Lone English-origin nouns in Spanish: The precedence of community norms

Participants and corpus

The data for this study are drawn from the New Mexico Spanish-English Bilingual corpus (NMSEB) (Torres Cacoullos & Travis, in preparation), constituted by approximately 340,000 words and 29 hours of spontaneous speech. The participants are speakers of a Spanish that has been spoken in New Mexico for centuries, and not descendants of recent immigrants. The Northern New Mexican communities represented in this corpus have been in contact with English for 150 years, and all of the speakers acquired both Spanish and English during childhood (Bills & Vigil, 2008). This diachronic depth affords us a wealth of information regarding the long-term effects of code-mixing within a community. NMSEB offers a particularly unique perspective on US Spanish phenomena, as it comprises spontaneous use of English and Spanish by a community of bilingual speakers who regularly use both languages, including during the same conversation (for more details on the corpus, see Torres Cacoullos & Travis, 2015; Travis & Torres Cacoullos, 2013). For this study, a sub-corpus of 18 interviews was examined, or approximately 180,000 words encompassing 17.5 hours of recorded speech from 20 speakers.

Single English-origin nouns occurred in 15 of these 18 interviews, and in the speech of 16 of the 20 speakers. (Interviews 4, 10 and 12 contained none, though these interviews did contain multi-word code-switches to English; see Wilson & Dumont, 2015. Interviewer data were not included.) ² Among the speakers who did use single English-origin items, rates of use were highly variable, with normalized frequencies ranging from 6 to 215 tokens per 10,000 words (see Appendix).

Data: Lone English-origin nouns

The data include all English-origin nouns and compound nouns surrounded by Spanish discourse by the same speaker (N=1,112 nouns, from 2,449 English-origin words, including adjectives, verbs, and discourse markers), irrespective of Intonation Unit breaks. From this initial extraction, proper nouns, including names and places (N=629), as in (4), and nouns that acted as a title, and which were thereby connected to a proper noun (N=32), as in (5), were excluded from the study because proper nouns may not be subject to the same processes of integration as common nouns (Poplack, Sankoff, & Miller, 1988, p. 99, n. 8).

(4)

(5)

Also excluded were examples in which the noun was used in a metalinguistic context in which the noun itself was the topic of conversation (N=18), as in (6).

(6)

Numbers, when treated as nominals as in (7), were included in this study; however, adjectival uses, as in (8), were excluded. Finally, items in which the noun of interest was truncated (N=12) were excluded.

(7)

Beyond these exclusions, the English-origin lone noun dataset was further refined to ensure, first, that the noun in question would not generally be considered part of a monolingual Spanish speaker’s lexicon, and second, that the compounds were conventionalized compounds among monolingual American English speakers (and therefore could be treated as single words). To satisfy the first criterion, the English-origin nouns that were listed in the most recent edition of the Diccionario de la Real Academia (DRAE, http://www.rae.es/rae.html) were excluded from these data (99 lexical types, N=102; see Appendix). Retained were cases where the definition provided in the DRAE was entirely different from the semantics found in the data (N=29). These uses were not judged to be equivalent to the DRAE listings, and included nouns such as brecas ‘brakes,’ carton (as a container), chanza ‘chance,’ craques ‘cracks,’ cricket (the insect), home (rest home), light (as in rays of light), machine, post, robe, semi (truck), trust, and yarda ‘yard’ (as in a piece of land next to a house).

To satisfy the second criterion, that the compounds were conventionalized compounds among monolingual American English speakers (and therefore could be treated as single items), I only included compounds found in Merriam-Webster’s English Dictionary (http://www.merriam-webster.com/), like alarm clock, chain saw, day care, duct tape, ghost town, mental illness, and monkey bars (see Appendix for excluded compounds). This left 820 lone English-origin nouns.

Though adjectives in general were excluded from this study, a handful (N=11) of lone adjectives were found with articles, and therefore functioning as nouns, as in (9). ³ This structure is generally acquired between the second and third year among monolingual Spanish speakers, which may be taken to indicate a certain level of syntactic complexity (Snyder, Senghas, & Inman, 2001, p. 162). Its presence in these data—both with Spanish-origin and English-origin nouns—speaks to the extent to which English-origin words have been incorporated into everyday language.

(9)

Baseline comparisons

There are several measures to determine whether an other-language-origin item is a borrowing or a code-switch, the former of which involves recourse to only one grammar, and the latter of which involves the juxtaposition of two grammars. Items that are frequent and that occur among many speakers (i.e. are diffuse) are expected to function as borrowings, but the primary criterion by which borrowing status is determined is on the basis of morphosyntactic integration: an established borrowing like troca ‘truck’ should pattern no differently from a Spanish-origin noun like camioneta ‘truck.’ Indeed, this morphosyntactic criterion applies regardless of whether the item is established; that is, any item that shows clear signs of morphosyntactic integration can be seen as a borrowing.

However, with languages as typologically similar as English and Spanish, morphosyntactic integration is often not easy to identify. This is particularly true because in many instances Spanish and English grammars show variability, such that any contrast is probabilistic, not absolute. To determine the grammatical status of lone other-language-origin nouns in code-mixed discourse on a larger scale, it is necessary to identify a conflict site, i.e. “a form or class of forms which differs functionally, structurally, and/or quantitatively across comparison varieties” (Poplack & Meechan, 1998, p. 132; Poplack & Tagliamonte, 2001, p. 101). In other words, the distributional patterns of contentious items (i.e. those that are neither frequent nor diffuse) must be compared with those of other items whose grammatical status is clear. Nouns in monolingual stretches of speech (e.g. Spanish-origin nouns surrounded by Spanish) are unambiguous. Moreover, established borrowings, as determined by diffusion within the speech community, also demonstrate morphosyntactic integration.

To identify conflict sites and examine distributional patterns, comparison datasets must be prepared. In this case, the NMSEB offers us the unique opportunity to explore the patterns found within the two languages of the speech community itself, instead of resorting to an idealized monolingual norm or some other regional variety. The use of Spanish and English from the same speakers allows us to avoid the “comparative fallacy” (Bley-Vroman, 1983), since we are assuming that the most appropriate measure of language norms, and therefore of integration or lack thereof, is within the bilingual community itself (see Poplack & Meechan, 1998; Pires & Rothman, 2009; Torres Cacoullos & Aaron, 2003).

To create comparison datasets for each of the languages in contact, a roughly equivalent number of Spanish-origin nouns surrounded by Spanish discourse (N=856) and up to the same number of English-origin nouns surrounded by English discourse (N=608) was extracted from the speakers, taking into account the number of occurrences provided by each in the lone English-origin dataset. ⁴ This ensured that the monolingual data were comparable with the English-origin data in terms of any individual biases or tendencies. For these datasets, extraction began at the 15-minute mark for each interview.

Frequency distribution of lone English-origin nouns

The 820 lone English-origin nouns represent a total of 405 types. Of the 820 occurrences, 30% (N=244) appeared only once in these data (these will be referred to as “Singleton” lone items), and another 18% (N=148, with 74 types) appeared only twice. Another 23% (N=186, with 52 types) appeared three to five times, and 11% (N=92, 13 types) appeared six to nine times. The remaining 20% of the data (N=150) includes six types, which occurred between 12 and 51 times. These six types were: uranium (N=12), troca/truck (N=15), grandpa (N=19), daddy (N=25), grandma (N=28), and dad (N=51). Figure 1 shows this distribution.

OPEN IN VIEWER

Figure 1.

Breakdown of lone English-origin nouns according to number of occurrences (N=820). For example, 30% of all tokens are of types that occurred only once.

However, given that 70 of the repeated items, such as uranium (N=12), were simply repeated multiple times by the same speaker, a better gauge for how widespread the English-origin words were is how many speakers used them. There were 314 types (N=433) used by only one speaker; of these, 244 occurred only once. There were 52 types (N=154) produced by two speakers. Finally, 12 types (N=51) were produced by three speakers, five types (N=53) by four speakers, and six types (N=129; grandpa, mom, weekend, troca ‘truck’, grandma, and dad) by six or more speakers. Those produced by three or more speakers will be considered “Diffuse” in this paper. Figure 2 shows this distribution both in terms of number of occurrences and number of types.

OPEN IN VIEWER

Figure 2.

Breakdown of lone English-origin nouns according to number of speakers (for “Occurrences,” N=820; for “Types,” N=389). For example, 129 of all 820 occurrences (i.e. 16% of the data) were produced by six or more speakers; however, these 129 occurrences are made up of only six types (i.e. distinct lexemes), making up under 2% of the 433 lexical items.

To recap, I will call those produced only once “Singletons,” and those produced by three or more speakers “Diffuse.” This allows us to clearly distinguish between those items that are commonly used within the community and are therefore likely loanwords (and may be transmitted by other speakers), from those whose loanword status is not established (and which therefore may entail active access to English). In the case of Diffuse items, we would expect these to pattern like Spanish-origin nouns (which, for the most part, they do). In contrast, the Singleton items may pattern like English nouns, revealing them to be, in the aggregate, single-word code-switches. Alternatively, they may pattern like the Diffuse English-origin nouns and like the Spanish nouns, revealing their aggregate status as nonce loans that are instantly integrated morphosyntactically (Poplack, 2012; Poplack & Meechan, 1998).

Semantic domains

It has often been argued that lexical borrowings tend to be drawn from certain semantic domains, particularly those that have to do with new cultural items that are associated with the contact culture. Smead (2000) identifies two types of loanwords: “unique” loanwords, defined earlier, and “synonymic” loanwords, i.e. “those that compete for the same semantic space with a native language term” (p. 292). Teschner’s (1974) annotated bibliography reveals several semantic domains that appear more likely to have English-origin words, including education, food, sports, and technology. It has often been assumed that highly frequent or core vocabulary items (Swadesh, 1971) are not likely to be borrowed (e.g. Smead, 2000, p. 282). However, Myers-Scotton and Okeju (1973) argue that (based on evidence from the East African language Ateso) “borrowings within the core vocabulary itself are also very common, given sufficiently extensive contact with another culture” (pp. 872–873).

The preliminary categories coded here were gleaned from the previous literature. These were then supplemented as the analysis was guided by the data. The 17 semantic categories coded are: kinship terms (e.g. dad); everyday items (e.g. bag); events and places (e.g. birthday); person (not kin) (e.g. firefighter); year or number; work- or money-related (e.g. spending money); related to the land or earth (e.g. mountain); technology (e.g. treadmill); academia (or school) (e.g. homework); food, drink, and smoke (e.g. popcorn); vehicle or transport-related (e.g. troca); institution (e.g. social services); animal (e.g. bird); abstract time/space concept (e.g. weekend); health and body (e.g. eyedrops); domestic life (e.g. sewing); and linguistics or language (e.g. Spanish). Nominalized adjectives and other items that did not fall into one of these domains were coded separately.

The tendency for both culture-specific and more universal (or “core”) terms to be borrowed in this community (cf. Torres Cacoullos & Aaron, 2003) is seen in Table 1. The last column shows the proportion of each domain that is produced as lone English items (based on a comparison with the Spanish and English samples described above).

OPEN IN VIEWER

Table 1.

Semantic domains per lone English-origin nouns vs. Spanish and English nouns in corresponding samples.

	Singleton		Diffuse		Spanish		English		Total	% English-origin
	N	%	N	%	N	%	N	%
Everyday items	34	14	4	2	84	10	65	11	187	20
Events and places	28	12	8	3	34	4	54	9	124	29
Other	28	12	18	8	86	10	78	13	210	22
Person	24	10	17	7	83	10	60	10	184	22
Year or number	19	8	4	2	4	<1	14	2	41	56
Technology	19	8	4	2	21	2	14	2	58	40
Earth	14	6	6	3	58	7	22	4	100	20
Food, drink, and smoke	14	6	0	–	41	5	25	4	80	18
Health	11	4	0	–	37	4	38	6	86	13
Academic	9	4	11	5	33	4	14	2	67	30
Nominalized adjective	9	4	0	–	19	2	4	<1	32	28
Animals	8	3	0	–	23	3	13	2	44	18
Domestic items	7	3	0	–	24	3	56	9	87	8
Institutional	7	3	0	–	10	1	17	3	34	21
Work or money	7	3	0	–	61	7	27	4	95	7
Vehicles	4	2	19	8	14	2	7	1	44	52
Abstract time/space	3	1	8	3	101	12	52	9	164	7
Kinship terms	2	<1	134	58	74	9	28	5	238	57
Linguistic	2	<1	0	–	40	5	20	3	62	3
Total	244	100	233	100	856	100	608	100	1941	25

This analysis shows that English-origin nouns do indeed occur relatively more often in particular semantic domains. These domains, however, are not all on the list of usual suspects, which, as mentioned, include technology, food items, and academia. The most favorable contexts for English-origin lone nouns in these data are kinship terms (which were produced as lone English-origin nouns 57% of the time), years and numbers (56%), technology (40%), and vehicles (52%). The items referring to events, technology, or vehicles, as in (10), (11), and (12), may indeed represent cultural novelties (though in the case of vehicles, 79% (15/19) of the Diffuse items are troca). However, the category of kinship belongs to the core vocabulary (Smead, 2000, p. 282) and cannot in this case be said to be culturally specific or represent lexical gaps.

(10)

The English-origin kinship terms found in these data exhibit certain lexical patterns that provide evidence that the conventionalization of English-origin nouns occurs at both the lexical and semantic levels (cf. Poplack et al., 1988), since we find both dad and grandma (but not other kinship terms) among the most frequent. For example, dad and daddy combined occurred 76 times as lone nouns, compared with four occurrences of papá among the Spanish nouns. In contrast, mom/momma occurred only eight times in the lone nouns, and mamá 15 times in the Spanish nouns. Grandma had 30 occurrences; abuela only one. Grandpa occurred 20 times, abuelo three. There were no occurrences of brother or sister, while mentions of hermano, hermana, and hermanos were made 11 times. Son and daughter did not appear (though son-in-law occurred once); hija and hijo occurred 18 times combined.

These patterns are mirrored in the corpus as a whole: English terms are used over 1.5 times as often as Spanish terms to refer to dads (305 vs. 182 tokens), while for moms, the breakdown is reversed (178 tokens of English terms vs. 219 in Spanish). ⁵ Given the overwhelming preponderance of kinship terms, they will be excluded from the analyses of morphosyntactic and discourse patterns in the Diffuse dataset, as they tend to display peculiar tendencies that would skew the results (for example, they show a greater tendency than other nouns to be referential, to occur in subject position, and with possessive marking).

In the datasets in this study, while the core items of kinship show evidence of high rates of English-origin nouns, the core items in the domain of abstract time or space, such as day, are much less likely to appear in English within Spanish discourse (2%, N=15), when compared with these types of items in Spanish (12%, N=101). Over half of these English-origin items are weekend(s) (N=8); items like day and year did not appear among the lone nouns, though there were 14 and 23 occurrences of their Spanish counterparts, día and año, in the Spanish, respectively. Nonetheless, the relatively high rate of specific years and numbers in the English-origin data shows that use of English in this related context is a local community norm. Indeed, the 36 occurrences of years (some as Singletons, some repeated across speakers, some Diffuse), as shown in (13), (14), and (15), are best understood as instances of a construction in which it is established practice to use English in this community.

(13)

(14)

(15)

A third area that at first blush seems to be disproportionately favorable to lone nouns is that of vehicles. However, 15 of the 23 occurrences are represented by troca, which indicates that this is a lexical effect, rather than a semantic class effect. Thus, while some use of English-origin nouns are domain-specific (e.g. numbers, kinship), there are also lexical effects.

Gender

Gender has been a primary target for the examination of the status of lone items because it can be understood as one indication of morphosyntactic integration: if the gender patterns are similar to those of Spanish-origin nouns, then it can be argued that they are morphosyntactically integrated, and thus not code-switches (e.g. Poplack, Pousada, & Sankoff, 1982; Zamora, 1975; Zamora Munnt & Btjar, 1987). A preference for masculine gender assignment has been suggested in several previous studies, among L2 learners (Martinez-Gibson, 2011), bilinguals (Chaston, 1996; Clegg, 2010: 16; Garcia, 1998; Montes-Alcalá & Lapidus Shin, 2011; Sánchez, 1995: 134-137), and monolinguals (Banfield, 1994; Pérez-Pereira, 1991; Smith, Nix, Davey, López Ornat, & Messer, 2003).

To determine gender for both Spanish-origin and English-origin nouns, overt cues beyond the noun itself were used. These cues included the gender of determiners and adjectival modifiers, as in (16), in which machine was coded as feminine due to the feminine marking on otra. If no overt cues were present, the noun was coded as unmarked for gender. In cases in which the semantics of the noun dictated one or the other gender, such as girls, a separate code was used, to ensure that all nouns coded as having been assigned a gender were, at least in theory, sites at which either gender could have been assigned. ⁶ Overt cue-based gender coding was necessary across all datasets to ensure comparability of the coding.

(16)

Table 2 and Figure 3 show the results regarding gender assignment for these data. Note that if we compare only masculine and feminine, we get a rate of 14% (16/117) feminine for Singletons. This is nearly identical to the results reported by others for other bilingual populations (Mota, cited in Otheguy & Lapidus, 2003, pp. 214–215; Otheguy & Lapidus, 2003; Poplack, 1982). Also notable here is the high rate of unmarked nouns, at 51% for Singletons. A superficial analysis of Singletons alone might lead us to take this as evidence of their lack of grammatical integration. However, a comparison with the rate of unmarked Spanish nouns, also high at 39%, discourages such an interpretation. This slightly elevated rate of the absence of gender marking can be attributed not to a lack of grammatical integration, but rather to the greater proportion of the Singletons used in nonreferential contexts, as will be discussed below.

OPEN IN VIEWER

Table 2.

Distribution of gender marking on nouns in Singleton, Diffuse, and Spanish datasets.

	No info		Masc.		Fem.		Total
	N	%	N	%	N	%	N
Singleton	122	51	101	42	16	7	239
Diffuse	31	35	39	44	19	21	89
Spanish	313	39	239	30	259	32	811

OPEN IN VIEWER

Figure 3.

Distribution of gender marking on nouns in Singleton, Diffuse, and Spanish datasets (percentages).

The preference for masculine gender reported in other studies of English-origin nouns in Spanish is found among these speakers as well, for both Singleton and Diffuse lone items, at 42% and 44%, respectively, compared with 30% for the Spanish nouns dataset. In contrast, only around 7% of the Singleton data was assigned feminine gender, compared with 21% for the Diffuse lone items and 32% for the Spanish nouns data. Figure 3 illustrates the contrast between Spanish and Singletons; Diffuse lone English-origin items fall in the middle.

The preference for masculine, however, likely has nothing to do with code-mixing tendencies per se, but may rather simply follow from patterns and preferences that are internal to Spanish. In a connectionist analysis of a longitudinal database of parental production, Smith et al. (2003) noted that “while regular feminine nouns were slightly more frequent than regular masculine nouns, irregular masculine nouns outnumbered irregular feminine nouns by roughly 2 to 1” (p. 306). Based on this input, which preserved type and token frequencies, a computer-generated model produced a similar bias toward masculine gender assignment to novel words, suggesting that the frequency distribution has a direct role in gender assignment. ⁷ When applied to lone English-origin items, such experiential patterns would likewise lead to a preference for masculine gender assignment.

Discourse function

Lacking from most previous studies of English-origin nouns in Spanish is consideration of patterns of referentiality (Hopper & Thompson, 1984; Thompson, 1997). This turns out to be a particularly profitable context where we may be able to measure the extent to which lone English-origin nouns are grammatically integrated, and can therefore be considered to be functioning as Spanish nouns and not as code-switches to English.

The data were coded for referentiality, which, in Hopper and Thompson’s (1984, p. 711) terms, refers to the “manipulability” of the referent; that is, “a noun phrase is referential when it is used to speak about an object as an object, with continuous identity over time” (Du Bois, 1980, p. 208). Thompson and Hopper (2001) identified three contexts in which nouns perform nonreferential functions: to form intransitive predicates with semantically weak verbs, to classify, and to orient. These contexts are illustrated in (17), (18), and (19), respectively.

(17)

(18)

(19)

Referentiality is key to the study of single other-language-origin nouns, as Torres Cacoullos and Aaron (2003) found that nonce-loan nouns in New Mexican Spanish often occurred in predicating functions, and were disproportionately frequent in classifying contexts, where they designated occupation or social status as predicate nominals. This led them to conclude that “It is not lack of grammatical integration, but these nonreferential uses in recipient-language predicates, that is manifested in bare nonce-loan nouns” (p. 293).

If we examine Table 3 and Figure 4, we see that both lone nouns datasets (Singleton and Diffuse) showed higher rates than both English nouns and Spanish nouns in the predicating role. This means that the former were more likely to appear as nonreferential objects of semantically weak verbs, thereby forming verbal compounds (Thompson & Hopper, 2001), as in (17). However, given that both Diffuse and Singleton lone nouns demonstrate this tendency, and this is not a conflict site between Spanish and English because nouns from both languages readily occur in this context, this cannot be understood as a measurement of grammatical integration. Instead, we may posit that this reflects the community practice of creating novel compounds, perhaps at times to express culturally specific concepts, as in (20). This could potentially be linked to the similar community practice of creating verbal compounds with hacer ‘do’ + V, which is a novel bilingual device (Wilson, 2013; Wilson & Dumont, 2015).

(20)

OPEN IN VIEWER

Table 3.

Distribution of nonreferential functions of nouns in Singleton, Diffuse, Spanish, and English datasets.

	Orienting		Classifying		Predicating		Total Non-ref.		Total
	N	%	N	%	N	%	N	%	N
Singleton	59	24	41	17	64	26	164	67	244
Diffuse	34	34	4	4	25	25	63	64	99
Spanish	259	30	69	8	159	19	487	57	856
English	170	30	85	14	95	16	350	58	608

OPEN IN VIEWER

Figure 4.

Distribution of nonreferential functions of nouns in Singleton, Diffuse, Spanish, and English datasets (percentages).

In fact, in terms of discourse referentiality, the only notable difference between English nouns and Spanish nouns is found in the classifying role, as in (18). Here, we find that English nouns and Singletons are more likely to fulfill this function. However, 22% (9/41) of the classifying Singletons refer to occupation, compared with 9% (8/85) of the English nouns, 9% (6/69) of the Spanish nouns, and none of the Diffuse nouns. The nonreferential use of Singletons in occupation-related classifying function was also found in Torres Cacoullos and Aaron (2003), and may explain in part some of the gender-marking prevarication noted by Otheguy and Lapidus (2003) in this context. For example, in (21), the lone item does not actually refer to the person herself, but only a category to which she belongs. Since the category itself—social worker—is unmarked for gender (i.e. both women and men can fulfill this role), its gender assignment would not necessarily be linked to that of the woman, who represents only one example of somebody in that category.

(21)

The explanation for the elevated occurrence of English nouns in classifying role, however, is not the same as for Singletons. Instead, it comes from variations of the expression that’s the thing, which accounted for another 9% (8/85) of the English nouns data; in contrast, cosa ‘thing’ occurs only once in classifying position in the Spanish nouns dataset. Finally, overall, Singletons are less likely to perform orienting functions (24%) than nouns in all other datasets; examples include numbers and words such as north, winter, month, and weekend.

A final site in which the ramifications of the discourse functions of English-origin nouns in New Mexico can be found is in the determiner patterns. Determiners have been cited as an ideal area in which transfer might be identified (e.g. Montrul & Ionin, 2010). However, the scant evidence that exists on the question of transfer involving determiners among Spanish-English bilinguals offers mixed results. One difficulty lies in the relative scarcity of obvious conflict sites in English and Spanish.

Two clear loci of inter-lingual difference do exist, however. First, in contexts of inalienable possession, as in (23), Spanish generally has a definite article, while English would use a possessive pronoun. Second, in generic contexts, as in (24), Spanish generally has a plural definite article, while English nouns in these contexts tend to be bare. Unfortunately, these two contexts were rare in the Singleton data (N=1 for inalienable possessions; N=20 for generics), and thus could not be analyzed here. In order to gain a fuller understanding of conflict sites between typologically similar languages, then, we must move beyond these clearer cases. Corpus studies are particularly well suited for more fine-grained analysis.

(23)

(24)

Determiner types considered here included bare (25), definite (26), indefinite (27), possessive (28), demonstrative (29), and quantifier (30).

(25)

(26)

(27)

(28)

(29)

(30)

Table 4 and Figure 5 show the distribution of determiner types for all four datasets. Regarding definiteness, it has been proposed that definite articles are more frequent, and indefinite articles less frequent, in Spanish than in English (see, e.g. Vargas-Barón, 1952, p. 410). This characterization is borne out here: the Spanish data show 43% definite articles, compared with 30% in English; conversely, Spanish shows 7% indefinite articles, compared with 16% in English. The Diffuse items line up with Spanish in terms of definite articles, at 43%, and are closer to Spanish than to English in indefinite article rates. Singletons, however, as shown clearly in Figure 5, are in line with English for both, at 29% definite articles and 17% for indefinite articles. This suggests that, at least sometimes, speakers may treat these items like English in terms of (Spanish) article usage. However, if possessives and definite articles are combined into one category, the English looks much like the Spanish, suggesting that the “conflict” may not be as great as it appears at first glance.

OPEN IN VIEWER

Table 4.

Distribution of determiner types of nouns in Singleton, Diffuse, Spanish, and English datasets.

	No Det.		Definite		Indefinite		Possessive		Quant.		Demon.		Total
	N	%	N	%	N	%	N	%	N	%	N	%	N	%
Singleton	114	47	70	29	42	17	7	3	8	3	3	1	244	100
Diffuse	30	30	43	43	11	11	3	3	7	7	5	5	99	100
Spanish	253	30	369	43	62	7	73	8	67	8	32	4	856	100
English	176	29	180	30	97	16	96	16	38	6	21	3	608	100

OPEN IN VIEWER

Figure 5.

Distribution of determiner types of nouns in Singleton, Diffuse, Spanish, and English datasets (percentages).

One context in which Singletons do indeed seem to align with English is in the predicating context. Only 1% (15/159) of the predicating nouns in Spanish occurred with indefinite articles, as in (31); much more common was zero marking, as in (32), produced by the same speaker only moments earlier. This compares with 19% (12/64) indefinite articles among predicating Singletons and 24% (23/95) in English, as shown in Figure 6. Notably, the indefinite article with Singletons in this context, as in (33), tends to occur with culture-specific nouns, such as guzzler, recreational vehicle, and B-47. Thus, the predicating noun construction may be a context in which we may see some one-word code-switches or transfer in this particular construction, pending investigation with more robust numbers.

(31)

(32)

(33)

OPEN IN VIEWER

Figure 6.

Percentage of indefinite articles among predicating nouns in Singletons, English nouns, and Spanish nouns (N=64, 95, and 159, respectively).

Nevertheless, Singletons do not generally align with English when it comes to bare marking. Singletons clearly stand out; at 47% bare, these nouns are marked with zero about 16 percentage points more than nouns in the other datasets, all of which hover near 30%. It is worth noting that 13% (15/114) of the bare singletons are numbers or years, compared with 7% (13/176) of the English nouns. So, while the prevalence of numbers in the Singletons may play some part in the elevated rate of bare nouns, the overall trend goes beyond lexical tendencies: it is linked to nonreferentiality. Nonreferentiality is common among bare nouns, both in Singletons, at 75% (N=86/114), and in Spanish nouns, at 83% (N=184/253). The higher proportion of nonreferential contexts with Singletons, then, would explain the higher rate of bare nouns in this dataset (see also Torres Cacoullos and Aaron 2003).

In sum, an examination of morphosyntactic factors that are generally understood to measure grammatical integration—and are therefore used to help determine the status of other-language-origin nouns as borrowings or code-switches—through the lens of discourse, semantics, and lexical patterns, has revealed the fingerprint of this community’s code-mixing norms. The morphosyntactic patterns found are in fact reflections not of a lack of grammatical integration, but rather of the use of English-origin words to perform certain discourse functions. These norms are established locally; there is nothing about English nouns in particular that makes them more or less suitable in these contexts. As we saw in section 3.2 above, the power of the community norm in linguistic usage is found at the semantic and lexical levels as well, with years and certain kinship terms generally expressed as English-origin nouns.