Chapter 2 Literature Review
2.4 Mandarin-speaking children’s acquisition of narrative connectivity
2.4.2 Recognition and expression of event relations
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31
Similar to Chang’s (2004) method, To et al. (2010) assessed narratives for
Cantonese-speaking children. With respect to narrative connectivity, they examined connectives produced in the narratives. It is worth stressing that Cantonese grammar is similar to that of Mandarin Chinese. The findings showed an increasing tendency to use connectives among younger groups. However, the use of connectives did not provide further predictions of later academic achievement in school. To et al. explained that connectives are not obligatory in Cantonese and intersentential relations could still exist by clause embedding,
which is often seen in Cantonese (Matthews & Yip, 1994). Therefore, the coding of the explicit use of cohesive devices cannot completely represent a Cantonese speaker’s
competence of language due to the optionality of connectives.
2.4.2 Recognition and expression of event relations
The connectivity of discourse, namely event relations in discourse, has been investigated by many researchers based on English data (e.g., Trabasso & Sperry, 1985; Trabasso & van dan Broek, 1985; Diehl et al., 2006). With respect to Mandarin Chinese data, fewer studies
have been conducted. Among those studies, Wu and Tsai (2006) and Sah (2007, 2013) presented children’s development of event relation recognition and expression.
Wu and Tsai (2006) investigated Mandarin-speaking children’s causal concept and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
32
relating language development, and how the patterns accounted for the children’s theory of mind. In the study, they recruited children aged three and six months to five and five months, and told them a story about a protagonist and his/her missing candy. The candy was missing because after the protagonist went out, it was moved by his/her mother. Then, the subjects
were asked to answer questions. They found that hardly any children younger than five years old were able to figure out how the protagonist’s understanding/mind was different from the children’s understanding/mind. The false belief understanding was found to be better in
children older than five, who also showed a better ability of explanation. Wu and Tsai adopted Piaget’s (1928, 1930) cognitive developmental stages to explain three-year-old children’s
inability in reasoning and explanation: they were at the preoperational stage and thinking in the way of pre-causal reasoning. Wu and Tsai further longitudinally examined two children’s explanation types, which included psychological explanation, physical explanation, social-traditional explanation, biological explanation, magic-fantasy explanation, and behavioral explanation. It was found that social-traditional explanation and psychological explanation were the two dominant types. Wu and Tsai concluded that children around three to five years of age develop their theory of mind critically.
Sah (2007) also studied Mandarin-speaking children’s reasoning and explanation by investigated whether children were able to point out a misconception and were able to relate
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
33
the events in a cause-effect situation using particular pictures in terms of particular scenes in the storybook Frog, Where Are You? (Mayer, 1969). Developmental differences were found between narrators in the groups of children five years and five months, five years and eight months, and five years and eleven months. The development of cognitive ability and the gaining of linguistic capacity contributed to the theory of mind development for older children, which facilitated the explanation of misconception. The Three Phase Model (Karmiloff-Smith, 1984) of children’s problem-solving was also adopted in the discussion.
Sah (2013) conducted research on narratives produced by Mandarin-speaking children aged five and nine and adults. The narrative data was elicited using the storybook Frog,
Where Are You? (Mayer, 1969). Sah adopted methods proposed by earlier researchers
(Trabasso & Sperry, 1985; Diehl et al., 2006) in the assessment of narrative connectivity. The results showed that in the five-year-old children’s narratives, a large percentage of events were discrete events and events with only one connection from or to the other events. For the nine-year-old children, the most frequently found event belonged to the type of events with two connections to or from it. As for the adults, their pattern was similar to that of the nine-year-old children; however, adults produced the highest proportion of events with three or more connections. This phenomenon implies that children gain the ability of inferring and constructing causal relations as they age. The results also corresponded to previous research
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
34
on English-speaking children. The theory of mind and working memory capacity were adopted in explaining younger children’s insufficiency in establishing causal relations: limited
working memory and the immature development of the theory of mind constrained their ability to encode planning components. Sah’s study provided evidence for the developmental
differences in language; moreover, it demonstrated the method of narrative representation proposed by Trabasso and Sperry (1985), using an alternative approach to the qualitative analysis of narratives.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
35
Chapter 3 Method
This chapter will present the methods adopted in the present study. Section 3.1 details how data was collected and grouped, and Section 3.2 shows the analytical frameworks used in dealing with the data.
3.1 Data collection
The present study used data from the Child Language Data Exchange System (CHILDES) Beijing2 corpora (Tardif, 1993, 1996; MacWhinney, 2000), with Dr. Twila Tardif’s
permission. The present study randomly selected 30 pieces of four-year-old children’s story narratives for further analysis. The children’s mean age was four years and five months, and
their mean MLU (mean length of utterance, Brow, 1973) was 6.68.
The data contributed from Dr. Twila Tardif (2002, 2003) consisted of transcribed speech samples of children speaking Mandarin in their home environment. The children were first presented with the picture storybook Frog, Where Are You? (Mayer, 1969) and were asked to read the story (24 pictures) silently with an investigator. Then, the children were asked to recount the story as they browsed through the book page by page. The children’s verbal
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
36
narration of the story was transcribed into written transcriptions following the transcription format of the Codes for the Human Analysis of Transcripts (CHAT) used by the CHILDES (MacWhinney, 2000).
Picture stories provide a controlled content with protagonists, problems, a set of actions that follow from these problems, and an outcome. Thus, the series of events allow narrators to infer causal relations, and narrators with different levels of cognitive maturity may relay different interpretations of the events (Berman & Slobin, 1994). Given this benefit, many studies have collected data using the picture storybook Frog, Where Are You?, which has
enabled the present study to make further comparisons.
Examining Tardif’s data, it was found that the investigator’s verbal feedback was seldom provided. The investigator gave neutral comments such as “uh-huh,” “yes,” “anything else?”, and “and…?”. This minimized and neutral feedback prevented the narrators from being
influenced in choosing particular words and structures. Thus, a narrator’s output could be
considered an independent and spontaneous narrative. The present study analyzed the children’s narratives without taking the investigator’s comments into consideration.
3.2 Analytical frameworks
Narrative connectivity was analyzed using the following procedures: linguistic unit
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
37
segmentation, causal connection coding, relation type coding, and linguistic device identification. Coding schemes and analytical frameworks are introduced below.
3.2.1 Communication units
A story event was represented by a communication unit (c-unit), which is defined as “an independent clause with its modifiers” (Loban, 1976) and roughly consists of a verb and its
arguments (Strong, 1998). Accordingly, the transcripts were segmented following c-unit segmentation conventions (Loban, 1976; Strong, 1998).2 An example of an utterance segmented into c-units is shown in (3.1) below:
(3.1) Pre-c-unit segmented sample:
*CHI: 貓頭鷹 出 來 了 給 小 孩兒
maotouying chu lai le gei xiao haier owl go_out come exclamation give small child
從 樹 上 踢 了 下 去
cong shu shang ti le xia qu
from tree above kick completive go_down go
“The owl came out and kicked the child off the tree.”
C-unit segmented:
(a) 貓頭鷹 出 來 了
maotouying chu lai le
owl go_out come exclamation
2
Two coders achieved 90% reliability in determining c-units. Disagreements were resolved through consensus.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
38
“The owl came out.”
(b) 給 小 孩兒 從 樹 上 踢 了
gei xiao haier cong shu shang ti le
give small child from tree above kick completive
下 去
xia qu go_down go
“[…] kicked the child off the tree.”
As seen in the pre-c-unit segmentation sample, this utterance produced by a child consists of two actions: “go-out come” and “kick.” According to the above definition, a verb
and its arguments—“The owl came out” —is an independent c-unit (see (3.1a)). Similarly, a verb and its arguments—“[…] kicked the child off the tree” —is another independent c-unit (see (3.1b)). These two c-units represent two events.
3.2.2 Causal connections
Causal connections between two linguistic units (c-units) were identified using Trabasso and Sperry’s (1985) procedures. A causal connection refers to the direct operative causal
relation between a pair of linguistic units (Trabasso & Sperry, 1985). The identification procedure works with a counterfactual test based on the criterion of necessity (Mackie, 1980).
The test provides a reliable, logical, and analytic procedure in the inference of cause and consequence (Trabasso & Sperry, 1985).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
39
A counterfactual argument was used in the following format: “If not A then not B.” That is, event A is the necessary element for the occurrence of event B. In other words, if event A has not occurred, it is impossible for event B to occur. A pair of linguistic units that passes this test is causally related, whereas any pair of linguistic units that fails to pass the test is not
causally related. The following pair of units in (3.2) exemplifies the counterfactual test. If event A “the frog jumps out” is to be causally related with event B “the frog is not there,” then their negated forms should be examined: “The frog did not jump out” makes “the little frog was there” possible, and thus “the frog jumps out” is said to be the reason that “the frog is not there.” In other words, if the frog “did not” jump out, then the frog “would be” there; because
the frog jumped out, subsequently, the frog is not there, the frog has gone. Accordingly, event A and event B in (3.2) are a causally related pair:
(3.2) *CHI: 小 青蛙 跳 出來 了。
xiao qingwa tiao chulai le little frog jump out-come ASP
“The little frog jumped out.” Event A
*CHI: 小 青蛙 沒有 了。
xiao qingwa meiyou le little frog no-have ASP
“The little frog is not there (anymore).” Event B
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
40
After identifying the related unit pairs,3 four levels of causal connectedness in the narratives were coded using Diehl et al.’s (2006) method, namely, C0, C1, C2, and C3+. C0
indicates an isolated unit; that is, the unit has no causal connection with other units in a narrative, and the event represented by the unit is a discrete event. C1 refers to an event that has only one connection with one other event in a narrative. C2 indicates an event that has two connections with other events in a narrative. C3+ refers to an event that has three or more connections with other events in a narrative. The levels of connectedness for three connections and above were collapsed because they were seldom found to be produced by children (Diehl et al., 2006). A connectivity network4 was derived as a visualization of all c-units and connections in a narrative, following Trabasso and Sperry’s (1985) procedures.
3.2.3 Connection types
After related pairs were identified, they were further coded with four types of causal relations5: motivation, psychological causation, physical causation, and enablement. This classification method follows Trabasso and Sperry’s (1985) procedures based on the taxonomy of relations (Warren et al., 1979). Definitions and examples are presented below:
3
Two coders achieved a mean K of 0.82 in identifying causal relations and coding connection types.
Disagreements were resolved through consensus.
4
Trabasso and Sperry (1985) referred to this as a “causal network.” The present study uses the term
“connectivity network” for clarity since both causal and temporal connections are shown in the network.
5
Two coders achieved a mean K of 0.82 in identifying causal relations and coding connection types.
Disagreements were resolved through consensus.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
41
(1) Motivation: Goal-directed actions. In example (3.3), the event stated in the clause “the little frog has gone” motivates the action stated in the clause “search for the little frog”; because the frog has gone, someone is going to search for the frog.
Thus, this clause pair is causally related in terms of motivation.
(3.3) *CHI: 小青蛙 沒有 了。
xiaoqingwa meiyou le little frog no-have ASP
“The little frog has gone.” Cause […]
*CHI: 找 小青蛙。
zhao xiaoqingwa search little frog
“Search for the little frog.” Effect
(2) Psychological causation: A non-goal-directed action in a statement involuntarily
causes a psychological effect in another statement. In example (3.4), the event stated in the clause “the little frog has gone” causes an involuntary psychological state “the little dog worries, too,” as seen in the second clause. Thus, this clause pair
is related in terms of psychological causation.
(3.4) *CHI: 小青蛙 沒有 了。
xiaoqingwa meiyou le little frog no-have ASP
“The little frog has gone.” Cause
‧
(3) Physical causation: Naïve interpretations of the physical world or mechanical
causality between objects and/or people. In example (3.5), the physical action stated in the clause “the little frog has jumped out” causes the outcome as stated in “the
little frog was not there.” Thus, this clause pair is related in terms of physical
(4) Enablement: Actions, occurrences, or states that are necessary but not sufficient to cause other actions or states. In example (3.6), it is the situation as stated in “[the frog] is raised in a bottle” that makes the following event in the statement “the little
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
43
frog has jumped out” possible; if the frog was never in the bottle, then it is
impossible for it to jump out. Meanwhile, the situation does not motivate any action or cause any physical or psychological state. Thus, this clause pair is merely related
in terms of enablement.
(3.6) *CHI: 養 在 瓶子 裡。
yang zai pingzi li keep at bottle inside
“[The frog] is raised in a bottle.” Cause […]
*CHI: 小青蛙 蹦 出來 了。
xiaoqingwa beng chulai le
little frog jump out-come ASP
“The little frog has jumped out.” Effect
3.2.4 Linguistic devices
After related pairs were identified, they were further coded with different types of linguistic devices.6 The categories of linguistic devices were based on Chang’s (2004) temporal coding scheme, with slight adjustments based on studies of Mandarin grammar (Li
& Thompson, 1981; Tiee & Lance, 1986; Yip & Rimmington, 2004) and evidence found in related studies (Hsu, 1996; Wu & Tsai, 2006; Chang & Su, 2012).
The main categories in the coding scheme are presented below. Types (1) through (3) are
6
After identifying related events, two coders achieved 100% reliability in recognizing linguistic devices.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
44
explicit linguistic devices; that is, certain forms in Chinese characters serve as linguistic devices. Type (4) is an implicit linguistic device.
(1) Temporal connectives: temporal conjunctions or adverbs, including sequential connectives.
Examples: …shi, …de shihou, …hou/yihou/zhihou, …qian/yiqian/zhiqian,
yi…jiu…, yushi, ranhou, jiezhe, jiu
(2) Causal connectives: causal conjunctions or adverbs.
Examples: jieguo, yinwei…, suoyi…, yinwei, youyu, yiner, yinci
(3) Aspectual expressions: perfective, durative, and experiential expressions Examples: le, yijing, zai, zhe, guo, qilai, and kaishi
(4) Zero connectives (juxtapositions): Zero connectives exist where two clauses are
related in meaning (causal or temporal) but come together without any explicit conjunction marker. The two clauses’ meanings will not change if they are equipped
with a conjunction marker. For example, the clauses in (3.7) are causally related and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
45
the clauses in (3.8) are temporally related. The meanings remain the same with or without the markers in the brackets.
(3.7) 時間 不 早 了， 咱們 走 吧。
shijian bu zao le zanmen zou ba time not early ASP we go Pt
“It is late, (so) let’s go.”
(3.8) 他 剛 想 出去， 忽然 下 起
ta gang xiang chuqu huran xia qi
he just think out-go suddenly fall begin
大雪 來 了。
daxue lai le heavy snow come ASP
“(When) he was just thinking of going out, it suddenly began to snow heavily.”
(Yip & Rimmington, 2004)
It is worth noting that clauses with causal relation may be close but not necessarily adjacent to each other—the hearer may insert attention using a minimal reactive token, and the causal utterance itself may span across several intonation units—however, the base meaning of these causal-clauses remains unchanged (Chang & Su, 2012).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
46
Chapter 4 Results
This chapter will present the results of the present study in the following sequence.
Section 4.1 shows the measures of narrative length and event connectedness. Section 4.2 displays various types of relations constructed between connected events. Section 4.3 presents the linguistic devices applied in connected event pairs. Section 4.4 presents the combined results of the linguistic devices and different connection types.
4.1 Narrative length and event connectedness
The overall story length of the four-year-old children’s story narratives was first calculated. From the calculation of the number of c-units, it was found that two values were
far from the mean.7 Given this finding, the present study conducted the following data analysis without the two outliers. The four-year-old children’s narrative performance was inferred from the remaining 28 pieces of data.
Table 1 displays the measure of story length and the mean attributes of the c-units. The
children produced between 18 and 57 c-units per story. Within a c-unit, the mean number of
7
The box plot (see R: Box Plot Statistics) defines outliers as any data point more than 1.5 interquartile ranges
(IQRs) below the first quartile (Q1) or above the third quartile (Q3). In measuring the total c-units, the outliers
are Story 11 with a total of 76 c-units and Story 25 with a total of 103 c-units.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
47
words ranged from 1 to 7.43. On average, a four-year-old child produced 35.89 c-units in his/her story and each c-unit consisted of 5.8 words. The number of causal connections per c-unit in each story ranged from 0.84 to 2.23, with a mean of 1.45. The statistics demonstrate that the four-year-old children used about 1.45 c-units to describe one story picture (24 in total).
Table 1. Story length and attributes of c-units
Variable Number of C-units Words per C-unit Causal Connections per C-unit
Range 18 to 57 1.00 to 7.43 0.84 to 2.23
Mean 35.89 5.8 1.45
SD 9.97 1.23 0.33
Story 3 had the closest statistics to the mean scores for story length (37 c-units) and attributes of c-units (5.92 words per c-unit, 1.41 causal connections per c-unit). All the c-units in Story 3 are displayed in Table 2. The connections among c-units in Story 3 are visualized as a connectivity network, as shown in Figure 2. Each c-unit is represented by an Arabic number, and each pair of connected c-units is linked with an arrow. The direction of the arrow indicates a cause to an effect.
‧ Table 2. List of story events (c-units) of Story 3
C-unit Content C-unit Content
1 小 男 孩兒 在 看 這 一 個 小 青蛙 . xiao nan hai er zai kan zhe yi ge xiao qing wa .
A little boy is looking at a little frog.
20 小 孩兒 倒 在 地 下 .
xiao hai er dao zai di xia . The little boy falls down on the ground.
2 等 他們 睡覺 時候 the little frog climbs out from here.
22 那 個 蜜蜂 在 後邊 追 .
na ge mi feng zai hou bian zhui . The bees fly after the dog.
4 他們 醒 來 了.
ta men xing lai le . They wake up.
23 然後 嗯 貓頭鷹 沒 看 見 .
ran hou en mao tou ying mei kan jian . Then, well, the owl doesn’t show up.
5 看 見小 青蛙 沒 了 .
kan jian xiao qing wa mei le . (They) find that the little frog has disappeared.
24 他 就 又 飛 回 去 了 . ta jiu you fei hui qu le . He then flies back.
6 後來 到處 找 小 青蛙 .
hou lai dao chu zhao xiao qing wa . Later, (they) search for the little dog around.
25 那 個 貓頭鷹 不 見 了 .
na ge mao tou ying bu jian le . The owl disappears.
7 小 狗 陽台 上面 .
xiao gou yang tai shang mian . The little dog (is) at the balcony.
26 他 又 找 小 青蛙 .
ta you zhao xiao qing wa . He searches for the little frog again.
8 然後 小 狗 跳 下 去 .
ran hou xiao gou tiao xia qu . Then the little dog jumps down.
27 然後 有 個 鹿 出 來 了.
ran hou you ge lu chu lai le . Then a deer appears.
9 小 孩兒 也 跳 下 去 .
xiao hai er ye tiao xia qu . The little boy jumps down, too.
xiao hai er ye tiao xia qu . The little boy jumps down, too.