Content uploaded by Li Sheng
Author content
All content in this area was uploaded by Li Sheng on Mar 12, 2018
Content may be subject to copyright.
JSLHR
Research Article
A Narrative Evaluation of Mandarin-Speaking
Children With Language Impairment
Ying Hao,
a,b,c
Li Sheng,
a,b,d
Yiwen Zhang,
a,b
Fan Jiang,
a,b
Jill de Villiers,
e,f
Wendy Lee,
f,g
and Xueman Lucy Liu
f,g
Purpose: We aimed to study narrative skills in Mandarin-
speaking children with language impairment (LI) to compare
with children with LI speaking Indo-European languages.
Method: Eighteen Mandarin-speaking children with LI
(mean age 6;2 [years;months]) and 18 typically developing
(TD) age controls told 3 stories elicited using the Mandarin
Expressive Narrative Test (de Villiers & Liu, 2014). We
compared macrostructure-evaluating descriptions of
characters, settings, initiating events, internal responses,
plans, actions, and consequences. We also studied general
microstructure, including productivity, lexical diversity,
syntactic complexity, and grammaticality. In addition, we
compared the use of 6 fine-grained microstructure elements
that evaluate particular Mandarin linguistic features.
Results: Children with LI exhibited weaknesses in
5 macrostructure elements, lexical diversity, syntactic
complexity, and 3 Mandarin-specific, fine-grained
microstructure elements. Children with LI and TD
controls demonstrated comparable performance on
2 macrostructure elements, productivity, g ramm atic ality,
and the remaining 3 fine-grained microstructure
features.
Conclusions: Similarities and differences are noted in
narrative profiles of children with LI who speak Mandarin
versus those who speak Indo-European languages.
The results are consistent with the view that profiles
of linguistic deficits are shaped by the ambient language.
Clinical implications are discussed.
Considerable research regarding language impair-
ment (LI) has been conducted with children
speaking Indo-European languages, for example,
English (e.g., Fey, Catts, Proctor-Williams, Tomblin, &
Zhang, 2004), Swedish (e.g., Reuterskiöld, Hansson, &
Sahlén, 2011), and Greek (e.g., Tsimpli, Peristeri, & Andreou,
2016). Compared to Indo-European languages, Chinese is
a typologically distinct language. Grammatical features in
Chinese are typically expressed by separate words, whereas
in many Indo-European languages, they are expressed by
affixes. The ambient language that a child with LI is exposed
to greatly shapes areas of ease and difficulty (Leonard,
2014b). More than a quarter of the world’s population speak
Chinese (Yip & Matthews, 2010). As one of the most im-
portant Chinese languages, Mandarin is not well studied in
the realm of LI (Leonard, 2014a). To equip Chinese speech-
language pathologists with more accurate assessments
and effective interventions, we first need to understand
the linguistic weaknesses of Mandarin-speaking children
with LI.
Narrative production incorporates various language
components into a complete story. Narrative evaluation
thus provides a rich description of children’s expressive lan-
guage. Assessing narrative production is informative in
understanding language manifestations of children with
LI. Using narrative analysis, previous studies have revealed
multiple deficits in children with LI who speak Indo-European
languages, including effects on the organization of com-
plete and coherent stories, and the diversity and complex-
ity of linguistic expressions (e.g., de Villiers, 2004; Reilly,
Losh, Bellugi, & Wulfeck, 2004; Tsimpli et al., 2016).
To the best of our knowledge, only three studies
(Cheung, 2009; Tsai & Chang, 2008; Zhang, 2009) evalu-
ated narrative production in Mandarin-speaking children
with LI. All of them had very small sample sizes (n≤6).
a
Shanghai Children’s Medical Center Affiliated with the Medical
School of Shanghai Jiao Tong University, China
b
MOE-Shanghai Key Laboratory of Children’s Environmental
Health, China
c
The University of Texas at Austin
d
University of Delaware, Newark
e
Smith College, Northampton, MA
f
Bethel Hearing and Speaking Training Center, Farmers Branch,
Corinth, TX
g
The University of Texas at Dallas
Correspondence to Xueman Lucy Liu: lucy.liu@ourbethel.com
Editor-in-Chief: Sean Redmond
Editor: Janna Oetting
Received September 18, 2016
Revision received April 17, 2017
Accepted August 28, 2017
https://doi.org/10.1044/2017_JSLHR-L-16-0367
Disclosure: The narrative evaluation is under development as a test and the authors
have financial interests in future royalties.
Journal of Speech, Language, and Hearing Research •1–15 •Copyright © 2017 The Authors
This work is licensed under a Creative Commons Attribution 4.0 International License. 1
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Though these studies revealed a number of difficulties in
Mandarin-speaking children with LI, most of these find-
ings have not been replicated. The current study intended
to replicate and extend previous findings and identify weak-
nesses in Mandarin-speaking children with LI using a
larger sample size.
Narrative evaluation typically comprises macro- and
microstructure analyses (Gagarina, Klop, Tsimpli, &
Walters, 2016; Justice, Bowles, Pence, & Gosse, 2010).
Macrostructure measures the global organization of a
story, including descriptions of characters, settings, initial
events, characters’internal states and plans, actions, and
consequences (Gillam, Gillam, Fargo, Olszewski, & Segura,
2017). Microstructure, however, takes into account fine-
grained language-internal properties, such as the usage of
different syntactic structures and specific types of words
(Gorman, Bingham, Fiestas, & Terry, 2016; Justice et al.,
2010). Microstructure evaluation also includes more gen-
eral measures providing information about overall oral
language productivity and complexity. For example, total
number of utterances (TNU) and total number of words
(TNW) reflect story length (Muñoz, Gillam, Peña, & Gulley-
Faehnle, 2003). The number of different words (NDW)
indicates lexical diversity (Muñoz et al., 2003). Mean length
of utterance (MLU) is related to stages of syntactic devel-
opment (Brown, 1973) and reflects grammatical complex-
ity (Scarborough, Rescorla, Tager-Flusberg, Fowler, &
Sudhalter, 1991). These measures can be generated by lan-
guage analysis software, such as the Systematic Analysis of
Language Transcript (Miller & Chapman, 2002) and Com-
puterized Language Analysis (MacWhinney, 2000). Com-
pared with fine-grained microstructure measures, these
measures provide more general indices of children’s linguis-
tic skills.
On the basis of previous literature, we categorized
children’s narrative skills into three components: macro-
structure, general microstructure, and fine-grained micro-
structure. In the following, we will review findings from
Indo-European languages under the three components.
Then, we will review findings from Mandarin-speaking
children with LI and Cantonese-speaking children with LI.
Together, these bodies of literature informed our selection
of narrative measures.
Narrative Skills in Indo-European Languages
Macrostructure
Across different Indo-European languages, children
with LI have demonstrated lower performance in macro-
structure components of story settings, topic maintenance,
problem resolution, event sequencing, and internal states
(e.g., Miranda, Mccable, & Bliss, 1998; Squires et al., 2014;
Tsimpli et al., 2016). The usage of evaluative devices is a
particularly vulnerable domain. Children with LI have dif-
ficulties using words to express diverse emotions, such as
mental states (e.g., happy), attention-seeking exclamations
(e.g., look!), and affective behaviors (e.g., He was crying;
Reilly et al., 2004). Both the deficits in language and theory
of mind (i.e., taking others’perspectives into account) may
contribute to the low performance on internal state descrip-
tions in children with LI (de Villiers, 2007). On the one
hand, understanding others’intentions and emotions is a
prerequisite for expressing internal responses; on the other
hand, improvement in language helps children to verbally
organize observations of internal responses.
Children with LI do not always perform more poorly
on all narrative macrostructure elements. In Norbury and
Bishop (2003), no difference was found between the English-
speaking, typically developing (TD) group and the group
with LI (6 to 10 years old) in the inclusion of initiating
events, attempts, and resolutions. While the two groups
achieved high scores in initiating events and attempts,
resolutions seemed to be challenging for both the TD
group and the group with LI. The authors claimed that
the ability to include resolutions developed late in both typ-
ical and atypical children. However, other later-acquired
elements, such as internal responses and plans, may have
provided a more complete picture of children’s narrative per-
formance (Heilmann, Miller, Nockerts, & Dunaway, 2010).
Heilmann et al. also pointed out that a sensitive macro-
structure evaluation should measure both quantity and
quality (i.e., whether children’s production is related to the
primary storyline) of macrostructure elements. The exami-
nation of both aspects should exhibit better sensitivity be-
cause a scoring rubric of this type is more stringent, hence
better equipped to identify areas of weaknesses in children’s
narrative skills.
General Microstructure
General microstructure measures have been exten-
sively studied in previous research. MLU and the usage of
complex sentences reflect syntactic complexity (Scarborough
et al., 1991), and both have been found to be indicators dis-
tinguishing children with and without LI across different
languages (e.g., English: Fey et al., 2004; Rice et al., 2010;
Greek: Tsimpli et al., 2016; Spanish: Restrepo, 1998). Rice
et al. (2010) found that both MLU in words and MLU in
morphemes were reliable indices that identified English-
speaking children with LI in a wide age range, from 3 to
9 years old. English-speaking children with LI displayed
lower usage of complex sentences (Reilly et al., 2004), which
were defined as containing two verb phrases, with one
coordinated or subordinated to the other (e.g., The boy
WANTS to OPEN the box).
Diversity measures (i.e., NDW) are more likely to
reveal language deficits than productivity measures (i.e.,
TNW and TNU). In English and Greek, children with LI
produced fewer NDW than children without LI (Hewitt,
Hammer, Yont, & Tomblin, 2005; Rezzonico et al., 2015;
Tsimpli et al., 2016). The two productivity measures, TNW
and TNU, do not seem to differentiate children with and
without LI. Children with LI were as productive as TD
age controls in TNW (Guo, Tomblin, & Samelson, 2008;
Hick, Joseph, Conti-Ramsden, Serratrice, & Faragher, 2002)
and total number of morphemes (Norbury & Bishop, 2003).
Similarly, the two groups of children were comparable in
2Journal of Speech, Language, and Hearing Research •1–15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
TNU (Cleave, Girolametto, Chen, & Johnson, 2010; Fey
et al., 2004; Guo et al., 2008; Norbury & Bishop, 2003).
Compared with the diversity measure (i.e., NDW), TNW
and TNU reflect verbosity. Language deficits may be more
likely to manifest in the quality (reflected by diversity mea-
sures) than in the quantity (reflected by productivity mea-
sures) of language.
Grammaticality has also been frequently studied. In
English, children with LI made more grammatical errors
than TD controls, such as omitting and misusing verb
auxiliaries, determiners, tense markers, and subject–verb
agreement (Leonard, 2014a). Reilly et al. (2004) calculated
the ratio of morphological errors to the total number of
prepositions from English-speaking children’s narrative
production. They found a sharp contrast between the TD
group and the group with LI in percentages of errors:
nearly 50% for the group with LI and less than 10% for
the TD group. Fey et al. (2004) found that English-speaking
children with LI produced a lower percentage of gram-
matical communication-units (C-units) than TD children
in oral narrative production. While TD fourth-grade chil-
dren achieved grammatical accuracy of 84%, the accuracy
of children with LI was 75%. In Swedish, a significant dif-
ference was also found between TD children and children
with LI in the percentage of grammatically correct C-units
(Reuterskiöld et al., 2011).
Fine-Grained Microstructure
In recent years, a sizable body of work has included
fine-grained microstructure measures in narrative analysis.
Justice et al. (2010) proposed a fine-grained microstructure
assessment protocol, including 18 English properties rang-
ing from sentence structure, phrase structure, modifiers,
and nouns to verbs. The authors reported good construct
validity and criterion-related validity of their assessment
protocol in 262 TD English-speaking children. The results
suggested that assessing fine-grained microstructure ele-
ments was potentially effective in monitoring child lan-
guage development and detecting children with LI. By
adapting the English protocol into Spanish, Gorman et al.
(2016) created a Spanish microstructure assessment pro-
tocol. Its usage in 67 preschool children also indicated
good construct and concurrent and predictive validity.
Reuterskiöld et al. (2011) compared Swedish-speaking
children with and without LI on several fine-grained micro-
structure elements. Differences were found in the diversity
of verbs. However, aside from the handful of studies, fine-
grained microstructure measures have not been widely used
to evaluate children with LI.
To summarize, in Indo-European languages, macro-
structure and general microstructure measures (lexical
diversity, syntactic complexity, and grammaticality) can
capture the differences between children with LI and with-
out LI, whereas verbosity measures may not. The emerg-
ing literature on fine-grained microstructure measures
suggest that these measures are potentially differentiating
between children with LI and without LI.
Narrative Skills in Mandarin
Mandarin and Indo-European languages are typo-
logically distinct languages. One prominent difference lies
in the area of morphosyntax. Indo-European languages
typically have richer morphosyntactic transformations,
whereas Mandarin lacks these changes. We take the com-
parison between Mandarin and English as an illustration.
While English uses “-s”to mark plurality, Mandarin does
not have a similar plural form. A bare noun can be singu-
lar or plural according to the context. While English marks
subject–verb agreement, Mandarin does not. English sen-
tences should agree in person and in number (e.g., I am,
you are, he is), whereas Mandarin does not have similar
rules. Given these Mandarin-specific features, we cannot
assume that all measures borrowed from Indo-European
languages will be differentiating in Mandarin. The likeli-
hood of finding grammatical errors may be smaller in a
narrative context in which children have much flexibility
in choosing language expressions. In the following, we
will review the three Mandarin narrative studies (Cheung,
2009; Tsai & Chang, 2008; Zhang, 2009) and see what dif-
ferentiating measures have been found in Mandarin.
Zhang (2009) compared two 4-year-old children with
LI and their age matches. Children with LI needed the
examiner to ask questions to elicit production and continue
storytelling. The group with LI produced fewer macro-
structure elements (e.g., actions, settings) and exhibited
lower usage of various expressions of internal state than the
TD matches. Likewise, Tsai and Chang (2008) observed
deficits in macrostructure in six Mandarin-speaking chil-
dren with LI at 8 to 9 years old when compared to age
matches. In addition, they found that general microstructure
measures revealed group differences, including word pro-
ductivity (i.e., TNW), lexical diversity (i.e., NDW), and
syntactic complexity (i.e., MLU; Tsai & Chang, 2008). The
two groups demonstrated comparable performance on sen-
tence productivity (i.e., TNU). This is different from find-
ings in many Indo-European language studies that did not
find a difference in either productivity measure (TNW and
TNU; e.g., Guo et al., 2008; Norbury & Bishop, 2003).
These findings from Tsai and Chang (2008) need to be rep-
licated using a larger sample size. In addition, fine-grained
microstructure evaluation examining the usage of specific
linguistic properties was not conducted.
Cheung (2009) focused on general and fine-grained
microstructure elements in an 18-month longitudinal study
of two 7-year-old Mandarin-speaking children with LI.
ThechildrenwithLIwerepairedinMLUwithtwocon-
trols who were 2.5-year-olds. Children with LI showed
more usage of various complex sentences than their MLU
matches. Regarding the two fine-grained microstructure
elements (i.e., classifiers and aspect markers), children with
LI did not outperform the TD controls who were almost
5 years younger. This suggests that the two microstructure
measures are likely to be delayed in Mandarin-speaking
children with LI in comparison to TD age matches. In the
following, we will describe the two Mandarin features and
children’s acquisition of the two features.
Hao et al.: Narratives of Mandarin-Speaking Children With LI 3
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Mandarin is extremely rich in the use of nominal
classifiers. The number of classifiers in Hanyu Liangci
Cidian (A Dictionary of Chinese Classifiers) reached 902
(Zhang, 2007). A nominal classifier is obligatory and must
appear between a number/demonstrative/quantifier and
a noun. The classifier typically indicates the property of
the entity the noun is referring to (Erbaugh, 2006; Li &
Thompson, 1981). In Example (1), “gen”is inserted be-
tween the number (two) and the noun (needle) indicating
a thin, slender, and stick-like object.
(1)
liǎng gēnzhēn
two classifier-gen needle
two needles
Erbaugh (2006) pointed out that children acquire
classifiers slowly as they expand their vocabulary. It takes
time for children to sort different semantic categories in-
grained in different classifiers. It also takes time for chil-
dren to learn the arbitrary usage of some classifiers. For
children with LI, the process of sorting semantic categories
and acquiring arbitrary usage may be longer.
Mandarin uses grammatical morphemes to mark
aspect. Instead of showing the absolute time of a situation,
aspect markers express different perspectives on the situa-
tion (Li & Thompson, 1981). Generally, aspect markers
can be categorized into perfective morphemes (–le, –guo)
and progressive morphemes (zai–,–zhe)
1
(Klein, Li, &
Hendriks, 2000). Perfective morphemes show that an ac-
tion is completed or a status is formed, whereas progres-
sive morphemes indicate that the action is in progress. See
Example (2) for the usage of perfective marker “le”and
Example (3) for the usage of progressive marker “zhe.”
In Example (2), the action “mai”(buy) is completed from
the speaker’s viewpoint, whereas in Example (3), the action
“chi”(eat) is in progress.
(2)
Xiǎonǚhaí maǐle yí gè dàngāo.
Little girl buy perfective-aspect one classifier-ge cake.
The little girl bought a cake.
(3)
Xiǎonǚhaí chīzhe yí gè dàngāo.
Little girl eat progressive-aspect one classifier-ge cake.
The little girl is eating a cake.
Aspect markers can be optional, and the omission of
aspect markers does not render a sentence ungrammatical
in Cantonese (Fletcher, Leonard, Stokes, & Wong, 2005).
Similar to Cantonese, Mandarin also allows the optional-
ity of aspect markers. However, in both languages, there
are contexts in which aspect markers are preferred if not
obligatory (Fletcher et al., 2005; Klein et al., 2000). It is
important to note that older Mandarin-speaking children
produced more aspect markers to specify perspectives of
time in a narrative setting than younger children (Shu,
2004).
Other Potential Indicators of LI in Mandarin
In addition to a narrative analysis, a growing body
of literature investigated the acquisition of Mandarin lin-
guistic features using cloze tasks. In this task, children are
asked to fill in words or phrases that are missing where
the context mandates them to use the target features. In a
study by He and Dai (2012), twelve 5-year-old, Mandarin-
speaking children with LI were less accurate in using
negations than TD age matches. They either did not use
any negation form in an obligatory context or used a wrong
negative form that was inappropriate in the context. In
Zeng, Li, Li, and He (2013), the same group of children
with LI displayed less production of “ba”sentences. The
“ba”structure is a Mandarin active sentence structure
with a noncanonical word order (“agent + ba + patient +
action”). The noncanonical word order may cause more
difficulty for children with LI. See Example (4) for an ex-
ample of “ba”structure.
(4)
Tābǎdàngāochīwán le.
She ba cake eat resultative-finish perfective-aspect.
She finished eating the cake.
Findings from Cantonese-speaking children with LI
motivate the investigation of additional fine-grained micro-
structure elements. The passive structure, also involving a
noncanonical word order, challenges Cantonese-speaking
children with LI. Compared with TD age matches, chil-
dren with LI demonstrated lower percentages of full and
grammatical passive sentences in contexts encouraging the
production of passive structure (Leonard, Wong, Deevy,
Stokes, & Fletcher, 2006). Mandarin passive structure is
similar to Cantonese passive structure. A passive structure
follows the noncanonical word order of “patient + bei +
agent + action.”See Example (5) for Mandarin passive
structure.
(5)
Tābeì gǒu zhuīle.
He bei dog chase perfective-aspect.
He was chased by a dog.
In summary, evidence from previous Mandarin or
Cantonese studies suggests that Mandarin-speaking children
with LI may demonstrate difficulties in the above six fine-
grained microstructure elements, including classifiers, per-
fective and progressive aspect markers, negation, active
“ba”structure, and passive “bei”structure. It should be
noted that except for aspect markers, all features are oblig-
atory if children chose to use the corresponding structures.
Though aspect markers are optional, they are more likely
1
To be more precise, “le”is the perfective marker, “guo”is the
experiential marker, “zai”is the progressive marker, and “zhe”is the
durative marker (Li & Thompson, 1981). However, in general, both
“le”and “guo”can be categorized as perfective markers, and both
“zai”and “zhe”can be categorized as progressive markers (Klein
et al., 2000).
4Journal of Speech, Language, and Hearing Research •1–15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
to be present in language production from more mature
Mandarin speakers (Shu, 2004). In addition to the fine-grained
microstructure measures, findings of macrostructure and gen-
eral microstructure deficits in Mandarin-speaking children
with LI need to be replicated. Grammaticality has not been
investigated in previous studies in Mandarin. The lack of
morphosyntactic features in Mandarin leads us to question
whether grammaticality can reliably differentiate Mandarin-
speaking children with and without LI in a narrative context.
Research Goals and Predictions
In this study, we intended to identify linguistic defi-
cits in Mandarin-speaking children with LI using narrative
analysis. We evaluated narrative production in a larger
sample of Mandarin-speaking children with LI (n=18)
and their age-matched controls (n= 18). On the basis of
a review of narrative evaluation in Indo-European lan-
guages and Mandarin/Cantonese, our narrative analyses
included three components: macrostructure, general micro-
structure (i.e., TNU, TNW, NDW, MLU, usage of com-
plex sentences, and grammaticality), and fine-grained
microstructure (i.e., classifiers, perfective and progressive
aspect markers, negation, “ba”structure, and “bei”structure).
We expected to see weaknesses in Mandarin-speaking
children with LI in macrostructure, as previous narra-
tive findings were consistent in showing the deficit (e.g.,
de Villiers, 2004; Tsai & Chang, 2008). Mandarin-speaking
children with LI would also demonstrate lower perfor-
mance in lexical diversity (i.e., NDW) and syntactic com-
plexity (i.e., MLU and proportion of complex sentences),
as shown in previous narrative studies across different lan-
guages (e.g., English: Fey et al., 2004; Mandarin: Tsai &
Chang, 2008; Cantonese: Klee, Stokes, Wong, Fletcher, &
Gavin, 2004; Greek: Tsimpli et al., 2016). Verbosity (i.e.,
TNU and TNW) would be comparable between children
with LI and without LI on the basis of previous literature
(e.g., Norbury & Bishop, 2003). Based on our analysis of
Mandarin-specific characteristics (the sparsity of morpho-
syntactic morphemes), grammaticality would also be
comparable between the two groups. Informed by existing
literature in Mandarin-speaking or Cantonese-speaking
children with LI (Cheung, 2009; He & Dai, 2012; Leonard
et al., 2006; Zeng et al., 2013; Zhang, 2009), the six fine-
grained microstructure elements hold the most promise
to differentiate the group with LI and the TD group. We
therefore predicted to see differences between the TD
group and the group with LI on the six measures.
Method
Participants
Participants in the current study were drawn from a
larger sample who participated in the validation study (Liu
et al., 2017) of the Diagnostic Receptive and Expressive
Assessment of Mandarin (DREAM; Ning, Liu, & de Villiers,
2014). The children were recruited from the Developmental
and Behavioral Pediatrics Clinic at Shanghai Children’s
Medical Center. All families with children between 2;6
(years;months) and 7;11 who visited the hospital during Oc-
tober 2014 were invited to participate. A total of 300 chil-
dren were initially recruited, but 70 of them were excluded
because they did not meet the following criteria: normal
hearing, normal or corrected-to-normal vision, and no
diagnosis of autism, neurological disorder, genetic disorder,
intellectual disability, or cerebral palsy. Children were
administered a battery of tests, including but not limited to
the DREAM, the Primary Test of Nonverbal Intelligence
(PTONI; Ehrler & McGhee, 2008), and either a spontane-
ous language (for children between 2;6 and 3;11) or a nar-
rative sample (for children between 4;0 and 7;11). Of the
230 children, 99 (age range: 4;0–7;11) produced narratives
elicited by the experimental version of the Mandarin Ex-
pressive Narrative Test (MENT; de Villiers & Liu, 2014).
2
We used a set of criteria to qualify children with LI
in this study: (a) children had at least one component stan-
dard score at or below 80 (1.3 SD below the mean or the
10th percentile) on the DREAM; (b) children had diffi-
culty with oral language according to pediatrician’s judg-
ment; and (c) children had a nonverbal IQ score that is
70 (2 SDs below the mean) or above on the PTONI.
3
The DREAM is a standardized oral language test
for children ages 2;6 to 7;11. It was normed on a nation-
ally representative sample of 969 children in Mainland
China. The test provides four component scores: receptive
language, expressive language, syntax, and semantics. As
reported by Liu et al. (2017), test–retest reliability was high
(r= .85). External validity was assessed using spontaneous
language indices (e.g., sentence complexity, vocabulary
diversity) and narrative indices (e.g., the inclusion of men-
tal verbs and connectives).
4
The correlation coefficient
between DREAM scores and spontaneous language was
.7 (p< .001) and that between DREAM and narrative
indices was .45 ( p< .01).
We used a cutoff score of 80 on any of the DREAM
components as one of our inclusionary criteria. According
to Liu et al. (2017), the DREAM demonstrated the best
sensitivity (73%) and specificity (82%) using this criterion
(< 80 on one component) when measured against a priori
2
MENT © 2014 by Bethel Hearing and Speaking Training Center Inc.
3
Different IQ cutoff scores have been adopted by researchers (Gallinat
& Spaulding, 2014). In our sample, five children with LI scored from
70 to 79. To see whether lower IQ scores changed our findings, we
conducted comparisons on all dependent variables between the reduced
sample (n= 13) of children with LI and their age matches (all children’s
IQ scores were above 80). The results from the reduced sample were
the same as those from the unreduced sample. Therefore, we included
children whose IQ scores were between 70 and 79 to maintain a larger
and a more representative sample.
4
Note that for the current study, we only included narrative samples
from the older children, not spontaneous samples from the younger
children. The measures used in the validation of the DREAM are
different from the measures we selected in the current study. From Liu
et al. (2017), they measured the usage of mental verbs and connectives,
but we focused on other narrative measures.
Hao et al.: Narratives of Mandarin-Speaking Children With LI 5
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
judgment by pediatricians, who usually make decisions of
this kind in China. The sensitivity of the DREAM improved
to 95% when spontaneous language sample (either sponta-
neous or narrative) performance and a priori judgment
were combined as the gold standard. Specificity remained
the same under the second approach.
Multiple researchers have advocated for the inclusion
of ecologically valid indicators of LI in addition to a statis-
tically driven definition. Bishop (1997) suggested the use
of a real-life concern expressed by those who are familiar
with the child. Bedore, Peña, Joyner, and Macken (2011)
and Peña, Resendiz, and Gillam (2007) reported that par-
ent, teacher, and clinician judgments provided reliable
indicators of language ability. In light of these recommen-
dations, we used the judgment of a developmental and
behavioral pediatrician as a converging indicator of LI.
Upon enrollment in the study, the caregiver of each child
was asked if they had any concerns about the child’s oral
language ability. If the answer was affirmative, the pediatri-
cian asked follow-up questions in an informal interview
(e.g., the child’s ability to ask and answer questions and
understand and follow directions, vocabulary, word retrieval,
sentence formulation skills, and conservational and peer
interaction skills). Based on the caregivers’answers, the
pediatrician formed a clinical impression.
For this study, the 18 children (age range: 4;3–7;11)
met all three criteria and were categorized as having LI. For
each child in the group with LI, we selected a TD child
who was within 5 months of age as a match (age range:
4;3–7;9). The TD controls achieved 80 or higher on all four
components of the DREAM, and there were no caregiver
concerns regarding language development. Children with
LI achieved lower performance than the TD children in
the four components and the total scores of the DREAM
(see Table 1). Children with LI in our sample more com-
monly displayed deficits in expressive language (17/18) and
in the domain of syntax (10/18).
5
In addition to age, we made an effort to match the
group with LI and the TD group on primary caregivers’
education and children’s nonverbal IQ scores. Education
was ranked into five levels ranging from 1 to 5. Five rep-
resented the highest education level (PhD/master’s), and
1 represented the lowest (middle school or lower). Nonver-
bal IQ was measured using the Chinese translation of the
PTONI (Ehrler & McGhee, 2008). The test included a
set of pictures for each testing item, and children were
required to point to a picture that was different from the
others. The PTONI was normed in a culturally diverse
sample, with 1,010 children from 38 states in the United
States. Though it was not normed in Mainland China, it
was considered to be largely free of cultural bias. The group
differences in caregiver education and PTONI scores were
not significant. Though not significant, the plevels were
close to .10. To address these potential confounds, we
examined the correlations between dependent variables
and education level/IQ scores before conducting group
comparisons. If they were significantly correlated, we would
enter them as covariates.
Materials and Procedure
Three stories were depicted on three wordless picture
series from the experimental version of the MENT. Each
story had at least two characters, a complication in the
events that encouraged talk of problems and resolutions,
visual presentation of characters’emotions, and depiction
of multiple actions. For example, one story contained two
characters. One character sees some food placed in one
location and leaves. The other character moves the food
to a new place out of sight and leaves. When the first char-
acter returns, he or she tries to look for the food. Thought
balloons were depicted to demonstrate characters’desires
and thoughts and to elicit production of internal responses
and plans. For instance, when the first character decides
to move the food, a balloon shows that the move is meant
to keep the food fresh. When the other character comes
back, there is a thought balloon showing that he or she is
thinking of the food. In order to tell this story, children
need to take the characters’perspective into account and
describe their internal states. Detailed information about
the MENT can be found in Appendix C of Liu et al. (2017).
An examiner presented pictures to the children one
by one in sequence. After viewing all pictures, the exam-
iner asked the children to tell the story with pictorial sup-
port. If a child did not follow the sequence in which the
pictures were presented, the examiner corrected the child
and required the child to follow the sequence. All stories
were video-recorded. A linguistics undergraduate who is a
native Mandarin speaker transcribed the recordings into
Chinese characters. The undergraduate was not aware of
the LI/TD status of the participants. The first author,
also a native Mandarin speaker, proofread all transcrip-
tions and corrected transcription errors. The first author
then coded the samples using the Computerized Language
Analysis software program (MacWhinney, 2000). At the
time of coding, the first author was also blind to the group
membership.
Coding
Macrostructure
On the basis of Monitoring Indicators of Scholarly
Language (Gillam et al., 2017), we coded seven macro-
structure elements, including character, setting, initiating
event, internal response, plan, action series, and conse-
quence. Each element was scored on a scale from 0 to 3.
As in the Monitoring Indicators of Scholarly Language,
we used a combination of quantitative and qualitative
analyses. We took into account the frequency of occurrence
of each macrostructure element, as well as the connection
of the production to the main storyline. For example, for
the element action series, 0 means that no action is taken
5
More specifically, eight children with LI were below 80 in only the
expressive component; five were below 80 in all components; four
were below 80 in expressive and syntax components; and one was
below 80 in all but the expressive component.
6Journal of Speech, Language, and Hearing Research •1–15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
by any character, 1 is including actions that are taken by
characters but not related to the story, 2 is including at least
a series of actions that is related to the main story but does
not form a complete episode, and 3 means that the action
series relates to the main story and forms a complete episode.
Table 2 presents macrostructure scoring rubric and examples.
General Microstructure Measures
We segmented sentences on the basis of C-unit (Loban,
1976) that consists of a main clause with all dependent
clauses attached to it. For example, “The little boy said ‘I
don’t know’” is a C-unit. It has a subordinate clause “I
don’t know”embedded in the main clause “The little boy
said.”“He found a ladder. He went there.”consists of
two C-units and should be counted as two utterances. The
second sentence does not attach to the first, so both are
independent clauses. Sentence segmentation formed the
basis of the TNU measure. Word segmentation was at the
word level instead of the morpheme level, due to the fact
that Mandarin grammatical features are realized at the word
level. Word segmentation formed the basis of the TNW
and NDW measures.
We coded complex sentences (C-units) and calculated
percentages of complex sentences (number of complex sen-
tences divided by TNU). A sentence with two or more than
two main verbs was regarded as a complex sentence. Two
verbs can have an embedding relation in which one clause
is embedded in another clause. In Example (6), the clause
with the verb “zou”(walked away/left) is embedded in the
clause with the verb “kan”(see). Two verbs can also have a
compounding relation. In Example (7), “huishou”(wave)
and “shuo”(say) are compounded showing a sequence of
actions. This structure is a serial verb construction (no con-
nective between the two verbs; Cheung, 2009).
(6)
Tākàn dào nánhaí zǒu le.
He see verb-resultative boy leave perfective-marker
He saw that the boy left.
(7)
Tāhuīshǒu shuōzaìjiàn.
He wave hand say goodbye
He waved and said goodbye.
In addition, we coded ungrammatical sentences
(C-units) in which ungrammaticality could arise from
any component of a sentence. Example (8) includes a loca-
tive phrase in which the preposition is missing. In addi-
tion, the sentence has an incomplete noun phrase in which
the head noun is missing. We calculated percentages of
ungrammaticality (number of ungrammatical sentences
divided by TNU).
(8)
Bàba shuōyào (zaì) wūdǐng shàng huà yí gè hóngsè (?)
Father say need (preposition “zai”missing) house
roof above paint one classifier red (?).
Father said that (they) need to paint a red (?) (on)
the roof of the house.
Fine-Grained Microstructure Measures
Based upon a review of Mandarin grammar (Li &
Thompson, 1981) and studies of Mandarin or Cantonese
microstructure and grammar acquisition (Cheung, 2009;
He & Dai, 2012; Leonard et al., 2006; Zeng et al., 2013;
Zhang, 2009), we counted the frequency of the “ba”struc-
ture, “bei”structure, negation, classifier, perfective aspect
marker “le/guo,”and progressive aspect marker “zai/zhe.”
All coded occurrences were accurate usage of these features
(inaccurate usage was included in the measure of gram-
maticality above). See Table 3 for descriptions and exam-
ples of each element.
Reliability
Following the same rubric, another research assistant
independently transcribed 20% of the narrative production
that were randomly selected from all narratives. Regarding
macrostructure, Cohen’sk(a statistic measuring interrater
Table 1. Participant information by typicality.
Measure
LI (n= 18) TD (n= 18)
tpvalueMSD MSD
Age (months;years) 6;2 1;2 6;2 1;1 0.14 .894
Gender: male/female 14/4 10/8
Education 2.33 1.14 3.00 1.28 1.65 .108
IQ 90.44 16.69 98.89 12.31 1.73 .094
DREAM–Total 83.61 9.77 100.61 8.47 5.85 <.001
DREAM–Receptive 84.65 10.57 100.72 9.08 4.53 <.001
DREAM–Expressive 72.44 7.64 98.89 9.88 7.79 <.001
DREAM–Syntax 79.33 7.67 94.06 6.70 6.08 <.001
DREAM–Semantics 87.50 13.00 106.89 11.85 4.99 <.001
Note. Education is presented in scales ranging from 1 to 5: 5 = PhD/master’s degree,4=bachelor’s
degree,3=associate degree,2=high school,1=middle school or lower. The IQ scores and DREAM
scores are standard scores. Independent-samples ttests were conducted to compare the two groups.
LI = language impairment; TD = typically developing; DREAM = Diagnostic Receptive and Expressive
Assessment of Mandarin.
Hao et al.: Narratives of Mandarin-Speaking Children With LI 7
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
agreement for qualitative data) indicated that agreement
between the two coders was substantial (k= 0.633, p< .0001;
Landis & Koch, 1977). Sentence and word segmentation
formed the basis for the general microstructure measures
of MLU, NDW, TNW, and TNU. Following Rezzonico
et al. (2015), reliability of sentence and word segmentation
was computed using the formula: number of agreements /
(number of agreements + number of disagreements). The
reliability was 89% for sentence segmentations (total: 263,
agreement: 234, disagreement: 29) and 93% for word segmen-
tations (total: 1,719, agreement: 1,605, disagreement: 114).
All disagreements were resolved by discussion. For the
other two general microstructure elements (i.e., complex
sentences, ungrammatical sentences) and the six fine-
grained microstructure elements, we followed Justice et al.
(2010): Agreement occurred when two coders scored within
one occurrence difference. For example, if one rater counted
the occurrence of an element four times and the other counted
it five times, their scorings were considered consistent. The
interrater consistency for these measures ranged from 91%
to 100%.
Results
We first conducted Pearson’scorrelationstoseewhether
education level and IQ scores were related to the depen-
dent variables. No significant correlations were found be-
tween education levels/IQ scores and any of the dependent
variables. Therefore, we did not enter education level and
IQ scores as covariates.
In the following, we conducted 18 group compari-
sons (all measures except proportions of complex sentences
and ungrammatical sentences) using independent-samples
ttests. To protect against the inflation of Type I error rate,
a Benjamini–Hochberg procedure (Benjamini & Hochberg,
1995) was conducted for the 18 comparisons. The Benjamini–
Hochberg procedure is less conservative than the Bonferroni
correction that may mask important clinical findings by
Table 2. Macrostructure scoring rubric and examples.
Macrostructure
elements
Score
01 2 3
Character No character is included or
only pronouns are used.
e.g., he, she
Includes characters
that are not related
to the storyline.
Includes one of the
two characters.
Includes both characters.
e.g., girl, mother
Setting No time or place is stated. Includes at least one
time or place that
is not related to
the storyline.
Includes one time or one
place that is related to
the storyline.
e.g., 1 day, in the fridge
Includes at least two times or
two places that are related
to the storyline.
Initiating
event (IE)
A problem or starting
event is not stated.
Includes events or
problems that are
not related to the
storyline.
Includes one event or
problem that is related
to the primary storyline.
e.g., The girl wants
to eat it.
Includes more than one event
or problem that is related
to the primary storyline.
Internal
response
No feelings, desires, or
thoughts are stated.
Includes expressions of
feelings, desires, or
thoughts that are not
related to the IEs.
Includes two feelings, desires,
or thoughts that are directly
related to the IEs.
e.g. , The mother thought
that it should be put into
fridge.
Includes more than two feelings,
desires, or thoughts that are
directly related to the IEs.
Plan No statement is provided
about the character’s
plan to solve the
problem.
Includes one statement
of plan that is not
related to the IEs.
Includes one statement
of plan that is directly
related to the IEs.
e.g., She is going to eat it.
Includes at least two statements
of plan that are directly
related to the IEs.
Action series No action is taken by
any of the characters.
There are a series of
random descriptions.
e.g., A girl. Another girl.
Actions are taken by
the characters but
are not directly related
to the IEs.
e.g., They are boiling
eggs. (no such action
depicted)
At least one action that is
directly related to the
IEs is taken by the main
character. However, the
action series do not form
a complete episode.
e.g., She gets the cake.
Includes a series of actions
that are directly related to
the IEs and can be formed
into a complete episode.
e.g., She is afraid the cake
may melt. She gets the cake
and moves it into the fridge.
Consequences No consequence is stated. There are statements of
consequences, but
they are not related
to the IEs.
Includes one consequence
that is directly related
to the IEs.
e.g., She puts the cake
into the fridge.
Includes at least two
consequences that are
directly related to the IEs.
Note. The examples are from one of the three stories. The rubric varies for different stories on the basis of the storyline and observations of
children’s overall performance.
8Journal of Speech, Language, and Hearing Research •1–15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
increasing Type II error (false negative; Glickman, Rao, &
Schultz, 2014; Perneger, 1998). This procedure controls the
false discovery rate (d), the expected proportion of errors
among all the rejected hypotheses, by ordering pvalues
from smallest to largest. In the Benjamini–Hochberg pro-
cedure, smaller pvalues are retained based on the number
of comparisons. If a pvalue is equal or less than d×(i/n)
(iis the rank of the pvalues, and nis the number of com-
parisons), the test is significant. We set the false discovery
rate at 0.05, a cutoff often used (Glickman et al., 2014).
At a false discovery rate of 0.05, a difference was sig-
nificant at p< .025. We also calculated Cohen’sdeffect
size (mean differences between the two groups divided by
pooled standard deviations) to reveal clinically important
discoveries. A Cohen’sdvalue of 0.8 is considered a
large effect; 0.5, a medium effect; and 0.2 a small effect
(Cohen, 1988).
Independent-samples ttests may not be appropriate
for percentage comparisons. For proportions of complex
sentences (number of complex sentences divided by TNU)
and ungrammatical sentences (number of ungrammatical
sentences divided by TNU), we used generalized linear
models with a binomial family and log link that compared
the two groups on the likelihood of using complex sentences
and ungrammatical sentences. In the models, odds ratios
were used to report effect estimates. The Benjamini–Hochberg
procedure was also implemented to reduce Type I error for
the two comparisons, and the false discovery rate was also
setat0.05.
Macrostructure
Independent-samples ttests were used to compare
the two groups on macrostructure elements. Table 4 pre-
sents the results. The two groups significantly differed on
macrostructure total scores. Regarding individual elements,
the group with LI displayed lower performance than the
TD group on character, setting, internal response, action
series, and consequence. Large effect sizes for the above dif-
ferences by group were found. Particularly, children with
LI displayed very little usage of internal responses, and the
averaged score of 0.07 indicated that there were almost no
statements about characters’internal states. No significant
difference was found for the element of plan, with both
groups showing very little production of the plan (both
groups’averaged scores were below 0.5). Between-groups
difference on initiating event approached significance with
a medium effect size.
General Microstructure Measures
We compared children’s performance on general micro-
structure measures, including MLU, NDW, TNW, and
TNU. MLU was calculated using the formula: MLU =
TNW/TNU. Independent-samples ttests were conducted
Table 3. Fine-grained microstructure definitions and examples.
Fine-grained microstructure
elements Definitions Examples
Ba Mandarin active sentence structure
with a word order “agent + ba +
patient + action”(“ba”is obligatory)
Dìdi bǎwūding tú chéng huángsè.
Brother ba roof paint resultative yellow.
The younger brother painted the roof yellow.
Bei Mandarin passive sentence structure
with a word order “patient + bei +
agent + action”(“bei”is obligatory)
Tīzi beì tāde bàba ná zǒu le.
Ladder bei he possessive-particle father
carry go-resultative perfective-marker.
The ladder was taken away by his dad.
Negation A negation word negating a word or a
sentence (negation words are obligatory) Zhè gè xiǎonǚhaí bú shuìjiào.
This classifier little girl no sleep.
This little girl is not sleeping.
Classifier A word in between a number and a noun
showing the property of the noun (the
generic classifier “ge”was not counted;
classifiers are obligatory)
Xiǎo meìmei bān laí le yì ba yǐzi.
Little sister move come-resultative perfective-
aspect one classifier chair.
The younger sister brought one chair.
Perfective aspect marker Mandarin perfective aspect “le”/“guo”
marker after either a verb or an
adjective showing an action or
status is completed (perfective
aspect markers are optional in
some cases)
Yīshēng laí le.
Doctor come perfective-maker.
The doctor came.
Dìdi gāo le.
Brother tall perfective-marker.
The younger brother became taller.
Progressive aspect marker Mandarin progressive aspect marker
“zai/zhe”showing an action is
progressing (progressive aspect
markers are optional in some cases)
Xiǎohuá zaì shuāqiáng.
Xiaohua progressive-marker brush wall.
Xiaohua is painting the wall.
Hao et al.: Narratives of Mandarin-Speaking Children With LI 9
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
to compare the group with LI and the TD group on these
measures (see Table 5). Children with LI demonstrated
shorter MLU and produced fewer NDW than their TD
peers with large effect sizes. The two groups did not differ
on TNW and TNU, with small to medium effect sizes. It
should be noted that both groups produced relatively short
stories (TNU: the group with LI ranged between three and
26 utterances per story; the TD group ranged between eight
and 21 utterances per story), indicating a limited amount
of data in the narrative samples.
We used generalized linear models to compare the
usage of complex sentences and ungrammatical sentences
between the two groups. In the models, dependent vari-
ables are binary (e.g., 0 indicates that a sentence is gram-
matical and 1 indicates that a sentence is ungrammatical ).
The fixed effect was the group, and participants were
treated as a random effect. The group had an effect on the
usage of complex sentences, F(1, 1244) = 20.25, p< .0001,
odds ratio = 0.55. The odds ratio indicated that the group
with LI was 45% less likely to produce complex sentences
than the TD group. On average, 38% (SD = 9%) of the sen-
tences produced by TD children were complex sentences,
and 24% (SD = 16%) of the sentences produced by chil-
dren with LI were complex sentences. No group differ-
ence was found for the usage of ungrammatical sentences,
F(1, 1244) = 0.708, p= .400, odds ratio = 1.12. The odds
ratioof1.12indicatedthatchildrenwithLIwere12%
more likely than TD children to produce ungrammatical
sentences. Both groups produced very few ungrammatical
sentences. On average, 2% (SD = 3%) of the sentences
produced by TD children were ungrammatical sentences,
and 4% (SD = 4%) of the sentences produced by children
with LI were ungrammatical sentences. Significance remained
the same after the Benjamini–Hochberg procedure.
Fine-Grained Microstructure Measures
Independent-samples ttests were also used to com-
pare fine-grained microstructure elements. See Table 6 for
comparison results. The group with LI displayed lower
production of Mandarin perfective aspect markers with a
large effect size. Though the differences between the two
groups on the “bei”structure and classifier were not sig-
nificant after the Benjamini–Hochberg correction, the ef-
fect size of the “bei”structure was large, and the effect
size of “classifier”approached large. Therefore, we con-
sidered the differences to be of clinical significance.
6
Chil-
dren with LI did not show any accurate usage of Mandarin
passive “bei”structure. They also displayed very low
usage of classifiers. Otherwise, all children, with LI or TD,
produced some instances of all examined microstructure
elements.
Discussion
This study compared narrative skills between
Mandarin-speaking children with LI and their age-matched
TD controls. Based on previous narrative studies in children
with LI, comparisons were conducted in three components:
macrostructure, general microstructure, and fine-grained
microstructure. For macrostructure, differences between
the two groups existed in the descriptions of characters,
settings, internal responses, complete action series, and
consequences. Regarding general microstructure, the group
with LI demonstrated shorter MLU, lower NDW, and
lower usage of complex sentences than the TD group. For
fine-grained microstructure, children with LI showed
lower usage of passive “bei”structure, classifiers, and per-
fective aspect markers. Group difference on these measures
displayed close to large to very large effect sizes (Cohen’sd
ranged from 0.75 for classifiers to 1.51 for internal response).
Macrostructure measures were sensitive to the pres-
ence of LI in Mandarin, as indicated by macrostructure
6
Even though the group difference in TNW was not significant, the
plevel was low ( p= .113), rendering it possible that lower frequencies
of the three microstructure elements in children with LI (“bei”sentence,
classifier and perfective aspect marker “le”) may have arisen from
lower TNW in the group with LI. We conducted group comparisons
of the three elements by entering TNW as a covariate. The plevels
were 0.055 (“bei”structure), 0.096 (classifier), and 0.006 (perfective
marker “le”), respectively, after controlling for TNW.
Table 4. Comparisons of macrostructure between the group with LI and the TD group.
Macrostructure
elements
LI TD
tpvalue Cohen’sdMSD MSD
Character 1.55 1.17 2.52 0.54 3.20 .004
a
1.065
Setting 1.87 0.79 2.43 0.57 2.42 .022
a
0.813
Initiating event 2.03 0.84 2.46 0.51 1.88 .071 0.619
Internal response 0.07 0.18 0.76 0.62 4.46 < .001
a
1.511
Plan 0.37 0.56 0.46 0.44 0.55 .588 0.177
Action series 2.02 0.46 2.52 0.46 3.24 .003
a
1.087
Consequence 1.25 0.96 1.98 0.74 2.56 .015
a
0.852
Total 9.16 3.80 13.13 2.46 3.73 .001
a
1.240
Note. All scores were averages of the three stories. All elements were scaled from 0 to 3. Independent-samples ttests were
conducted to compare the two groups. LI = language impairment; TD = typically developing.
a
Shows significance after conducting a Benjamini–Hochberg procedure for 18 comparisons. The false discovery rate was at .05.
10 Journal of Speech, Language, and Hearing Research •1–15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
total scores and performance on five of the seven individ-
ual macrostructure elements. This is consistent with previ-
ous narrative findings in LI (Miranda et al., 1998; Reilly
et al., 2004; Tsai & Chang, 2008; Zhang, 2009). In our
study, the macrostructure element plan was rarely used
by children, both with LI and TD, the oldest of whom
were almost 8 years old. This suggests that the plan is a
higher level macrostructure element acquired late by chil-
dren. As we discussed in the beginning, macrostructure
assessment should include later-acquired elements so that
the assessment is sufficiently challenging and can cover a
wider age range. Future investigations of macrostructure
may include the assessment of the plan for children in later
elementary grades. It is possible that the plan will surface
as an element of deficit for children with LI in higher grade
levels.
We found that among other things, Mandarin-speaking
children with LI showed significant deficits in the produc-
tion of internal responses. Their production of internal re-
sponse elements was close to floor. In Mandarin-speaking
children with LI, it is likely that the poor performance in in-
ternal responses resulted from impairment in both language
and theory of mind (de Villiers, 2007). Group differences
in the element of initiating event did not reach statistical
significance (p= .07). However, the medium effect size
(d= 0.62) suggests that children with LI may still need
some enhancement in this macrostructure element. Future
research is also needed to see if the production of initiat-
ing event in narratives is indeed relatively intact in children
with LI.
While Mandarin syntactic complexity (i.e., MLU
and usage of complex sentences) and lexical diversity (i.e.,
NDW) were sensitive to the presence of LI, productivity
measures (i.e., TNU and TNW) were not. These are con-
sistent with findings from previous studies in many Indo-
European languages (e.g., Guo et al., 2008; Hewitt et al.,
2005). These findings may suggest that differences between
children with and without LI mainly lie in linguistic so-
phistication of narrative production, not verbosity. Even
though children with LI produced a similar number of
sentences and words as their TD peers, their production
included less-complex sentences and less-diverse vocabu-
lary. We did not replicate Tsai and Chang (2008), in
which TNW was different between Mandarin-speaking
childrenwithandwithoutLIinaverysmallsample
size.
While grammaticality is very sensitive to the presence
of LI in languages with richer morphology (e.g., Reilly
Table 5. Comparisons of general microstructure between the group with LI and the TD group.
General
microstructure
elements
LI TD
tpvalue Cohen’sdMSDMSD
MLU 6.18 1.85 7.52 1.36 2.46 .019
a
0.825
NDW 35.19 16.43 47.57 13.22 2.49 .018
a
0.830
TNW 71.38 49.53 93.48 28.79 1.64 .111 0.546
TNU 11.45 5.24 12.70 3.42 0.85 .402 0.283
Note. All scores were averages of the three stories. Independent-samples ttests were conducted to compare the two
groups. LI = language impairment; TD = typically developing; MLU = mean length of utterance; NDW = number of
different words; TNW = total number of words; TNU = total number of utterances.
a
Shows significance after conducting a Benjamini–Hochberg procedure for 18 comparisons. The false discovery rate
was at .05.
Table 6. Comparisons of fine-grained microstructure between the group with LI and the TD group.
Fine-grained
microstructure
elements
LI TD
tpvalue Cohen’sdMSDMSD
Ba structure 0.83 1.03 1.04 0.68 0.70 .490 0.240
Bei structure 0.00 0.00 0.13 0.23 2.36 .030 0.810
Negation 1.01 0.66 1.20 0.85 0.73 .472 0.250
Classifier 0.03 0.12 0.37 0.63 2.25 .037 0.750
Perfective aspect marker 2.56 2.11 4.57 1.31 3.45 .002
a
1.145
Progressive aspect marker 0.89 1.11 0.87 0.76 −0.09 .927 −0.021
Note. All scores were averages of the three stories. Independent-samples ttests were conducted to compare the two
groups. LI = language impairment; TD = typically developing.
a
Shows significance after conducting a Benjamini–Hochberg procedure for 18 comparisons. The false discovery rate
was at .05.
Hao et al.: Narratives of Mandarin-Speaking Children With LI 11
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
et al., 2004), Mandarin-speaking children with LI did not
produce more ungrammatical sentences than their TD
peers in a narrative setting. The low percentages (4% for
LI and 2% for TD) of ungrammatical sentences are in sharp
contrast to Fey et al. (2004), wherein up to 16% of the
utterances produced by TD fourth graders were ungram-
matical. Since Mandarin lacks morphosyntactic features,
grammatical errors are less likely to be manifested in
Mandarin compared with languages with richer morpho-
syntactic features. In addition to the lack of complex conju-
gationsystems,Mandarinallowstheomissionofaspect
marking (e.g., perfective aspect “le/guo,”progressive aspect
“zai/zhe”). The optionality in aspect marking may also con-
tribute to better grammaticality performance in Mandarin-
speaking children with LI than children with LI who speak
Indo-European languages.
It should be noted that even though aspect markers
can be omitted in Mandarin, more-mature language learners
tend to use aspect markers to be clear about perspectives of
time in narratives (Shu, 2004). Therefore, the lower usage of
perfective aspect markers reflects true deficits in Mandarin-
speaking children with LI. It should also be noted that other
fine-grained Mandarin linguistic features examined in the
current study are not optional. In the “ba”or “bei”struc-
ture, the word ba or bei must be included (Li & Thompson,
1981). Similarly, when describing the quantity of an object,
classifiers must be present between the number and the noun
(Erbaugh, 2006; Li & Thompson, 1981). Thus, the differ-
ences between the two groups on the usage of the “bei”
structure and classifiers are not confounded by optionality
but reflect real differences in children’s deployment of these
linguistic features.
Another potential reason for low ungrammaticality
resides in the spontaneous language sampling context. The
narrative context gives children the flexibility to choose
expressions that they have already mastered and to avoid
expressions that they are not confident with. Differences in
grammaticality between children with and without LI may
become more prominent in an obligatory context where
children are required to use certain language features. Also,
this may explain why we did not find a difference by group
on negation and “ba”structure, whereas Zeng et al. (2013)
and He and Dai (2012) found a difference using cloze tasks.
However, we think that the influence of context is second-
ary compared to the impact of language typology. In a
narrative context, performance on grammaticality differed
significantly between the TD group and the group with
LI in many Indo-European languages (e.g., Reilly et al.,
2004; Reuterskiöld et al., 2011). For these languages, the
flexibility afforded by the spontaneous language context
cannot mask the deficits of the group with LI in the mark-
ing of verb or noun-related morphological features.
The three fine-grained microstructure elements (i.e.,
Mandarin passive “bei structure, Mandarin perfective
marker “le/guo”and classifiers) are Mandarin-specific fea-
tures not shared with Indo-European languages. These
results resonate with Leonard’s (2014b) proposal that
the particular language determines domains of ease and
difficulty in children with LI. English-speaking children
with LI show particular deficits in English verb conjuga-
tions (e.g., tense markers and subject–verb agreement;
Leonard, 2014a). Children with LI who speak Italian and
Spanish are found to be more likely to demonstrate deficits
in noun-related morphology (e.g., clitics, plural inflection;
Bedore & Leonard, 2001; Leonard & Dispaldro, 2013).
Mandarin-speaking children with LI, however, show a dif-
fuse array of deficits ranging from sentence structure to
verb particles and noun modifiers.
Mandarin classifiers are known to be later acquired
as a result of their vast number and complex usage
(Erbaugh, 2006). Mandarin-speaking children with LI
may need extended time learning these forms and their
corresponding semantic categories. Regarding passive
structure, children generally exhibit late mastery across
a number of languages (Leonard et al., 2006; Perovic,
Vuksanović,Petrović,&Avramović-Ilić, 2014). Mastery
of the passive structures may require more-sophisticated
linguistic knowledge and cognitive ability because the
forms involve noncanonical word order and are used in
special circumstances.
Differences in perfective aspect markers were associ-
ated with a very large effect size (d= 1.15). It is possible
that children with LI may opt to omit aspect markers when
they are not sure of the correct use. However, the lower
frequency of perfective marker in the children with LI can-
not be fully attributed to the optionality of this grammati-
cal feature because even after we factored out the potential
influence of TNW, the difference was still highly signifi-
cant (p= .006). These uncertainties notwithstanding, we
can confidently say that the TD children used perfective
aspect marking more frequently, indicating better mas-
tery and clear evidence of perspective taking in various
contexts.
Conclusion, Clinical Implications, Limitations,
and Future Directions
To conclude, this study compared narrative skills
between Mandarin-speaking children with and without LI
on macrostructure and microstructure (general and fine-
grained) measures. Results revealed that Mandarin-speaking
children with and without LI showed comparable perfor-
mance in productivity and grammaticality in a narrative
context. These children exhibited weaknesses in most macro-
structure elements, lexical diversity, syntactic complexity,
and three Mandarin-specific microstructure elements.
The study has important clinical implications for
speech-language pathologists who work with children
speaking Mandarin Chinese. First, the measure of gram-
maticality was not differentiating between the TD group
and the group with LI in Mandarin-speaking children. This
is different from children speaking many Indo-European
languages where grammaticality was typically sensitive
in detecting group membership (e.g., Reilly et al., 2004).
When choosing measures to monitor children’s language
12 Journal of Speech, Language, and Hearing Research •1–15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
performance, researchers and speech-language pathologists
should always take language typology into account and
bear in mind that a language with very sparse morphology
may inherently lead to fewer opportunities for grammatical
violations. Second, clinicians should focus their attention
on the areas of identified weaknesses in Mandarin-speaking
children with LI, including macrostructure, lexical diver-
sity, syntactic complexity, and the three fine-grained micro-
structure elements. These areas should be directly targeted
when screening children for potential LI and when setting
intervention goals. Last but not the least, clinicians should
be aware that verbosity is secondary compared to the qual-
ity of language production. In storytelling, children with
LI may produce a comparable number of utterances and
even a comparable number of total words as their age
peers, but they are more likely to repeatedly use the same
words (i.e., lower lexical diversity), and their sentences may
be less densely packed with information (i.e., lower syntac-
tic complexity).
This study has several limitations. Children pro-
duced short story samples elicited by the three stories. This
yielded a relatively low TNU (about 36 C-units from each
child). As such, the current findings should be interpreted
with caution and need replication through larger numbers
of children and longer narrative samples. Moreover, we
conducted multiple comparisons, which increased Type I
error. However, we took the Benjamini–Hochberg proce-
dure to protect against Type I error. The findings, thus,
should reliably reflect deficits in Mandarin-speaking chil-
dren with LI.
Future research could explore the effects of different
narrative tasks on narrative performance. In this study,
we did not model the task of storytelling for children.
There is debate about whether a model should be provided
to familiarize children with how to tell a story. Heilmann,
Miller, and Nockerts (2010) claimed that the provision of
a model enables children to display their best narrative
performance. However, Norbury and Bishop (2003) be-
lieved that storytelling without a model provides a more
realistic impression of children’s abilities to plan and orga-
nize stories as it is not shaped by an adult model. Future
research could also investigate speech disruption rates (e.g.,
pauses, vocal hesitations, revisions, and repetitions). Defi-
cits in these measures reflect processing difficulties in re-
trieving vocabulary and syntactic frames, and children with
LI have been found to display higher speech disruption
rates in word and sentence retrievals (Guo et al., 2008).
Acknowledgments
This study was supported by the Shanghai Municipal Edu-
cation Commission (D1502) awarded to Li Sheng, Yiwen Zhang,
and Fan Jiang and a Pudong One Hundred Award to Li Sheng.
The authors wish to thank all the participating families for
their time; Jiandan Huang, Jiaolong Yang, Huilin Chen for their
assistance with data collection; and Zijing Yu, Yuxiang Wang,
and Li Chen for their assistance with data transcription and
interpretation.
References
Bedore, L. M., & Leonard, L. B. (2001). Grammatical morphol-
ogy deficits in Spanish-speaking children with specific lan-
guage impairment. Journal of Speech, Language, and Hearing
Research, 44, 905–924.
Bedore, L. M., Peña, E. D., Joyner, D., & Macken, C. (2011).
Parent and teacher rating of bilingual language proficiency
and language development concerns. International Journal of
Bilingual Education and Bilingualism, 14(5), 489–511.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false dis-
covery rate: A practical and powerful approach to multiple
testing. Journal of the Royal Statistical Society. Series B
(Methodological), 57, 289–300.
Bishop, D. V. (1997). Uncommon understanding: Comprehension in
specific language impairment. East Sussex, United Kingdom:
Psychology Press.
Brown, R. (1973). A first language: The early stages. Cambridge,
MA: Harvard University Press.
Cheung, H. (2009). Grammatical characteristics of Mandarin-
speaking children with specific language impairment. In S. Law,
B. Weekes, & A. M.-Y. Wong (Eds.), Language disorders in
speakers of Chinese. Bristol, United Kingdom: Multilingual
Matters.
Cleave, P. L., Girolametto, L. E., Chen, X., & Johnson, C. J.
(2010). Narrative abilities in monolingual and dual language
learning children with specific language impairment. Journal
of Communication Disorders, 43, 511–522.
Cohen, J. (1988). Statistical power analysis for the behavioral sci-
ences (2nd ed.). Hillsdale, NJ: Erlbaum.
de Villiers, J. (2007). The interface of language and theory of
mind. Lingua, 117, 1858–1878.
de Villiers, J. G., & Liu, X. L. (2014). Mandarin Expressive Narra-
tive Test (MENT). Dallas, TX: Bethel Hearing and Speaking
Training Center.
de Villiers, P. (2004). Assessing pragmatic skills in elicited produc-
tion. Seminars in Speech and Language, 25(1), 57–71.
Ehrler, D. J., & McGhee, R. L. (2008). Primary Test of Nonverbal
Intelligence (PTONI) manual. San Antonio, TX: Pro-Ed.
Erbaugh, M. (2006). Chinese classifiers: Their use and acquisi-
tion. In P. Li, L. H. Tan, E. Bates, & O. J. L. Tzeng (Eds.),
Handbook of East Asian psycholinguistics (Vol. 1, pp. 39–51).
Cambridge, United Kingdom: Cambridge University Press.
Fey, M. E., Catts, H. W., Proctor-Williams, K., Tomblin, J. B., &
Zhang, X. (2004). Oral and written story composition skills
of children with language impairment. Journal of Speech, Lan-
guage, and Hearing Research, 47, 1301–1318.
Fletcher, P., Leonard, L. B., Stokes, S. F., & Wong, A. M.-Y.
(2005). The expression of aspect in Cantonese-speaking chil-
dren with specific language impairment. Journal of Speech,
Language, and Hearing Research, 48, 621–634.
Gagarina, N., Klop, D., Tsimpli, I. M., & Walters, J. (2016). Nar-
rative abilities in bilingual children. Applied Psycholinguistics,
37, 91–122.
Gallinat, E., & Spaulding, T. J. (2014). Differences in the perfor-
mance of children with specific language impairment and
their typically developing peers on nonverbal cognitive tests:
A meta-analysis. Journal of Speech, Language, and Hearing
Research, 57, 1363–1382.
Gillam, S. L., Gillam, R. B., Fargo, J. D., Olszewski, A., &
Segura, H. (2017). Monitoring Indicators of Scholarly Lan-
guage: A progress-monitoring instrument for measuring nar-
rative discourse skills. Communication Disorders Quarterly,
38(2), 96–106.
Hao et al.: Narratives of Mandarin-Speaking Children With LI 13
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Glickman, M. E., Rao, S. R., & Schultz, M. R. (2014). False discov-
ery rate control is a recommended alternative to Bonferroni-type
adjustments in health studies. Journal of Clinical Epidemiology,
67(8), 850–857.
Gorman, B. K., Bingham, G. E., Fiestas, C. E., & Terry, N. P.
(2016). Assessing the narrative abilities of Spanish-speaking
preschool children: A Spanish adaptation of the narrative
assessment protocol. Early Childhood Research Quarterly, 36,
307–317.
Guo, L., Tomblin, J. B., & Samelson, V. (2008). Speech disrup-
tions in the narratives of English-speaking children with spe-
cific language impairment. Journal of Speech, Language, and
Hearing Research, 51, 722–738.
He, X., & Dai, H. (2012).
[The acquisition of negation in Mandarin-speaking children with
SLI]. ,23(4), 16–20.
Heilmann, J., Miller, J. F., & Nockerts, A. (2010). Sensitivity of
narrative organization measures using narrative retells pro-
duced by young school-age children. Language Testing, 27(4),
603–626.
Heilmann, J., Miller, J. F., Nockerts, A., & Dunaway, C. (2010).
Properties of the narrative scoring scheme using narrative
retells in young school-age children. American Journal of
Speech-Language Pathology, 19, 154–166.
Hewitt, L. E., Hammer, C. S., Yont, K. M., & Tomblin, J. B.
(2005). Language sampling for kindergarten children with and
without SLI: Mean length of utterance, IPSYN, and NDW.
Journal of Communication Disorders, 38, 197–213.
Hick, R. F., Joseph, K. L., Conti-Ramsden, G., Serratrice, L., &
Faragher, B. (2002). Vocabulary profiles of children with
specific language impairment. Child Language Teaching and
Therapy, 18, 165–180.
Justice, L. M., Bowles, R., Pence, K., & Gosse, C. (2010). A scal-
able tool for assessing children’s language abilities within a
narrative context: The NAP (narrative assessment protocol).
Early Childhood Research Quarterly, 25, 218–234.
Klee, T., Stokes, S. F., Wong, A. M. Y., Fletcher, P., & Gavin,
W. J. (2004). Utterance length and lexical diversity in Cantonese-
speaking children with and without specific language impair-
ment. Journal of Speech, Language, and Hearing Research, 47(6),
1396–1410.
Klein, W., Li, P., & Hendriks, H. (2000). Aspect and assertion
in Mandarin Chinese. Natural Language & Linguistic Theory,
18(4), 723–770.
Landis, J. R., & Koch, G. G. (1977). The measurement of observer
agreement for categorical data. Biometrics, 33(1), 159–174.
Leonard, L. B. (2014a). Children with specific language impairment
(2nd ed.). London, United Kingdom: MIT Press.
Leonard, L. B. (2014b). Children with specific language impair-
ment and their contribution to the study of language develop-
ment. Journal of Child Language, 41(Suppl. 1), 38–47.
Leonard, L. B., & Dispaldro, M. (2013). The effects of production
demands on grammatical weaknesses in specific language
impairment: The case of clitic pronouns in Italian. Journal
of Speech, Language, and Hearing Research, 56, 1272–1286.
Leonard, L. B., Wong, A. M.-Y., Deevy, P., Stokes, S. F., &
Fletcher, P. (2006). The production of passives by children
with specific language impairment: Acquiring English or
Cantonese. Applied Psycholinguistics, 27, 267–299.
Li, C., & Thompson, S. A. (1981). Mandarin Chinese: A functional
reference grammar. Berkeley, CA: University of California
Press.
Liu, X. L., de Villiers, J., Ning, C., Rolfhus, E., Hutchings, T.,
Lee, W., . . . Zhang, Y. W. (2017). Research to establish the
validity, reliability, and clinical utility of a comprehensive lan-
guage assessment of Mandarin. Journal of Speech, Language,
and Hearing Research, 60(3), 592–606.
Loban, W. (1976). Language development: Kindergarten through
grade twelve (Vol. 18). Urbana, IL: National Council of
Teachers.
MacWhinney, B. (2000). The CHILDES project: Tools for analyz-
ing talk (3rd ed.). Mahwah, NJ: Erlbaum.
Miller, J., & Chapman, R. (2002). Systematic Analysis of Language
Transcript. Madison: University of Wisconsin—Madison,
Waisman Center, Language Analysis Laboratory.
Miranda, A. E., Mccable, A., & Bliss, L. S. (1998). Jumping around
and leaving things out: A profile of the narrative abilities of chil-
dren with specific language impairment. Applied Psycholinguis-
tics, 19, 647–667.
Muñoz, M. L., Gillam, R. B., Peña, E. D., & Gulley-Faehnle, A.
(2003). Measures of language development in fictional nar-
ratives of Latino children. Language, Speech, and Hearing
Services in Schools, 34(4), 332–342.
Ning, C. Y., Liu, X. M., & de Villiers, J. (2014). The Diagnostic
Receptive and Expressive Assessment of Mandarin. Dallas, TX:
Bethel Hearing and Speaking Training Center.
Norbury, C. F., & Bishop, D. V. M. (2003). Narrative skills of chil-
dren with communication impairments. International Journal
of Language & Communication Disorders, 38(3), 287–313.
Peña, E., Resendiz, M., & Gillam, R. (2007). The role of clinical
judgments of modifiability in the diagnosis of language impair-
ment. Advances in Speech Language Pathology, 9, 332–345.
Perneger, T. (1998). What’s wrong with Bonferroni adjustments.
British Medical Journal, 316, 1236–1238.
Perovic, A., Vuksanović, J., Petrović, B., & Avramović-Ilić,I.
(2014). The acquisition of passives in Serbian. Applied Psycho-
linguistics, 35, 1–26.
Reilly, J., Losh, M., Bellugi, U., & Wulfeck, B. (2004). “Frog,
where are you?”Narratives in children with specific language
impairment, early focal brain injury, and Williams syndrome.
Brain and Language, 88, 229–247.
Restrepo, M. A. (1998). Identifiers of predominantly Spanish-
speaking children with language impairment. Journal of
Speech, Language, and Hearing Research, 41(6), 1398–1411.
Reuterskiöld, C., Hansson, K., & Sahlén, B. (2011). Narrative
skills in Swedish children with language impairment. Journal
of Communication Disorders, 44(6), 733–744.
Rezzonico, S., Chen, X., Cleave, P. L., Greenberg, J., Hipfner-
Boucher, K., Johnson, C. J., . . . Girolametto, L. (2015). Oral
narratives in monolingual and bilingual preschoolers with
SLI. International Journal of Language & Communication Dis-
orders, 50(6), 830–841.
Rice, M. L., Smolik, F., Perpich, D., Thompson, T., Rytting, N.,
& Blossom, M. (2010). Mean length of utterance levels in
6-month intervals for children 3 to 9 years with and without
language impairment. Journal of Speech, Language, and Hear-
ing Research, 53, 333–349.
Scarborough, H. S., Rescorla, L., Tager-Flusberg, H., Fowler, A. E.,
& Sudhalter, V. (1991). The relation of utterance length to gram-
matical complexity in normal and language-disordered groups.
Applied Psycholinguistics, 12(01), 23–46.
Shu, K. L. (2004). The acquisition of aspect markers by Mandarin-
Speaking children (Doctoral dissertation). Boston University,
MA. Retrieved from http://search.proquest.com/docview/
305215146?pq-origsite=summon&accountid=7118
Squires, K. E., Lugo-Neris, M. J., Peña, E. D., Bedore, L. M.,
Bohman, T. M., & Gillam, R. B. (2014). Story retelling by
bilingual children with language impairments and typically
14 Journal of Speech, Language, and Hearing Research •1–15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
developing controls. International Journal of Language & Com-
munication Disorders, 49(1), 60–74.
Tsai, W., & Chang, C. (2008). “But I first…andthenhekept
picking”Narrative skill in Mandarin-speaking children with
language impairment. Narrative Inquiry, 18(2), 349–377.
Tsimpli, I. M., Peristeri, E., & Andreou, M. (2016). Narrative pro-
duction in monolingual and bilingual children with specific lan-
guage impairment. Applied Psycholinguistics, 37(1), 195–216.
Yip, V., & Matthews, S. (2010). The acquisition of Chinese in
bilingual and multilingual contexts. International Journal of
Bilingualism, 14, 127–146.
Zeng,T.,Li,H.,Li,K.,&He,X.(2013).
[On acquisition of
ba-construction in Mandarin-speaking children with specific lan-
guage impairment]. ,1(49), 10–18.
Zhang, F. F. (2009).
[Narrative development in Mandarin-speaking children with
SLI]. In J. Zhou & J. R. Zhang (Eds.),
[Research in Mandarin-speaking children’s language develop-
ment (1st ed., pp. 127–147). .
Zhang, H. (2007). Numeral classifiers in Mandarin Chinese. Jour-
nal of East Asian Linguistics, 16(1), 43–59.
Hao et al.: Narratives of Mandarin-Speaking Children With LI 15
Downloaded From: http://jslhr.pubs.asha.org/ by University of Delaware, Li Sheng on 02/12/2018
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
