Borrowing or code-switching? Single-word English prepositions in Hong Kong Cantonese

1 Introduction: Revisiting the distinction between single-word code-switching and lexical borrowing

This article re-examines the controversial distinction between single-word code-switching and lexical borrowing, deploying a dataset of English prepositions in Hong Kong Cantonese.^[1] Tremendous efforts have been invested into classifying various language contact phenomena; yet, definitions are still under debate, and views diverge as to the precise criteria for differentiating these phenomena. One major issue of contention is whether single words adopted from another language are instances of lexical borrowing or code-switching. A prominent position is that all single words transferred from another language – whether they are established (i.e. the loanwords that appear frequently) or ephemeral (i.e. the nonce forms that are not recurrent) in a speech community – are borrowings (Poplack 2012, 2018, Poplack and Dion 2012, Poplack et al. 2020) insomuch as they undergo morphosyntactic integration with the recipient language. In this view, borrowing can be less frequent (i.e. the nonce forms) or more frequent (i.e. established borrowings), and therefore, frequency is not the (only) criterion with which a single-word transfer is classified as lexical borrowing (which is supposed to be more frequent) or code-switching (which is supposed to be less frequent). Neither is phonological adaptation an indicator of the borrowing/code-switching status of single words from another language (or ‘Lone Other Language Items/LOLI’ in Poplack’s terminology), since variability is high among speakers in the phonetic realization of these elements (Poplack et al. 2020). Multi-word elements (or Multiword Units/MWUs – Treffers-Daller (2023)) from another language that invoke the grammar of the donor language are treated as code-switches (Poplack and Meechan 1995, Poplack 2018), which is less controversial and consistent with other approaches (e.g. Myers-Scotton 1992, 1993), although most recently Treffers-Daller (2023) has argued that these multiword units can be borrowings which are formulaic and listed in the lexicon of the recipient language.

A different paradigm is presented by Myers-Scotton (1992, 1993). In her African data, both phonological and morphosyntactic integration may vary among single words drawn from another language (i.e. ‘LOLIs’), and so frequency and meaning are brought to the fore as the main defining criteria for distinguishing single-word code-switches and lexical borrowing. More specifically, compared with established loanwords which recur and introduce a new concept to the recipient language (i.e. the so-called ‘cultural B forms’ – Myers-Scotton (1993)), single-word code-switches, which often have meaning equivalents in the recipient language, are rare. Nonetheless, for Myers-Scotton (1992, 1993), these singly occurring morphemes may be ‘core B forms’, in other words, borrowings. The major difference between single-word code-switches and lexical borrowing (i.e. ‘core B forms’) is that the former are drawn from the lexicon of the donor language (or ‘Embedded Language/EL’), while the latter (either ‘core B forms’ or ‘cultural B forms’) are drawn from the recipient language (or ‘Matrix Language/ML’ – see Myers-Scotton (1992, 33) and Myers-Scotton (1993, 207)). Nevertheless, as Myers-Scotton (1992, 1993) argues, since both singly-EL morphemes (i.e. code-switches) and the B forms enter a sentence by the same process of lexical insertion, the difference is not directly observable. As a singly-occurring EL morpheme (i.e. a code-switch) comes to be used repeatedly and entrenched in a bilingual community, it may become a ‘core B form’ and eventually a ‘cultural B form’ in the bilingual community; hence, Myers-Scotton’s (1992, 30, 1993, 171) conception of single-word code-switching and lexical borrowing is more of a ‘continuum’ rather than a categorical distinction.

At first glance, the difference between these two approaches lies mainly in naming or terminology. More concretely, what Poplack (2012, 2018) has called ‘nonce forms’ are treated as ‘core B forms’ or single-word code-switches (singly-occurring EL morphemes) by Myers-Scotton (1992, 1993). Despite the different treatments, empirically speaking, these single words from another language (or LOLIs) may undergo morphosyntactic integration with the recipient language (or Matrix Language/ML) on a par with its native words, which is acknowledged in both approaches. The key to understanding the underlying schism between the two approaches – which has often been overlooked in more recent discussions (except Matras (2009)) – is that they are intertwined to the very different syntactic models or constraints which these researchers proposed for code-switching. In Myers-Scotton’s (1992, 1993, 2002, 2006, etc.) Matrix Language Frame (MLF) Model, according to the ‘System Morpheme Principle’, ‘content morphemes’ (largely content words such as nouns and verbs) may be drawn from the Embedded Language (i.e. donor language or EL) but ‘system morphemes’ (largely function words and bound morphemes) are only drawn from the Matrix Language (ML). Accordingly, a single word from the donor language (or Embedded Language/EL) can of course be a code-switch, provided that it is a content morpheme. Morphosyntactic integration is accounted for by the other tenet of the System Morpheme Principle; that is, this word (or ‘root’) may be inflected with affixes or bound morphemes, that is, system morphemes, from the ML. Furthermore, the word is inserted into a slot in a grammatical structure derived from the ML; particularly, its word order follows ML grammar in accordance with ‘The Morpheme Order Principle’. Indeed, these single-word code-switches – content morphemes inserted into an ML-framed sentence – are the core or prototypical cases which the MLF Model seeks to explain, as Myers-Scotton (2002) calls them ‘classic code-switching’. On the other hand, the multi-word elements from another language, which are uncontroversial code-switches for both Myers-Scotton and Poplack, do not fit comfortably into the MLF Model. Specifically, such phrases may show a grammatical structure or word order absent in the ML, thus flouting the ‘Morpheme Order Principle’. To capture this potential contradiction, the MLF Model invokes another principle, that is, the ‘EL Island Principle’ (Myers-Scotton 1993, 2002) according to which a phrase with all words from the EL is formed in the grammar of EL rather than ML.

Crucial in Poplack’s approach is the idea that borrowing is morphosyntactically integrated into the recipient language whereas code-switching is not. The presumption was formalized as the two well-known constraints, namely, the Free Morpheme Constraint (Poplack 1980), which rules against word-internal code-switches, and the Equivalence Constraint, which allows code-switching only where the structure or word order is shared by the two participating languages. Nonce forms which may be morphologically inflected with affixes from the recipient/ML potentially violate the Free Morpheme Constraint, but there would be no violation if these nonce forms are treated as a phenomenon distinct from code-switching. Poplack (2012) recalls that Poplack et al. (1988) is the first attempt to differentiate code-switching and borrowing in a large French-English corpus collected in Ottawa-Hull. The concept of ‘nonce borrowing’ is operationalized in the study as single words from another language (or ‘lone other language items/LOLIs’) which occur only once in the corpus (Poplack et al. 1988, 57). However, there are very few instances of nonce borrowing in the Ottawa-Hull corpus (Poplack et al. 1988, 57), and the study focuses on the variability of borrowings (established and nonce, but mainly established borrowings) in relation to different social variables rather than the necessity of nonce borrowing. As Poplack et al. (1988, 93) recalled, “[i]n French-English bilingualism, however, there are often no morphological or syntactic criteria for determining whether a single lexical item is following English or French rules; noun morphology is most frequently null in both languages, and word order is very similar. Is there any point, then, in attempting to conform to the theoretical distinction between code-switching and borrowing at the single-word level?” In this connection, a full-length exposition of ‘nonce borrowing’ is in fact to be found by Sankoff et al. (1990) who analyse a corpus of Tamil–English data. In the corpus, there are apparently many single-word code-switches with English nouns appearing in pre-verbal object position and inflected with Tamil case markers, as with Tamil nouns in a typical Tamil SOV sentence. These data defy not only the Free Morpheme Constraint but also the Equivalence Constraint (Poplack 1980), which rules out code-switching between verb and object in Tamil, an SOV language, and English, an SVO language. In this context, nonce borrowings, as distinguished from single-word code-switches, seem to be primarily driven by an attempt to reconcile the Tamil-English data with the Free Morpheme Constraint and the Equivalence Constraint, although Sankoff et al. (1990, 96) emphasized “that the goal in this study has not been to rescue the code-switching constraints.” Putatively, Sankoff et al.’s (1990) suggestion that these English elements are nonce borrowing is independently grounded in quantitative analysis; that is, the English nouns, whether established or nonce, are shown to closely parallel the native Tamil nouns in both syntactic distribution and case-marking. ‘The Nonce Borrowing Hypothesis’ is proposed to capture the idea that there is no difference between nonce and established borrowings in their being morphologically and syntactically assimilated with the recipient language (Sankoff et al. 1990, 94). More recently, Poplack and Dion (2012, 311) echoed the claim that nonce borrowing is empirically based but not theoretically driven, reiterating that the necessity of nonce borrowing does not ‘emerge from any theory-internal considerations’. Notwithstanding, Poplack and Meechan (1995) do make explicit reference to the Equivalence Constraint (Poplack 1980) in their analysis of lone words and multi-word fragments from French in Wolof and Fongbe.

It is at this juncture that Stammers and Deuchar (2012) pose a serious challenge to the Nonce Borrowing Hypothesis according to which established borrowings and nonce forms undergo the same kind and rate of morphosyntactic integration. In their Welsh-English corpus, there is a statistical difference between English-origin verbs which are established borrowings (as listed in dictionaries) and those which are not, in other words, the nonce forms. Succinctly, while the former tend to undergo soft-mutation as native Welsh verbs, the latter tend not to, although frequency of a verb (Welsh or English-origin, listed or unlisted) in the corpus has proven to be another determining factor. The distinction between established borrowings and nonce forms is more in line with the alternative approach – the MLF Model (Myers-Scotton 1992, 1993) – in which the unintegrated words (e.g. the English-origin words which do not undergo soft mutation in Stammer and Deuchar’s (2012) corpus) are more appropriately analysed as single-word code-switches but the integrated ones as (established) borrowing. It is, however, important to note that the key point underlying Myers-Scotton’s (1992, 1993) ‘continuum’ is not that lexical borrowing and single-word code-switches are different in their degree of morphosyntactic integration. As Myers-Scotton (1993, 183) states, “[a] major reason to argue against morphological integration as the categorical criterion which will separate CS and B forms is that there are many cases of partial integration for B forms as well as for CS forms.” Morphosyntactic assimilation and frequency can be used as possible criteria, but they cannot differentiate code-switching and borrowing in a clear-cut, unequivocal manner, especially between single-word code-switches and ‘core B forms’ which are similarly rare. What Myers-Scotton (1992, 30, 1993, 176) means by ‘continuum’ is that a single-word code-switch (CS) can become a B form when it recurs and becomes gradually more frequent in bilingual discourse.

In response to Stammers and Deuchar (2012), Poplack (2012, 645) stressed that the Nonce Borrowing Hypothesis does not treat all single words from another language (or ‘lone other language items/LOLIs’) as borrowings. Only those which are morphosyntactically integrated with the recipient language are, but single-word code-switches are supposed to be overwhelmingly rare (also see Poplack 2018, 211, Poplack and Meechan 1998). Poplack (2012, 646–7) further argued that variability in soft mutation of English-origin verbs may be some kind of phonological variation which is susceptible to confounding factors other than frequency, and listedness is not a good indicator of the borrowing/code-switching status of the English items.

In a follow-up study, Deuchar and Stammers (2016) broadened their scope of investigation and examined other types of morphosyntactic integration of the English-origin verbs in their Welsh corpus in addition to soft mutation. While most of these verbs take the Welsh derivational suffix ‘-io’ as the Welsh verbs do, none of them appear in the synthetic construction (i.e. inflected with Welsh tense/agreement suffixes) where most Welsh verbs do. Instead, these English-origin verbs appear in a periphrastic construction with another Welsh dummy auxiliary verb taking up the tense/agreement suffixes, where few Welsh verbs do. Put in Poplack’s framework, according to different yardsticks, the same set of English-origin verbs would be counted as lexical borrowing (by virtue of ‘-io’ inflection) or single-word code-switches (by their appearance in synthetic/periphrastic constructions). Deuchar and Stammers (2016, 14–5) concluded that lexical borrowing may be of various degrees of frequency but is highly integrated with the recipient/ML; on the other hand, single-word code-switches are low in both frequency and integration with the recipient/ML.

Revisiting a dataset of single-word English prepositions in Hong Kong Cantonese, this article adds weight to the idea that morphosyntactic integration cannot be the criterion for distinguishing lexical borrowing and single-word code-switches, countering Poplack (2018). This set of data is intriguing in multiple ways: First, previous studies have focused on corpora where the single-word transfer (i.e. borrowing or code-switching) examined is primarily nouns (e.g. the Ottawa-Hull corpus as in Poplack et al. (1988)) or verbs (e.g. Stammers and Deuchar (2012)), but the behaviour of single-word prepositions from another language has received relatively little attention. Second, contrary to nouns and verbs, prepositions (or adpositions) are more akin to function words or a functional category, but in some languages, the word class may well be a mixed bag. In English, for instance, some prepositions are more contentful (e.g. for in English) than others (e.g. of in English), the former ones being closer to content words whereas the latter ones being more similar to function words. Thirdly, in Cantonese, most adpositions which indicate temporal or spatial relationships are postpositions or ‘localizers’, a term widely used in the literature of Chinese/Cantonese grammar (e.g. Matthews and Yip 2011). Following an NP complement, a localizer is reasonably treated as a postposition. Interestingly, in Cantonese, a localizer is optionally preceded by a nominalizer, and hence, it may be treated as nouns, albeit of a special type (Chan 2018). There may be a few prepositions in Cantonese (e.g. jau4 (from)), but then another predicate is required after the prepositional phrase (i.e. NP PP VP), and hence these ‘prepositions’ are also called ‘co-verbs’ (i.e. they have to be accompanied by another verb – Matthews and Yip 2011). On the other hand, the structure (NP COP PP – e.g. The book is on the table), which is grammatical and quite common in English, is absent in Cantonese (Chan 2018). These typological differences between Cantonese and English enable us to identify whether a mixed sentence involving an English preposition is morphosyntactically adapted to Cantonese or introduces an English structure. More elaboration of these typological differences between Cantonese and English regarding prepositions and related constructions that encode spatial(-temporal) relationships is presented in Section 3.

The primary observation is that these English prepositions are indeed integrated into Cantonese morphology and/or grammar in some constructions where they would be classified as borrowing, but the same prepositions may also introduce distinctive structures from English where they would be counted as code-switching or, better still, structural borrowing (Chan 2018) or constructional borrowing (Chan 2022). Accordingly, it is impossible to judge whether these English prepositions are borrowing or code-switching without considering the actual grammatical constructions in which they are used. Furthermore, the fact that these English prepositions may bring along an English structure to Cantonese defies the MLF Model (Myers-Scotton 1993, etc.). Last but not least, the repetitive occurrences of for as a recurrent form in Hong Kong Cantonese makes it plausible to analyse the preposition as lexical borrowing (see Table 1 and discussion below), although for may bring along English structures, challenging the idea that lexical borrowing must be integrated with the recipient/ML (Deuchar and Stammers 2016).

2 Single-word English prepositions in Hong Kong Cantonese: History and data

Single-word English prepositions were observed in Hong Kong Cantonese in very early studies of ‘code-mixing’, the term that used to be more popular than code-switching for intra-sentential alternation of languages (e.g. Gibbons 1987), and ever since they have appeared in different works on Cantonese-English code-mixing/code-switching (Chan 1992, Leung 2001). Despite their relative infrequency in comparison with single-word English nouns, verbs and adjectives (B. Chan 1992, A. Chan 2023, Gibbons 1987, Lam 2005, Y-B. Leung 1987, T. Leung 2001), the enduring presence of single-word English prepositions and their appearance in different constructions (see more discussion below) suggests that the single-word English prepositions (and related constructions) have been entrenched as a linguistic or communicative resource among Hong Kong speakers.

The data for analysis here are based on naturalistic speech documented in the existing literature on Cantonese-English code-switching that are accessible, including published books (Gibbons 1987) and papers (Chan 1998, Luke 1998) and, to a greater extent, unpublished dissertations (Chan 1992, Lam 2005, Y-B. Leung 1987, T. Leung 2001, etc.). It turns out that there are not a lot of examples in which a single-word English preposition appears in a code-switched Cantonese-English sentence (i.e. altogether 23 tokens). This dataset is then compared to another mini-corpus which was compiled from June 2012 to June 2013 by the diary method with which code-switched sentences with an English preposition were recorded. The source of this mini-corpus is primarily from two Hong Kong radio stations, namely, Radio Television Hong Kong Channel Two (RTHK2) and Commercial Radio Two (CR2), which I tuned to regularly from 11 am to 1 pm and from 11 pm to 1 am during the weekdays within the above-mentioned period of time. The radio programmes all involved conversation among two or more participants who were either hosts or guests, as in conversation the bilingual speakers are presumed to speak more spontaneously and to engage in code-switching more frequently. These data were then double-checked by reviewing the sound files of the programmes on the websites of respective radio stations. A second source is spontaneous conversations in which I was either a participant or an eavesdropper, and the examples were jotted down by pen and transcribed on the computer as quickly as I could afterwards. Admittedly, there are various limitations with this kind of data collection method; for instance, the transcription may not be totally accurate, especially of those examples from spontaneous conversations which cannot be double-checked. Nevertheless, considering the relative infrequency and unpredictability of the constructions, in addition to the observer’s effect (these constructions may well be seen as some kind of language mixing or code-switching which may draw negative attitudes) and the likelihood that some bilingual speakers do not use these constructions at all, the diary method is perhaps more realistic in eliciting naturalistic language data than tape-recording a conversation in which the targeted constructions may not be attested.^[2]

Other instances of English prepositions in Cantonese have not been documented in the literature in more recently collected corpora except Chan (2023). There are two examples involving in and from (Chan 2023, 184–5), respectively, in which the speakers were indicating directions by using these spatial prepositions in a map task. Occasionally, English prepositions can be found in Hong-Kong-based Facebook forums in posts written in the style of spoken vernacular Cantonese, suggesting that the English prepositions are still in use in Hong Kong Cantonese currently.^[3] Due to their sporadicity, these more recent examples are discarded from the current analysis which is based on the older datasets.

In the 2012/2013 mini-corpus, the preposition for is significantly more frequent than other prepositions (Table 1). Not only does the preposition for top the frequency list in the mini-corpus, but also its tokens outnumber those of others significantly. The emergence of for as a recurrent form raises the possibility of treating it as an instance of lexical borrowing rather than code-switching in Hong Kong Cantonese (Section 4).

Holistically, the single-word English prepositions offer various contrasts with other major English word classes in Hong Kong Cantonese, namely, nouns, verbs, and adjectives (Chan 2018). Not only are English prepositions less frequent than English nouns, verbs, and adjectives in various corpora (B. Chan 1992, A. Chan 2023, Gibbons 1987, Y-B. Leung 1987, T. Leung 2001, etc.). They also emerged more recently than the latter categories in Cantonese-English contact. More specifically, before Gibbons (1979, 1987) first documented instances of English prepositions in Hong Kong Cantonese, many established loanwords borrowed from English had been widely used in Cantonese. These borrowed words are clearly phonologically assimilated into Cantonese (Cheung 1972, Luke and Lau 2008), and many of these loanwords even have standard written forms in Chinese characters as used in newspapers and popular books in Hong Kong (Bauer 1996, Chan and Kwok 1982, Wong et al. 2009), as illustrated in Table 2.

In all lists of established loanwords borrowed from English, nouns, verbs, and adjectives are attested; however, prepositions are conspicuously absent (Bauer 2006, Chan and Kwok 1982, Luke and Lau 2008, Wong et al. 2009). In terms of language ecology, there is a sense that the English prepositions emerged in the late 1970s (Gibbons 1979, 1987) when Hong Kong bilingual speakers became linguistically more creative in drawing from English more boldly; not only was there less phonological adaptation with the nouns, verbs and adjectives, the speakers also mixed other English categories such as the prepositions, conjunctions and adverbs, including longer fragments (i.e. phrases and sentences), albeit less frequently (Chan 1992, 1998). If code-switching (or code-mixing) implies more extensive deployment of linguistic resources – including words and constructions – from the donor language, English prepositions apparently lean towards the latter.

Whereas loanwords that are phonologically adapted to Cantonese and ascribed Chinese characters document established borrowing from English in an earlier phase, there may be more recent borrowings from English which are not phonologically integrated and/or assigned Chinese characters. Drawing on large corpora of informal writing mostly in the style of spoken Hong Kong Cantonese, Li et al. (2016) found that most single-word English insertions are printed in English alphabets rather than Chinese characters. Although Li et al. (2016) call these insertions ‘transfer’ (following Clyne (2003)) and avoid differentiating single-word code-switches from lexical borrowing, many of these examples may well be analysed as lexical borrowing on grounds of morphosyntactic integration (with Cantonese) and frequency in the corpus. In particular, Li et al. (2016) attested far more monosyllabic English words (i.e. high, cool, band, keep, send, etc.) than polysyllabic ones, which they explain by the idea of ‘monosyllabic facilitation’. That is, as Cantonese words tend to be monosyllabic, the monosyllabic English words are more easily transferred (or borrowed) to Hong Kong Cantonese with less phonological adaptation. The longer English words would probably be truncated to the preferred word length, a kind of phonological adaptation too, when borrowed into Hong Kong Cantonese (e.g. interview becomes in – also see Luke and Lau 2008). In this light, the English prepositions, most of which being monosyllabic (e.g. for, in, on, at, from, through, of, etc.) are also predisposed to transfer or borrowing on phonological grounds. However, single-word English prepositions are rare in comparison with other English word classes such as nouns, verbs, and adjectives (B. Chan 1992, A. Chan 2023, Gibbons 1987, Y-B. Leung 1987, T. Leung 2001, etc.), and hence, ‘monosyllabic facilitation’ (Li et al. 2016) does not seem to apply to the former. In this light, again, the English prepositions appear to be quite deviant from borrowing and, by default, closer to single-word code-switches. As mentioned above, what makes English prepositions particularly interesting in Cantonese-English contact is that the syntax of adpositions and related constructions are radically different in the two languages in contact, which will be elaborated below.

3 Morphosyntactic behaviour of single-word English prepositions in Hong Kong Cantonese

4 Frequency and meaning

Apart from morphosyntactic integration, Myers-Scotton (1992, 1993) propose that frequency and meaning may be appealed to as criteria for differentiating single-word code-switches and lexical borrowing, and these two criteria are now applied to the single-word English prepositions in Hong Kong Cantonese.

As Table 1 (Section 1) indicates, for became much more recurrent than the other single-word English prepositions, which is non-trivial, as in other contact varieties the preposition has also emerged as a regular feature (e.g. Nigerian Pidgin and Ghanian Pidgin English – Corum 2015) or even the most frequent English preposition among others (e.g. isiXhosa – Simango 2019). By the criterion of frequency, it is plausible to treat for as an instance of lexical borrowing in Hong Kong Cantonese (and in the above-mentioned contact varieties too), whereas others attested in the datasets can be analysed as single-word code-switches. For Myers-Scotton (1992, 1993), the ‘cultural B forms’ are much more frequent as they introduce new concepts into the recipient language; on the other hand, the ‘core B forms’ or the ‘CS forms’ (i.e. code-switches) are less frequent and they often have equivalents in the recipient language. However, for does not introduce new concepts to the language/variety. In many examples involving for, the word or the sentence may be roughly translated into Cantonese. In (9a), for introduces a BENEFICIARY (Pullum and Huddleston 2002, 654) and may be translated by a coverb and a predicate VP (wai6 NP ji4-cit3). In (9b), for marks FUNCTION (Lindstromberg 2010, 229) and may be substituted by a verb jong6-zok3 (used for). In (9c), for implies SUPPORT for the human participant represented by the prepositional object (Lindstromberg 2010, 227, Pullum and Huddleston (2002, 654) describe this meaning as IN FAVOUR OF), but it means CORRESPONDENCE in (9d) (Lindstromberg 2010, 227). In both examples, the meaning of for is typically expressed by verbs in Cantonese (i.e. zi1-ci4 (support) in (9c) and fu1-jing3 (match/correspond to)).

In these senses, for is more like a ‘core B form’, if not code-switches, but it could behave as a ‘cultural B form’ introducing new concepts to Cantonese. There are instances in which for does not seem to have an exact equivalent in Cantonese, as exemplified in (10). In this example, for indicates PURPOSE (Lindstromberg 2010, 229, Pullum and Huddleston 2002, 655) but the purpose (i.e. drinking coffee only) is arguably also the FUNCTION and even one of the TERMS of the subject (i.e. the coupon).

The examples in (9) and (10) suggest that for is frequent because of its polysemy.^[7] After all, logically speaking, it is possible that for is frequent and yet it expresses the same meaning in the data, as in other established English borrowings in Cantonese (e.g. baa1-si2 (bus)). Pullum and Huddleston (2002, 655) comment that “[f]or is perhaps the most polysemous of the prepositions of English.” Therefore, it becomes much more convenient and economical to use the same form (i.e. for) to express a wide range of meanings in various contexts. Whereas polysemy may explain the higher frequency of for in comparison with most other English prepositions, it does not explain the rarity of of which is also polysemous (see Lindstromberg 2010, 205–13; Pullum and Huddleston 2002, 658–9; it does appear in some Hong Kong Cantonese data albeit very rarely – Chan 2015a, see (6a) above). For and of are in fact similar in many ways; they are both non-prototypical English prepositions in that their meanings are largely non-spatial/temporal, and yet both are frequent and polysemous in English. The reason why for is more frequent than of in Hong Kong Cantonese is indeed intriguing. One difference between the two prepositions appears to be their distribution. In the data, for appears in a PP that is a predicative complement (e.g. (3c), (9a)–(9c), (10)), or a postmodifier (e.g. (6b)). In addition, for may behave as a verb (e.g. (7b)) or even a complementizer between a verb and a clause (e.g. (11)).

On the other hand, of only appears in a PP postmodifier (e.g. (6a)). These distributions largely parallel those of the two prepositions in English. Whereas for expresses various meanings in a wider range of constructions, of most often appears in a complex noun phrase (i.e. N of NP) expressing various meanings (Pullum and Huddleston 2002, 659).^[8] In particular, of rarely appears in a PP predicative complement, a position where for is often used in English and most frequently found in the Hong Kong Cantonese data here (e.g. (3c), (9a)–(9c), (10)). In sum, syntactic versatility may well be another factor contributing to the frequency of for in our data.

At any rate, for can be regarded as borrowing due to its frequency. One major issue is that for may introduce English syntactic structure to Hong Kong Cantonese (e.g. (3c), (6b)), which is inconsistent with the popular conception that borrowings are always morphosyntactically integrated with the recipient language (i.e. Cantonese in our case). Another caveat is that borrowings (e.g. dik1-si2 (taxi)) seem seldom as polysemous and syntactically versatile as for in Hong Kong Cantonese.

5 Discussion

6 Conclusions

The differential behaviours of single-word English prepositions in Hong Kong Cantonese make it impossible to pin them down as lexical borrowing (established or nonce) or single-word code-switches solely based on morphosyntactic integration with Cantonese, the recipient language, as espoused in the Nonce Borrowing Hypothesis (Poplack 2018, Poplack and Dion 2012, Sankoff et al. 1990). On the other hand, the fact that these English prepositions may bring along an English grammatical structure to a Cantonese-English code-switched sentence defies the MLF Model (Myers-Scotton 1993, 2002, etc.) in which a code-switched construction follows the grammar of the ML (i.e. Cantonese in our case) rather than that of EL (i.e. English in our case). Whereas the defiance of prepositions may well be due to their being function words, the single-word English prepositions examined in this study (except of) are treated as content morphemes in the MLF Model which partake the same morphosyntactic process or procedure as those of nouns or verbs in being inserted into a code-switched sentence. An alternative picture emerges if we take frequency as the criterion in characterizing these single-word English prepositions as lexical borrowing or single-word code-switches. While most of these prepositions are relatively rare (even though examples of them do appear in different datasets over time), for has become much more repetitive, and hence, it is possible to treat it as an instance of borrowing. One problem of this analysis is that for does take part in code-switched sentences whose structure is more likely to be drawn from English (mostly the NP COP PP structure), contradicting the assumption that lexical borrowing does not alter the grammar of the recipient/ML.

Instead of providing a straightforward answer to whether single-word English prepositions in Hong Kong Cantonese are lexical borrowing or single-word code-switches, this article argues that borrowing and code-switching are overlapping categories. Different criteria may be applied to characterize single-word transfer/replication as more towards borrowing or code-switching (Matras 2009), but there are always indeterminate cases, as illustrated by the single-word English prepositions in Hong Kong Cantonese. The fuzziness is partly because ‘languages’ are not rigidly separate from one other in the mind of bilinguals or multilinguals, a notion that has permeated into conceptions of ‘borrowing’ and ‘code-switching’ (Myers-Scotton 1992, 1993, Poplack 2018, etc.). Freed from a pre-occupation with distinguishing lexical borrowing and single-word code-switching, a general picture emerges in which morphosyntactic integration and overall frequency in single-word transfer/replication correlate with the word class of the transferred/replicated material (i.e. nouns are most integrated and frequent; verbs are less so and prepositions are even less integrated and frequent). While this looks like a restatement of the Borrowing Hierarchy (Thomason and Kaufman 1988), the generalization may well be due to two factors; first, the varying strengths in which different word classes are associated with morphosyntactic patterns in respective languages, and second, different ways in which they are stored and (co-)activated in the mental lexicon (Chan 2015b). Additionally, as indicated in our exploration of different frequencies of for and other English prepositions in Hong Kong Cantonese, there can be variations within the same word-class in different language contact settings (e.g. prepositions in Hong Kong Cantonese).

Abbreviations

1

first-person pronoun

2

second-person pronoun

3

third-person pronoun

ASP

aspect marker

BENEF

beneficiary

CL

classifier

COP

copula verb

COV

coverb

DEM

demonstrative

EXIST

existential marker

FOC

focus marker

LNK

linking particle

LOC

locative marker

MOD

modal verb

N

noun

NEG

negation marker

NOM

nominalizer

P

preposition/postposition

PERM

permissive

PL

plural marker

PP

prepositional phrase/postpositional phrase

PRT

verbal particle

QUAN

quantifier

SFP

sentence-final particle

V

verb