Torres Cacoullos, Rena and Catherine E. Travis: Bilingualism in the community. Code-switching and grammars in contact.

1 Introduction

One of the most debated topics in research on linguistic diversity (historical, typological, and sociolinguistic alike) is how language and population contact affect (the evolution of) grammar. Bilingualism in the community by Torres Cacoullos and Travis addresses this question from the perspective of observed patterns of variation in bilingual language use, and through a principled method of analysis which rests upon (1) comparisons between usage patterns in distinct language varieties used by the same group of bilingual speakers, and (2) comparisons between these bilingual varieties and their monolingual benchmarks. The contact scenario investigated in the book is the Spanish and English-speaking community of northern New Mexico, in the southwest of the United States. The linguistic variable under investigation is subject pronoun expression.

While Spanish is a typical instance of a null-subject language, where subject pronouns often go unexpressed, English tends to overwhelmingly express them. This striking crosslinguistic difference in usage preferences has been taken as an ideal test case for the study of contact-induced change, under the assumption that Spanish varieties with English as a contact language would display a much stronger preference for subject pronoun expression, in alignment with the contact language. While this is a widely investigated issue in the field of Spanish contact studies (Silva-Corvalán 1994; Otheguy & Zentella 2012), what Torres Cacoullos and Travis take as the starting point for their own quest is the need for a better contextualization of the bilingual data. They argue that this can only be attained through a more sociolinguistically balanced population of target speakers than what has been used in previous studies of the phenomenon, better data on bilingual practices as experienced by these speakers, and a more rigorous baseline to test and assess influences between languages in contact. The proposed research design aims to address all three issues via quantitative analyses of over 10,000 instances of the variable of interest (1st and 3rd person singular subjects) as distributed over a community-based bilingual speech corpus of New Mexican Spanish and English, plus three corpora documenting an earlier variety of New Mexican Spanish, monolingual Spanish, and monolingual English, respectively.

The overarching questions addressed in the book are: are grammars in contact different from grammars not in contact? (p. 11) and, more specifically, does code switching boost contact-induced change? Answers to these questions are sought through the variationist comparative method, which focuses on the study of the probabilistic constraints that condition speakers’ choices among competing variants in one and the same domain of linguistic encoding (the variable context). The variable context investigated in the book is the one pertaining to subject pronoun expression. The contents of the book are distributed over 11 chapters, which I summarize in Section 2.

2 Synopsis

Chapter 1 introduces the topics under study, research questions and methods.

Chapter 2 presents the community whose bilingual practices are investigated in the book. This is the Spanish-English bilingual community of northern New Mexico, where Spanish has been spoken for over 400 years and English for over 150, and thus the two languages have been in contact for at least 150 years. This qualifies New Mexican Spanish as an ideal object of study for testing hypotheses about contact-induced change, under the assumption that the longer the time of contact the higher the likelihood of convergence between the languages in contact. The Spanish-speaking community of the Santa Fe area was for centuries one of the most isolated areas within the whole Spanish speaking world (p. 14), and is the oldest truly bilingual community of the United States. Established as a Spanish settlement in 1693, first as a New Spain colony and later on as part of independent New Mexico (henceforth NM), the area became part of the US in 1848, but maintained its pronounced Hispanic identity. The Hispanic population of NM is still for the most part autochthonous (with only 16% of Spanish-speaking immigrants, mostly from Mexico). NM Spanish is generally described as retaining archaic features, which are otherwise lost in the standard, such as the archaic forms of the preterit of the verbs for ‘see’ and ‘bring’, vide and truj-, which coexist with the standard variants vi and traj- (p. 16). In addition, even though NM Spanish is historically related to Mexican Spanish, it has developed its own lexical, phonetic, and morphological features as, for instance, variable aspiration of syllable initial /s/, and the form ha, for the first person perfect of ‘to have’, corresponding to he in Mexican and standard Spanish. The acquisition of NM as a territory of the US severely affected the use of NM Spanish as a primary language of communication, especially after the imposition of English at school. This paved the way to a dramatic process of stigmatization culminating today, when the population of NM Spanish speakers has diminished and there is evidence for ongoing shift to English, especially in the southwestern part of the state. NM Spanish is also under pressure from other non-local Spanish varieties (Mexican Spanish or standard Spanish), which speakers perceive as more prestigious. In this context of general language endangerment, those bilingual speakers of NM Spanish and English who still use both languages in daily life and freely alternate between them are the target population of the study pursued in the book. The sample consists of 40 speakers, who are all third generation New Mexican Hispanics, and whose demographics vary in terms of age, sex, occupation, educational level, and place of residence. A first investigation of rates of subject pronoun expression across demographic features shows that there is no effect of social factors on the distribution of the linguistic variable of interest, which is in line with findings from other Spanish varieties (Silva-Corvalán 2017). The linguistic practices and identities of the sampled population are studied through a unique data set of bilingual speech, the New Mexican Spanish-English Bilingual corpus (henceforth NMSEB), whose structure is the topic of Chapter 3.

The book follows a “community-based” (p. 35) method of data collection, which is presented in Chapter 3. The method aims at capturing community-specific patterns in bilingual language use through a large corpus of spontaneous speech, the NMSEB corpus. The NMSEB corpus is based on data stemming from sociolinguistic interviews conducted by experienced fieldworkers who are themselves speakers of the bilingual variety which is being documented.^[1] The interviews comply to the “narrative of personal experience” format, which enhances the probability of the speakers’ involvement in the topics under discussion and is likely to trigger more naturalistic patterns of language use, including code switching. The interviews mostly consist of conversations with individual speakers or between pairs of speakers; some group discussions are also included. As a result of this data collection procedure, the corpus captures an impressive amount of spontaneous code switching, which occurs both within and across conversational turns, and yields an overall even distribution of Spanish and English sentences. The transcriptions of the recorded data presuppose at least five rounds of revisions for each recording, and are orthographically and prosodically-informed so that intonation units can be identified. This is essential to the study of code-switching, whose occurrence in bilingual discourse tends to be delimited by prosodic boundaries. Prosodically-informed transcriptions are also remarkably useful for syntactic analysis and, more specifically, for the study of subject pronoun expression given that subject pronouns and verbs tend to occur in the same intonation unit. Data are thoroughly anonymized in order to ensure confidentiality.

Chapter 4 characterizes the sociolinguistic profile and language attitudes of the community under study by means of content analyses of the recorded conversations. These show that such notions as language dominance or first and second language are less meaningful to the linguistic identity of bilingual speakers than suggested by proficiency tests and/or sociolinguistic questionnaires, traditionally employed as a means to characterize the demographics and history of bilingual communities. The impact of three independent measures of contact as inferred through the corpus – self-reported preference of English vs. Spanish, self-rating of English proficiency, and proportion of English clauses produced by individual speakers – on the rate of subject pronoun expression is assessed through independent statistical analyses. These, however, do not bring support to the hypothesis of contact-induced change. The remaining chapters put this hypothesis to the test through the in depth analysis of the bilingual speech documented in the corpus and against the many facets of subject pronoun expression in both Spanish and English.

Chapter 5 sets the stage for the NMSEB corpus analyses by reviewing and revisiting the conditioning factors for subject pronoun expression in (monolingual) Spanish. The chapter draws upon published work as well as the authors’ own analyses of the Corpus of Conversational Colombian Spanish and, to a lesser extent, the NMSEB corpus. The analyses show that while contrast, emphasis, and ambiguity resolution are most commonly argued to be the main functional factors that motivate subject pronoun expression in Spanish, speakers’ usage does not reflect these generalizations, which, in fact, account for a small portion of the dataset. The most robust effects documented by the authors are instead those engendered by accessibility/subject continuity and syntactic priming. The probability of pronominal subject expression is directly proportional to distance from previous mentions of the coreferential subject, and further increases if the intervening non-coreferential subjects are human, as in example (1), taken from chapter 5, p. 82 (example 66 in the reference text).

In (1), the first 3rd person subject in (a) has zero pronominal expression, but the second in (f), which directly follows the intervening non-coreferential human subject yo ‘I’, is expressed pronominally. These effects pertain to reference accessibility. In addition, pronominal subject expression is favored by previous mention in pronominal form, which is an effect related to syntactic priming. This is illustrated in example (2), which is taken from chapter 5 p. 89 (example 75 in the reference text).

In (2), previous pronominal expression favors subsequent pronominal expression. Subject continuity is enhanced by previous unexpressed subjects, which means that the two effects closely interact with each other.

The authors further show that clause linking disfavors subject pronoun expression, and that the strength of this effect is maximal in coordinated clauses, followed by subordinate clauses. In contrast, non-coordinated main clauses are more likely to favor pronominal subjects. Perfective aspect and dynamic verbs, that is, predicates encoding activities, are two additional factors that disfavor pronominal subjects, along with the quotative construction decir, ‘say’. Conversely, the first person form of the present tense of the verb for ‘believe’ (yo) creo highly favors subject expression across monolingual Spanish varieties. Finally, while 1sg subjects are likely to occur at greater distance between each other, 3sg pronominal subjects and unexpressed subjects tend to cluster closer together, as full noun phrases are more likely to be featured at greater distances. The probabilistic constraints on subject pronoun expression examined in the chapter all stem from “general cognitive mechanisms and wider discourse patterns”, which are not only specific to Spanish (p. 110). It is against this background, grounded in general and crosslinguistically common usage preferences, that comparisons across varieties, and in the light of divergent sociolinguistic scenarios, become meaningful.

Chapter 6 studies similarities and differences between Spanish and English in the domain of subject pronoun expression. The two languages are usually classified as representing two opposite types, with Spanish being defined as a typical null subject language and English as a typical non-null subject language. While unexpressed subjects occur in English, they are much more rare than in Spanish, and it is generally assumed that the probabilistic constraints that condition their distribution in English categorically differ from those operating in Spanish. Chapter 6 puts this assumption to the test by studying similarities and differences between the two languages based on observed patterns of use. A dynamic method of crosslinguistic comparison, whereby languages are compared with respect to how they vary in relationship to the usage distribution of the variable of interest rather than in terms of discrete structural types, is put forward under the label variationist typology. The case study is based on the analysis of 1st and 3rd person singular subjects as occurring in comparable conversational corpus data of Spanish and English, which are prosodically transcribed based on the same criteria as the ones used for the NMSEB corpus. While it is generally assumed that the preference for unexpressed subjects in VP coordination sets English apart from other languages, the study conducted by the authors shows that (1) VP coordination is not a discrete trigger of unexpressed subjects, but rather a gradient type of phenomenon which operates at its highest in the occurrence of prosodic and syntactic linking with the preceding coreferential subject-clause, and (2) this pattern of variation is not specific to English but also attested in Spanish. What is more English-specific is the fact that unexpressed subjects tend to always occur at the beginning of an intonation unit. In addition, the analyses show that while probabilistic constraints related to accessibility, coreferential subject priming, and verb class operate in both languages, the strength of these effects varies across languages, with accessibility (measured through prosodic and syntactic linking) being the strongest constraint in English. Finally, lexically-determined differences between the two languages are also found. In English, constructions of the type [Motion Verbiand Ø Verbi] (She goes and Ø gets the mother) and [Verbiand Ø Verb-of-speechi] (He called me last week, and Ø said he wanted to be with me) strongly favor unexpressed subjects. In Spanish, constructions of the type [(yo) + Cognition verb\textsc{1sf}] (yo no se ‘I don’t know’) strongly disfavor unexpressed subjects. By identifying differences in the probabilistic constraints that condition subject expression in English and Spanish, testable hypotheses about contact-induced change can be formulated. If subject pronoun expression in NM Spanish is changing under the influence of English, we should expect to observe: lower rates of expressed subject in prosodically initial positions (as in English), a stronger effect of accessibility (as in English), an overextension of unexpressed subjects in coordinated clauses linked by y ‘and’ (as in English) vis à vis a lower effect of constructions involving cognition verbs (which are common triggers of pronominal subjects in Spanish). These hypotheses are tested in Chapter 8.

Chapter 7, in turn, compares probabilistic constraints on the expression of subject pronouns in NM Spanish with an earlier stage of the variety, allegedly less exposed to contact with English, and with monolingual Spanish varieties. Data on earlier stages of NM Spanish are drawn from a sample of the oldest, and less prone to code switching, speakers of the New-Mexico-Colorado Spanish Survey (Bills & Vigil, 2008). Data on monolingual Spanish are taken from corpora of San Juan (Puerto Rico) and Madrid (Spain) use. Both case studies, which are based on a combination of multivariate analyses and cross-tabulations of the probabilistic constraints accounting for the distribution of subject pronoun expression, provide evidence “for continuity rather than change” (p. 159). With respect to the grammar of subject pronoun expression, present-day NM Spanish does not significantly differ from earlier stages of the variety nor from monolingual benchmark data.

Chapter 8 tests the hypothesis of convergence between bilingual Spanish and English, as formulated at the end of Chapter 6, through comparisons between four parallel datasets of prosodically annotated spontaneous speech: monolingual Spanish (through the Corpus of Conversational Colombian Spanish, CCCS), monolingual English (through the Santa Barbara Corpus of Spoken American English (SBCSAE), bilingual Spanish (through the NMSEB corpus), and bilingual English (also through the NMSEB corpus). The analyses presented in the chapter align with each other, showing no sign of convergence of bilingual Spanish and English along the three main diagnostic features established in chapter 6: prosodically initial position, accessibility, and language-specific verbal constructions. This pattern of negative evidence also holds for bilingual English in relationship with Spanish. The overarching finding of the chapter is thus that both bilingual varieties comply to their respective monolingual varieties in the distribution of expressed and unexpressed subjects. What remains to be tested is the hypothesis that code-switching boosts convergence effects in this domain. This is the topic of Chapter 9.

The NMSEB is ideal for testing synchronic hypotheses about the role of code switching in contact-induced change, as it abounds with instances of both single and multi-word incorporations from English. In Chapter 9, the authors test whether the presence of code switching in syntactic contexts which are maximally close to candidates for variable subject expression, and which are uttered by the same speakers, boosts the probability of occurrence of pronominal subjects. Once again, the results produce no evidence of convergence between the grammars in contact, in that they show no effect of proximity of English on the rate of Spanish pronominal subjects, and no changes in the probabilistic constraints that condition subject expression in bilingual discourse as compared with monolingual discourse. This is a remarkable result, which goes against the expectation that at least reference accessibility (the strongest predictor of unexpressed subject in English) could have been impacted by code switching, either in the direction of weakening (as an outcome of simplification of pragmatic constraints under language contact) or strengthening (in alignment with English).

“Speakers’ choices are not independent of their prior choices” (p. 189). Priming is a driving force in the structuring of discourse, which has been argued to also impact bilingual discourse, as parallel language structures across languages in use may boost convergence. Chapter 10 tests this hypothesis by investigating how coreferential subject priming, whereby pronominal subjects are favored in the presence of a coreferential pronominal subject in previous mention, changes in the presence of code switching. The study shows that the effect of coreferential subject priming and its interaction with accessibility do not vary in the presence of code switching as compared with its absence. What instead varies under the pressure of code switching are the contextual distributions of the variable of interest. In the absence of code switching, only 27% of the data occurs with a pronominal prime (mostly a Spanish pronoun). Conversely, in the context of proximate code switching, pronominal subject primes go up to 61%, and the majority of the primes are English pronouns. This means that while the rate of subject pronoun expression stays flat, the probability of within-language (Spanish pronoun-to-pronoun) priming decreases, and the distribution of the conditioning factors for the variable of interest in turn changes (in this case, the proportion of pronominal primes, which favor subsequent subject pronouns, increases). Ultimately then, the study shows that, even under code switching, the grammars of the languages used in bilingual discourse are kept distinct.

Chapter 11 provides some concluding remarks and evaluates the main methodological and theoretical contributions of the study against the two initial research questions (cf. Section 1). The results of the investigation show no sign of convergence between NM Spanish and English in the domain of subject expression (question 1, are grammars in contact different from grammars not in contact?), and they also indicate that code-switching does not boost contact-induced change in subject expression, but only affects its contextual distribution (question 2, does code switching boost contact-induced change?).

Contributions to typology

Bilingualism in the community is a ground-breaking study on language variation and change and a must-read for anyone interested in the internal and external dynamics of linguistic diversity. As I see it, this book makes two major contributions to linguistic typology, which I briefly summarize in the following and further discuss in Sections 3.1 and 3.2.

1. Defying the assumption that patterns of language structures can be readily captured into discrete types. The book promotes a methodology of crosslinguistic research, variationist typology, which instead aims to identify and assess the probabilistic contribution of a set of competing conditioning factors to the distribution of a given typological variable, and on the basis of the observation of language(s) in use.

2. Putting at stake common beliefs about language and population contact and, in particular, the idea that long-term bilingualism (always) leads to convergence between languages in contact. The book rejects this idea by demonstrating its lack of validity for the bilingual community under study. This is done through a series of related case studies, which rely on rigorous sampling of bilingual and monolingual speech data, and comparisons thereof, as well as on cutting-edge statistical methodologies.

While (1) has general implications for linguistic typology as a theory and method of crosslinguistic comparison, (2) specifically contributes to understanding the interplay between language structures and the sociolinguistic contexts in which languages are learned and used, which lies at the foundation of one of the most recent and fastest growing lines of research in our field, sociolinguistic typology.