Chapter 3 Method
3.2 Experimental Tasks
In this study, we tested children with their expressive vocabulary, receptive vocabulary, phonological perception, phonological production and NWR. In addition, we tested children’s nonverbal intelligence with the Leiter-R as a control for their cognitive abilities. However, concerning the broad age ranges we covered in this study, some constructs were tested with different tests that were more appropriate for the age.
For example, vocabulary knowledge among children at age 3 or above was measured with the Receptive and Expressive Vocabulary Test (REVT, for expressive vocabulary) and Peabody Picture Vocabulary Test-Revised (PPVT-R, for receptive vocabulary).
However, for children below age 3, their expressive vocabulary and receptive vocabulary were tapped with the Mandarin-Chinese Communicative Development Inventory (MCDI-T). Also, on account of younger children’s smaller attention span, we downsized the number of trials children below age 3 had to complete in some of the tasks. These adjustments would be specified in the following introduction of the tasks.
3.2.1 Mandarin-Chinese Communicative Development Inventory (Taiwan)
(MCDI-T). The MCDI-Taiwan (Liu & Tsao, 2010) is a parent-report evaluation of
infants’/toddlers’ word production, communicative functions, sentence complexity, and44
the mean length of the three longest utterances. This assessment was applied in this study to collect the data on very young children’s expressive vocabulary size. Since there was no parallel measure of receptive vocabulary for children below 3 in Mandarin, this form was also used to collect our children’s receptive vocabulary.
3.2.2 Peabody Picture Vocabulary Test-Revised (PPVT-R). The PPVT-R (Lu &
Liu, 1994) is often used to measure age 3 to age 12 children’s receptive vocabulary knowledge. Different from the MCDI-T, this test is administered by the experimenter. In this study, it was also applied to test children’s receptive vocabulary at age 2.5.
The test has form A and form B. In this study, form A was adopted. In the test, the child responded by pointing to one of four line drawings that corresponds to the word spoken by the experimenter. The test has 125 items. The child started from the item appropriate to his or her age. Testing was stopped when the child made 6 errors in 8 consecutive items. The total number of items the child accurately answered was counted as his or her PPVT-R score.
3.2.3 Receptive and Expressive Vocabulary Test (REVT). The REVT (Huang,
Jian, Zhu, & Lu, 2011) is designed for children at the age from 3 to 6, and consists of the scale of receptive vocabulary and the scale of expressive vocabulary. Each scale is comprised of four subtests, including label, category, definition, and reasoning. This test takes into consideration the specific linguistic features of Mandarin, in comparison with the linguistic structure of English. We adopted the scale of expressive vocabulary to assess children’s ability to encode or express language based on concept.3.2.4 Productive phonology task. This task tapped the children’s productive
phonology, which reflected the quality of phonological representation (Anthony et al.,45
2010). Colored pictures of familiar objects were used to elicit the children’s production of 21 Mandarin syllable-initial consonants. Since words in Mandarin are mostly disyllabic, the position constraints were considered by testing all the target phonemes in the first syllable and the second syllable (see Appendix B).
In order to promote young children’s participation, we also designed a
“find-the-figure” game to elicit children’s production. Five pictures were created incorporating the figures of the target words. Children were given pieces of the figures, and were asked to paste each to its corresponding figure on the pictures. While pasting, children were required to name the figure.
Each child would be given a score on onset production and a score on rhyme production. Onset and rhyme were scored separately because previous studies suggest that they play asymmetric roles in speech processing and language acquisition (Nazzi, 2005; Nespor, Peña, & Mehler, 2003). With regard to the scoring of onset, we first calculated the accuracy rate of the target phoneme in each lexical item; then we took the average of the accuracy rates across all the lexical items that contained the target phoneme. For example, for the phoneme /b/ at the word initial position, one child produced bei1-zi (cup), bai2-tu4 (white rabbit), and bang1-mang2 (help). For the lexical item bei1-zi, 5 out of his 6 productions of /b/ were accurate; therefore, the accuracy rate of /b/ in this lexical item was 0.83. The accuracy rates for /b/ in bai2-tu4 and
bang1-mang2 were 0.7 and 1, respectively. Therefore, the average accuracy rate of /b/ at
word-initial position was 0.8433. One point was given when the target phoneme was pronounced correctly in all the lexical items that contained it. The children’s systematic mispronunciations for a single phoneme were considered inaccurate but were noted for scoring NWR. Children’s scores of each of the phoneme at the word-initial position and the non-word-initial position were summed up to yield a productive phonology score, with the maximum score of 42.46
The scoring of rhymes followed a similar procedure. We first calculated an average accuracy rate for each of the rhymes. Then we took the average of the rhymes with the same rhyme structure. For example, we averaged the accuracy rates of the rhymes ai, ei,
ao, and ou to get a score of the VG rhyme structure. There are 6 rhyme structures in
Mandarin, including V, VN, VG, GV, GVN, and GVG. The maximum score for rhyme production was 6.3.2.5 Word discrimination. This task was designed to tap children’s
discrimination ability at the phonemic level. There were four practice trials and 24 test trials. The test trials were composed of 6 pairs of sound contrasts. Two pairs of the sound contrasts differed in three features (3-f difference, e.g. t-m and b-q); another two pairs differ in two features (2-f difference, e.g. h-p and x-j); and the other two pairs differ in only one feature (1-f difference, e.g. d-g and n-l) (see Appendix C for the stimuli). A female Taiwan Mandarin speaker produced spoken version of these stimuli in a sound-attenuated booth, and recording were made by using a DAT-recorder.In this task, children were instructed to answer “yes” or “no” in response to the correspondence between the picture they saw and the audio label they heard. For example, for the n-l pair, children would see a picture of a bird. In the “yes” trial, the children were asked “zhe4-li3 you3-mei2-you3 niao3? (Is there a bird here?)” However, in the “no” trial, the children were asked “zhe4-li3 you3-mei2-you3 liao3? (Is there a knotweed?)”
The test was divided into four blocks. The presentation order of the six pairs of stimuli would be counterbalanced across the four blocks. Therefore, for each sound pair, there would be four times of discrimination, in which each of the sound stimuli would appear twice in the audio, and the presentation order of the sound stimuli would be counterbalanced. However, considering that younger children might have shorter
47
attention span, children below age 3 only had to complete the first two blocks.
Each accurate response was credited with one point. The maximum score for the task was 24. However, for children at age 2 and age 2.5, the maximum score was 12.
3.2.6 Nonword repetition. Two types of nonwords were constructed based on the
characteristics of Mandarin phonology, following the Mandarin NWR studies (Lee, 2005; Li & Cheung, 2014). The first type is nonce words, which are nonsense words consisting of two existing syllables. Though a nonce word cannot be mapped onto any existing semantic representation, each of its constituent syllables can be mapped onto a lexical representation. The second type is gap words, which are formed by conjoining two phonotactically legal but non-existing syllables. Gap words are absent in real-life language use and therefore have no stored representations at either the semantic level or the lexical level. Based on their compositions, the two types of nonwords are distinctive in their lexicality. While nonce word repetition is supposed to be more related to vocabulary knowledge, the repetition of gap words may tap phonological analysis.Some gap syllables may be real syllables in other dialects in Taiwan, and children’s familiarity with these dialects may affect their performance. In fact, we have checked the lexical status of the Mandarin gap syllables in Southern Min and Hakka, two of the major dialects in Taiwan. It is found that thirty-three syllables out of the 108 gap syllables are real syllables in Southern Min (despite the fact that there might be subtle difference in the actual pitch of the tone). As in the case of Hakka, sixty of the Mandarin gap-syllables were real syllables in Hakka (also disregarding the subtle difference in the actual pitch of the tone).
The nonwords were composed of early-acquired consonants only (Cheung, 2000;
Zhu & Dodd, 2000) to avoid articulatory difficulties. The NWR task contained 36 disyllabic nonce words and 36 disyllabic gap words. Each type of nonword was equally
48
divided into the one-word list, the two-word list, and the three-word list (see Appendix D).
In the task, the child was told to repeat some strange words. The task was administered in the form of a live presentation for an optimal condition for the perception of the stimuli (Chiat & Roy, 2007). The repetition of nonce words was administered first.
Concerning that the younger children’s shorter attention span and their need for more motivation to engage in this task, we not only downsized the number of nonword items, but also adopted and modified the testing procedure of Stokes and Klee (2009) for children at age 2 and age 2.5. In the testing, the child was asked to imitate each nonword said by the experimenter, and then rolled a ball down a chute as a reward, whether the word was correctly imitated or not. At age 2, the child was required to complete only half of the stimuli in the one-word list and the two-word list of each NWR task (i.e. 12 disyllabic nonce words and 12 disyllabic gap words). At age 2.5, the child had to complete half of the stimuli in the one-word list, the two-word list and the three-word list of each NWR task (18 disyllabic nonce words and 18 disyllabic gap words). For children at and above age 3, they were instructed to repeat after puppies for the all the nonwords.
The recall accuracy of the nonwords was scored at the syllable level. A syllable received one point if it was correctly repeated. Any omission, deletion, addition, or substitution would be considered errors because they signaled an inability to encode or maintain the original phonological representation. However, children’s systematic substitutions, as observed in their performance in the productive phonology task, were counted as correct to minimize the effect of articulatory constraint on their NWR performance. The maximum score was 72 for the nonce word repetition or the gap word repetition.
49
3.2.7 Leiter International Performance Scale-Revised (Leiter-R). The Leiter-R
(Roid & Miller, 1997)is an intelligence test designed to assess nonverbal cognitive abilities in children and adolescent aged 2 to 20. It includes two groupings of subtests:the Visualization and Reasoning Battery and the Attention and Memory Battery. For the purpose of this study, we used 4 subtests in the Visualization and Reasoning Battery as a rapid estimate of global intellectual ability, including Figure Ground (FG), Form Completion (FC), Repeated Pattern (RP) and Sequential Order (SO). The raw scores on each of the measure would be converted to normalized scale scores. The composite of the subtest scaled scores was regarded as a Brief IQ estimate.