclusion: Types of learners distinguished by a code-based
out intervention (early resolvers, n = 11), children who met RTI model showed differences in their narrative language. First
criterion following 4 weeks of intervention (good responders, graders who responded well to code-based reading interven-
n = 8), and children who failed to meet criterion following tion retold stories that contained more language and better story
4 weeks of intervention (poor responders, n = 13). A narrative grammar than first graders who did not respond well to inter-
retell and a norm-referenced language test were obtained vention. These results indicate the need to evaluate narrative
before intervention. language performance within RTI, especially for early resolvers.
Results: There were no significant differences between the
3 learner types on the language test. However, the narratives
of the good responders were significantly higher than the nar- Key Words: response to intervention, reading, narratives,
ratives of the other 2 groups on total number of words, number language

he identification of children who are at risk for aca- difficulties. This study examined the narrative language per-

T demic difficulties is changing with the emergence of


response to intervention (RTI). The current focus of
RTI is to identify and ameliorate children’s decoding and
formance of three types of learners who had been identified
as being at risk for reading through RTI procedures in first
grade.
reading fluency difficulties before referral to special educa-
tion. These aspects of reading involve phonological process- RTI Procedures
ing and other aspects of language (Cunningham & Stanovich,
1998). However, much of language is not examined in the The 2004 reauthorization of the Individuals with Disabil-
current decoding- and fluency-focused RTI models. As a ities Education Act (IDEIA; PL 108-446) specifically per-
result, the language needs of this new population of RTI- mitted local education agencies to use an RTI process to
identified at-risk readers may be overlooked, particularly determine whether a child has a specific learning disability.
subgroups that may appear to have resolved their initial Alternative reading evaluations that focus on the learning
process are used within the RTI framework. RTI is intended
to significantly improve the reading performance of the
a nation’s children, which is accomplished through multiple
University of Wyoming, Laramie
b
STRIDE Learning Center, Cheyenne, WY avenues: identifying early reading difficulties, resetting the
Correspondence to Melissa M. Allen: mallen20@uwyo.edu learning trajectories of children with mild deficits, determin-
Editor: Janna Oetting ing which children have inherent significant reading dis-
Associate Editor: Phyllis Schneider abilities, and reducing the number of children who need
Received April 21, 2011 special education. Research evidence suggests that RTI mod-
Accepted September 28, 2011 els can accomplish all of these purposes (Denton & Mathes,
DOI: 10.1044/0161-1461(2011/11-0024) 2003; Gersten et al., 2008). However, the application of

LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012 * American Speech-Language-Hearing Association 205
RTI in the schools is still in the early stages, with many Decoding skills and reading comprehension come together in
variations in administration and implementation challenges fluent, skilled reading. However, there are important differ-
(Fuchs, 2003; Marston, 2005; Mellard, McKnight, & Woods, ences between the two domains.
2009; Ukrainetz, 2006a). Decoding means “breaking the code” and deals with rec-
Children who are struggling with reading are identified ognizing the printed word. It involves the print-specific skills
early in their schooling as having a trajectory that will not of alphabet knowledge, letter-sound correspondence, and
allow them to meet year-end benchmarks. In the RTI process, spelling conventions. Decoding requires a degree of lan-
educators intervene by teaching particular aspects of reading guage facility. Reading a printed word involves the language
and continuously monitoring a child’s response to this in- processes of holding phonological representations in mem-
struction (Fuchs & Fuchs, 2006; Fuchs, Fuchs, & Compton, ory, separating and manipulating phonemes, and retrieving
2004; Fuchs, 2003). Although instruction can be individual- the meaning attached to a phonological representation (Adams,
ized and optimized based on learner response, akin to lan- 1990). Deciding how to read a single printed word taps the
guage treatment, it is more typically delivered for set periods linguistic skills of vocabulary, syntax, and discourse (e.g.,
of time in a standardized manner with high instructional deciding whether to read the word read as /rid/ or /rEd/).
fidelity (Troia, 2005). The multitier RTI process begins by Reading comprehension involves putting the recognized
evaluating the performance of all children within a grade words together to gain sentential and textual understand-
level. In Tier I, at-risk readers initially receive classroom ing. Reading comprehension has cognitive, attentional, and
instruction with regular evaluation. Progress for some of the situational processes involved, but it is essentially language
children accelerates so they meet benchmarks by the next (Kamhi, 2009). Language is a huge domain that is diffi-
evaluation. These children are considered to have been solely cult to delimit; most educational learning involves written
experientially deficient and continue with only regular class- or spoken language (American Speech-Language-Hearing
room instruction. The children still not meeting expecta- Association [ASHA], 2001; Ukrainetz & Fresquez, 2003).
tions then receive supplementary instruction in Tier II. This Language-in-print differs in some significant ways from
instruction may be provided as differentiated small-group spoken language, such as in patterns of syntactic complexity
instruction with more intense monitoring. Children not re- and decontextualization (Scott, 1995; Westby, 1985), but
sponding sufficiently to the second tier of instruction may essentially language is still language: a complex, dynamic,
be provided with the even more intense Tier III intervention. multidimensional, rule-governed, conventional representa-
Children who do not respond satisfactorily to intensive tional system that crosses representational modalities.
supplementary support are considered for identification as The two aspects of reading, decoding and comprehension,
reading disabled and eligibility for long-term special edu- differ in their length and nature of acquisition. Decoding and
cational support (Swanson & Howard, 2005). fluency are constrained skills in that they are learned fairly
Valid decisions about learning responses in RTI can only quickly and thoroughly (Paris, 2005). Early code-related
be made if quality instruction has been provided. Quality skills are relatively easy to teach, and children can show large
instruction involves research-validated curriculum and gains in letter names, letter sounds, and phonemic awareness.
practices. Tier I instruction addresses all five “pillars” of By third grade, typical achievers have generally acquired
instruction: phonemic awareness, phonics, vocabulary, fluent decoding skills.
comprehension, and fluency (National Institute of Child In contrast, the language skills and world knowledge that
Health and Human Development [NICHD] National Read- comprise reading comprehension begin in infancy with con-
ing Panel, 2000). Tier II (and III) intervention is primarily versational interactions, grow in the school years with oral
code based. In kindergarten, instruction focuses on letter- and print experiences, and, depending on life experiences
sound correspondence, phonemic awareness, and basic word and learner inclinations, continue to improve through life.
decoding skills. In first and second grade, more advanced These skills are called unconstrained because of the lack of
word decoding skills and rapid, accurate oral reading are a specified mastery criterion: More is usually better (Paris,
addressed. This selective instructional focus is needed to 2005). In addition to a much longer learning trajectory, lan-
provide the intensity and therapeutic attention needed for guage is different from decoding in its modifiability. Specific
at-risk readers (Ukrainetz, 2006a). However, it may also vocabulary may be acquired easily through lived experi-
result in overlooking children who may be at risk for other ence (e.g., acquiring a dog leads to an array of canine-related
important reading skills, for example, comprehension. vocabulary). However, the rest of language is more resis-
tant to change, particularly through school lessons (see
Reading and Language in a Code-Based Weaver, 1996, for the many fruitless efforts over decades to
improve students’ written grammar).
RTI Model
Based on the differences between decoding and com-
To read fluently and understand what is read, one must prehension, RTI models concentrated on one domain may
have competency in two broad domains: decoding and not predict performance on the other domain, thus missing
comprehension (Gough & Tunmer, 1986; Perfetti, 1985). important information from half of the reading equation. The

206 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
only study located that examined this is Al Otaiba and Fuchs allow quick and reliable administration and scoring. Global
(2006). Al Otaiba and Fuchs examined children’s respon- norm-referenced tests often provide separate receptive and
siveness to code-based instruction and investigated the learner expressive scores, but expressive skills are usually limited
characteristics associated with responsiveness. Measures of to single word and sentence responses.
fluency, letter sounds, and phoneme segmentation in kinder- An alternative approach to measuring children’s language
garten and of oral reading in first grade were used to iden- skills is to use discourse tasks that evaluate these compo-
tify responsiveness by considering growth from pretreatment nents in integrated, functional wholes. Narratives are an
to posttreatment. Children were identified as always respon- important genre of discourse that can tap multiple features
sive, meeting the end-of-year criterion for kindergarten and levels of language simultaneously (Hoffman, 2009a;
and first grade; never responsive, never meeting the end-of- Ukrainetz, 2006b). A variety of narrative analyses have dif-
year criterion for kindergarten or first grade; or sometimes ferentiated children with typically developing (TD) language
responsive, meeting the end-of-year criterion for kinder- from those with language impairment (LI) (e.g., Fey, Catts,
garten or first grade, but not both. Receptive vocabulary Proctor-Williams, Tomblin, & Zhang, 2004; Gillam &
and sentence imitation, along with problem behavior, naming Johnston, 1992; Liles, 1985; Newman & McGregor, 2006;
speed, and amount of intervention, predicted reading fluency Scott & Windsor, 2000). Narrative performance is typically
responsiveness after 2 years of intervention. The extended evaluated through language sampling procedures that may
period of intervention time within the Al Otaiba and Fuchs be standardized and may even have age comparison infor-
study differed from the RTI goal of short-term identification mation (e.g., Miller & Chapman, 2004; Strong, 1998), but
and amelioration; however, it suggests that language may the open-ended responses and the detailed descriptive results
be predictive of reading performance within an RTI model. are not typically amenable to the strict psychometric proce-
An important consideration is that the language measures dures of norm-referenced testing. An exception to this is
and scoring systems used in the study provided a very limited the Test of Narrative Language (TNL; Gillam & Pearson,
view of language performance. 2004), which samples whole narratives, but performance is
Reading competence involves both a set of decoding still reduced to overall receptive and expressive scores.
skills and the large domain of comprehension. Because of One type of narrative task, a retell of a previously pre-
the differing natures of these two domains, code-based RTI sented story, has both comprehension and production demands.
models may be missing the linguistic side of reading. The model narrative presents the events, plot structure, and
words that the narrator is to retell, which allows more reliable
scoring than a generated story that can go in many directions.
Language and Reading
Retelling requires language, attention, comprehension, and
Young or poor readers may try to bypass decoding diffi- memory for the original story. Despite the support of the
culties by tapping into the linguistic side of reading through source story, as narrators retell the story, they filter the
the spoken modality. There are strong positive relations events through their own linguistic and cognitive resources
between oral language and both decoding and reading com- (Mandler & Johnson, 1977; Slobin & Welsh, 1973). Thus,
prehension (Bishop & Adams, 1990; Catts, Fey, Tomblin, & narrative retells may be a particularly revealing elicitation
Zhang, 1999, 2002; Catts, Fey, Zhang, & Tomblin, 2001; task for investigating relations between children’s reading
Nation, Clarke, Marshall, & Durand, 2004; Roth, Speece, & and their oral language.
Cooper, 2002; Snowling, 2005): Vocabulary has been con- Microstructure analysis. Narrative retells can be evalu-
sistently implicated in reading performance (NICHD National ated at microstructural and macrostructural levels. Micro-
Reading Panel, 2000); isolated measures of syntax have been structure is the sentence-internal structure of a narrative
found to predict reading difficulties (Badian, McAnulty, (Justice et al., 2006). It involves measures of vocabulary
Duffy, & Als, 1990; Catts et al., 2001; Vellutino et al., 1996); (e.g., quantity and variety of words), morphological structure
and a measure of grammatical complexity obtained during (e.g., inflections and derivations), and syntax (e.g., con-
conversational samples was found to contribute unique vari- junctions, noun phrases, and dependent clauses). Micro-
ance to later reading skills (Scarborough, 1990). In addition, structural measures that have differentiated children with
children with and without reading difficulties have been language difficulties from children with TD language include
differentiated on narrative episodic structure (e.g., Feagans total number of words (TNW), number of different words
& Short, 1984; Ripich & Griffith, 1988; Roth & Spekman, (NDW), number of utterances, mean length of utterance
1986). (MLU), and grammatical accuracy (e.g., Fey et al., 2004;
Children’s language skills can be measured in many Newman & McGregor, 2006; Scott & Windsor, 2000).
ways. Norm-referenced tests typically take a discrete skill These individual measures have shown some utility in de-
approach that elicits components of language in isolation scribing language and distinguishing groups; nonetheless,
and combines the results into global age-referenced scores, grouping outcomes from two or more measures may produce
thus evaluating children’s general language skills. Norm- more robust group differentiation (e.g., Puranik, Lombardino,
referenced tests involve strictly standardized procedures that & Altmann, 2008). A common grouping is by the language

Allen et al.: Narratives of At-Risk Readers 207


domains of semantics and syntax, though some frequently of causally related elements (Mandler & Johnson, 1977;
used measures draw on both domains. MLU, for example, will Stein & Glenn, 1979). These story elements (i.e., setting, ini-
be affected by both the size of the child’s vocabulary and tiating event or complication, internal response, plan, attempt,
his or her phrasal structure available to construct utterances. consequence, and reaction) are organized into problem-
There may be ways of combining these measures that are solution units called episodes. A complete episode is a three-
more revealing of patterns of language use, such as that of part structure that consists of a complication, an attempt to
Justice et al. (2006). resolve the complication, and a consequence (Peterson &
Justice et al. (2006) empirically investigated how a com- McCabe, 1983).
prehensive array of microstructural measures could be Story grammar studies have shown mixed findings con-
parsimoniously combined in a few summative indices to cerning children with LI. In story generation and retell tasks,
provide a comprehensive and convenient measure of micro- Merritt and Liles (1987) found that the number of com-
structure for children’s oral narratives. During the norming plete episodes and number of story grammar elements differ-
process for the TNL, narratives were obtained from 250 chil- entiated 9- to 11-year-old children with and without LI. Other
dren ages 5 to 12 years (Gillam & Pearson, 2004). The oral studies have failed to find reliable story grammar differ-
narratives were generated based on a single fanciful pic- ences (Liles, Duffy, Merritt, & Purcell, 1995; Newman &
ture. Scores on eight measures were calculated: TNW, NDW, McGregor, 2006). Differences may be due to the sensitiv-
number of T-units, mean length of T-units in words, num- ity of the measure of episodic structure. For example, the
ber of complex T-units, number of coordinating conjunctions number of complete episodes may not be as sensitive as
(COORD), number of subordinating conjunctions (SUBORD), measuring the presence or frequency of later occurring com-
and proportion of complex T-units. The scores were then ponents, such as motivating states or plans.
subjected to an exploratory factor analysis that revealed two Language skills can be evaluated through norm-referenced
major factors: linguistic productivity and complexity. These tests or narrative sample analyses. Narrative sample analyses
two summative indices were together termed the index of may be a more sensitive measure of subtle differences in
narrative microstructure (INMIS). The formulae and indices performance due to the natural context, which does not con-
reported at 1-year intervals were not applied to another strain linguistic productions. Narrative analyses provide
independent sample, but the INMIS appears to have potential information on microstructural features that reflect linguistic
as an efficient and clinically useful summative measure of productivity and complexity and on macrostructural organi-
narrative microstructure. zational features, such as story grammar. Evidence has been
Hoffman (2009b) tested whether the INMIS or an error found that both types of features distinguish children with
measure better differentiated children with and without LI. LI from their TD peers.
Twenty-four children between the ages of 8 and 10 years
with LI were matched for age, gender, and nonverbal cogni-
The Current Study
tive ability with 24 children with typical language. The
children narrated a wordless picture book. The narratives This study investigated the narrative language perfor-
were analyzed for INMIS productivity, INMIS complexity, mance of first graders who had been identified as being at
and the proportion of utterances with syntactic or semantic risk for reading difficulties through an RTI process. The
errors. The results indicated that there were no significant RTI model of reading instruction and assessment is intended
differences between the two groups for productivity, but to redirect experientially based learning trajectories and
there were significant differences for complexity and errors. identify children with persistent learning problems. RTI
The effect size for the error measure was much larger than that instruction in first grade is primarily code based, with little
for complexity, indicating the importance of evaluating attention to comprehension. Many young children who are
grammatical accuracy. The lack of sensitivity of INMIS identified with code-based difficulties will also present with
productivity did not correspond with the consistent finding language difficulties that are manifested concurrently or later
that children with LI produce less language than their TD in oral language and reading comprehension (Bishop &
peers (e.g., Fey et al., 2004; Newman & McGregor, 2006; Adams, 1990; Catts et al., 2002; Nation et al., 2004). The
Scott & Windsor, 2000). However, the elicitation technique primary purpose of this study was to determine whether
of narrating a set number of pages may have provided narrative language performance differentiates learner types
support for weaker narrators, reducing variation and group who have been identified through code-based RTI procedures.
differences in story length. In this study, first graders were identified as being at
Macrostructure analysis. In addition to sentence-internal risk for reading difficulties in January. The children’s oral
microstructural analyses, narratives can be analyzed for how reading fluency was reassessed in March, at which point the
sentences are organized and linked into a discourse unit. early resolvers were identified. Early resolvers were children
These are macrostructural analyses. A common macro- who had shown poor performance at entry to first grade
structural analysis is story grammar. Story grammar analysis but who resolved their difficulties before intervention. These
examines how agents resolve complications through a series children were considered typical learners who got off to a

208 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
poor start, possibly from poor test-taking behavior or mini- the norm-referenced test to potentially subtle differences in
mal experiential deficits. In April, the narrative language linguistic competence of the three learner types.
performance and general language skills of all children The purpose of this study was to examine the narrative
participating in the study were assessed before code-based language performance of first graders who had been iden-
Tier II intervention was initiated, which also occurred in tified as being at risk for reading difficulties. The specific
April. Once the intervention was completed, good respond- research questions were:
ers and poor responders were identified. & Do these three learner groups differ on their total,
Good responders and poor responders, who had more receptive, or expressive language as measured by a
persistent deficits than early resolvers, responded differently norm-referenced test of oral language?
to Tier II intervention. Good responders accelerated their & Do these three learner groups differ on individual
growth trajectories and met criterion growth expectations. measures of microstructure obtained during a narrative
These children were considered experientially deficient but retell?
good learners who were not expected to require further sup-
& Do these three learner groups differ on indices of
plementary reading instruction. Poor responders consisted
productivity and complexity obtained during a narrative
of the children who failed to meet criterion growth expecta-
retell?
tions following intervention. These children were expected
to have continued reading deficits and possibly inherent & Do these three learner groups differ on macrostructural
learning disabilities. measures of episodic complexity obtained during a
There were several possible outcomes for this study, de- narrative retell?
pending on the relationship between decoding and compre-
hension, the latter being essentially language. If decoding
and language are parallel in the early grades, so a child who
is low in decoding will be low in language, then children METHOD
identified by code-based RTI processes should also have low
language. As such, poor responders would show the weakest
Design
narrative language performance; early resolvers, who do A retrospective group comparison design was employed.
not require supplemental decoding attention, would show the The data were obtained in 2001–2002 as part of a larger
highest narrative language performance; and good respon- project (Allen, Nippold, & Simmons, 2011), with the current
ders would fall between these two groups with their initial narrative analyses designed and applied following data col-
deficits but a good capacity for learning given supplementary lection. All first graders in the district had been assessed
instruction. on reading at the beginning of the school year in September.
However, if decoding and language travel different paths, Study participation was solicited for children who were iden-
then any of the three groups may demonstrate language def- tified as being below expectations in January of first grade and
icits regardless of their performance in decoding and read- reading was reassessed in March. Participants were admin-
ing fluency (Bishop, McDonald, Bird, & Hayiou-Thomas, istered an assessment battery in April and then underwent
2009). If so, additional identification practices need to be 4 weeks of reading intervention and progress monitoring. Only
instituted along with code-based RTI. The outcome of this the measures relevant to the current study are reported here.
study matters most for early resolvers. If these children are
discovered to have low narrative language performance, then
Participants
they are at particular peril because early resolvers are con-
sidered capable learners who, despite an initial misstep, do The participants were selected from a suburban school
not need any attention beyond regular classroom instruction. district encompassing six elementary schools in the Pacific
We employed two measures of language in our study: Northwest. Five of the elementary schools qualified for
a norm-referenced test to measure children’s general lan- Title I services. In this school district, 43% of the children
guage skills and a narrative retell to investigate their narra- qualified for free and reduced-price meals. The majority of
tive language performance. The norm-referenced expressive the children were White (87.2%); of the remaining children,
score was only minimally expressive, with no tasks required 6.8% were Latino, 2.2% were Asian, 2.1% were African
beyond single sentence production. Narrative language American, and 1.7% was Native American.
performance was measured in terms of eight microstructural School district personnel completed early literacy bench-
measures, INMIS productivity and complexity indices, and mark testing for every child in the district three times across
two measures of episodic structure for a narrative retell. Use of the academic year using the Dynamic Indicators of Basic
this general language test allowed determination of whether Early Literacy Skills (DIBELS; Good & Kaminski, 1996)
a narrative retell taps language skills other than those obtained and the Test of Reading Fluency (TORF; Deno, Deno, Marston,
with conventional norm-referenced measures. It was ex- & Marston, 1987). District personnel assessed an original
pected that the narrative retell would be more sensitive than pool of 398 children in kindergarten, who were then followed

Allen et al.: Narratives of At-Risk Readers 209


into first grade. During January of their first-grade year, The pattern of performance on reading fluency at the
these children were assessed on reading fluency using the three testing points supported the learner group assignments
TORF. (Table 2): Early resolvers consistently outperformed good
Across the school district, 42 children (10.6%) read responders and poor responders, and by the end of the inter-
<20 correct words per minute (CWPM) on the TORF in vention, good responders outperformed poor responders.
January of their first-grade year. They were not on target to These results confirm that the three identified groups repre-
reach the end-of-year first-grade benchmark and were iden- sented different types of at-risk readers.
tified as being at risk for reading difficulties. Their par-
ticipation was sought for a set of investigations, including Measures
the current one. Thirty-four children returned consent forms
and received an assessment battery. Two children did not Oral reading fluency. The TORF is an individually ad-
complete the testing, resulting in a total of 32 children iden- ministered criterion-referenced reading measure of accuracy
tified as the participants who read <20 CWPM in January and rate that consists of equivalent grade-level passages. The
of first grade (M = 12.50, SD = 7.01). At the time of data child is shown a passage but is not told or shown any pictures
collection, none of the children was receiving special educa- about it. The child is instructed to read aloud each word. If
tion services for reading or instruction in English as a second a child hesitates or struggles with a word for 3 s, then the tester
language; three participants were receiving services for provides the target. Errors include substitutions, omissions,
speech. and hesitations longer than 3 s. The child is required to stop
Following testing in January, the 32 children continued to reading after 1 min. The total score is the number of CWPM.
receive regular reading instruction without any supplemen- Children read different passages at each testing point.
tary assistance (i.e., intervention). In March, these children Oral language. The Test of Language Development—
were retested with the TORF. The children who met the Primary (TOLD–P:3; Newcomer & Hammill, 1997) was
target performance of 30 CWPM were considered to be administered following standardized instructions. Six sub-
performing at a level equivalent to their nonidentified peers. tests examine semantic and syntactic production and com-
These 11 children were identified as early resolvers and prehension: Picture Vocabulary, Oral Vocabulary, Relational
did not receive supplementary reading intervention. The Vocabulary, Grammatical Understanding, Sentence Imita-
21 children who read <30 CWPM received supplementary tion, and Grammatical Completion. The Spoken Language
reading intervention for 4 weeks. (r = .96, content sampling average; r = .92, time sampling),
Oral reading fluency was assessed once a week using Listening (r = .91, content sampling average; r = .88, time
the TORF over the 4 weeks of intervention. Deno, Fuchs, sampling), and Speaking (r = .92, content sampling average;
Marston, and Shin (2001) identified an appropriate increase r = .82, time sampling) composite standard scores were
in words read correctly per minute for TD first graders as obtained to represent the children’s total, receptive, and
1.8 words per week (7.2 WPM increase over 4 weeks). Eight expressive language.
participants met this standard and were identified as good Narrative. A narrative language sample was elicited using
responders. Thirteen participants exhibited a TORF increase the Strong Narrative Assessment Procedure (SNAP; Strong,
score of <7.2 WPM growth and were identified as poor 1998). In this procedure, a child is presented with a book
responders. The three learner groups were similar for age, and is told to look at the book while he or she listens to the
gender, and ethnicity (Table 1).
Table 2. Mean and standard deviation for reading fluency
(correct words per minute on the Test of Reading Fluency
Table 1. Age, gender, and ethnicity of the study participants. [TORF; Deno, Deno, Marston, & Marston, 1987]) by time
and group.

Early Good Poor


resolvers responders responders Early Good Poor
Characteristic (n = 11) (n = 8) (n = 13) resolvers responders responders
M SD M SD M SD
Age
In months 84.2 85.8 84.7
In years;months 4;9 4;7 5;2 Januarya,b 19.00 6.57 12.25 2.43 7.15 4.20
Gender Pretesta,b 45.18 10.48 14.75 5.47 13.00 8.94
Male 6 6 8 Posttesta,b,c 53.73 11.43 32.56 11.99 15.15 7.85
Female 5 2 5
Ethnicity Note. Comparisons were significant at p < .05.
White 11 7 11 a
Latino 0 1 2 Early resolver and good responder; bEarly resolver and poor
responder; cGood responder and poor responder.

210 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
story. In our study, each child listened to an audiotaped story (Hunt 1965). However, from the instructions and examples,
and looked at the pictures from the wordless picture book, along with communication with the authors, it was apparent
A Boy, a Dog, and a Frog (Mayer, 1972). The child turned that C-units were actually used.
the page when he or she heard a specific sound on the record- The hand-coded measures were identified following
ing (i.e., a frog’s croak). An examiner sat with the child Justice et al. (2006). COORD consisted of any of seven coor-
during the story to aid in page turning. After listening, the dinating conjunctions (e.g., and, for, but, so, nor, or, yet) that
child retold the story without the book. To create a reason for were used to connect two clauses in a C-unit when a sub-
the retelling, the child was asked to retell the story into a tape ject was not repeated. For example, “but” was counted in
recorder so that the teacher could listen later (this is an The dog ran but tripped, but not in The dog ran/but the
alternative procedure that is allowed in the SNAP manual). cat tripped. The participants frequently used and, and then,
and then as cohesive tools at the beginning of a C-unit, so
these were not included in the COORD count. SUBORD
Procedure
consisted of any of the 26 subordinating conjunctions pro-
Measurement timeline. The pretest TORF, TOLD–P:3, vided by Justice et al. (see Appendix B in Justice et al., 2006,
and SNAP were administered in April of first grade as part for a complete list). COMPLEX was coded when a C-unit
of a larger assessment battery. Testing was completed in contained one or more dependent clauses. All C-units previ-
three, 45-min sessions in each child’s home school. The ously coded with a subordinating conjunction as well as all
assessment battery was administered to the children in the other C-units with subordinate clauses, including relative
same order. The TORF was also administered once per week clauses (e.g., The girl who has red hair is my friend), nomi-
during the 4-week intervention to the children who received nal clauses (e.g., The frog said, “I am sad.”), and infini-
treatment. The TORF was only administered at pretest and tive clauses (e.g., I want to go), were coded as complex
posttest for the early resolvers. utterances. PROPCOMPLEX was the numerical value for
Narrative analyses. The audiotaped narrative transcripts COMPLEX divided by LENGTH.
were transcribed, segmented, and analyzed at microstructural The INMIS formulae from Justice et al. (2006) were used
and macrostructural levels. to calculate the productivity and complexity of the narratives
Transcription and segmentation. The narrative language produced by each participant. The formulae used to assess
samples were transcribed and segmented by the third author the microstructure of the oral narratives were as follows:
according to guidelines of the Systematic Analysis of Lan-
Productivity ¼ 1:60 þ ð0:001  MLC-WÞ þ ð0:21
guage Transcripts (SALT; Miller & Chapman, 2004). The  PROPCOMPLEXÞ þ ð0:017  NDWÞ
narratives were segmented using communication units þ ð0:00054  TNWÞ þ ð0:014  COORDÞ
(C-units; Loban, 1963, 1976). A C-unit is “each independent þ ð0:0072  SUBORDÞ þ ð0:0094  LENGTHÞ
clause with its modifiers” (Loban, 1976, p. 9) or an elliptical þ ð0:068  COMPLEXÞ
response. Elliptical responses could be either an answer to
a question whether it only lacked part of the sentence ele- Complexity ¼ 2:84 þ ð0:27  MLC-WÞ þ ð0:85
ments in the question or each yes or no (Loban, 1976). For  PROPCOMPLEXÞ þ ð0:012  NDWÞ
the purposes of this study, each yes or no response that was in þ ð0:0027  TNWÞ þ ð0:028  COORDÞ
response to glossing by the examiner (E: You said the frog þ ð0:026  SUBORDÞ þ ð0:085  LENGTHÞ
þ ð0:14  COMPLEXÞ:
jumped? C: Yes) was omitted from the transcript. Any utter-
ance that did not contribute to the narrative was omitted to Macrostructure analysis. Episodic structure was analyzed
avoid an inflated word count (e.g., Are we done?). using guidelines from the SNAP (Strong, 1998). In the SNAP,
Microstructure analysis. Each transcribed language sam- a complete episode consists of a complication (e.g., He
ple was individually coded and was run through the SALT wanted a pet frog), an attempt to resolve the complication
program. The automated SALT analysis provided TNW, (e.g., He went to find a frog), and a consequence of the
NDW, number of C-units (LENGTH), and mean length of attempt (e.g., He tripped over a tree and fell into the water).
C-units in words (MLC-W). Four measures were hand coded To be considered a complete episode, the three elements had
by the third author: total number of coordinating conjunctions to be presented in that order. The source story consisted of
(COORD), total number of subordinating conjunctions five complete episodes. Each complete episode from the
(WUBORD), total number of complex C-units (COMPLEX), story was scored as 1 point. In addition, the number of epi-
and proportion of complex C-units (PROPCOMPLEX). sodic elements was scored. One point was assigned for each
The INMIS formulae from Justice et al. (2006) were used statement of setting, initiating event, internal response, at-
to further analyze the microstructure. It should be noted tempt, consequence, plan, or reaction.
that Justice et al. reported using minimal terminable units Narrative scoring reliability. A trained graduate student
(T-units) to segment utterances, resulting in a mean length of in speech-language pathology independently transcribed
T-units in words (MLT-W). A T-unit is similar to but more and coded one (33%) of the narratives. Word-to-word
restricted than a C-unit, excluding all nonpredicated utterances match transcription was 91%. Point-to-point agreement

Allen et al.: Narratives of At-Risk Readers 211


on identification of C-units was 93%, subordinating clauses Fidelity of intervention implementation. The interven-
was 94%, coordinating conjunctions was 89%, and complex tionists were provided training before conducting intervention.
utterances was 100%. Point-to-point agreement on identi- Fidelity observation forms (previously developed for other
fication of complete episodes was 94%. studies) were used to ensure that the instructors were cor-
Reading intervention. Two reading interventions were rectly implementing the programs. Each component was
provided. Children who read ≥10 CWPM were randomly scored as 2 for always implemented, 1 for sometimes imple-
assigned to either an accuracy or a fluency intervention group. mented, and 0 for never implemented. Each interventionist
The original purpose of these two groups ( but not part of was formally observed twice between the second and fourth
the current study) was to see if these different intervention weeks of intervention. On average, the accuracy intervention
focuses had differential effects on assessment measures of was implemented with 90% fidelity (range of 72%–100%),
accuracy and fluency. These two groups showed no signif- and the fluency intervention was implemented with 92%
icant difference in reading fluency at posttest ( p = .11). fidelity (range of 80%–100%).
Children who read <10 CWPM were only placed into the
accuracy intervention group. Good and poor responders were
present in both intervention groups. Although there were no
significant differences between the intervention groups in
RESULTS
regard to their reading score, the fluency group represented a This study examined the narrative language performance
greater proportion of the good responders than did the accu- of three types of at-risk first-grade readers. Analyses of
racy group (71% vs. 29%). Initial reading status may be variance (ANOVAs) with pairwise follow-up comparisons
important for entry into the good responder group because using the Tukey HSD procedure were conducted, with an
29% of the children who read <10 CWPM entered this group, overall alpha level of .05. Effect sizes were calculated as the
whereas 43% of the children who read ≥10 CWPM entered standardized mean difference (d ) for pairwise comparisons
this group. and eta-squared (h2) for multiple comparisons. Values >0.80
The reading accuracy intervention consisted of the Read- for d and >0.14 for h2 were considered large (Cohen, 1977).
ing Mastery program (Englemann & Bruner, 1998) or the
Proactive Beginning Reading program (a version based on
General Language Skills
Open Court; Mathes, Torgesen, Menchetti, Wahl, & Grek,
2001). If the school curriculum was Open Court, then the The first research question addressed whether there were
child received the Proactive Beginning Reading program. group differences among learner groups for total, receptive,
All other children received the Reading Mastery program. A and expressive language as measured by a norm-referenced
placement test was used for each child. Both reading pro- discrete skill test of oral language. Total mean score perfor-
grams follow a direct instruction approach (Carnine, Silbert, mance was almost 1 SD below the mean for all three groups
Kame’enui, & Tarver, 2004) with a focus on word identifi- for total and expressive language. Receptive language was
cation skills. Children were taught phonemic awareness close to the mean for all three groups (Table 3).
skills, letter-sound correspondences, sounding out words, A factorial ANOVA failed to show significant differences
and recognition of high-frequency irregular words. among the three groups for total, F(2, 29) = 0.06, p = .94;
The reading fluency intervention consisted of the Read
Naturally materials (Ihnot, 1999). Each child was placed into
graded materials (e.g., Level 0.8, 1.0). A child first read the Table 3. Mean standard score and standard deviation for
story without any practice and charted progress. The child general language skills by learner type.
then listened to a recording of the story twice while follow-
ing along with the text. Next, the child orally read the text
Early Good Poor
four times. Then the child completed a few comprehension resolvers responders responders
questions and read the story to a partner. The child read
the story a final time aloud to the interventionist and charted Measure M SD M SD M SD
this performance. The purpose of this intervention was to
improve rapid word identification skills in connected text. Total language 87.2 7.9 86.1 16.1 87.8 9.6
Reading intervention procedures. The reading interven- Receptive language 97.5 6.5 102.3 13.4 96.5 11.5
tions were provided in individual or paired 30-min lessons Expressive language 85.5 7.9 83.1 11.7 86.8 10.4
daily in addition to regular classroom reading instruction. All
children attended at least 80% of the sessions. The children Note. Total language = Spoken Language Quotient of the TOLD–
received either small-group or one-to-one instruction pro- P:3; Receptive language = Listening Quotient of the Test of
Language Development—Primary (TOLD–P:3; Newcomer &
vided by two district educational assistants, two research Hammill, 1997); Expressive language = Speaking Quotient of the
assistants, or one graduate student during the regular school TOLD–P:3. Standard scores based on Mtest of 100 and SD = 15.
day. All instruction took place in the children’s home schools. No significant differences at p < .05.

212 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
receptive, F(2, 29) = 0.76, p = .48; and expressive language, across the eight measures. Four measures showed signifi-
F(2, 29) = 0.35, p = .71. The standard deviations were ob- cant differences: TNW, F(2, 29) = 7.17, p = .003, h2 = 0.33;
served to be noticeably larger for the good responders, so NDW, F(2, 29) = 5.06, p = .013, h2 = 0.26; LENGTH,
Levene’s Test of Equality of Error Variances was calculated. F(2, 29) = 6.81, p = .004, h2 = 0.32; and COORD, F(2, 29) =
Results showed that there were no significant differences 5.65, p = .008, h2 = 0.28. No significant differences were
in variance between the groups (total p = .27; receptive obtained for the other four measures: MLC-W, F(2, 29) =
p = .09; expressive p = .50). 0.07, p = .930, h2 = 0.01; SUBORD: F(2, 29) = 0.89, p = .422,
An additional post hoc test was conducted on the observed h2 = 0.06; COMPLEX: F(2, 29) = 1.15, p = .331, h2 = 0.07;
difference between the receptive and expressive language and PROPCOMPLEX: F(2, 29) = 2.06, p = .146, h2 = 0.12.
means. A paired-samples t test indicated that the scores for Follow-up pairwise comparisons were conducted on the
receptive language were significantly greater than the scores four significant measures (Table 5). Good responders per-
for expressive language for all three learner groups: early formed significantly better than early resolvers and poor
resolvers, t(20) = 3.89, p = .0009; good responders, t(14) = responders for TNW, NDW, and LENGTH. Good respond-
3.05, p = .009; and poor responders, t(24) = 2.26, p = .03. ers performed significantly better than early resolvers for
These results indicated that the three learner types did not COORD. All of these comparisons showed large effect sizes.
differ significantly on norm-referenced language test scores, The results indicated that good responders had relatively
but all three learner types showed significantly lower expres- strong narrative microstructure compared to the similarly
sive than receptive results. low performance of early resolvers and poor responders.

Narrative Microstructure Productivity and Complexity Indices


The next research question addressed whether there were The third question addressed whether there were group
group differences in narrative language performance for differences on two summative measures of microstructure,
eight microstructural measures (Table 4). A multivariate productivity and complexity. A factorial ANOVA on pro-
analysis of variance (MANOVA) was conducted across the ductivity showed significant differences, F(2, 29) = 4.43,
eight measures and three groups. Wilks L of .36 was signifi- p = .02, h2 = 0.23 (Table 4). Follow-up pairwise comparisons
cant, F(16, 44) = 1.86, p = .05, h2 = 0.40. A conservative revealed the differences to lie between good responders
alpha level of .01 was set to control for family-wise error and early resolvers: Mdifference = 0.60, SE = 0.21, p = .022,
Cohen’s d = 1.29. The differences between good responders
Table 4. Mean and standard deviation for microstructural and poor responders narrowly failed to reach significance:
measures by learner type. Mdifference = 0.50, SE = 0.20, p = .06, Cohen’s d = 1.11. There

Table 5. Inferential results and effect sizes for the significant


Early Good Poor comparisons of the microstructural measures.
resolvers responders responders
Measure M SD M SD M SD
Mean
a,b
Measure Comparison Difference SE p Cohen’s d
TNW 60.0 32.3 137.9 62.3 77.2 43.8
NDWa,b 34.6 15.8 57.1 17.4 39.1 14.9
LENGTHa,b 9.0 4.7 20.3 8.8 11.5 6.8 TNW Early–Good* 77.88 24.06 .003 1.75
MLC-W 6.8 2.1 7.1 1.0 6.8 1.2 Good–Poor* 60.72 25.15 .016 1.24
COORDa 0.1 0.3 1.00 0.8 0.4 0.7 Early–Poor 17.15 15.57 .635 0.46
SUBORD 1.3 1.2 1.4 1.4 0.8 0.9 NDW Early–Good* 22.49 7.36 .013 1.11
COMPLEX 3.1 2.3 4.8 3.7 3.2 2.0 Good–Poor* 18.05 7.12 .043 1.17
PROPCOMPLEX 0.3 0.2 0.2 0.1 0.3 0.2 Early–Poor 4.44 6.49 .774 0.31
Productivity indexa –0.82 0.42 –0.22 0.58 –0.72 0.40 LENGTH Early–Good* 11.25 3.14 .003 1.78
Complexity index –0.75 0.79 –1.43 0.46 –0.98 0.81 Good–Poor* 8.71 3.04 .020 1.22
Early–Poor 2.54 2.77 .635 0.44
COORD Early–Good* 0.91 0.28 .006 1.69
Note. TNW = total number of words; NDW = number of different
Good–Poor 0.62 0.32 .067 0.85
words; LENGTH = number of C-units; MLC-W = mean length of
Early–Poor 0.29 0.20 .450 0.56
utterance in words; COORD = number of coordinating conjunctions;
SUBORD = number of subordinating conjunctions; COMPLEX =
number of C-units containing at least one dependent clause; Note. Only comparisons that were significant on a prior factorial
PROPCOMPLEX = COMPLEX/LENGTH. analysis of variance are listed. Good = good responders, Poor =
a
Early resolver and good responder comparison significant at poor responders, early = early resolvers. Cohen’s d was calculated
p < .05; bGood responder and poor responder comparison signifi- with means and standard deviations.
cant at p < .05. *significant at p = .05.

Allen et al.: Narratives of At-Risk Readers 213


were no significant differences between the early resolvers test, differed for three types of first graders who had been
and the poor responders: Mdifference = 0.10, SE = 0.19, identified as being at risk for reading difficulties. General
p = .847, Cohen’s d = 0.26. language skills were evaluated using composite scores on
A factorial ANOVA on complexity showed no signifi- total language, receptive language, and expressive language.
cant differences: F(2, 29) = 2.00, p = .154, h2 = 0.12. No There were no significant differences between the three
follow-up pairwise comparisons were conducted. Good re- types of at-risk readers for any of the measures of general
sponders did not produce more complex language than the language skills. In regard to overall language, all three groups
other two learner types, despite the finding that they produced were performing almost 1 SD below the mean. Interestingly,
more narrative language than the other two learner types. the receptive language of all three groups was near the
mean, whereas their expressive language performance was
Narrative Macrostructure significantly lower than the mean. A review of the individ-
ual expressive test scores indicated that 50% of the good
The final question required an examination of group responders (n = 4) performed more than 1 SD below the
differences for the macrostructural measures of number of mean, as did 31% of the poor responders (n = 4) and 55% of
complete episodes and number of episodic elements. The the early resolvers (n = 6). The expressive subtests on the
mean number of episodes was highest but most variable TOLD–P:3 were limited to vocabulary and sentence structure
for the good responders (Table 6). A factorial ANOVA for tasks with limited production demands, but this was enough
episode showed no significant differences among the groups, to tap a different set of skills from the receptive subtests.
F(2, 29) = 2.29, p = .120, h2 = 0.14. The expressive subtests showed that all three learner types
A factorial ANOVA for total number of episodic elements were weak in expressive language. However, the norm-
showed significant differences among the groups: F(2, 29) = referenced measure failed to reveal differences among the
4.53, p = .019, h2 = 0.24. The good responders provided three groups.
significantly more episodic elements than the early resolvers:
Mdifference = 9.23, SE = 3.13, p = .02, Cohen’s d = 1.53. There
Narrative Language Performance for Three
were no significant differences between the good respond-
ers and the poor responders: Mdifference = 6.88, SE = 3.03,
Types of At-Risk Readers
p = .076, Cohen’s d = 0.95; nor between the poor responders The second goal of this study was to determine whether
and the early resolvers: Mdifference = 2.34, SE = 2.76, p = .677, narrative language performance differed among three types
Cohen’s d = 0.40. These results indicate that the good re- of first graders who had been identified as being at risk for
sponders provided more story grammar elements in their reading difficulties through an RTI process. Narrative
narrative retellings than did the early resolvers. retells were evaluated at microstructural and macrostruc-
tural levels.
Microstructure. Eight microstructural measures were
evaluated across the three types of at-risk readers. Good
DISCUSSION
responders performed significantly better than the other two
General Language Skills for Three Types groups on measures related to amount of talk: TNW, NDW, and
LENGTH. Good responders performed significantly better
of At-Risk Readers
than early resolvers for COORD and the overall productivity
The first goal of this study was to determine whether index, which is a variably weighted combination of eight
general language skills, as measured by a norm-referenced microstructural measures. Effect sizes for these differences
were all large (d values of 1.11 to 1.78).
Table 6. Mean and standard deviation for episodic structure The microstructural measures of complexity generally
by group. did not show significant differences among the three learner
types. Good responders were significantly stronger on one
measure of complexity, NDW, which is primarily a semantic
Early Good Poor measure. Good responders provided, on average, È57 differ-
resolvers responders responders ent words of the 150 possible; the other two groups pro-
M SD M SD M SD vided <40 words each.
The other contributors to complexity were primarily syn-
tactic (MLC-W, COORD, SUBORD, COMPLEX, and
Number of complete 0.18 0.40 1.00 1.60 0.23 0.60
episodes PROPCOMPLEX). These, along with the complexity index,
Number of episodic 9.27 4.56 18.50 8.32 11.62 7.23 did not show even a consistent trend toward superiority for
elementsa any of the groups. Syntax is the most consistent marker of
specific LI (Leonard, 1998) but did not differentiate among
a
Early resolver and good responder comparison significant at p < .05. these at-risk groups.

214 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
For COORD, or clausal coordination, all three groups, on difference between the episodic elements of good respond-
average, produced fewer than one clausal coordinator. The ers and early resolvers was significant, but the overall
source story only provided five examples of clausal coordi- pattern suggested a relative strength for the good responders.
nation using and within a C-unit. For the most part, the The episodic structure analysis scored only episodes that
children did not produce a modified utterance but simply were retold from the source story, not from any original epi-
omitted the compound sentence from their retell. The lack of sodes. Examination of the narratives showed that the chil-
utterances containing and in the participant retells contrasts dren produced only modeled episodes. Good responders
with its early developmental appearance (Owens, 2008; were also distinguished on how many participants attempted
Retherford, 2000). However, this may be due to the clausal all five episodes modeled (5 of 8 children), compared to the
coordinator scoring system. C-unit segmentation requires other two groups (early resolvers = 0 of 11 children; poor
an utterance division at the clausal coordinator if a second responders = 3 of 13 children).
subject is present (Loban, 1976), resulting in an utterance- Most children attempted at least three episodes but often
initial and for the second utterance. This use of and was failed to complete the episodes. When participants addressed
considered a discourse component, as a continuer and marker an episode, they tended to omit the complication, initiating
of the main story line (Peterson & McCabe, 1991), not an their retell with an attempt to solve the complication (e.g.,
intrasentential connector. Additionally, the COORD variable so he followed the footprints) followed by a consequence.
was scored for the coordination of verbs (e.g., running and These results are similar to those for third- and fifth-grade
jumping) but not nouns (e.g., frog and friend ) that served children who retold stories that included more attempts and
to conjoin clauses within a C-unit. These restrictions lowered outcomes than complications (Goldman & Varnhagen, 1986)
the number of coordinating uses of and that were scored. and for 6-, 7-, and 8-year-olds who also tended to omit com-
But was also used to initiate a C-unit but rarely appeared plications in generated narratives (Peterson & McCabe, 1983).
despite its presence three times in the source story. In sum, in our study, microstructural narrative produc-
Other than COORD, the syntactic measures are different tivity measures and the macrostructural measure of num-
but overlapping aspects of subordination. The source story ber of episodic elements were shown to distinguish among
used 22 tokens (only 1 token per C-unit could be counted, three types of at-risk learners. Where significant differences
resulting in 20 as the maximum for COMPLEX) that contrib- were found, good responders consistently outperformed the
uted to COMPLEX (6 subordinating conjunctions, 2 nominal other two learner types.
dependent clauses, and 14 infinitive clauses). The children
reproduced few of these subordinating structures, with mean Language Performance Compared
performance across the three groups at È3.5 tokens. Infini- to Typical Learners
tive clauses, which are early appearing structures (Owens,
2008; Retherford, 2000) and are not counted as measures of The research questions in this study concerned patterns of
complexity by some authors (e.g., Hughes, McGillivray, & narrative language performance across RTI-identified learner
Schmidek, 1997; Justice & Ezell, 2002), were rarely used types. However, the use of a norm-referenced test and a
by the participants. Across the three groups, the early resolv- standardized narrative sampling procedure that has age com-
ers provided a mean of only 1.6 (SD = 1.6) infinitives, the parison data allowed for consideration of how the language
good responders provided a mean of 3.6 (SD = 2.7) infinitives, of these at-risk learners compared to that of typical learners.
and the poor responders provided a mean of 2.5 (SD = 1.8). The norm-referenced TOLD–P:3 showed marginal over-
It should be noted that the INMIS indices may be of all language ability for all three learner groups in regard to
unexpected magnitudes, and care should be taken in inter- mean standard score performance of 87.03. Sixty-three per-
preting them. The good responders exemplified this: Their cent of the children (n = 20) showed performance in the
complexity score was the lowest of the three groups, al- lower quartile (standard score of ≤90) that RTI is intended
though their performance on the eight contributing measures to flag. Thirty-one percent (n = 10) scored <85, or less than
was as high or higher than those of the other two groups. the 16th percentile, which warrants referral for a special
Their lower complexity score was due to the good respond- education evaluation.
ers having produced more narrative language than the other The narrative analysis showed similar low performance for
two groups, but with similar complexity. It appears that all three learner types. On average, the three groups of children
the expectation of the INMIS complexity index is that there were missing most of the story in their retells. There were
should be more evidence of complex narrative language almost 400 words in the source story (Appendix). TNW was
within the longer narratives. 137.9 for the good responders, 60.0 for the early resolvers, and
Macrostructure. The narratives were also evaluated for 77.2 for the poor responders. This was still only approximately
macrostructural episodic organization. Good responders one-third of the source story vocabulary. Appendix J1 of
provided a mean of 1.0 complete episodes and 18 episodic
elements, compared to <0.25 complete episodes and 1
Two versions of SNAP were published in 1998. One version contains
<12 episodic elements for the other two groups. Only the 198 pages and the other contains 206 pages with Appendix J.

Allen et al.: Narratives of At-Risk Readers 215


the SNAP manual presents field-test data for 7-year-olds, for demands of producing and organizing expressive discourse.
which a similar elicitation procedure, including the same As a result, it was not surprising that it did not show the same
storybook, was used. For 13 TD children, the SNAP reported pattern across learner types as the narrative retell results.
a TNW of 166.7 words (SD = 70.1), and for 13 children with For narrative language performance, good responders,
LI, a TNW of 82.8 words (SD = 31.9). Although the good who were children with initially low reading scores who met
responders’ performance appropriately fell between that of grade-level expectations for growth, stood out above the
the two SNAP groups, the performance of the poor respond- other two groups. Compared to the other two learner types,
ers and early resolvers was like that of the children with LI. good responders recalled and produced more of the stories
NDW was not reported for the SNAP data, so that comparison they heard, showed greater semantic diversity, and presented
could not be made. better episodic structure. Verbal output and vocabulary have
For MLC-W, the three learner types produced utterance consistently differentiated language and reading disabili-
lengths of approximately seven words. This was slightly ties (e.g., Fey et al., 2004; Scott & Windsor, 2000), as has
below the level of both children with typical and language episodic structure (Merritt & Liles, 1987; Ripich & Griffith,
impaired performance in the SNAP field-test data. Interest- 1988; Roth & Spekman, 1986).
ingly, considering the emphasis in language sampling on Good responders did not differ significantly on the com-
average sentence length (whether as MLU, MLC-W, or plexity measures, but the other areas of relative strength
MLT-W), the SNAP showed essentially no difference be- suggest that good responders had some of the linguistic abili-
tween typical development and LI (M = 7.6, SD = 0.9; ties needed for success in reading, especially as emphasis
M = 7.4, SD = 1.5). No comparison was made for the other switches to comprehension. These findings are consistent
complexity measures because of differences in how the with an RTI model that predicts that such children were only
scoring was conducted. experientially deficient rather than having an inherent learn-
Performance of these participants on episodic elements ing disability. However, the comparisons with the data
can again be compared to the SNAP field-test data. For reported in Strong (1998), and the wide variability on the
children with TD language, the data showed a mean of 2.1 norm-referenced test, suggest that even the language skills
complete episodes (SD = 1.3); for children with LI, the data of the good responders still warrant further monitoring and
showed a mean of 0.6 complete episodes (SD = 0.8). The possibly supplementary support.
episodic structure of the narratives from the study participants, Poor responders showed a clear language profile of con-
particularly the early resolvers and the poor responders, cern. However, they are likely to eventually receive the
resemble the episodic structure of the narratives from the attention they need. Not only did they perform inadequately
group with LI. at the end of intervention, but they also performed inade-
This exploration of comparison data, obtained in a similar quately at the outset of intervention, as more poor responders
manner but outside this study, showed weaknesses in lan- than good responders scored in the lowest level of read-
guage for children who had been identified using code-based ing fluency. Five of the seven children (71%) who read
measures. Despite the relative strengths in productivity <10 CWPM became poor responders, whereas eight of the
and vocabulary, even the good responders reproduced only 14 children (57%) who read ≥10 CWPM became poor
one-third of the source story. However, performance of the responders. These persistently low performers either will be
other two learner types closely resembled that of children flagged for further prereferral Tier III intervention and then
with LI, which puts these learners, who have not been a referral for special education or may be referred directly
identified as language impaired, at risk for linguistic difficul- for evaluation. Either of these routes is likely to lead to a
ties in addition to decoding and reading fluency. comprehensive examination of language as well as decoding
performance.
Subtypes of At-Risk Readers: More Than The results did not support the hypothesis that the group
of children who did not require reading intervention, early
Just Decoding
resolvers, would perform significantly better than the other
First graders who did not meet grade-level reading fluency two groups on measures of narrative skills. Despite their
benchmarks were grouped into three learner types. The learning strengths—they resolved their initial reading fluency
findings suggest that, in addition to decoding differences, difficulties with only regular classroom instruction—the
these groups have distinctive narrative language patterns. early resolvers’ linguistic performance was significantly
These patterns provide an important additional dimension below that of the good responders, with large effect sizes
to the performance of young at-risk readers. for the differences. These results are similar to those found
General language skills did not differ among the learner by Bishop et al. (2009) in that there are samples of chil-
types. In addition, receptive and expressive norm-referenced dren with LI who present with accurate word reading skills.
scores showed similar performance. The expressive compo- Early resolvers were similar in language performance to
site, based on tasks with only word and sentence construc- the group of clear concern, poor responders, but the early
tion or imitation requirements, does not really reflect the resolvers were in an even more problematic situation. The

216 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
group’s on-track status for code-based skills overshadows differences between TD children and children with LI or
their current linguistic needs. The risk for future difficul- reading impairment, it provides relatively low power for a
ties appears high for this group due to the absence of a three-way comparison among more subtly different groups
“language-track” RTI model (Ukrainetz, 2006a). Their of children. This would have been difficult to remedy; an
weaknesses will not be identified until reading and academic entire school district was needed to locate this number of
performances are affected sufficiently for a special education children. Only 10.6% of the children in the district qualified
referral. The nonconsideration of linguistic skills puts this as being at risk for reading difficulties and moved into
group of children into a “wait to fail” model. Tier II instruction. A large-scale multiyear evaluation of
The central implication of this study is that RTI models schoolwide universal screening showed 15% of children
should consider language in addition to decoding. A sub- advancing to Tier II instruction (VanDerHeyden, Witt, &
group of at-risk readers, the early resolvers, met the bench- Gilbertson, 2007). The lower percentage of Tier II chil-
mark criterion for accurately and fluently decoding connected dren in this district contrasts with a report suggesting the
text. If narrative performance had been used to determine opposite: that children are being overidentified with the
responsiveness, then these children may have continued to high expectations of the current DIBELS (Catts, Petscher,
be identified as at risk and may have been provided supple- Schatschneider, Bridges, & Mendoza, 2008). The reasons
mentary support. These findings suggest that a sensitive for the lower number identified may be due to the strength
measure of language performance should be added to evalu- of the core curriculum provided within the school district,
ations of early risk for reading difficulties. but no matter the reason, this did result in an undesirably
small sample.
The INMIS scores obtained in this study were not com-
Study Limitations and Future Directions
pared to those of Justice et al. (2006) because of the differ-
There were several limitations of this study. A major ences in the narrative elicitation procedures. Age-referenced
limitation was in the use of extant data, which limited the language sample information is often used without refer-
study design choices. A comparison group of typical readers ence to elicitation context (e.g., SALT reports separately for
from the same district would have been more informative conversation and narrative but collapses across data col-
and procedurally robust than the norm-referenced scores and lected using diverse procedures). However, elicitation proce-
SNAP age reference information. dures do have an impact on narrative language. Merritt and
Language performance following intervention would Liles (1987) found that children with and without language
have been very helpful in understanding the implications of disorders produced narratives of a similar length with story
the narrative language patterns obtained and the effects of generation tasks but were differentiated by length for a story
code-based intervention on language performance. It is pos- retell task. In regard to macrostructure, Schneider and Dube
sible that the therapeutic attention that the children received (2005) found that children produced more story grammar
during the intervention, even though it did not explicitly units for a story retell task than a story generation task sup-
target narrative skills, may have indirectly and positively ported by pictures. Thus, although the INMIS was helpful
impacted their language skills. The effects of active controls in making within-study comparisons, it could not be used
in recent studies suggest the power of repeated opportunities across different elicitation procedures. To better understand
for motivated and attentive language interactions (Gillam and validate the INMIS as a clinical tool, studies should
et al., 2008; Ukrainetz, Ross, & Harm, 2009). Another con- use the same narrative elicitation contexts or ones that sys-
sideration is that the type of intervention provided may have tematically vary from the original.
a differential impact on narrative language performance. Finally, investigators need to be clear on the segmentation
If a second narrative retell had been elicited after the inter- unit and rules used. T-units and C-units were developed more
vention was completed, then one would have been able to than 40 years ago by Hunt (1965) for analyzing writing
investigate the impact of a fluency-based intervention that and by Loban (1963) for analyzing spoken language. T-units
uses a complete discourse unit versus an accuracy-based inter- are limited to grammatically complete sentences, whereas
vention that mostly uses single word and sentence level units. C-units additionally include ellipses and fragments. To ana-
The use of a norm-referenced measure of narrative abil- lyze by T-units, many utterances (e.g., A boy, a dog, and a
ity such as the TNL may have been more informative in frog; Yeah; Me too; Maybe she was; The end ) must be omit-
regard to the performance of the children in comparison to ted from the transcript. Confusion between segmentation
a peer group. The TNL requires full narrative productions units will have a significant effect on conversation, with its
from children, so it might have revealed these subtle group heavy use of elliptical utterances, with less effect on mono-
differences more than the TOLD–P:3, but that possibility logic narration (Allen, Ukrainetz, & Petersen, 2010). How-
remains speculative. At the time of the data collection, the ever, even for narrative studies, if inconsistent application
TNL had not been published. occurs, such as if the SALT program is used to automatically
A significant concern involved the sample size. Although count all of the transcribed utterances (i.e., C-units) for “total
a sample of 32 participants may be sufficient to detect number of utterances” and hand coding is used to count

Allen et al.: Narratives of At-Risk Readers 217


the proportion of complex to grammatically complete utter- and T-units. Poster presented at the annual conference of the
ances (i.e., T-units), internally inconsistent results will be American Speech-Language-Hearing Association, Philadelphia, PA.
obtained. American Speech-Language-Hearing Association. (2001).
Roles and responsibilities of speech-language pathologists with
respect to reading and writing in children and adolescents.
Conclusion Rockville, MD: Author.
This study investigated the oral language performance Badian, N., McAnulty, G., Duffy, G., & Als, H. S. (1990).
of first graders who had been identified as being at risk Prediction of dyslexia in kindergarten boys. Annals of Dyslexia,
for reading difficulties through an RTI process. A norm- 40, 152–167.
referenced language test did not reveal differences across Bishop, D. V., & Adams, C. (1990). A prospective study of the
learner types for general language skills. However, narrative relationship between specific language impairment, phonological
analyses were sensitive to differences: Children who re- disorders and reading retardation. Journal of Child Psychology
sponded well to intervention showed stronger narrative and Psychiatry, 31, 1027–1050.
language performance than children who did not respond Bishop, D. V., McDonald, D., Bird, S., & Hayiou-Thomas,
adequately to intervention or children who resolved without M. E. (2009). Children who read words accurately despite lan-
intervention. Good responders retold more of the source guage impairment: Who are they and how do they do it? Child
story, with greater semantic diversity and more episodic Development, 80, 593–605.
elements. There were no significant differences among the Carnine, D. W., Silbert, J., Kame’enui, E. J., & Tarver, S. G.
groups for syntactic complexity. Norm-referenced expres- (2004). Direct instruction reading (4th ed.). Upper Saddle River,
sive language performance and comparison to field-test NJ: Merrill.
data from Strong (1998) suggested that the three groups, but Catts, H. W., Fey, M., Tomblin, J. B., & Zhang, X. (1999).
most especially the early resolvers and poor responders, Language basis of reading and reading disabilities: Evidence
performed similar to children with LI. from a longitudinal investigation. Journal of Scientific Study
In conclusion, the results of this preliminary retrospective of Reading, 3, 331–361.
small sample study indicate that children who have been Catts, H. W., Fey, M. E., Tomblin, J. B., & Zhang, X. (2002).
identified as different kinds of learners through an RTI pro- A longitudinal investigation of reading outcomes in children
cess have some significant differences in narrative language with language impairment. Journal of Speech, Language, and
Hearing Research, 45, 1142–1157.
performance. Children of greatest concern are those who,
despite initial difficulties, meet growth expectations in their Catts, H. W., Fey, M. E., Zhang, X., & Tomblin, J. B. (2001).
reading fluency without intervention. These children may Estimating the risk of future reading difficulties in kindergarten
have language deficits that will not be flagged until they are children: A research-based model and its clinical implications.
Language, Speech, and Hearing Services in Schools, 32, 38–50.
in serious educational distress. Evaluation of narrative lan-
guage performance, in particular analyses that capture produc- Catts, H. W., Petscher, Y., Schatschneider, C., Bridges, M., &
tivity and episodic elements, should be added to RTI models. Mendoza, K. (2008). Floor effects associated with universal
screening and their impact on the early identification of reading
disabilities. Journal of Learning Disabilities, 42, 163–176.
Cohen, J. (1977). Statistical power analysis for the behavioral
ACKNOWLEDGMENT sciences (rev. ed.). New York, NY: Academic Press.
Thanks to Amy Peterson for assisting with reliability. Cunningham, A. E., & Stanovich, K. E. (1998). What reading
does for the mind. American Educator, 22, 8–15.
Deno, S. L., Deno, D., Marston, D., & Marston, D. (1987).
Test of Reading Fluency: Measures for screening and progress
REFERENCES monitoring. Minneapolis, MN: Children’s Educational Services.
Adams, M. J. (1990). Beginning to read: Thinking and learning Deno, S. L., Fuchs, L. S., Marston, D., & Shin, J. (2001). Using
about print. Cambridge, MA: MIT Press. curriculum-based measurement to establish growth standards for
children with learning disabilities. School Psychology Review,
Al Otaiba, S., & Fuchs, D. (2006). Who are the young children
30, 507–524.
for whom best practices in reading are ineffective? An experi-
mental and longitudinal study. Journal of Learning Disabilities, Denton, C. A., & Mathes, P. G. (2003). Intervention for strug-
39, 414–431. gling readers: Possibilities for change. In B. R. Foorman (Ed.),
Allen, M. M., Nippold, M. A., & Simmons, D. C. (2011). A Preventing and remediating reading difficulties: Bringing
comparison of assessment tools and methodologies for identify- science to scale (pp. 229–251). Timonium, MD: York Press.
ing children’s responsiveness to early reading intervention. Englemann, S., & Bruner, E. C. (1998). Reading mastery I:
Manuscript submitted for publication. Distar reading. Chicago, IL: Science Research Associates.
Allen, M. M., Ukrainetz, T. A., & Petersen, D. B. (2010, Feagans, L., & Short, E. J. (1984). Developmental differences in
November). An examination of the differences between C-units the comprehension and production of narratives by reading-disabled

218 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
and normally achieving children. Child Development, 55, Ihnot, C. (1999). Read naturally. Saint Paul, MN: Read Naturally.
1727–1736. Individuals with Disabilities Education Improvement Act of 2004,
Fey, M. E., Catts, H. W., Proctor-Williams, K., Tomblin, J. B., Pub. L. No. 108-446, 118 Stat. 2647 (2005).
& Zhang, X. (2004). Oral and written story composition skills Justice, L. M., Bowles, R. P., Kaderavek, J. N., Ukrainetz,
of children with language impairment. Journal of Speech, T. A., Eisenberg, S. L., & Gillam, R. B. (2006). The index
Language, and Hearing Research, 47, 1301–1318. of narrative microstructure: A clinical tool for analyzing school-
Fuchs, D., & Fuchs, L. S. (2006). Introduction to response to age children’s narrative performance. American Journal of
intervention: What, why and how valid is it? New Directions Speech-Language Pathology, 15, 177–191.
in Research, 41(1), 93–99. Justice, L. M., & Ezell, H. K. (2002). The syntax handbook:
Fuchs, D., Fuchs, L. S., & Compton, D. L. (2004). Identifying Everything you learned about syntaxI but forgot. Austin, TX:
reading disability by responsiveness-to-instruction: Specifying Pro-Ed.
measures and criteria. Learning Disability Quarterly, 27, Kamhi, A. G. (2009). The case for the narrow view of reading.
216–227. Language, Speech, and Hearing Services in Schools, 40, 174–177.
Fuchs, L. S. (2003). Assessing intervention responsiveness: Con- Leonard, L. B. (1998). Children with specific language impair-
ceptual and technical issues. Learning Disabilities Research and ment. Cambridge, MA: MIT Press.
Practice, 18, 172–186.
Liles, B. Z. (1985). Cohesion in the narratives of normal and
Gersten, R., Compton, D., Connor, C. M., Dimino, J., Santoro, language-disordered children. Journal of Speech and Hearing
L., Linan-Thompson, S., & Tilly, W. D. (2008). Assisting Research, 28, 123–133.
students struggling with reading: Response to intervention and
Liles, B. Z., Duffy, R. J., Merritt, D. D., & Purcell, S. L.
multi-tier intervention for reading in the primary grades. A
(1995). Measurement of narrative discourse ability in children
practice guide (NCEE 2009-4045). Washington, DC: National
with language disorders. Journal of Speech and Hearing
Center for Education Evaluation and Regional Assistance.
Research, 38, 415–425.
Retrieved from http://ied.ed.gov/ncee/wwc/publications/
practiceguides/. Loban, W. (1963). The language of elementary school children.
Champaign, IL: National Council of Teachers of English.
Gillam, R. B., & Johnston, J. R. (1992). Spoken and written
language relationships in language/learning-impaired and nor- Loban, W. (1976). Language development: Kindergarten through
mally achieving school-age children. Journal of Speech and grade twelve. Urbana, IL: National Council of Teachers of
Hearing Research, 35, 1303–1315. English.
Gillam, R. B., Loeb, D. F., Hoffman, L. M., Bohbman, T., Mandler, J. M., & Johnson, N. S. (1977). Remembrance of
Champlin, C. A., Thibodeau, L., . . . Friel-Patti, S. (2008). things parsed: Story structure and recall. Cognitive Psychology,
The efficacy of Fast ForWord language intervention in school- 9, 111–151.
age children with language impairment: A randomized controlled Marston, D. (2005). Tiers of intervention in responsiveness to
trial. Journal of Speech, Language, and Hearing Research, 51, intervention: Prevention outcomes and learning disabilities iden-
97–119. tification patterns. Journal of Learning Disabilities, 38, 539–544.
Gillam, R. B., & Pearson, N. (2004). Test of Narrative Language. Mathes, P. G., Torgesen, J. K., Menchetti, J., Wahl, M., &
Austin, TX: Pro-Ed. Grek, M. (2001). Proactive beginning reading. Unpublished
Goldman, S. R., & Varnhagen, C. K. (1986). Memory for document.
embedded and sequential story structures. Journal of Memory Mayer, M. (1972). A boy, a dog, and a frog. New York, NY: Dial
and Language, 25, 401–418. Books.
Good, R. H., & Kaminski, R. A. (Eds.). (1996). Dynamic Mellard, D. F., McKnight, M., & Woods, K. (2009). Response to
Indicators of Basic Early Literacy Skills (4th ed.). Eugene, OR: intervention screening and progress-monitoring practices in 41 local
Institute for the Development of Educational Achievement. schools. Learning Disabilities Research & Practice, 24, 186–195.
Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, and Merritt, D. D., & Liles, B. Z. (1987). Story grammar ability in
reading disability. Remedial and Special Education, 7, 6–10. children with and without language disorder: Story generation,
Hoffman, L. M. (2009a). Narrative language intervention inten- story retelling, and story comprehension. Journal of Speech and
sity and dosage: Telling the whole story. Topics in Language Hearing Research, 30, 539–552.
Disorders, 29, 329–343. Miller, J. F., & Chapman, R. S. (2004). Systematic Analysis
Hoffman, L. M. (2009b). The utility of school-age narrative of Language Transcripts (Version 8) [Computer software].
microstructure indices: INMIS and the proportion of restricted Madison: University of Wisconsin–Madison, Waisman Center,
utterances. Language, Speech, and Hearing Services in Schools, Language Analysis Laboratory.
40, 365–375. Nation, K., Clarke, P., Marshall, C., & Durand, M. (2004).
Hughes, D., McGillivray, L., & Schmidek, M. (1997). Guide to Hidden language impairments in children: Parallels between poor
narrative language: Procedures for assessment. Eau Claire, WI: reading comprehension and specific language impairment? Journal
Thinking Publications. of Speech, Language, and Hearing Research, 47, 199–211.
Hunt, K. W. (1965). Grammatical structure written at three grade National Institute of Child Health and Human Development
levels. Champaign, IL: National Council of Teachers of English. National Reading Panel. (2000). Teaching children to read:

Allen et al.: Narratives of At-Risk Readers 219


An evidence-based assessment of the scientific research litera- Slobin, D. I., & Welsh, C. A. (1973). Elicited imitation as a
ture on reading and its implications for reading instruction. research tool in developmental psycholinguistics. In C. A.
Washington, DC: Author. Retrieved from http:www.nichd.nih. Ferguson & D. I. Slobin (Eds.), Studies of child language devel-
gov/publications/nrp/findings.htm. opment (pp. 485–497). New York, NY: Hold, Rinehart, & Winston.
Newcomer, P. L., & Hammill, D. D. (1997). Test of Language Snowling, M. J. (2005). Literacy outcomes for children with oral
Development—Primary (3rd ed.). Austin, TX: Pro-Ed. language impairments: Developmental interactions between
Newman, R. M., & McGregor, K. K. (2006). Teachers and lay- language skills and learning to read. In H. W. Catts & A. G.
persons discern quality differences between narratives produced Kamhi (Eds.), The connections between language and reading
by children with or without SLI. Journal of Speech, Language, disabilities (pp. 55–75). New York, NY: Erlbaum.
and Hearing Research, 49, 1022–1036. Stein, N., & Glenn, C. (1979). An analysis of story comprehen-
Owens, R. E. (2008). Language development: An introduction sion in elementary school children. In R. O. Freedle (Ed.), New
(7th ed.). Boston, MA: Pearson Education. directions in discourse processing (pp. 53–120). Norwood, NJ:
Ablex.
Paris, S. G. (2005). Reinterpreting the development of reading
skills. Reading Research Quarterly, 40, 184–202. Strong, C. J. (1998). The Strong Narrative Assessment Procedure.
Eau Claire, WI: Thinking Publication.
Perfetti, C. A. (1985). Reading ability. New York, NY: Oxford
University Press. Swanson, H. L., & Howard, C. B. (2005). Children with reading
disabilities: Does dynamic assessment help in classification?
Peterson, C., & McCabe, A. (1983). Developmental psycho- Learning Disability Quarterly, 28(1), 17–34.
linguistics: Three ways of looking at a child’s narrative.
New York, NY: Plenum. Troia, G. A. (2005). Responsiveness to intervention: Roles for
speech-language pathologists in the prevention and identification
Peterson, C., & McCabe, A. (1991). Linking children’s con- of learning disabilities. Topics in Language Disorders, 25,
nective use and narrative macrostructure. In A. McCabe & 106–119.
C. Peterson (Eds.), Developing narrative structure (pp. 29–54).
Hillsdale, NJ: Erlbaum. Ukrainetz, T. A. (2006a). EBP, RTI, and the implications for
SLPs: Commentary on L. M. Justice. Language, Speech, and
Puranik, C. S., Lombardino, L. J., & Altmann, L. J. P. (2008). Hearing Services in Schools, 37, 298–303.
Assessing the microstructure of written language using a retelling
paradigm. American Journal of Speech-Language Pathology, Ukrainetz, T. A. (2006b). Teaching narrative structure: Coherence,
17, 107–120. cohesion, and captivation. In T. A. Ukrainetz (Ed.), Contextual-
ized language intervention: Scaffolding PreK–12 literacy
Retherford, K. S. (2000). Guide to analysis of language tran- achievement (pp. 195–246). Austin, TX: Pro-Ed.
scripts (3rd ed.). Eau Claire, WI: Thinking Publications.
Ukrainetz, T. A., & Fresquez, E. F. (2003). What isn’t language?:
Ripich, D. N., & Griffith, P. L. (1988). Narrative abilities of
A qualitative study of the role of the school speech-language
children with learning disabilities and nondisabled children: pathologist. Language, Speech, and Hearing Services in Schools,
Story structure, cohesion, and propositions. Journal of Learning 34, 284–298.
Disabilities, 21, 165–173.
Ukrainetz, T. A., Ross, C. L., & Harm, H. M. (2009). An
Roth, F. P., Speece, S. L., & Cooper, D. H. (2002). A longitu- investigation of treatment scheduling for phonemic awareness
dinal analysis of the connections between oral language and early with kindergartners at risk for reading difficulties. Language,
reading. Journal of Educational Research, 96, 259–272. Speech, and Hearing Services in Schools, 40, 86–100.
Roth, F. P., & Spekman, N. J. (1986). Narrative discourse: Spon- VanDerHeyden, A. M., Witt, J. C., & Gilbertson, D. (2007). A
taneously generated stories of learning-disabled and normally multi-year evaluation of the effects of a response to intervention
achieving students. Journal of Speech, Language, and Hearing (RTI) model on identification of children for special education.
Research, 47, 816–834. Journal of School Psychology, 45, 225–256.
Scarborough, H. S. (1990). Very early language deficits in dys- Vellutino, F. R., Scanlon, D. M., Sipay, E. R., Small, S. G.,
lexic children. Child Development, 61, 1728–1743. Pratt, A., Chen, R., & Denckla, M. B. (1996). Cognitive
Schneider, P., & Dube, R. V. (2005). Story presentation effects on profiles of difficult-to-remediate and readily remediated poor
children’s retell content. American Journal of Speech-Language readers: Early intervention as a vehicle for distinguishing
Pathology, 14, 52–60. between cognitive and experiential deficits as basic causes of
Scott, C. M. (1995). A discourse approach to syntax teaching. In specific reading disability. Journal of Educational Psychology,
D. F. Tibbits (Ed.), Language intervention: Beyond the primary 88, 601–638.
grades (pp. 435–463). Austin, TX: Pro-Ed. Weaver, C. (1996). Teaching grammar in context. Portsmouth,
Scott, C. M., & Windsor, J. (2000). General language perfor- NH: Boynton/Cook.
mance measures in spoken and written narrative and expository Westby, C. (1985). Learning to talk—Talking to learn: Oral–literate
discourse of school-age children with language learning dis- language differences. In C. S. Simon (Ed.), Communication
abilities. Journal of Speech, Language, and Hearing Research, skills and classroom success (pp. 181–218). San Diego, CA:
43, 324–339. College-Hill.

220 LANGUAGE, SPEECH, AND HEARING SERVICES IN SCHOOLS • Vol. 43 • 205–221 • April 2012
APPENDIX. OCCURRENCES OF MACROSTRUCTURAL AND
MICROSTRUCTURAL FEATURES IN THE SOURCE STORY

Feature Value

TNW 396.00
NDW 150.00
LENGTH 44.00
MLC 9.00
COORD 5.00
SUBORD 6.00
COMPLEX 20.00
PROPCOMPLEX .45
Complete episodes 5.00
Episodic elements 54.00

Note. TNW = total number of words; NDW = number of different words; LENGTH =
number of C-units; MLC = mean length of utterance in words; COORD = number
of coordinating conjunctions; SUBORD = number of subordinating conjunctions;
COMPLEX = number of C-units containing at least one dependent clause;
PROPCOMPLEX = COMPLEX/LENGTH; Complete episodes = number of
complete episodes; Episodic elements = number of episodic elements.

Allen et al.: Narratives of At-Risk Readers 221


The Narrative Language Performance of Three Types of At-Risk First-Grade
Readers

Melissa M. Allen, Teresa A. Ukrainetz, and Alisa L. Carswell


Lang Speech Hear Serv Sch 2012;43;205-221; originally published online Jan 9,
2012;
DOI: 10.1044/0161-1461(2011/11-0024)

This information is current as of April 5, 2012

This article, along with updated information and services, is


located on the World Wide Web at:
http://lshss.asha.org/cgi/content/full/43/2/205

View publication stats

You might also like