Lexicography for loanwords and words with special orthography

Abstract

Loanwords in Modern Standard Chinese have attracted increasing attention in recent years. We try to unravel the complexity that stems from the properties of the Chinese writing system itself, its variant systems (Japanese Kanji), the rich history of the Chinese lexicon, and its interactions with neighboring languages. We argue that the status of a loanword is not binary, but rather dynamic on a developmental continuum, due to the various intentional choices and cumulative efforts involved in loanword integration. We provide a metalanguage for loanword notation to make explicit throughout the study all relevant kinds of relationships between source and target languages during the borrowing process from any language into Chinese. Finally, we show examples for all major loanword categories, including a modified and notated categorization of loanwords in Chinese, based on Shi ([Chinese loanwords] Hanyu wailaici. Commercial Press, Beijing, 2013).

1 Introduction

The Chinese lexicon today counts roughly 20,000 loanwords, many of which have entered the lexicon since the late Qing dynasty. Among these we find roughly 3000 Japanese loanwords (Shi 2019). Huang and Liu (2017) have recorded over 50,000 alphabetic words.

A good part of this development can be exemplified by how the very definition of what counts as a loanword has changed over the years, which is discussed in Sect. 2.4. We argue that the status of a loanword is not binary, but dynamic on a developmental continuum, due to the various intentional choices and cumulative efforts involved in loanword integration. We introduce a new way of conceptualizing this continuum, i.e. how well loanwords are integrated into the standard vocabulary by measuring their degree of ‘Integration’ and by how much a loanword has structurally and orthographically changed compared to its source counterpart (‘Distance’) in Sect. 2.4.2.

We are able to do so due to a notation metalanguage we developed, i.e. a method for making explicit the most important processes during borrowing. We can now elucidate the qualitative differences that are created in the relationship between source and target words.¹

The article is organized in the following way: try to give a general overview over all major types of loanwords in Chinese in a systematic fashion, that is by focusing on three major aspects of the borrowing process: (a) orthographic integration based on the choice of orthographic representation, (b) phonetic integration, including material borrowing as well as updated and replaced transliterations, (c) semantic transliteration, here focusing on the fundamental differences in terms of morphemic function: iso-morphemic transliteration; various complex transliterations; conceptual remakes, as well as sense extensions. Section 2.4 addresses the aforementioned categories with an admittedly short history of the development of Chinese loanword theories. Here we attend to such questions as: Which categories are mostly contested in the literature? Or, is it reasonable that Japanese loanwords form a separate class of their own? Regarding Japanese loanwords, we take the liberty of shortly discussing a claim by Shen (2017) and others that a considerable amount of so-called Japanese loanwords arguably had been created by Western missionaries and dictionary makers in China, before being transferred to Japan during the Meiji period – evidence pointing to a Western origin of said words. We close this paper with an annotated table of loanword categories, based on (Shi 2013).

2 Methods of borrowing

Methods of borrowing refer to the three basic aspects of a word that can be borrowed: its orthographic properties, sound, and word meaning. These properties can all be selectively borrowed, and selectively combined with autochthonous morphemes. In the outline of this research, we have organized various borrowing strategies under different segments, according to the salience of each aspect’s borrowing strategy; however, this simplicity could be misleading. In fact, borrowing is the flip side of the process of word-creation. And word-creation employs the entire arsenal of linguistic strategies—to various degrees and successes—to integrate a loanword into the vocabulary. Borrowing is an active linguistic process not only of receiving a loanword, but of actively transforming it, which often involves a high degree of effort on behalf of the ‘receiving’ language community. This is even more so for Chinese, due to the specific properties of the Chinese script that express form,² meaning and sound separately.³

Chinese is different from Western alphabetic languages in that a Chinese word, once communicated via script, will take on one additional layer that gains and actuates some semantic function—the orthographic layer. Due to its orthographic properties a word always has at least one sense associated with its form that may or may not stand independently from the sense that is actuated by its pronunciation, e.g. ‘photograph, picture’ zhaopian is written in standard form as , but can be humorously written—without changing its pronunciation—as , whereas ( ) means ‘to fool s.o.’, creating the meaning ‘fake photograph’. The whole 2-character combination gains an additional semantic layer due to its potential for exploiting a large number of homonyms in Chinese. In written form, the relationship between a morpheme or a syllable and its representing character is not trivial, but rather a fully self-contained subject matter. The takeaway is that due to its graphic representation, borrowing foreign words into Chinese is more complicated than into alphabetic languages. But it also means that Chinese, by nature of its script, is more sensible to loanwords; hence, more efforts involved in loanword integration.

2.1 Borrowing orthographic properties

The general discussion about loanwords focuses on words. Yet, numerals and scientific units are also some kind of ‘words’, just written in another character set—the set of numerals or specially assigned letters.⁴ Hence, in the following subsections we discuss words with special or mixed orthographies. Section 2.1.1 explores alphabetic words; in 2.1.2 we argue that Arabic numerals also count as (written) loanwords in Chinese, also mentioning a small set of characters coined to suit Western scientific notations. Graphic loans is a confusing term, since only a very small number of graphs really have been borrowed from other languages into Chinese (see 2.1.3). We argue that graphic loans are better subsumed under the term silent borrowing if only involving orthographic and semantic properties. However, those from Japanese, should be labeled Japanese loanwords and dealt with separately, as discussed in the final section.

2.1.1 Alphabetic words

There is an ongoing discussion as to the degree where alphabetic words count as loanwords, since native speakers of Chinese often create those words themselves (see Cook (2018); Ding et al. (2017)). Huang and Liu (2017) argue that the linguistic (syntactic) behavior of Mandarin alphabetic words (MAW) is similar to that of character words, even in light of rules for lexical word formation. In terms of pronunciation, however, alphabetic words are heavily influenced by the accent of the speaker (Ding et al. 2017). The IPA representations given in the tables below therefore serve as a reminder that they are not read with a standard English rules of pronunciation.

(a) Alphabetic words in Chinese starting with a variable letter or number The Table 1 below lists words with a variable letter or number as part of the word. These are a subclass of Mandarin alphabetic words. Their number is much lower than words with a fixed alphabetic letter. Any such word can be represented in the format: $[x_{a} \mu _b] = [\langle \mu _{a}^{pc}\rangle _{\textsc {L}} \langle \mu _{b}^{c}\rangle _{\textsc {G}}]$, where the alphabetic component of the source word is kept in the transliteration $\langle \rangle _{\textsc {L}}$, or $\langle \rangle _{\textsc {S}}$ for any symbol, also preserving the sense and approximate pronunciation $\mu ^{pc}$. Source morpheme $\mu _{b}$ is mapped onto a Modern Standard Chinese morpheme written as a character $\langle \mu _{b}^{c}\rangle _{\textsc {G}}$, agreeing with its source counterpart only in semantic content, not in pronunciation (see Table 11:1).

Table 1 Chinese words starting with a variable letter

(b) Words beginning with a fixed Latin and Greek letter Words in this category contain a fixed alphabetic component as part of the word. Shi (2019) catalogs roughly 1,700 alphabetic words, most being purely alphabetic ones—abbreviations such as ITA, which stands for ‘Information Technology Agreement’, or words with a character component, such as OK bianlidian (OK convenience store). Zhang et al. (2018) show significant differences in the selection and reception of Mandarin alphabetic words, listing such words as byebye rou (bye-bye flabby arms waving goodbye), or band-zai (band- band member), O-zui (O- O-shaped mouth). Huang and Liu (2017) collect 56,833 MAWs combining the Sinica Corpus (Chen et al. 1996) and the Chinese Gigaword Corpus 2.0 (Huang 2009). In general, words beginning with a letter are more common than any other configuration (Huang and Liu 2017) (see Table 11:2).

(c) Words starting with a fixed numberShi (2019) finds 36 words starting with a fixed number. Huang and Liu (2017) do not collect this type of words. Any such word can be approximated as: $[x_{a} x_{b} m_c] =$$[\langle \mu _{a}^{c}\rangle _{\textsc {S}}$$\langle \sigma _{b}^{p}\rangle _{\textsc {L}}$$\langle \mu _{c}^{c}\rangle _{\textsc {G}}]$ , whereas S represents any symbol, including numerals. Numbers are transliterated as morphemes, mapping onto their source counterparts only in terms of conceptual content c, but read in their native pronunciation, written in numeric form. The letters in 3C or 4K are kept as such $\langle \rangle _{\textsc {L}}$. Some numbered words are: san [ ] chanye (3C ) for ‘3C industry’, san [ti] dongzuo (3 ) for ‘3D action’, san [ti] xiaoguo (3D ) for ‘3D effect’, etc. (see Table 11:3).

2.1.2 Numerals, units and scientific notation

Arabic numerals from 1 to 9, as well as 0, can be considered loangraphs in Chinese (Xu and Zhang 2006, p. 206). Although Arabic numerals entered China much earlier during the 13th and 14th century, only one or two centuries after Fibonacci introduced Arabic numerals to Europe, their usage had not been widespread in China until the turn of the 20th century. Early translations of mathematical and scientific works used Arabic numerals, such as Xu Guangqi’s ( ) translation of Euclid’s Elements, but his translation (and those of others) was mainly driven by the wish to promote the spread of Catholicism in China, and met with disinterest or rejection (Tian 2005, p. 57). Arabic numerals were first officially published in a mathematics textbook in 1892, translated by Calvin and Zou Liwen ( ). Shortly before that, in 1885, a book for beginners in Western Calculation was published, only in Shanghainese (Xu and Zhang 2006, p. 75). Although the decimal system was a Chinese invention, the development of wan ( 10,000) as a separate counting unit was probably an influence from translations of early Indian Buddhist scriptures, such as the Avata.saka Sūtra or Flower Garland Sutra (Li and Qian 1998, p. 230), and can maybe count as a loanword itself (see Table 11:4).

Mathematical symbols ($+,-,\times ,\div ,\%,=,x,\pi ,<,\subset ,\in ,\infty ,\int$, etc.), units (mm,  cm,  m,  km,  $m^2,$$m^3,$s,  h, etc.) and scientific notation in general can all be considered loangraphs. Every scientific notation is basically a word with a specific conceptual association and a pronunciation. The borrowing process of a scientific notation is similar to that of numerals: $[x] = [\langle \mu ^{c}\rangle _{\textsc {S}}]$, where pronunciations mostly observe the native tradition, maybe with the exception of $\pi$ and a few others (see again Table 11:4).

In Taiwan, the basic unit for measuring the size of living space is called ping⁵ ( ), originally meant ‘flat wide space’, and is a loanword from Japanese. Characters such as haoke ( , milligram), qianwa ( , kilowatt), or jialun ( , gallon), have all been created⁶ at the beginning of the 20th century in response to Western units of Gram and Watt. These can count as a special subgroup of loanwords, under which entirely new characters have been invented.⁷

2.1.3 Non-Chinese loangraphs

Arguably the best known loangraph in Chinese is the Buddhist symbol wan , also written as . Some controversy exists about the origin of the loangraph Pu (2001)⁸ (see Table 11:5).

More recently, a couple of Japanese characters have gained recognition in the Chinese lexicon, e.g. the character dan ( ), originally meaning ‘dwell’, and mentioned in the Shuowen as a variant of jing (dwell), despite its rare usage. Dan ( ) gained a new life, a new pronunciation (dōng or dòng) and a status of an independent character, differentiated from jing ( ), only after the Japanese used dan for referring to donburi, the name of a kind of Japanese food in a large rice bowl. With the global spread of Japanese cuisine, the character also gained popularity in Chinese communities. Another example is xian ( gland), etc. (see Table 11:6).

The borrowing of non-Chinese graphs as new graphs into the Chinese character set can be expressed as $[\langle {m}\rangle _{g}] = [\langle \mu ^{c}\rangle _{g}]$. During the process, the borrowed graph g is linked with a native Chinese morpheme $\mu$, i.e. its associated concept c. The orthographic form g is copied. Only rarely can it be found that a Japanese character is borrowed and combined with a Chinese ontological suffix, as in (see Table 11:7).

2.1.4 Japanese loanwords

The special borrowing process between Japanese Kanji and Chinese characters has been labeled as ‘graphic loans’ (Masini 1997), or jiexing hanzici ( , cf. Shi (2013)), or xingyici ( , cf. Yang (2007)). In this paper we regard this process as ‘silent borrowing’ (see Table 11:8–10). The total number of Japanese loanwords in the Modern Chinese lexicon can figure between 1500, up to 3000 words (Shi 2019).

Strictly speaking, the term graphic loan is a misnomer. You cannot borrow what you already have in possession. The novelty of this type of borrowing process is the concept associated with the characters (which may be expressed by a novel combination of characters), but not the characters themselves (Wang 2015, p. 180). In case of any orthographic differences between Japanese and Chinese graphs, the Chinese do the exact opposite of borrowing, namely they replace Japanese graphs with their own graphs, that is selecting the Chinese set-equivalent form, i.e. the ‘variant’. This can be represented as: $[\langle {m}\rangle _{g}] = [\langle \mu ^{c}\rangle _{g'}]$, whereas $g'$ refers to the substitution of a character with its set-equivalent variant.⁹ For someone unaware of the substitution, $g'$ is equivalent to G.

What is meant with the terminology ‘graphic loans’, is the theoretical intention to differentiate between a silent borrowing process (borrowing meaning via the characters without the associated sound), and a material borrowing process that deals with sound and meaning.

Now, the importance of silent borrowing plays directly into the question of how to determine the status of Japanese loanwords. Two points are thus brought up for discussion: (a) how much emphasis is placed on sound for defining loanwords?—A steadily closing gap is clear in the academic discussion that recognizes the borrowing of word meaning as equally important as sound; (b) how many of those loanwords labeled ‘Japanese’ are of real Japanese origin?—a question more related to historical facts. Chen (2011); Feng (2016) argue that some Japanese loanwords might have Western origins. Generally speaking, the influence of Western (mostly) missionaries on some words in Chinese is a historical fact. Zhao (2016) shows for example that some mathematical terms like xing ( form, shape), jiao ( angle), xian ( arc) can be traced back to early translations from the Elements of geometry by Xu and Ricci, published in 1607. However, it has long been overlooked that the Western influence since the industrial revolution and the rise of the West is a much greater factor. Schmidt (forthcoming) argues that roughly a quarter of Japanese loanwords in Modern Chinese—as recognized by the current literature—are of Western origin¹⁰. Although oral borrowing and borrowing via informal writing can be fairly assumed to have happened, but the more traceable pathway of this transmission was by early English-Chinese dictionaries upon which Japanese translators heavily relied. Many of these dictionaries included new words (compounds, creations) for new concepts. One of those traceable impacts is Wilhelm Lobscheid’s English-Chinese Dictionary ( ), published between 1866 and 1869, in which dozens of words made their first debut—words he and his Chinese co-lexicographers arguably created themselves (see Table 2).

Table 2 Selected words first attested in the English-Chinese Dictionary by Lobscheid

In Japan, received and home-made loanwords were part of a large-scale social and linguistic modernization, strongly appealing to the Chinese people as well. The starting date for back borrowing from Japanese into Chinese can be set around the year 1900.¹¹ While it took nearly 30 years for the Japanese vocabulary to complete modernization, Chinese achieved its own ‘second wave’ of modernization¹² in roughly half the time.

2.2 Borrowing phonetic properties

Phonetic transliteration is the single uncontested category of loanword borrowing. The borrowing process of phonetic properties is understood in the following way: Ideally, each source morpheme can be mapped onto one or more target syllables in Chinese. That is, in a Chinese phonetic loanword, each syllable has no more and no less than a single phonetic function, referencing back to a source word. The actual process of phonological adjustment during phonetic borrowing may vary, according to various perception-based, phonetic or phonological rules (see Peperkamp et al. 2008; Paradis and Tremblay 2009; Kang 2011).

Prior to more discussion about simple phonetic loans in more detail, please note that Table 11 contains a rather large number of specific subcategories related to phonetic borrowing. Due to limited space, we cannot go into detail for all of these. We regard these types of words as phonetic loans because their most salient aspect of borrowing is based on phonetic representation. This main function can then be combined with other lexical processes. However, to establish a neat classification, we explicitly deny any morpheme-level semantic processes to participate in any borrowing process called ‘phonetic’. In other words, as soon as sub-word semantic processes are combined with any phonetic processes, the entire categorization shifts towards complex transliteration. This being said, the classification does allow the interaction of orthography with phonetics in the following way: (a) complex orthography (karaoke OK (see Table 11:12); also asir ‘sir’ (see Table 11:13); as well as ‘byebye’ $\rightarrow$ 88 (see Table 11:14); and (b) phonetic transliteration partly imitating source orthography, as in [t$^{h}$i]xu T- ‘T-shirt’ (see Table 11:15). The mapping from a source syllable to target syllable during borrowing can be represented as $[s_{a}] = \langle \sigma _{a}^{p}\rangle$, using p if the target syllable approximately resembles the pronunciation of the source syllable.¹³ When Chinese borrows polysyllabic source words, it is almost forced to apply a polysyllabic phonetic representation/transliteration, which motivates the language to adopt a more semantics-orientated transliteration strategy. In fact, phonetic borrowing is often only the first step of the borrowing process, followed by a collective effort from the language community to figure out better substitutes via either orthographic adjustment (i.e. by adding classifiers, prefixes, by shortening), or re-creating the word using some semantics-oriented approach (conceptual remake).

2.2.1 Material borrowing (simple phonetic loans)

Material borrowing refers to the borrowing of sound-meaning pairs, irrespective of orthographic properties (see Table 11:11). Specifically, material borrowing refers to word-level semantic borrowing only. If any morpheme-level semantic processes are involved, the category changes to complex transliteration. In the Western literature the range of material borrowing might include lexemes, or lexeme stems¹⁴ (including also sometimes affixes¹⁵), to even entire phrases (Haspelmath and Tadmor 2009). However, the concept of material borrowing is complex in the Chinese context, because phonological adjustment often results in polysyllabic constructions.¹⁶ If such polysyllabic constructions, however, are supposed to be fully integrated into the native vocabulary, then they require any syllable to have a tone, which effectively turns the respective syllable into a morpheme. Once we start to deal with morphemes, we cross the border into complex transliteration territory. This does not mean that there cannot be material borrowings in Chinese, it means that because monosyllabic morphemes often take part in compounding polysyllabic loanwords, the function of each syllable is more than just syllabic. The morphemic quality of each unit plays an important part in supporting or influencing the entire word sense. Hence, it is often very difficult to say that each unit is solely restricted to a phonetic function, and only the entire phonetic form of the word is then equated with the entire word sense. If so, then this is most likely to happen with proper names. Finally, in the absence of morphology, compounding plays a very dominant role in Chinese, which always involves the semantics at the morpheme-level. Yet, sub-word level compounding processes are generally excluded from loanword theories, a mistake any Chinese loanword theory should carefully avoid.

Simple phonetic loans in Chinese are easily noticeable for a native speaker because of their phonetic resemblance to the foreign source words, and/or, in terms of their written form, due to their apparent lack of a semantic relationship with the chosen characters. In light of a multiplicity of Chinese homonyms, the criteria of unanalyzability—as stressed by Haspelmath (2009)—can be especially useful on the orthographic level, rather than on the sound level.¹⁷ This can be represented as: $[m_{a}m_{b}] = \langle \sigma _{a}^{p}\rangle _{\textsc {G}}\langle \sigma _{b}^{p}\rangle _{\textsc {G}}$.¹⁸

Generally speaking, in any given scenario of linguistic contact between Chinese and a language without script or with a non-Han-related script, it would be reasonable to expect the way of vocabulary transmission to be mainly phonetic. This certainly is true throughout history, although the linguistic footprint for various Mongolian tribes was astonishingly limited, e.g. phonetic loanwords from the Xianbei ( ) during the Late Han dynasty, such as kehan < qaghan ( Supreme leader), or the proto-Mongolian Khitan ( ) at the times of Northern Liao, Eastern Liao and Later Liao, e.g. mouke < mäke ( organization of Khitan families), and the Tungusic-speaking Jurchen ( ) during the Jin dynasty, e.g. the name of the Jurchen itself stemming from nüzhen < jurchin.

During the Mongolian-reigned Yuan dynasty, Chinese received a few phonetic loans such as chengjisihan < čingizhan ( Genghis Khan), gebi < gobi ( desert), hutong < gudum ( alley), laba < labai ( trumpet), or haba (gou) < xaba, halban ( (hairy lion) dog), etc.¹⁹ Despite the 300 years of lexical impact accompanying Manchu rule (Qing dynasty) over China²⁰, only a couple of words were left behind, such as bashi < baksi, bagši ( accomplished technician, master)²¹, saqima < sacima ( Sachima, a common Chinese pastry), mahu < mahuu ( Manchu hood, wrapped around the ears and forehead, but open on the top), etc.

This demonstrates that even phonetic loanword transmission from source to target languages is to some degree facilitated by some sort of cultural influence of the donor over the target culture, whereas a lack thereof results in a low number of loanwords. This explains why phonetic loanwords from Sanskrit and Indian culture are comparatively more common in number and usage, e.g. emituofo < amitābha ( ), luohan < arhat, arhān ( Rohan), fan < brahmā, brahman ( clean, Sanskrit), fo < Buddha ( ), zhina < čina ( China), heshang < khosha ( teacher, monk), yujia < yoga ( ), etc.

Phonetic borrowing has also been the standard approach by Yan Fu and other early Chinese foreign-trained scholars (see Table 11:16). ‘Zebra’ was phonetically translated into zhibula (today semantically banma ‘striped-horse’), or ‘philosophy’ into feiluosufei (today zhexue ‘wise-learning’), or ‘revolution’ as lifoliuxuan (today geming ‘change-heavenly mandate’).²²

A couple of reasons made this approach difficult. Firstly, words of modern science often have multiple sources: English, French, German, Russian, etc. A phonetic representation makes back-tracking from sound to meaning difficult. Secondly, phonetic transliteration often requires the use of multiple syllables to convey a single concept, which is unnatural in Chinese. Moreover, many of those words use difficult characters to stress their phonetic function (within that word) in contrast to a literal semantic reading. This, however, makes the word harder to write and remember. Lastly, using complicated polysyllabic words in Chinese might be acceptable in specific religious texts, e.g. in translations related to Buddhism. However, it would be rather discouraging to use obscure characters and long multisyllabic words²³ for new everyday-concepts such as ‘telephone’ (delüfeng ) or ‘president’ (bolixitiande ) etc.

Early simple phonetic transliterations might be adjusted over time. Indeed, the most common used fall under the categories of (a) orthographic adjustment, not changing pronunciation, e.g. ‘cacao’ keke$\rightarrow$keke , or ‘curry’ gali²⁴$\rightarrow$ , or ‘Italy’ yidali (as used in Taiwan) $\rightarrow$ (as used in Mainland China, see Table 11:17); (b) shortening, with optional addition of ontological suffix, e.g. ‘America, USA’ in meilijian$\rightarrow$mei + ‘country’ ( ); (c) replacement by semantic transliteration, e.g. professor$_{a}$ from its original phonetic loan bufeise ( foretell$_{a_{1}}$ not$_{a_{2}}$ color$_{a_{3}}$) to later re-coined jiaoshou ( teach$_{a_{1}}$ give$_{a_{2}}$). For more examples, see Zeng (2017, p. 130–132).

2.2.2 Mediated material borrowing

Mediated material borrowing refers to a scenario where a target language borrows a sound-meaning pair via an intermediate third-party language. There can be more than one intermediate language (see Table 11:18). For example, ‘Aspirin’ was materially borrowed from German to English, then to Japanese (asupirin), and lastly into Chinese (asipilin). Japanese loanwords, written in Katakana or Hiragana, serve as mediator for materially borrowing into Chinese (see Table 3). If Japanese words are written in Kanji, Chinese speakers would prefer to borrow the orthographic units (characters) over sound, resulting in graph-meaning borrowing, which is labeled herein as silent borrowing, or otherwise as graphic loans (Tranter 2009), and symbolic loans (Cook 2018).

Table 3 Examples of loanwords into Taiwanese Mandarin from English via Japanese by material borrowing

In some cases, the borrowing from Japanese into Modern Standard Chinese is mediated by a dialect of Mandarin, e.g. Southern Min, as this is sometimes the case in Taiwan, for example tenpura²⁵ ( ) was mediated by Southern Min thian$_{35}$ pu$_{55}$ lah$_{3}$ and then adopted further into Chinese tianbula ( ) (see Table 11:19).

2.3 Borrowing semantic properties

Discussing loanwords in Chinese, we often see a differentiation between phonetic transliteration (yinyi ) and semantic transliteration (yiyi ), as in Zhao (2016, p. 3) However, what counts as a semantic transliteration though has not been strictly defined. For example, the words miyue ( honeymoon), bairimeng ( daydream), and dianhua ( telephone) are all considered semantic loanwords of some sort in the current literature, either implicitly as jiexing hanzici ( , cf. Shi (2013))²⁶, or explicitly as yiyi ( , cf. Yang (2007)), duiyi ( , cf. Feng 2004, p. 27), or hezhihanyu ( , cf. Takeyoshi 1996). However, if we take a closer look at the morpheme level, we can see a different type of relationship between the source-morpheme structures and target-morpheme structures for each of the above-mentioned examples. In the case of honey$_{a}$ moon$_{b}$, each morpheme of the source word is represented in the target word , while also preserving its morphosyntactic order. Due to said property, we call this type of transliteration iso-morphemic.²⁷

In the second case, day$_{a}$dream$_{b}$ has two morphemes, but its transliteration bairimeng has three ($\mu _1, \mu _2, b$), while only morpheme b is directly projected onto the target word. The first morpheme day $_{a}$ is split into a ‘two-syllable-two-morpheme’-construction, roughly meaning ‘white $_{\mu {_1}}$ day $_{\mu {_2}}$’.²⁸ Due to this mismatch in the morphemic structure – based either on a different count of total morphemes, a partial morphemic mapping, or the addition of ontological suffixes to the target word, we call this type complex transliteration.

The third common type of transliteration is a conceptual remaking, most commonly observed in Japanese loanwords, as in the case of tele $_{a}$phone$_{b}$, which, if using an iso-morphemic projection, would be transliterated as *yuansheng  $^{*}$ for tele = far$_{a}$ and phone = sound$_{b}$. However, the modern Chinese version is dianhua , which reads as ‘electric$_{\mu {_1}}$ talk$_{\mu {_2}}$’. The modern word is connected to the source word neither by morpheme structure nor by phonetic properties; it is an brand-new word created based on the shared concepts of both words, namely what a telephone is (Table 4).

Table 4 Comparison of transliterations

2.3.1 Iso-morphemic transliterations

An iso-morphemic transliteration is a near-synonymous mapping relationship between morphemes of the source and target words. Morphosyntactic order is often preserved. Since strict synonyms are rare, some degree of sense variation should be allowed, for example, German ‘Blind$_{a}$darm$_{b}$’ (English: ‘blind$_{a}$ gut$_{b}$’, appendix, caecum), was the source for Chinese mangchang ; however there is arguably a slight variation in the scope of the senses Darm$_{b}$ and chang ( ), because the Chinese pendent can also refer to ‘thought’ and ‘feeling’. However, since the core senses of both morphemes are near-synonyms, we can still consider this transliteration to be iso-morphemic. The following list shows some similar examples of loanwords, patterned in an iso-morphemic fashion. This can be written in a general form as: $[m_{a}m_{b}] = [\langle \mu _{a}^{c}\rangle _{\textsc {G}}\,\langle \mu _{b}^{c}\rangle _{\textsc {G}}]$, whereas each morpheme has an external direct conceptual mapping relationship with its source counterpart $\langle x^{c}\rangle$ (see Table 11:20) (Table 5).

Table 5 List of loanwords by Direct Iso-morphemic transliteration

2.3.2 Complex transliterations

Complex transliteration refers to an imperfect or partial mapping relationship between the morpheme structure of a source word and a target word. During borrowing, as long as phonetic and semantic processes are involved at the morpheme-level, or if any division of labor takes place between a syllable and a semantically relevant graph, then we have a case of complex transliteration. While material borrowing only refers to a semantic mapping relationship between source and target word at the word level, in complex transliterations the morphemes of the target language play a much bigger role. They attempt to exploit the vast number of Chinese homonyms, but also interplay with the semantic functions of Chinese characters, often purposefully create a witty second layer of meaning. They do not intend to be exact transliterations. This part of the discussion is hardly touched upon outside of the Chinese context.

Now, the mismatch between the source and target morpheme structures can result from several reasons: (a) A set of functionally separated morphemes used in the same word, as in jiuba ( ‘bar’), here jiu (‘wine, alcoholic’) has a purely semantic function, ba a purely phonetic one. Or as in kache ( ‘orig. car; truck’, see Table 11:21). (b) A partial projection of the morphemes, as in Consul$_{a}$ general$_{b}$ transforming to zonglingshi ( ). Here, only the morpheme maps onto general$_{b}$ while is a conceptual remake of Consul$_{a}$, because both words are not related to each other in light of morpheme structure or phonetic properties. The same is true for motuoche ( ‘motorcycle’), here one part of the source word (‘cycle’) was substituted for a better fitting morpheme (che ‘vehicle’), creating a mismatch on the morpheme level (see Table 11:22). (c) A partial projection with an addition of affixes in the target word, mostly loan-affixes but also autochthonous affixes. For example, ‘assimilation’ is projected into Chinese as similar$_{a}$ change$_{b}$, or tonghua ( ). The core sense of assimilation is represented by the morpheme tong ( ), and the second morpheme hua ( ) is a loan-suffix – a way of translating the English suffix -tion into Chinese. Another example would be moshui ( ) for ink, where shui ‘water, fluid’ is the added suffix (see Table 11:23). In rare cases, a prefix can be used which normally only goes together with native Chinese morphemes, such as a in ashe ( ‘sir’), whereas the a is an honorary prefix (see Table 11:24). (d) A fusion of phonetic and semantic function for all morphemes (fully fused). Here, the complex transliteration can be so successful that it is hard to spot: jiapan < ‘chapan’ from the Turkish chapgan, meaning ‘sewn together’, whereas the pronunciation, the word meaning and the sense of each individual morpheme are all reasonably well mapped to the source word (see Table 11:25). (e) A fusion of phonetic and semantic functions for some units of the target word (partially fused). ‘Idol $_{a}$’ is projected to aidou , where $\mu _1$ and $\mu _2$ both imitate the pronunciation of the source word, but $\mu _1$ also carries the sense of ‘love’, to express the positive feelings one has towards an idol. Shuqiu ( ) for ‘shoot’ would be another good example. Here, $\sigma _{1}$ is only phonetic, but $\mu _{2}$ is phonetic and somehow loosely semantic (see Table 11:27).

An important property of many but not all complex transliterations is the mismatch of the unit (morpheme or syllable) count, generalized as $[m_{a}] = [\langle \mu _{a}\rangle _{\textsc {G}}\, \langle \mu \rangle _{\textsc {G}}]$, whereas the difference of units is at least one, either one more or one less. For example, ‘gene$_{a}$’ has one morpheme, but its Chinese equivalent jiyin ( ) counts two; ‘sauna$_{a}$’ counts three in sanwennuan ( ).

Given that this type of borrowing process has been widely gone unacknowledged in the literature, we think it is useful to discuss it here in slightly more detail:

(a) Complex transliteration with functionally separated morphemes The simplest situation of complex transliteration happens if all morphemes of the target word are functionally separated. During this process, the target language opts for using different morphemes establishing either a phonetic or a semantic relationship to the source word.

‘Ballet’ is transliterated into baleiwu ( bā$^{p}_{a_1}$lěi$^{p}_{a_2}$ dance$^{c}_{b}$), whereas balei has a phonetic function and wu a semantic function; or lace$_{a}$ border$_{b}$ into leisi huabian ( lěi$_{a_1}$ sī$_{a_2}{\!}$ flower$_{\mu }$ border$_{b}$), whereas leisi is phonetic and huabian semantic. In some cases, this works even for abbreviations, such as ‘AIDS’, here transliterated into aizibing (²⁹ ài$^{p}_{a_1}$zī$^{p}_{a_2}$ sickness$^{c}_{b}$), whereas aizi refers to AIDS phonetically, and bing (‘sickness’) semantically.³⁰

(b) Complex transliteration with suffixes During transliteration, Chinese lexical word-building rules allow the addition of either ontological suffixes, such as -che ($\sim$ car), -fang ($\sim$ house), or the addition of more abstract suffixes, most of which are loans themselves from Japanese, such as xing ($\sim$ -tion), or -shi ($\sim$ kind of, style), etc. Here, it is important to note that the sense of those suffixes have undergone some significant semantic shift (cf. Table 8)³¹ (see Table 11:23 and 24) (Table 6).

Table 6 Examples of complex transliteration with classifiers

(c) Complex transliteration with functionally fused morphemes A more complicated situation surfaces, if a single target morpheme has more than one function, most often a fusion of phonetic and semantic functions. The fusion of phonetic and semantic functions happens in two different directions. Firstly, a syllable of the source word is phonetically related to one or more target morphemes $[s_{a}] = [\mu _{a_{1}}^{p}\mu _{a_{2}}^{p}]$. Secondly, a morpheme (or shadow morpheme³²) of the target word is semantically related to other morphemes (or shadow morphemes) of the target word $[\mu _{a}^{c'}\mu _{b}^{c'}]$, creating an additional word-internal layer of meaning, engaging in a conceptual reinterpretation ($c'$) of the source morpheme $[m]^{c}$.³³ Fusion can take place fully, as well as fully based on an abbreviated source word, or partly. An example for partial fusion would be miniqun for ‘mini$_{a}$skirt$_{b}$’ ( ), here the first two syllables are used phonetically, yet the characters used to represent those syllables create a second-layer meaning, roughly something like ‘fan of you’ or ‘follow you’, which obviously is not an intended transliteration of ‘mini’ and also not a conceptual remake, rather a play of meaning, given the constraints that the syllables are supposed to primarily represent the sound ‘mini’.

2.3.3 Conceptual remakes

Conceptual remaking comprises the third subclass of semantic transliterations. These words are most detached from a direct morpheme-to-morpheme transliteration. In fact, conceptual remakes arguably are better described as words constructed in such a way that their morphemes capture the same idea of the source word in a completely new fashion, nevertheless referring to the same idea (see Table 11:30). ‘Baseball’ is a word named for some sort of ball game. The word highlights the importance of ‘base’ and ‘ball’ as two key features of the game. In Chinese, however, the word bangqiu ( ) is used. It highlights bat$_{a}$ and ball$_{b}$ instead.³⁴ Using ‘batball’ instead of ‘baseball’ is a conceptual remake of the same idea of that very ball game. We can compare both approaches more generally in the following way: [base$_{a}$ ball$_{b}] =$ [bat$_{\mu }$ ball$_{b}^{c}]^{c'}$, where source morpheme base$_{a}$ is substituted with morpheme bat$_{\mu }$ in Chinese; ball$_{b}$ remains in the transliteration; the new word denotes the same concept by highlighting one different and one identical aspect of its original. The fact that bat$_{\mu }$ and ball$_{b}^{c}$ create a conceptual variation of the same idea is captured by superscript $c'$, which stands for ‘conceptual remake’. Yet, not all conceptual remakes are loanwords. As pointed out in Cook (2018), the Western literature deals with conceptual remakes often as examples of compounding. She suggests labeling conceptual remakes (in her terminology called ‘holistic calques’) as creations and not as loanwords. Although we agree that modern autochthonous creations are common, it is also true that many of the words that would fit into the category of ‘holistic calques’ are factually borrowed, which can be verified by comparing historic (bilingual) dictionaries of English, Chinese and Japanese. A categorization can not simply solve the problem based on some lexical principle (Table 7).

Table 7 Examples of conceptual remakes

2.3.4 Sense extensions

Sense extension is the semantic process by which a target morpheme acquires an additional sense via borrowing without losing its original sense(s). Such an ‘enhanced’ suffix is then combined with standard morphemes to form new compounds. The enhancement stems from a period of social and linguistic modernization in Japan. Many of these words can already be found in classic Chinese texts, but not with their new meanings. The sense extensions are often semantically natural extensions of their original meanings. Note that many suffixes are highly productive (see Table 11:31).

The following list shows examples of suffixes which have undergone some sort of semantic extension by the Japanese, compared to their Chinese counterparts. Their number far exceeds one hundred. For example, bao ( ) originally meant ‘execute, answer, report’ during the Han dynasty. The word was then borrowed into Japanese and acquired there, much later though, the sense ‘newspaper’. After back-import into Chinese in the early 20th century, this new sense of ‘newspaper’ was added to the original sense ‘answer, report’, effectively creating a loan meaning extension. Today we find compound words such as huibao ( ‘answer’) with the old meaning, next to zhoubao ( ‘weekly paper’) with the new meaning (Table 8).

Table 8 Selected loan suffixes from Japanese

2.4 Development of Chinese loanword theories

Generally speaking, phonetic borrowing, especially its most prototypical cases, such as material borrowing and phonetic transliteration of names, are academically accepted as criteria for loanwords across the board. Complex transliterations with fused or separated phonetic and semantic functions are also relatively undisputed, because phonetic features in the target word are partially reflective of the source language, either being phonetic (loanblend) in nature but adding a conceptual suffix, or due to the selection of graphs that express phonetic and semantic features simultaneously or separately. Calques are more often disputed, because they focus on the semantic representation of the source morphemes and are therefore less or not at all reflective of the phonetic properties of the source word.

The real challenge for any theory of loanwords in Modern Standard Chinese is when it comes to Japanese loanwords, because they predominantly are conceptual remakes and sense extensions. The following discussion focuses on the theoretical implications of identifying Japanese loanwords and grating them loanword status.

First of all, Japanese loanwords are highly integrated into the Chinese lexicon and syntactically not at all obviously different from any other words of Chinese origin. The dispute can be simplified down to two points: (a) the creatorship of the writing system and the characters pose the question of who came first and what matters most: the new words (semantic content, conceptual associations) that have been created in Japan, or the characters (orthographic medium) by which those new words are recorded in writing, which are of Chinese origin; and (b) the peculiar properties of the Chinese characters to carry phonetic and semantic functions, which can be independent of each other activated or read.

In the European tradition, the concept of ‘loanword’ was developed based on languages written in some sort of alphabet. Being alphabetic/phonetic in nature, a word can be judged as loanword by directly observing its phonetic and morphological properties, i.e. whether it is a ‘phonological loan’ (Haugen 1972), also called ‘transfer’ (Weinreich 1953), or whether only the meaning has been borrowed in some way, called ‘loanshift’ (Haugen 1972), further differentiated by Weinreich (1953), whether all morphemes were borrowed, called ‘transliteration loan’, or just a single morpheme, called ‘semantic loan’. Following the Western tradition, Wang (1958) and others (Liu 1984; Lü 1992; Li and Yu 2005), propose a strict loanword definition:

Strict Definition A word is granted loanword status if, and only if, both of the following conditions are true:

• the word is not analyzable

• the word is pronounced resembling the source language.

The same logic is applied by many Chinese scholars of Japanese loanwords (Zhang 1958; Wang 1980; Liu et al. 1984; Zhou 1995). For instance, (Wang 1980) argues that ‘qudi, yindu, shouxu... do not really count as loanwords, because their sound has not been borrowed,’ followed by Zhou (1995) differentiating sense extensions of existing words and new words created by adhering to Chinese word formation rules. Many Western scholars follow some flavor of the strict phonological definition, unwilling to grant Japanese loanwords full loanword status, labeling them ‘graphic loans’ instead (Masini 1997, p. 153; Tranter (2009); Wiebusch and Tadmor 2009, p. 581).

Theoretical advancements were possible by focusing more on borrowing sense than sound. Consequently, the restriction of phonetic borrowing as a necessary condition of granting loanword status, was softened (Zhou 1998; Shi 2000, 2019; Yang 2007; Xue 2007). Shi (2000) acknowledges this, stressing form and authorship over phonetic properties: “Japanese loanwords in the Chinese language need to be given a special status. One can say, these words are positioned between Chinese words and real loanwords. They are Japanese-made words, ... they borrow the word form (i.e. characters), but not the pronunciation, it is just the written form of the word.” And also: “[F]or Japanese-made words, no matter whether created by semantic transliteration or newly constructed ones, the authorship of creation lies in Japan.” Zhou (1998) argues that because Japanese loanwords are written in Chinese characters, this does not reduce their loanword status, but in fact allow them to be borrowed in the first place. Furthermore, orthographically speaking, the argument that any borrowing via Chinese characters does not qualify loanword status was also questioned Xue (2007): “... therefore, similarity in terms of the writing system does not count as an argument for saying Japanese loanwords are not loanwords. The Japanese only adopted Chinese characters to express words from other languages or to create words by themselves based on their own new words.”

A relaxed definition of Japanese loanwords can be stated as below:

Relaxed Definition A word is granted loanword status if the following conditions are true:

• either the sense and/or the pronunciation and/or the word form cíxíng ( ), has been borrowed (at least one aspect of the three),

• the origin of the word can be traced back to a different people and culture

• the loanword is in general use for some time.

Now, because Japanese loanwords have been coined by the Japanese, these words express Japan’s reaction to modernization; and because the combination of Chinese characters was previously not part of the Chinese lexicon, or if so, the whole word did not have the modern meaning; ergo, Japanese loanwords are loanwords.

In reaction to the under-appreciation of graphic loans, Tranter (2009) formulates a so-called Graphic Loan Theory. He argues correctly that graphic loans have long been neglected in loanword theories, especially those influenced by the Western tradition. His criticism is based on the argument that in modern literate societies words are exchanged verbally and in written form. Exchanging words across language borders can happen, again, verbally (by phonological reference), by translation and via the borrowing of graphs. We label this process silent borrowing.

An important point of Tranter’s theory is to demonstrate that graphic loaning even occurs between languages which do not share the same writing system, e.g. between English and Japanese. He discusses the example of US English ‘jitterbug’, which was arguably used by GIs stationed in Japan (Tranter 2009, p. 25), and loaned verbally into Japanese, transliterated as . Normally, we would expect the orthographically visible “tt” also to be transliterated as or (underlines added for clarity), if—as in most other English loanwords into Japanese—the borrowing would have happened by “eye route” (Tranter 2009, p. 25), that is copying via spelling, not verbally. He uses this counter example to demonstrate that in fact the standard way of borrowing from English into Chinese, Japanese and other Asian languages has happened by spelling, that is by graphic loaning. He further argues that graphic loans deserve more attention, since they are common, systematic and regular.

According to Tranter, graphic loaning happens in a three-step process: introducing a new idea in form of writing (stage 0); the language community produces different phonological realizations of the graph, based on the orthographic and phonological rules of the language community (stage 1); filtered by phonological constraints, the phonological form is then transformed into the script system of the target language (stage 2), compare Table 9 below.

Table 9 3 step process of Graphic Loans by Tranter (2009)

Tranter demonstrates this three-step process in different language settings: twice from English to Japanese (Katagana), from Chinese to Japanese (Katagana), from Chinese to Vietnamese, from Japanese to Korean, from Italian to English, and from Japanese to English. Yet, in all those cases, the target language uses a phonological script system for step 2. Unfortunately, he withholds from us the most interesting example of Chinese as target language. In fact, the idea of graphic loaning expressed in Tranter (2009) is still basically a process of phonological borrowing, but with the difference that the phonological information is not obtained directly via verbal actualization but indirectly via spelling. Spelling introduces deviation errors compared to the actual phonological realization. The theory itself, indeed, incorporates a step of phonological adjustment (step 1); and another step that takes this phonological adjustment as input for writing systems (step 2), which in all given examples are again phonological in nature.

A recent compelling approach has been undertaken by Cook (2018). She differentiates between (a) calques, (b) silent borrowing (‘symbolic loans’, e.g. Japanese loanwords), (c) material borrowing (‘transliterations’, i.e. borrowing sound-meaning pairs), (d) orthographic loans (cases of borrowing orthography and sound, but not the pronunciation—‘graphic loans’), (e) a class which we do not recognize as loanwords, which are labeled ‘wholesale loans’, that are foreign unaltered words in Chinese; and (f) so-called ‘hybrids’, which are any words with mixed borrowing properties. Laudable is Cook’s differentiation between ‘partial calques’ (e.g. daoban ‘pirate copy’) by which each morpheme of the source word is mapped onto the target word, and ‘holistic calques’ (e.g. diannao ‘computer’), by which only the word sense is being borrowed, but differently realized in the target language, a process which we label conceptual remakes (Cook 2018, p.8-9 and footnote 8). Unfortunately, she stops shortly before also recognizing that the so-called partial calques hint towards the more general property of Chinese characters as being able to fully or partially represent sound and meaning of native words as well as of borrowed words, with important ramifications for a theory of loanwords in Chinese. This is a topic we have explored here in terms of iso-morphemic and complex transliteration, as well as conceptual remakes.

At this point we want to steer our more historically oriented discussion into the home stretch, aiming at a general classification of loanwords in Chinese. The first purpose of such a classification is to tell loanwords from native words. As Haspelmath (2009) pointed out, for identifying loanwords one needs to be looking for a smoking gun, a similar ‘shape and meaning’ of words from genealogically different languages. If we take Haspelmath literally and substitute graph for shape, Japanese loanwords immediately fall overboard, simply because their shape and meaning are often very similar, yet they are historically proven loanwords. Hence, some extra-linguistic knowledge will be needed for this task.

The three fundamental principles upon which we build the classification are already present in the very structure of the previous section, namely, the capacity of written Chinese to express sound and meaning separately and selectively due to the use of Chinese characters, that is (a) the principle of representing sound via association of graph (symbol) with word pronunciation; (b) the principle of representing meaning via association of graph with word sense, and (3) the principle of representing a second layer of meaning due to the specific properties of Chinese characters.

Now, let us shortly summarize the state of the art. The following few paragraphs hopefully show the reader where most approaches so far have been insufficient. Firstly, the Chinese literature parts loanwords most often simply into phonetic loans (yinyi ) and semantic loans (yiyi ). Semantic loans are often equated with calques, either iso-morphemic or complex transliterations, although this distinction is almost never drawn. In fact, most approaches suffice to look only at the word level to draw conclusions about any processes of borrowing. The morphological level is often ignored.

Secondly, a triadic categorization adds hybrid loans, that is loanwords that carry a native suffix, e.g. -che ($\sim$ car) or -gou ($\sim$ dog). Again, which specific function the suffix and/or the other morphemes fulfill, either towards each other or towards the source word is almost never further investigated.

Scholars often regard Japanese loanwords as a separate class of loanwords altogether, which are labeled, unfortunately, graphic loans in the Western literature, because it was assumed that Japanese Kanji would be part of the elements being borrowed. The necessary adjustments that take place during borrowing point to the fact that the most important element of the copying process is the idea behind the Kanji combination³⁵, not so much the Kanji themselves, which are in any case written in autochthonous Chinese characters, a process we label silent borrowing instead. The Chinese literature calls those either hanzici ( ) as a general term, or jiexing hanzici ( , see Shi (2013)), hezhi hanyu ( , see Takeyoshi (1996)), or huiguici ( ), if those words are regarded to be of Chinese origin, or riben jieci ( ), if those words are regarded as genuine Japanese words not modeled after Western concepts or words.

In terms of orthography, some researchers explicitly mention the addition of a semantic pianpang³⁶ during the process of selecting (or in rare cases, creating) characters during transliteration, such as jin ( ) for metals in tai ( titanium), or qi ( ) for gases as in yang ( oxygen) (Liu et al. 1984, Cen 2015, Chu 2006, Fu 2011, Yang 2007). In terms of phonetic loans, Japanese phonetic loans (e.g. tatami) are sometimes also explicitly mentioned.

The minimal consensus for any classification of loanwords goes arguably to a set containing just phonetic loans and hybrid loans (complex transliteration in various styles), after Wang (Wang 1958, p. 9) famously rejected the idea of accepting calques as loanwords. He holds on to the Western-influenced principle that loanwords, just as words in general, are phonetic-semantic units, leaving out one aspect or the other from the borrowing process consequently would disqualify the word as loanword. Lü (1992) and Shi (2000, 2019) also follow this principle; however, Shi makes a differentiation whether the calque is a direct transliteration of a Western concept, in which case it counts as a semi-loanword only, or whether it was mediated via Japanese and counts as a graphic loan (this is marked by the asterisk * in Table 10 below). One step further, allowing for a more relaxed definition, scholars would also include some sort of semantic loans (Wang 2015; Masini 1997; Feng 2004; Yang 2007; Zhao 2016). However, in these approaches subtypes of semantic loans are often not clearly separated.

The setup in Table 10 below captures most categories used by most studies of Chinese loanwords. Of course, sometimes we come across extremely elaborated categorizations, such as the ones by Wang (1958), Novotná (1967), mentioned by Masini (1997), Shi (2013), or recently by Cook (2018).

Table 10 Simple categorization of loanwords

2.4.1 Detailed classification

The following Table 11 lists in detail all most relevant categories of loanwords in Chinese. It is based on a classification by Shi (2013), but modified³⁷ in two notable ways. We expanded and changed the categories mentioned by Shi, including the category organization, names, examples, explanations, and order. Furthermore, we added our notation to make explicit the relevant processes that underlie each category.³⁸

Table 11 Detailed classification of Chinese loanwords (based on Shi (2013), modified)

2.4.2 A Continuum of loanwords in Standard Modern Chinese

The notation system is useful for elucidating various aspects of the borrowing processes and for showing how and why loanword categories differ. But the notation system does not show us, by how much they differ. Hence, we might ask how to transform the qualitative differences into quantifiable differences to conceptualize a numerical continuum of loanword categories. In fact, the question of how much loanword categories differ from each other can be expressed in two ways: (a) to which degree is a loanword different from a native Chinese word, in terms of orthographic, phonetic and semantic integration, and (b) to which degree is a loanword different from its respective source word, again in terms of orthography, phonology and semantics. The first question is about the degree of linguistic integration and a measurement of how much effort a language community has put into making the loanword look and feel like a standard native word. The second question is about some sort of distance between source and target word, i.e. how much the target words has been reshaped in comparison to its original source word.³⁹

Both dimensions create the conceptual space within which we measure and compare every loanword category, and in fact each loanword. Each dimension (‘Degree of Integration’ and ‘Distance’) is calculated by assigning specific values to four parameters⁴⁰: (a) unit: syllable or morpheme; (b) function of each unit: phonetic, semantic, or both; (c) word length; (d) orthography: Chinese characters or alphabetic letters, also considering if phonetic and semantic functions are outsourced from morpheme to graph. For ‘Integration’, the value for each loanword is measured against a typical Chinese word; for ‘Distance’, it is compared to the values of the respective source word.

To give a slightly simplified example, source word ‘car; (truck)’, is rewritten as $\texttt {<m K-pc>L}$⁴¹, target word kache ‘ ’, written as $\texttt {<s.P[i;b]>G, <m.K[b;t]>G}$⁴², whereas after assigning specific conceptual values to each parameter, we compute two values between 0 and 1 (for the target word only): Degree of Integration = 0.71, Degree of Distance = 0.52.

The higher the degree of integration, the more similar the word to a native Chinese word. The higher the value for distance, the greater the difference between source and target words. For each loanword category, we computed a various amount of words.⁴³ We used k-means for clustering.

Fig. 1

Continuum of loanword categories

If plotted with ‘Distance’ as the x-axis and degree of ‘Integration’ on the y-axis, loanword categories form three clusters: (a) Group 1 in the top left sector: high integration—in many cases not easily recognizable as loanwords. (b) Group 2 in the right sector—‘half-way’ integration (with the exception of category 7), generally far detached from their original sources—the words have been transformed, but not very successfully so. They are all still to some degree ‘loanwordy’, either due to a use of non-Hanzi orthography or by a non standard use of morphemic functions. (c) Group 3 in the bottom sector: these words are not very well integrated (which does not contradict high usage), they still feel and look very ‘loanwordy’, either because they contain alphabetic elements or exhibit a non-standard morphosyntactic functionality. Some of these words are probably going to be substituted by better integrated words over time (Fig. 1).

We think some cautious generalizations about the relationship between these clusters of loanwords can be drawn: (a) loanwords can change their category; e.g. going from phonetic towards a more semantic approach. This has been widely observed. This change would be reflected in an upward move on the continuum; (b) the more effort Chinese makes in transforming a loanword—by selecting native syllables, morphemes and graphs—the more the loanword shifts to the right side of the continuum, increasing the distance from its source word. This right-handed movement does not necessarily lead to integration though; (c) the best integrated loanwords are words that follow Chinese morphosyntactic rules, especially fusing phonetic and semantic functions, either by selecting specific morphemes or by a division of labor between syllable and graph, making use of the vast quantity of homonyms and the capacity of Chinese characters to selectively express sound and meaning.

3 Summary

This paper gives a general overview about loanwords in Modern Standard Chinese. Based on the fundamental functions of written language, we differentiate between borrowing orthographic, phonetic and semantic properties. We argue that alphabetic words, numerals, units of measurement and foreign graphs have been neglected under the ordinary definition of loanwords; hence, we have included them under orthographic borrowing. Due to large numbers of homonyms, characters can express phonetic and semantic values selectively; therefore processes of borrowing into Chinese are not trivial. Borrowing from Japanese into Chinese often triggers an exchange of set-equivalent variant characters. Phonetic borrowings are very common and most recognized, but they also can be updated over time by alternating one or more graphs of a word. In some cases the entire borrowing strategy can switch from phonetic to semantic or complex transliteration.

In terms of semantic borrowing, questions may arise that are unknown in Western languages. A written Chinese word has not only a semantic value (referent) associated with its entire word meaning, but also an additional second layer of meaning that is created due to the literal reading of the semantic function of each character. Each character always has at least one semantic value associated with itself (the pictographic meaning or ‘root meaning’ of the character) which may or may not be independent of the modern word sense in which the character takes part. This fact largely increases the—often playfully exploited—ability to create ambiguity.

We define loanwords by virtue of the three fundamental mapping principles that are relevant during the borrowing process between any source language and Chinese: (a) the principle of representing sound via association of graph (symbol) with word pronunciation; (b) the principle of representing meaning via association of graph with word sense, and (c) the principle of representing a second layer of meaning due to the use of Chinese characters. This investigation also includes, on a morpheme level, observing how the subparts of a Chinese word functionally relate to each other, to the entire word meaning, and to each counterpart of the source word. Based on this thorough assessment, we summarize our findings in a detailed classification of loanwords.

With great detail often comes confusion; however, the devil is in the details. Why, for example, do we regard julebu ( ) as silent borrowing, not a phonetic loan in relation to ‘club’, or even as some flavor of a semantic borrowing process, since all three characters read together may trigger a different meaning? Here extra-linguistic knowledge is needed to decide whether a factual borrowing instance may or may not conflict with linguistic principles. An even more difficult case is the differentiation between phonetic loans (with complex sense, Category 19) and complex transliteration (Categories 27–33). For example, leida ( radar) is considered to be a complex transliteration after a careful weighting of the triadic relationship mentioned earlier. However, it could be argued that neither thunder$_{a}$ nor reach$_{b}$ are in any shape or form semantically associated with the word sense of radar, hence leida could be safely considered phonetic. Yet, we reply to this argument that, given the phonological mapping constraints, thunder$_{a}$ and reach$_{b}$ are not supposed to be directly associated with the word sense ‘system of detecting the presence of objects by radio waves’, yet somehow ‘thunder’ encapsulates a metaphorical sense of ‘wave’ and ‘reach’ probably some sort of ‘detecting’. There surely is room for discussion. A definite answer can only be given, if we are able to trace back the very first attested occurrences in all relevant languages and make a decision based on time and date.

In terms of the theoretical development, most Chinese scholars today acknowledge the importance of Japanese loanwords, which has softened their standpoint on calques and semantic loanwords in general, but there exist exceptions nonetheless. The times that a loanword had to be phonetic—as earlier Western-influenced Chinese scholars argued—are over. In recent years, interesting theoretical developments have been undertaken to harmonize Western loanword theories with Chinese theories. Tranter’s Graphic Loan Theory, Shi’s investigation into Japanese loanwords and Cook’s research, among others, have definitely benefited the field. Future research should take on the problem of compound words in Chinese. Currently, compound words are excluded by principle from any loanword status, influenced by Western theories, but in Chinese, compounding is a strong force in the development of new words, as well as in integrating loanwords and must not be excluded. Rather, compounding can trigger interesting phenomena on the morpheme level, especially when they are created by intent, which happens during active loanword integration. Hence, our senses should be alerted when looking out for selective use of orthographic, phonetic and semantic functions that occur during any mapping between source and target words.

Although Chinese has doubtless the longest written history of the world, traceable impact of neighboring languages—with the exception of Japanese—upon its lexicon is astonishingly limited. Now, Japan on the other hand, due the course of history, was at least 30 years ahead in terms of absorbing and processing Western knowledge, before the Chinese started to massively borrow from the Japanese lexicon. The argument has often been brought forward that Japanese loanwords are not loanwords because they are written in Chinese characters and ‘feel Chinese’; hence, they could just as well have been made in China. By introducing the concept of silent borrowing, opposed to early ‘graphic loaning’, we tried to elucidate all relevant processes during this specific type of borrowing and can conclude that loanwords from Japan count as loanwords. Even more so, we forward the argument that a considerable number of Japanese loanwords are of Western origin, created by Westerners in China before the Meiji Restoration began.

Finally, after a detailed categorization of all major loanword categories and many minute details between them, we sum up this investigation by providing a large picture. We map the loanword categories provided in Table 11 on a conceptual 2-dimensional space, measuring for any category (a) by how much a loanword differs from a standard native Chinese word (‘Degree of Integration’), and (b) by how much a loanword is different to its respective source word (‘Distance’). This measurement is done for various numbers of loanwords in each category. Since different classes have a different count of members, this calculation is better understood as a conceptual assessment rather than an exact numeric evaluation. However, by doing so, we hopefully exemplify not only how loanword categories relate to each other, but also the kind of relationship that initiates a loanword into the Chinese lexicon in general.