LSHSS
Research Article
The Rules of the Game: Properties of a
Database of Expository Language Samples
John Heilmanna and Thomas O. Maloneb
Purpose: The authors created a database of expository oral
language samples with the aims of describing the nature of
students’ expository discourse and providing benchmark
data for typically developing preteen and teenage students.
Method: Using a favorite game or sport protocol, language
samples were collected from 235 typically developing students
in Grades 5, 6, 7, and 9. Twelve language measures were
summarized from this database and analyses were completed
to test for differences across ages and topics. To determine
whether distinct dimensions of oral language could be
captured with language measures from these expository
samples, a factor analysis was completed.
Results: Modest differences were observed in language
measures across ages and topics. The language measures
I
ndividuals with specific language impairment (SLI)
have significant difficulty with the comprehension
and production of language in the absence of signifi-
cant cognitive and perceptual deficits (Leonard, 2014). For
most individuals with SLI, the disorder presents in early
childhood and persists into adulthood (Stothard, Snowling,
Bishop, Chipchase, & Kaplan, 1998). Much of the research
on children with SLI has focused on the preschool and
early elementary school–age years, when children are mak-
ing the critical (and language-heavy) transition to literacy
(Scarborough, 2009). However, in subsequent school years
and into adolescence, children with SLI continue to experi-
ence significant difficulties with oral and written language
(Stothard et al., 1998) and require the services of speech-
language pathologists (SLPs). In particular, SLPs need to
pay attention to how these children’s deficits affect academic
performance because language skills are central to much of
the junior high and senior high school curricula (National
Governors Association Center for Best Practices, 2010).
a
University of Wisconsin–Milwaukee
b Disclosure: The data were collected by practicing clinicians (who have no financial
School District of Brown Deer, Wisconsin
Correspondence to John Heilmann: heilmanj@uwm.edu
Editor: Marilyn Nippold
Associate Editor: Lynne Hewitt
Received July 17, 2013
Revision received March 17, 2014
Accepted June 2, 2014
DOI: 10.1044/2014_LSHSS-13-0050
Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014 • © American Speech-Language-Hearing Association
were effectively classified into four distinct dimensions:
syntactic complexity, expository content, discourse difficulties,
and lexical diversity.
Conclusion: Analysis of expository data provides a functional
and curriculum-based assessment that has the potential to
allow clinicians to document multiple dimensions of children’s
expressive language skills. Further development and testing
of the database will establish the feasibility of using it to
compare individual students’ expository discourse skills to
those of their typically developing peers.
Key Words: language sample analysis, expository discourse,
assessment, language, diagnostics
Although SLPs working with older students rarely
face new referrals for SLI, under the requirements of the
Individuals with Disabilities Education Act (IDEA; Indi-
viduals With Disabilities Improvement Act of 2004), they
must regularly conduct evaluations for determining con-
tinuing eligibility as well as assessments for determining
treatment objectives and monitoring continuing progress.
Under IDEA, such testing must gauge the extent of a stu-
dent’s progress within the curriculum and be racially and
culturally unbiased. Although numerous norm-referenced
language tests are available for older students (hereafter
referring to students in Grades 5–9), SLPs cannot rely on
them solely when assessing their students. Norm-referenced
tests can be biased against children from cultural and linguis-
tic minorities (Battle, 2011). Furthermore, norm-referenced
tests are most effective when identifying the presence of a
disorder but have limited utility in providing a description
of the nature of the disorder (McCauley & Swisher, 1984).
In developing a profile of a child’s relative strengths and
interests in SALT Software, LLC) and were transcribed at the transcription lab
of SALT Software, LLC. Dr. Jon Miller and the employees of SALT Software
provided the transcription as an in-kind service for this research. The data analysis,
interpretation, and manuscript preparation were completed by the two authors
listed in the byline, neither of whom has a financial interest in SALT Software,
LLC. No members of SALT Software, LLC, were involved in the writing of this
manuscript. The authors believe that there is no real conflict of interest, but they
wish to acknowledge the role of SALT Software, LLC, to address any perceived
conflict of interest.
277
weaknesses, SLPs can instead draw upon naturalistic mea-
sures to provide a rich description of language ability within
a functional context (Damico, Secord, & Wiig, 1992).
Because it provides a method for systematically analyz-
ing children’s production of naturalistic discourse, language
sample analysis (LSA) is a recommended best practice for
clinical assessments (American Speech-Language-Hearing
Association, 2004; Paul & Norbury, 2012). However, accord-
ing to survey data, SLPs who work with junior high or senior
high school students are less likely to use language sampling
than their colleagues who work with preschool or elementary
school–age children (Hux, Morris-Friehe, & Sanger, 1993).
These authors speculated that one cause of the discrepancy
might be the lack of widely available language sampling pro-
cedures for students in the more advanced educational levels.
In recent years, several investigators have documented
that LSA continues to be a powerful assessment tool through-
out the junior high and senior high school years, particularly
when examining children’s use of language to convey infor-
mation through expository discourse (see Nippold, 2014).
Planned expository discourse, both spoken and written, is
directly related to the curriculum for older students and is
thus a prominent feature of the Common Core State Stan-
dards for English Language Arts (National Governors
Association Center for Best Practices, 2010), which have
been adopted by 45 states,1 four territories, the District of
Columbia, and the Department of Defense. For example,
students in Grades 9–10 are expected, when speaking, to
“present information, findings, and supporting evidence
clearly, concisely, and logically” (p. 50). Writing standards
for Grades 9–10 similarly mandate that students produce
“informative/explanatory texts, including the narration of
historical events, scientific procedures/experiments, or tech-
nical processes” (p. 65).
When choosing a task to elicit a language sample, SLPs
need to consider not only its connection to the curriculum
but also how engaging the task is for the child. For instance,
whereas play-based language samples can elicit complex
language in preschool-age children, Evans and Craig (1992)
documented that play-based samples generated less com-
plex language when compared to structured interviews in
8- to 9-year-old children with SLI. Both spoken and written
expository tasks have been found to elicit more complex
language from older students when compared to the lesser
demands of conversation and some narrative tasks, thereby
providing a more challenging and motivating communica-
tive task for older students (Nippold, Hesketh, Duthie, &
Mansfield, 2005; Scott & Windsor, 2000).
1
Since the submission of this article, the number of states adopting the
Common Core Standards has dropped from 45 to 44. In addition,
Missouri and South Carolina have each adopted a process that may
eventually replace the standards we reference, but, for now, we still
count both states as adopters, as do the authors of the Common Core
(http://www.corestandards.org/standards-in-your-state/). We have
included Minnesota as one of the 44 adopting states in the current
discussion; Minnesota has adopted the English Language Arts
Standards, which is the focus of this article, but not the Math Standards.
278 Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014
Studies of expository discourse have documented its
effectiveness in distinguishing children with SLI and non-
specific language impairment (NLI) from their typically
developing peers, thus providing an evidence base for its
use in clinical assessments (Nippold, Mansfield, Billow, &
Tomblin, 2008; Scott & Windsor, 2000). However, a major
obstacle to clinical application of expository LSA is deter-
mining what constitutes an age-appropriate level of perfor-
mance. To get such comparative data, clinicians can refer to
published studies describing children’s performance on ex-
pository tasks. However, many of the studies’ samples were
drawn from different populations using different elicitation
protocols, which can have a significant impact on the mea-
sures (Westerveld & Heilmann, 2012). In addition, the time
requirement to generate tables of normal performance is pro-
hibitive for most busy clinicians.
An alternative approach is to establish a large-scale
normative database. Using databases developed with the
software program Systematic Analysis of Language Tran-
scripts (SALT; Miller & Iglesias, 2012), Heilmann, Miller,
and Nockerts (2010) compared measures from the conver-
sational language samples of children with SLI to those of
typically developing speakers and thereby correctly identi-
fied 81% of the children who had a true impairment. Rec-
ognizing the strength of analyzing children’s expository
language data within the clinical context and the power of
large databases of normal speakers, Nippold, Mansfield, and
Billow (2007) and Scott (2010) have recommended the crea-
tion of large-scale expository databases to provide bench-
mark data for typically developing adolescents and young
adults. In serving as a school-based SLP at the middle and
high school level, the second author observed repeatedly that
his students with language difficulties struggled to produce
expository discourse, including how to play their favorite
game or sport. These clinical experiences led to his interest in
creating the expository database described in this manuscript.
Elicitation Considerations for Expository LSA
The first step in creating a database is to develop a
protocol for eliciting the samples that facilitates reliable
comparisons to the database. The goal is to create a proce-
dure that generates meaningful data related to the curricu-
lum while being repeatable across children. In collecting
their expository samples, Scott and Windsor (2000) had all
children describe the same content by prompting them to
retell events depicted in a 15-min film about life in the de-
sert. In contrast, Nippold et al. (2005) allowed their par-
ticipants more freedom to individualize their expositions
by prompting children to describe their favorite game or
sport. Having each child select his or her favorite game
or sport would appear to introduce additional variability
across children (who are describing different content); how-
ever, this individualization presumably increases the mean-
ingfulness of the productions due to the child’s personal
knowledge and interest in the activity. Moreover, descrip-
tions of a favorite game or sport are likely to be culturally
unbiased because all cultures have games and sports particular
to them, and the participants were free to draw on their cul-
tural experience in making their selection.
For our expository database, we too chose to prompt
children to describe their favorite game or sport. We be-
lieved this task would motivate students by allowing them
to draw on their personal experiences. Moreover, it aligns
well with academic expectations. For example, Wisconsin’s
Model Academics Standards for Physical Education require
students in Grades 6–8 to “describe the critical elements of
a sport-specific skill (e.g., a basketball free throw or forearm
pass).” Students are also expected to “explain at least two
game tactics involved in playing net/wall sports (e.g., tennis,
badminton, volleyball, etc.) . . . [and] . . . at least two game
tactics involved in invasion sports (e.g., soccer, basketball,
handball, etc.”; Wisconsin Department of Public Instruction,
Student Services/Prevention, and Wellness Team, 2010,
p. 55). For Grades 9–12, the standards call on students to
“describe offensive, defensive, and transitional skills and
strategies in team and individual sports” (p. 67). In addition,
students are expected to “describe the impact of new skills
and tactics . . . [and] . . . explain appropriate tactical decisions
in a competitive activity” (p. 69).
The elicitation protocol in the current study varied
from Nippold et al. (2005) in two important ways: by
introducing planning time and materials and by using
writing to support the production of oral language samples.
Whereas the participants in Nippold et al. began speaking
immediately after hearing the instructions and were asked
only a single follow-up question about strategies, we pro-
vided the students with an outline identifying the key com-
ponents of the expository (see Appendix A). Students were
given the opportunity to plan for each aspect of their ex-
planation by writing and/or drawing. After completing their
plan, students were then encouraged to refer to it while
speaking. These procedures were motivated by current
trends in authentic performance assessments that encourage
teachers to make curricular expectations clear to students
(Wiggins & McTighe, 2005). In this case, we wished to
convey that a detailed and comprehensive explanation was
expected. In a study of written narratives, Sun and Nippold
(2012) took a similar approach to planning by giving their
subjects “a detailed outline” that could be used “to address
a set of key story grammar elements” (p. 5) as they wrote.
Moreover, we believed that using the written modal-
ity to support the oral language samples aligned better with
the standard curriculum when compared to previous methods
of eliciting expository discourse. Our approach follows
Scott and Stokes (1995), who suggested that with older stu-
dents, challenging and curriculum-based language sampling
contexts often involve a mix of speaking and writing. Fur-
thermore, we believed that using writing to support oral
production was consistent with the philosophy of the Com-
mon Core Standards. Although separate grade-level strands
were devised for speaking, listening, reading, and writing
objectives, the Common Core Standards presume that “the
processes of communication are closely connected” and thus
call for “an integrated model of literacy” (National Gover-
nors Association Center for Best Practices, 2010, p. 4).
Testing the Properties of the Database
Having a large database of expository language sam-
ples affords the opportunity to test the properties of various
language measures, which will provide guidance for SLPs
wishing to use expository data in their clinical assessments.
Changes with increasing age. Previous cross-sectional
studies of spoken exposition have documented slow but
steady growth in syntactic complexity from childhood into
adulthood. Syntactic measures found to increase with age
include mean length of utterance (MLU; as measured in
T-units), clausal density (i.e., the average number of clauses
per T-unit), and the frequency of relative clauses (Berman
& Verhoeven, 2002; Nippold et al, 2005; Nippold et al., 2007;
Verhoeven et al., 2002). In addition, age-related increases for
exposition have been reported for several nonsyntactic mea-
sures: (a) type–token ratio, which has been interpreted as a
measure of lexical diversity (Berman & Verhoeven, 2002);
and (b) sample length, which has been viewed as a measure
of language output (Berman & Verhoeven, 2002, using total
words; Nippold et al., 2005, using T-units). By examining
a wider range of language measures in our study, we hoped
to determine which ones are most sensitive to age-related
changes.
Effect of topic. In addition, we were interested in de-
termining if the freedom to choose a topic in the favorite
game or sport (FGS) protocol had a significant effect on
the resulting language measures. Previous studies examin-
ing the FGS protocol allowed children to select any game
or sport (team or individual) and presumably assumed that
the topic had minimal influence on the measures. In the
present study, we tested for clinically significant differences
across three topic choices: team sports, such as football or
basketball; individual sports, such as golf or bowling; and
games, such as card games like 500 rummy or board games
like Monopoly.
Effects of elicitation procedures. We were next inter-
ested in how children’s performance using our modified
FGS protocol would compare to existing reports in the lit-
erature. In particular, we questioned whether adding extra
planning time and combining spoken and written modali-
ties in the FGS protocol would affect the resulting language
measures. Differences in language sample measures have
been observed within children’s narrative productions based
on the presence or absence of pictures to support the child’s
story production (Schneider & Dubé, 2005; Westerveld &
Heilmann, 2012) and the amount of modeling provided
prior to initiating the narrative task (Gummersall & Strong,
1999). We know of no studies to date that have directly
examined the effects of variant elicitation procedures on
language measures for expository discourse.
Using expository data to describe children’s profiles.
Clinicians are encouraged to use descriptive and naturalis-
tic assessments, such as LSA, when describing the nature
of a child’s impairment and identifying relative strengths
and weaknesses across language dimensions (McCauley,
2001; Paul & Norbury, 2012). These rich descriptions can
be used to characterize a student’s relative strengths and
Heilmann & Malone: Properties of Expository Samples 279
weaknesses and assist with planning individualized treat-
ments. However, there is limited empirical support for the
practice of using assessment data to develop clinical profiles.
Tomblin and Zhang (2006) completed a thorough
testing of language dimensionality using multiple norm-
referenced language assessments and documented that
their measures could be broken into the two dimensions
of vocabulary and sentence use. Because measures from
standardized tests are not designed to provide detailed de-
scriptions of children’s relative strengths and weaknesses
(McCauley & Swisher, 1984), analyses of language dimen-
sionality should be improved by using descriptive lan-
guage assessments, such as measures from language samples.
In fact, Westerveld and Gillon (2010) documented that mea-
sures from oral narratives could be broken into three distinct
dimensions: content, grammatical ability, and grammatical
complexity. Having a large database of expository samples
provides the opportunity to test for the dimensionality of
children’s expository language. If there are indeed multiple
dimensions of language that can be captured from exposi-
tory language measures, SLPs would have support for the
use of expository LSA to develop profiles of older students’
language skills and plan appropriate treatments based on
the students’ relative strengths and weaknesses.
Upon compiling a database of expository language
samples, we had the opportunity to examine the nature of
expository discourse in older students. In a first step toward
developing a validated clinical protocol, we documented
the properties of this database by addressing the following
research questions:
1. For samples elicited with a modified FGS protocol,
were there significant age-related changes within a
wide range of syntactic and nonsyntactic language
measures?
2. Does the topic selected (i.e., team sport, individual
sport, or game) have a substantial effect on the resulting
language measures?
3. Were the language measures from the current database
substantially dissimilar from those found in the
expository literature for children of comparable age?
4. When a factor analysis is applied to language
measures from the current database, do they capture
unique dimensions of children’s spoken exposition?
Method
Participants
The data reported in this manuscript are combined
from two separate projects involving typically developing
children who attended public schools in Wisconsin. The
first project collected expository samples from 7th (n = 41)
and 9th (n = 39) grade students (Malone et al., 2008), and
the second project added samples from 5th (n = 86) and
6th (n = 69) grade students (Malone, Heilmann, Miller,
DiVall-Rayan, & Rolland, 2010). For both projects, admin-
istrators at the school or district level approved the study
280 Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014
and allowed the research team to send a written request for
participation to all families of students who met the inclu-
sionary criteria. Children were ineligible for the study if they
were currently receiving special education services or English-
language-learning services, as determined by a review of the
school records. Samples were collected from a total of
235 students who had a mean age of 12;6 (years;months;
SD = 1;5; range = 10;7–15;9). Children were evenly split
for gender; 49% of the participants were female and 51%
were male.
On all students for whom informed consent was ob-
tained, demographic information was collected and ana-
lyzed. Our aim was to have the students in this database
be representative of the broader population of monolingual
English-speaking preteen and teenage students. However,
because the present study was a grassroots effort with a
limited budget, we relied on SLPs who volunteered to col-
lect samples from as diverse a population as was available
to us in Wisconsin. Earlier studies by the first author have
documented that geographical differences had a modest
effect on children’s narrative language sample measures
when comparing children from Wisconsin to children from
San Diego (Heilmann, Miller, & Nockerts, 2010) and when
comparing children from the United States to children
from New Zealand (Westerveld & Heilmann, 2012). There-
fore, we attempted to closely match the racial and socio-
economic demographics of the state of Wisconsin. The
children’s socioeconomic status was documented by deter-
mining the percentage that received free or reduced-price
lunches (Ensminger et al., 2000), which revealed that 24%
of the children were eligible for the program. The race and
ethnicity breakdown of the participants was 80% White,
11% African American, 7% Hispanic or Latino, and 2%
Asian. This is slightly more diverse than the state demo-
graphics reported in the 2010 United States Census, which
were 83% White, 7% African American, 6% Hispanic or
Latino, and 3% Asian.
Procedure
Samples were collected by 70 school-based SLPs from
students who attended 55 different schools located within
seven Wisconsin school districts. Each SLP collected 3.0
(SD = 1.6) samples on average. All examiners stated that
they were familiar with the language sampling procedures,
held a Certificate of Clinical Competence from ASHA, and
held an educator’s license from the Wisconsin Department
of Public Instruction. Two SLPs provided training on the
recruitment and elicitation procedures. The second author
was the trainer for approximately half of the clinicians. An-
other SLP with over 20 years of experience collecting and
analyzing language samples was responsible for training the
remaining clinicians. Examiners first reviewed a written
overview of the project that was followed by a 1-hr training
session. The SLPs contacted the second author for clarifica-
tion whenever questions arose.
Participants were asked to explain how to play a fa-
vorite game or sport to a naïve listener. Examiners read
from a script (see Appendix B) that provided explicit in-
structions for completing the task and established the expec-
tation that students provide a comprehensive and detailed
explanation that was at least 5 min in length. Once the partic-
ipants had selected a game or sport, they were given a plan-
ning sheet that listed the eight expected components of the
explanation and included a description of what information
was to be presented for each component (see Appendix A).
Participants were then asked to take a few minutes to plan
their explanation either by taking notes in the space provided
or by drawing a diagram or graphic organizer on the back
of the planning sheet. Examiners reported that the average
time to produce a sample was approximately 20 minutes,
with students devoting the majority of their time toward
completing the planning sheet.
Data Transcription and Analysis
All expository samples were digitally recorded and
sent to SALT Software, LLC, for transcription by trained
research assistants. Each research assistant (RA) was re-
quired to complete at least 10 hr of training on the standard
transcription conventions for the SALT program. The tran-
scribers for this project completed an additional 5 hr of
training and practice for coding the Expository Scoring
Scheme (ESS), a rubric that will be explained below. At the
end of the training, each RA transcribed and coded five
complete language samples, which required at least 90%
agreement with a master transcript.
Each sample was transcribed and coded via consensus
transcription, where one RA completed an initial transcript
that was checked by a second RA who listened to the sample
and noted any disagreements. These discrepancies were dis-
cussed and resolved, with the final decision made by the
first transcriber. Utterances were segmented into communi-
cation units (C-units; Loban, 1976). A C-unit is defined as
an independent clause plus all associated subordinate clauses.
The ESS was coded by hand after completing the initial tran-
scription, as was clausal density (see below for details).
Language Measures
The SALT program was used to generate multiple
language measures for each transcript. All measures were
limited to C-units that were complete and intelligible. Al-
though SALT can generate dozens of measures, we limited
our analysis to those measures that have been described
in the literature as being developmentally robust in older
students and/or sensitive to differences between children
with SLI and their typically developing peers.
Mean length of C-unit (MLCU). Calculating the
average number of words in a child’s C-unit provides an
overall index of children’s language growth and is particu-
larly relevant to syntactic development (Scott & Stokes,
1995). Studies of expository language samples, which chal-
lenge older children to produce complex language, have re-
vealed that measures of utterance length (such as MLCU)
did in fact increase during the school-age years and into
adulthood (Nippold et al., 2005).
Clausal density (CD). Scott (1988) described how
school-age children’s utterances lengthened and became in-
creasingly complex by using a greater number of subordinate
clauses. Although children’s use of subordinate clauses is
reflected in MLCU, calculating the frequency of subordinate
clause production provides a more direct assessment of chil-
dren’s use of complex syntax. Previous studies have docu-
mented that young children rarely use subordinate clauses
(Scott, 1988) and that there are small incremental gains from
school-age years into adulthood (Loban, 1976; Nippold
et al., 2005, 2007). In the current study, CD was calculated
for each transcript following the procedures described in
Nippold et al. (2005), whereby we summed the total number
of independent clauses and finite subordinate clauses and
then divided that total by the number of C-units.
In addition, each finite subordinate clause was coded
according to its syntactic function (as described in Nippold
et al., 2005). An adverbial clause functions as an adverb
by describing a cause or condition; a nominal clause func-
tions as a noun phrase in either the subject or object posi-
tion; and a relative clause functions as an adjective by
describing the noun that precedes it. The following C-unit
is drawn from a student’s explanation of a penalty in hurl-
ing, a Gaelic field game (Worth, 1990); it consists of seven
clauses and features all three types of subordinate clauses:
And if it’s by the goalie [adverbial clause]
or a point is scored [adverbial clause]
or it goes in the goal [adverbial clause],
it’s a puck-out [independent clause],
which is [relative clause]
the goalie just hits it [nominal clause]
as far as he can to his teammates [adverbial clause].
For each clause type, we controlled for sample length
by dividing the total number of clauses by the number of
C-units in the sample. Nippold et al. (2005) documented
that preadolescent children predominantly used adverbial
and nominal clauses, with a substantially lower frequency
of relative clauses. However, relative clauses were found to
have the strongest age-related growth.
Lexical diversity. Measures of lexical diversity docu-
ment the richness of vocabulary that children use and are
thought to reflect overall vocabulary skills (Leadholm &
Miller, 1992). Watkins, Kelly, Harbers, and Hollis (1995)
documented that measuring the number of different words
(NDW) for a predetermined sample size was most effective
for identifying children with SLI. In the current study, we cal-
culated the NDW for the first 150 words in each child’s sam-
ple, a procedure comparable to Scott and Windsor (2000).
Productivity. Measures of productivity document the
amount of information produced by the child and the effi-
ciency with which the child produced the sample. With a
predetermined language sampling task, such as telling a
story or describing a procedure, there is an expectation that
children produce an adequately detailed description with
sufficient length. However, children with SLI tend to
Heilmann & Malone: Properties of Expository Samples 281
produce significantly shorter samples than their typically
developing peers in both narrative (Strong & Shaver, 1991)
and expository discourse (Scott & Windsor, 2000). Be-
cause the children in the present study were instructed to
talk for at least 5 min, we also expected that length would
be an indicator of a child’s overall skill in generating ex-
pository discourse.
The number of total words (NTW) was our first
measure of sample length and was calculated by summing
the number of main body words. The number of total
C-units (TCU), our second measure of length, was calcu-
lated by summing the number of complete and intelligible
C-units produced by the child. Although we expected NTW
and TCU to be closely related, we believed that TCU would
be less constrained by children’s use of complex syntax. That
is, even if the child produces short, simple utterances with
minimal subordination, he or she may still relate a large
amount of information that would likely be better captured
using TCU. Conversely, another child may pack a consider-
able amount of information into a relatively small number
of utterances by using complex syntax; for that child NTW
may serve as a better measure of productivity.
Expository Scoring Scheme (ESS). In addition to the
utterance-level microstructural measures described above,
children’s connected discourse often has concepts that tran-
scend the individual utterance and convey text-level con-
cepts. These discourse-level measures have been well studied
in children’s narrative productions (e.g., Heilmann, Miller,
Nockerts, & Dunaway, 2010; Petersen, Gillam, & Gillam,
2008) but have received limited attention with expository
discourse. The literature to date on expository rubrics has
been limited to written exposition (e.g., Koutsoftas & Gray,
2012) and case study examples demonstrating how children
increased the sophistication of their spoken and written ex-
positions by repeatedly referring to the central themes or
points of the process they are describing and by deploying a
variety of cohesive devices (Scott, 2010).
The ESS rubric (see Appendix C) was designed to
provide a more in-depth analysis of children’s ability to
clearly and comprehensively convey the key points of their
explanation. Its format was modeled after the Narrative
Scoring Scheme (NSS), a rubric developed as a hybrid
method of documenting the overall quality of narrative
performance. To score the NSS, examiners identify specific
macrostructural features while employing subjective judg-
ment (Heilmann, Miller, Nockerts, & Dunaway, 2010).
Heilmann, Miller, and Nockerts (2010) documented that
incorporating examiner judgment into this type of hybrid
scoring system was more sensitive than discrete coding
schemes that simply identified the presence of individual
macrostructural aspects of children’s narrative retells. For
the ESS, eight of its 10 components were drawn from com-
mon features of sports discussed in Worth (1990). To en-
sure that these features also reflected what is expected for
explanations of games, the website Card Game Rules
(www.pagat.com) was consulted.
The transcribers coded each sample using the ESS
rubric to document how well the child conveyed each of
282 Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014
these eight components of the expository sample presented
on the planning sheet. In addition, transcribers completed
two additional global ratings on the appropriateness and so-
phistication of terminology used in the sample (terminology)
and the level of cohesiveness throughout the production
(cohesion; Halliday & Hasan, 1976). Each of the 10 charac-
teristics was scored on a 0–5 scale. Anchors were provided
for performance that was rated as follows: 1 = immature/
minimal use, 2 = emerging/inconsistent, and 5 = proficient. If
the RAs judged that the child’s performance was between
two of the anchors, they were instructed to give a score of
2 or 4. A score of 0 was given if the student failed to address
a particular component of the rubric.
Verbal fluency. We used two measures to document
how fluently children produced connected discourse.
Words per minute (WPM) represents the speed with which
a child produced the sample and was calculated by divid-
ing all the words produced in the sample by the elapsed
time. Whereas an abnormally high speaking rate can neg-
atively affect a child’s communicative ability, a faster
speaking rate suggests that a speaker was able to organize
and generate his/her thoughts coherently and fluently
(Leadholm & Miller, 1992). The second measure of verbal
fluency was derived from coding breakdowns in produc-
tions. In natural discourse, most speakers spontaneously
produce reduplications, false starts, and reformulations that
Loban (1976) collectively termed mazes. Guo, Tomblin,
and Samelson, (2008) documented that children with SLI
produce significantly more mazes than their typically devel-
oping peers.
Errors and omissions. Children with SLI have been
shown to demonstrate significantly more grammatical errors
when producing expository text when compared to their
typically developing peers (Scott, 1995) as well as their
language-matched peers (Scott & Windsor, 2000). The SALT
software was used to generate four separate measures reflect-
ing children’s discourse errors: errors at the word level (e.g.,
overgeneralizations and incorrect word choice), syntactic
errors at the utterance level (i.e., utterances that do not
make syntactic sense), omissions of bound morphemes that
occur in obligatory contexts, and omissions of words that
occur in obligatory contexts. The Errors and Omissions
measure was calculated by summing these four measures.
Transcription and Coding Agreement
Twelve of the transcripts were randomly selected to
be transcribed and coded by a separate transcriber, and the
second set of transcripts was compared to the original set.
Percent agreement was calculated by dividing the total
number of agreements by the sum of the total agreements
and disagreements. High levels of agreement were ob-
served: 97.8% for morphemes (18,465/18,890), 95.1% for
C-unit segmentation (1,328/1,396), and 95.6% for placement
of mazes (778/814). Krippendorff’s alpha (Krippendorff,
1980), which accounted for the degree of difference in ordi-
nal data, was .91 for coding of CD and .81 for coding of
the ESS.
Results
Properties of the Database
To test for age-related changes in language sample
measures, a series of bivariate correlation coefficients was
calculated between the children’s age and their performance
on each of the expository measures. There was a small and
nonsignificant correlation with age for most of the measures
(see Column 1 of Table 1). Significant age-related correla-
tions with a small association (Cohen, 1988) were observed
for MLCU, relative clauses, NDW, WPM, ESS, and mazes.
To illustrate how the measures differed across age groups,
we summarized each measure across the four grades in
Columns 2–5 of Table 1.
We next completed a series of one-way analysis of
variance (ANOVA) equations to determine if each of the
language measures differed significantly due to the topic
selected (i.e., team sport, individual sport, or game). Of the
12 measures, significant differences were observed for only
three: NDW, F(2, 232) = 4.2, p = .017, h2 = .04; NTW,
F(2, 232) = 4.0, p = .019, h2 = .03; and TCU, F(2, 232) = 4.8,
p = .009, h2 = .04. Table 2 summarizes descriptive statistics
broken down by topic for these three measures. We com-
pleted a series of Scheffé post hoc analyses to determine
where significant differences occurred across topics for each
of the measures. For NDW, children’s values from the game
samples were significantly lower than team sport ( p = .02);
no significant differences were observed between game and
individual sport ( p = .11) and individual sport and team
sport ( p = .98). For both NTW and TCU, measures from
the game samples were significantly lower than individual
sport ( p = .02 and .01 for NTW and TCU, respectively); no
significant differences were observed between game and
team sport ( p = .30 and .35) and between team sport and in-
dividual sport ( p = .14 and .06). For the other nine language
measures, nonsignificant differences ( p ≥ .10) with small
effect sizes (h2 ≤ .02) were observed.
Comparison to the Literature
The performance of the children in the current study
was compared to previous reports in the literature that an-
alyzed the expository discourse of typically developing chil-
dren who were similar in age to children from the present
study (see Table 3). We first compared our data to three
studies that also used an FGS protocol, but without the
graphic organizer (Nippold et al., 2005, 2008; Westerveld &
Moran, 2011). In addition, we compared language measures
from our samples to expository interviews with children who
have chess knowledge (Nippold, 2009) and the retellings
of an expository video described in Scott and Windsor
(2000). Although all of the children were typically develop-
ing, there were slight differences in ages between the stud-
ies. However, the differences were modest, and previous
research has documented that expository measures show
relatively small cross-sectional age-related growth during
the preadolescent years (e.g., Nippold et al., 2005). Mod-
est age-related growth was further confirmed in the cross-
sectional analyses completed in the current study. Because
the biggest difference across these studies was the method
used to elicit the expository samples, we predicted that any
observed differences in measures were due to differences
in the elicitation method. For these comparisons, the vari-
ables were limited to TCU, MLCU, and CD, which were
consistently included across the studies.
As seen in Table 3, TCU, MLCU, and CD were con-
siderably higher for the children in the current study when
compared to the three other studies that used an FGS proto-
col (Nippold et al., 2005, 2008; Westerveld & Moran, 2011).
In comparison to the present study, interviews of children

Table 1. Relationship between language measures and age and descriptive data (M and SD) for each measure, broken down by grade.

Grade
Language measure r with age 5th 6th 7th 9th

MLCU .16* 11.4 (1.9) 11.7 (1.6) 11.9 (1.7) 12.1 (2.0)
CD .01 1.7 (0.2) 1.7 (0.2) 1.7 (0.2) 1.7 (0.2)
Adverbials –.04 35.9% (15.9%) 34.4% (14.2%) 33.4% (15.3%) 34.5% (15.0%)
Nominals .03 13.6% (8.0%) 13.8% (7.9%) 13.7% (9.3%) 13.6% (6.9%)
Relatives .19** 12.8% (8.5%) 16.6% (11.3%) 16.9% (9.3%) 18.5% (11.1%)
NDW .12* 74.6 (7.4) 75.1 (7.2) 75.7 (6.1) 76.9 (7.3)
NTW .08 610.6 (279.1) 583.3 (238.6) 706.1 (404.2) 634.5 (214.8)
TCU .03 53.9 (23.5) 50.2 (19.5) 60.4 (34.3) 52.4 (16.0)
ESS .14* 32.7 (5.7) 32.5 (5.6) 34.2 (5.1) 34.6 (5.2)
WPM .22** 126.0 (27.6) 129.0 (19.6) 135.4 (23.7) 140.8 (27.0)
Mazes –.15** 10.2% (5.0%) 10.5% (4.4%) 8.5% (3.4%) 8.4% (4.2%)
Errors and omissions –.01 6.5 (4.3) 6.4 (5.2) 7.5 (4.5) 5.7 (4.7)

Note. MLCU = mean length of C-unit; CD = clausal density; adverbials = adverbial clauses as a percentage of C-units; nominals = nominal clauses
as a percentage of C-units; relatives = relative clauses as a percentage of C-units; NDW = number of different words; NTW = number of total words;
TCU = number of total C-units; WPM = words per minute; ESS = expository scoring scheme; mazes = percentage of maze words to total words;
errors and omissions = total number of word errors, utterance-level errors, and omissions of words and morphemes that occur in obligatory contexts.
*p < .05. **p < .01.

Heilmann & Malone: Properties of Expository Samples 283

Table 2. Descriptive statistics (M and SD) of measures for which topic had a significant effect.

Topic
Language measure Team sport (n = 151) Individual sport (n = 30) Game (n = 54)

NDW 76.0 (7.5) 76.2 (5.6) 72.9 (6.4)

NTW 624.6 (228.4) 738.2 (527.2) 555.3 (223.8)
TCU 52.5 (18.6) 64.6 (43.4) 48.1 (19.8)

Note. Results of Scheffé post hoc testing revealed that for NDW, game < team sport and for NTW and TCU, game < individual sport.

on how to play chess (Nippold, 2009) generated substan-
tially longer samples but similar MLCU and CD values.
The present study generated considerably longer samples
than Scott and Windsor’s (2000) video retellings, but similar
MLCU and CD values.
Factor Analysis
We next completed a factor analysis to discover if
there were in fact distinct dimensions of language being cap-
tured by the expository samples and to identify which lan-
guage measures were associated with the same underlying
construct. When completing a factor analysis, the variance
structure of the measures is examined in order to identify
variables that share a significant amount of variance. Factor
analyses assist in identifying variables that are strongly in-
terrelated (i.e., have substantial covariance) while at the same
time are distinct from other groups of variables (i.e., have
limited covariance with other measures). See Tabachnick and
Fidell (2013) for an overview of factor analysis.
The first decision to make when completing a factor
analysis is whether it should be confirmatory (where the
examiner tests how well variables adhere to a known factor
structure) or exploratory (where the factor structure is un-
known and is empirically derived through the analysis). Al-
though we had some initial ideas on which variables would
covary, there were limited data describing the factor struc-
ture of measures from expository language samples collected
from older students. Therefore, we chose to complete a prin-
cipal components analysis (PCA), which is one method of
completing an exploratory factor analysis. In the current
study, the 12 expository measures were analyzed using a
PCA with varimax rotation. The varimax rotation manipu-
lates the data to strengthen the loadings onto their respective
factors, which affords a clearer observation of the factor
structure. The final choice when completing a PCA is deter-
mining how many factors to include in the final model. If a
small number of factors were retained (e.g., a single-factor
model), we could not determine if there were in fact distinct
dimensions of language (i.e., everything would load on a
“general language” dimension) and we would not explain
much of the variability in our measures. If, on the other
hand, we retain too many factors, the model would become
too watered down and would prohibit a parsimonious
interpretation of the factor structure. One approach is to
“eyeball” the data by completing multiple models (e.g., one-,
two-, three-, and four-level models) and judging which is
the best fit. However, such approaches are prone to bias.
We chose to use a more empirical approach by including
only factors with eigenvalues greater than one.
The data were best explained by a four-factor model,
which is summarized in Table 4. The strongest loadings for
Factor 1 were from MLCU, CD, and the three types of
subordinate clauses. Together, these variables explained
24% of the variability in language measures. For Factor 2,
the strongest loadings were observed for NTW and TCU,
with additional strong loadings observed for WPM and
ESS. Variables in Factor 2 uniquely explained 22% of the
variability in the measures. For Factor 3, strong loadings
were observed for children’s use of mazes and errors and
omissions, which together explained 11% of the variability.
Finally, NDW and the use of relative clauses had a strong
and unique loading onto Factor 4, which explained 10% of
the variability. In addition, children’s use of relative clauses
had a notable loading onto Factor 4.
Discussion
The need for authentic language assessments for
older students motivated the creation of an expository

Table 3. Summary of measures (M and SD) collected from the present study and comparable studies in the literature.

Task Study n Age M (range) TCU MLCU CD

FGS (with outline) Present study 235 12;6 (10;7–15;9) 53.5 (23.7) 11.7 (1.8) 1.7 (0.2)
FGS (no outline) Nippold et al., 2005 20 11;4 (10;9–11;8) 35.3 (17.4) 9.3 (0.9) 1.5 (0.2)
Nippold et al., 2008 254 13;11 (13;1–15;5) 29.2 (17.2) 10.0 (2.1) 1.5 (0.2)
Westerveld & Moran, 2011 20 11;4 (11;0–11;11) 32.4 (15.9) 9.1 (1.3) 1.4 (0.1)
Chess interview Nippold, 2009 32 10;11 (7;3–15;4) 88.1 (56.8) 11.5 (2.3) 1.6 (0.2)
Video retell Scott & Windsor, 2000 20 11;6 (9;10–12;11) 29.4 (9.1) 11.4 (2.1) 1.8 (0.2)

Note. FGS = favorite game or sport protocol.

284 Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014
Table 4. Summary of data from the exploratory factor analysis.
topic only accounted for 2% to 4% of the variance in the
a b c d
measures. Examiners should use caution when collecting
Variable Factor 1 Factor 2 Factor 3 Factor 4
samples describing games, because NDW, TCU, and NTW
MLCU .86 .15 –.16 .11 values may be underestimated. Given the small effect sizes,
CD .96 .09 .01 .05 we believe that these differences among the three topic
Adverbials .78 .15 .01 –.24 choices will have minimal impact on the interpretation of
Nominals .46 –.13 .05 .01 results within clinical protocols; further study of children
Relatives .50 .12 –.08 .52
NDW –.14 –.01 .03 .86 with LI will be required to determine language measure dis-
NTW .05 .93 .14 –.03 crepancies that constitute clinical significance.
TCU –.23 .89 .19 –.06 We do believe that the risk of a topic effect is out-
ESS .23 .60 –.03 .25 weighed by the benefits of allowing students flexibility to
WPM .12 .69 –.10 –.01
Mazes .03 –.13 .81 –.09 choose a meaningful topic. By choosing from among the
Errors and omissions –.07 .28 .75 .09 total range of games and sports, students are more likely
to identify one with which they are familiar and that they
Note. The explained variance estimates (Row 3) document the are motivated to discuss. In addition, when collecting multi-
amount of variability explained by each respective factor. The
loadings (Rows 4–15) document the correlation between each ple samples over time to monitor a student’s progress, an
measure and its respective factor. Boldface values indicate the SLP can allow a student to explain a wide variety of differ-
variables that were judged to have substantial loadings onto their ent games or sports, thereby avoiding artificial increases
respective factors. in measures due to an increased familiarity with explaining
a
Eigenvalue = 2.9; explained variance = 24.0%. bEigenvalue = 2.7; a particular game or sport.
explained variance = 22.2%. cEigenvalue = 1.3; explained variance =
11.0%. dEigenvalue = 1.2; explained variance = 9.7%.
Comparison to Previous Studies
database that can be used to compare the performance of Another goal was to compare our protocol to elicita-
individual clients to a large sample of typically developing tion procedures used in other studies of expository discourse.
speakers. We collected a large corpus of samples from typi- Because these comparisons relied on descriptive statistics
cally developing children using a protocol that we believed alone and there was no way to determine statistical signifi-
best reflected the demands of the curriculum for preteen cance, we estimated whether the differences across the studies
and teenage students. Having this database allowed us to appeared to have clinical significance. There were substantial
document the nature of expository discourse in typically differences in sample length (TCU) and utterance complexity
developing students and provides motivation for the further (MLCU and CD) when comparing the present study to the
testing of the use of expository LSA in assessing children three previous studies examining the FGS protocol. We at-
with a range of language impairments, including both SLI tributed the increased length and complexity in our study to
and NLI. the extra support and scaffolding provided by our elicitation
task. Allowing students to use a planning sheet to organize
and support their verbal productions likely prompted them
Age-Related Growth to include more detail in their productions. The planning
Half of the measures significantly increased (MLCU, process may also have reduced the processing load required
relative clauses, NDW, WPM, ESS, and mazes) with chil- for organization of their productions, thus freeing up cogni-
dren’s advancing age (albeit with a small effect), whereas tive resources to facilitate the production of longer and more
the rest demonstrated nonsignificant age-related effects. It complex samples.
may seem somewhat surprising that there weren’t greater To put the comparisons in a broader perspective, we
age-related changes because the students covered a fairly next compared the results from our study to additional
wide range (from 5th through 9th grades). Yet our findings studies of expository discourse that did not use the FGS
were consistent with the slow age-related growth documented protocol. When comparing measures from the present study
in previous studies of children’s expository discourse (e.g., to those that were elicited through interviews about chess
Nippold et al., 2005; Westerveld & Moran, 2011). (Nippold, 2009) and through retellings of a science video
(Scott & Windsor, 2000), we observed similar MLCU and
CD values. The extra support provided by the questioning
Effect of Expository Topic of the interviewers and the children’s strong content knowl-
To our knowledge, we completed the first analyses of edge in Nippold (2009) may explain why their measures of
topic effects for an expository task. The results of our anal- utterance complexity were comparable to the present study.
yses revealed that there were minimal differences in lan- The retell context in Scott and Windsor may have also
guage measures across the topics of team sports, individual provided additional scaffolding for the children, requiring
sports, and games. Significant differences were observed fewer cognitive resources for discourse planning and facili-
only for NDW and for the two measures of sample length tating greater ability to generate more complex syntax. A
(i.e., TCU and NTW). Although there were statistically similar advantage for narrative retells over self-generated
significant differences across topics for these measures, narratives was documented by Merritt and Liles (1989).

Heilmann & Malone: Properties of Expository Samples 285

 Although no definitive conclusions could be reached
regarding the differences in measures across studies, the
data suggest that the elicitation methods did indeed have an
effect on the expository measures. This result was consistent
with previous studies documenting the significant effect of
elicitation procedures on narrative measures (Gummersall
& Strong, 1999; Westerveld & Heilmann, 2012). Our pur-
pose in comparing elicitation contexts was not to identify
one as superior to the others, but rather to acknowledge the
potential effects of elicitation procedures and to reiterate
the importance of consistently following the same protocol
that was used to compile the comparison samples. To con-
firm the influence of different elicitation procedures on stu-
dents’ performance, a future study would need to make
comparisons using the same participants.
Factor Analysis
The factor analysis revealed that measures from expos-
itory LSA successfully captured multiple dimensions of
oral language. The current study has extended the work of
Tomblin and Zhang (2006) and Westerveld and Moran
(2011) to provide empirical evidence for the presence of mul-
tiple language dimensions. The interpretation of the factors
was relatively straightforward. Factor 1 revealed that MLCU,
CD, and the three clause types (adverbials, nominals, and rel-
atives) were documenting the same underlying construct, one
likely related to syntactic complexity. This result confirmed
that MLCU continues to provide important and unique in-
formation about older students’ oral language skills. Its close
and unique relationship with CD suggests that it is primarily
capturing the advancing use of complex syntax in expository
discourse of preteen and teenage students.
Given the covariance among MLCU, CD, and the
three clause types, it may be tempting to save the extra time
required to code these measures of complex syntax and sim-
ply analyze MLCU. Although MLCU appears to provide
an overall estimate of children’s complex syntax use, it
doesn’t provide a detailed description of the types of clauses
the child is using. By coding for CD and each of the clause
types, a clinician can develop a profile detailed enough to
determine if a student has met the Language Standards of
the Common Core Standards (National Governors Associ-
ation, 2010), which as early as 6th grade call on students to
“vary sentence patterns for meaning, reader/listener inter-
est, and style” in both speaking and writing (p. 52). By
Grades 9–10, students are expected to “demonstrate com-
mand of . . . various types of . . . clauses (independent, depen-
dent; noun, relative, adverbial) to convey specific meanings
and add variety and interest to writing or presentations”
(p. 54). Armed with information about a student’s CD and
clause variety, a clinician can write treatment objectives and
implement intervention programs that target complex syntax
(Nippold, 2014).
Factor 2 had a wide variety of measures that together
appeared to be documenting expository content. Two mea-
sures of sample length (NTW and TCU) had the strongest
loadings onto the second factor. It is not surprising that
286 Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014
the WPM also loaded onto this factor, because it is also a
measure of general productivity; when children have the
capacity to provide longer explanations, they tend to for-
mulate them efficiently. Although we expected the ESS to
form its own distinct language dimension, we can account
for its association with Factor 2. For most of the items on
the ESS, children received a higher score if they provided
greater detail in their productions. When children produced
samples that were lengthy (i.e., a large number of C-units
and words) and fluent (i.e., a high number of WPM), they
were likely as well to produce well-formed and well-organized
explanations. Although these four measures have strong co-
variance, we suggest that clinicians routinely examine the
child’s ESS scores in addition to the productivity measures.
On its face, the ESS provides the most descriptive account
of a child’s overall ability to produce an expository sample.
In completing the ESS rubric for a given sample, an SLP
must critically evaluate its content and organization, fea-
tures that are closely aligned with curricular expectations.
Factor 3 included mazes and errors and omissions,
which together reflected children’s discourse difficulties.
Children’s use of mazes has previously been found to have
limited developmental sensitivity, yet provides a rich de-
scription of children’s discourse difficulties (Leadholm &
Miller, 1992). This factor analysis confirms that children’s
use of mazes is part of a distinct dimension of language
that could be used to describe a child’s profile. Given the
covariance between errors and omissions and mazes, this
result provides further documentation that children’s use
of mazes is indicative of discourse production difficulties.
Although the participants in this study did not have many
errors or omissions in their samples, this measure still helped
to explain a portion of the variance and can be used to help
describe a child’s expository profile. We anticipate that the
Errors and Omissions measure will be even more instru-
mental in describing the expository skills of students with LI
(Scott & Windsor, 2000).
In the final factor, children’s lexical diversity was
predominantly explained by NDW, which was calculated
with a consistent transcript cut (i.e., the first 150 words in
the transcript). In a separate analysis, we calculated NDW
using each child’s entire transcript. In that factor analysis,
which was limited to a three-factor solution, NDW loaded
onto Factor 1. Therefore, if clinicians want to use NDW
as a distinct measure of lexical diversity, they will need to
control for sample length.
Children’s use of relative clauses also had a notable
loading onto Factor 4 in addition to its strong loading on
Factor 1. This reflects the dual function of relative clauses.
They act not only to increase the syntactic complexity of
an utterance; their adjectival status also allows speakers to
embed additional descriptive information into a post-noun
phrase. To do so, speakers must coordinate the “lexicon-
syntax interface” (Berman & Nir, 2010). Examination of
children’s relative clause use thus affords an additional op-
portunity to describe children’s lexical skills.
Together, these data from the factor analysis confirmed
that children must effectively integrate multiple aspects of
the language system to successfully engage in expository dis-
course. Therefore, clinicians and teachers wishing to improve
a student’s expository language skills should take a compre-
hensive view of expository discourse and address multiple
dimensions during instruction. Because these data were com-
piled on typically developing students, they provide strong
empirical support for SLPs consulting with classroom
teachers to help facilitate improved expository discourse in
their classrooms. Although data describing children with
LI would be required to test whether these language dimen-
sions are maintained, the present study provides emerging
evidence for the expository language structure that could be
used in treatment planning for children with LI. Further re-
search will determine the nature of linguistic profiles in chil-
dren with LI, which can be used to describe a student’s
relative strengths and weaknesses and lead to planning indi-
vidualized treatments. These types of descriptive data are
urgently needed for older students. SLPs have a large num-
ber of validated instruments that are effective at making a
differential diagnosis, but require a greater number of valid
and sensitive descriptive assessments to plan for treatments
directly related to the curriculum.
Clinical Implications and Future Directions
For clinicians working with junior high or senior high
school students, descriptive assessments related to the curric-
ulum are some of the most powerful tools in their arsenal.
SLPs are continually required to demonstrate how their
services contribute to their students’ academic success. Rec-
ognizing the power of expository discourse, several investi-
gators have called for the creation of a large database of
samples from typical speakers (Nippold et al., 2007; Scott,
2010). Our research team answered the call for a large data-
base by first developing an elicitation protocol that was
consistent with academic expectations and could be used
to document performance on various requirements in the
Common Core State Standards (National Governor’s Asso-
ciation, 2010). We gave thorough consideration to creating
a protocol that could be administered quickly and consis-
tently by multiple clinicians and that would encourage chil-
dren to engage in high-level expository discourse. Having
such a standardized protocol facilitated the creation of the
database described in this manuscript.
The ultimate goal of creating this large database was
to strengthen required descriptive assessments by allowing
SLPs to compare their individual clients to a large group
of typically developing students. In our previous work, we
have demonstrated that databases of typically developing
speakers can be effective when used to identify children
with SLI (Heilmann, Miller, & Nockerts, 2010). The cur-
rent study provides a critical first step toward fully vali-
dated implementation of expository analysis.
The next step in this line of research is to collect and
analyze language samples from children with LI using the
same FGS protocol. Given that Nippold et al. (2008) and
Scott and Windsor (2000) have documented differences in
expository language measures between children with LI and
their typically developing peers, we anticipate that there will
indeed be differences in the performance of children with
LI when compared to the database. Further research will
provide a better understanding of the value of using a data-
base to assist with clinical decision making when analyzing
children’s expository discourse. This future work will help to
determine which measures are most effective in documenting
the presence of LI, what levels of performance constitute
clinical significance, and which measures are most sensitive
for documenting the profiles of children with LI. Having
data from children with LI may also modify the composi-
tion of the database itself; because children with LI were
not included in the database, there is a chance that the per-
formance of students with LI may be slightly underestimated
(see McFadden, 1996). Until those data are available, SLPs
comparing their clients to the performance of children in
this expository database should interpret their findings with
caution.
Acknowledgments
Special thanks to Jon Miller, Ann Nockerts, Joyelle DiVall-
Rayan, and Karen Andriacchi of SALT Software, LLC, for their
assistance with this project, including provision of transcription
services. Neither author has a financial interest in SALT Software,
LLC. Additional thanks go to Mary-Beth Rolland and Elizabeth
Schoonveld of the Madison (Wisconsin) Metropolitan School District.
We gratefully acknowledge the speech-language pathologists and
students who generously contributed the samples described in this
study; they are affiliated with the following Wisconsin public school
districts: Brown Deer, Fox Point-Bayside, Madison, Shorewood,
Waukesha, Wauwatosa, and West Allis-West Milwaukee.
References
American Speech-Language-Hearing Association. (2004). Preferred
practice patterns for the profession of speech-language pathol-
ogy. Available from www.asha.org/policy
Battle, D. (2011). Communication disorders in multicultural and
international populations (4th ed.). St. Louis, MO: Mosby.
Berman, R., & Nir, B. (2010). The language of expository discourse
across adolescence. In M. Nippold & C. Scott (Eds.), Expository
discourse in children, adolescents, and adults (pp. 99–121).
New York, NY: Psychology Press.
Berman, R., & Verhoeven, L. (2002). Cross-linguistic perspectives
on the development of text-production abilities: Speech and
writing. Written Language and Literacy, 5, 1–43.
Cohen, J. (1988). Statistical power analysis for the behavioral
sciences (2nd ed.). Mahwah, NJ: Erlbaum.
Damico, J., Secord, W., & Wiig, E. (1992). Descriptive language
assessment at school: Characteristics and design. In W. Secord
(Ed.), Best practices in school speech-language pathology
(pp. 1–8). San Antonio, TX: Psychological Corporation.
Ensminger, M., Forrest, C., Riley, A., Kang, M., Green, B.,
Starfield, B., & Ryan, S. (2000). The validity of measures of
socioeconomic status in adolescents. Journal of Adolescent
Research, 15, 392–419.
Evans, J. L., & Craig, H. K. (1992). Language sample collection
and analysis: Interview compared to freeplay assessment con-
texts. Journal of Speech, Language, and Hearing Research, 35,
343–353.
Heilmann & Malone: Properties of Expository Samples 287
Gummersall, D. M., & Strong, C. J. (1999). Assessment of complex
sentence production in a narrative context. Language, Speech,
and Hearing Services in Schools, 30, 152–164.
Guo, L., Tomblin, J. B., & Samelson, V. (2008). Speech disruptions
in the narratives of English-speaking children with specific lan-
guage impairment. Journal of Speech, Language, and Hearing
Research, 51, 722–738.
Halliday, M., & Hasan, R. (1976). Cohesion in English. London,
England: Longman.
Heilmann, J., Miller, J., & Nockerts, A. (2010). Using language
sample databases. Language, Speech, and Hearing Services in
Schools, 41, 84–95.
Heilmann, J., Miller, J. F., Nockerts, A., & Dunaway, C. (2010).
Properties of the narrative scoring scheme using narrative re-
tells in young school-age children. American Journal of Speech-
Language Pathology, 19, 154–166.
Hux, K., Morris-Friehe, M., & Sanger, D. (1993). Language sam-
pling practices: A survey of nine states. Language, Speech, and
Hearing Services in Schools, 24, 84–91.
Individuals with Disabilities Education Improvement Act of 2004.
Pub. L. No. 108446. § 118 Stat. 2647 (2004)
Koutsoftas, A. D., & Gray, S. (2012). Comparison of narrative
and expository writing in students with and without language-
learning disabilities. Language, Speech, and Hearing Services in
Schools, 43, 395–409.
Krippendorff, K. (1980). Content analysis: An introduction to its
methodology. Thousand Oaks, CA: Sage.
Leadholm, B. J., & Miller, J. F. (1992). Language Sample Analysis:
The Wisconsin guide. Madison: Wisconsin Department of
Instruction.
Leonard, L. B. (2014). Children with specific language impairment
(2nd ed.). Cambridge, MA: MIT Press.
Loban, W. (1976). Language development: Kindergarten through
grade twelve. Champaign, IL: National Council of Teachers of
English.
Malone, T., Heilmann, J., Miller, J., DiVall-Rayan, J., & Rollan, M.
(2010, November). Reaching the tweeners: Extending two
SALT databases to Grades 5–6. Paper presented at the annual
meeting of the American Speech-Language-Hearing Association,
Philadelphia, PA.
Malone, T., Miller, J., Andriacchi, K., Heilmann, J., Nockerts, A.,
& Schoonveld, L. (2008, November). Let me explain: Teenage
expository language samples. Paper presented at the annual
meeting of the American Speech-Language-Hearing Association,
Chicago, IL.
McCauley, R. (2001). Assessment of language disorders in children.
Mahwah, NJ: Erlbaum.
McCauley, R., & Swisher, L. (1984). Use and misuse of norm-
referenced tests in clinical assessment: A hypothetical case.
Journal of Speech and Hearing Disorders, 49, 338–348.
McFadden, T. U. (1996). Creating language impairments in typically
achieving children: The pitfalls of “normal” normative sampling.
Language, Speech, and Hearing Services in Schools, 27, 3–9.
Merritt, D. D., & Liles, B. Z. (1989). Narrative analysis: Clinical
applications of story generation and story retelling. Journal of
Speech and Hearing Disorders, 54, 438.
Miller, J. F., & Iglesias, A. (2012). Systematic Analysis of Language
Transcripts. Middleton, WI: SALT.
National Governors Association Center for Best Practices. (2010).
Common Core Standards for English Language Arts & Literacy
in History/Social Studies, Science, and Technical Subjects.
Washington, DC: National Governors Association Center for
Best Practices, Council of Chief State School Officers. http://
www.corestandards.org
288 Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014
Nippold, M. A. (2009). School-age children talk about chess: Does
knowledge drive syntactic complexity? Journal of Speech, Lan-
guage, and Hearing Research, 52, 856–871.
Nippold, M. (2014). Language sampling with adolescents: Implica-
tions for intervention (2nd ed.). San Diego, CA: Plural.
Nippold, M. A., Hesketh, L. J., Duthie, J. K., & Mansfield, T. C.
(2005). Conversational versus expository discourse: A study
of syntactic development in children, adolescents, and adults.
Journal of Speech, Language, and Hearing Research, 48,
1048–1064.
Nippold, M., Mansfield, T., & Billow, J. (2007). Peer conflict ex-
planations in children, adolescents, and adults: Examining the
development of complex syntax. American Journal of Speech-
Language Pathology, 16, 179–188.
Nippold, M., Mansfield, T., Billow, J., & Tomblin, J. B. (2008).
Expository discourse in adolescents with language impairments:
Examining syntactic development. American Journal of Speech-
Language Pathology, 17, 356–366.
Paul, R., & Norbury, C. (2012). Language disorders from infancy
through adolescence. St. Louis, MO: Elsevier.
Petersen, D., Gillam, S., & Gillam, R. (2008). Emerging procedures
in narrative assessment: The Index of Narrative Complexity.
Topics in Language Disorders, 28, 115–130.
Scarborough, H. (2009). Connecting early language and literacy
to later reading (dis)abilities: Evidence, theory, and practice.
In F. Fletcher-Campbell, J. Soler, & G. Reid (Eds.), Ap-
proaching difficulties in literacy development: Assessment,
pedagogy, and programmes (pp. 23–39). Thousand Oaks,
CA: Sage.
Schneider, P., & Dubé, R. V. (2005). Story presentation effects on
children’s retell content. American Journal of Speech-Language
Pathology, 14, 52–60.
Scott, C. (1988). Producing complex sentences. Topics in Language
Disorders, 8, 44–62.
Scott, C. (1995). Syntax for school-age children: A discourse per-
spective. In M. Fey & J. Windsor (Eds.), Language learning
disabilities in school-age children and adolescents: Some under-
lying principles and applications (pp. 49–95). San Diego, CA:
College Hill Press.
Scott, C. (2010). Assessing expository texts produced by school-
age children and adolescents. In M. A. Nippold & C. M. Scott
(Eds.), Expository discourse in children, adolescents, and adults:
Development and disorders (pp. 191–214). New York, NY:
Psychology Press.
Scott, C., & Stokes, S. (1995). Measures of syntax in school-age
children and adolescents. Language, Speech, and Hearing Services
in Schools, 26, 309–319.
Scott, C., & Windsor, J. (2000). General language performance
measures in spoken and written narrative and expository
discourse of school-age children with language impairment.
Journal of Speech, Language, and Hearing Research, 43, 324–339.
Stothard, S. E., Snowling, M. J., Bishop, D. M., Chipchase, B. B.,
& Kaplan, C. A. (1998). Language-impaired preschoolers: A
follow-up into adolescence. Journal of Speech, Language, and
Hearing Research, 41, 407–418.
Strong, C., & Shaver, J. (1991). Stability of cohesion in the spo-
ken narratives of language-impaired and normally developing
school-age children. Journal of Speech and Hearing Research, 34,
95–111.
Sun, L., & Nippold, M. A. (2012). Narrative writing in children and
adolescents: Examining the literate lexicon. Language, Speech,
and Hearing Services in Schools, 43, 2–13.
Tabachnick, B., & Fidell, L. (2013). Using multivariate statistics
(6th ed.). Boston, MA: Allyn & Bacon.
Tomblin, J. B., & Zhang, X. (2006). The dimensionality of language
ability in school-age children. Journal of Speech, Language, and
Hearing Research, 49, 1193–1208.
Verhoeven, L., Aparici, M., Cahana-Amitay, D., van Hell, J., Kriz,
S., & Viguié-Simon, A. (2002). Clause packaging in writing
and speech: A cross-linguistic developmental analysis. Written
Language and Literacy, 5, 135–162.
Watkins, R., Kelly, D., Harbers, H., & Hollis, W. (1995). Measur-
ing children’s lexical diversity: Differentiating typical and
language-impaired learners. Journal of Speech and Hearing
Research, 38, 1349–1355.
Westerveld, M. F., & Gillon, G. T. (2010). Oral narrative context ef-
fects on poor readers’ spoken language performance: Story
retelling, story generation, and personal narratives. Interna-
tional Journal of Speech-Language Pathology, 12, 132–141.
Westerveld, M. F., & Heilmann, J. J. (2012). The effects of geo-
graphic location and picture support on children’s story retell-
ing performance. Asia Pacific Journal of Speech, Language,
and Hearing, 15, 129–143.
Westerveld, M., & Moran, C. (2011). Expository language skills
of young school-age children. Language, Speech, and Hearing
Services in Schools, 42, 182–193.
Wiggins, G., & McTighe, J. (2005). Understanding by design
(2nd ed.). Alexandria, VA: ASCD.
Wisconsin Department of Public Instruction, Student Services/
Prevention and Wellness Team. (2010). Wisconsin
Standards for Physical Education. Retrieved from www.
wellnessandpreventionoffice.org/WI_Standards_for_PE-1.pdf
Worth, S. (Ed.). (1990). Rules of the game. New York, NY:
St. Martin’s Press.

Appendix A
Expository Planning Sheet

Component What’s covered Notes

Object What you have to do to win

Preparations Playing area and setup
Equipment and materials
What players do to get ready
Start How the contest begins, including who goes first
Course of play What happens during a team or player’s turn, including any special plays,
positions, or roles, both offensive and defensive
Rules Major rules, including penalties for violations
Scoring Different ways to score, including point values
Duration How long the contest lasts, including how it ends and tie-breaking procedures
Strategies What smart players do to win, both offensively and defensively

Appendix B
Examiner Script for Eliciting Expository Samples

I’m interested in finding out how well you do at giving explanations. I’m going to make a recording so I can remember
what you say. If you want, you can listen to the recording when we’re finished.
I want you to imagine that I am a student about your age. I’m visiting the United States from another country and I want
to learn as much as I can about life in the U.S. You can help me by explaining how to play your favorite sport or game. You
have lots of choices. For example, you could pick a sport, such as basketball or tennis. You could pick a board game, such
as Monopoly or chess. Or you could pick a card game, such as poker or rummy. What sport or game do you want to pick?
The student offers an appropriate choice. If a choice is not offered or is inappropriate (such as a video game), reread
the examples given above and/or add more examples to aid the student in making an appropriate choice. If the student is still
having difficulty making a selection, suggest picking a game or sport recently played in the student’s physical education class.
Assume that in my country we don’t play [name of sport or game]. I’d like you to explain everything I would need to
know to so I could learn to play. I’ll expect you to talk for at least five minutes. To help you organize your thoughts, here’s a
list of topics I’d like you to talk about [hand the student a copy of the planning sheet]. Please take the next few minutes to
plan your explanation by taking notes in the blank spaces [indicate empty column on the right]. But don’t waste time writing
sentences. Just write some key words to remind you of what you want to say. You can talk about the topics in the order they
are listed, or else you can number the topics any way you wish. If you don’t want to take notes, you can use the backside of
the list to draw a diagram or make a graphic organizer. Do you have any questions?
If student expresses difficulty with reading any portion of the checklist, read the unclear portions aloud. If the student has
difficulty with understand the vocabulary, give an example from a sport or game different from the one the student has chosen.
Go ahead and start planning.
Allow enough time for student to write something for each topic on the checklist or to complete a diagram or graphic
organizer. If the student stops writing or drawing before planning is finished, prompt with, “Please do some planning for [topic
name(s)].”
I’m ready to turn on the recorder. You will be doing all the talking. I’m going to listen to what you have to say. Take as
much time as you need to give a complete explanation. Remember, I expect you to talk for at least five minutes.
Turn on recording device and have the student begin speaking. After the student has finished speaking from his or her
planning sheet, turn off recording device. If the student finishes speaking before five minutes has elapsed, prompt with, “Is
there anything else you can tell me?” Review the recording for quality before releasing the student.

Heilmann & Malone: Properties of Expository Samples 289

Appendix C
Expository Scoring Scheme Rubric

Characteristic Proficient Emerging Minimal/Immature

Object Full description of the main objective Mention of the main objective Mention of winner but no or limited
description how that is determined
OR
Description of another aspect of the
contest, such as strategy or scoring
Preparations 1) Playing area 1) Playing area 1) Playing area
Labels place and provides details Labels place and provides limited Labels place but no details about
about shape & layout details about shape & layout shape & layout
AND/OR OR OR
2) Equipment 2) Equipment 2) Equipment
Labels items and provides detailed Labels items with limited description Labels items with no description
description, including function
AND/OR OR OR
3) Player preparations 3) Player preparations 3) Player preparations
Provides detailed description Provides some description Provides limited description
Start Describes initial situation and how Describes initial situation or how play Limited description of the initial situation
play begins begins, but not both or how play begins
Course of Detailed description of: Some description of: Limited description of:
play A unit of play A unit of play A unit of play
AND/OR OR OR
Major roles Major roles Major roles
AND/OR OR OR
Major plays Major plays Major plays
Rules Clear statement of major rules and, Mentions major rules and, when Minimal or no mention of major rules or
when applicable, consequences applicable, consequences for consequences for violations
for violations violations but without full detail
Scoring Full description of ways to score and Incomplete description of ways to Limited description of ways to score or
point values score and point values point values
Duration Clear description of: Some description of: Limited description of:
How long the contest lasts, including, How long the contests lasts How long the contests lasts
when applicable, the units in which
duration is measured
AND/OR OR OR
How the contest ends How the contest ends How the contest ends
AND/OR OR OR
Tie breaking procedures Tie breaking procedures Tie breaking procedures
Strategy Full description of some ways to win Mention of some ways to win the Vague or incomplete mention of some
the contest that are not required by contest that are not required by ways to win the contest that are not
the rules but are what competent the rules but are what competent required by the rules but are what
players do players do, competent players do
Terminology Terms of art are clearly defined Some terms of art defined, but not Terms of art introduced but not further
whenever introduced consistently or clearly defined
Cohesion Topics follow a logical order Topics follow a logical order Little discernable order to topics;
AND OR Much jumping between topics;
Topics are completely covered before Topics are completely covered AND
moving on to another; before moving on to another
AND OR Abrupt transitions between topics
Smooth transitions between topics Smooth transitions between topics.

Scoring: Each characteristic receives a scaled score 0–5. Proficient characteristics=5, Emerging=3, Minimal/Immature=1. Scores in between
(e.g., 2, 4) are undefined, use judgment. Significant factual errors reduce the score for that topic. Scores of 0, NA are defined below. A com-
posite is scored by adding the total of the characteristic scores. Highest score=50.
A score of 0 is given for student errors (e.g., not covering topic, explaining a different game or sport, not completing/refusing task, student
unintelligibility, abandoned utterances).
A score of NA (nonapplicable) is given for mechanical/examiner/operator errors (e.g., interference from background noise, issues with
recording (cut-offs, interruptions), examiner quitting before student does, examiner not following protocol, examiner asking overly specific or
leading questions rather than open-ended questions or prompts).

290 Language, Speech, and Hearing Services in Schools • Vol. 45 • 277–290 • October 2014
Copyright of Language, Speech & Hearing Services in Schools is the property of American
Speech-Language-Hearing Association and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.