Abstract
This study examines the bilingual compound verb hacer ‘to do’+
Introduction
The investigation of bilingual behavior has often been approached through a comparison of bilingual and monolingual behavior (Edwards & Gardner-Chloros, 2007; Gafaranga & Torras, 2002) with the underlying assumption that the cognitive processes of acquiring, storing, accessing, and using language may be markedly different (and potentially problematic, difficult, or confusing) for speakers of more than one language as compared with monolinguals. This bias is evident in the characterization of certain bilingual behaviors as being motivated by the ‘cognitive load’ of having twice as many linguistic elements to keep track of (Silva-Corvalán, 1994, p. 6).
The present research contributes to the understanding of bilingual behavior by examining the bilingual compound verb (BCV) hacer+ (1) hacer+ Miguel pero he was the one who, ‘but he was the one who, … lo and it was square. and it was square.’
In order to test the cognitive load hypothesis we examine disfluencies in the proximity of intonation units (IUs) containing the construction hacer+
One of the first linguists to document BCVs was Moravcsik in her 1975 article “Verb borrowing”. Found in many contact situations, BCVs typically “consist of a light or ‘operator’ verb, frequently translatable as ‘do’ or ‘make’, which appears with a lexical element (the main verb) which contributes the core semantic content of the construction”; the light verb contributes tense, aspect, and mood (TAM) to the construction (Edwards & Gardner-Chloros, 2007, p. 74). Edwards and Gardner-Chloros (2007) list over 15 different contact situations described in the literature in which BCVs occur, including pairs as diverse as Popoloca-Spanish (Veerman-Leichsenring, 1991) and Japanese-English (Stanlaw, 1982). In Spanish-English contact situations, usage of hacer+
In hacer+
It has been assumed that one of the primary motivations for the use of hacer+
Data and participants
The data come from the New Mexico Spanish-English Bilingual (NMSEB) corpus, (Torres Cacoullos & Travis, in preparation). This is a unique bilingual corpus in that it comprises large amounts of speech in Spanish and English, with the participants freely switching between the two, produced by a community of Spanish-English bilinguals. Participants in this corpus are all bilinguals living in Central or Northern New Mexico who acquired both languages in childhood and use both languages on a daily basis, including within the same conversation (cf. Travis & Torres Cacoullos, 2013; Torres Cacoullos & Travis, 2015). The subset of the corpus used for this analysis comes from speech of 24 participants, representing over 16 hours of speech. All participants were between 18 and 89 years of age at the time of the recording, with a mean age of 66 years. For this study, every token of the BCV hacer+
The conversations were transcribed in ELAN 3 (cf. Sloetjes & Wittenburg, 2008) following the conventions outlined by Du Bois, Scheutze-Coburn, Cumming, and Paolino (1993), in which each IU (cf. Chafe, 1987) is transcribed on a separate line, and interactional and other production phenomena (pauses, truncations, overlap, laughter, vocal noises, inhalations, etc.) are systematically noted. As others have highlighted (Chafe, 1987; Croft, 1995; Shenk, 2006), the empirical study of IUs provides insight into the way that linguistic structure is evidenced in patterns of speech, and in the case of bilinguals, the degree to which two languages are combined (Torres Cacoullos & Travis, 2015). The non-speech phenomena that are included in the transcription (including those we focus on here, namely pauses and truncations) provide us with means of quantifying production-related phenomena that can in turn inform us about the on-line processes experienced by the speakers at the moment of speech.
The following sections address three research questions in order to achieve an understanding of individual variation in language combining patterns: (1) is hacer+
Research question 1: Is hacer+Verb Eng associated with a greater cognitive load?
In this section we test the proposal that hacer+ (2) Pausing before hacer+ Me, me corrí . . . me
Silva-Corvalán (1983, 1994) has used the term cognitive load to explain code-switching and borrowing as a linguistic strategy among bilinguals. In studying bilingual behavior among English-dominant Spanish/English bilingual adolescents, she suggested that these speakers switched as a compensation strategy due to memory lapses reporting that “there was a high percentage of switches preceded by editing phenomena such as false starts, hesitation, and pauses” (1983, p. 78). Following up on cognitive issues in her in-depth study of bilinguals in Los Angeles, Silva-Corvalán (1994, p. 6) writes: The general hypothesis investigated is that, in language contact situations, bilinguals develop strategies aimed at lightening the cognitive load of having to remember and use two different linguistic systems.
In testing the cognitive load hypothesis here, we use Dumont (2010) as a point of departure. In examining disfluencies in bilinguals, Dumont (2010, p. 4) proposed the following operationalization of the cognitive load hypothesis in another study of New Mexican Spanish: … if more bilingual speakers have a heavier cognitive load than less bilingual speakers, then the more bilingual speakers will demonstrate higher overall rates of repair, different types of repair, repair affecting different parts of speech, and increased rates of repair preceding other language items.
Examining rates of disfluencies in terms of pauses and truncation used by more and less bilingual speakers of Spanish and English, she found that pauses did not “serve as additional markers of processing difficulty” (2010, p. 20) for more bilingual speakers and that truncation did not occur more frequently when code-switching was present. Her study thus found no evidence to suggest that code-switching was motivated by production difficulties, nor that these disfluencies serve as flagging devices in this speech community (see Gardner-Chloros, McEntee-Atalianis & Paraskeva (2013, p. 3) for an overview of pauses as possible flagging in bilingual behavior).
Results from other studies have been conflicting. Hlavac (2011, p. 3803–4) found a disproportionally higher rate of hesitation and monitoring phenomena (e.g. pauses and false starts) around code-switching and borrowing in Croatian/English speakers and proposed that disfluencies may perform the interactional function of integrating imported forms. On the other hand, in an examination of rates of disfluencies in ‘bilingual’ and ‘monolingual modes’ of Hungarian bilinguals with a variety of L2s, Navracsic (2004, p. 108) did not find a higher rate of unfilled pauses or repetitions (including with truncation) in bilingual mode. Gardner-Chloros et al. (2013, p. 3) focused their examination of ‘flagging’ on pauses in the vicinity of code-switching, proposing that it is the most comparable type of hesitation and monitoring phenomenon between monolingual and bilingual speech. The rate of pauses around code-switching was not significantly different between the more monolingual Greek Cypriot speakers and the more bilingual Greek/English group who participated in their study. This was taken as evidence that the processing of code-switching was not more cognitively demanding than monolingual processing.
Psycholinguists concerned with the cognitive costs associated with bilingualism have also failed to achieve consensus on this issue. On the one hand, there is some evidence to suggest that bilinguals are at a disadvantage, relative to monolinguals, in lexical retrieval (cf. Gollan & Acenas, 2004; Gollan, Montoya, Cera, & Sandoval, 2008; Gollan, Montoya, Fennema-Notestine, & Morris, 2005; Gollan, Montoya, & Werner, 2002). On the other hand, Kroll, Dussias, Bogulski, and Valdés Kroff (2012, p. 230) report that there are “relatively few processing costs to bilinguals”, and “that proficient bilinguals do not make inadvertent and random errors of language” (2012, p. 232). This is supported by recent psycholinguistic research (cf. Hartsuiker, Pickering, & Veltkamp, 2004; Kroll et al., 2012; Loebell & Bock, 2003; Tily et al., 2009), which points to the simultaneous activation of two linguistic systems, with no evidence that this is problematic for bilinguals. Such work indicates that the practice of regularly using two languages does not come with a higher cognitive load.
The findings of the present study, which examines naturally occurring data, suggest that linguistic patterns are best interpreted through a larger scale understanding of linguistic practices of a speech community and that two linguistic systems do not appear to impose significant difficulties upon the bilingual speaker.
In order to test the cognitive load hypothesis in relation to the production of hacer+
The comparative data come from the first 1000 IUs of the 12 recordings in which hacer+
Truncation
We examine truncation prior to the occurrence of hacer+ (3) Truncation 1 Ivette sí tengo .. este sentimiento que yo sabía, ‘I do have .. this feeling that I knew, –TRNC 2 .. que yo podía 3 … yo digo que en mi mente yo -- … I say that in my mind I -- 4 … yo
IUs 2 and 3 were coded as having no preceding truncation, while IU 4 was coded as preceded by truncation. In addition, the second and fourth IUs were coded as having instances of hacer+
Table 1 shows the rate of truncation in these four environments. We see that of the 96 IUs that contained hacer+
Truncation in the preceding Intonation Unit (IU).
IU internal pauses
The second disfluency examined was the presence of an IU-internal pause,
4
as in the following example: (4) IU internal pauses 1 Ivette no puedo estar yendo pa’ atrás y pa’ adelante. ‘I can’t be going back and forth. ABSENT 2 and then -- and then -- ABSENT 3 es cuando .. hicimos decide de, that’s when .. we decided to, 4 jun- de, toge- to, ABSENT 5 … hacer rent. … to rent.’ ABSENT
In example (4) two pauses are noted: in IUs [3] and [5]. The type of pause in IU [5] was not included in our analysis, because it occurs at the beginning of an IU. The rationale for examining IU-internal instead of IU-initial pauses is that by occurring in the middle of the IU, they are more indicative of a production-motivated interruption to speech, whereas IU-initial pauses may have interactional significance, such as lack of turn-uptake at a transition-relevance place (cf. Ford and Thompson, 1996). As in the case of preceding truncation, we excluded cases where overlap or unclear speech could have made the determination of a pause uncertain. We also excluded cases where hacer+
Table 2 shows the rate of occurrence of IU-internal pauses in IUs containing hacer+
Intonation Unit (IU) internal pauses.
The similar rates of occurrence of pauses within IUs containing hacer+
Research question 2: hacer+Verb Eng as a productive verbal construction – a discourse-morphosyntactic-prosodic profile
In order to obtain a measure of productivity through a discourse profile of the BCV hacer+
Firstly, we find evidence for the productivity of the construction in the wide range of English verb types that occur. The 116 tokens of hacer+
At the same time, it shows a continuum of conventionalization. There are eight types that are used by two or more speakers, listed in Table 3. Only one combination could be considered widespread in the community according to these data – hacer retire, which was found in four separate conversations, with a total of six tokens being produced across four different speakers. The Spanish-origin closest equivalent jubilar(se) occurs only once in the NMSEB, while forms based on English ‘retire’ and retirar occur 52 times. 5 Jenkins (2003) also reported hacer retire in an earlier corpus of New Mexican Spanish, signaling that it has been in use for at least two decades. Examining the remaining seven types, each used by two speakers, we see that they account for less than 10% of all bare English infinitive types (7/73), and 19% of all tokens (22/116).
Collocations.
Evidence for productivity is also seen in the wide variety of tense-aspect-moods. While the majority of the data occur with the present indicative (40/116), imperfect (32/116), and preterit (27/116), as would be expected, there are tokens in a range of other forms, including the future, the perfect, subjunctive, and infinitive following a preposition (17/116).
An important observation concerns the distribution of the construction over prosodic units, the IU. The overwhelming majority (105/116) of hacer+
(5) hacer+ Gabriel y qué hacían? ‘and what did they do?’ Francisco .. papas=. ‘.. potatoes=.’ Gabriel .. oh[=]. ‘.. oh [=].’ Francisco [hacían], ‘[they], cultivate papas [2o=2], cultivated potatoes [2or=2],’ Gabriel [2yeah2]. ‘[2yeah2].’ Francisco … (H) áhi iban todos a trabajar. ‘… (H) they all went to work there.’
Strong morphosyntactic evidence for the unitary status of the construction and its integration into Spanish grammar is seen in the fact it freely occurs with clitic pronouns (43/116 occur with some kind of clitic, including direct and indirect objects and reflexives). Example (6) illustrates a reflexive pronoun; example (1) above illustrates a direct object pronoun, here with clitic climbing, which has been taken as evidence of auxiliary status and the fusing of the two items into a single constituent (Torres Cacoullos, 1999).
(6) hacer+ Ivette no sabe cómo
Research question 3: hacer+Verb Eng in the community – who uses it?
Variability in bilingual behavior can be indicative of community norms (cf. Poplack, 1988) or individuals’ bilingual ability (cf. Poplack, 1980). The speech community under consideration is one in which code-switching is widespread (Gonzales, 1999), yet hacer+
In order to identify tendencies correlated with speakers’ language use, each IU was tagged for language. Multiword IUs were chosen for this analysis so as to restrict the IUs under consideration to only those contexts in which a bilingual IU could occur; apparent multiword IUs where no language change was likely or possible (e.g. collocations like I mean, I guess, ¿que no?) were excluded. (In restricting the count to multiword IUs, we depart from Shenk, 2006, p. 188, who includes single-word IUs that are precluded from being bilingual.) The multiword IUs were assigned one of three codes depending on the language(s) used: English (E), Spanish (S), and bilingual (B). Example (7) illustrates this coding.
(7) Language tagging Fabiola …(1.0) to get the problems fixed, ‘…(1.0) to get the problems fixed, porque the, because the, …(0.7) ese dientista, …(0.7) that dentist, .. que estaba haciendo plastic surgery, .. that was doing plastic surgery, …(0.9) he would just, …(0.9) he would just,
The criteria used to classify an IU as bilingual was intentionally as broad as possible in order to capture the whole range of bilingual phenomena and counted the presence of any lexemes deriving from both languages in one IU as bilingual. We excluded established borrowings that are attested in monolingual varieties of Spanish or English outside the US Southwest (as evidenced in their appearance in monolingual dictionaries), such as rifle or adobe, but we included borrowings that are likely established within this community but are not found outside this variety (dad, grandma, numbers) (see Aaron, 2015, for a discussion of such items). Spanish IUs containing semantic calques, such as aplicar meaning ‘apply for (a job, benefits)’ and planta meaning ‘factory or plant’, were not coded as bilingual, as the lexemes themselves have not been borrowed.
As shown in Table 4, not all participants used hacer+
Speaker usage of hacer+
Shaded cells indicate that over 70% of a speaker’s intonation units (IUs) occurred in one language.
Individual rates of hacer+
IU: intonation unit.
Table 6 shows the language profiles of the attested and non-attested producers of hacer+
Language use and hacer+
When we look at the proportion of bilingual IUs produced per speaker, however, differences emerge between the two groups. As can be seen in Table 4, there is a wide range of individual variability in the proportion of bilingual IUs produced, from 1% of all multiword IUs to 23%. We note that the six attested users of hacer+
We now turn to consider all speakers, taking into account the full range of types of bilingual IUs employed. As Table 7 shows, speakers who have overall rates of <5% bilingual IUs do not produce the full range of bilingual IUs in these data. They all produce IUs with single other-language word items (nouns, adjectives), and nearly all use discourse markers (so, anyway, pues, etc.) and multiword switches (… se llenaba la troca with seventy dollars, ‘he would fill up his pick-up with seventy dollars,’ (09 La salvia, 0:36:26–0:36:28). But no speaker in this group incorporated an English origin verb with Spanish morphology (e.g. wreckiar ‘to wreck (a car), misstir ‘to miss’, flunkear ‘to flunk’), which in fact were very infrequent in all speakers (<20 tokens overall). In the intermediate group, those who display overall rates of 5–10% bilingual IUs, all bilingual IU types were attested, but not all members displayed all types. When the group with the highest rates of bilingual IUs is examined, the full range of bilingual IU types is seen, and all speakers incorporate lone verbs using at least one of two strategies. Again, because the amount of data from individual speakers varies widely and given the relative infrequency of English-origin verbs, our statements are offered cautiously.
Type of bilingual intonation unit (IU) by % of overall bilingual IUs (N = 24 speakers).
The findings presented in Table 7 are suggestive of a continuum of bilingual behaviors. All of these speakers regularly make use of both languages within a conversation but they do not do so in the same ways. The more bilingual IUs a speaker produces, the more types of bilingual IUs he or she will employ, with what are arguably the most morphologically complex strategies (those involving the incorporation of a lone English origin verb) occurring among those with higher overall rates of bilingual IUs. Another way to represent the distribution might be something like that shown in Figure 1.
Continuum of emergence of bilingual Intonation Unit (IU) types.
While there are both synthetic and periphrastic strategies for incorporating lone English origin verbs with Spanish morphology into bilingual discourse (as well as some calques, such as aplicar ‘apply’ and retirarme ‘retire’), hacer+
Discussion
The production of hacer+
We have also provided quantitative evidence that hacer+
The language patterns observed suggest that, contrary to the view that hacer+
Footnotes
Funding
This work was partially supported by National Science Foundation (NSF) grant #1019112/1019122 awarded to Rena Torres Cacoullos and Catherine E. Travis to support the development of the NMSEB corpus.