Jump to ContentJump to Main Navigation
BorrowingLoanwords in the Speech Community and in the Grammar$

Shana Poplack

Print publication date: 2018

Print ISBN-13: 9780190256388

Published to Oxford Scholarship Online: November 2017

DOI: 10.1093/oso/9780190256388.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (oxford.universitypressscholarship.com). (c) Copyright Oxford University Press, 2021. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use. date: 15 October 2021

Distinguishing borrowing and code-switching

Distinguishing borrowing and code-switching

Why it matters

Chapter:
(p.141) 9 Distinguishing borrowing and code-switching
Source:
Borrowing
Author(s):

Shana Poplack

Publisher:
Oxford University Press
DOI:10.1093/oso/9780190256388.003.0009

Abstract and Keywords

This chapter confronts the structure of borrowed items explicitly with that of multiword code-switches produced by the same French-English bilinguals. Speakers are shown to imbue switches with the morphosyntactic structure of the donor language while integrating borrowings into that of the recipient language, to the extent of mirroring its variable patterning. Also measured is speakers’ relative propensity to engage in these mixing types, to determine whether those who make copious use of one are equally likely to use the other. No such correlation could be established, further attesting to the distinction among these strategies. Corroborating evidence comes from three additional language pairs and one triplet, in which, regardless of diagnostic or language, lone donor-language items, nonce and more frequent, are seen to behave in parallel in their adoption of recipient-language structure, and differently from multiword code-switches, which retain donor-language structure.

Keywords:   borrowing, code-switching, donor language, recipient language, integration, variable patterns, propensity to code-switch, propensity to borrow, replications, distinguishing mixing strategies

9.1. Introduction

Our focus thus far has been on the process of borrowing. Zeroing in on the quantitatively predominant but highly contentious lone LD items, and in particular those among them that are unattested, or nonce, forms, we sought to ascertain their status. This involved analyzing their behavior on a variety of morphosyntactic diagnostics in a dozen typologically similar and distinct language pairs and systematically comparing it to that of their (unmixed) LR counterparts, as well as to like forms in LD where available. The consistent result, for every diagnostic and language pair, was that in the aggregate, lone LD-origin items patterned with LR, leading to the conclusion that they had been borrowed into that language. Also confirmed were earlier observations that, linguistically, nonce borrowings behaved like attested loanwords.

But the variationist perspective on language mixing, described in chapter 2, entails the further demonstration that along with parallels with LR, the contentious lone items display systematic differences from LD, not only in its unmixed guise but, more tellingly, in code-switches to that language. Accordingly, having now characterized the lone LD-origin items empirically in their own right, in this chapter we confront them explicitly with code-switches.

We noted above that the recognition and identification of nonce borrowings has been the subject of acrimonious debate. Distinguishing them from code-switched elements is often claimed to be impossible. Code-switches composed of a single word (the canonical unit of both established loanwords and nonce borrowings) are of course the most contentious in this regard, but blanket refusals to differentiate all kinds of code-switches, single and multiword, from borrowed forms persist. Some argue that attempts to make such a distinction must fail because the methodology is lacking (e.g., Eastman 1992; Eliasson 1989; Gardner-Chloros 2009; Johanson 1993; Thomason 2001; Winford 2003). The findings of the preceding chapters should lay those concerns to rest; we now have well-established procedures that have been shown to yield replicable results. More insidious is the claim that there is no reason to do so, because the distinction is inherently fuzzy or because the phenomena are instantiations of the same process or represent different points on the same continuum (e.g., Bentahila and Davies 1991; (p.142) Boumans 1998; Boyd 1993; Boztepe 2003; Clyne 2003; Eliasson 1989; Field 2002; Haspelmath 2009; Heath 1989; Myers-Scotton 1993, 2002, 2006; Thomason 2003; Treffers-Daller 2005; Winford 2009). We summarize this position as the Identity Assumption: at any given point in time, (single-word) code-switches cannot be distinguished from (nonce) borrowings. The analyses undertaken in this chapter constitute a severe test of this assumption. Here we examine the treatment the same French-English bilinguals accord the various language-mixing strategies they employ, namely 1) nonce borrowing, 2) lexical retrieval of more frequent lone LD-origin items, and 3) unambiguous code-switching of multiword sequences of LD. Their relative degree of integration (or lack thereof) to LR structure is measured on diagnostics previously selected to reveal which grammar is operating at the moment the LD-origin item is accessed.

At the root of the abiding controversy over the Identity Assumption is the question of how to classify the vast quantities of lone LD-origin items that make up such a disproportionate part of any synchronic corpus of language mixing. We have been at pains in preceding chapters to demonstrate how this can be achieved. Still, those who claim that single-word code-switching cannot be distinguished from (nonce) borrowing seem to take this (in the first instance, methodological) credo as an article of faith. Indeed, because insertional code-switching theories (e.g., the Matrix Language Frame model [Myers-Scotton 1993]) are built on the Identity Assumption, it is in the interests of their proponents to endorse it. Accordingly, rather than further belabor the (largely theory-internal) controversy over whether analysts can distinguish among mixing strategies, in this chapter we focus on how bilingual speakers themselves handle them in spontaneous production, addressing in the process doubts that the categories of code-switching and borrowing have a discrete psychological or social reality (Gardner-Chloros 2009).

9.2. Data

The data for this study come from the subset of Quebec speakers in the Ottawa-Hull French Corpus sample (Table 8.1) who were ascertained to engage, with some regularity, in each of the three major mixing types. Locating such speakers was complicated by the fact that even in a corpus as large as this one, a quarter of the participants rarely, if ever, code-switched (via multiword stretches; see also section 9.4). The goal is to discover, to our knowledge again for the first time, whether the same speakers distinguish LD-origin material when it occurs as nonce forms, more frequent LD items, or code-switches. The three utterances in (1), all produced by speaker 082, illustrate the categories on which the following analyses are based. Under the Identity Assumption, nonce forms (1a) and counterpart items in code-switches (1c) should display parallel linguistic behavior, and this in turn should differ from that of more established borrowings (1b), which we know to be governed by LR grammar.

(p.143)

Distinguishing borrowing and code-switchingWhy it matters

9.3. Results

9.3.1. Lexical constitution of mixing strategies

We begin by considering the lexical constitution of the three mixing strategies.

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.1 Relative proportions of parts of speech in four categories of English word.

Source: This and the tables and figures in section 9.3 are adapted or reproduced with permission from those in Poplack and Dion (2012).

Figure 9.1 charts their distribution across syntactic categories, with content words clustered at the left and function words at the right. Both nonce LD items (the grey bars) and more frequent forms (the white bars) are almost entirely made up of content words. The odd function word that does occur among them is used meta-linguistically, as with the pronouns in (2a) and the article in (2b).

Distinguishing borrowing and code-switchingWhy it matters

(p.144) Words within multiword code-switches to English, however, show no restriction as to part of speech. A full half of them are function words. Pronouns, for example, make up 20% of the words in code-switches, but these never occur in otherwise French discourse. Even content words are distributed differently across the two mixing strategies. While nouns account for up to two-thirds of all lone LD-origin items, consistent with both the well-documented preference for borrowing nouns and their preponderance in monolingual nonce formations more generally (Hohenhaus 2005, 2007; Lipka 1994), they make up only 14% of multiword segments. And this distribution is not an isolated quirk but, as shown in Figure 9.2, has remained stable over time. This is the canonical lexical constitution of mixing strategies. If nonce items originate as code-switches, these discrepancies remain unexplained.

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.2 Distribution across parts of speech of English-origin words over time.

Under the hypothesis that nonce forms are not code-switches but borrowings, however, the disproportionate distributions across parts of speech are explicable by the fact that fragments drawn unaltered from LD, as is the case with code-switches, are governed by the grammar of that language. Consistent with this scenario, the English words contained within them are distributed in the same way as those in unmixed English,1 as represented in Figure 9.1 by the black bar. On the other hand, the distribution of the lone LD items, whether nonce or more frequent, resembles neither that of code-switches nor of unmixed English. Rather, even in the early stages, English-origin nonce items pattern with their more frequent English-origin counterparts, and both are readily distinguishable in this respect from English items contained in longer stretches of English.

(p.145) 9.3.2. Linguistic integration

We now examine the linguistic treatment the same speaker accords each of her language-mixing strategies, revisiting the diagnostics we examined in earlier chapters, and introducing a new one, adjective placement. If nonce borrowings are indistinguishable from code-switches, speakers should treat their own nonce tokens of English no differently on each of these measures from the way they treat the corresponding English words contained in their multiword fragments of English. On the other hand, if borrowing lone LD items is essentially different from code-switching multiword LD fragments, the former should behave just like the established loanwords contained within the category of more frequent items.

9.3.2.1. Verb inflection

Figure 9.3 compares verbal morphology across the three categories of LD-origin verb: nonce, more frequent, and code-switched. The comparison lends no support to the Identity Assumption. Instead, where discernable, speakers make both nonce and more frequent lone LD verbs conform to the French morphological patterns detailed in chapter 8. By contrast, within multiword fragments of English, only English grammar is operative. On the measure of verb morphology then, speakers treat their lone LD incorporations, nonce and more frequent, differently from their multiword code-switches.

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.3 Language of affix on verbs in three language-mixing types (N = 721).

9.3.2.2. Plural marking

We showed in chapter 8 that semantically plural lone nonce English-origin nouns virtually always feature the French null plural affix. How do they compare with those implicated in the other mixing categories? Figure 9.4 again shows remarkable parallels between both sets of lone LD nouns, nonce and more frequent. Almost all of them are rendered with the phonetically null French plural marker, while those occurring within code-switches tend rather to feature the appropriate overt English marker. (p.146)

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.4 Plural marks on plural nouns in three language-mixing types (N = 173).

In each case there are a few outliers, however. For the nonce forms, we already saw in chapter 8 that most involved false plurals of the type illustrated in (9) in that chapter. The same is true of the more frequent lone LD items in (3) below.

Distinguishing borrowing and code-switchingWhy it matters

There is also a small component of phonetic variability, illustrated with the lone LD-origin item shorts in (3c) and (4).2 This same variability also accounts for the 75% rate of overt plural marking in these speakers’ multiword code-switches to English (5), where (in an ideal prescriptive world) we would expect 100%. We return to the issue of phonetic variability in chapter 10.

Distinguishing borrowing and code-switchingWhy it matters

(p.147) When calculations of marking rates are restricted to cases where [s]‌ actually fulfills a grammatical function, we find that for both sets of lone LD-origin items, nonce and more frequent, integration into French is nearly categorical (Figure 9.5).

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.5 Plural marks on plural nouns in three language-mixing types (false plurals excluded).

Only in two cases, the nonce form we saw in (10) in chapter 8 and the more frequent lone LD item in (6), did an overt [s]‌ on a lone English-origin noun function as a plural marker, as do their counterparts in multiword code-switches to English. Barring the possibility that the observed variability between [s] and [Ø] is phonetically rather than grammatically motivated (as will in fact be suggested by the analyses of chapter 10), we would conclude, based on the criterion of retention of LD grammar, that these two tokens are (single-word) code-switches at the equivalence site between determiner and noun. By the same criterion, the remaining lone LD-origin items are not.

Distinguishing borrowing and code-switchingWhy it matters

9.3.2.3. Determiner realization

Because, as noted in chapter 8, English admits many more null determiners than French, the Identity Assumption would require that this disproportion be evident in the treatment of nonce items versus code-switches as well.

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.6 Determiner usage on nouns in three language-mixing types (N = 3,676).

Figure 9.6 shows that nearly all of the more frequent lone LD-origin items do feature an overt determiner, as expected if they were being treated as French. The moderately lower rate of overt determiners on nonce nouns would appear to lend some support to the Identity Assumption, were it not for the fact that, as in the diachronic analysis presented in the last chapter, these all occurred in contexts where determiner absence is expected, if not required, in French. The same is true of their more frequent counterparts, as illustrated in (7). Indeed, all of the 183 lone English-origin nominals with a null determiner occurred in one of these legal contexts.

(p.148)

Distinguishing borrowing and code-switchingWhy it matters

Once the structural facts are taken into account (Figure 9.7), it is plain that all lone LD-origin items, whether nonce or more frequent, are made to conform to LR grammar, with no distinction between them, while nouns in multiword code-switches are treated as English.

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.7 % determiner usage consistent with patterns of French and English.

9.3.2.4. Adjective placement

In English, attributive adjectives precede their head, whereas in French they canonically follow. As with null and overt determiners, however, the conflict is only partial, since the English preposed order is also possible in French, albeit (p.149) under heavy lexical restrictions. The Identity Assumption predicts that nouns contained in multiword code-switches to English should feature only adjectives consistent with English word order (preposed), and the same should be true of English-origin nouns occurring as nonce items (Figure 9.8).

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.8 Placement of adjectives with respect to nouns in three language-mixing types (N = 300).

Comparison of adjective placement patterns in all three sets of English words shows that nouns in multiword code-switches are in fact always preceded by their modifying adjectives, as illustrated in (8).

Distinguishing borrowing and code-switchingWhy it matters

But so are over half the nonce forms and, perhaps more surprisingly, the large majority (87%) of the more frequent lone items as well. Closer inspection reveals that here again, every one of the apparently anomalous preposed tokens involves the small class of adjectives admitting (if not requiring) preposing in French. Adjectives belonging to the preposable class, illustrated in (9), are in fact always preposed; the others are always postposed (10).

Distinguishing borrowing and code-switchingWhy it matters

(p.150)

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.9 % adjective placement consistent with patterns of French and English.

On this diagnostic as well, as shown in Figure 9.9, speakers’ treatment of lone LD items, whether nonce or more frequent, is dictated by the grammar of LR and categorically differs from that of LD words contained within code-switches.

9.3.2.5. Consistency in gender assignment

The final conflict site to be analyzed in this connection involves grammatical gender. English, of course, supplies no indications as to noun gender, nor do English nouns contained in code-switches to English show any. Thus, if the Identity Assumption were correct, consistent gender assignments would only be expected in the category of more frequent lone LD-origin items—if at all.

We could not test this assumption by comparing nonce tokens to more frequent lone LD-origin items, because consistency can only be assessed on the basis of two or more occurrences, and nonce forms are by definition non-recurrent. We therefore compared gender assignments in more frequent lone LD words to those of idiosyncratic LD words, defined in chapter 4 as lone LD-origin items uttered at least twice, but only by a single speaker.

Table 9.1 Consistency in gender assignment to idiosyncratic English items

Consistency in gender assignment

% types

N types

100%

96

47

80%

2

1

50%

2

1

Total

100

49

Table 9.1 charts consistency levels among idiosyncratic LD lexical types with overt indications of gender. Nearly all (96%) show categorical consistency. Another type (diamond drill) displays a lesser but still extremely high consistency rate of 80%, as illustrated in (11). At 50% consistency, barbershop (12), the sole type featuring apparently robust variability, has only two mentions. In short, even idiosyncratic LD nouns, from the time of their second mention, are consistently assigned the same gender.

(p.151)

Distinguishing borrowing and code-switchingWhy it matters

Their polar opposites, the 43 widespread types (defined in chapter 4 as 10+ tokens, 10+ speakers) display a comparatively lesser, though still very high, consistency rate: 81% are categorically marked for the same gender (Table 9.2). All but two of the others show close to 90% consistency. Only one widespread type (baseball) is truly variable—half of its 22 tokens are assigned feminine gender, and the other half masculine, with no discernable change in meaning.

Table 9.2 Consistency in gender assignment to widespread English itemsa

Consistency in gender assignment

% types

N types

100%

81

35

97%

2

1

94%

2

1

91%

2

1

90%

2

1

89%

5

2

78%

2

1

50%

2

1

Total

98a

43

(a) Due to rounding.

The slight shortfall displayed by the more frequent lone LD-origin items does not impeach the level of consistency achieved by idiosyncratic nonce types. If anything, the result requiring explanation is this extraordinary level of gender agreement altogether, given the robust precedent for gender variability in unmixed French discussed in chapters 4 and 8. The answer no doubt resides in the nature (p.152) of nonce or near-nonce formations. It is simply easier to achieve consistency (on anything) between two tokens than among 217 (shop), and within a single speaker as opposed to many (although cf. speaker 095 in example (12)).

Taken together, these findings confirm that gender is all but fixed at the (near-) nonce stage. As the grammar of LD (which gives rise to code-switches) provides no instruction in this regard, speakers can only be drawing on the grammar of LR to achieve this result. Here again, then, nonce borrowings are clearly distinguishable from code-switches.

9.3.3. Speaker propensity to code-switch and nonce borrow

Whatever difficulties analysts may experience in teasing (nonce) borrowings apart from code-switches, these are manifestly not shared by bilingual speakers. On every linguistic measure examined here, individuals consistently distinguished their borrowings (nonce and more frequent) from their code-switches. This should allay any lingering doubts as to the psychological reality of these language-mixing types. In this section we show that the strategies are also differentiated extra-linguistically, in terms of which speakers code-switch and which nonce borrow. If (single-word) nonce borrowing (1a) can be equated with multiword code-switching (1c), speakers who make copious use of one should be equally likely to use the other (e.g., Field 2002). Figure 9.10 plots the relative propensities to code-switch (calculated for each individual from raw numbers of multiword switches per recording) and to nonce borrow (based on the proportion lone nonce LD items (p.153) represent of each speaker’s borrowed types), distinguishing among occasional, moderate, or copious use of each.

Distinguishing borrowing and code-switchingWhy it matters

Figure 9.10 Speaker propensity to produce lone English-origin nonce items relative to code-switches.a

a Given the amount of available data and the number of subdivisions required for this analysis, the category of nonce items was expanded to include idiosyncratic types (produced more than once but by no more than a single speaker). Further restrictions would have resulted in insufficient data. Propensity levels were defined as follows: to code-switch: occasional = 0–3 (28 speakers, 25% of sample); moderate = 4–17 (58 speakers, 50% of sample); copious = 18–132 (30 speakers, 26% of sample). To nonce-borrow: occasional = 0–6% (34 speakers, 28% of sample), moderate = 7–13% (56 speakers, 47% of sample), copious = 14–39% (30 speakers, 25% of sample).

This shows that the two strategies are clearly not correlated. We do observe an increase in copious nonce use (the black bar) as we progress from occasional to copious switchers. But the same correlation obtains for occasional nonce users (the white bar), where the inverse would be expected. Thus copious code-switchers are as likely to be occasional as copious (nonce) borrowers, while occasional switchers are least likely to be occasional borrowers—or copious ones, for that matter. In other words, there is no difference between sparse and copious nonce borrowers in terms of their propensity to code-switch. Given that both strategies require active recourse to the other language (and therefore a certain degree of bilingual proficiency), this finding is inexplicable under the assumption that code-switching is equivalent to (nonce) borrowing. The results confirm, on the contrary, that these are alternative, not correlated strategies.

9.4. Corroborating evidence

Replications of this work using the same accountable methodology confirms the basic distinction between code-switching and borrowing, based on other diagnostics and language pairs. We noted earlier that in a number of datasets quantitatively studied (e.g., Budzhak-Jones and Poplack 1997; Ghafar Samar and Meechan 1998; Moinzadeh 1999; Yoshizumi and Poplack 2012), few, if any, multiword code-switches occurred. But in datasets that do contain them, LD words in multiword code-switches could be seen to differ consistently from lone LD words, virtually always patterning with unmixed LD instead. By way of illustration, in this section we review a few of these.

Consider first Table 9.3, which reprises the Igbo-English nominal mixing results (Eze 1997, 1998) discussed in chapter 6, now contextualized with respect to unambiguous code-switches. On all four diagnostics (involving differential preferences in [unmixed] LR and LD for null determiners, generic reference, modified nouns, and position of modifiers), English nouns in code-switches to English differ from lone English-origin nouns. The former are systematically aligned with unmixed English, while the latter (as we have seen throughout) pattern with Igbo.

Table 9.3 Comparison of four diagnostics involving nouns in Igbo-English discourse (adapted from Tables 6, 7, 8, and 10 of Eze [1998])a

Distinguishing borrowing and code-switchingWhy it matters

(a) Here and in subsequent tables, shading indicates parallels.

Another example comes from Mustafawi’s (2002) five-way comparison of LR- and LD-origin nouns in Gulf Arabic-English bilingual discourse according to the conflict sites of word order (chapter 7), determination, and gender marking. Table 9.4 shows that here too the behavior of nouns internal to code-switches differs from that of lone English-origin nouns, which always align with established loanwords and unmixed Arabic.

Table 9.4 Comparison of word order patterns involving nouns in Gulf Arabic-English discourse (adapted from Tables 2–4 of Mustafawi [2002])a

Distinguishing borrowing and code-switchingWhy it matters

(a) Mustafawi (2002) offered no reason for the discrepancy between unmixed English and code-switches to English specifically in the indefinite context. There may simply be fewer code-switches here, which may in turn derive from the fact that Arabic features no overt indefinite determiner, reducing the number of equivalence sites for code-switching (see chapter 6, section 6.3.1 for the opposite situation). Examination of the variable conditioning of the expression of indefinite reference in LR, of the type illustrated in section 6.4, may also help elucidate this issue.

Adalar and Tagliamonte’s (1998) study of Turkish-English mixing extends the comparison beyond the usual benchmark categories to include not only lone (p.154) (p.155) English-origin nouns in Turkish but also lone Turkish-origin nouns in English. Illustrating with the syntactic diagnostic of word order in the NP, Table 9.5 again shows that in each case lone LD-origin items in LR differ from LD items in longer stretches.

Table 9.5 Comparison of word order in Turkish-English discourse (adapted from Tables 6a, 6b, 8a, and 8b of Adalar and Tagliamonte [1998])a

Distinguishing borrowing and code-switchingWhy it matters

(a) Totals may exceed 100% because tokens consistent with the syntax of both languages count in both categories.

An even more complex example comes from Huynh’s (2009) study of the mixing strategies of Vietnamese-French-English trilinguals. Focusing on the behavior of both lone English- and French-origin nouns incorporated into Vietnamese, she situated each with respect to like items produced in multiword switches to each of those languages as well as to unmixed Vietnamese. Table 9.6 illustrates with three conflicts involving the relative positioning of demonstratives, adjectives, and possessors vis-à-vis the noun. Lone LD nouns incorporated into Vietnamese, whether of English or French origin, show the by now familiar parallels in nominal modification patterns with LR counterparts. They simultaneously differ—at times categorically!—from nouns in code-switches to each of the donor languages. These examples can be multiplied. (p.156)

Table 9.6 Comparison of the relative positioning of modifiers in Vietnamese-French-English discourse (adapted from Huynh [2009])

Distinguishing borrowing and code-switchingWhy it matters

9.5. Discussion

In this chapter we considered the claim that at any synchronic moment in time, single-word code-switches cannot be distinguished from nonce borrowings. The hypotheses, summarized in chapter 5, were that if LR grammar could be shown to be operating, the lone LD item could be inferred to have been borrowed, whereas if LD grammar were activated, it would have been switched. These hypotheses were tested by appealing to diagnostics straightforwardly associated with one or the other of the languages in contact. Our results are damaging to the Identity Assumption.

Particularly novel here is the evidence we have presented that French-English bilinguals who employ all three language-mixing types systematically differentiate their borrowings from their code-switches. Operationally defining nonce forms as unattested lone LD items uttered once by exactly one speaker, more frequent lone LD items as those occurring twice or more, and (unambiguous) code-switches as multiword fragments of LD, we examined how the same speakers treat each of these categories of LD-origin item. In all areas but their overall frequency and level of diffusion, not only do lone LD-origin items, both nonce and more frequent, parallel each other’s behavior, but they always differ from that of multiword code-switches: in the parts of speech that constitute them, in speaker propensity to use them, and, as shown repeatedly in preceding chapters, in their linguistic behavior, where discernable. With only two exceptions among over 7,000 tokens examined in this connection, the behavior of lone LD items is always consistent with LR grammar. And these distinctions receive robust corroboration at the community level, on the basis of different language pairs and different diagnostics. Indeed, the rate parallels across lone items and respective LRs on the one hand and code-switched items and LDs on the other are simply remarkable, considering that each category may contain elements that independently favor or disfavor the linguistic phenomenon being measured (as illustrated in chapter 7).

Granted, a more compelling comparison would have confronted lone nonce borrowings with single-word code-switches. But applying objective measures to detect them in lieu of theory-internal criteria, here we were unable to locate more than two (potential) examples of the latter! The comparison with multiword code-switches is rather intended as a heuristic in response to claims that they are the same, or are at least located on the same continuum. If either of these claims were correct, the lone LD-origin items would have displayed at least some commonalities with their counterparts contained in multiword code-switches. Regardless of diagnostic analyzed, they do not. This is because the vast majority of lone LD items are borrowed.3

(p.157) We conclude that beyond their etymological origin, nonce borrowings do not share the linguistic properties of code-switches, single-word or longer. Their only common characteristics are extra-linguistic, involving levels of recurrence and diffusion. As the analyses presented throughout this volume converge in showing, their linguistic structure is rather that of established loanwords, which in turn mirror those of the language into which they are incorporated. These characteristics are summarized in Table 9.7.

Table 9.7 Characteristics of three language-mixing types (adapted from Table 9 of Poplack and Dion [2012])

Distinguishing borrowing and code-switchingWhy it matters

It follows that the LD-origin nonce words are not code-switches, despite the fact that both require knowledge of LD to produce. They are borrowings, albeit ones that are (still) devoid of the social characteristics of established loanwords. As such, they manifestly must be distinguished from code-switches prior to attempting to build a theory of either. The foregoing chapters have demonstrated how this can be accomplished.

Notes:

(1.) The unmixed English data on which these calculations are based consist of 1,400-word stretches of speech provided by each of 14 speakers subsampled from the Quebec English Corpus (Poplack, Walker, and Malcolmson 2006).

(2.) Note that these alternate realizations of [s]‌ can only be phonetically motivated, since there is no singular form with these meanings.

(3.) It is worth repeating here, as we have observed elsewhere, that this is an empirical finding, not a theoretical prerequisite. There is no constraint against single-word code-switching. The method proposed here is fully capable of detecting them (e.g., Meechan and Poplack 1995; Poplack and Meechan 1995; Turpin 1998); nonetheless, they are only attested sporadically.