Professional Documents
Culture Documents
Nippold 2005 PDF
Nippold 2005 PDF
Nippold 2005 PDF
L
ater language development is characterized by growth in the abil-
ity to communicate in flexible ways for diverse purposes (Ravid
& Tolchinsky, 2002). This includes being able to use different dis-
course genres, depending on the situation. One important genre is expository
discourse, the use of language to convey information (Bliss, 2002). Highly
valued in academic settings, this occurs, for example, when a high school
student explains the purpose, steps, and outcome of a chemistry experiment or
instructs a group of peers in the use of a telescope for an astronomy meeting.
This contrasts with conversational discourse, a more interactive and less
formal genre (Crystal, 2002). Highly valued in social settings, conver-
sation occurs when a group of high school friends discuss the latest
fashions, movies, or CDs, exchanging opinions as well as sharing
1048 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005 AAmerican Speech-Language-Hearing Association
1092-4388/05/4805-1048
information. Both genres require speakers to tap into words per clause or the number of clauses per T-unit.
their cognitive and linguistic resources as they com- In some studies, the communication unit (C-unit) has
municate their ideas in a clear and organized fashion. been used to measure growth in syntax instead of the
However, expository discourse, with its emphasis on T-unit. These two units are identical with the excep-
speaking in a precise, monologic fashion (Berman & tion that the C-unit includes responses that lack an
Verhoeven, 2002), may heighten these demands, pos- independent clause when answering questions (Loban,
sibly resulting in greater linguistic complexity. In any 1976; for examples, see Appendix A).
case, the ability to communicate in both genres—by Using the C-unit, Loban (1976) investigated syn-
considering the setting, purpose, and needs of the lis- tactic development in the context of conversational
tener and readily adjusting to the situation—is a mark discourse in 211 children who were followed longi-
of speaker competence. tudinally from kindergarten through the 12th grade.
We designed the present study to compare the Every year, each child participated in a conversation
linguistic complexity of conversational versus exposi- with an adult who recorded the child’s language sam-
tory discourse in children, adolescents, and adults and ple for detailed analyses. The results showed that mean
to identify broad developmental patterns. A primary C-unit length increased as the children aged but at a
contributor to linguistic complexity is the speaker’s very slow rate. In a subgroup of students who repre-
use of syntax, the structural foundation of sentences sented a range of language proficiency levels, includ-
(Crystal, 1996). Although little is known about the ing high, low, and average (‘‘Random Group,’’ n = 35),
development of syntax in expository discourse, much Loban reported the following mean C-unit lengths at
has been learned about its development in conversa- Grades 3, 5, 8, and 11, respectively: 7.62, 8.82, 10.71,
tion through longitudinal (e.g., Brown, 1973; Loban, and 11.17 (p. 27). Despite its slow progression, mean
1976) and cross-sectional (e.g., Leadholm & Miller, C-unit or T-unit length is considered to be an impor-
1992; Miller, 1981; Paul, 1981) research. For exam- tant marker of later syntactic development (Hunt,
ple, Miller (1981) has shown that by age 5, most chil- 1970; Loban, 1976; Nippold, 1993; Scott, 1988; Scott
dren are conversing in complex sentences, have a mean & Stokes, 1995; Scott & Windsor, 2000).
utterance length of at least 6.0 morphemes, and make
few grammatical errors. Despite the impressive attain- Another marker of later syntactic development is
ments of a typical 5-year-old child, syntactic develop- clausal density, defined as ‘‘the average number of clauses
ment in conversation continues well into adolescence, (main and subordinate) per T-unit’’ (Scott, 1988, p. 58).
albeit at a more gradual pace than that which occurred Also called the subordination index, it is measured by
previously (Berman, 2004; Loban, 1976; Nippold, 1998). summing the total number of clauses (independent +
As Bates (2003) explained, beyond the preschool years, subordinate) and dividing by the total number of T-units
speakers are ‘‘becoming more fluent and efficient in the or C-units produced in a language sample. This marker
process by which words and grammatical constructions also increases gradually during the school-age and
are accessed in real time, and [are] learning how to use adolescent years. Analyzing data from Loban’s (1976)
the grammar to create larger discourse units (e.g., longitudinal study, Scott (1988) reported the following
writing essays, telling stories, participating in a long mean subordination indexes for Grades 3, 5, 8, and 11,
and complex conversation)’’ (p. 15). Consistent with respectively: 1.22, 1.29, 1.39, and 1.52 (p. 59).
this perspective, Loban (1976) argued that cognitive The appropriate use of subordination can enhance
development and intellectual stimulation are ‘‘far more the efficiency with which ideas are expressed. For
likely to accelerate syntactic growth than grammar example, instead of producing a monotonous string of
knowledge’’ (p. 36) in older children and adolescents. independent clauses (e.g., ‘‘I went to California. I saw a
Two key markers of later syntactic development movie. It was a new movie. Jeremy had recommended
include sentence length and clausal density (Nippold, it’’), a competent speaker can combine them into one
1993; Scott & Stokes, 1995). Beyond the preschool complex sentence that is rich in information yet clear
years, sentence length is often measured with the in meaning (e.g., ‘‘When I went to California, I saw the
terminable unit, or T-unit (Scott, 1988). Hunt (1970) new movie that Jeremy had recommended’’). By age 5,
defined the T-unit as ‘‘one main clause plus any sub- typically developing children are able to use all types
ordinate clause or nonclausal structure that is at- of subordinate clauses in conversation (Paul, 1981),
tached to or embedded in it’’ (p. 4; for examples, see including relative, adverbial, and nominal (for exam-
Appendix A). Examining written language develop- ples, see Appendix A). However, additional refine-
ment in school-age children and adolescents, Hunt ments continue to occur beyond the preschool years.
reported that the mean number of words per T-unit In Loban’s (1976) longitudinal study, all three types
was a more sensitive index of syntactic growth during of subordinate clauses were used more frequently as
this period than other measures, such as the number of the students grew older, thereby contributing to the
1050 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005
adolescent years and that key markers of growth 1996; Scott & Stokes, 1995). For example, academic
include increases in sentence length and clausal den- tasks, such as explaining the solution to a physics
sity, particularly in the use of relative, adverbial, and problem or the steps involved in proving a theorem,
nominal clauses (Loban, 1976; Scott, 1988; Scott & can be expected to challenge a speaker’s linguistic com-
Stokes, 1995). We predicted that similar age-related petence more fully than conversation about simpler
changes would be evident in the expository genre but topics that have been discussed many times before.
that syntactic complexity would be greater in exposi- Information on how a child or adolescent performs in
tory than in conversational discourse because of the this arena would be helpful in identifying subtle yet
increased communicative demands the former seems to troublesome difficulties in using a type of discourse
place on the speaker. We also predicted that expository that is highly valued in school. Gillam, Pena, and
discourse would have a lengthy developmental time Miller (1999) discussed the role of expository discourse
course, perhaps even longer than that for conversa- as a language skill that contributes to academic
tional discourse. For this reason, it was important to success. As such, they recommended that speech-
examine development beyond adolescence and well language pathologists assess the syntactic complexity
into adulthood. Little is known about language develop- of expository discourse in upper elementary and high
ment in adults, because this has not been widely school students. However, in the absence of a norma-
investigated. However, Hunt (1970) reported that tive database, it is impossible to know what levels of
mean length of the T-unit continued to increase beyond complexity are reasonable to expect at different ages.
adolescence and into early adulthood in written In summary, in the present study we addressed
language, and Berman and Verhoeven (2002) reported the following questions:
similar findings with respect to spoken language.
1. How does syntactic development differ in conver-
However, no studies of syntactic development have
sational versus expository discourse?
included middle-aged adults. In the present study, two
groups of adults—one younger (age 20–29 years) and 2. What markers of syntactic growth exist for con-
one older (40–49 years)—were included. The findings versational versus expository discourse?
were expected to contribute to the knowledge base in 3. What changes occur in syntactic development
later language development, a topic of expanding in- during early and middle adulthood?
ternational interest (Berman, 2004).
4. What levels of performance can be expected of
The present study also represents an initial children, adolescents, and adults?
attempt to build a normative database in expository
discourse, one that could be used by speech-language
pathologists in their work with school-age children and
adolescents suspected of having language disorders. Method
When evaluating these young people, speech-language
pathologists frequently administer standardized lan- Participants
guage tests and engage the students in conversation A total of 120 individuals participated in this
for the purpose of examining oral language develop- investigation, including two groups of children, two
ment (Evans & Miller, 1999). Often, the conversational groups of adolescents, and two groups of adults. With
sample will be transcribed and analyzed using a 20 participants in each group, the mean ages (and
computer program such as Systematic Analysis of ranges) were as follows (in years;months): 8;1 (7;8–
Language Transcripts (SALT; Miller & Chapman, 8;7), 11;4 (10;9–11;8), 13;9 (13;1–14;4), 17;3 (16;9–
2003), which calculates key syntactic, lexical, and 17;11), 25;6 (20;8–28;5), and 44;8 (40;8–49;9). The
morphological variables, including mean utterance children and adolescents were public school students
length, total number of words and utterances pro- who were attending an elementary school (Grade 2 or
duced, and the use of bound morphemes. A reference 5), middle school (Grade 8), or high school (Grade 11)
database is available through SALT for comparing a located in a low- to lower middle–income neighborhood
speaker’s conversational performance with that of age- in western Oregon. According to their teachers, all stu-
matched, typically developing peers. dents were considered to be typical achievers who were
Although this database can be helpful in docu- free of deficits in language, learning, and cognition. All
menting oral language deficits and planning interven- were of the appropriate age for their grade, and none
tion, there are cases where school-age children and were receiving special education services. The adults were
adolescents perform adequately on standardized lan- residents of western Oregon who were individually re-
guage tests and exhibit normal conversational skills, cruited by the graduate students who conducted the in-
yet demonstrate difficulties with more complex speak- terviews. Often, the adults were acquaintances of the
ing tasks in natural settings (Nippold & Schwarz, students or were referred to them by friends or relatives,
1052 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005
clauses, the guidelines shown in Appendix A were used. for differences in sample length. SAS also determined
To simplify the coding process, only finite clauses the subordination index, a measure of clausal den-
(both independent and subordinate) were identified. sity. This consisted of the total number of clauses
As defined by Quirk and Greenbaum (1973), a finite (independent + RC + AVC + NOM) divided by the to-
clause ‘‘always contains a subject as well as a predi- tal number of T-units in a sample (Scott, 1988). Again,
cate’’ (p. 310). Thus, in the present study, all nonfinite in determining the total number of clauses, only fi-
clauses (those that did not contain a subject) were nite clauses were counted (see Appendix A).
ignored. After coding a sample, the same investigator
double-checked all codes and made any necessary cor-
rections. A third investigator performed a final check
of all codes and flagged any disagreements with the
Results
initial codings. For each type of clause, the percent- Both tasks were effective in eliciting discourse
age of disagreement between the two investigators from all age groups. Regarding the favorite game or
was as follows: independent clauses = 1%; RCs = 1%; sport task, a wide range of activities was discussed,
AVCs= 1%; and NOMs = 6%. All disagreements were including games such as poker, chess, and Clue, and
resolved through discussion so that 100% agreement sports such as basketball, football, baseball, track and
was attained for all clause types. In addition to the field, cross-country running, swimming, wrestling,
independent and subordinate clauses, SALT counted hockey, tennis, soccer, golf, and rowing. For each of
the total number of T-units and calculated the mean the six age groups, the following variables are reported
length of T-unit in words for each sample. The data in Tables 1 and 2, respectively, for conversational and
then were entered into the SAS system (SAS Institute, expository discourse: total number of T-units, mean
2001), which determined the percentage of T-units length of T-unit in words, percentage of T-units con-
that contained each type of subordinate clause. Percent- taining RCs, percentage of T-units containing AVCs,
ages rather than raw numbers were used to control percentage of T-units containing NOMs, and clausal
Total T-units
M 33.95 25.45 38.25 61.50 50.80 60.10
SD 13.31 17.20 20.98 31.54 23.88 34.45
Range 9–62 4–68 6–84 19–140 6–96 17–151
Mean length of T-unit
M 6.74 7.31 6.88 8.33 9.86 9.56
SD 0.86 1.62 0.93 1.27 2.09 2.01
Range 4.42–8.44 3.67–10.56 5.56–8.52 5.83–10.32 6.00–13.44 6.88–15.16
Relative clause usea
M 3.42 5.00 4.43 8.29 10.20 8.68
SD 3.61 6.41 3.78 4.64 8.12 8.22
Range 0–11.11 0–22.22 0–12.12 0–20.93 0–22.92 0–36.84
Adverbial clause usea
M 7.28 10.84 5.50 10.99 10.43 12.04
SD 4.67 9.37 4.77 4.63 7.73 6.19
Range 0–16.67 0–33.33 0–16.98 0–18.60 0–22.58 0–25.00
Nominal clause usea
M 7.04 9.42 6.58 11.05 18.13 17.71
SD 7.00 10.86 6.03 6.01 11.99 10.14
Range 0–23.33 0–43.75 0–20.75 0–22.73 0–53.16 5.56–38.33
Clausal density
M 1.18 1.25 1.17 1.30 1.39 1.38
SD 0.09 0.22 0.09 0.11 0.21 0.19
Range 1.00–1.37 1.00–1.75 1–1.42 1.08–1.56 1.08–1.82 1.12–1.77
a
Reported as percentage of T-units per sample.
Total T-units
M 33.05 35.30 36.15 44.00 51.55 60.55
SD 25.36 17.40 19.12 27.28 31.95 23.30
Range 10–97 16–77 9–88 12–113 9–128 11–109
Mean length of T-unit
M 8.59 9.29 8.68 10.59 11.04 11.46
SD 1.89 0.90 1.64 1.60 1.39 1.11
Range 6.07–13.62 8.04–12.05 6.53–14.05 7.58–13.65 8.21–13.24 10.08–15.17
Relative clause usea
M 6.50 6.14 5.20 11.27 11.60 14.44
SD 6.91 4.84 5.10 6.29 6.20 8.28
Range 0–21.43 0–19.61 0–18.18 3.70–30.43 0–22.97 0–29.55
Adverbial clause usea
M 25.62 22.61 23.50 27.22 22.51 25.60
SD 19.49 10.58 13.97 18.12 9.80 10.77
Range 0–76.19 6.25–41.18 3.33–54.55 4.35–79.49 7.59–47.06 10.20–48.94
Nominal clause usea
M 10.24 16.67 13.11 17.85 19.79 19.06
SD 8.78 7.68 12.37 17.62 9.89 6.99
Range 0–33.33 2.70–33.33 0–44.44 3.85–83.33 0–39.36 8.77–33.03
Clausal density
M 1.42 1.45 1.42 1.56 1.54 1.59
SD 0.23 0.16 0.20 0.28 0.15 0.14
Range 1.00–1.90 1.26–1.75 1.12–2.05 1.24–2.33 1.24–1.89 1.30–1.89
a
Reported as percentage of T-units per sample.
density. To assist the reader in visualizing these data, and mean length of T-unit in words, raw scores served
they are also presented in Figures 1 through 6. as the dependent measure; however, for each type of
To examine the effects of age and discourse genre, subordinate clause, raw percentage scores served as
a repeated measures analysis of variance was per- the dependent measure. Bonferroni corrections were
formed on each variable. For total number of T-units used for multiple comparisons (adjusted a = .008). Ef-
fect sizes were computed using the eta coefficient
(h; Meline & Schmitt, 1997) and were interpreted as
Figure 1. Language output (total T-units produced) for
conversational and expository discourse for each age group.
Figure 2. Mean length of T-unit (in words) for conversational and
expository discourse for each age group.
1054 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005
Figure 3. Relative clause use for conversational and expository Figure 5. Nominal clause use for conversational and expository
discourse for each age group. discourse for each age group.
Con Exp Con Exp Con Exp Con Exp Con Exp Con Exp
Mean length of T-unit 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
Relative clause use .46* .41 .48* .08 .21 .57** .61** .14 .67** .26 .81**** .18
Adverbial clause use .59** .79**** .72*** .40 .40 .56** .85**** .58** .65** .20 .36 .27
Nominal clause use .02 .07 .60** .56** .24 .47* .10 .46* .45* –.07 .49* .36
Clausal density .48* .83**** .75**** .57** .49* .83**** .67** .69*** .75*** .19 .75*** .49*
for all groups. As shown in Figure 2, the growth For clausal density, a statistically significant main
rate for mean T-unit length was nearly identical in effect was obtained for group, F(5, 114) = 7.78, p G
both genres. .0001, h = .50, and genre, F(1, 114) = 100.07, p G .0001,
For relative clauses, statistically significant main h = .68. The effect size was large for both group and
effects were obtained for group, F(5, 114) = 9.00, p G genre. The interaction between group and genre was
.0001, h = .53, and genre, F(1, 114) = 11.28, p = .0011, not statistically significant, F(5, 114) = 0.60, p > .05, h =
h = .30. The effect size was large for group and medium .16. Nevertheless, Tukey tests indicated that for
for genre. The interaction between group and genre conversation, the 44- and 25-year-old adults outper-
was not statistically significant, F(5, 114) = 1.02, p > formed the 13- and 8-year-old children, but for ex-
.05, h = .21. Tukey tests indicated that for conversa- planation, the groups did not differ. Again, this result
tion, the 25-year-old adults outperformed the 13- and may stem from the conservative nature of the Tukey
8-year-old children. For explanation, the 44- and test (Winer, 1971). Thus, although clausal density
25-year-old adults outperformed the 13-year-old chil- showed an age-related increase only in conversation,
dren. Thus, in addition to an age-related increase in it was greater in explanation than in conversation for
RC production, all groups showed greater use in ex- all groups.
planation than in conversation. Because mean length of T-unit has long been
For AVCs, a statistically significant main effect regarded as a key marker of syntactic development,
was obtained for genre, F(1, 114) = 120.69, p G .0001, and one that reflects the use of subordination (Hunt,
h = .72, but not for group, F(5, 114) = 0.87, p > .05, h = 1970), we decided it was important to examine the
.19. The effect size for genre was large. The interaction relationship of T-unit length to the different types of
between group and genre was not statistically signifi- subordinate clauses that were examined in this study.
cant, F(5, 114) = 0.76, p > .05, h = .18. For every group, For each genre, Pearson product–moment correlation
AVCs were used more often in explanation than in con- coefficients were calculated for each age group sepa-
versation, but there were no age-related increases in use. rately, using each participant’s mean length of T-unit
For NOMs, a statistically significant main effect and raw percentage score on each type of subordinate
was obtained for group, F(5, 114) = 6.57, p G .0001, h = clause and the clausal density score. As shown in Table 3,
.47, and genre, F(1, 114) = 12.62, p = .0006, h = .32. The many of the coefficients were statistically significant,
effect size was large for group and medium for genre. particularly for AVC production and for clausal
The interaction between group and genre was not density. This indicates that mean length of T-unit
statistically significant, F(5, 114) = 0.77, p > .05, is a good predictor of the use of subordination in both
h = .18. Tukey tests indicated that for conversation, conversational and expository discourse genres.
the 44- and 25-year-old adults outperformed the Given the usefulness of the T-unit as a general
13-, 11-, and 8-year-old children; however, for explana- index of syntactic proficiency, independent t tests
tion, the groups did not differ. The failure to find sig- comparing the performance of males and females were
nificant group differences for explanation may be due conducted for both genres. The results were not sta-
to the conservative nature of the Tukey test, which tistically significant for either genre: conversation,
tightly guards against Type I error (Winer, 1971). t(118) = –0.28, p = .7795, and explanation, t(118) =
Thus, the findings indicated an age-related increase –0.57, p = .5727. These findings are consistent with past
in NOM production only in conversation and that all research in expository discourse (Berman & Verhoeven,
groups used NOMs more often in explanation. 2002) that has indicated no gender differences.
1056 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005
Finally, it is enlightening to examine what the It was interesting that AVCs, used by all six
participants actually said when speaking in the con- groups more frequently in explanation than in con-
versational and expository genres. Appendix B con- versation, did not show any age-related increases. It is
tains portions of transcripts recorded from speakers notable that even the youngest children in the study
representing each of the six age groups. Each type of used this type of clause as often as the adults, re-
clause is indicated immediately after each main verb. gardless of genre. Similarly, the youngest group used
Mean length of T-unit obtained by the speaker, on the NOMs as often as the adults when speaking in the
basis of the complete sample, is shown. Differences in expository genre. Thus, the expository discourse task
utterance length and syntactic complexity can be ob- appeared to stimulate even the youngest children to
served between genres and, to a lesser degree, across employ these types of clauses when explaining their
age groups. It is important to note that these exam- favorite game or sport.
ples illustrate how the favorite game or sport task Clausal density, recognized as a key marker of
appears to stimulate the speakers to use a level of syntactic development in conversation (Scott, 1988;
syntactic complexity that is not apparent from their Scott & Stokes, 1995), also proved sensitive to growth,
conversational discourse. but not in expository discourse. Although unexpected,
this finding seems to have resulted from the fact that
all six age groups used all three types of subordinate
clauses to a greater extent in expository than in
Discussion conversational discourse. Thus, it cannot be assumed
The purpose of this cross-sectional study was that key markers of syntactic development are identi-
to examine syntactic development in conversational cal for both genres. It is important that this lack of
and expository discourse over a wide age range (7– sensitivity to growth in the expository genre not be
49 years). The main objective was to identify and com- interpreted as a negative result. On the contrary, it is
pare key markers of syntactic growth across participant positive in showing that even young, school-age
ages and discourse genres and to determine if growth children are able to use subordination as frequently
continued beyond adolescence and into early and as middle-aged adults, but young children require a
middle adulthood. We expected that syntactic com- task that is cognitively challenging to reveal their
plexity would be greater in expository than in con- syntactic competence. It is possible that the favorite
versational discourse but that growth would be game or sport task is especially effective in eliciting
evident in both genres. subordinate clauses as speakers explain the special
conditions of their chosen activity (e.g., ‘‘If your
The findings indicated that for both genres, syntax
opponent hits the ball out of bounds, you get a point’’).
continued to develop beyond adolescence and into early
For this reason, it cannot be assumed that other types
adulthood (20–29 years) and remained stable into
of expository tasks would yield the same outcome. For
middle age (40–49 years). Although we predicted that
example, Verhoeven et al. (2002) reported an age-
expository discourse might undergo a longer develop-
mental time course than conversational discourse, related increase in the use of subordinate clauses when
there was no evidence for continued growth beyond 10-year-old children and young adults were compared
early adulthood in either genre. The two best indica- in their descriptions of interpersonal conflicts. Thus,
tors of growth were mean length of T-unit and RC we emphasize that the results of the present study
production. In contrast to the other measures (i.e., the pertain to one type of expository discourse, explaining
use of AVCs, NOMs, and clausal density), these two the rules and strategies of a favorite game or sport. It is
measures showed age-related increases into adulthood possible that different results might be obtained with
in both genres, and the effect sizes for group were other types of expository tasks (e.g., summarizing the
large. Berman and Verhoeven (2002) and Verhoeven outcome of an election campaign; comparing and
et al. (2002) also reported that mean length of T-unit contrasting two brands of candy, clothing, or soda;
and RC production were particularly sensitive to de- explaining why babies cry). Hence, it will be important
velopmental gains in expository discourse in their in future research to examine syntactic development
cross-linguistic study. In the present study, another by comparing different types of expository discourse
variable that was sensitive to age-related growth was tasks.
the total number of T-units produced, a measure of As predicted, expository discourse elicited greater
language output. In general, older speakers talked syntactic complexity than conversational. For each of
more than younger ones regardless of genre. NOM the six age groups, performance in the expository
production also increased in relation to chronological genre exceeded performance in the conversational
age but in conversational discourse only. genre on each of the relevant variables: production of
1058 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005
syntactic complexity as a function of increasing chro- We emphasize the importance of recruiting larger
nological age. Indeed, despite the statistically signifi- numbers of participants in future research, given some
cant group effects, speakers within every age group of the findings of the present study. For example, it is
demonstrated considerable variability in performance. surprising that the 13-year-old adolescents appeared
For example, inspection of the ranges and means to perform below the 11-year-old children, especially
reported in Table 2 indicates that there were 11-year- in mean T-unit length and the use of subordination
old children whose expository performance (e.g., mean (see Figures 2–6). Although not statistically significant,
length of T-unit) resembled that of the average 25- this slight drop in performance is counter to what has
year-old adult, just as there were speakers in their 20s been observed in past research. For example, Loban
and 40s whose performance (e.g., RC use) resembled (1976) found gradual increases in mean C-unit length
that of the average 8- or 11-year-old child. Similar and subordination when comparing Grades 5 and 8,
patterns can be observed in all age groups. Because and other investigators have reported that other
speakers of any age appear to differ widely in the use of aspects of language (e.g., semantics) undergo rapid
complex syntax during the years between middle growth as children make the transition into adoles-
childhood and middle adulthood, the means reported cence (e.g., Nippold & Haq, 1996; Nippold, Hegel,
in Tables 1 and 2 should be thought of as estimates Sohlberg, & Schwarz, 1999; Nippold & Rudzinski,
rather than as standards or norms. This is particularly 1993). Larger sample sizes would help to determine
true considering the means are based on fairly small the validity of this unexpected drop.
sample sizes and represent the performance of English- Developmental studies of expository discourse are
speaking children, adolescents, and adults living in just emerging. It is clear that many additional inves-
western Oregon. In addition, it is unknown how these tigations remain to be conducted in this arena, par-
speakers might perform on other types of expository ticularly in building a database that can be used by
tasks. speech-language pathologists to assess the develop-
Nonetheless, it is notable that this pattern of wide ment of discourse in children, adolescents, and adults.
individual differences within groups has been observed Studies that attempt to account for the wide individual
in early childhood as well. Bates and her colleagues differences observed within age groups will be most
(Bates, 2003; Bates, Dale, & Thal, 1995) reported strik- informative, particularly when they include speakers
ing examples of this in their studies of the grammatical who exhibit patterns of delayed, average, and even
development of healthy, middle-class toddlers. As they superior language development.
emphasized, these large individual differences pose
serious difficulties for the notion of identifying deficits
in the language development of young children. As the
Acknowledgments
present study indicates, large individual differences
pose similar challenges for identifying deficits in older This project was partially supported by Grant
children, adolescents, and adults. To solve this prob- 2P50DC02746-06A1 from the National Institute on Deafness
lem, it will be necessary to recruit large numbers of par- and Other Communication Disorders and a Summer Faculty
ticipants for each age group (e.g., 100+) so that valid Research Award from the University of Oregon, awarded
to Marilyn A. Nippold.
percentiles (e.g., 10th, 25th, 50th, 75th, 90th) and stan-
We express sincere gratitude to the children,
dard scores on each measure can be established. It will
adolescents, and adults who participated in this research
be important also to recruit participants from diverse project and to the teachers and administrators who granted
cultural, linguistic, and socioeconomic backgrounds permission for the testing to take place at their schools.
and to build those differences into the normative The assistance of Communication Disorders and Sciences
database. Then, to account for individual differences, graduate students in collecting and transcribing the
it will be necessary to examine expository discourse in language samples is also greatly appreciated.
conjunction with other cognitive and linguistic skills
such as verbal reasoning, spoken and written language
comprehension, and academic achievement. It will be References
interesting also to perform more fine-grained analyses
of the data. This could include an examination of non- Bates, E. (2003). On the nature and nurture of language.
finite clauses, noun- and verb-phrase elaboration, and In R. Levi-Montalcini, D. Baltimore, R. Dulbecco, F. Jacob,
E. Bizzi, P. Calissano, & V. Volterra (Eds.), Frontiers of
later developing infinitive, gerundive, and participial biology: The brain of Homo sapiens (pp. 241–265). Rome:
phrases. The interface between syntactic and lexical de- Istituto della Enciclopedia Italiana fondata da Giovanni
velopment could also be investigated by charting the use Trecanni.
of later developing adverbial conjuncts (e.g., moreover, Bates, E., Dale, P. S., & Thal, D. (1995). Individual
similarly) to achieve cohesion across clauses. differences and their implications for theories of language
Crews, F. (1977). The Random House handbook (2nd ed.). Quirk, R., & Greenbaum, S. (1973). A concise grammar of
New York: Random House. contemporary English. New York: Harcourt Brace.
Crystal, D. (1996). Rediscover grammar (2nd ed.). Essex, Ravid, D., & Tolchinsky, L. (2002). Developing linguistic
England: Longman. literacy: A comprehensive model. Journal of Child
Language, 29, 417–447.
Crystal, D. (2002). The English language: A guided tour of
the language (2nd ed.). London: Penguin Books. SAS Institute. (2001). SAS system [Computer software].
Cary, NC: Author.
Evans, J. L., & Miller, J. (1999). Language sample analysis
in the 21st century. Seminars in Speech and Language, Scott, C. M. (1988). Spoken and written syntax. In M. A.
20, 101–116. Nippold (Ed.), Later language development: Ages nine
through nineteen (pp. 49–95). Austin, TX: Pro-Ed.
Gillam, R. B., Pena, E. D., & Miller, L. (1999). Dynamic
assessment of narrative and expository discourse. Topics Scott, C. M. (1994). A discourse continuum for school-age
in Language Disorders, 20, 33–47. students. In G. P. Wallach & K. G. Butler (Eds.), Language
learning disabilities in school-age children and adolescents:
Hunt, K. W. (1970). Syntactic maturity in school children Some principles and applications (pp. 219–252). New York:
and adults. Monographs of the Society for Research in Child Macmillan.
Development, 35((Serial No. 134, No. 1), ).
Scott, C. M., & Stokes, S. (1995). Measures of syntax in
Leadholm, B. J., & Miller, J. F. (1992). Language sample school-age children and adolescents. Language, Speech,
analysis: The Wisconsin guide. Madison: Wisconsin and Hearing Services in Schools, 26, 301–319.
Department of Public Instruction.
Scott, C. M., & Windsor, J. (2000). General language
Loban, W. (1976). Language development: Kindergarten performance measures in spoken and written narrative and
through grade twelve. Urbana, IL: National Council of expository discourse of school-age children with language
Teachers of English. learning disabilities. Journal of Speech, Language, and
Meline, T., & Schmitt, J. F. (1997). Case studies for Hearing Research, 43, 324–339.
evaluating statistical significance in group designs. Verhoeven, L., Aparici, M., Cahana-Amitay, D., van
American Journal of Speech-Language Pathology, Hell, J., Kriz, S., & Viguie-Simon, A. (2002). Clause
6(1), 33–41. packaging in writing and speech: A cross-linguistic
Miller, J. F. (1981). Assessing language production in developmental analysis. Written Language and Literacy,
children: Experimental procedures. Baltimore: 5, 135–162.
University Park Press. Winer, B. J. (1971). Statistical principles in experimental
Miller, J. F., & Chapman, R. S. (2003). SALT: Systematic design (2nd ed.). New York: McGraw-Hill.
Analysis of Language Transcripts [Computer software].
Madison: University of Wisconsin—Madison, Waisman
Research Center, Language Analysis Laboratory. Received May 25, 2004
1060 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005
Appendix A. Definitions and examples of T-units, C-units, fragments, and clauses.
T-Unit
A T-unit contains one independent (main) clause and any dependent (subordinate) clauses or nonclausal structures that are attached to it or
embedded within it (Hunt, 1970). For example, the utterance ‘‘Bill bought a new bicycle before he went to Europe’’ is one T-unit that contains an
independent clause (‘‘Bill bought a new bicycle’’) and a dependent clause (‘‘before he went to Europe’’). In contrast, the utterance ‘‘Bill went to
France and then he went to Italy’’ consists of two T-units because it contains two independent clauses joined by the coordinating conjunction ‘‘and.’’
Whenever a coordinating conjunction (e.g., ‘‘and,’’ ‘‘but,’’ ‘‘so’’) initiates an independent clause, that clause is considered to be a new T-unit.
C-Unit
A C-unit is identical to a T-unit but includes responses that lack an independent clause when answering a question (Loban, 1976). For example,
the response ‘‘yes’’ to the question ‘‘Did Jack drive?’’ is one C-unit.
Fragment
A fragment is an utterance that lacks a main verb and/or a subject; therefore, it is not an independent clause (Crews, 1977). It does not answer a
question. For example, the following utterances are fragments: ‘‘going down the road,’’ ‘‘the other day,’’ ‘‘2 weeks later.’’
Independent (Main) Clause
An independent clause contains a subject and a main verb and makes a complete statement (Crews, 1977). For example, the following are both
independent clauses: ‘‘Mother rode her bicycle to work today,’’ and ‘‘It started to rain late last night.’’
Dependent (Subordinate) Clauses
A dependent clause contains a subject and a main verb but does not make a complete statement; therefore, it cannot stand alone. There are three
main types of dependent clauses: relative, adverbial, and nominal (Crews, 1977; Quirk & Greenbaum, 1973):
1. A relative clause (i.e., adjective clause) acts like an adjective and modifies the noun that precedes it: for example, ‘‘The cat that was sleeping
on the couch was content.’’
2. An adverbial clause acts like an adverb and modifies a verb. It often describes a condition or cause and begins with a subordinate conjunction:
for example, ‘‘Unless we can reach Los Angeles by eight o’clock, we’ll miss the concert.’’
3. A nominal clause is a noun-like element that can serve as either the subject of a sentence (e.g., ‘‘Whatever she told you about the wedding
was a great exaggeration’’) or its object (e.g., ‘‘I told her what she needed to hear’’). Nominal clauses often begin with wh-words: For example,
‘‘I never know where I should park ’’; ‘‘My desire to become a nurse is why I study so hard ’’; ‘‘Checkmate is when your opponent’s king
cannot escape.’’
Each type of clause is coded immediately after the main verb. If a clause contains more than one main verb, the clause code occurs after the first main
verb. MLTU is based on the complete sample.
1062 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005
Appendix B (p. 2 of 3). Portions of transcripts recorded from children, adolescents, and adults, speaking in
conversational and expository genres (MLTU = mean length of T-unit, IC = independent clause, RC = relative clause,
AVC = adverbial clause, NOM = nominal clause, FRG = fragment).
1064 Journal of Speech, Language, and Hearing Research Vol. 48 1048–1064 October 2005