This paper presents a revised version of the semantic map for oblique roles originally proposed by Haspelmath (2003) for instrumental and related functions and expanded by M. Zhang (M. Zhang 2008a, 2010) to accommodate more roles, particularly the Chinese-specific oblique role generally known among Chinese linguists as chùzhìshì shòushì 處置式受事 “pretransitive/disposal patient.” My purpose is twofold: (1) to demonstrate the usefulness of a “bottom-up” approach to the semantic map model as an innovative tool for the study of comparative Chinese dialectal and diachronic grammar; and (2) to discuss particular issues pertaining to the construction and revision of the oblique-role map, and to illustrate how such a map can contribute to our understanding of the typology and diachrony of oblique markers across regional varieties of Chinese.
為確定漢語特有的「處置式受事」角色的類型學地位，筆者（張敏2008–2010）曾將 Haspelmath（2003）提出的工具語及相關功能語義地圖擴充為一個初步的間接題元語義地圖。本文對該圖作出進一步修訂,旨在（1）為漢語方言比較語法和歷史語法學界推介一種簡便易行的「自下而上」的語義地圖建構方法，（2）探討間接題元語義地圖建構及修訂中遇到的一些疑難問題，並說明該圖在漢語方言間接題元標記的共時及歷時研究中的作用。(This article is in English.)
* This paper was originally presented at the Li Fang-Kuei Society Young Scholars Symposium in Celebration of the 10th Anniversary of the Establishment of the Li Fang-Kuei Society for Chinese Linguistics, University of Washington, Seattle, August 11-13, 2013. The research is supported by the Hong Kong rgc General Research Fund (grf) Grant # 645510 for the project “A Semantic Map Approach to Comparative Chinese Dialectal Grammar.”
1 Background: Methodological Underpinning
Multifunctionality or polysemy of one and the same grammatical form (i.e., a grammatical morpheme or a grammatical construction, the former abbreviated as “gram” henceforth, following Haspelmath 2003) is a complex and fascinating phenomenon that has attracted substantial attention from linguists studying Chinese and other languages, and typologists alike, for the asymmetrical, non-isomorphic correspondence between form and meaning (i.e., a one-to-many mapping) may open a door to linguistic inquiries from a wide array of dimensions.
To start with, let us take, for instance, the grams in various Chinese dialects which share the same phonetic form of the verb meaning “to give” (M. Zhang 2000). In Xuzhou 徐州 Mandarin, the same form [ke55] 給 has as many as nine gram uses, marking various oblique roles: (a) causee: gěi kàn bù gěi kàn 給看不給看? “Allow me to take a look?”; (b) passive agent: xiǎo yúr gěi māo tōu chī le 小魚兒給貓偷吃了 “The fish was eaten by the cat”; (c) disposal patient: gěi chuānghu kāikāi 給窗戶開開! “Open [marker] the window!”; (d) recipient: jì gěi tā 寄給他 “Mail (it) to him”; (e) beneficiary: gěi wǒ bànshì給我辦事 “Do it for me”; (f) comitative: gěi wǒ yīqǐ qù 給我一起去! “Go with me!”; (g) conjunctive: Xiǎolǐ gěi Xiǎowáng 小李給小王 “Li and Wang”; (h) locative: gěi hé nà yár wánr lái 給河那涯兒玩兒來 “(Someone) is playing at the river bank”; (i) source: gěi zhèr wǎng nán zǒu 給這兒往南走 “Walk from here toward the south.”
Multifunctional grams are complex, in part because describing any one such gram within any one language, like the Xuzhou [ke55], is a difficult task in itself. The easiest and simplest way to approach it is to take the homonymist position and treat the different uses as unrelated and merely homonymous. The treatment may not be tenable in this particular case, as the same multifunctional pattern is also found elsewhere: in the Yuyao (Ningbo) 餘姚 (寧波) dialect (Southern Wu), the form [tɐʔ55] 得, which is apparently etymologically unrelated to the Xuzhou [ke55], has all the same functions from (a) to (i) as well as of the verb use of “give.” If the multiple uses are not coincidentally coded by a homonymous form, then, how do we account for the interrelation? One possible solution from the monosemist position, which is often adopted by synchronic grammarians, is to claim that all such uses can be subsumed under a single “core” or common grammatical meaning. But such a “core” meaning, if it exists, would be too abstract and too vague to be meaningful at all (c.f. Haspelmath’s 2003 discussion). Alternatively, one may take the intermediate polysemist position, recognizing that (a) to (i) are different but interrelated meanings (without a common “core”) attached to the same gram, likely linked in a Wittgensteinian fashion of “family resemblance.” In other words, the different uses are “chained” to each other. But what exactly is the chaining pattern of the interrelation and how can we empirically justify its configuration? Moreover, does the “chain” represent a diachronic pathway of development (i.e., a series of extensions of meaning or grammaticalization)? Is the chaining pattern or cluster of related meanings, if proven to be common across languages, rooted in the mind and reflective of a universal conceptual structure? These questions show that the study of multifunctional grams has far-reaching implications in many subdisciplines, including synchronic syntax and semantics, historical grammar, cognitive linguistics, and the study of human conceptualization.
In the study on multifunctional grams, “comparison”—including intra-dialectal/intra-linguistic and inter-dialectal/cross-linguistic comparison—plays a critically important role. Comparison would inevitably provoke further complexity, but it is exactly the compounded complexity that holds the key to the discovery of hidden regularities. Table 1 below contains some exemplary data from our sample of multifunctional grams that have the same form of verb “give” in Chinese dialects and other languages of the world.
It reveals to us an intriguing picture of unity-in-diversity: the patterns of overlap vary, but the variation is by no means random. Clearly, comparison can lead us to see far more than what could be seen from a single gram in a single dialect/language. (For instance, we come to know that the nine uses of Xuzhou [ke55] can be grouped into a few clusters, one of them being [a][b][c][d][e].)
A question naturally arises from the observation: how do we maximally make sense of the data? What needs to be discovered here is a suitable analytical tool with which to conduct the comparison. Scientists facing a multifactorial issue may conceive it as a problem of metrics—that is, distance measurement (e.g., p-norm distance in Euclidean space)—and readily available tools to this end include clustering analysis, topology, combinatorics, etc.; but mathematics alone has little to offer in terms of the linguistic significance of the patterns. Chinese linguists, on the other hand, may approach the problem from various humanistic perspectives: structural (syntactic/semantic attribution), social-areal (language contact and diffusion), historical (pathways of change, grammaticalization), and cognitive (conceptualization, categorization, and metaphorical extension). In actual fact, with the recent upsurge of interest in comparative Chinese dialectal grammar and renaissance of diachronic syntax in the past two decades, such approaches have been extensively applied to the study of multifunctional grams, yielding a number of important findings (Zhu 1985 and 1990; Yue-Hashimoto 1991 and 1993; also refer to Chappell, Li, and Peyraube 2007 for a comprehensive overview). I myself have used these approaches, particularly in earlier works (M. Zhang 1998, 1999, 2000, 2001, and 2003). Nonetheless, the current paradigm is not ideal in that it lacks mathematical precision, and the claims are not always formulated in an explicitly falsifiable way. I believe that an optimal tool for handling multifunctionality of Chinese grams is one that has the potential to combine humanistic insights with scientific rigor. The Semantic Map Model—developed by functional typologists who care as deeply about language variation and change as do historical linguists and who, like social scientists, work with large-size probability samples—is exactly such a tool.
2 A “Bottom-Up” Approach to the Semantic Map Model
2.1 The Semantic Map Model (smm)
The smm is a technique that aims to represent non-isomorphic form-meaning correspondences in the study of linguistic typology. Though the model was first put forward more than 30 years ago (Anderson 1982), it was not put to great use until research in the last couple decades by Haspelmath (1997a, 1997b, and 2003) and Croft (2001, 2003; Croft and Poole 2008), among others. A semantic map is a geometrical representation of multifunctionality of grams across languages, thereby linking language-particular formal categories (cross-linguistically or intra-linguistically different grams, as variables) to universal semantic categories (“meanings/uses/functions,” as constants). Nodes on the map represent meanings/uses/functions associated with one or more grams, and are linked by connected lines representing relations between the uses of grams. The resulting configuration of functions in a certain domain (i.e., a “semantic map” or “conceptual space” for that domain) is claimed to be universal. It serves as a grid—comparable to a cartographic base map, upon which actual uses of a specific gram in a particular language can be projected or mapped—forming a language-particular and gram-specific semantic map. The construction of a semantic map is the product of empirical research involving laborious consideration of cross-linguistic data: the first task is to identify the polyfunctionality and overlap patterns in individual languages. The second task is to rearrange the relevant nodes so that all the uses carried by a gram are geometrically contiguous in the conceptual space, as required by the “Semantic Map Connectivity Hypothesis” which states that “any relevant language-specific and/or construction-specific category should map onto a connected region in conceptual space” (Croft 2001).
A prime example of the smm research is Haspelmath’s (1997a) cross-linguistic investigation of indefinite pronouns. Haspelmath identifies nine uses of indefinite pronouns found in languages of the world, which are carried by three series of pronouns in English (see Fig. 1 below): nobody which only has the use of (9) “direct negation” (as in “I have seen nobody”), somebody (or someone) which has the uses of (1)–(5) (e.g.,  “specific unknown” as in “I heard somebody, but I couldn’t tell you who,”  “irrealis nonspecific” as in “Please ask somebody else,” and  “question” as in “Did somebody tell you about it?”), and anybody which has the uses of (4)–(9) (e.g.,  “conditional” as in “If you see anybody, tell me immediately,” and  “comparative” as in “He is better than anybody else in the group”). With nine uses there are 29–1 = 511 combinations possible. However, in a sample of 140 languages, Haspelmath finds 133 indefinite pronouns showing only thirty-nine different combinations of uses, and he ultimately manipulates the thirty-nine combinations into a single configuration (as shown in Fig. 1), which is claimed to be a cross-linguistically universal “conceptual space.” Fig. 2 below shows the boundaries of three indefinite pronouns in English: somebody, anybody, and nobody. Note that the regions are all continuous on the universal conceptual space. The same is true for all indefinite pronouns in 40 languages examined by Haspelmath (1997a), without counterexamples.
The prowess of smm is multifaceted. First, it is a powerful tool to help us to uncover and represent language universals, and do it “in batch.” As maintained by Haspelmath (2003), a semantic map “embodies a series of implicational universals.” For example, Haspelmath’s (1997a) conceptual space for indefinite pronouns (Fig. 1) entails as many as seven implicational universal hypotheses: If a gram in a language has uses (1) and (3), it must have (2); if it has (3) and (8), it must have (5); etc. Second, semantic maps can also be used as an important tool for diachrony, as the connection patterns, when the lines are rendered into arrows, generally reflect historical paths of grammaticalization. Finally, semantic maps can be taken as “a direct representation of the relationships between meanings in speakers’ minds” (Haspelmath 2003: 233), or “the geography of the human mind, which can be read in the facts of the world’s languages in a way that the most advanced brain scanning techniques cannot even offer us” (Croft 2001: 364).
2.2 A Fieldworker-Friendly “Bottom-Up” Approach to smm
The SMM was originally intended as a typological tool for cross-linguistic comparison, rather than for the study of an individual language like Chinese. However, the essence of this methodology—to empirically uncover regularities in partially overlapping patterns of form-meaning correspondence through comparison—makes it applicable to any grammatical inquiries involving comparison, whether the comparison is among unrelated languages of the world or languages/dialects in the same taxon. The study of comparative Chinese dialectal grammar is no exception in this regard.
Nevertheless, applying a typological method to the study of an individual language or a particular taxon of languages/dialects may not always be the same as “doing typology” per se. Researchers of an individual language and typologists differ in their research interests, aims, and agenda. They also have different strengths and constraints in their studies. Working with large and well-balanced samples of world’s languages, typologists are better positioned to make reliable universal claims and are relatively free of genealogical and geographical bias, but the data they rely on are largely secondhand and practically difficult to thoroughly verify. Researchers of an individual language are in a much weaker position to make universal claims, but they enjoy better knowledge of and easier access to the comparative (and generally firsthand) data. In full recognition of the differences, we have been actively advocating a “bottom-up” approach to the SMM study of Chinese at various venues since 2008 (M. Zhang 2008a, 2008b, 2009b, 2010, and 2011). By “bottom-up” I mean that the work of semantic-map building starts with comparison of genealogical and areally related dialects, data that Chinese linguists are more familiar with and have better control over and easier access to. The initial results would be a network of implicational hierarchies that are at least “universal” within the limited range, say the Sinitic taxon, which may have direct bearings on Chinese linguistics. Then, more data can be gathered from non-Sinitic languages that Chinese linguists are more familiar with, like other Sino-Tibetan languages or Sinosphere languages—and ultimately, from other languages of the world—turning the initial maps into genuine universal claims. The work may terminate at any point of convenience; nonetheless it is valuable within the specific realm of the investigation.
The feasibility of the “bottom-up” approach has already been suggested by Haspelmath (2003), who notes from his extensive experience in building semantic maps that “it is not uncommon for different grams of the same language to overlap in their distribution…. So strictly speaking, one does not even need cross-linguistic comparison to construct a semantic map, because all we need is different grams with an overlapping distribution.” Comrie (1989), using relative clauses as an example, also hints that it is not impossible to establish an implicational universal using data from only a single individual language, given that the language has more than one grammatical form to convey the same function. In recent years, the “bottom-up” approach has been successfully tested with data from Chinese varieties to build semantic maps in various domains, such as oblique roles (M. Zhang 2008a, 2009b, and 2010; D. Zhang 2010; Sheng 2012); oblique roles and coordination (Phua and Ong 2011); resultatives (Wu 2009); tense, aspect, and modality (Fan 2009, 2011, 2012, and 2014; D. Zhang 2010); functions derived from “drop”-type verbs (Weng and Li 2010); adverbs (Guo 2010 and 2012; Y. Zhang 2013); spatial motion domain and related functions (Wang 2012); resultative complements (Chen, Guan & Wei 2012); conjunctions (Zhai 2012); interrogative and non-interrogative uses of WH words (Noda 2012); conditionals in Pekinese and other languages (Mao 2013); ditransitive constructions in Cantonese (Phua 2009a), Archaic Chinese (Phua 2009b and 2015), and Xiang dialect (Ding and Zhang 2012); among many others (Li, Zhang, and Guo 2015).
What follows is a concrete example to illustrate this approach. In order to provide a showcase of feasibility for the “bottom-up” approach, I (M. Zhang 2009b and 2010) purposefully chose a genealogically and geographically unbalanced sample of Chinese dialects to rebuild the semantic map for oblique roles in M. Zhang (2008a) (fig. 5a below), and I find that a large portion of it can be derived with the limited data, which are almost exclusively from Chinese dialects in a relatively small region, the Human Province. It is further noted that, with addition of data from only Archaic Chinese, the whole map in fig. 5a can be generated with no need of data from other languages (see Phua 2009b). Fig. 3 below is a fraction of the whole map as shown in fig. 5a. It consists of six nodes: C(omitative), human G(oal) (labeled as “direction/recipient” in fig. 5a), B(eneficiary), disposal P(atient), Co(njunctive), and I(nstrument). The comparison can start from grams with two and only two functions. For example, [põ33] 幫 in Xinhua 新化 (Xiang) has the functions of disposal patient marker and instrumental marker, and therefore the link “B-P” can be established. Similarly, the dual functionality of [tã44] 擔 as disposal patient marker and instrumental marker in Anren 安仁 (Gan) helps to determine the link “P-I.” Put the two minimal links together, and we will have the smallest one-dimensional semantic map “B-P-I.” We have tried another way to build the map, in which each round of comparison involves three nodes (the minimal number for deriving an implicational hierarchy) (M. Zhang 2009b and 2010). Any three of the six nodes form a unit, which can be tabulated in a Tic-Tac-Toe fashion for comparison (see tables 2 and 3 below). For example, of the three possible ways to arrange the three nodes in Unit I (i.e., cgb, cbg, bcg), two are eliminated by the connectivity requirement (“*” indicates that the pattern is unattested in the data), and only cgb remains. The resulting map is depicted in Fig. 3:
Note that each round of comparison results in an implicational hypothesis; for example, CGB can be read as the following: “If a gram is used as a marker for Comitative and Beneficiary, it must also be used to mark human Goal.” Hence, the semantic map in Fig. 3 forms a network of six implicational claims. Moreover, except for “with” from English and [ke55] 給 from Xuzhou, all the data from which Fig. 3 is generated come exclusively from a sample of fifty dialects spoken in a single province (Hunan) of China. Even without the help of English, the Hunan data alone are still sufficient to generate a one-dimensional map in the form of “Co—C—G—B—P—I,” which is largely on a par with the map in Fig. 3.
As indicated above, the smm has a great potential for application in the study of dialectal and historical grammar of Chinese. It is hoped that the “bottom-up” approach to smm could encourage more Chinese dialectologists to engage in typologically informed research. China has arguably the largest number of field-working dialectologists in any single country (as evidenced by the number of journal articles in cnki with the keyword “dialect”—7,918 in the recent two decades, or 396 articles annually), but contributes to linguistic typology far less proportionally, conceivably because typology appears to be a forbidden enterprise to Chinese fieldworkers since the majority of them have no easy access to comparative resources like balanced corpora and reference grammars of world’s languages generally published in the West; the language barrier also presents an additional problem. The bottom-up approach we advocate here presents a user-friendly typological methodology: echoing William Blake’s Auguries of Innocence, a fieldworker in Hunan may not have a chance “to see the world,” but he or she can see it “in a grain of sand” from his/her native province, and “hold infinity” if he or she just has a palmful of neighborhood data.
3 The Original Semantic Map for Oblique Roles: Its Achievements and Deficiencies
3.1 The Original Semantic Map for Oblique Roles
The semantic map for oblique roles (Fig. 5a below) proposed in M. Zhang (2008a, 2008b) is nothing but a slight expansion of Haspelmath’s (2003) map for instrumental and related functions (Fig. 4) as well as a portion of his map for typical dative functions. It was tentatively drafted as a showcase to advocate the application of smm in the study of Chinese in some informal venues; it was not meant for publication. It was not until the map was cited by several published articles that I felt the need to have a comprehensive revision.
This extension is motivated by two related objectives: (1) to pinpoint the location of the role commonly known as chùzhìshì shòushì 處置式受事 “pretransitive/disposal patient” (a role that is arguably unique to Chinese as well as some neighboring Tai-Kadai and Hmong-Mien languages) in the conceptual space of oblique functions; and (2) to see if Haspelmath’s map, with some additional nodes, could be regenerated with only Sinitic data. To a large extent, both the objectives have been achieved satisfactorily. First of all, the node of “pretransitive/disposal patient” is placed in a central position with only three immediately adjacent nodes—instrumental, beneficiary, and causative (or more precisely, causee). See the following section for a more detailed discussion. Secondly, though Haspelmath’s map (Fig. 4 above), which is built largely on the basis of data from Michaelis and Rosalie (2000), involves no data from Chinese itself, it has been verified by data from a significant number of Chinese varieties, leaving only a very limited number of exceptions.
The map in Fig. 5a was created from the Sinitic data from (i) ninety-six dialects (in the sense of “dialect points”) with polysemous grams involving “comitative,” (ii) ninety-five dialects with polysemous grams involving “disposal patient,” (iii) 229 dialects with polysemous grams involving “passive agent” or the verb “give,” and (iv) several historical varieties of Chinese. Below are examples of syncretistic patterns of relevant grams found in the Sinitic data; what follows is largely based on the data from which the map in Fig. 5 is constructed.
instrumental—disposal patient: [la21] 拿 in Bijie 畢節 (swm), [pa21] 把 in Yangxin 陽新 (Gan), [tã44] 擔 in Anren 安仁 (Gan), [tã55] 擔 in Longhui 隆回 (Xiang), [lau45] in Hengyang 衡陽 (Xiang), [to33] in Dongan 東安土話, etc.;
beneficiary—disposal patient: [po213] 把 in Jixi 績溪 (Hui), [põ33] 幫 in Xinhua 新化 (Xiang), [paɯ44] 幫 in Chenxi 辰溪 (Xiang), [paŋ311] 幫 in Xiuning 休寧 (Hui), and the same morpheme bāng 幫 in a number of Wu varieties in Zhejiang;
recipient—beneficiary—disposal patient—causee: [paŋ55] 幫 in Jishou 吉首 (swm), [pɒ23] 把 in Xupu 漵浦 (Xiang), [po24] 把 in Shuangfeng 雙峰 (Xiang), [pəʔ] 撥 in various Northern Wu dialects, etc.;
comitative—human goal: [lian53] 連 in Tongxin 同心 (nwm), etc.;
comitative—human goal—beneficiary: [tʻɔŋ35] 同 in Rugao 如皋 (SEW), hé 和 in a large number of Mandarin varieties, etc.;
comitative—human goal—beneficiary—disposal patient: [kaŋ31] 共 in Quanzhou 泉州 (Min);
conjunctive—comitative—human goal—beneficiary: [tʻoŋ11] 同 in Cantonese;
conjunctive—comitative—human goal—recipient—beneficiary: [kei42] 給 in Wenshang 汶上 (M), etc.;
conjunctive—comitative—human goal—beneficiary—disposal patient: [æ44] 挨 in Kunming 昆明 (swm), [kɛ55] 跟 in Yuanlin 沅陵鄉話, [ken44] 跟 in Ruijin 瑞金 (Gan), [tʻuŋ11] 同 in Meixian 梅縣 (Hakka), etc.;
conjunctive—comitative—human goal—beneficiary—disposal patient—instrumental: [na11] 拿 in Linwu 臨武土話, etc.;
beneficiary—cause: [tʻiɛ31] 替 in Shicheng 石城 (Hakka), etc.;
causee—passive agent: ràng 讓, jiào 叫, jiào 教 in a number of Mandarin varieties such as Pekinese;
disposal patient—causee: bă 把 in a number of Mandarin varieties such as Pekinese;
disposal patient—causee—passive agent: [tɕʻiau312] 叫 in Yexian 葉縣 (M);
beneficiary—cause—passive agent: wèi 為 in Archaic Chinese.
The functions of each gram mentioned above occupy a contiguous region on the map in Fig. 5a. For more examples of specific grams whose functions are represented as bounded regions on the map, see the figures in the Appendix. Fig. 5b (for four grams in three varieties of Southern Min), Fig. 5c (for five grams in Xuzhou, Suzhou, and Xinhua), and fig. 5d (for five grams in five non-Sinitic languages of China) are taken from M. Zhang (2008a, 2008b). Phua (2009a) examines the gram pei 俾 in early Cantonese as recorded in 1828, and finds that its multifunctional pattern also fits well with the map (Fig. 5e in the Appendix). Phua (2009b) demonstrates that the map can reveal not only polysemous patterns of individual oblique markers, but also those of constructions; he finds that the semantic extensions exhibited by both “V+yu+Indirect Object” and “V+Direct Object+yu+Indirect Object” constructions in Archaic Chinese occupy different contiguous areas on the oblique-role map (Fig. 5e in the Appendix). Phua and Ong (2011) further extend Fig. 5a into Fig. 5f (see the Appendix), with several new nodes introduced by Haspelmath (2004), and find that the preposition/conjunction yǐ 以 in Archaic Chinese also forms a contiguous area on the extended map, despite that it has as many as eleven functions.
3.2 Problems to be Resolved
Though the original map for oblique roles passes the test in various geographical and historical varieties of Chinese as well as some non-Han languages, it suffers from several salient problems, which motivate this revision. There are at least three major problems, the loci of which are marked in Fig. 6.
Problem 1: The linking pattern “direction–recipient–beneficiary” is suggested by Haspelmath (2003) in his map of typical dative functions. However, it faces a difficulty when it comes to the data from Chinese, which puts us in a dilemma. On one hand, a number of grams in various Chinese dialects, such as gēn 跟 and hé 和 in Southwestern and Southeastern Mandarin, share the functions of direction and beneficiary, but not that of recipient, thus resulting in a discontiguous region. On the other hand, the linking pattern receives wide support from languages of the world, and it is also attested in some southern Chinese varieties like certain Gan 贛and Min 閩 dialects, where the syncretism of direction and recipient and that of recipient and beneficiary coexist in one and the same dialect, suggesting that the three nodes have to be arranged in Haspelmath’s way. The tentative merger of the two nodes in the either-or form “direction/recipient” in Fig. 5a is not meant as a solution to this problem, but rather as a work-around to be corrected later.
Problem 2: As indicated earlier, one of the main purposes of our extension of Haspelmath’s map is to promote the application of SMM in Chinese dialectology by providing a showcase illustrating that a map built on non-Sinitic data can be readily used to account for distributional patterns of Chinese grams, only with some slight amendment by inserting certain missing nodes into the map. Therefore, we must try to retain the original configuration of Haspelmath’s map as much as possible. This attempt creates some unexpected problems. One of them lies in a missing but necessary link between the nodes “source” and “direction.” Note that “source” has to be put adjacent to “passive agent,” as evidenced by the multifunctional pattern of the gram von in German. But in many languages, Chinese included, the role “source” is intimately linked with some other spatial roles, one of which is “direction.” Take, for example, the preposition zǒu 走 in Chinese, which is grammaticalized from the verb meaning “walk, run.” In quite a number of Chinese dialects (e.g., Hengyang 衡陽, Bijie 畢節, and Chengdu 成都), the gram can be used to mark direction and source, and sometimes location as well, but not any of the non-spatial roles on the map. This is not a problem for Haspelmath’s map, which can be rearranged as in Fig. 7 (see the Appendix), without losing any original insights. But it is problematic for our extended map. If the nodes “passive agent” and “source” keep their original positions in the lower right corner, the addition of the nodes “pretransitive/disposal patient” and “causative/cause” makes it impossible for the similar rearrangement to rescue the map from violating the connectivity requirement of SMM.
Problem 3: This problem is inherent to Haspelmath’s map (Fig. 4 above), which posits the linking pattern of “instrumental–passive–cause.” The syncretism of the three functions is indeed found in languages of the world (though in Chinese we can only find the syncretism of passive and cause); nevertheless Narrog and Ito (2007) point out that there are languages which lack a passive but still have the syncretism of the functions of instrumental and cause, and there are also languages where the roles of instrumental and cause are marked by one and the same gram, but where passive agent is marked by another gram. Therefore, Narrog and Ito (2007) argue that the passive node should not be placed between the nodes of instrumental and cause.
4 Revision of the Map
4.1 Resolving the Problems
Problem 3 is the easiest to solve, and it can be readily rectified by turning the one-dimensional link “instrumental–passive–cause” into a two-dimensional loop by closing up the three nodes.
Problem 2 is the most difficult one. To a certain extent it is not even solvable if we persist in (1) taking into account a large enough sample of languages of the world, and (2) restricting the dimensionality of maps to 2d—that is, disallowing crossing or curved lines to connect the nodes on a map. See below for a more detailed discussion.
I now turn to the solution of Problem 1, which is relatively more complicated than Problem 3, but still soluble. The problem actually arises from an under-differentiation of the relevant functions. The role “recipient” is also known in the literature as “human-goal”—the endpoint of an event of transference (thus “goal”) of possession (thus “human”). The role “direction” is also referred to as “goal,” prototypically the goal of a motion event. However, there is yet another type of “direction” or “goal” that is neither the goal of a motion nor the recipient of a transferred object. It is yet another kind of “human goal” in that a human or an animate being, while not physically affected by an action, is the target of a non-motion action.
Fig. 8 indicates that both “direction” and “human goal” have two subtypes, which partially overlap at the function of “to” as in “nod to him.” In various typological studies involving obliques, the above roles are invariantly divided bipartitely into “recipient” (or “dative” as it is the prototypical function of the dative case) and another role, labeled as “goal” by Heine (1990) and Malchukov and Narrog (2009), “allative” by Luraghi (2001), “goal/allative” by Yamaguchi (2004), and “direction” by Haspelmath (2003). This fact suggests that it may be rare for languages of the world to separately code the three-way distinction as shown in Fig. 8 (note that all the three functions are carried by the single gram “to” in English). However, Chinese is exactly such a language. Compare the following:
a. jièqiángěi tā借錢給他 “lend money to him”
lend money to3sg
b. gěi tālǐ-fà給他理髮 “do a haircut for him”
for 3sg cut-hair(Mandarin varieties)
a. duì wǒ shuō對我說 “speak to me”
b. duìwǒ diǎn-diǎn-tóu對我點點頭 “nod to me”
toward 1sgnod-nod-head(Mandarin varieties)
a. gēntādiǎn-tóu跟他點頭 “nod to him”
with 3sg nod-head
b. gēn tālǐ-fà跟他理髮 “do a haircut for him”
for3sg cut-hair(Mandarin varieties)
tɕi21 mai23 tɛʔit5taʔ54 ȵiu21 tau45 ŋo21 li 渠賣得一隻牛到我俚
3sgsellRES one CLcowtomyfamily“He sold a cow to my family.”(Anyi variety of Gan 安義贛語)
vuai213 ɦiok3 xu35 i213賣肉去伊 “sell meat to him”
Sellmeat to3sg(Tunchang variety of Min 屯昌閩語)
In a number of Mandarin varieties, the gram gěi 給 (< verb “give”) marks recipient (1a) and beneficiary (1b), but not direction, in whatever sense. The gram duì 對 carries the function of direction in the second sense (i.e., [ii] in Fig. 8), as in (2), but it does not mark recipient, nor direction in the first sense. The gram gēn 跟 shows a syncretism between direction in the second sense (interchangeable with duì in  in this sense) and beneficiary, as in (3a) and (3b) respectively. The function of “direction of a motion” may be marked by grams like dào 到 “to” or wǎng 往 “toward” as in wǎng qián zǒu 往前走 “go forward (lit. toward front go)” in Mandarin, with no other functions. Following the common practice in SMM studies, if a function, say, “direction,” can be subdivided into two functions that are expressed by two separate morphemes in some language, then we have to postulate two nodes rather than one (de Haan 2004). Therefore, the tripartite scheme in Fig. 8 becomes a necessity in order to accommodate the Chinese data.
I proposed the following solution to Problem 1 (M. Zhang 2010). Following the convention in van der Auwera and Plungian (1998), we use a big oval X to represent the broad category of “human goal,” which includes two subsumed categories (i.e., Y and Z). (The arrowed line stands for “generalization.”)
Though this solution can easily account for the observed multifunctional patterns, including those found in Chinese, it is nevertheless undesirable for two reasons. (1) It does not specify which nodes on the left and right Y and Z are closest to; and therefore, the predictive power of the map is inevitably weakened. (2) The way X, Y, and Z are represented is inconsistent with the convention by which the other nodes are arranged on the same map.
The data presented in (4) and (5) may give us a clue for a better solution. In the majority of modern dialects, the recipient marker is derived from the general-purpose verb of “giving,” and normally syncretistic with beneficiary, rather than with direction. However, the syncretism between direction of motion and recipient can nevertheless be found in some southern dialects, particularly Gan and Wu varieties, as well as some Min varieties in Hainan. (See M. Zhang 2009a and 2011 for a comprehensive survey.) For example, the gram [tau45] 到 (< “arrive, reach”) in Anyi 安義 is used primarily as a marker for direction of motion, but it also serves as the recipient marker in a dative construction, as shown in (4). Similarly, [xu35] 去 (< “go, leave”) in Tunchang 屯昌 is used to encode recipient in (5), but it can also mark direction of motion, as in 駛去海口 “drive to Haikou (lit. drive go Haikou)”. Given that these two grams carry no other function (unlike to in English and yú 於/于in Archaic Chinese), it becomes evident that the nodes “direction of motion” and “recipient” have to be placed immediately adjacent to each other on the map.
We are now inclined to give up our previous solution in M. Zhang (2010), and return to the norm of node representation in the smm tradition. In what follows, we use the following terms to refer to the three functions in fig. 9 on the last page: (1) “direction,” which is reserved to refer to the direction of motion in space only (i.e., function [i]); (2) “human goal,” in an ad hoc sense referring to function (ii) (i.e., direction of a non-motion action toward a human or other animate being); and (3) “recipient” (i.e., function [iii]).
The above mini-map consists of two closed loops, one linking up “direction–human goal–recipient,” and the other “human goal–recipient–beneficiary.” It passes the verification checks with all data available to us, as exemplified by the four gram-specific semantic maps in Fig. 10, which are all connected subgraphs of the underlying generic map (or, conceptual space, using Croft’s 2001 term).
4.2 Redrawing the Map
The revised version of our semantic map (conceptual space) for oblique roles is presented as Fig. 11 below.
The redrawing of the map is greatly facilitated by the attainment of additional Sinitic data from the RGC-funded project on semantic maps (645510) and recent smm-related research involving oblique roles, particularly Stolz et al. (2006), Malchukov and Narrog (2009), Narrog and Ito (2007), Narrog (2010), D. Zhang (2010), and Wang (2012). Among them, research conducted by D. Zhang (2010) and Wang (2012), both containing sizeable Chinese data and both adopting the bottom-up approach to construct their maps, provide the most pertinent inputs for the redrawing. D. Zhang (2010) provides a semantic map study on polysemous grams that share at least the “instrumental” or “comitative” function, and his data originate from thirty Chinese dialects and eighty-six languages of the world. Fig.12 (in Appendix) is D. Zhang’s (2010) conceptual space for “instrument/comitative” grams in Chinese dialects, and Fig. 13 above is his conceptual space for world’s languages. Wang (2012) is a comprehensive survey of multifunctional grams that mark thematic roles in the spatial motion domain. The data this study relies upon come from 148 Chinese dialects (449 grams in total) and 63 languages of the world (201 grams in total). Wang’s semantic map, drawn from data exclusively from Chinese dialects, can be found in Appendix (Fig. 14), and her map based on data from both Chinese dialects and other languages of the world can also be found there (Fig. 15).
Note that the three maps (Fig. 11, Fig. 13, and Fig. 15) are largely compatible with one another, but nonetheless they exhibit obvious differences in linking patterns and modes of visualization. We discuss but a few of the most methodologically significant issues, which are germane to the differences between our map (Fig. 11) and the other two maps, particularly D. Zhang’s.
What follows are the major differences between Fig. 11 and Fig. 13: (1) The node “possessor” has a direct link with “beneficiary” in Fig. 11, which is absent in Fig. 13. (2) Fig. 13 follows the traditional dichotomy between “direction” and “recipient,” while Fig. 11 adopts the tripartite distinction of goal-related roles, as discussed in 4.1. (3) Fig. 13 posits a direct link between “recipient” and “causee,” which is absent in Fig. 11.
For difference (1), the most robust evidence supporting our map comes from the recurrent polysemous patterns of ditransitive constructions in world’s languages, including ancient and modern varieties of Chinese. One such pattern is the syncretism among possessor, beneficiary, recipient, and direction (see Malchukov et al. 2007; M. Zhang 2011; and Phua 2015 for details). For example, the double-object construction (DOC) in Archaic Chinese exhibits the same syncretism among the aforementioned four roles (Phua 2005). Consider the following:
cì Qí hóu zuò 賜齊侯胙 (左傳：僖.9 Zuozhuan: Xi.9)
bestow Qi duke meat “bestow meat to the Duke of Qi”(recipient)
tóu gān dōng-hǎi 投竿東海 (莊子：外物 Zhuangzi：Waiwu)
throw pole east-sea “throw the pole into the East Sea”(direction)
tiān shēng mín ér lì zhī jǖn 天生民而立之君 (左傳：襄.14 Zuozhuan: Xiang.14)
heaven bear folk and establish 3pl monarch
“Heaven gives birth to the people and sets them rulers.”(beneficiary)
duó zhī niú 奪之牛 (左傳：宣.11 Zuozhuan: Xuan.11)
rob 3sg buffalo “rob him of the buffalo”(possessor/ maleficiary)
The same form of DOC can encode a constructional meaning that involves either “recipient” (6), or “direction of motion” (7), or “beneficiary” (8), or “beneficiary/maleficiary” (9). Note that the function node “beneficiary” on our map actually stands for both “beneficiary” and “maleficiary.” Example (9) is notorious for its ambiguity between a maleficiary reading (parallel to modern Chinese qiǎng tā yītóu niú 搶他一頭牛 “rob a buffalo from him”) and a possessive reading (=duó qí niú 奪其牛 “rob his buffalo”); it has triggered a hot and long-lasting debate among Chinese historical grammarians on whether Archaic Chinese has this type of malefactive DOC or not. The ambiguity of (9) clearly indicates that there must be a direct connection between “possessor” and “beneficiary” (“maleficiary” included).
Difference (3) merits special attention, as it points to a methodological pitfall one should try to avoid in drawing semantic maps.
4.2.1 Polygrammaticalization as a Source of Noise
D. Zhang’s (2010) postulation of the direct link between “recipient” and “causee” is primarily based on the extensive existence of the recipient-causee syncretism in Chinese dialects. For example, in a modern dialect, if its recipient marker is derived from the general-purpose verb of “giving,” chances are that the same form can also be used as a causative marker, such as in sòng gěi tā 送給他 “give (it) to him” (gěi as the recipient marker) and bù gěi wǒ kàn 不給我看 “do not allow me to look” (gěi as the causative marker). We have made the same observation (c.f. M. Zhang 2000), but are not led by superficial observation to the conclusion that the two nodes must be immediately adjacent to each other. We believe that the two nodes have no direct conceptual connection, nor derivational relationship, since the syncretism is the result of polygrammaticalization from the same lexical source—that is, the general-purpose verb of “giving.”
Chappell and Peyraube (2006) propose that “give” verbs in modern dialects and historical varieties of Chinese have at least two different pathways of grammaticalization, explicable in terms of polygrammaticalization (see also Chappell, Li, and Peyraube 2007). The pathways are the following:
V [+ “give”] > dative marker
V [+ “give”] > V [+ causative] > passive marker
The argument is based on the observation that (1) it is impossible for a dative marker to directly develop into a causative marker; and (2), vice versa, it is impossible all the same for a causative marker to evolve into a dative marker, as the grammaticalization into the two functions occurs at distinct sites in two different kinds of serial verb construction. Our investigation of some other languages also supports the analysis by Chappell and Peyraube (2006). Besides Chinese, the “recipient–causee” syncretism is also found in some Southeast Asian and African languages, such as Betta Kurumba (Dravidian), Thai, Lao (Tai-Kadai), Plang, Tetun Dili (Austronesian), So, Kasong (Austroasiatic), and Ewe (Niger–Congo). Interestingly, in these languages the recipient-cum-causee markers are also homonymous with their verbs of “giving,” suggesting that the “recipient–causee” syncretism in such languages is also the result of the same polygrammaticalization as found in Chinese varieties.
As pointed out by Malchukov (2010), “viewed diachronically, polysemies arise from meaning extensions of individual categories. In most cases, such meaning extensions are gradual and proceed stepwise along the network of functions on a map without incurring a contiguity violation (Croft et al. 1987). However, there are more complex diachronic scenarios, . . . . . . , which might be problematic for semantic maps.” The “polygrammaticalization effect” is one of the diachronic factors which may cause violations to the contiguity of semantic maps, according to Malchukov (2010). We note that the polygrammaticalization effect could also be misleading even when it does not result in a contiguity violation.
Take, for example, the “instrumental–passive agent” syncretism. It is widely attested in world’s languages, and the conceptual relatedness between the two functions is beyond any doubt. Therefore, the two nodes are immediately linked to each other on our original map for oblique roles (Fig. 5a), the revised map (Fig. 11), and D. Zhang’s map (Fig. 13). Consider the following examples in Haikou (Min) 海口閩語:
[io55] 杓 来 探 糜。
io spoon come scoop rice “scoop some rice with a spoon” (instrumental)
车 [io55] 侬 偷 喽。
car io 2sg steal prt “The car was stolen by him.” (passive agent)
[io55] 两 個 碗 来!
io two cl bowl come “Bring two bowls here!” (verb: “to take”)
大姊 [io55] 一 支 笔 去 伊。
big-sister io one cl pen to 3sg (verb: “to bring”)
“The big sister gave him a pen. (lit. The big sister brought a pen to him.)”
[io55] 一 叶 纸 我。
io one cl paper 1sg “Give me a piece of paper.” (verb: “to give”)
天 落 大 雨 [io55] 我 无 成 去。 (causative marker)
sky drop big rain io 1sg neg achieve go
“The heavy rain prevented me from going. (lit. The heavy rain caused me not be able to go.)”
One may be misled by examples (1) and (2) into believing that the “instrumental–passive agent” syncretism also occurs in Chinese. This is an illusion rather than the reality, and a genuine syncretism of the two functions has never been attested in any variety of Chinese. The illusion is caused by the “polygrammaticalization effect.” If the other sentences above are taken into consideration, it becomes clear that the word [io55] is first of all used as a full-fledged verb (in ) meaning “to take, to hold, to bring” (which is etymologically related to what is written as yuē/yào/è 約/要/搤, etc., in a number of southern Wu, eastern Gan, and some Min dialects; see M. Zhang 2011), and the same verb can be used in (4) to express a ditransitive meaning, and then evolves into a general-purpose verb of “giving” (as in ). The general-purpose verb of “giving” [io55] may then follow the common path of grammaticalization to develop into a causative marker (as in ), and then a passive marker (as in ).
Therefore, we concur with Malchukov (2010) that polygrammaticalization effects should be viewed and treated as noises that must be filtered out. After removing the direct link between “recipient” and “cause,” we are able to relocate the node of recipient to the exterior side of the map, thus making the connection between “possessor” and “beneficiary” possible on the map.
4.2.2 Dimensionality of the Semantic Map: How to Deal with the Crossing Lines
Let us finally return to the unsolved Problem 2, which, unfortunately, remains unsolved after the revision. Note that on all three maps—our revised map (Fig. 11), D. Zhang’s world map (Fig. 13), and Wang’s world map (Fig. 15 in Appendix)—the crossing lines persist.
it may well turn out that in a number of functional domains non-vacuous maps are impossible to construct, but experience shows that there are many areas in which there are very strong universal restrictions, so that interesting maps can be drawn. As soon as more than three functions are considered, a vacuous map would have to involve crossing or curved connecting lines, i.e., in fact more than two dimensions. So it is a general rule that the fewer dimensions and the fewer connecting lines a map shows, the more predictions it makes and the more interesting it is.
The Interface Hypothesis:
On a largely non-vacuous map, if three-dimensionality is unavoidable, the third dimension must occur at the interface nodes linking two separate domains.
Note that on our revised map (Fig. 11), the three crossing lines occur at the juncture of two separate domains: the spatial domain (involving “direction,” “location,” and “source”) and the non-spatial oblique role domain; so do the two crossing lines on D. Zhang’s (2010) world map. The theoretical significance of this hypothesis lies in the future growth of the SMM endeavor, from building individual semantic maps to weaving a network of semantic maps covering the whole cognitive system underlying grammatical structure. If what Croft (2001) claims is true, that “the conceptual space is the geography of the human mind, which can be read in the facts of the world’s languages in a way that the most advanced brain scanning techniques cannot ever offer us,” then subcomponents of the conceptual system must be connected with one another in a principled way. In this regard, the Interface Hypothesis may be reinterpreted as yet another Semantic Map Connectivity Hypothesis. The difference between the old and the new connectivity hypothesis is that the former governs the connectivity within a semantic map (conceptual space), while the latter controls the connectivity between two or more semantic maps (conceptual spaces).
ChenYing, SiyiGuan, and Wei Hang 陳穎，關思怡，魏航. 2012. “Cóng yǔyì dìtú móxíng kàn yǔfǎ kuòzhǎn-xìng: yǐ chéngdù bǔyǔ wéilì” 從語義地圖模型看語法擴散性：以程度補語為例 [Functional expansion of grams viewed from the semantic map perspective: A case study of resultative complements in Chinese]. Paper presented at the Symposium for Semantic Map Studies on Multifunctional Grammatical Forms in Chinese, April 6–8, 2012, Peking University. [A revised version appears in Li, Zhang, and Guo eds. (2015).]
Ding, Jiayong 丁加勇, and Zhang Min 張敏. 2012. “Cóng Xiāng fāngyán dòngcí jùshì kàn shuāngjíwù jiégòu yǔyì dìtú” 從湘方言動詞句式看雙及物結構語義地圖 [The semantic map of ditransitives as viewed from verbal constructions in Xiang]. Paper presented at the Symposium for Semantic Map Studies on Multifunctional Grammatical Forms in Chinese, April 6–8, 2012, Peking University. [A revised version appears in Li, Zhang, and Guo eds. (2015).]
Fan, Xiaolei 范曉蕾. 2009. “Cóng Hànyǔ fāngyán de duōyì qíngtàicí kàn qíngtài gàiniàn de yǔyì guānlián” 從漢語方言中的多義情態詞看「能性」情態概念的語義關聯 [The semantic relatedness among notions of possibility modality as revealed from multifunctional grams in Chinese dialects]. MA thesis, Peking University.
Guo, Rui 郭銳. 2010. “Fùcí de bǔchōng yì yǔ xiàngguān yìxiàng de yǔyì dìtú” 副詞的補充義與相關義項的語義地圖 [The semantic map of Chinese adverbs in additive and other functions]. Paper presented at the International Symposium for Comparative and Typological Research on Languages of China, May 8–9, 2010, Hong Kong University of Science and Technology.
Kimura, Hideki 木村英樹. 2005. “Běijīnghuà ‘gěi’-zìjù kuòzhǎn wéi bèidòngjù de yǔyì dòngyīn” 北京話「給」字句擴展為被動句的語義動因 [Semantic motivation for the extension of gei in Pekinese into a passive marker]. Hanyu Xuebao, no. 2.
Noda, Hiromichi 野田寬達. 2012. “Shenme kuà yǔyán fènxī: Jīyú yǔyì dìtú” 「Shenme」跨語言分析: 基於語義地圖 [A cross-linguistic study of shenme ‘what’ based on the semantic map model]. Paper presented at the Symposium for Semantic Map Studies on Multifunctional Grammatical Forms in Chinese, April 6–8, 2012, Peking University. [A revised version appears in Li, Zhang, and Guo eds. (2015).]
Phua, Chiew-Pheng 潘秋平. 2009. “Cóng yǔyì dìtú kàn Guǎngdōngsheng Tǔhuà Zìhuì zhōng de ‘bǐ” 從語義地圖看《廣東省土話字彙》中的 “俾” [Pei in the Vocabulary of the Canton Dialect: A semantic map approach]. Proceedings of the 13th International Conference on Cantonese and Other Yue Dialects. Hong Kong: City University of Hong Kong Press.
Sheng, Yimin 盛益民. 2012. “Yǔyì dìtú de bù liánxù hé lìshǐ yǎnbiàn: Yǐ Shàoxìng Kēqiáo xūcí ‘zuò’, ‘bō’ wéilì” 語義地圖的不連續和歷史演變:以紹興柯橋虛詞「作」、「撥」為例 [Semantic map disconnectivity and historical change: A case study on the grams [tso33] and [peʔ5] in Shaoxing (Keqiao) variety of Wu]. Paper presented at the Symposium for Semantic Map Studies on Multifunctional Grammatical Forms in Chinese, April 6–8, 2012, Peking University. [A revised version appears in Li, Zhang, and Guo eds. (2015).]
WengShanshan, and Li Xiaofan 翁姍姍，李小凡. 2010. “Cóng yǔyì dìtú kàn xiàndài hànyǔ ‘diào’-lèi cí de yǔyì guānlián hé xūhuà guǐjì” 從語義地圖看現代漢語「掉」類詞的語義關聯和虛化軌跡 [The semantic relatedness and the pathway of grammaticalization of the “dào”-type of verbs in Modern Chinese from the semantic map perspective]. Yuyanxue Luncong 42. Beijing: Commercial Press.
Wu, Fuxiang 吳福祥. 2009. “Cóng ‘dé’-yì dòngcí dào bǔyǔ biāojì—Dōngnányà yǔyán de yīzhǒng yǔ fǎhuà qūyù” 從「得」義動詞到補語標記—東南亞語言的一種語法化區域 [From the verb meaning “get, acquire” to the resultative construction marker: A kind of grammaticalizaion area in the Southeast Asian languages]. Zhongguo Yuwen 3: 195–211.
Zhai, Yun 翟贇. 2012. “Cóng yǔyì dìtú móxíng kàn ‘hé’ yì cí de gōngnéng” 從語義地圖模型看「和」義詞的功能 [Functions of ‘and’-type of grams from the semantic map perspective.] Paper presented at the Symposium for Semantic Map Studies on Multifunctional Grammatical Forms in Chinese, April 6–8, 2012, Peking University. [A revised version appears in Li, Zhang, and Guo eds. (2015).]
Zhang, Min 張敏. 1999. “Hànyǔ fāngyán tǐcí chóngdiéshì yǔyì móshì de bǐjiào yánjiū” 漢語方言體詞重疊式語義模式的比較研究 [A comparative study on semantic models of reduplicative forms across Chinese dialects]. In Hànyǔ Fāngyán Gòngshí yǔ Lìshí Yǔfǎ Yántǎo Lùn wénjí [Synchronic and diachronic grammar of Sinitic languages], ed. WuYunjin, 1–31. Guangzhou: Jinan University Press.
ZhangMin. 2001. “Syntactic Change in Southeastern Mandarin: How Does Geographical Distribution Reveal a History of Diffusion?” In In Memory of Professor Li Fang-Kuei: Essays of Linguistic Change and the Chinese Dialects, ed. Ting Pang-Hsin and Anne O. Yue, 197–242. Taipei and Seattle: Academia Sinica and University of Washington, 2000b.
Zhang, Min 張敏. 2003. “Cóng lèixíngxué kàn shànggǔ hànyǔ dìngyǔ biāojì ‘zhī’ yǔfǎhuà de láiyuán” 從類型學看上古漢語定語標記「之」語法化的來源 [The grammaticalization of attributive marker zhi in Archaic Chinese: From a typological perspective]. In Yǔfǎhuà yǔ Yǔfǎ Yánjiū [Studies in Grammar and Grammaticalization], vol. 1, ed. Wu Fuxiang and Hong Bo. Beijing: Commercial Press.
Zhang, Min 張敏. 2008a. “Kōngjiān dìtú hé yǔyì dìtú shàng de ‘cháng’ yǔ ‘biàn’: Yǐ Hànyǔ bèi dòng, shǐyì, chùzhì, gōngjù, shòuyìzhě děng guānxì biāojì wéilì” 空間地圖和語義地圖上的「常」與 「變」: 以漢語被動、使役、處置、工具、受益者等關係標記為例 [Universals and particulars manifested on spatial maps and semantic maps: On passive, causative, disposal, instrumental, beneficiary markers across Chinese dialects]. Academic talk delivered on January 10, 2008, at the Linguistics Institute, the Chinese Academy of Social Sciences, Beijing.
ZhangMin. 2009a. “How to Give in a Language without ‘Give’: Towards a New Typology of Ditransitives in Sinitic Languages.” Paper presented at the 17th Annual Meeting of the International Association of Chinese Linguistics (IACL-11), École des Hautes Études en Sciences Sociales, Paris, July 2–4.
Zhang, Min 張敏. 2009b. “Rúhé cóng yīgè shěng de Hànyǔ fāngyán yǔliào dǎochū rénlèi yǔyán gòngxìng guīlǜ—Húnán fāngyán jiècí de yǔyì dìtú yánjiū” 如何從一個省的漢語方言語料導出人類語言共性規律—湖南方言介詞的語義地圖研究 [Deriving language universals from dialects of a single Chinese province: A semantic-map study on prepositions in Hunan dialects]. Paper presented at the 5th International Symposium on Grammaticalization in Chinese, August 20–23, 2009, Shanghai Normal University, Shanghai.
Zhang, Min 張敏. 2010. “Yǔyì dìtú móxíng—yuánlǐ, cāozuò jí zài Hànyǔ duōgōngnéng yǔfǎ xíngshì yán jiū zhōng de yùnyòng” 語義地圖模型—原理、操作及在漢語多功能語法形式研究中的運用 [The semantic map model: Theory, methodology, and its application in the study of multifunctional grams in Chinese]. Yuyanxue Luncong 42: 3–63. Beijing: Commercial Press.
Zhang, Min 張敏. 2011. “Hànyǔ fāngyán shuāngjíwù jiégòu nánběi chàyì de chéngyīn: Lèixíngxué yánjiū yǐnfā de xīn wèntí” 漢語方言雙及物結構南北差異的成因: 類型學研究引發的新問題 [Revisiting the alignment typology of ditransitive constructions in Chinese dialects]. Bulletin of Chinese Linguistics, 4 (2): 87–270.