2.1 Heritage Language Studies: An Evolving Field
Heritage Language (HL) Studies is an emerging area within linguistics, seeking to better understand how HL speakers differ both from monolingual and other bilingual speakers. The field developed from psycholinguists’ interest in language acquisition and its ability to inform our understanding of linguistic and mental structures. It developed primarily through experiments and controlled elicitation to collect data about what HL speakers know about their languages. More recently, interaction with sociolinguistic approaches has developed, allowing the incorporation of data collected in more ecologically valid contexts and analyzed using multivariate statistical methods. One motivation for this interaction is the belief that variation is an important part of grammar, as discussed in Nagy (Reference Nagy, Preston and Stanford2009b) and exemplified in Nagy’s (Reference Nagy2000a) “sociogrammar” of Faetar and Jauncey’s (Reference Jauncey1997) grammar of Tamambo. Including variable patterns in descriptive works, or recognizing “orderly homogeneity” (Weinreich et al., Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968, p. 100), means that the descriptions
are more accurate than traditional “static homogeneous models” (Bailey & Shuy, Reference Bailey and Shuy1973, p. xiv; Weinreich et al., Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968, p. 99);
may represent usages of marginalized groups (Nagy, Reference Nagy, Preston and Stanford2009b, p. 398);
reflect historical diversity in the language as well as social diversity among the speakers (Nagy, Reference Nagy, Preston and Stanford2009b, pp. 398–399);
may provide documentation of stylistic variation (Nagy, Reference Nagy, Preston and Stanford2009b, p. 406);
help us understand the range and limitations of formal grammars (Barbiers, Reference Barbiers, Cornips and Corrigan2005, p. 253; Chambers, Reference Chambers2000; Cheshire, Reference Cheshire and Corrigan2005, p. 19; Hale, Reference Hale, Grenoble and Whaley1998, p. 194; Wilson & Henry, Reference Wilson and Henry1998, p. 1);
allow authors to avoid (or at least examine) value judgments about whose speech and which styles to represent (Nagy, Reference Nagy, Preston and Stanford2009b, p. 411); and
represent more types of knowledge (Goebl et al., Reference Goebl, Nelde, Starý and Wölck1996).
This broader reach may help support speakers’ attitudes toward languages, particularly important if acquisition is at issue. Despite recent efforts, coordination between acquisition and pedagogy for HLs remains limited. For discussion of this issue, please see Kisselev et al., (Reference Kisselev, Dubinina and Polinsky2020) and Laleko and Dubinina (Reference Laleko, Dubinina, Kresin and Bauckus2018).
2.2 The Experimental and the Comparative Variationist Approaches
In addition to pedagogical research, two distinct approaches for analyzing the structure of HLs exist. One, the experimental approach, is based in language acquisition and psycholinguistic methods and the other, the comparative variationist approach, is based in variationist sociolinguistics. While both have generated a wealth of data, their findings are often disparate, even contradictory, regarding the question of whether and when convergence between bilinguals’ two languages is inevitable – see the discussion in Torres Cacoullos and Travis (Reference Torres Cacoullos and Travis2015, p. 365).
2.2.1 The Experimental Approach
The first approach, the experimental approach, is based in experimental psycholinguistic methods (cf. Benmamoun et al., Reference Benmamoun, Montrul and Polinsky2010). The experimental approach seeks to better understand how languages are acquired and evolve in contact situations. Interest lies in the evolving behaviour of both speakers and languages, addressing these questions:
(1) What is the developmental course of HL acquisition?
(2) Do heritage speakers follow the developmental trajectory of monolingual acquisition, of second language learners, or does their acquisition follow a different course?
(3) What language-internal and language-external factors contribute to variation in the acquisition and use of HLs?
This approach seeks to determine areas of strength and vulnerability in HLs by contrasting them to “baseline” varieties that are presumed to be the target of HL speakers. Identifying and accounting for such areas informs our general understanding of natural language, so this approach crucially intersects with theoretical linguistics.
Common experimental methods of data collection include elicitation, usually of short, decontextualized utterances; grammaticality judgment tasks; repetition tasks that are not communicatively oriented; descriptions of pictures or videos and prompted story-retelling tasks that elicit connected speech. Comparison is made between groups of heritage-language speakers and either monolinguals or second-language learners (sometimes both). Where differences are found, they are interpreted as areas of vulnerability for HLs.
2.2.2 The Comparative Variationist Approach
The second approach, comparative variationist sociolinguistics (cf. Labov, Reference Labov, Baugh and Scherzer1984; Poplack & Levey, Reference Poplack, Levey, Auer and Schmidt2010), quantifies the effects of contextual forces on the selection of linguistic forms in naturalistic/conversational speech, aiming to account for both categorical and variable patterns within the same constructs (e.g., rules, constraints) that are used to describe a mental grammar.
This approach also compares groups of speakers for evidence of change, expecting to find variability in all groups of speakers, homeland and heritage alike, and seeking similarities and differences in the patterns of variation, rather than just the choice of forms produced. Comparisons across generations, between HLs and English, and between heritage and homeland varieties, allow innovations to be identified and the source(s) of the innovations identified. This, in turn, contributes to our understanding of how languages evolve. In particular, this method, which prioritizes analysis of naturalistic, vernacular speech (Labov, Reference Labov1972b), allows us to grapple with the constraints on change, the phases of language change, the social and linguistic embedding of a language change, how speakers evaluate linguistic variation, and the factors that actuate change (Weinreich et al., Reference Weinreich, Labov, Herzog, Lehmann and Malkiel1968, pp. 100–102) as we work to best describe and understand the systematic nature of linguistic evolution.
These concerns have been re-expressed by Hilpert & Gries (Reference Hilpert, Gries, Kytö and Pahta2016, p. 37) as follows:
(1) When and how does a given change happen?
(2) Can a process of change be broken down into separate phases?
(3) Do formal and functional characteristics of a linguistic form change in lockstep or independently from one another?
(4) What are the factors that drive a change, what is their relative importance, and how do they change over time?
(5) How do cases of language variation in the past compare to variation in the present?
Variationist sociolinguists have devoted relatively little attention to the description of variable linguistic structure in minority languages (Meyerhoff & Nagy, Reference Meyerhoff and Nagy2019; Nagy & Meyerhoff, Reference Nagy, Meyerhoff, Meyerhoff and Nagy2008), but if we accept that language variation is a human phenomenon, then studying changes in smaller, less well-described languages should contribute equally to the enterprise as larger, better documented languages. In a similar vein, varieties spoken by mobile groups, such as communities of people who recently immigrated, have been overlooked in favour of so-called NORMS (non-mobile, older, rural male speakers). Yet
mobility has become such a central feature of human existence in the age of globalization that any kind of linguistics that is not able to address its effects will be in danger of falling out of step with reality.
Heritage speakers are, by definition, part of this overlooked group, although immigrants and their children feature prominently in HL studies. The Heritage Language Variation and Change in Toronto (HLVC) project is not the only generator of HL studies applying variationist sociolinguistic methods. Recent collections that include variationist studies of minority or HLs include Aalberse et al. (Reference Aalberse, Backus and Muysken2019); Bayley et al. (Reference Bayley, Preston and Li2022); de Leeuw and Celata (Reference de Leeuw and Celata2019); Flores et al. (Reference Flores, Gürel and Putnam2020); Guijarro-Fuentes and Schmitz (Reference Guijarro-Fuentes and Schmitz2015); Halmari and Backus (Reference Halmari and Backus2020); Hildebrandt et al. (Reference Hildebrandt, Jany and Silva2017); Kasstan and Nagy (Reference Kasstan and Nagy2018); Mayr and Morris (Reference Mayr, Morris, Mayr and Morris2021); Polinsky and Scontras (Reference Polinsky and Scontras2020a and Reference Polinsky and Scontras2020b); Rao (Reference Raoin press); Schmid and Köpke (Reference Schmid and Köpke2017); Shea and Kim (Reference Kim2021), and a special issue of Spanish in Context, 13.1. Additional recent studies of variation in lesser-studied languages include Davidson (Reference Davidson2015), Mooney (Reference Mooney2019), Morris (Reference Morris2013, Reference Morris2017), Muxika-Loitzate (Reference Muxika-Loitzate2017, Reference Muxika-Loitzate2020), Nance (Reference Nance2013, Reference Nance2014), and Strandberg (Reference Strandberg2022).
However, the quantity of heritage-language research in the variationist sociolinguistic framework remains small, and currently we do not know exactly what accounts for the contradictory findings between experimental and variationist approaches. The next sections consider the differences.
2.2.3 Differences in Methods
The goal of the research presented in this book, then, is to bring together the object of study of experimental, psycholinguistic heritage-language studies with the methods of comparative variationist sociolinguistics. As noted in Section 2.2, this enterprise has revealed differences in outcomes. Thus, in this next section, we delve into the methodological differences that might account for these differences.
2.2.3.1 Participant Differences Based on Definitions of Heritage Speakers
A first important difference relates to the concept of a baseline. Under the variationist approach, the grammar of any language variety (e.g., heritage versus homeland, Gen1 versus Gen2) is considered a complete system that stands on its own. Comparisons between systems (e.g., between generations or between heritage and homeland varieties) are based on identical methods (described in Chapter 4) to identify innovations and track the contexts that favour these innovative forms.
Because of their interest in how HL-speakers diverge from other speakers, psycholinguistically oriented research often excludes HL-speakers who are “too fluent,” that is, exhibiting linguistic behaviour that is deemed too similar to that of monolinguals. The categorical exclusion of speakers based on comparison with any standard or other group risks excluding innovative speakers (or contexts favouring innovations), exactly the type of data that is the target of sociolinguistic investigations of language change. Yet the experimental approach often does exactly this, comparing HL data to an idealized “standard homeland” variety, although this trend is decreasing in recent work. For instance, (Rothman, Reference Rothman2009, p. 156) defines HLs as follows:
A language qualifies as a heritage language if it is a language spoken at home or otherwise readily available to young children, and crucially this language is not a dominant language of the larger (national) society…
However, this definition continues:
There is the possibility that quantitative and qualitative differences in heritage language input, influence of the societal majority language and differences in literacy and formal education can result in what on the surface seems to be arrested development.
This leads us directly to a second difference in approaches: who may be included, based on the researchers’ definition of a HL speaker. In the experimental approach, the definition often relates to performance. For example, Polinsky (Reference Polinsky2011, p. 2) states that
Heritage languages are spoken by early bilinguals […] whose L1 (home language) is severely restricted because of insufficient input. […] they can understand the home language and may speak it to some degree but feel more at ease in the dominant language of their society.
Similarly, Montrul (Reference Montrul2009, p. 241) notes that “many aspects of grammar may not reach full development and remain incompletely acquired.” Kagan (Reference Kagan2007) notes that features of HLs include limited vocabulary, incomplete morphology, impoverished syntax, spotty socio-cultural knowledge, and registers that are not fully developed, while Benmamoun et al. (Reference Benmamoun, Montrul and Polinsky2010, pp. 10–11) note that HL speakers are “more likely to speak English … [whose] comfort in English increases … often at the expense of the home language,” following Cho et al. (Reference Cho, Cho and Tse1997), and that “the crucial criterion is that the heritage language was … not completely acquired.” While Benmamoun et al. (Reference Benmamoun, Montrul and Polinsky2010) acknowledge that HL speakers may be as proficient as monolinguals, the research focus is on HL speakers who are not.
Not all experimental studies of HLs employ these types of deficit-based definitions of HL speakers. Exceptions, in which participant selection is determined by autobiographical characteristics, such as when each language was acquired and when/how often they are used, include Cuza & Frank (Reference Cuza, Frank and Heijl2010), Cuza et al. (Reference Cuza, Pérez-Leroux and Sánchez2013, p. 14), and Liceras and Senn (Reference Liceras and Senn2009, p. 42). In such approaches, a HL speaker is one who acquired a first language that is not the socially dominant language in a given geographical area.” They still stipulate (ibid.) that it is not the speaker’s dominant language.
In the variationist approach, on the other hand, HL speakers are defined strictly by their linguistic autobiography and not by their linguistic performance. For the HLVC Project, this follows from the Canadian government definition of a HL as a mother tongue that is neither an official language nor an indigenous language (Cummins, Reference Cummins2005; Harrison, Reference Harrison2000).
Definitional differences of who is considered a HL speaker determine the population to be investigated, forcing a difference between methods. However, methodological issues also impact the participant samples selected by these approaches. The participants who produced the speech that constitutes the HLVC data agreed that they were willing to talk in their HL for an hour, as a condition of inclusion. The degree of fluency and confidence necessary to assert this about oneself may produce a very different participant sample than one that comes from a language-learning classroom where all students are invited to participate in an experiment, let alone a sample based on excluding those who are “too fluent.”
2.2.3.2 Population Differences
Studies may sample from different populations. Most experimental HL studies have been conducted in the United States and in Germany, while these sociolinguistic studies using spontaneous speech were conducted in Toronto, Canada. While some experimental work has also been conducted in Canada, we are not aware of variationist work conducted in the US with heritage speakers except for Spanish speakers (Flores-Ferrán, Reference Flores-Ferrán2004; Geeslin & Gudmestad, Reference Geeslin and Gudmestad2016; Lapidus & Otheguy, Reference Lapidus and Otheguy2005; Raña Risso, Reference Raña Risso2010; Schwenter, Reference Schwenter and Campos2011; Shin, Reference Shin2014; Silva-Corvalán, Reference Silva-Corvalán1994; Torres Cacoullos & Travis, Reference Torres Cacoullos, Travis, Rivera-Mills and Villa2010). There are important political differences in the acceptability of multilingualism and of immigrant populations in Toronto versus the United States. Over time, this may well affect how people speak their HLs, so comparison across populations from different countries must be undertaken with care.
A further difference in the populations sampled is that experimental studies have often included children and adolescents, while the HLVC project primarily targets adults (though it includes a handful of adolescents). Additionally, experimental studies often recruit from language classrooms and are thus limited to speakers with university classroom exposure, as well as overrepresenting participants in their late teens and early twenties. This age restriction was avoided in the HLVC project by casting a wider net. Although we recruited primarily through personal networks of HL-speaking university students, they were encouraged to locate speakers across ages, professions, and geographic space (within the Greater Toronto Area).
2.2.3.3 Task Differences
The tasks that HL speakers are asked to complete also differ. Experimental approaches often include school assignment- or school test-like activities such as judging the grammaticality of a set of sentences as an organized task, manipulating or describing pictures, and controlled elicitation activities where the focus is not on the communication of information. Such tasks allow maximal comparability across responses and target the very specific parts of the language that are considered relevant.
In contrast, the HLVC Project uses standard Labovian sociolinguistic methods (as described in Labov, Reference Labov, Baugh and Scherzer1984) to collect conversational speech in which topics of mutual interest to both the interviewee and interviewer are discussed. This is followed by an informally administered oral questionnaire to collect information about speakers’ language learning experiences, usage patterns and preferences, and cultural preferences related to their ethnicity. A brief-picture description task is also administered, which allows for testing some hypotheses about the effects of free speech versus semi-controlled elicitation and produced more similar vocabulary and syntactic structures across speakers (see Methods in Chapter 4).
There are differences also in the ecology of the two approaches, that is, a different physical place where data is collected. HLVC data is collected in speakers’ homes, cultural centers, clubs, or cafes frequented by speakers of the HL, that is, spaces where participants are accustomed to using their HL. In contrast, experimental studies are often conducted in a classroom or in the research laboratory at a university. Heritage-language speakers have spent their whole lives (at least post-migration, in the case of Gen1 speakers) being told that their HLs do not belong in the classroom. Thus, it is plausible that when we ask people to participate in a test–like activity in a place where they have never felt welcome to use their HL, it influences their performance, perhaps making the tasks more difficult. This would be exacerbated in any instances where the experimenter does not speak the HL, further diminishing the sense of a shared communicative goal during the data-collection activities. Such classroom environments may also be contexts that make HLs more subject to influence from English, that is, triggering the use of more English-like variants, as English is the prescribed language of the space. This relates to standard language ideology, that there is a standard language variety that is appropriate to particular places, such as schools.
To summarize, task differences may include:
(1) comprehension/judgment/manipulation tasks versus open-ended conversational production;
(2) interactions among speakers with different native varieties versus speakers with the same native variety;
(3) salience of the “test-taking” laboratory setting with inherent concepts of right and wrong answers versus conversation with communicative intent, in a familiar setting, in which the interviewers are trained to make efforts to create a casual setting (e.g., how they speak, how they dress, body language, location);
(4) requirement to choose an answer, structure, or form versus options to avoid a particular structure that exists in spontaneous speech;
(5) unfamiliar experimental tasks versus typical everyday conversational behavior.
There is one other way that data collection methods matter: In conversational data, we expect bigger effect sizes than in reading task data. That is, Labov (Reference Labov1972a), among others, demonstrated that we find greater effect sizes in less monitored speech styles.
2.2.3.4 Analytic Differences
In addition to differences in how the data is collected, we acknowledge differences in how it is analyzed. Experimental design requires a focus on a single structure at a time and normally on the effect of one predictor at a time. (Owing to the strictly controlled nature of tasks, other factors are controlled for.) In contrast, the speaker sample design and the sample of all relevant utterances (tokens within the envelope of variation) in a variationist analysis require multivariate analysis in order to look at interactions and overlaps across multiple factors varying simultaneously, as well as data manipulations to control for multicollinearity. In addition to examining genre or style effects, as alluded to in Section 2.2.3.3, differences in relevant linguistic context (e.g., phonological, morphological, syntactic, pragmatic) in which tokens appear, differences across speakers (e.g., generation, sex, ethnic orientation) and, in some analyses, the language must be entered as factors in order to accurately determine the contexts in which the innovative variant is favoured over the conservative one.
Owing to the labour-intensive nature of the data collection process, it is fortunate that the variationist approach, by collecting untargeted conversational speech samples, can use the same samples to examine many different linguistic variables from the same speakers collected under the same conditions. This ensures greater inter-speaker and inter-group comparability than is possible for many experimental tasks, where speakers are recruited for a single or small set of tasks.
2.3 Looking Forward: Can We Determine Which Differences in Methods Account for the Differences in Outcomes?
Chapters 5–8 illustrate several differences in the outcomes reported by these contrasting approaches. Because of the co-occurrence of population, sample, task, and analytic differences between the two methods, we are not able to disentangle their effects at this time. Ongoing work in which we collect data via both methods from the same participants may help disentangle this issue. This includes work initiated within the HLVC project but also elsewhere. For example, Maria Polinsky has worked primarily in the experimental paradigm, but has recently been overseeing the construction of corpora of connected speech in heritage varieties of Bulgarian, English, Japanese, Korean, Mandarin, Russian, and Spanish (Rakhilina et al., Reference Rakhilina, Vyrenkova and Polinsky2016). The speech is not conversational, but rather elicited by requesting speakers to tell stories based on a series of pictures or video prompts, and is produced in a laboratory setting, which is less natural than typical contexts in which variationist sociolinguists collect data. A similar project by Heike Wiese in Germany incorporates two speech styles but relies on video clips for elicitation in Heritage German, Greek, Russian, and Turkish (Wiese, Reference Wiese and Ramisch2020). Other recent efforts are also aiming to bridge gaps discussed in this chapter. Peter Muysken and Suzanne Aalberse’s group, for example, examines a broad range of HLs via a range of methods (cf. Aalberse et al., Reference Aalberse, Backus and Muysken2019).
