Krott & Nicoladis (2005)

In this paper, Krott & Nicoladis (2005) examined whether family size would affect children’s segmentation of compounds. Twenty-five English-speaking children between 3;7 and 5;9 were asked to explain the meaning of existing compounds with constituents of varying family size to an alien puppet. As just mentioned, twenty-five English-speaking children took part in this study. Two children were excluded from the sample because they were bilingual and another one had to be left out because information about her knowledge of compounds was not available. Therefore, the analyses were based on 22 children aged 3;7 to 5;9.

In the experiment, the children were introduced to a puppet named Mork by a native speaker of English. The researcher told the children that Mork came from another planet, who did not speak English very well and was interested in why some words were used. She provided the example of ‘blueberries,’ saying‘We say blueberries because they are berries that are blue.’ Krott & Nicoladis (2005) decided not to use a noun–noun compound as an example so that it was not possible for the children to repeat the semantic relation of the example compound for all experimental compounds. The purpose of the example was rather to show that a compound could be split up into two parts that were related to each other. The experimenter asked the children to explain to Mork 25 noun–noun compound words, the components of which had either high family size or low family size (see Table 2-8).

25

Table 2-8 Stimuli by Family Size (High or Low) of Modifiers and Heads (Krott &

Nicoladis 2005, p. 144)

High-High High-Low Low-High Low-Low

Baby book tape-recorded and all speech relevant to the task was transcribed.

Because the number of compounds that the preschool children knew was rather small and most of the compounds they knew did not have a large family, the set of possible experimental items was restricted. Furthermore, Krott & Nicoladis (2005) wanted to restrict the compounds they selected to those in which each constituent played a transparent role in the meaning and each constituent was a concrete word the children might know.

Note that there were equal numbers for each category except for the low-family-size modifier category and the low-family-size head category. Krott &

Nicoladis (2005) included an extra item here because they were concerned that the children might not know the constituents of these compounds. Four items contained a

26

compound word as the modifier (e.g. breakfast cereal and grapefruit juice). All of these compounds, though, were semantically opaque. Krott & Nicoladis (2005) included them because even the adults did not always see the underlying structure of opaque compounds (Elliot 1997).

To determine the modifier families and head families of the target compounds, Krott & Nicoladis (2005) first gathered all noun–noun compounds available in the CELEX lexical database (Baayen, Piepenbrock & Gulikers, 1995). The database was constructed from adult language, so Krott & Nicoladis (2005) modified the compound list only to include compounds children were likely to hear. Two native English speakers, both trained in psycholinguistics and with extensive experience with children, went through the database, adding and deleting words according to what they thought children might know.

The analyses of Krott & Nicoladis’ (2005) experiment showed that their subjects were more likely to mention modifiers of compounds with large family sizes than small family sizes. This was true, independent of whether or not the children knew the compounds. There was no difference between the four-year-olds and the five-year-olds, implying that the modifier families were active as soon as they existed.

In contrast to modifiers, the scores for heads showed a less robust effect on the family size. That is, the family size effect was more evident for unknown compounds than for known compounds.

The results showed that the children were more likely to mention the modifier of compounds if they came from large constituent families than from small constituent families. Other variables were also shown to have some, but smaller effects on the children’s parsing, including the frequency of the constituent words and the compounds, whether the compounds were already known, and age. Taken together, these studies rather showed an effect of the modifier than of the head. This might be

27

due to the way compounds are processed, with the modifier playing a relatively more important role than the head.

2.2.4 Krott, Gagné & Nicoladis (2009)

Krott, Gagné & Nicoladis (2009) investigated various ways in which frequency might affect children’s selection of thematic relations for novel noun–noun compounds. Twenty-seven monolingual British-English-speaking children (age range 4;9 to 5;8, mean 5;4, SD 0;3) took part in the experiment. Krott, Gagné & Nicoladis (2009) constructed thirty novel noun–noun compounds using constituents that they expected children to know. Novel compounds can be interpreted in various ways and constituent families support these interpretations to different degrees. In order to investigate whether the family support has an effect, they made one likely interpretation for each compound against which they could compare the participants’

(both adults’ and children’s) responses. They determined likely interpretations for the compounds by presenting the adult participants with a randomized list of thirty compounds and asking them to write down their interpretations. The relations that dominated their responses were selected as the most likely interpretation.

Choosing the adults’ preferred relations as the comparison relations also allowed them to directly compare the children’s and adults’ interpretations. The next step was to determine the support that the dominant relations receive from the modifier and head families.

The experimental compounds were constructed with varying head family

and modifier family support in the child compound list because they were interested in the effect of modifier and head families on children’s interpretations. One third of the compounds had a dominant modifier–head relation that was supported by both a bias of the modifier family in the database (mean number of supporting family members 6.0, SD 1.6) and a bias of the head family in the database (mean number of

28

supporting family members 8.5, SD 5.4) (condition High–High). Another third (condition Low–High) had a dominant relation that was supported by only a bias of the head family (modifier family: 1.0, SD 1.7; head family: 10.0, SD 4.0) and the remaining third (condition High–Low) had a dominant relation that was supported only by a bias of the modifier family (modifier family: 6.6, SD 1.4; head family: 0.4, SD 0.7). Note that there was no condition like Low–Low because it was very difficult to construct easily interpretable compounds that did not have any support of the modifier or head families.

The subjects were tested individually and the experimenter asked them to explain the meaning of the experimental compound words. The children were asked what bear trousers were and corrected the child if he/she did not give a response that explained the relation between the head and the modifier (e.g. ‘trousers for bears’). The experimental compounds were presented to each child in random order. For each compound, the experimenter asked ‘What does X mean?’ If a child did not respond with an explanation for the meaning of the compound, the same question was repeated later. As mentioned, the adults were tested with a written version of the task, i.e, they were given instructions and the two examples above in writing and asked to write down their interpretations.

Both the children’s and adults’ interpretations were coded by two native English speakers and one fluent English speaker. The agreement among the three coders for the children’s responses was 71% (Fleiss’ K=0.68, p<0.001), for the adults’ responses it was 86% (Fleiss’ K=87.1, p<0.001). The codes for the remaining responses were decided by taking the majority code (15% of all the child responses and 13% of all the adult responses) or by discussion.

Of all the child responses, 39.7% were dominant interpretations, i.e., interpretations that were preferred by the adults in the adult pretest of the materials. In

29

18.5% of the responses the children reversed the roles of the modifier and head in their explanations (e.g. ‘an animal what has got a helmet on’ for an animal helmet or

‘it’s cheese and I think it is made out of fish’ for cheese fish), showing that even the five-year-olds still had some difficulties distinguishing between the head and the modifier when exposed to novel compounds.

Some of their findings are as follows: (1) when interpreting novel compounds, both the children and the adults responded with highly frequent relations within compound families; (2) the children and the adults relied on families related to different constituents; (3) in addition to using the most frequent thematic relation of a family, the children had a strong tendency for HAS or LOCATED.

2.2.5 Summary

The related issues in children’s comprehension and production of noun-noun compounds have been reviewed in this section. Table 2-9 is a brief summary of these previous studies.

30

Table 2-9 Major Findings and Limitations of the Previous Studies

Major Findings Limitations

Nicoladis (2003)

1. The younger children were less likely to understand that a compound referred to two objects.

2. There were age differences in comprehending and producing

The children had difficulty in selecting and integrating

properties of the modifier onto the head, a process required by

property interpretations.

The children were more likely to mention the modifier of a

noun-noun compound with large than with small family sizes.

1. Subjects: only two age groups with a group of native controls

2. Task: only a production task

31

Krott, Gagné &

Nicoladis (2009)

1. The children and the adults used similar but different strategies to interpret compounds.

2. Only the children’s interpretations revealed an overuse of visually perceivable relations.

1. Subjects: only two age groups with a group of native controls

2. Task: only a production task

In the above studies, we can see that there were age differences in the production of noun-noun compounds. On the other hand, we have also found the inadequacy of the subject numbers (Nicoladis 2003, Parault, Schwanenflugel & Haverback 2005, Krott

& Nicoladis 2005, Krott, Gagné & Nicoladis 2009). In addition, almost all the studies employed only a production task. To supplement this, we will include a comprehension task in our experiment.

2.3 Summary of Chapter Two

Several issues in Chinese noun-noun compound interpretation, L1 acquisition of noun-noun compounds in other languages, and types of noun-noun compound

relations in Mandarin have been reviewed in this chapter. The three studies on Chinese noun-noun compounds reviewed have demonstrated that adults’

interpretation of noun-noun compounds may depend on certain and existing relations.

Furthermore, studies on L1 acquisition of noun-noun compounds have indicated that children to some degree were more likely to mention the modifier of a noun-noun compound with large family sizes than with small family sizes. In addition, studies have shown that the three-year-olds and the four-year-olds were equally likely to produce compounds to name two interacting objects. However, the three-year-olds

32

were less likely than the four-year-olds to understand that a compound referred necessarily to two objects.

33

CHAPTER THREE RESEARCH DESIGN

This chapter presents the research design of the present study. Section 3.1 describes the participating subjects. Section 3.2 introduces the methods and materials adopted in the present study. The experimental procedures are exemplified in Section 3.3. In the end, this chapter is summarized in Section 3.4.

3.1 Subjects

One hundred and twenty subjects participated in the present study. They were further divided into 6 age groups. Each group consisted of twenty subjects (10 males, 10 females). We started from Age 4 since the previous L1 literature (Nicoladis 2003, Parault, Schwanenflugel & Haverback 2005, Krott & Nicoladis 2005, Krott, Gagné &

Nicoladis 2009) have reported Age 4 or older a good start to acquire noun-noun compounds.

The children of Groups 1-2 were preschoolers of one kindergarten in Da-an District of Taipei City, and those in Groups 3-5 were first, second, and third graders of an elementary school in Wenshan District of Taipei City. In addition, a group of native Chinese controls participated in the experiment. The control group consisted of twenty adults. A summary of the subjects is shown in Table 3-1:

34

Table 3-1 A Summary of the Subjects

Group Number Age range

The children were randomly selected by the teachers of the kindergarten and the elementary school to participate in the experiment. The control group were mostly undergraduate or graduate students of National Taiwan Normal University in Taipei.

The preschoolers attending the kindergartens spent five days a week in school.

They usually participated in indoor activities, such as reading, drawing pictures, and playing with building blocks in the morning and outdoor activities in the afternoon.

English and Taiwanese Southern Min classes were offered one hour per week. They use Mandarin to communicate with each other in the kindergarten.

The children in the elementary school also attended school from Monday to Friday. All of them arrived at school around 7:45 a.m. every weekday, and they usually left school at 12:00 p.m., except that some of them may attend extra classes and left in the afternoon on some of the weekdays. Although they speak Mandarin in school, they had two periods of English and one period of Taiwanese Southern Min per week. They also offered various courses such as mathematics, social science, and Mandarin Chinese.

35

3.2 Methods and Materials

Linguistic data can generally be collected from three sources (Bennet-Kastor 1988): indirect or anecdotal evidence; native speaker ‘intuitions,’ especially as judgments of the acceptability of utterances; and ‘raw’ data actually manifested in conversational and other naturally occurring forms. As Bennet-Kastor (1988) commented, the scientific value of the first source is limited because of its low validity. The linguistic data in the present study hence will mainly be obtained from the second and third sources.

The way of collecting linguistic data can also be distinguished by diachronic or synchronic collection. Diachronic data collection requires researchers to observe subjects over weeks, months, and usually several years. By contrast, synchronic data are collected not through long-term observation but through a number of subjects’

short-term participation in a study. Each method may be preferably applied in different research designs depending on the characteristics of a study. In the present study, synchronic data collection will be employed for the convenience to conduct a quantitative research.

Since this study aims to investigate children’s interpretation of novel noun-noun compounds, a production task and a comprehension task was designed. For the production task, it was an act-out task in which the subjects need to name the compound of two nouns in each question. In addition, a comprehension task was employed, in which all the subjects will be given two examples of novel noun-noun compounds to decide which one is the right interpretation. The design of the two tasks was elaborated in the following two subsections. We classified the test items in terms of animacy of the modifier(N1) and the head(N2).

The production task and the comprehension task were based on the similar structure, but using separate test items. Table 3-2 presents the test items for the two

36

tasks of the present study:

Table 3-2 Test Items Designed for the Two Tasks

Modifier Head Example Production Comprehension

Type 1:

In addition to the above 32 items, 4 fillers were included in each task. Thus, these two tasks consisted of 40 test items.

37

3.2.1 The Production Task

The production task was built on a story created by the researcher. The story depicted a central character, named Xiaomei, who subsequently saw different novel creatures or objects in a museum. There was a trial section in which the subjects were asked to name a familiar compound composed of two nouns. For each question, the subjects were asked to name the novel combinations of two nouns to examine their tendency of producing noun-noun compounds.

Table 3-3 An Example of the Test Scenario in the Production Task The subject would see:

Q2

The subject would hear:

Yuyitien zhaoshang, xiaomei gen mama qu canguan yijian xinqi buowuguan, zai guan nei, ta kanjienle zhege dongxi. Xiaopengyu, ni zhidao ta shi sheme ma?

‘One morning, Xiaomei and her mom went to visit “Novel Museum.” Inside the museum, she saw this thing. Do you know what it is?’

More examples of the production task can be found in Appendix A.

3.2.2 The Comprehension Task

The comprehension task was also built on a story created by the researcher, but this story would depict another central character, named Xiaoming, who saw different

38

creatures or objects in an art gallery. Two factors were taken into consideration when the items were designed: animacy of the modifier and of the head. Xiong (1998) and Huang (2008) both classify test items according to three variables: animal, plant, and artifact. To get a more general view, the three variables were reclassified into two main categories: animate and inanimate.

In the experiment, each subject saw animations on a computer screen and they would be told a recorded story comprised of the test items in different scenarios. After one scenario, the subjects will be asked the question “What do you think the mentioned item is?” Each subject was provided with two options and they choose one.

They were told in advance that there was no right or wrong answers. All of the subjects’ responses were audio-recorded. All utterances were coded based on the taxonomy discussed in Section 2.3. The coding procedure is reported in Section 3.3.3.

Table 3-4 An Example of the Test Scenario in the Comprehension Task The subject saw: tebie. Xiaopengyu, ni zhidao nayizhang hua shi Xiaoming kandao de ma?

‘One afternoon, Xiaoming and his dad went to an art gallery. He saw many novel pictures there. He first saw a “car elephant,” which looked special. Can you point to the picture that Xiaoming saw?’

39

More examples of the comprehension task can be found in Appendix B.

3.3 Procedures

Procedures employed in the present study were elucidated in this section. Section 3.3.1 summarizes the pilot study and discusses its inadequacies, and Section 3.3.2 will present the procedures of the formal study. The scoring and statistical analysis will be discussed in section 3.3.3.

3.3.1 Pilot Study

The pilot study aimed to probe into novel noun-noun interpretation strategies used by Mandarin-speaking children. The subjects of the study consisted of ten kindergarten children and ten adults. The children were divided into two age groups:

5-year-olds (Group 1), 6-year-olds (Group 2). The adult group as a control group (Group 3) was mainly comprised of university students whose mean age was 23 years old.

The subjects participated in a task in which sixteen novel noun-noun compounds needed to be explained. Of the sixteen compounds, in addition to four fillers, twelve compounds were further divided into three classes: High-Low, Low-High, and High-high. Modifiers of the High-Low compounds were found to have higher power than heads in affecting the children’s final interpretation. Analogically, heads of Low-High compounds had higher power in deciding the final compound meaning.

The High-High compounds were compounds in which the heads and modifiers had an identical relation bias, thus leading to the High-High condition.

After examining all the statistics, it was found that the children preferred

After examining all the statistics, it was found that the children preferred