View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by HKU Scholars Hub

An index of syntactic development for Cantonese-Chinese

Title preschool children

Author(s) Wan, Ka-hei, Emily; 溫嘉曦

Citation

Issued Date 2011

URL http://hdl.handle.net/10722/192908

The author retains all proprietary rights, (such as patent rights)

Rights and the right to use in future works.
 An index of syntactic development for Cantonese-Chinese preschool children

Wan Ka Hei, Emily

A dissertation submitted in partial fulfilment of the requirements for the Bachelor of Science
(Speech and Hearing Sciences), The University of Hong Kong, June 30, 2011.

1
 An index of syntactic development
Abstract

This research study aimed to develop an index of syntactic development for Cantonese-

speaking children. Language samples taken from 14 normal children aged from 4;1 to 5;0, 16

normal children aged from 5;1 to 6;5 and 15 SLI children aged from 5;1 to 6;4 were analyzed

and credited according to the framework developed. Normal children aged from 4;1 to 5;0

performed poorer on the index than those aged from 5;1 to 6;5 with the same clinical status.

Children with language difficulty performed poorer than their normal age peers on the index

as well. The index was validated against MLU and the two indices moderately correlated with

each other. A linear combination of age, D and the index was entered into discriminant

analysis, yielding a classification accuracy of 86.7% of all the children. The index was found

to be a potentially useful clinical marker of SLI yet replication is needed to confirm the

findings. Further modification of the index was discussed. The age and language growth

sensitivity of MLU was discussed as well.

2
 An index of syntactic development
Introduction

Language sample analysis has been widely adopted to document language growth and

assess preschoolers‟ language ability. Though there are lots of psychometric tests available to

achieve these purposes, their ecological validity are always being compared against language

samples. Ecological validity refers to how well a measure resembles everyday conversation

settings (Costanza-Smith, 2010). Language sample analysis is being considered as more

ecologically valid than standardized tests. In language sample analysis, children can decide

topic that they want to elaborate and use language in a naturalistic context highly similar to

daily conversation (Dunn, Flax, Sliwinski & Aram, 1996). They have to assemble their own

sentences using their syntactic knowledge, choose appropriate vocabulary and use language

properly to interact with communicators with regard to their pragmatic knowledge (Costanza-

Smith, 2010). Getting back to psychometric measures, they are administrated in restricted and

controlled environments that hardly resemble daily communication (Crais, 1995). Moreover,

children‟s productions are elicited from sentence completions or unnatural sentence creations

(Costanza-Smith, 2010). Psychometric tests can only assess what children know about the

syntax and the semantics of a language, while no information about how they integrate form,

content and use of language in daily conversation can be obtained (Lahey & Bloom, 1994).

Several indices used of quantifying language development have been established for

English-speaking preschoolers using language sample analysis. Among these indices, MLU

(Brown, 1973) and D (Malvern & Richards, 1997) are two widely investigated norm-

referenced grammatical indices. Owing to their global features, they have been adopted in

language sample analysis of Cantonese speaking preschoolers as well. MLU is a calculation of

mean morpheme numbers per utterance (Brown, 1973). Studies have shown that MLU

correlated with age in normal English-speaking preschoolers (Klee, Gavin & Stokes, 2007;
3
 An index of syntactic development
Blake, Quartaro & Onorati, 1993; Miller and Chapman, 1981), English-speaking preschoolers

with language difficulties (Klee, Schaffer, May, Membrino & Mougey, 1989), and normal

Cantonese-speaking preschoolers (Klee, Stokes, Wong, Fletcher & Gavin, 2004). As for

another quantitative language measure, D, it is a relatively new calculation of lexical diversity.

Unlike previous calculation of lexical diversity such as TTR, D was not affected by sample

size (Richards & Malvern, 1998). It correlated with age as well in the studies of normal

English-speaking preschoolers (Klee et al., 2007) and normal Cantonese-speaking

preschoolers (Klee et al., 2004). These two indices were said to be developmentally sensitive.

Apart from being sensitive to age, MLU and D could differentiate SLI children from their

normal peers. MLU exhibited group difference between Cantonese speaking children with and

without language difficulty (Wong, Klee, Stokes, Fletcher & Leonard, 2010). The same

pattern held in D as well in English study (Owen and Leonard, 2002) and Cantonese studies

(Wong et al., 2010). The two measures, MLU and D, in combination with age could be a

diagnostic marker of SLI both for English-speaking preschoolers (Klee et al., 2007) and

Cantonese-speaking preschoolers (Klee et al., 2004). However, a replication study of Klee et

al. (2004) did not yield supporting results (Wong et al., 2010).

Grammatical development quantification of Cantonese-speaking preschoolers is limited to

MLU and D only. However, in studies of English-speaking preschoolers, criterion-referenced

procedures such as LARSP (Language assessment, remediation and screening procedure)

(Crystal, Fletcher & Garman, 1976, 1981), Assigning Structural Stage Procedure (Miller,

1981) and Index of Productive Syntax (IPSyn; Scarborough, 1990) are available. LARSP is a

qualitative language profile of preschoolers. It analyzed clause, phrase, word and expansion in

language samples. In the study by Blake et al. (1993), the phrasal and clausal stages in LARSP

were adopted to develop an index. Frequency counts of the targeted phrases and clauses were
4
 An index of syntactic development
computed for each child using language samples. As for Assigning Structural Stage Procedure

and IPSyn, both of them were criterion-referenced quantitative grammatical index for English-

speaking children. They looked for noun phrase expansions, verb phrase expansions,

questions, negation and complex sentence structures in language samples. Quantification

procedures were similar to that in Blake et al. (1993). The shared features between the

syntactic complexity measure of Blake et al. (1993) and IPSyn was that, they correlated with

age and MLU in normal children. The same relationship was found in language delayed

children using IPSyn as well (Scarborough, Rescorla, Tager-Flusberg, Fowler, & Sudhalter,

1991). These indices‟ sensitivity to age did not decrease with age growth while MLU did so

after utterance length of 4.5 (Blake et al., 1993) and 3 (Scarborough et al., 1991). So it was

concluded that quantitative grammatical index offered more sensitive age differentiation than

MLU (Blake et al., 1993; Scarborough et al., 1991).

Qualitative language profile was developed for Cantonese-speaking preschoolers in the

study by Fletcher, Leung, Stokes and Weizman (2000). In their study, types and tokens of

closed and open class vocabulary, syntax and sentence structures used by two- to five-year-old

children were documented from language sample analyses. But this profile was not used for

quantifying grammatical development in Cantonese population. Given that MLU and D, the

only two quantifying language indices in Cantonese, may not be a sensitive language growth

index after utterance length exceeds 3, a development of a syntactic development index will

be necessary. It is hoped that the index can document language growth quantitatively and

assess language ability of preschoolers as well.

Another call for development of a Cantonese syntactic complexity index stems from a lack

of studies in Cantonese-speaking preschoolers isolated form and structure expression

development. Only a few studies have investigated into this issue, such as aspect markers
5
 An index of syntactic development
(Leung, 1995), serial verb construction (Cheng, 2006), connectives (Ma, 2006) and classifiers

(Tse, Li & Leung, 2007). The study done by Fletcher et al. (2000) was a comprehensive one

but it did not identify the qualitative grammatical development. These findings are relatively

scarce to define and quantify preschoolers‟ syntactic complexity using isolated grammatical

forms and structures, compared with studies in English. Hence the value of conversational

sample in assessing language-impaired children is limited to the use of MLU and D only

If a syntactic complexity index is to be developed for Cantonese preschoolers, its validity

should be tested against another language measure. MLU and D are the only norm-referenced

indices derived from language sample analysis in Cantonese. However, MLU has been

validated against other similar quantitative grammatical summaries (e.g. syntactic complexity

measure, Blake et al., 1993; IPSyn, Scarborough, 1990). So MLU is a better option than D. It

was found to be highly correlated with quantitative syntactic measures of English-speaking

preschoolers, such as IPSyn in the study of Scarborough (1990) (r=.92) and syntactic

complexity measure in the study by Blake et al. (1993) (r=.88). Though these reported

correlations in English cannot be totally generalized to Cantonese as its grammatical

morphemes are not obligatory (To, Stokes, Cheung & T‟sou, 2010), they support the use of

MLU in validation.

Establishing a valid syntactic complexity index from language sample analysis can

contribute to the identification of SLI preschoolers. In Hong Kong, the only available

expressive language assessment for preschoolers is The Reynell Developmental Language

Scales-Cantonese Version (Hong Kong Society for Child Health and Development, 1987).

The ecological validity of psychometric tests in examining preschoolers‟ language ability has

been compared against language sample analysis previously. To be in short, psychometric

tests elicit children‟s language productions in either vocabulary or sentence form in isolation.
6
 An index of syntactic development
Children need not make up their own sentences (Costanza-Smith, 2010). Testing

environments hardly resemble real communication situations (Crais, 1995). Language

difficulty encountered in daily communication may not be truly reflected and language

problems may be left undetected (Miller, 2005). Moreover, standardized tests do not examine

error productions in conversation. In the study by Dunn et al. (1996), SLI children produced

more errors in spontaneous speech than normal age peers. As for language sample analysis,

children‟s use of language in natural context is assessed. Communication problems caused by

morphosyntactic deficit of SLI children can be truly and fully revealed. Qualitative difference

in morphosyntactic system between SLI children and normal age peers can be addressed. For

instance, SLI children showed less facility in utilizing aspect markers in conversation (Stokes

& Fletcher, 2000). Experimental studies also showed that SLI children were less proficient

than their age peers in constructing passive sentences (Leonard, Wong, Deevy, Stokes and

Fletcher, 2006) and who-object questions (Wong, Leonard, Fletcher & Stokes, 2004). Error

pattern produced during conversation can also be examined in language sample analysis.

This project aimed to address the following research questions:

1. Was the index of grammatical development developed in this study sensitive to

developmental changes in Cantonese-speaking preschoolers?

2. How strong was the correlation between the index of grammatical development developed

in this study and MLU in Cantonese-speaking preschoolers?

3. Did children in the SLI group receive lower score on the index of syntactic development

than the normal age-matched peers?

4. Did children in the SLI group produce more syntactic errors than the group of normal

children of the same age?

7
 An index of syntactic development
5. How well could the index of syntactic developmental, age, MLU and D in combination

discriminate children with SLI from their normal age peers?

Method

Forty five language samples from the language archives in the study of Wong et al. (2010)

were analyzed for this study. The forty five samples were divided into three groups. Fourteen

normally developing children aged from 4;1 to 5;0 were classified into the TD 4-5 group,

sixteen normally developing children aged from 5;1 to 6;5 were in the TD 5-6 group and

fifteen SLI children aged from 5;1 to 6;4 were in the SLI 5-6 group. Among these samples, the

MLU and D of the language samples from the TD 4-5 group were reported in Wong et al.

(2010).

According to Wong et al. (2010), the typically-developing children did not have any history

or parental concerns of language, speech & hearing problem. All the children in the TD 4-5

group and the TD 5-6 group scored above -0.5 SD and -0.3 SD respectively on the Receptive

subtests of the Cantonese version (Hong Kong Society for Child Health and Development,

1987) of the Reynell Developmental Language Scales – Revised (RDLS-R, Reynell &

Huntley, 1985). All the children in the SLI 5-6 scored below -1.2 SD on the mean of RDLS-R,

with thirteen children scoring below -1.25 SD. All the children passed Columbia Mental

Maturity Scale. They have also passed a pure-tone audiometry screening tested at 0.5, 1.0, 2.0

and 4.0 kHz with loudness at 25-30 dB HL. A history of seizure, neurological or

psychological problems should not be present in all the children.

To collect a language sample, each child was involved in a 15- to 20-min conversation with

one or two speech therapist research assistants trained in language sampling previously. Then

a team of eight students majoring in Speech and Hearing Sciences, Psychology and Chinese

Linguistics transcribed the samples after training. For further details of language sampling and

transcription, please refer to Wong et al. (2010).

8
 An index of syntactic development
Four subtests of syntactic structure were identified: noun phrase elaboration (NP), verb

phrase elaboration (VP), complex sentence structure (SS), question (Q) and connectives &

sentence adverbs (CON/SA). These categories were adopted and validated in syntactic

development studies of English-speaking preschoolers (Scarborough, 1990; Miller, 1981).

Items in each category are as follows: A noun phrase comprises a head noun with an adjective,

a relative clause and a classifier as modifiers (Zhang, 2007; Matthews & Yip, 1994); a verb

phrase can be elaborated by an aspect marker, a verbal complement, a modal auxiliary, a

manner predicate and a clausal complement; SS includes pivotal construction, prepositional

phrase and serial verb construction; Q includes yes/no question and wh-question; CON/SA

includes connectives and sentence adverbs. For details and examples of the items included in

each category, please refer to Appendix A.

The chosen items should possess potential in illustrating developmental progress in

preschoolers. For instance, two-year-old children used “result + zo2” (laan6 zo2) to describe a

consequence while three-year-old children used resultant verb complement (verb + result +

zo2) (dit3 laan6 zo2) and older children can produce it with greater diversity (Cheung, 1991).

The types of classifier employed by preschoolers increased with age, from three to five years

old (Tse et al., 2007). Children started to use more general classifier “go3” to more specific

sortal classifiers (e.g. tiu4, bun2), and then advanced to mensural classifiers (e.g. deoi3, hap6)

and measurement classifiers (e.g. jat6, jyut6). These items should also be potentially

discriminative between SLI and typically-developing children. For instance, SLI children

produced qualitatively and quantitatively less aspect markers than their age-matched

counterparts during conversation (Fletcher, Leonard, Stokes & Wong, 2005). In some

experimental studies, they were also less capable of producing correct shape and function

classifiers and omitted classifiers more frequently than age-matched peers (Stokes & So,

9
 An index of syntactic development
1997). Construction of passive sentences posed more difficulty to SLI children than to normal

age-matched peers as well (Leonard et al., 2006).

Refinement was made to eliminate items showing weak changes, i.e. those rarely produced

or equally well-mastered by children in both age groups (Scarborough, 1990). For example,

negation was removed from the index as it was found to be equally well-mastered by

typically-developing children in both age groups. Novel structures identified occasionally will

be added to the index for analysis. To reveal the syntactic development among the two normal

age groups, items that emerge as later developmental form will receive more credits after

refinement. For instance, why- and how-questions receive one more credit than what-, who-

and where-questions, as the study by Fletcher et al. (2000) showed that the former question

types emerged later than the latter types.

Targeted items were then identified from each utterance in the samples. Three kinds of

productivity criteria (contextual criterion, context criterion and phrasal criterion) were

adopted form IPSyn (Scarborough, 1990) for crediting a second example of the same items. If

a form appears for its second time, it must meet the contextual criterion, i.e. be in a distinct

adjacent context for crediting of a second point. The lexical criterion should be satisfied as

well, i.e. to be expressed in a different form compared with the first appearance. If the item is

a sentence/phrase, the second example should meet the phrasal criterion, i.e. half of its

wordings should be different from the first example. A grammatical item can be credited for a

maximum of three times if the above-mentioned criteria can be satisfied. Syntactic errors will

be identified in each sample and being classified into addition, substitution, omission and

wrong order of word. To address inter-rater reliability, 10% of the samples, i.e. 1 sample from

the TD 4-5 group, 1 from the TD 5-6 group and 2 from the SLI 5-6 group, would be marked

by another rater again. The percentage of agreement is 95.1%, revealing high accuracy of

agreement between the two raters.

10
 An index of syntactic development
Results

In this study, all of the group comparisons were conducted using one-way ANOVA

procedures followed by Tukey HSD post hoc comparison, conducted at p<.05 level. Pearson‟s

r tests were used to quantify the strength of the correlations among the variables of interest.

According to Cohen (1988), r ranging from 0.9 to 1 represents a very high correlation, 0.7 to

0.9 represents a high correlation, 0.5 to 0.7 indicates a moderate correlation, and 0.3 to 0.5

indicates a low correlation.

For our questions of interest, the reports focus on the comparisons between the TD 4-5

group and the TD 5-6 group, and between the TD 5-6 group and the SLI 5-6 group.

Comparison between the TD 4-5 group and the SLI 5-6 group will be presented only when

there were significant differences.

Results of The Standardized Tests

The descriptive statistics for CMMS, CRVT, RDLS-E and RDLS-R of each group are

summarized in Table 1, accompanied with the results of Tukey HSD post-hoc comparisons in

which significant difference was marked at p< .05.

Table 1. Sample size, age (mean and range) and scores of the standardized language test (Mean, SD and range)

for each group

Group TD 4-5 TD 5-6 SLI 5-6

N 14 16 15

Age(month) M (SD) 55.71 a, b 67.25 a 67.13 b

Range 49 - 60 61 – 77 61 – 76

CMMS M (SD) 108.93c (5.99) 113.06d (7.21) 100.07c,d (9.55)

Range 98 – 120 101 – 126 83 – 111

CRVT M (SD) 58.79e (4.30) 61.69f (2.27) 53.80 e, f (4.41)

(max=65) Range 52 -64 56 – 65 45 – 63

11
 An index of syntactic development
RDLS-E M (SD) 57.57 g (5.02) 62.69g,h (4.22) 54.73 h (5.95)

(max= 73) Range 49 – 66 56 – 70 44 – 66

RDLS-R M (SD) 55.64 i, j, k (3.23) 59.81 j, i, k (2.86) 48.47 k, i, j (4.47)

(max=67) Range 50 – 62 56 – 65 41 – 54

Note: Means having the same superscript differ significantly at p<.05.

CMMS = Columbia Mental Maturity Scale presented in standard score; RDLS-E = Reynell

Developmental Language Scales- Expressive raw score; RDLS-R = Reynell Developmental

Language Scales- Receptive raw score; CRVT= Hong Kong Cantonese Receptive

Vocabulary Test presented in raw score

According to the ANOVA results, there was a significant difference in age among the three

groups, F(42, 2) = 29.28, p<.0001, and the follow up comparisons confirmed that the TD 5-6

group and the SLI 5-6 group did not differ in age, p= .997.

ANOVA results revealed significant group differences in all standardized tests ([CMMS]:

F (2, 42) = 11.26, p<.0001; [CRVT]: F (2, 42) = 17.43, p<.0001; [RDLS-E]: F (2, 42) = 9.74,

p<.0001; [RDLS-R]: F (2, 42) = 39.60, p<.0001). For CMMS, which is a non-verbal cognitive

test, post-hoc comparison indicated that no significant difference was found between the two

TD groups, p=.321, while the TD 5-6 group (mean= 113.01, SD= 7.21) scored significantly

higher than SLI 5-6 group (mean= 100.01, SD= 9.55), p< .0001. The TD 4-5 group (mean=

108.93, SD= 5.99) has significantly higher score than the SLI 5-6 group as well, p=.01. The

same pattern across the three groups was found in CRVT, i.e. the two TD groups do not differ

significantly, p= .099; whereas the TD 6-5 group (mean= 61.69, SD= 2.27) scored

significantly higher than the SLI 5-6 group (mean= 53.8, SD= 4.41), p<.0001. The TD 4-5

group (mean= 58.79, SD= 4.30) scored significantly higher than the SLI 5-6 group as well,

p= .003.

12
 An index of syntactic development
With regard to the expressive subtest of RDLS, the TD 5-6 group (mean= 62.69, SD= 4.22)

scored significantly higher than their normal younger peers (mean= 57.57, SD= 5.02),

p= .024; and the SLI 5-6 group (mean= 54.73, SD= 5.95) respectively, p <.0001. For the

receptive subtest, given that the scores were used as the inclusion criteria for the SLI group, as

expected, the TD 5-6 group (mean= 59.81, SD= 2.86) had significantly higher score than the

TD 4-5 group (mean= 55.64, SD= 3.23), p=.008; and the SLI 5-6 group respectively (mean=

48.47, SD= 4.47) , p< .0001. And the younger TD group scored significantly higher than the

SLI 5-6 group as well, p< .0001.

Results of the Syntactic Complexity Measure, the Component Measures, MLU and D

The descriptive statistics of MLU, D, the index and its component measures (noun phrase

expansion measure [NP], verb phrase expansion measure [VP], sentence structure measure

[SS], connective and sentence adverb measure [CON/SA] and question measure [Q]) of each

group are presented in Table 2.

Table 2. Mean, SD and range of MLU, D, the syntactic complexity index and the subtests of each group

Group TD 4-5 TD 5-6 SLI 5-6

MLU M (SD) 4.73 (0.96) 4.57 (0.60) 4.06 (0.68)

Range 3.47 - 6.59 3.62 – 5.62 2.82 – 5.22

D M (SD) 58.95 (13.08) 66.39a(14.32) 48.76a (8.48)

Range 39.65 – 85.72 40.01 – 90.68 29.33 – 59.90

Index M (SD) 38.64b,c,d (6.81) 47.3c,b,d (9.46) 27.73d,b,c (8.88)

(max= 96) Range 25 – 51 30 – 62 15 – 45

NP M (SD) 3.79e (1.63) 3.88f (1.41) 2.27e,f (1.71)

(max= 12) Range 1–6 1–6 0–5

VP M (SD) 17.5g (3.46) 21.13h (5.06) 12.87g,h (4.27)

13
 An index of syntactic development
(max= 42) Range 10 – 23 11 – 29 7 – 23

SS M (SD) 7.43 (3.72) 9.25i (3.99) 5.2i (3.19)

(max= 18) Range 2 - 14 2 – 14 0 – 10

CON/SA M (SD) 6.57 (3.63) 7.23j (3.17) 3.73j (3.45)

(max =12) Range 0 - 12 4 – 12 0 – 12

Q M (SD) 3.36k (1.69) 5.68 k, l (2.91) 3.67l (1.91)

(max= 12) Range 0–7 1 – 11 0–7

Note: Means having the same superscript differ significantly at p< .05.

NP= Noun phrase expansion; VP= Verb phrase expansion; SS= Sentence Structure;

CON/SA= Conjunction and sentence adverbs; Q= Question

In the following reports, one-way ANOVA results on the abovementioned language

measures will be presented first, followed by the post-hoc comparisons.

Group Comparisons for MLU, D, the Index and the component measures

Differences among the three groups in MLU were approaching significance (F (2,

42)=3.18, p=.052).The mean MLU of the SLI group (mean=4.06, SD=0.86) was lower than

the TD 4-5 group (mean= 4.23, SD= 0.96). Post-hoc analysis confirmed that this is the only

comparison that was approaching significance, p=.055.

With regard to D, significant group difference was found, (F(2, 42)=8.051, p=.001). Post-

hoc comparison revealed that the TD 5-6 group (mean=66.39, SD=14.32) had higher D than

the TD 4-5 group (mean=58.95, SD= 13.08) but the difference was not statistically significant,

p=.232. The TD 5-6 group has significantly higher D than the SLI 5-6 group (mean= 48.76,

SD= 8.48), p=.001. This is the only significant group difference found in D. So the TD 4-5

group did not score significantly higher than SLI 5-6 group, p=.077.

As for the index, significant group differences were found (F(2, 42)=8.07, p=.008). The

14
 An index of syntactic development
post-hoc analyses revealed that, the TD 5-6 group (mean=47.3, SD=9.46) scored significantly

higher than the TD 4-5 group (mean=38.64, SD=6.81) (p= .021). The TD 5-6 group has

significantly higher score than the SLI 5-6 group as well (mean=27.73, SD=8.88) (p< .0001)

With regard to the subtests, significant group differences were found in all the indices

([NP]: F(2, 42)=4.92, p=.012; [VP]:F(2, 42)=13.98, p< .0001; [SS]: F(2, 42)=4.83, p=.013;

[CON/SA]: F(2, 42)=4.54, p=.016; [Q]: F(2, 42)=4.79, p=.013). Results of post-hoc

comparisons are as follows:

For NP, no significant difference was found between the two normal groups in NP

(p= .987). However, TD 5-6 group (mean= 2.27, SD=1.71) scored significantly higher than

SLI 5-6 group (mean= 3.88, SD=1.41) (p=.019). The TD 4-5 group (mean=3.79; SD=1.63)

had significantly higher score than the SLI 5-6 group as well (p=.035). The same pattern was

found across the three groups in VP. The two TD groups did not score significantly different

from each other (p= .070) while TD 5-6 group (mean=21.13; SD=5.06) scored significantly

higher than the SLI 5-6 group (mean= 12.87, SD= 4.27) (p<.0001). The TD 4-5 group

(mean=17.5; SD=3.46) scored higher than the SLI 5-6 group as well (p=.017).

In SS, significant difference was only found between the TD 5-6 group (mean= 9.25, SD=

3.99) and their SLI age peers (mean=5.2, SD=3.19), p=.009. The same pattern of group

difference was also found in SA/CON, i.e. TD 5-6 group (mean=7.23; SD=3.17) scored

significantly higher than the SLI 5-6 group (mean=3.73; SD=3.45) (p=.017). This was the

only significant difference found across the groups in this subtest. The last subtest to be

reported is Q. The TD 5-6 group (mean=5.68; SD=2.91) scored significantly higher than their

normal younger peers (mean=3.36; SD=1.69) (p=.850). The TD 5-6 group had significantly

higher score than the SLI 5-6 group (mean=3.67; SD=1.91) as well (p=.045), whereas no

significant difference was found between the TD 4-5 group and the SLI 5-6 group (p=.928).

The results to be reported next is the correlation between variables of interest

15
 An index of syntactic development
Relationship between Age, MLU and the index

The correlations among age and MLU, index and D were conducted on the 30 children in

two normal groups only.

Relationship of between Age and other language measures (MLU, the index)

Age and MLU weakly, negatively and insignificantly correlated with each other, r(30)= -

.122, p=.522, two-tailed. The negativity can be attributed to the fact that some children from

the TD 5-6 group produced utterances that were on average shorter than those of the TD 4-5

group, resulting a slightly lower mean MLU in the TD 5-6 group compared with that in the

TD 4-5 group ([TD 4-5]: mean=4.73; [TD 5-6]: mean= 4.57). Age and the index mildly

correlated with each other, r(30)=.345, p=.062, two-tailed. The relationship was approaching

significance.

Relationship between the index and MLU

The index and MLU moderately correlated with each other, r(30)= .502, p= .005, two-

tailed. The correlation between them was studied again after removing the effect of age. The

partial correlation was significant, r(30)=.584, p=.001, two-tailed, showing that the

relationship between MLU and the index was significant even if the variance due to age was

removed. The relationship is depicted in Figure 1 in Appendix B.

Since the index exhibited group difference between the three groups and D did so between

the TD 5-6 group and the SLI 5-6 group, discriminant analysis was carried out as proposed to

investigate if a linear combination of age, MLU, D and the index could distinguish the three

groups of children from each other, and the degree of accuracy that these measures can

correctly classify the children into the respective groups. However, since MLU did not show

significant group difference, it was not entered into the analysis.

16
 An index of syntactic development
Discriminant Analysis
Age, the index and D were entered into linear discriminant analysis. Since the index and D

correlated with age, they were entered into the analysis simultaneously. The Box‟s M was

12.01 with significant value of .551, indicating that the groups do not differ significantly from

the covariance matrices. So the analysis could be carried on. The overall Wilks‟s Lambda was

significant, Λ= .194, χ2(6, N = 45) =67.27, p<.0001, showing that the three groups of children

were being successfully differentiated by the three predicators variables (age, D and the

index). The residual of Wilks‟s Lambda was significant as well, Λ= .194, χ2(6, N = 45)

=67.27, p<.0001, indicating that the predicators could differentiate among the groups after

eliminating the effects of the first discriminant function. Since these tests showed statistical

significance, interpretation of both functions was carried out.

The resulting discriminant function equation of the first function was (0.204 x age) + (-

0.038 x index) + (-0.027 X D) -9.982 and that of the second function was (0.082 x age) +

(0.092 x syntactic complexity score) + (0.018 x D) -9.725. The first function was more

strongly correlated to age than to the language measures, as its correlations with age, the index

and D were .951, -.320 and -.331 respectively. The second function had a stronger language

component, with higher contribution from the index (.786) than from D (.215). The

correlation of the second function with age was .380.

Then every child‟s group membership was predicted using the two functions. This leaded

to correct classification of 39 out of 45 (86.7%). The classification accuracy in the TD 4-5

group, the TD 5-6 group and the SLI 5-6 group were 92.9%, 81.3% and 86.7% respectively.

Six cases were misclassified in total. One child in the TD 4-5 group was misclassified into the

SLI 5-6 group; two children out of three in the TD 5-6 group were misclassified into the SLI

5-6 group whereas the remaining one was misclassified into the TD 4-5 group; two children in

the SLI 5-6 group were misclassified into the TD 5-6 group.

17
 An index of syntactic development
To validate the discriminant function, “leave-one-out classification” was done for

reclassification of the same cases. A slightly different result was produced: 38 out of 45

(84.4%). One more case was misclassified. One more children from the TD 5-6 group was

misclassified into the TD 4-5 group.

Given that seven cases were being misclassified after validation, the clinical usefulness of

the composite measure was being evaluated. These evaluation procedures went beyond the

early work. All the misclassified cases which belonged to the two normal groups originally

were regarded as true negative cases, whereas those being misclassified in the SLI group

originally were regarded as false negative cases. The composite measure‟s sensitivity was

83.3%; specificity was 86.7%; positive likelihood ratio was 5.2 whereas negative likelihood

was 0.16.

Syntactic Errors Analysis

The final results to be reported are the syntactic error analysis. The number of tokens and

the percentage of occurrence of each error type in the TD 5-6 group and the SLI 5-6 group

were presented in Table 3.

Table 3. No. of tokens (% of occurrence) of each error type in the TD 5-6 group and the SLI 5-6 group (N=31)

TD 5-6 SLI 5-6

Addition 4 (16%) 12 (21.1%)

Substitution 14 (56%) 23 (40.4%)

Omission 6 (24%) 17 (29.8%)

Error in word order 1 (4%) 5 (8.8%)

Total number of errors 25 57

Among all these error types, substitution was found to be dominating in both groups.

Further analysis revealed that substitution of classifiers was dominant in both groups. So the

content of these classifier substitutes was further analyzed and presented in Table 4.

18
 An index of syntactic development
Table 4. No. of tokens (% of occurrence) of each classifier substitute in the TD 5-6 group and the SLI 5-6 group

(N=31).

Substitute content TD 5-6 SLI 5-6

Default go3 5 (71.4%) 7 (70%)

Other classifier 2 (28.6%) 3 (30%)

- zek3 1 3

- gaa3 1 0

According to the percentage of occurrence, the TD 5-6 group and the SLI 5-6 group used

the default go3 and other classifiers as substitute with more or less the same frequency

respectively.

Discussion
This study aimed to investigate the age-sensitivity of the index of grammatical

development, its validation against other language measure (MLU), its potential to

differentiate the SLI children from their normal peers, its accuracy in giving a diagnostic label

to each child and the error pattern of the SLI children and their age peers. The results were

interpreted so as to address these research questions accordingly.

Age Sensitivity of the Index of grammatical development

The primary aim of this study was to examine whether the index of grammatical

development was sensitive to age or not. The goal was met, as the TD 5-6 group scored

significantly higher than the TD 4-5 group in the measure. The results were supportive to the

previous hypothesis that there was a quantitative difference between the two normal age

groups in syntactic development. The TD 5-6 group scored higher than the TD 4-5 group in all

the subtests but statistical significance was only found in the subtest Q (Question). The

following discussion will investigate the contributing factors to the quantitative difference

observed in both the measure as a whole and each discrete subtest.

19
 An index of syntactic development
Let us recall that the index developed in this study analyzed the children‟s ability in noun

phrase expansion (NP), verb phrase expansion (VP), constructing specific sentence structure

(SS) and question (Q), and employing connectives and sentence adverb (CON/SA). Items in

these subtests include the utilization of closed class words, open class words and construction

of specific sentence forms.

In the two subtests NP and VP, some of the items involved employment of closed class

words, such as classifier, modal auxiliary and aspect marker to elaborate the meaning of noun

or verb phrases respectively (Matthews & Yip, 1994). In the subtests Q and CON/SA,

question words and connectives, which are close class words as well, were used to request

information and connect sentences or ideas respectively. Previous study has shown that five-

year-old children could use more varieties of close class than four-year-old children. For

instance, in the study by Ma (2006), the five -year-old children utilized more varieties of

connectives when compared with the four-year-old peers in spontaneous speech. And the

children of the TD 5-6 group in this study did use more types of connectives than their

younger normal peers. Examples of connectives used by the two groups were summarized in

Table 5.

Table 5. Examples of connectives in English translation used by the TD 4-5 group and the TD 5-6 group

TD 4-5 TD 5-6

Categories Coordinative/Additive tung4maai4 tung4maai4,

of jyut6…jyut6

Connectives Causal jan1wai6, so2ji5 jan1wai6, so2ji5

Concessive bat1gwo3, bat1gwo3, daan6hai6,

daan6hai6, seoi1jin4 kei3 sat6

Conditional jyu4gwo2 jyu4gwo2

20
 An index of syntactic development
Temporal sau2sin1, zung1jyu1

Note: Connectives being underlined were the unique productions of the corresponding

group

The TD 5-6 group used temporal connectives while the TD 4-5 group did not. The TD 5-6

group could produce more types of connectives in the category of coordinative. The findings

echoed the study by Ma (2006) that types of connective production increased by age.

Apart from connectives, the TD 5-6 group produced more types and tokes of simple and

compound directional verbal complement than the TD 4-5 group in this study. The older

group could use six and ten simple and compound directional verbal complements (DVC)

respectively, such as “dit3 lok6 lei4”, “juk1 lei4 juk1 heoi3”, “ling1 hoi1 nei1 gau6 je5”,

“pou5 hei2 nei5”, whereas the younger group could only use five and eight simple and

compound DVCs respectively with smaller tokens. Recall that two examples have to be

expressed in two different forms for crediting two points in the measure (lexical criterion),

and that each items in the measure could be fully credited with three examples. Being able to

use more different types of closed class, the TD 5-6 group could obtain a higher score in the

measure.

Even if the diversity of certain closed class was comparable between the two groups, the

TD 5-6 group could use them more frequently and skillfully than the TD 4-5 group. For

instance, in the study by Fletcher et al. (2000), the 4-year-old children produced more or less

the same number of types of question words compared with the 5-year-old group. Both groups

could produce what-, where-, who-, why- and how-question. However, in this study,

significant group difference was found between these two age groups in the subtest Q. Let us

recall there are three categories in the subtest Q. Producing advanced questions such as why-,

how- and when- question (Q-Wh-2) could receive one more credit than what-, where-, who-,

which- (Q-Wh-1) and yes/no question (Q-Y/N). Further examinations showed that the TD 5-6
21
 An index of syntactic development
group as a whole raised questions more frequently than the TD 4-5 group. Moreover, the older

group produced nearly double the number of Q-Wh-2 than the TD 4-5 group (TD 5-6: 17

times; TD 4-5: 10 times). So the significant group difference between them could be attributed

to the higher frequency of all question words and advanced question words employment. Or

the TD 4-5 group was less skillful and produced more errors in the questions than the older

peers. Erroneous productions could not be credited.

Apart from using the more diversified types of closed class words, the increase in the

types of nouns and verbs, which are open class words, has also attributed to the growth of the

measure from the younger normal group to the older normal group. The study by Fletcher et

al. (2000) found that the five-year-old children could use more types of nouns, transitive and

intransitive verb than those in the four-year-old group. Recall that two examples should be in

distinct surrounding contexts in order to be credited twice (content criterion). For instance, the

two examples “hoi1 zo2 dou6 mun4” and “hoi1 gan2 dou6 mun4” could not be credited twice

for the use of the aspect markers (underlined with dotted line) since both the aspect marker

expand the meaning of the verb “hoi1”, i.e. with the same contexts. Being able to use more

different types of nouns and verbs, the five-year-old normal group can produce exemplars

with higher chance to be credited in subtest NP and VP.

Validation of the index using MLU

Another purpose of this study is to validate the index of syntactic development using

MLU. The correlation between them was moderate (r=.502) with statistical significance. The

correlation strength is weaker than those between grammatical productivity indices in English

and MLU, such as IPSyn in Scarborough (1990) (r=.938); and the syntactic complexity

measure (r=.88) in Blake et al. (1993). The moderate correlation indicated that utterance

length may not always be valid in estimating the index. For instance, a child in the TD 4-5

group with MLU of 5.6 scored 36 in the index of syntactic development, while another one in

22
 An index of syntactic development
the TD 5-6 group with similar MLU (=5.62) scored 62. This phenomenon echoed the findings

of the previous studies that the sensitivity of MLU to grammatical development will decrease

after a mean length of roughly 3.0 is reached (Scarborough, Rescorla, Tager-Flusberg, Fowler

& Sudhalter, 1991; Scarborough, 1990; Rondal, Ghiotto, Bredart & Bachelet, 1987). In the

studies by Scarborough et al. (1991) and Scarborough (1990), the correlations between MLU

and IPSyn were always lower for higher MLU.

And since the MLU and age correlation is also lower in Cantonese sample (r=.44) (Klee

et al., 2004) than in English samples (r=.66) (Klee et al., 2007), MLU might not be a good

index of language growth, especially with the older age group. In the study by Klee et al.

(2004), there were two instances where MLU declined as age increased (from 35.5 to 41.7

months, from 60.2 to 65.7 months). Such decline was found in this study as well, from the TD

4-5 group and the TD 5-6 group. This echoed the conclusion in the study by Scarborough et

al. (1991) that MLU is a less sensitive language measure than IPSyn.

Therefore, using MLU to explain and predict grammatical complexity may not be valid

and adequate enough for children with MLU longer than 3. However, it still can be a useful

validation tool for language measure provided that children are at early language

developmental stage.

Performance of the 5-year-old normal and SLI group on the measure

Given that the index could show grammatical development in the normal groups, the

group difference found between the SLI 5-6 group and their age-matched peers is of interest.

The results indicated that the index could differentiate the two groups from each other and all

the subtests could accomplish the task as well with statistical significance. The results are not

surprising. It has been mentioned previously that lots of studies in Cantonese showed that the

morphosyntatic system of SLI children was not as well-developed as their normal age peers.

Since it is more important on the clinical perspective to find out if an index can make accurate

23
 An index of syntactic development
diagnosis than if the groups performed differently on it (Wong et al., 2010), the discussion

will proceed to the diagnostic accuracy of the linear combination of age, D and the index.

Diagnostic accuracy of the linear combination of age, D and the index

Recall that the discriminant analysis correctly classified 25 of out 30 cases in the two

normal groups and 13 out of 15 in the SLI group. One less child in the normal group was

correctly classified in the “leave-one-out” validation procedure. The sensitivity and specificity

of the index were 83.3% and 86.7% respectively. Sensitivity and specificity levels of 90% and

above are regarded as good and 80% as fair (Plante & Vance, 1994). Upon these two indices,

it can be concluded that the diagnostic accuracy of the composite measure is fair.

To evaluate the clinical usefulness of the composite measure as a diagnostic or screening

tool, the positive and negative likelihood ratio (LR+ and LR-) should be considered. The LR+

and LR- of the composite measure were 5.2 and 0.16 respectively. According to Dollaghan

(2007), a clinically-useful instrument should possess a LR+ greater than 10 and a LR- smaller

than 0.1. So the test cannot achieve such purpose.

Although the clinical usefulness of the composite measure could not be confirmed from

LR+ and LR-, it is yet a potential clinical marker of SLI since its sensitivity and specificity

levels were fair. Replication of the study in an independent sample will be necessary to

confirm the findings.

Error pattern
The last research question to be addressed is the comparison of error tokens between the

TD 5-6 group and the SLI 5-6 group. As a group, the SLI group produced more errors than the

normal group during conversation. This confirms the result of a study by Dunn et al. (1996)

that SLI children produced more syntactic errors in spontaneous speech than normal children.

Classifier substitution error will be the principal discussion area in this session as it was found

to be dominating in both groups. Both groups used a similar proportion of default go3 (TD 5-

24
 An index of syntactic development
6: 71.4%; SLI 5-6: 70%) and other classifiers as substitute (TD 5-6: 28.6%; SLI 5-6: 30%) of

all their errors. However, it should be noted that only one child in the SLI group used other

classifiers as substitutes and that child used zek3 in all three occasions of substitution. So it is

quite possible that the SLI child was using zek3 as his own default substitute. Therefore, it can

be concluded that the SLI children used general classifiers as substitute more frequently than

the normal group. The results did not support the findings of the study by Stokes et al. (1997),

which concluded that language-impaired children as a group used other classifier substitutes

more frequently than the normal age-matched peers. The discrepancy may be due to the

different methods adopted by the two studies. The study by Stoke et al. (1997) elicited

children‟s production of classifiers using questions or sentence completion, while errors of

classifier use were noted during conversation in the present study. If children had to assemble

their own sentences during conversation, they behaved differently in two studies.

Future Modification

Given that the index was able to illustrate developmental difference between two age

groups while such difference could not be observed in most of the subtests, it should be

modified to develop an index of syntactic development with higher sensitivity to grammatical

complexity. The modification can be done on the selection and categorization of items,

crediting system of the index and implementing the index on more participants.

In terms of selection, certain grammatical items may not be good indicators of syntactic

development as both normal groups used them with equal abilities. These items were not

eliminated in the pilot study as previous studies have shown that the older group either

produced more types of that item or produced that item more frequently than the younger

group. For instance, in the study by Fletcher et al. (2000), the five-year-old group produced

similar number of types of aspect markers as the four-year-old group but the older group used

them more frequently than the younger group. So this item was kept in the index. However,
25
 An index of syntactic development
developmental trend could not be observed in the aspect marker usage in this study. Similar to

the study by Fletcher et al. (2000), both the TD 4-5 and the TD 5-6 groups could use the same

number of types of aspect marker (zo2, gan2, zyu6, gwo3 and haa5). But unlike the previous

study, the two groups produce comparable tokens of the item (TD 4-5: 49 tokens; TD 5-6: 48

tokens). No subtle grammatical development can be shown by this item. It should be

eliminated if the study is to be replicated on children of the same age groups.

With regard to the further modification of the crediting system, more items should have

adopted a crediting hierarchy. For some of the items in the subtests, such as aspect marker,

modal auxiliary and question words, more credits were given to the exemplars which were

known to emerge later in the developmental milestone. For instance, usage of question word

“dim2gaai2” (why) received 1 more credits than question word “mat1je5” (what). But such

system was not applied throughout the index since the pilot study did not cover every sample

employed in this study. Therefore, certain developmental trends could not be identified. After

every language sample had been analyzed, it was found that the crediting hierarchy should

have been adopted in connectives, classifiers and prepositional phrase as well. For instance,

referring to Table 5, temporal connectives were only used by the 5-year-old children. In the

study by Ma (2006), the five-year-old children used more temporal connectives than the four-

year-old group and they first emerged when a child was 4-year-old. These findings justified

that more credits should be given to temporal connectives, or less credits should be given to

other common connectives equally used by these two groups, such as jan1wai6 (because),

tung4maai4 (and). For prepositional phrase, it was found that the four-year-old group received

credits from this item using hai2- and tong4- construction only. However, the five-year-old

group was using more varieties of prepositional phrase such as zoeng1-, dang2- and caang1-

construction. Again, more credits should be given to the late emerging items as well.

26
 An index of syntactic development
Clinical implication

Though the composite measure is not a clinically-useful instrument according to the value

of the LR+ and LR-, the fact that the diagnostic accuracy of the measure is fair gives insight

into the development of assessment procedures of the SLI children. Spontaneous speech

samples may supplement the diagnosis of SLI children and compensate the drawbacks of

psychometric tests. There have been lots of controversies regarding the reliance on

psychometric tests. These tests may tend to emphasize a single aspect of language (American

Speech-Language-Hearing Association, 1989). For instance, the most complex language form

being assessed in the RDLS-R is multiple-step command embedded with locative and colour

concepts. No narrative element is included. For the RDLS-E, the most complex production

required is at sentence level which focuses on the semantic content of sentences and their

form. Any cohesiveness in daily dialogue and other uses of the language (e.g. request) are not

assessed.

To ascertain the value of language sample analysis in assessing preschoolers‟ language

ability, the above-mentioned improvements of the index should be take into account and

replication of the study can help. The listing of structures that makes up the index provides a

framework for examining language development for descriptive purpose over time. Once can

examine if a child has made progress in using a more complex form of an item, or using a

form more productively, during conversations or elicited productions.

Acknowledgement

I would like to thank my adviser, Dr. Anita Wong for her precious time and invaluable advice

on this research project. This project will be impossible to be finished without her guidance

and encouragement. My heartfelt appreciations go to Miss Feiona Man Ka Yee as well for her

contribution of time to be the inter-rater in this study. Finally, I would like to thank my fellow

classmates for their words of emotional support over the course of my study.
27
 An index of syntactic development
References:

American Speech-Language-Hearing Association. (1989). Issues in determining eligibility for

language intervention. Asha, 31, 113-118.

Brown, R. (1973). A first language: the early stages. Cambridge, MA: Harvard University
Press.

Blake, J., Quartaro, G. & Onorati, S. (1993). Evaluating quantitative measures of grammatical
complexity in spontaneous speech samples. Journal of Child Language, 20, 139-152.

Cheung, S. L. (1991). The notion of “Result” in Cantonese Children. Papers and Reports on
Child Language Development, 30, 17-24.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences. (2nd ed.). Hillsdale,
N.J.: L. Erlbaum Associates.

Costanza-Smith, A. (2010). The clinical utility of language samples. Perspectives on language

learning and education, 17(1), 9-15.

Crais, E. R. (1995). Expanding the repertoire of tools and techniques for assessing the
communication skills of infants and toddlers. American Journal of Speech-Language
Pathology, 4, 47-59.

Dollaghan, C. A. (2007). The handbook for evidence-based practice in communication

disorders. Baltimore, MD: Brookes.

Dunn, M., Flax, J., & Sliwinki, M. (1996). The use of spontaneous language measures as
criteria for in identifying children with Specific Language Impairment: an attempt to
reconcile clinical and research incongruence. Journal of Speech and Hearing Research, 39,
643 – 654.

Fletcher, P., Leonard, L. B., Stokes, S. F., & Wong, A. M.-Y. (2005). The expression of
aspect marker in Cantonese-Speaking children with Specific Language Impairment.
Journal of Speech, Hearing and Language Research, 48, 621-634

Fletcher, P., Leung, C. S., Stokes, S., & Weizman, Z. (2000). Cantonese pre-school language
development: A guide. Hong Kong: University of Hong Kong, Department of Speech and
Hearing Science.

Hong Kong Society for Child Health and Development (1987). Manual of the Reynell
Developmental Language Scales, Cantonese (Hong Kong) version. Hong Kong-SAR-
China; Author.

Klee, T., Gavin, W. & Stokes, S. (2007). Utterance length and lexical diversity in American-
and British-English speaking children: What is the evidence for a clinical marker of SLI.
In R. Paul (Ed.), Language disorders from a developmental perspective: Essays in Honor
of Robin S. Chapman. Mahwah, N. J.: London: Lawrence Erlbaum Associates, Inc.

28
 An index of syntactic development
Klee, T., Scahffer, M., May, S., Membrino, I., & Mougey, K. (1989). A comparison of the
age-MLU relation in normal and specifically language-impaired preschool children.
Journal of Speech and Hearing Disorders, 54, 226-233.

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 Impairment. Journal of Speech, Language and Hearing Research, 47, 1396-
1410.

Lahey, M., & Bloom, L. (1994). Variability and language learning disabilities. In K. G. Butler
(Ed.), Language learning and disabilities in school-age children & adolescents: Some
principles and applications. Needlham Heights, MA: Allyn & Bacon.

Leonard, B. L., Wong, A. M.-Y., Deevy, P., Stokes, S., & Fletcher, P. (2005). The production
of passives by children with SLI acquiring English or Cantonese. Applied Linguistics, 27,
267-299.

Leung, C. S. (1995). The development of aspect markers in a Cantonese-speaking child

between the ages of 21 and 45 months. Unpublished doctoral dissertation, University of
Hawaii, Manoa.

Matthews, S. & Yip, V. (1994). Cantonese: A Comprehensive Grammar. London: Routledge.

Miller, J. F. (1981). Assessing language production in children: Experimental procedures.

Baltimore: University Park Press.

Miller, J. F. & Chapman, R. S. (1981). The relation between age and mean length of utterance
in morphemes. Journal of Speech and Hearing Research, 24, 154-161.

Miller, J. F. (2005). Language Sample Analysis: The Wisconsin Guide. Madison, Wisconsin:
Wisconsin Dept. of Public Instruction.

Owen, A. J., & Leonard, L. B. (2002). Lexical diversity in the spontaneous speech of children
with specific language impairment: Application of D. Journal of Speech, Language and
Hearing Research, 45, 927-937.

Paul, R. (2007). Language Disorder From Infancy Through Adolescence: Assessment and
Intervention. (3rd ed.). St. Louis, Missouri: Mosby Elsevier.

Plante, E. & Vance, R. (1994). Selection of preschool language tests: A data-based approach.
Language, Speech, and Hearing Services in Schools, 25, 15-24.

Richards, B. J., & Malvern, D. D. (1997). The New Bulmershe Papers: Quantifying lexical
diversity in the study of language development. Reading, England: University of Reading.

Richards, B. J., & Malvern, D. D. (1998). A new research tool: mathematical modeling in the
measurement of vocabulary diversity (Award reference no. R000221995). Final Report to
the Economic and Social Research Council, Swindon, UK.

29
 An index of syntactic development
Rondal, J. A., Ghiotto, M., Bredart, S., & Bachelet, J. (1987). Age-relation, reliability and
grammatical validity of measures of utterance length. Journal of Child Language, 14, 433-
446.

Scarborough, H. S. (1990). Index of Productive Syntax. Applied Psycholinguistics, 11, 1-22.

Scarborough, H. S., Rescorla, L., Tager-Flusberg, H., Fowler, A. & Sudhalter, V. (1991). The
relation of utterance length to grammatical complexity in normal and language-disordered
group. Applied Psycholinguistics, 12, 23-45.

Stokes, S. F. & Fletcher, P. (2000). Aspectual forms in Cantonese children with specific
language impairment. Linguistics, 41(2), 381-405.

Stokes, S. F. & So, L. K.-H. (1997). Classifier use by language-disordered and age-matched
Cantonese-speaking children. Asia Pacific Journal of Speech, Language and Hearing, 2,
83-101.

To, C. K.-S., Stokes, S. F., Cheung, H.-T., & T‟sou, B. (2010). Narrative assessment for
Cantonese-Speaking Children. Journal of Speech, Language and Hearing Research, 53,
648-669.

Tse, S.-K., Hui, L. & Leung, S.-O. (2007). The acquisition of Cantonese classifiers by
preschool children in Hong Kong. Journal of Child Language, 34, 495-517.

Wong, A. M.-Y., Klee, T., Stokes, S. F., Fletcher, P. & Leonard, L. B. (2010). Differentiating
preschool children with and without SLI using MLU and lexical diversity (D). Journal of
Speech, Language, and Hearing Research, 53, 794-799.

Wong, A. M.-Y., Leonard, L. B., Fletcher, P., & Stokes, S. F. (2004). Questions without
movement: A study of Cantonese-speaking children with and without specific language
impairment. Journal of Speech, Language, and Hearing Research, 47, 1440-1453.

Zhang, H. N. (2007). A Grammar of Cantonese as Spoken in Hong Kong. Hong Kong: The
Chinese University of Hong Kong.

30
 An index of syntactic development
Appendix A
Examples and definitions of the syntactic structures examined in this study

The definitions and some of the examples of the items were from To et al., (2010), Zhang,

(2007), Leung, (1995), Matthews & Yip, (1994).

A. Noun Phrase Expansion (NP)

Definition: The meaning of a noun phrase can be modified and expanded by demonstrative,

articles, adjectival modifier (including possessives and quantifiers) and relative clause. In this

study, only structures exist in Cantonese linguistics and start emerging in the developmental

stages of participants will be credited. Their definition and examples are as follows:

(NP-CL) Classifier (loeng4 ci4)

Definition: Each noun has a specific assigned classifier. There are mainly two types of

classifiers: measural classifiers and sortal classifiers. The former specifies quantity of the

head noun while the latter denotes the intrinsic quantity (e.g. shape) of the object.

Examples: yat1 baan1 yan4 (measural classifiers)

One CL people (a group of people)

sam1 tiu4 jy2 (sortal classifiers)

Three CL fish (Three fish)

(NP-A) Adjectival modifier

Definition: They are adjectives which proceed and modify meaning of a noun phrase.

Example: daai6 fei1 kei1

Large airplane (large airplane)

31
 An index of syntactic development
(NP-RC) Relative clause

Definition: Relative clause is predicative clause elaborating a noun phrase. It always serves

to restrict the identity of a noun phrase. It is ahead of the noun phrase and can be connected to

it with linking particle ge3 (LP) or go2 go3.

Example: ngo5dei6 soeng5 tong4 go2zan6si4 waan2 caai2 daan1ce1 gaa3

We have lesson when play ride bike PRT

(We rode on the bike when we were having lesson)

B. Verb Phrase Expansion (VP)

Definition: According to Matthews and Yip (1994), modal auxiliary, predicate adverb,

manner predicate (gum2-construction) can serve the same function as well. Clausal

complement can modify the meaning of perceptual and mental verbs (To et al, 2010).

Verbal Complement (wai1bou2)

Definition: The former is an action or description which is described by the latter complement.

The complement can be a verb or an adjective. A verb particle may be present.

Different types of verbal complement

(VP-RVC) Resultant complement

Definition: The second predicate indicates the result of the first predicate.

Example: sai3 lo2 zing2 laan6 sai3 di1 syu1

Younger brother make broken all PRT book

(Younger brother has broken all the books)

(VP-DVC) Directional complement

Definition: Simple and compound direction complement can be assembled using these items:

lei4 (come) , heoi2 (go) soeng5 (up), lok6 (down), hoi1 (away), mai4 (close), ceot1 (out), jap6

(in), gwo3 (over), hei2 (up), dou6 (here), faan1(back)

32
 An index of syntactic development
Example: Simple direction complement

han4 seong5 lok6 lau2

walk up six floor

(walk up to the sixth floor)

Compound direction complement

cung1 lok3 heoi6 wing6 ci4

rush down come pool

(Rush into the swimming pool)

(VP-OVC)Other verbal complement

I. Phase complement

Definition: The second predicate indicate the special status of the first predicate. The second

predicates are always zoek6, dou2, hei2, gwo3

Example: keoi5 fan3 zoek3 zo2

He/she sleep finish ASP

(He/she is asleep)

II. Intensifying complement

Definition: An adverb is added before a stative verb to intensify the status of that

adjective/verb

Example: nau sei2 nei5

angry die (as intensifier) you

(Very angry with you)

III. Potential complement

Definition: It describes the possibility of an event or action. An infix is inserted into the

complement to express the possibility. The adverb „dak1‟ indicates positivity while „m4‟
33
 An index of syntactic development
indicates negativity.

Example: faan3 m4 zoek3

Sleep NEG finish (Cannot be asleep)

sik6 dak1 baau2

eat ADV full ((Somebody is) full)

IV. Descriptive complement

Definition: Its position is less restricted that it does not necessarily follow the predicate

immediately.

Example: keoi5 ceong3 dak1 hou2 hou2 teng1 aa3

He/she sing ADV good good listen PRT

(He/she sings very well.)

V. Complement of Extent

Definition: It expresses the extent of the action in the predicate using the special structure

“predicate + dou3”

Example: siu3 dou3 lok1 dei2

laugh V-PRT trundle floor

(Somebody laughs to the extent that he/she trundle along the floor)

(VP-AS) Aspect Markers

Definition: Aspect marker indicates the duration of time for which an event lasts while does

not specify the exact time that the event happens. Aspect markers in category VP-AS-1 appear

before 3 years and 3 months old (Leung, 1995).Their emergence is earlier than those in

category VP-AS-2.

34
 An index of syntactic development
Different types of aspect marker:

(VP-AS-1)

I. Perfective marker zo2

II. Imperfective marker gan2, zyu6

III. Experiential marker gwo3

(VP-AS-2)

IV. Habitual hoi1

V. Inchoative hei2 (seong6) lai4

VI. Delimitative haa5

Example: tai2 gan2 din6si6

Watch ASP TV

((Somebody is) watching TV)

(VP-VA) Verbal Adverbs

Definition: Verbal adverbs are used to elaborate and describe verbs, with respect to their

extent, manner, tone, time and effective area etc. So their linguistic functions are very limited.

Example: m4 tung1 yat1 zaai1 fan3 me1

Is-it-possible together sleep PRT

(Is it possible to sleep together?)

(VP-MA) Modal Auxiliary

Definition: Modal verbs are used to express the extent of possibility, ability, wishes,

obligation and other related concepts. They appear in front of a verb phrase.

(VP-MA-1)

I. wui3 (will/would)

II. ho2 ji5 (can, may)

35
 An index of syntactic development
III. yiu3 (want)

IV. seong2 (wish to)

(VP-MA-2)

I. zeon2 (allow)

II. jing1 goi1 (should , ought to)

III. m4 sai2 (no need)

IV. ling4 jyun4 (rather)

V. sik1 (know)

VI. bei2 (let)

VII. zung1ji3 (love, like).

Example: ngo5 soeng2 fan3 kaau3

I want sleep

(I want to sleep)

m zeon2 nei5 waan2

not allow you play

((I) Don‟t let you play)

(VP-CC) Clausal Complement

Definition: They are predicative or verbal clausal complement appearing after perceptive or

mental verbs. Clauses following verbs such as hoi1 ci2 (begin), zeon2 bei6 (get ready), gai3

zuk6 (continue) are also regarded as clausal complement.

Example: ngo5 zeoi3 geng1 sik6 laak6 jyu2

I most scared eat spicy fish

(I am most scared of spicy fish)

36
 An index of syntactic development
(VP-MP) Manner predicate

Definition: It appears before a verb to indicate or explain the manner of the action. It can be

indicated and preceded by adverbial gum2 or gum2 jyn2.

Example: koei2 ho2 daai6 lek6 gum2 daa2 ngo5

He/she very big force thus hit me

(He/she hits me hard.)

ngo5 dei6 jat1 cai4 co5 aa1

We together sit PRT

(We sit together.)

C. Sentence Structure (SS)

(SS-SV) Serial Verb Construction (lin4wai6)

Definition: The subject is connected to two or more verbs without any intervening

conjunction in between them.

Example: ze4 ze1 caak6 a4 sai2 min6

Elder sister brush teeth wash face

(Elder sister brushes teeth and washes face.)

(SS-PC) Pivotal construction (gim1jyu5)

Definition: The pivotal noun phrase (the underlined NP in the examples) serves as the object

of the former verb and the subject of the latter verb simultaneously.

Example: dong3 zek3 sau2 hai6 jan4 aa3

pretend CL hand is person PRT

(Pretend that the hand is a person.)

37
 An index of syntactic development
(SS-PP)Prepositional phrase

Definition: Prepositional phrase usually appears before the verb(dotted lined) (there are

exceptions such as hai2-construction). The phrase comprises a preposition and a noun phrase.

(The verb is dotted underlined, the preposition is doubly underlined and the noun phrase is

singly underlined in the examples)

I. bei2-passive construction

dai4 dai2 bei2 ze4 ze1 cou4 seng2 zo2

younger brother by elder sister noise wake ASP

(The younger brother is waken by the noise made by the elder sister)

II. zeong1-construction

zeong1 syu4 zai2 fong3 lok6 heoi3 zyu2

place potato put down go cook

(To put the potato into something and cook)

III. deoi3-construction

zek6 gau2 deoi3 zyu6 go3 caak6 fai6

CL dog face ASP CL thief bark

(The dog bark face-to-face to the thief.)

IV. lin4-construction

lin4 sai3 lo2 dou1 gok3 dak1 m4 gau3 baau2

Even younger brother also feel NEG enough full

(Even the younger brother is not full)

V. hai2-construction

zek3 gau2 hai2 toi2 dai2 dou6 fan3 gaau3

CL dog at table bottom sleep

(A dog sleeps under the table.)

38
 An index of syntactic development
VI. tong4-construction

tong4 keoi5 dei2 waan4

with they play

(Play with them.)

D. Question

(Q-YN) Yes/No question

Definition: There are seven types of yes/ no questions. They are particle questions, A-not-A

questions, copula questions, mei6 questions, existential questions, tag question and echo

question. Only those used in the samples will be listed. Their definitions and examples are as

follows:

I. Particle question

Definition: It is the simplest question form by addition of a particle to a declarative

sentence.

Example: mat1 nei5 m4 zi1 gaa3 me1?

What you NEG know PRT PRT

(Don‟t you know?)

II. A-not-A questions

Definition: The construction of this question involves a negative morpheme m4 in

between a repetition of the verb or adjective

Example: nei5 zung1-m4-zung- ji3 sik6 tong aa3?

You like-not-like eat sweet PRT

(Do you like sweet?)

III. Copula questions

Definition: Copula question is signified by copula verb hai6 mai6 which precedes the

39
 An index of syntactic development
man verb. hai6 mai6 is a kind of explicit A-not-A

Example: nei5 hai6-mai6 yiu3 aa3?

You be-not-be want PRT

(Do you want it?)

IV. mei6-questions

Definition: Addition of the negative morpheme mei6 to a declarative sentence can form a

question about whether an event has occurred or not.

Example: nei3 sik6 zo2 faan6 mei6 aa3?

You eat ASP food not-yet PRT

(Have you had your meal yet?)

V. Existential questions

Definition: Existential question uses the words yau5 mou5 to ask for existence of an

object.

Example: le1 dou6 yau5 mou5 ce1 ce1 aa3?

PRT here have-not-have car car PRT

(Is there any car here?)

Wh-question

Definition: The wh-word appears in similar order as in an English question if it is a subject

question. However, the wh-word acts as an object and appears after the verb like a direct

object in an object question.

Subject question Object question

Bin1go3 wan2 ngo5 a3? nei2 wan2 bin1go3 a3

Who seek me PRT you seek who PRT

(Who is looking for me?) (Who are you looking for?)

40
 An index of syntactic development
(Q-Wh-1) Question word Type 1

Definition: The wh-question words in this category emerge from 24 to 30 months old.

I. ‘Who’ question – bin1go3

Example: Bin1go3 daa2 din6waa2 lei4 a4?

who call-phone come PRT

(Who was it that called?)

II. ‘What’ question – mat1ye5

Example: Sik6 di1 mat1ye5 ho2 a?

Eat SL what good PRT

(What shall we eat?)

III. ‘Which’ question – bin1 + CL + noun

Example: Nei5 gok3dak1 bin1 zek6 gau2 zui3 leng3 a4?

You feel which CL dog most nice PRT

(Which dog you think is the nicest?)

IV. ‘Where’ question – bin1dou6/ bin1syu3

Example: ngo5 dei6 heoi3 bin1dou6 waan2 a4?

We go where play PRT

(Where shall we go for dim sum?)

(Q-Wh-2) Question word Type 2

Definition: The wh-question words in this category emerge from 30 to 48 months old.

I. ‘When’ and time question

Example: nei2 gei2si4 faan1 lei4 a4?

You when return come PRT

(When will you be back?)

41
 An index of syntactic development
II. ‘Why’ and reason question

Example: dim2gaai2 wui2 gam2 ge?

How come would thus PRT

(Why is it like this?)

III. ‘How’ question

Example: nei2 dim2 yeung2 jing2 saa1loet2 gaa3?

You how way make salad PRT

(How do you make salad?)

E. Sentence adverbs and connectives

Sentence adverbs (SA)

Definition: Sentence adverbs are distinct from verbal adverbs in the way that the former

modify the meaning of the whole sentence instead of the verb phrase or the predicate only.

Their position are less restricted than verbal adverb in the way that they can be positioned (a)

between the subject/topic and the verb; (b) at the beginning of the sentence; (c) at the end

of the sentence, by way of afterthought.

Example: ngo5 waak6 ze2 m4 lei4 dak1

I perhaps no come able

(I may not come)

Sentence adverbs can be classified according to their functions. The functions include time

(present), such as yi1 gaa1 (now); time (future), such as haa6 ci3 (next time); frequency, such

as do1 sou2 (by that time), gan1 zyu6 (next); quantity, such as zing6 hai6 (only);

presupposition, such as dim2 zi1 (out of expectation)

42
 An index of syntactic development
Connectives/Conjunctions (CON)

Definition: There are two major categories of conjunction. The first one is used to connect

two words or even larger linguistics units, such as phrases or clauses. They are used to

reveal the relations between the two units. The second one is used to connect idea other than

the main clause of the speaker. It can be used to connect idea of the previous speaker.

There are different types of connectives classified into coordinative, clausal, concessive,

hypothetical and temporal.

Example: maa4 maa1 tung1 ze4 ze1 dou1 heoi3 gaai1

Mother and sister also go street

(Mother and sister also go out.)

Daan6 hai6 yiu3 maai5 faan1 lei4 lo1

But need buy back come PRT

(But (we) have to buy it back.) (A concessive connectives)

43
 An index of syntactic development

An index of syntactic development – Charting Sheet

Items Description Exemplar1 Exemplar2 Exemplar 3

(credit/s)

Noun phrase elaboration /12

NP-CL(1) Classifier

NP-A(1) Adjectival modifier

NP-RC(2) Relative Clause

Verb phrase elaboration /42

Verbal complement

VP-RVC(1) Resultant Verb Complement

VP-DVC(1) Directional Verb

Complement

VP-OVC(1) Other verbal complement

Aspect Marker

VP-AS-1(1) Perfective marker: zo2

Imperfective marker: gan2,

zyu6

Experiential marker: gwo3

VP-AS-2(2) Habitual: hoi1

Inchoative: hei2 (seong6)

lai4

Delimitative: haa5

VP-VA(1) Verbal Adverb

44
 An index of syntactic development
Modal Auxiliary

VP-MA-1(1) wui3, ho2 ji5, yiu3, seong2

VP-MA-2(2) zeon2, jing1 goi, m4 sai2,

ling4 jyun4, sik1, bei2,

zung1ji3

VP-CC(2) Clausal Complement

VP-MP(2) Manner predicate

Sentence structure /18

SS-PC(2) Pivotal construction

SS-SC(2) Serial verb construction

SS-PP(2) Prepositional phrase

Questions /12

Q-YN(1) Yes/No question

Wh-question

Q-Wh-1(1) What, who, where, which

Q-Wh-2(2) How, why, when

Sentence Adverb and connectives /12

SA(2) Sentence adverbs

CON(2) Connectives

45
 An index of syntactic development
Appendix B
Figure 1. Scatter plot of the index of grammatical development and MLU, for typically developing Cantonese

speaking children (N=30)

60.00
Index of syntactic development

50.00

40.00

30.00

3.00 4.00 5.00 6.00 7.00

MLU100

46