中文之容納語意認知: 以量化語料庫方法研究中文空間詞「裡」、「內」、「中」

全文

(1)i 碩士論文 Master’s Thesis Department of English National Taiwan Normal University. 中文之容納語意認知: 以量化語料庫方法研究中文空間詞「裡」、「內」、「中」 Conceptualization of “Containment” in Chinese: A Quantitative Corpus-based Study on Chinese Spatial Particles “Li” “Nei” “Zhong”. 指導教授: 陳正賢博士 Advisor: Dr. Alvin Cheng-Hsien Chen 研究生: 蘇洪寬 Student: Hung-Kuan Su 中華民國一百零七年 2018.

(2) ii Chinese Abstract 中文之容納語意認知: 量化語料庫方法研究中文空間詞「裡」、「內」、「中」研究生: 蘇洪寬. 指導教授: 陳正賢博士國立臺灣師範大學英語研究所摘. 要. 認知語義學領域最常討論的話題之一就是人類如何將空間經驗概念化。而在空間概念中，容納(CONTAINMENT)和支撐(SUPPORT)這兩個重要的空間概念，在過去十年中備受關注。雖然語言共享許多共同的空間概念，但不同語言的人們在如何編碼這些空間關係上卻是十分多樣的。本研究旨在探討中文母語使用者如何通過三個接近同義詞的空間詞構式意即[在…NP…裡/內/中]來概念化容納語意關係。過去大都採用質性分析來理解這些空間構式上的語義差異，而本研究意在透過以量化語料庫的角度，檢視容納空間構式與其地標詞(landmark)之間的共現模式。更具體地說，本研究藉由兩個統計分析，即多重鑑別詞彙連接分析(multiple distinctive collexeme analysis)和事後語義分析(post-hoc analysis)，從量化語料庫的角度來研究這些概念上類似的容納語意構式。目的在於探討每個空間構式，鑑別詞彙連接(distinctive collexeme)所形成之語義場(semantic field)，從而識別構式間之語義差異。研究結果發現，三種空間構式對其地標詞選擇具有不同的偏好。這些差異反映了中國人使用容納的意象基模(image schema)來概念化一系列抽象概念，如時間和事件。 [在…NP…裡]此空間構式被認為是容納關係的原型和無標記形式(unmarked)。因此，它具有更多樣化的模式，吸引了具體名詞，時間名詞，靜態抽象名詞。另一方面，當使用容納的意象基模來指定時間事件的閾值時，[在…NP…內] 此空間構式則被用於指定預先計劃.

(3) iii 的目標的截止日期。當容納的意象基模擴展到事件類的概念化時，[在…NP…內] 此空間構式可能在中文較為偏好，且通常會激發對地標的未完成(imperfective)解讀。. 關鍵詞: 構式搭配分析法, 語料庫分析法, 空間詞.

(4) iv English Abstract The Conceptualization of “Containment” in Chinese: A Quantitative Corpus-based Study on Chinese Spatial Particles “Li” “Nei” “Zhong” Student: Hung-Kuna Su. Advisor: Dr. Alvin Cheng-Hsien Chen Department of English. National Taiwan Normal University. Abstract How humans conceptualize space is one of the most commonly discussed topics in the field of cognitive semantics. Two important spatial concepts, namely CONTAINMENT and SUPPORT, have received much attention in the past decades. While languages share many common spatial concepts, studies have suggested cross-linguistic diversity in the categorizatgion of these spatial relations. In particular, this study focuses on how Chinese conceptualize the CONTAINMENT relation through three near-synonymous spatial particle constructions [zai…NP…li/nei/zhong]. While previous works mostly adopt a qualitative analysis on the semantic differences of these spatial particles, this study provides a quantitative corpus-based analysis of these CONTAINMENT constructions by examining the relationship between space particles and their co-occurring landmarks. More specifically, we investigate the semantic commonalities and differences of these conceptually similar particles of CONTAINMENT in Chinese through two quantitative analyses, i.e., multiple distinctive collexeme analysis and posthoc semantic analysis. Our results show that each construction has its own semantic preference for the landmarks. These differences reflect how Chinese speakers utilize the image schema of CONTAINMENT to conceptualize a range of abstract concepts. For li construction, it tends to.

(5) v attract more diverse types of landmarks, including concrete objects, temporal units, static abstract entities. It is the only particle that shows a connection to a prototypical CONTAINMENT landmark, i.e., a three-dimensional bounded entity. Our result suggests that li construction may be the prototypical and unmarked form to encode the CONTAINMENT. For nei construction, it shows a strong preference for landmarks denoting temporal concepts. This metaphorical use of CONTAINMENT with temporal landmarks often implies a preplanned objective with the landmark as an intended deadline for achieving the goal. Finally, for zhong construction, it shows a strong connection to landmarks denoting events of higher dynamicity. This metaphorical use of CONTAINMENT with dynamic abstract entities as landmarks often comes with a marked specificity, motivating an aspectual reading of the landmark.. Keywords: collostructional analysis, corpus analysis, spatial particles.

(6) vi Acknowledgment I would first like to thank my thesis advisor Dr. Cheng-Hsien Chen. The door to Prof. Chen’s office was always open for me whenever I needed any suggestion on my research, writing, or even life. He guided me to explore the world of corpus linguistics, and cognitive linguistics, which were totally new to me. Frustrated as I sometimes felt when encountering difficulty, most of time, the process of writing a thesis was fun and inspiring for me since I knew Alvin was always there and could give me much constructive advice whenever he though I needed. I would also like to acknowledge Dr. Miao-Hsia Chang, Dr. Fu-hui Hsieh as the second reader of this thesis. Lots of constructive comments, suggestion, and encouragement were received during proposal presentation, and thesis defense. It is really my greatest pleasure to have the two warm-hearted and supportive committee members. My thanks also go to all the teachers in the Department of English in NTNU. They offered a lot inspiration on my research. Finally, I must express my gratitude to my parents and my classmates for providing me Twith consistant support and encouragement. Throughout my years of study in NTNU, they were always beside me and were always willing to talk with me and cheer me up when I was stressed out. Everything I have done so far would not have been possible if it were not for them. Thank you..

(7) vii Table of Content Chinese Abstract ......................................................................................................................................... ii English Abstract ......................................................................................................................................... iv Acknowledgment ........................................................................................................................................ vi Table of Content........................................................................................................................................ vii 1. 2. Introduction ......................................................................................................................................... 1 1.1. Motivation ................................................................................................................................... 1. 1.2. Theoretical Framework................................................................................................................ 6. 1.3. Collostructional Analysis ............................................................................................................ 8. 1.4. Research Questions ................................................................................................................... 10. 1.5. Organization of the Proposal ..................................................................................................... 10. Literature Review ............................................................................................................................. 11 2.1. Conceptualization of “CONTAINMENT” across Languages ................................................... 11. 2.2. Conceptualization of “CONTAINMENT” within Languages................................................... 17 Conceptualization of “CONTAINMENT” in English .......................................................... 17 Conceptualization of “CONTAINMENT” in Chinese .......................................................... 20. 2.3 3. Studies on Collostructional Analysis ......................................................................................... 33. Methodology ...................................................................................................................................... 39 3.1. The Corpus in the Present Study ............................................................................................... 39. 3.2. Data Extraction .......................................................................................................................... 39 Extracting the Constructions ................................................................................................. 39 Identifying the landmark head nouns .................................................................................... 42.

(8) viii 3.3. Data Analysis............................................................................................................................. 42 Multiple Distinctive Collexeme Analysis ............................................................................. 42 Semantic Analysis of Distinctive Collexemes ...................................................................... 46. 4. Results ................................................................................................................................................ 53 4.1. The Distinctive Collexemes in Each Construction: Results from Multiple Distinctive. Collexeme Analysis ................................................................................................................................ 53 The Top 20 Distinctive Collexemes in Each Construction ................................................... 53 4.2. The Correlation Between Noun Types and Constructions: Results from Chi-square test ......... 64 The Correlation Between CONCRETENESS and Constructional Choice ................................ 64 The Correlation Between TIME-RELATEDNESS and Constructional Choice .......................... 66 The Correlation Between DYNAMICITY and Constructional Choice .................................... 67 Internal summary .................................................................................................................. 68. 4.3 5. 6. Summary.................................................................................................................................... 69. Discussion........................................................................................................................................... 71 5.1. CONTAINMENT with Concrete Entities ................................................................................. 72. 5.2. CONTAINMENT with Temporal nouns ................................................................................... 74. 5.3. CONTAINMENT with Events of different degree of dynamicity ............................................ 76. Conclusion ......................................................................................................................................... 83 6.1. Summary.................................................................................................................................... 83. 6.2. Limitations of the present study and suggestions for future study ............................................ 84. References .................................................................................................................................................. 86 Appendix .................................................................................................................................................... 92.

(9) 1 1 1.1. Introduction. Motivation What sets human apart from other species is our ability to use different kinds of linguistic. representation to categorize different spatial relations in our daily life. It reflects how we interact with this world. The issue of how humans use their cognitive ability to encode the spatial experience has been one of the major focuses in the field of cognitive semantics. Interestingly, there are a set of common patterns in how people with different language backgrounds conceptualize the world. When people describe a spatial scene, two entities will normally be identified, i.e., the trajector (TR) and the landmark (LM) . The characteristics of the two entities seem to be cross-linguistically similar: TR usually possesses characteristics of being smaller, moveable, and more dependent, while LM tends to be permanently located and more independent and tends to specify the geometric information (Talmy, 1983). For example, in the sentence the candy is in the jar, the candy is defined as the trajector, and the jar as the landmark, and the LM is preceded by a space-characterizing preposition in to form a prepositional phrase. In other languages like Atsugewl, languages may encode the LM and the spatial information on the verb (Talmy, 1983). For instance, the verb suffix -cis carries the meaning of ‘into the fire’. The instances all show a strong relation between LM and the geometric information. In spite of the universality, researchers have pointed out that there is diversity in how people from different languages encode the spatial relations (Evans & Chilton, 2010; Landau, Johannes, Skordos, & Papafragou, 2017; Levinson & Wilkins, 2006; Tyler & Evans, 2003). The most widely studied issue focuses on the difference between the two topological terms IN and ON in English and other languages (Choi, 2006; Coventry, Carmichael, & Garrod, 1994; Feist, 2000; Feist & Gentner, 2003; Landau, 2017; Levinson & Wilkins, 2006; Tyler & Evans, 2003; Vandeloise,.

(10) 2 2017; Zhang, Segalowitz, & Gatbonton, 2011). IN represents the spatial concept of CONTAINMENT and ON represents the spatial concept of SUPPORT. A typical example of CONTAINMENT is shown in Figure 1-1. In English, it is linguistically realized as apples in the bowl. The preposition in is used to encode this kind of CONTAINMENT relation. On the other hand, Figure 1-2 is an example of SUPPORT relation and is linguistically encoded by the preposition on in English as in apples on the plate. Furthermore, the same spatial scene can be conceptualized as either CONTAINMENT or SUPPORT in different languages. For example, as observed in Zhang et al. (2011), a spatial scene in which an English speaker describes as a bird in the tree turns out to be yizhi niao zai shu shang ‘there is a bird on the tree’ in Chinese. While English speakers perceive the spatial relation as CONTAINMENT, Chinese speakers conceptualize this spatial configuration as one of SUPPORT.. Figure 1-1. Figure 1-2.

(11) 3 Zhang et al. (2011) is one of the only few studies that address the two topological relations in Chinese. How Chinese conceptualize CONTAINMENT and SUPPORT has not been widely investigated in the line of research. In English, the preposition in is used to encode the CONTAINMENT relation, and on is for SUPPORT relation. However, in addition to in, there are also into, inside and within and all of them can also encode some CONTAINMENT scene as well. In Chinese, the concept of CONTAINMENT is also encoded by more than one spatial particle including ‘li’, ‘nei’, and ‘zhong’. While Zhang et al. (2011) suggest that these should be all treated as the particles that encode CONTAINMENT, it remains unclear how these spatial particles differ from one another in encoding the CONTAINMENT relation. Most of the available sources like the dictionary from the Ministry of Education treat the three particles interchangeably. The definitions of these particles in the dictionary of the Ministry of Education are provided in Table 1-1.. Table 1-1 Definitions for the three terms. word. definition. 裡. 1. 內部 2. 特定的時間空間. 內. 1.裡，相對於「外」. 中. 1.內、裡面。 2.泛指某一時期以內.

(12) 4 Moreover, in Sinica Corpus (Chen & Ma, 2010), we can also find the instances in which the three particles co-occur with the same head noun as the LM, as in (1).. (1) a.. 在. 未來. zài ZAI. 五. 年. 內. 增加. 到. 十億. 人口. wèilái wǔ. nián. nei. zēngjiā. dào. shíyì. rénkǒu. future five. year. in. increase. to. billion population. 的. 力量. ‘…increase to 10 billion people in future five years’ b.. 在. 過去. 的. 一. 年. 中. zài. guòqù de. yī. nián. zhōng. ZAI. past. one. year. in. DE. 希拉蕊. 發現. 了. 她. 內在. xīlāruǐ. fāxiàn. le. tā. nèizài de. lìliàng. Hillary. discover. LE. her. inner DE. power. ‘in the past one year, Hillary discovered her inner power’ c.. 在. 過去. 的. 六. 年. 裡. zài. guòqù de. liù. nián. lǐ. ZAI. past. DE. six. year. in. 伴. 著. 我們. 成長. 的. 母校. bàn. zhe. wǒmen. chéngzhǎng. de. mǔxiào. accompany. ZHE. we. grow. DE. alma.mater. ‘the alma mater which accompanied us in the past six years’.

(13) 5 In (1), the three spatial particles all modify the same head noun nian ‘year’. However, in some cases, the three spatial particles are not interchangeable. For example, from Sinica Corpus, we can find head noun rizi ‘day’ cooccurring with li and zhong but we cannot find the noun with nei as in (2). Several native speakers were consulted to see whether they accept the sentence in which the spatial particles in (2) were changed to nei, and all of them considered the sentences with nei cooccurring with rizi to be unacceptable.. (2) a.. 只. 能. 靠. 小. 朋友. 在. 平常. 日子. 裡. zhī. néng. kào. xiǎo. péngyǒu. zài. píngcháng. rìzǐ. lǐ. only. can. rely. small friend. ZAI. ordinary. days. in. 多. 讀. 多. 寫. 多. 聽. duō. dú. duō. xiě. duō. tīng. more read. more write more listen. ‘…can only rely on children to read, write and listen more in their daily life’ b.. 其實. 在. 她們. 每天. qíshí. zài. tāmen měitiān. 的. 日子. 中. 至少. 有. de. rìzǐ. zhōng zhìshǎo. yǒu have. actually ZAI they. everyday. DE. days. in. 三分之一. 以上. 的. 是. 和. 同學. 相處. sānfènzhīyī. yǐshàng de. shíjiān shì. hé. tóngxué. xiàng chù. one.third. up. time. with. classmate. get.along. DE. 時間. SHI. at.least. ‘Actually, in their everyday life, they have at least one third of the time getting along with their classmates’.

(14) 6. The low acceptance of those sentences with nei indicates the possible difference among the three spatial particles. The three spatial particles might be near-synonymous; however, they are not completely interchangeable as demonstrated in (2). In linguistic terms, they are “neither in free variation, nor in complementary distribution” (Divjak & Gries, 2006, p. 21). Some studies have tried to account for the semantic difference of the three spatial particles (Liu & Zhang, 2008; Xing, 1996; Yang, 2008; Zheng, 2005). Nevertheless, most of the previous analyses are qualitative and based on introspection. Therefore, this study aims to investigate these conceptually similar particles of CONTAINMENT in Chinese from a quantitative corpus-based perspective. More specifically, we hope to uncover the subtle semantic differences of these near synonymous space particles by examining their frequently co-occurring landmarks in these locative constructions. By pining down the conceptual difference among the three CONTAINMENT-related spatial particles with a more statistically sophisticated approach, we aim to uncover the differences in the ways of how Chinese conceptualize the CONTAINMENT relation.. 1.2. Theoretical Framework The whole study is conducted under the framework of Construction Grammar (CG). The. concept of construction, which is key to Construction Grammar, is defined as follows: “A construction is … a pairing of a form and meaning/use such that some aspect of the form or some aspect of the meaning is not strictly predictable from the component parts or from other constructions already established to exist in the language.” (Goldberg, 1996, p. 68).

(15) 7 In other words, a construction can be any linguistic expression, no matter how concrete or abstract, as long as it has a specific semantic or pragmatic function in its own right. Therefore, various levels of abstraction like words, semi-fixed phrases, argument structures, tense, aspect and modality can be a construction themselves. Moreover, Construction Grammar makes two important assumptions. First of all, if a word can occur in a construction, then the word must be semantically compatible with the meaning of the construction; second, if two constructions are syntactically distinct, then they must be semantically or pragmatically distinct (Goldberg, 1995). The two assumptions are very helpful in identifying the mechanisms behind alternation phenomenon for near-synonymous linguistic units. For example, in Gries and Stefanowitsch (2004b), they apply the CG framework to investigate the possible difference in the dative alternation, i.e. to-dative (i.e. John sent a book to me) and ditransitive (i.e. John sent me a book), which are often thought of as synonymous. According to Gries and Stefannowitsch (2004b), these two constructions differ considerably in the verbs attracted to each construction. The todative tends to attract verbs that denote a larger distance between the agent and the recipient like send, bring, etc. On the other hand, the ditransitive tends to attract verb that denotes a shorter distance, e.g., give and tell, which involve face to face interaction. Gilquin (2006) also adopts the CG framework and examines the differences among ten causative constructions like [X CAUSE Y Vto-inf], [X GET Y Vto-inf], etc, and proposes that these constructions are not totally synonymous and interchangeable since they show very strong preferences for particular verbs in the non-finite V slot. For example, [X CAUSE Y Vto-inf] construction attracts verbs that denote a transformation or specific movement like rise, become, while [X GET Y Vto-inf] attracts verbs that are used to elicit words or agreement like talk, say, and agree..

(16) 8 The present study also adopts the CG framework and explores the possible semantic difference among three spatial particles. The three locative constructions formed by the particles are defined as follows [zai NP li], [zai NP nei], and [zai NP zhong]. Based on the CG assumptions, these three constructions may be differentiated by looking at the nouns in the NP slot that are strongly associated with these constructions. Furthermore, by examining the semantic fields formed by these nouns, we are able to underpin the meaning/function of the three constructions. Therefore, we need a proper tool to measure the association strength between NP and the cooccurring construction in order to determine whether a noun is ‘attracted’ to a construction at a statistically significant level.. 1.3. Collostructional Analysis Construction Grammarians contends that if a word can occur in a construction, then the. word must be semantically compatible with the meaning of the construction. Collostructional analysis emerge as a practical methodological framework to examine the association between a word and a construction. in corpus-based research with sophisticated statistics (Stefanowitsch & Gries, 2003). The afore mentioned two studies (Gilquin, 2006; Gries & Stefanowitsch, 2004b) gain fruitful insights with the aid of the analysis. They all look into the slot variation among synonymous constructions. Gries and Stefanowitsch (2004b) looks into the VP slot in the todative construction and ditransitive construction, and Gilquin (2006) looks into the VP slot in the causative constructions. The analysis has the advantage of increasing “the adequacy of grammatical description by providing an objective way of identifying the meaning of a grammatical construction and determining the degree to which particular slots in it prefer or are restricted to a particular set of lexemes” (Stefanowitsch & Gries, 2003, p. 211)..

(17) 9 Therefore, the present study will adopt this approach to measure the association strength between NP and the spatial particle constructions [zai NP li], [zai NP nei], and [zai NP zhong]. Terminology in this approach needs explanation. In the collostruction analysis, a “collostruction” refers to the syntagmatic relationship between a collexeme and a construction, where a word (lexeme) that is attracted to a construction is termed as a collexeme. This study is going to look for the distinctive NP collexemes in the spatial particle constructions. There are three variant techniques of Collostructional analysis: the simple collexeme analysis, distinctive collexeme analysis and covarying collexeme analysis. Simple collexeme analysis studies co-occurrence relation between a lexical construction (i.e. a word) and a grammatical construction, e.g., the V slot in the [X think nothing of Vgerund] (Stefanowitsch & Gries, 2003). Distinctive collexeme analysis compares the collexemes in two semantically or functionally similar constructions, such as the verbs in the ditransitive and to-dative constructions (Gries & Stefanowitsch, 2004b). Covarying collexeme analysis studies the interaction between two slots within a construction, such V1 and V2 in the [X V1 into V2gerund] causative construction (Gries & Stefanowitsch, 2004a). The technique used in the study is multiple distinctive collexeme analysis, which is an extension of distinctive collexeme analysis. The distinctive collexeme analysis is used to compare the frequency distribution of words between two synonymous constructions. If there are more than two constructions under investigation, then the multiple distinctive collexeme construction is applied. The three spatial particles construction inquired in the study are semantically similar, all of which are related to the conceptualization of CONTAINMENT. Given this semantic similarity, multiple distinctive collexeme analysis can therefore be applied to the three constructions to analyze the variation in NP slot in the three constructions [zai NP li],.

(18) 10 [zai NP nei], and [zai NP zhong]. More detailed explanation of the application of techniques will be introduced in Chapter 2.. 1.4. Research Questions Based on the theoretical framework of Construction Grammar and with the aid of. multiple distinctive collexeme analysis, this study addresses three research questions: 1. For the three conceptually similar spatial particle constructions [zai NP li], [zai NP nei], [zai NP zhong], what are the distinctive collexemes of each type of space particle constructions? 2. What kind of semantic field is formed by the collexemes from each type of construction? 3. What are the semantic commonalities and differences that we can observe from the three semantic fields?. 1.5. Organization of the Proposal The present study is organized as follows. Chapter 2 gives a brief overview of the related. studies on spatial relations, and collostructional analysis. In Chapter 3, the processes of data collection and the data analysis are introduced. In Chapter 4, the results of the multiple distinctive collexeme analysis of the three spatial particles constructions are presented..

(19) 11 2. Literature Review. In this chapter, previous studies related to CONTAINMENT across languages will be presented in 2.1. Section 2.2 will introduce how CONTAINMENT is realized within a language. Section 2.3 will provide case studies applying the collostructional analysis.. 2.1. Conceptualization of “CONTAINMENT” across Languages There are infinite possible configurations between two objects in space; however,. languages only have limited set of spatial terms to describe those spatial relations, which forces languages to categorize the configurations into different sets in order to express them linguistically in spatial terms. Furthermore, different languages might consider different spatial configurations to be put into the same category. For example, Bowerman and Pederson (1992) investigated speakers of 38 languages and asked them to describe different spatial scenes including CONTAINMENT, SUPPORT, ATTACHMENT, and so on. Here are some of their findings. First of all, Spanish uses the spatial particle en to encode spatial scenes that are encoded by different spatial particles in English, as in apple on twig and apple in bowl. Moreover, in Dutch, CONTAINMENT (i.e. an apple in a bowl) is encoded by in, which is the same as that in English. However, there are three kinds of SUPPORT terms in Dutch. Solid support from below (i.e. an apple on a table) is encoded by op, and tenuous support (i.e. clothes on line) is encoded by aan, and om is for encirclement or surrounding (i.e. a necklace on a neck). In English, these spatial configurations all fall into the category of on. Even when two languages have particular spatial terms for the two previously mentioned spatial relations, their categorizations of the spatial configuration may not necessarily be the same. For example, Bowerman and Pederson (1992) investigated Berber, an Afroasiatic language spoken in north Africa. They found out that.

(20) 12 in Berber, x and di stand for the concept of SUPPORT and CONTAINMENT respectively. Berber and English do have some similarity in the two concepts. The typical SUPPORT scene for on in English is encoded by x in Berber as in a cup on the table and the typical CONTAINMENT scene for in in English is encoded by di in Berber. However, there are also some variations as well. For instance, the scene describing “an apple on a twig” is encoded by di in Berber, which is a SUPPORT particle. All in all, Bowerman and Pederson’s (1992) findings point out that there are some overlapping and variation in how different languages categorize the configuration in space, especially for the concept of CONTAINMENT and SUPPORT. In fact, the conceptual connection between CONTAINMENT and SUPPORT has received a lot of attention. Previous studies have analyzed this link from a cross-linguistic perspective (Bowerman, 1996; Bowerman & Choi, 2001; Feist, 2000; Feist & Gentner, 2003; Landau et al., 2017; Levinson & Wilkins, 2006). Researchers have shown that typological concept of CONTAINMENT is not a universal holistic concept, but one composed of a set of semantic properties (Feist, 2000; Feist & Gentner, 2003; Levinson & Wilkins, 2006). These studies try to describe the semantics of different spatial terms in languages and investigate the ranges of configurations that are categorized as CONTAINMENT and SUPPORT in a language and further identify the important semantic properties of those configurations categorized under the SUPPORT and CONTAINMENT terms. The identified attributes may then shape the basis for the meaning of the spatial terms. Since these spatial terms encode different kinds of configurations, studies on these terms mostly start with a geometry-based approach, which focuses on the geometry of the spatial scene described. In addition to the geometry-based approach, there are also studies taking on a.

(21) 13 function-based approach, which claims that the function of the landmark or the functional relation between the trajector and landmark determines the proper spatial terms for a scene. First of all, the geometry-based approach looks into the geometrical information encoded by different spatial terms (Bowerman, 1996; Bowerman & Choi, 2001; Bowerman & Pederson, 1992; Choi, 2006; Feist, 2000; Feist & Gentner, 2003; Landau et al., 2017; Talmy, 1983; Tyler & Evans, 2003; Zhang et al., 2011). The geometry description of the landmark and trajector is the most straightforward way to identify the semantics of a spatial term and is useful for us the identify the proto-scene in a language. Tyler and Evans (2003) provide the proto scene of in to explain the primary semantics of the spatial term as in Figure 2-1. We can see that the trajector is located within a landmark. Three elements are identified in the scene: an interior, an exterior, and a boundary.. Figure 2-1 Proto-scene for in. The proto-scene for in corresponds to the definition of the non-linguistic spatial concept of CONTAINMENT in psychology, which is defined as follows: “a perceived container has a particular size, shape, and so forth, but the spatial primitive container used to interpret it does not: It consists only of a thing with an inside and an outside and a boundary between” (Mandler, 2007, p. 747). The boundary which establishes the inside and the outside is an important feature to this type of relation. These geometric descriptions help explain why the CONTAINMENT.

(22) 14 particle in is used in Figure 1-1. The bowl forms a boundary and separates the interior as well as exterior of the space. The apples are located in the interior of the bowl; therefore, in is the proper candidate for the scene. In addition, other geometric factors also play a role in determining the spatial terms. For example, the contact between the landmark and trajector is regarded as an important feature for the use of English on and the inclusion of the trajector in the ground is an important feature for the use of English in (Feist, 2000; Feist & Gentner, 2003). Note the two characteristics for in and on in English have some overlaps. In the spatial scenes like Figure 1-1 and Figure 1-2, both trajectors have contact with the landmark and could be a possible candidate for English on. However, what sets them different is that in Figure 1-1, the landmark is concave enough to form the interior and the English in can be applied. Concavity, which is closely related to the geometry of the LM, is an important factor to distinguish the CONTAINMENT and SUPPORT terms in English (Feist & Gentner, 1997). The geometrical factor is also influential for the use of Chinese CONTAINMENT and SUPPORT terms. Zhang et al. (2011) identified the systematic similarities and differences between the ways that English and Chinese speakers encode the spatial scenes of CONTAINMENT and SUPPORT. The two concepts, as shown above, do not have a clear-cut boundary among languages; instead, the connection of the two concepts may be argued to from a conceptual continuum. They found that geometrical and functional factors can account for the selection of the spatial particles li IN and sheng ON in Chinese. However, the geometrical factor is more influential than the functional factor. For example, when describing the scene where a flower is in the hair, Chinese speakers tend to use shang ON to describe the spatial scene while the English speakers are likely to use the spatial particle in. The reason for English speakers.

(23) 15 using in for the spatial scene is based on a geometry ground. They see the hair as a container, which allows the penetration of the flower’s stem, while Chinese speakers focus more on the flower, which has contact with the surface of the hair. To sum up, the geometric factor to configuration is closely related to the schematization of the landmark and trajector. This geometric-based approach provides many important insights into the similarity and variation among spatial terms in languages. However, there are some limitations with the geometry approach. For example, in according to geometry approach, must involve inclusion. However, in Figure 2-2, not all apples are located at the interior of the barrel. In this way, the geometry approach may need to argue that the apples form an aggregate, part of which is in the interior part of the barrel.. Figure 2-2. Another possible limitation of the approach is that the same spatial scene can be encoded by different spatial terms within a language with the geometry property of the scene remaining the same. For example, when describing a scene like Figure 2-3, a Chinese speaker can have two alternatives, as in (3). A geometry approach cannot explain the applicability of both the CONTAINMENT and SUPPORT particle to the same spatial configuration and how do people make choices between the two possible spatial particles..

(24) 16. Figure 2-3. (3) a.. 麵包. 在. 盤子. 上. miànbāo. zài. pánzi shàng. bread. ZAI. plate. on. ‘The bread is on the plate’ b.. 麵包. 在. 盤子. 裡. miànbāo. zài. pánzi lǐ. bread. ZAI. plate. in. ‘The bread is in the plate’. In response to the limitations in geometry approach, some scholars propose the function approach as a possible solution (Coventry et al., 1994; Feist, 2000; Feist & Gentner, 2003). The typical function of the landmark or the functional relation between the trajector and the landmark may play a role in the selection of the spatial terms. The function approach does not reject the geometry approach but considers it to be an important factor since the function of the landmark is closely related to its geometric properties. For example, if we want to say an apple is in a bow, the bowl must be concave enough to function as a container. Coventry et al. (1994) contended.

(25) 17 that English preposition in is appropriate if the LM is conceived of fulfilling its CONTAINMENT function. The CONTAINMENT function is closely related to the locational control because as the container moves, the containees should move with it. In this way, we can justify why in Figure 2-2, all of the apples are considered to be in the barrel. The barrel is fulfilling its function as a container even though not all the apples are in the interior of the barrel. This explanation is more straightforward than arguing that the apples form an aggregate (Feist, 2000). Feist and Gentner (2003) also stress the functional aspect of the CONTAINMENT relation. They point out that different landmarks may prefer different prepositions even though the geometry information remains the same. The choice of the English preposition might be related to our world knowledge about the landmark.. 2.2. Conceptualization of “CONTAINMENT” within Languages Conceptualization of “CONTAINMENT” in English This section will explore how the spatial concept of CONTAINMENT is linguistically. realized within languages. In English, CONTAINMENT is linguistically encoded by the preposition in. Tyler and Evans (2003) adopt a principled polysemy approach to analyze the multiple senses related to the preposition. In this approach, the relationship between these senses is argued to be strongly connected with each other and are systematically motivated. It is suggested that the senses are grounded in spatial-physical experience, which will give rise to situated inference. For example, the state sense for in like he is in love is based on our daily experience. When we, as trajectors (TR), are in a particular landmark (LM), we often experience certain kinds of emotions. For example, when we are in our lover’s arms, we will feel secure and love. By repeating this coupling experience of a particular emotional state in a specific location,.

(26) 18 a strong correlation between the location and the mental state may be established, thus deriving the state sense of the preposition in. In the framework of principled polysemy, it is important to decide the primary meaning of in, to which other related senses are conceptually linked. Evans and Tyler (2004) defined the primary sense of these spatial particles, i.e., proto-scene, as “a highly abstract representation of a recurring spatial configuration between two objects”(Evans & Tyler, 2004, p. 166). The proto scene of in provided in Evans and Tyler (2004) is reproduced in Figure 2-4. From the figure, we can see that the dark circle (TR) located within a landmark formed with bold lines with three salient structural elements: an interior, an exterior, and a boundary. In addition, the proto-scene also designates the functional elements. For example, the bounded LM can constrain the movement of TR. In our daily experience, when we move the glass, the water in it will also be moved. Another functional element related to preposition in is the function of support (i.e. in the sentence the flower in the vase, the flower is supported by the vase in order to stand upright).. Figure 2-4 Proto-scene for in(Evans & Tyler, 2004, p. 168). In in English carries different senses. Previous studies have tried to account for the multiple senses of in. There are mainly two approaches to this issue — the homonymy approach and the polysemy approach. The homonymy approach assumes that the senses are arbitrary. There are no connections between those senses. However, this perspective is counter-intuitive..

(27) 19 The same word can derive different senses depending on the context, which indicates that these senses have some connections to each other in our conceptual structure (Evans & Tyler, 2004; Langacker, 1987). For example, in (4), in carries different senses. In (4a), Amy is the landmark and house is the trajector, and in carries the meaning of being in a location. In (4b), Amy is still the landmark but despair is not a concrete place for TR. The world knowledge of depression, i.e. a kind of emotions, may prompt the reading that the undergoer (TR) of the state has difficulty leaving the emotional state (LM). Therefore, in in (4b), through the interaction with the context and the world knowledge, derives the sense of being in a state. That shows that the senses are not arbitrary as the homonymy approach claims but organized in a systematic way and grounded in our spatio-physical experience as the polysemy approach argues (Evans & Tyler, 2004; Tyler & Evans, 2003).. (4) a. Amy is in the house b. Amy is in depression. The polysemy approach states that meaning extension is a highly experiential-motivated process (Evans & Tyler, 2004; Tyler & Evans, 2003). Through our interaction with all kinds of entities in our daily life, a word can derive distinct senses. A new meaning might emerge in a particularized context and then conventionalized as a distinct sense through recurring configuration. Following this principle, Tyler and Evans (2003) proposed 27 senses for in in English. Lu (2017) provides a parsimonious account for the polysemy of in, and he narrows down the 27 senses in into only two senses. The primary meaning is the physical.

(28) 20 CONTAINMENT and the other sense is way/means sense. In his study, the contextualization is considered to be an important device to decide the core meaning of in. If a sense can be inferred from the context, then it will not be counted as a distinct sense. The method follows the principle of parsimony, which tries to be as economical as possible. So far, we can see that most of the CONTAINMENT scene can be linguistically encoded by in in English. Nevertheless, there are also other terms for CONTAINMENT, including inside and within. Lindstromberg (2010) claims that inside is an emphatic alternative to in. The emphatic tone with inside suggests a complete inclusion. Therefore, in the case of partial CONTAINMENT, like the flower is in the vase, it is hard for us to say that the flower is inside the vase. Within is softer than inside in tone and is often used with abstracted landmarks, such as range denoting nouns or time-related nouns like within three months. As for in, Lindstromberg (2010) argues that it has more senses than the other two alternatives. It seems that in can cover all the cases in which within and inside can appear since they only differ in tone.. Conceptualization of “CONTAINMENT” in Chinese In Chinese, the concept of CONTAINMENT is linguistically encoded by three spatial particles, i.e., ‘li’, ‘nei’, and ‘zhong’. In terms of the semantic differences of the three spatial particles, Xing’s (1996) analysis is found to be the most fundamental and influential. Analyzing two of the three near-synonymous particles, li, and zhong, he proposed that both can be used to express something or someone being in a specific place, time, or event, such as xinli and xinzhong ‘in one’s heart’, cunzili and cunzizhong ‘in the villiage’, wunianli and wunianzhong ‘in five years’. However, what makes li different from zhong is that li is more grammaticalized and can thus sometimes be omitted. For instance, neixinli ‘in one’s heart’ can be understood as.

(29) 21 neixin. Furthermore, li can set a boundary which is opposite to wai ‘out’, such as wuli wuwai dou shi ren ‘in the house and out the house are all people’. As for zhong, zhong can be used to express something or someone being in a process (i.e. tanpanzhong ‘in the process of negotiation’), a certain state (i.e. fennuzhong ‘in anger’), or a place without boundary (i.e. kongzhong ‘in the air). Furthermore, he added that li tends to co-occur with a concrete place while zhong with collective and abstract nouns. Deng (2006) extended Xing’s (1996) analysis by including another spatial particle ‘nei’, which shares similar meanings with ‘li’ and ‘zhong’. She starts the discussion from a diachronic perspective by referring to the definitions of these spatial terms in the classic etymology dictionary of Chinese, Shuowen Jiezi. The etymological origins of the three spatial particles are summarized in Table 2-1. According to Deng (2006), zhong and nei appeared earlier than li. Zhong and nei were more frequently used before Chin dynasty while li first appeared in West Han dynasty. Zhong, at that time, tended to co-occur with a larger landmark like di ‘ground’ and landmarks referring to a particular state like xian ‘danger’. Early in Chin dynasty, Zhong cooccurred with both concrete and abstract nouns, often without specifying a clear boundary. The original meaning of nei focuses on the transition from the outside to the inside, and thus the landmarks co-occurring with nei must have a clear boundary like men ‘door’, ting ‘yard’ or some temporal nouns with a clear range like qinian ‘seven years’. It not only emphasizes the inside of the landmark but also tries to show a contrast with the outside. Li, since it appeared later than the two particles, takes some senses from the previous two spatial particles. It can co-occur with landmarks with or without a clear boundary, and it can occur with both concrete nouns and abstract nouns..

(30) 22 Table 2-1 The Etymological Origins of the Three Spatial Particles (Deng, 2006, pp. 10-15). particle. original meaning. li. Referring to the inner side of cloth. nei. Emphasizing on the process of entering from the outside to inside. zhong. Referring to process of setting a bar through the drum and carrying the meaning of “in the middle”. After an etymological account of these three spatial particles, Deng (2006) conducted a comparative analysis from a synchronic perspective. Analyzing examples from many different sources, including 15 modern Chinese literature, online resources like Baidu, Xinlang News, and corpora from Beijing University, and some self-created example, she examined the semantic coherence of the cooccurring nouns (i.e. the landmarks) and divided them into five main categories..

(31) 23 Table 2-2 summarizes the association of the three space particles in relation to her five semantic categories. The first column lists five main senses encoded by the spatial particles. Note that the range sense overlaps with the location sense. Deng (2006) argued that when the cooccurring noun is followed by a determiner, the configuration will carry the sense of a specific range. For example, fanzi ‘house’ is a common place noun, so fanzili ‘in the house’ is referring to a place while zhegefanzili ‘in this house’ refers to a range set by the house. As for the sense ‘referring to the people in the N’, it is demonstrated in (5).. (5) Deng (2006, p. 44) 門. 裡. 遲疑. 了. 一下. 便. 答…. mén. lǐ. chíyí. le. yīxià. biàn. dá…. door. in. hesitate LE. one-CLF. so. answer. ‘the person behind the door hesitated for a while and answered…’. As for the result, Deng (2006) didn’t specify the token frequency of each configuration; instead, she listed all the combinations observed in her corpus data. As for the features of the cooccurring nouns, she investigated at least four factors of the cooccurring nouns, including dimensionality, boundedness, dynamicity, and concreteness. According to Deng’s observations, li can be used with all types of nouns. Zhong cannot appear with the nouns that are used to set a boundary and cannot refer to the people in the place. Nei cannot take most of abstract nouns such.

(32) 24 as nouns denoting abstract ideas, dynamic nouns, event nouns, and cannot refer to people in the place as well..

(33) 25 Table 2-2 A Summary Table of Deng’s (2006) Findings. “X zhong/nei/li”. Information about “X”. Concrete. Concrete, bounded, two/three dimensional nouns (i.e. V. place. wu ‘house’). V. V. V. X. Be able to set the boundary (i.e. men ‘door’). X. V. V. Abstract. Concrete but not a place (i.e. jingzi ‘mirror’). V. V. V. place. Abstract idea (i.e. jingshen ‘spirit’). V. V. X. Dynamic nouns (i.e. lunshu ‘statement’). V. V. X. Collective nouns (i.e. renqun “crowd”). V. V. V. Place nouns (i.e. zhegewuzhi ‘this house’). V. V. V. A point of time (i.e. yuji ‘rain season’). V. V. V. A range of time (i.e. jigeyueli ‘several months’). V. V. V. With an initial and end point, and set a specific. X. V. V. Event (i.e. wenhua geming ‘Cultural Revolution’). V. V. X. Adjectives ( i.e. beishang ‘sadness’). V. V. X. Place noun ( i.e. men ‘door’). X. V. X. to a range. Referring. Referring to a place. tiankong ‘sky’). Referring to time. temporal boundary (i.e. anfa qian jitain li ‘in the few. the N. people in. to states. days before the incident’). to the. Referring. nei. V. Unbounded, either two or three dimensional (i.e.. Referring. zhong li.

(34) 26 The previous two studies are found to be quite inspirational to the present study. Xing (1996) provides a general picture of how these similar spatial particles could be semantically differentiated, and Deng (2006) analyzes the difference and similarity in a more systematic way. However, both studies have limitations in several aspects. First of all, in terms of the data resource for analysis, the representativeness of the examples may be a concern. Both Deng (2006) and Xing (1996) incorporate intuition data for analysis. The intuition data may be influenced by one’s dialect or style, and sometimes are difficult to verify as introspection is not observable, and may not necessarily capture the general tendency of language use (McEnery & Wilson, 2001). Second, although Deng (2006) conducted a corpus-based study, she didn’t use a more rigorous quantitative approach to identify the distribution of the nouns in different spatial particle construction. Third, the semantic categories of the nouns are ambiguous, especially for the type of abstract nouns in Deng (2006). For example, for the nouns that denote dynamicity like taolun ‘discussion’, Deng (2006) claims that when we say taolunzhong, it has the meaning of being in an abstract location. She didn’t explain why an activity like noun is regarded as a place instead of a process noun as Xing (1996) proposed. Following Deng (2006), Yang (2008) also adopts a diachronic perspective to analyze the contemporary uses of the three near-synonymous spatial particles in question. What makes Yang (2008) different from the previous two studies is that he provides a mental representation of the particles and tries to explain how the mental representation influences their daily uses. While Yang has based his claims on native intuition and self-created data, Yang’s (2008) observations are fruitful and inspiring. There are four major findings. First of all, in terms of the cooccurring concrete nouns, Yang (2008) suggests that li tends to modify nouns that are larger than those that nei can modify and smaller than those that zhong can modify. For example, he argues that native.

(35) 27 speakers are more likely to use zhong to modify hai ‘sea’ instead of li. In addition, li also carries the meaning of being full of something. The reason for the additional meaning might result from its original meaning since the inner side of the clothing is always full of other materials like cotton. As a result, we are more likely to say xiangzili dou shi shu ‘in the box, there are all books’ instead of xiangzizhong dou shi shu because we want to emphasize that the box is full of books. Second, as for zhong, since its original meaning is about setting a flag in the middle of the ground, it tends to emphasize the centrality of the landmark. Furthermore, it does not carry the meaning of being full of something; instead, it allows the entity to move around, thus often cooccuring with a larger space. Yang (2008) also points out that there are some cases in which both li and zhong are somewhat interchangeable but native speakers are more likely to use zhong like huiyizhong ‘in the meeting’, jinglizhong ‘in the experience’ because zhong does not carry the meaning of being full of something. Third, as for nei, since its original meaning is about going from the outside into the inside by crossing the boundary, the boundary needs to be specified and it is always used to show contrast with wai ‘out’ like xiaonei ‘in the school’ and xiaowai ‘out the school’. In the previous example, Yang (2008) claims that a native speaker will not use zhong and li if they want to show such a contrast. Finally, he provides mental representations of the three spatial particles as in Figure 2-5. The mental representations help us to capture the prototypical meaning of the three spatial particles, and to understand how the meaning extends through the interaction between the mental proto-scene and our daily experience..

(36) 28. Figure 2-5 The mental representation for the three spatial particles. There are several methodological concerns in Yang’s analysis that may need further discussion. First, some of the semantic criteria for the analysis of landmarks may be more difficult to operationalize. For example, Yang (2008) claims that zhong is more likely to cooccur with larger place while li is more likely to cooccur with a smaller place. Yet it remains unclear how large a place could be to be counted as a larger place. Second, Yang (2008) argues that native speakers are more likely to use zhong like huiyizhong ‘in the meeting’, jinglizhong ‘in the experience’ because zhong does not carry the meaning of being full of something. The author doesn’t explain why words such as huiyi ‘meeting’, jingli ‘experience’ does not carry the meaning of full of something. Different from previous studies, Liu and Zhang (2008) examined the geometrical information carried by the three spatial particles from a quantitative perspective. They carried out a corpus-based analysis on the three spatial particles li, nei and zhong by analyzing the cooccurrences of nouns with these three particles in a 1,000,000-character corpus. It should be noted that they only focus on physical nouns (i.e. physical entities serving as the landmarks) and investigate geometrical features of the cooccurring nouns, including dimensionality, boundedness and porousness. They analyze the frequency distribution of different types of nouns in relation to the three particles. Their observations of the preferences for cooccurring noun in the three spatial particles are summarized in Table 2-3. From the table, we can know that.

(37) 29 porousness may not be crucial in differentiating the three spatial particles. Nei tends to cooccur with nouns with clear boundary such as xuxian nei ‘in the dotted line’. As for the dimensionality, three spatial particles show different preferences.. Table 2-3 Preferences for Cooccurring Nouns in The Three Spatial Particles. li. nei. zhong. dimensionality. three-dimensional. two dimensional. two/three-dimensional. boundedness. +/-. +. +/-. porousness. +/-. +/-. +/-. Liu and Zhang’s (2008) analysis provides a more quantitative evaluation of the tendency of which types of nouns is more likely to cooccur with one of the three spatial particles. Their findings in general support claims made in the previous studies (Deng, 2006; Xing, 1996; Yang, 2008). Nevertheless, there are some limitations in Liu and Zhang’s (2008) study. First of all, their analysis only included physical landmarks. The study would be more complete if the abstract landmarks can also be taken into consideration. However, if abstract nouns had been included in the analysis, the dimensionality would not work for the abstract landmarks since dimensionality is based on the geometric information of a physical object (Talmy, 1983). New criteria for data coding may be needed if we want to incorporate all the concrete nouns and abstract nouns in present study. Second, as Liu and Zheng have based their generalization on the frequency distribution of the co-occurrences between nouns and the particles, the method could be more rigorous by incorporating more statistical techniques in order to determine the important relationship between the space particles and a particular semantic field..

(38) 30 So far, we have seen studies focusing on different aspects of the three spatial particles. A summary table of the major findings and the limitation of the previous researches is in Table 2-4. Table 2-4 A Summary of The Previous Studies on Chinese CONTAINMENT Spatial Particles. Xing. Strength & major findings. Limitations. 1. Zhong and li can be used to express that something or. 1. Personal observation might be. (1996) someone is in a specific place, time, or event. biased and some observations can. 2. Li is more grammaticalized than zhong. be supported from a quantitative. 3. Li can set a boundary which is opposite to wai. perspective.. 4. Zhong can be used to express that something or someone is in a process, a certain state, or a place without boundary 5. Li tends to appear with a concrete place while zhong tends to cooccur with collective and abstract nouns. 6. Factors investigated: Boundedness, Dynamicity, Concreteness. Deng. 1. The diachronic perspective of the three spatial. (2006) particles is investigated.. 1. Some observations can be supported from a quantitative. 2. A more systematic analysis of the three spatial. perspective.. particles is presented in five aspects.. 2. The study describes all the. 3. Different degree of abstraction of noun can play a role. possible combinations but does. in differentiating the three particles.. not specify the possible. 4. Factors investigated: Dimensionality, Boundedness,. preference for each particle.. Dynamicity, Concreteness.

(39) 31 3. Some categorizations of the nouns are hard to justify.. Yang. 1. The diachronic perspective of the three spatial. (2008) particles is investigated and their meaning extension.. 1. Personal observation might be biased. Some observations can be. 2. Li tends to modify things larger than those can be. supported from a quantitative. modified by nei, but things smaller than zhong and also. perspective.. carries the meaning of being full of something.. 2. Some phenomenon need further. 3. Zhong allows trajector to move around in the. clarification especially about the. landmark and thus tends to cooccur with landmarks. compatibility between the abstract. denoting larger space.. landmark and the spatial particles.. 4. Nei specifies the boundary and is always used to show contrast with wai ‘out’. 5. The mental representations of the three spatial particles are provided. 6. Factors investigated: Boundedness, Dynamicity, Concreteness. Liu. 1. It is more quantitative since the authors count on the. 1. The study only focuses on. and. cooccurrence between different types of nouns and the. physical landmarks.. Zhang. cooccurring spatial particles.. 2. Relying on the observed. (2008). 2. Li tend cooccurs with three-dimensional landmarks.. frequency is not enough to decide. 3. Zhong has near equally distribution in two or three. that one type of noun is attracted. dimensional landmarks.. to a spatial construction. 4. Nei tends to cooccur with two dimensional landmarks. significantly.. and it emphasizes the frame or the boundary of the medium.

(40) 32 5. Factors investigated: Dimensionality, Boundedness. Generally speaking, there are three main concerns in previous research on CONTAINMENT particles in Chinese. The first limitation is related to the methodology. Most of the previous studies are qualitative in nature, using intuition data for analysis. A more quantitative corpus-based approach may be needed to render the findings of the studies more reliable and generalizable for cross-study comparisons. The second is related to the issue of statistical analysis. Few studies deal with the issue from a rigorous quantitative perspective. Liu and Zhang (2008) is the only quantitative corpusbased analysis, utilizing the frequency distribution of the landmarks in analyzing the semantics of the CONTAINMENT particles. However, the raw frequency is not enough for us to decide whether a noun is attracted to a construction or not. The raw frequency of a word in a construction can be biased by the marginal frequencies of the word (i.e. the noun) or the construction frequency (i.e. the spatial particle). If a word has high token frequency in the corpus, it would be expected to show a high raw frequency in the construction as well. Therefore, a more quantitative approach may take into consideration the potential influence of the marginal frequencies and further determine the importance of the observed frequency distribution by comparing it with the expected frequency distribution. This would ensure a more generalization account for the patterns. The third issue is on the semantic categorization of the landmarks. While Liu and Zhang (2008) have only analyzed the relation between spatial particles and physical landmarks, most of the previous studies do not have a consistent account for the categorization of other abstract.

(41) 33 nouns. Deng’s (2006) categorization is more descriptive, such as dynamic nouns, abstract idea, event nouns, which covers only a small set of abstract nouns. The present study hopes to analyze the abstract landmarks based on a more general and operational semantic property. Of particular importance to the semantics of abstract entities is the notion of boundedness discussed in Langacker (1987). According to Langacker (1987), bounded entities tend to have internal organization and are conceptually heterogenous while unbounded entities are viewed as homogenous or internally undifferentiated. To be more specific, the heterogenous nouns refer to the nouns that can be further divided into internal stages and possess temporal duration inherently. It is more event-like and is made up of different activities with internal temporal duration. For example, taolun ‘discussion’, consists of a series of actions and has a temporal duration and is regarded as a heterogeneous noun. For abstract nouns like guocheng “process’, although it might not exhibit strong dynamicity, it refers to a general concept of successive activities and is thus regarded as a heterogeneous noun. For the nouns like hope or faith, in which we cannot identify the constitutive states or activities, they are regarded as unbounded entities. The heterogeneity may serve as one possible factor to distinguish the cooccurring landmarks with the spatial particle constructions.. 2.3. Studies on Collostructional Analysis Collostructional analysis is grounded in two frameworks—one theoretical and one. methodological (Gries & Stefanowitsch, 2004a). First of all, the theoretical framework is based on the Construction Grammar, which views various levels of formal abstraction (i.e. morphemes, words, idioms, etc.) as a meaningful linguistic unit in its own right. The two important assumptions in Construction Grammar, introduced in Section 1.2, are repeated here: (1) if a word.

(42) 34 can occur in a construction, then the word must be semantically compatible with the meaning of the construction (Goldberg, 1995). (2) if two constructions are syntactically distinct, then they must be semantically or pragmatically distinct (Goldberg, 1995). Therefore, collostructional analysis emerges as a useful method to study the relationship between words and the grammatical structures they cooccur with (Stefanowitsch & Gries, 2003). The methodological framework is based on the quantitative corpus linguistics, featuring three methodological characteristics. First, naturally occurring linguistic data from a representative and balanced corpus is always needed for the linguistic investigation with this approach. Second, the linguistic phenomena in question should be extracted exhaustively from the corpus. Third, the data should be submitted to strict quantification and inferential statistics to determine the importance of the relationships. The Collostructional analysis highlights the importance of measuring the association strength between a construction and a word. It is quite different from the simple frequency-count method, since collostructional analysis builds a model of probability and can be used for generalization. (Gries & Stefanowitsch, 2004a, 2004b; Stefanowitsch, 2013; Stefanowitsch & Gries, 2003). There are three major variant techniques of collostructional analysis: the simple collexeme analysis, distinctive collexeme analysis and covarying collexeme analysis. The multiple collexeme analysis, which is the technique used in the present study, is an extension of the distinctive collexeme analysis. The three techniques in the collostructional analysis are briefly summarized in (6)..

(43) 35 (6) a. Simple collexeme analysis: the analysis that studies co-occurrence relation between a lexical construction (i.e. a word) and a grammatical construction, e.g. the V slot in the [X think nothing of Vgerund] (Stefanowitsch & Gries, 2003) b. Distinctive collexeme analysis: the analysis that studies one slot in two semantically or functionally similar constructions, e.g. the different semantics of verbs in the ditransitive and to-dative constructions of (Gries & Stefanowitsch, 2004b) b’ Multiple Distinctive collexeme analysis: the extended analysis that studies one slot in more than two semantically or functionally similar constructions, e.g. the V-slot in the causative construction (Gilquin, 2006). c. Covarying collexeme analysis: the analysis that studies the interaction between two slots within a construction, e.g. V1 and V2 in the [X V1 into V2gerund] causative construction (Gries & Stefanowitsch, 2004a). Simple collexeme typically studies co-occurrence relation between a lexical construction (i.e. a word) and a grammatical construction. For example, Stefanowitsch and Gries (2003) investigate the V slot in the [X think nothing of V]. The meaning of the construction refers to ‘to have low opinion of something’. The study adopts a simple collexeme analysis to uncover which verbs are more strongly attracted to the particular slot of this construction. Then those collexemes are ordered according to their association strength to the construction. In terms of interpreting the result, the collostructional strength refers to the association strength between the construction and the verb and is represented by p-value (how the p value is calculated will be demonstrated in Chapter 3). The smaller the p-value, the stronger the attraction. The ranking.

(44) 36 does pick out verbs denoting possible risk like mortgage, confine, motor, leap and fly, and verbs denoting an undesirable situation like haggle, rip, abandon. As a result, the simple collexeme analysis provides a more statistics-based result than purely intuitive analysis by showing the degree of attraction to a particular construction. The covarying collexeme analysis investigates the interaction between two slots within a construction. Two terms that are attracted to the slots can be further analyzed to uncover the semantic coherent patterns of the constructions using exploratory techniques, such as cluster analysis. For example, Gries and Stefanowitsch (2004a) look into the interaction between V1 and V2 slot in the [X V1 into V2gerund] causative construction. Most of the previous studies only discussed about the semantics of the verbs appearring in V1 slot, which often fall into categories like TRICKERY (i.e. trick, fool), PHYSICAL FORCE (i.e. force, bully), and VERBAL PERSUASION (i.e. coax, goad) (Stefanowitsch & Gries, 2003). The study aims to explore whether different semantic types of V1 might prefer different types of verbs in the V2 slot. They proposed a covarying collexeme analysis for identifying pairs of verbs that are strongly associated with each other in that construction, and the results support the assumption that different kinds of verbs in V1 slot would attract different kinds of verbs in V2 slot. Take the second significant covarying collexeme (i.e. torture and confess) for example. Both torture and confess are semantically related because they are both related to an act of communication that is not out of the speaker’s willingness. By using the covarying collexeme technique, we are able to understand the possible interaction between two potential collocates within a construction. Distinctive collexeme analysis studies one slot in two or more semantically or functionally similar constructions. For example, Gries and Stefanowitsch (2004b) use the method to investigate the dative alternation in English. They look into the ditransitive construction like.

(45) 37 John sent Mary the book and the to-dative construction like John sent the book to Mary. With the aid of distinctive collexeme analysis, we learn that in ditransitive construction, give is the verb with the strongest association strength to the construction. The verb encodes a face to face interaction between the agent and the recipient. Similar examples found in the distinctive collexemes are tell, show, offer, teach, ask. On the other hand, in to-dative construction, the verb with strongest attraction to the construction is bring, which indicates some distance between the agent and the recipient and the agent must take some time to complete the action. Similar examples found in the distinctive collexemes are take, pass, supply. To sum up, the distinctive collexeme analysis helps us to identify the possible difference between two or more semantically similar construction from a corpus-driven quantitative perspective. The multiple collexeme is an extension of the distinctive collexemes analysis, where more than two constructions are under investigation. Gilquin (2006) adopts this technique to investigate the 10 causative constructions as shown in Table 2-5. She pointed out that little is done on the lexical preference with the respect to V slot in these causative constructions. The result from multiple distinctive collexeme analysis indicates that constructions show very strong preference for different groups of verbs in the non-finite verb slot. [X MAKE Y Vinf] construction attracts verbs that cause a process that is not directly dependent on the CAUSEE. In other words, they are descriptive verbs like look, appear, seem and verbs denoting mental processes or psychological states like think and wonder without too much volition involved. The multiple distinctive collexeme analysis again shows its power in determining the subtle yet important differences among near-synonymous constructions. The present study will also adopt the same analysis method to identity the semantic difference among the three spatial particle constructions in Chinese..

(46) 38. Table 2-5 The Causative Constructions And The Examples. Construction. Example. [X CAUSE Y Vto-inf]. cause the price to fall. [X GET Y Vto-inf]. get Jes to do it. [X GET Y Vpp]. get everything sorted out. [X GET Y Vprp]. get these earphones working. [X HAVE Y Vinf]. have Roche inspect the hut. [X HAVE Y Vpp]. have the blade sharpen. [X HAVE Y Vprp]. have a blow lamp going. [X MAKE Y Vinf]. make our home seem so small. [X BE made Vto-inf]. be made to pay. [X MAKE Y Vpp]. make their voice heard.

(47) 39 3. Methodology. This chapter illustrates the procedures of our data analysis. Section 3.1 gives a brief overview of the corpus adopted in this study. Section 3.2 introduces the steps of data extraction. Section 3.3 presents two analyses (the multiple collexeme analysis, and post-hoc semantic analysis) to the extracted data. Section 3.4 is the summary of the whole chapter.. 3.1. The Corpus in the Present Study. The data analyzed in the present study came from the Academia Sinica Balanced Corpus of Modern Chinese 4.0, the so-called Sinica Corpus (Chen & Ma, 2010), which is the first and largest balanced corpus in Taiwan, developed by the Chinese Knowledge and Information Processing group at Academia Sinica. Sinica Corpus consists of 10,000,000-word writing samples collected from 1981-2007, covering different topics (i.e. philosophy, literature, society, science, and art) and genres (i.e. newspapers, fictions, drama scripts, stories, and conversations, etc.). Every text in the corpus has been automatically segmented into words and annotated with part-of-speech (POS) tags.. 3.2. Data Extraction Extracting the Constructions. The three constructions were extracted using the self-developed script written in R language. The script made use of the function ‘exact.matches.2.r’, developed by Gries (2009). The function requires a regular expression, which defines the target constructions to be extracted. To retrieve the three spatial particle constructions, i.e. [zai NP li], [zai NP nei], and [zai NP zhong], we defined a regular expression to extract strings with zai in the beginning, followed by noun.