A TIME TO REFLECT ON OUR TERMINOLOGY

Recently, criticism of certain choices of terminology has arisen in the area of heritage language acquisition. The main criticism that concerns us here can be found in Otheguy (Reference Otheguy, Tortora, den Dikken, Montoya and O’Neill2016), Pascual y Cabo and Rothman (Reference Pascual y Cabo and Rothman2012), and, especially, Kupisch and Rothman (Reference Kupisch and Rothman2018). These authors warn against the use of the term incomplete acquisition when describing the process of language acquisition experienced by heritage speakers, broadly defined as bilingual speakers of an ethnic or immigrant minority language. Incomplete acquisition is now widely used in our field to account for the possibility that some property of the target grammar fails to be acquired due to insufficient input (Montrul, Reference Montrul2002, Reference Montrul2008, Reference Montrul2011, Reference Montrul2016; Polinsky, Reference Polinsky2006, Reference Polinsky2008; Silva-Corvalán, Reference Silva-Corvalán2018a, Reference Silva-Corvalán, Lapidus Shin and Erker2018b).^{Footnote 1} Broadly speaking, the term incomplete acquisition has come to describe a process of heritage language acquisition (distinct from attrition) that is often characterized by grammatical reductions, simplifications, and reanalyses (see overview in Benmamoun, Montrul, & Polinsky, Reference Benmamoun, Montrul and Polinsky2013a).^{Footnote 2} For instance, Spanish-English bilinguals in the United States have been found to exhibit a reduced tense-aspect-mood verbal system, omit object clitics, and realize overt subject pronouns with different distributions (Silva-Corvalán, Reference Silva-Corvalán2018a); Polinsky (Reference Polinsky2008) argues that Russian-English bilinguals develop a reanalyzed gender-marking system that is different from monolingual Russian.

In our view, incomplete acquisition does not necessarily have to describe the outcome of bilingual acquisition. That is, a grammar of a heritage speaker could be incomplete due to the acquisition conditions that are specific to that speaker, regardless of whether it resembles the grammars of their parents. Our assumption is that incomplete acquisition can arise when input conditions change during the course of acquisition (and, crucially, before a mature and stable grammar has fully developed) and that change impacts upon the overall pattern of acquisition and resulting grammar.

Although it may appear that the outcome of incomplete acquisition is a simplified or reduced version of the full grammar,^{Footnote 3} heritage speakers' grammars are nevertheless regarded as coherent in their own right (Polinsky, Reference Polinsky2008, p. 41) and obey the same principles of language as any other grammatical system (Montrul, Reference Montrul2008, p. 167). These authors have nothing to say about the (in)ability of heritage speakers to function as bilingual speakers or about any other issue that is not strictly linguistic (see also Silva-Corvalán, Reference Silva-Corvalán, Lapidus Shin and Erker2018b). This is important because, when discussing the inappropriateness of using the term incomplete acquisition, Kupisch and Rothman ask:

It seems to us that this line of criticism is based on the mistaken assumption that terms such as deficient, attrited, or incomplete may be taken to represent an evaluation of the linguistic abilities of a bilingual speaker as a person. Kupisch and Rothman question the theoretical appropriateness of incomplete acquisition as a hypothesis to explain heritage speakers' acquisition, but their criticism seems to focus on the appropriateness of the term as well as its potential offensive connotations (Kupisch & Rothman, Reference Kupisch and Rothman2018, p. 567). In our view, this criticism seems to be about beliefs and ideology (i.e., what the bilingual speaker should or should not be able to do). Yet given that generative approaches to second language acquisition (SLA) and bilingualism are only interested in the grammatical representations in a speaker’s mind, they do not (and indeed, cannot) make evaluations of speakers. However, heritage language acquisition researchers clearly do have a duty to address the terminology problem highlighted by Kupisch and Rothman, with three main considerations making this matter currently relevant. Firstly, in the United States, the language proficiencies of heritage speakers are often evaluated formally on an institutional basis. Secondly, unlike instructed second language learners, heritage speakers often face issues of identity and belonging, and this is particularly critical when learning their minority language in a classroom (Carreira & Kagan, Reference Carreira and Kagan2011). For instance, in the United States many Spanish heritage speakers find it difficult to reconcile negative attitudes and beliefs of their own variety of Spanish when compared to the standardized version of Spanish as a prestigious world language (Carreira, Reference Carreira and Pascualy Cabo2016, p. 164). Thirdly, pressure to make our results impactful to society means that communication of scientific results to a general public is flourishing. Bilingual speakers, as well as other observers outside our field, are discovering our research and without the relevant grounding in the methodological and terminological conventions adopted in the field, may mistakenly interpret our findings as either evaluation or a judgment of failure.

By extension, bilingual educational programs may also be misinformed by this terminology. In these respects, Kupisch and Rothman correctly claim that: “Once offered to the public domain, terminology can have far reaching and long-lasting effects, even—perhaps especially—when these are unintended by their original promoters” (Kupisch & Rothman, Reference Kupisch and Rothman2018, p. 574). Likewise, Rothman, Tsimpli, and Pascual y Cabo (2016) argue that because theoretical discussions on heritage speakers should be used to inform educational debates concerning these speakers, referring to them as incomplete learners is detrimental.^{Footnote 4}

This article is organized around the arguments for the two opposing positions regarding terminology in L2 acquisition and heritage speaker research, which we identify in response to the preceding criticism:

1. 1. As a scientific enterprise, L2/heritage speaker research should apply whatever the most descriptively transparent terms are to describe and explain the object of our study: if any persons or communities feel offended, threatened, or further marginalized by our terminology, then this is a misconstrual or misinterpretation on their part.

2. 2. The risk of even indirect offense and marginalization should be avoided at all costs: L2/heritage speaker research should avoid any terms that have the potential to do that, even if only through misconstrual or misinterpretation.

We then identify and argue for an “intermediate” third position in this debate that we hope offers a guideline for minimizing any negative impacts as far as possible, while prioritizing the effective communication of our research assumptions, methods, and findings. Specifically, in reaching this position we make the following points:

1. 1. Ostensibly “negative” terms exist in all acquisition contexts regardless of socioeconomic status, education, or other sociolinguistic variables. They do not constitute a judgment or evaluation of the speaker; to do so would be to misunderstand or misrepresent the discipline.

2. 2. We need to be mindful and clarify that heritage speakers are not “incomplete” bilinguals, but some aspect(s) of their grammars can be, although this is not the only outcome. More scrutiny on the sources of problems for these speakers is necessary by researchers.

3. 3. We need to communicate our findings to heritage speakers and external bodies in a way that is positive and easy to understand in order to avoid misinterpretation, misconstrual, and even mistrust of our research.

THE OBJECT OF OUR STUDY

Generative linguistic research provides principled scientific explanations to questions regarding the nature of linguistic representations and how these are acquired (see details in Roberts, Reference Roberts2017). It is interested in examining the acquisition and development of grammars, or Internalized-language (I-language): the internal, individual, and implicit knowledge of the grammar represented in the mind of a speaker (similar, although not completely identical, to linguistic competence). Following one of the fundamental scientific developments of the Chomskian approach to the study of language, I-language is formally divorced from Externalized-language (E-language), which covers aspects of language use. This distinction does not serve to set adjacent disciplines against each other or establish I-language as inherently worthier or more suitable for study, but it does formalize and emphasize a key distinction in the object of study for research into language structure and research into language use. This is significant as it may not be immediately apparent to (even interested) observers of our field because:

1. 1. Even though our research focuses on I-language properties, just like research in formal linguistic theory, we can only access these using inference from E-language properties, either in naturalistic or experimental conditions.

2. 2. In instructional contexts, learners’ E-language typically is formally evaluated in a way that native speakers’ E-language typically is not.

The distinction shown in the preceding text delimits the research agenda within the field of generative SLA, underpins our research questions, and shapes the kinds of explanations we seek.

Generative SLA is often criticized because of its native (or monolingual) bias, namely that it tends to evaluate L2 learners’ grammars against the grammatical knowledge of native speakers (and sometimes only educated ones, at that).^{Footnote 5} Critics of the field have often taken the native versus nonnative comparison to imply that generative researchers have expectations that L2 learners should aspire to behave like monolingual speakers, when this is clearly not the case. As clarified previously, in generative SLA research, the object of study is not the learner but the learner’s mental grammar, that is, a system of tacit internalized linguistic representations, decoupled from contextual considerations, acquired using an inferential learning procedure under exposure to external linguistic input. Generative research into L2 grammars does not, then, concern itself either with the sociocultural context surrounding the act of learning or using a language, nor with the personal characteristics of its users (such as status, personality, education, and motivation). For this reason, the description of learner grammars makes no implicit or explicit evaluation of such characteristics, and indeed has nothing to say about whether learners should aim to become nativelike.

Under the preceding working assumptions of our field, the comparison between second language learners and native speakers is warranted by the scientific method we use. Any scientific study of some well-defined properties of a target population or experimental group (in our case, the mental grammars of second language learners) is likely to measure those properties against those of a control group that serves to establish a baseline. In language acquisition, this is necessary to confirm or refute hypotheses stemming from theoretical considerations and based on different variables, to test whether interventions can alter linguistic knowledge or behavior, to test the validity of a test or task, and so on. The control group typically consists of adult native speakers not because of any inherent value we assign to those speakers, but because their grammars are scientifically useful: typically they are relatively stable, broadly homogeneous within a given variety of a language (and in many respects even across them), and often well documented within linguistic theory concerning the formal grammatical mechanisms underlying key properties (such as verbal inflection, wh-constructions, and the binding conditions).

This valid and necessary research method of comparison against the control group, alongside some of the key findings of generative second language research, is borne out in some of the terminology we use to describe nonnative grammars. These have sometimes been taken to be controversial, particularly—though not exclusively—by those outside our immediate research field. Yet a grammar described as “deficient” cannot be equated with a speaker who is somehow deficient. By the same token, we argue that a grammatical representation described as “incomplete” in our field does not entail that the person is somehow an incomplete learner. It simply describes the fact that the grammar does not share the features or properties of the input of the full grammar (Rothman, Reference Rothman2007; White, Reference White2003). On this issue, White (Reference White2003) argues that:

These considerations constitute the thrust of the defense of position 1 as outlined in “A Time to Reflect on Our Terminology.” We next consider the fallacy of position 2.

INCOMPLETE GRAMMARS EXIST IN ALL ACQUISITION CONTEXTS

WHY “DIFFERENT” IS NOT ENOUGH

Both Montrul (Reference Montrul2016) and Polinsky (Reference Polinsky2018b) have recently advocated avoiding the term incomplete acquisition altogether, as it can “unintentionally lead to a negative interpretation and portrayal of the ethnic minorities who speak these languages” (Montrul, Reference Montrul2016, p. 225).^{Footnote 13} This is also reflected in Benmamoun, Montrul, and Polinsky (Reference Benmamoun, Montrul and Polinsky2013b) when the authors admit that although they think that incomplete acquisition is scientifically valid, it would be better not to use it. For us, the important question is whether there is another term that can be an adequate substitution.

Benmamoun et al. (Reference Benmamoun, Montrul and Polinsky2013b) and Montrul (Reference Montrul2016) argue that divergent is a better choice than incomplete when referring to the grammars of heritage speakers, as it is more neutral and does not carry the same negative connotations. For their part, Kupisch and Rothman (2018, p.16) propose using differential acquisition instead of incomplete acquisition. Similarly, Scontras, Fuchs, and Polinsky (Reference Scontras, Fuchs and Polinsky2015) have argued in favor of using divergent attainment instead of incomplete acquisition. We understand that by divergent these authors mean different, which is in line with Kupisch and Rothman’s line of argument.

We see some potential problems with these three proposals (see also Silva-Corvalán, Reference Silva-Corvalán2018a). One issue is that divergent means different indeed, but different and absent (incomplete) are two different things (see Sorace, Reference Sorace1993). If a feature is not acquired, then it is not completely appropriate to say that the grammar is divergent or different, as the term does not provide an accurate description of the phenomenon. If Kupisch and Rothman think that incomplete acquisition should not be used because it is theoretically inaccurate, then the same should be said about different, which lacks explanatory import and is descriptively nebulous in that it does not provide an accurate depiction of the phenomenon. In our view, the term different is not an effective substitute for incomplete because a grammar that is different from another might be incomplete or complete (see discussion in “Heritage Language Grammars”). In other words, we argue against adopting a position that approaches position 2 on the grounds that adopting a more “neutral” but descriptively diluted term ultimately risks downplaying key insights of generative research in identifying and accounting for grammatical properties of speakers in monolingual, bilingual, and L2 contexts, both in adults and children.

Nevertheless, the question remains to be addressed: Is it legitimate to still use terms such as incomplete or attrited? The preceding discussion presents a problem in that scientific considerations lead us to position 1 over position 2, while we must recognize that using terms that may appear to be outwardly negative or judgmental to a certain extent invites misconstrual and misinterpretation of our field. Moreover, there may be potential for unintended and unjustified negative impact upon the very people whose language we want to study and understand. If the alternative is the descriptively inferior characterization of bilingual/multilingual grammars as simply “different” while we know that there are alternatives that do a better job, then it is surely the right time to have an open discussion about the grounds on which we make that terminological shift.^{Footnote 14}

A possible resolution may lie in the intermediate position, as follows:

This position concedes that we as researchers have a duty to be aware of potential indirect effects of (misinterpretation of) our terminology, and minimize this as far as possible by avoiding them, if there are alternatives that will do just as well or better. For instance, we invite scholars to consider the potential negative impact of using incomplete in titles and headlines whose exact scientific meaning can easily be misinterpreted. It does not mean, however, that we dilute our theoretical vocabulary and put this ahead of our scientific objectives. The benefit is that even the more “pragmatic” position 3 remains consistent with the matters raised in “The Object of Our Study” (concerning the object of study in the field being mental grammars and not speakers) and in “Why ‘Different’ Is Not Enough” (concerning the risk of adopting descriptively inferior terms). The difference is that this position, rather than position 1, would entail a degree of comparison of some of the terms available. This could be done more or less formally, and our aim in this article is only to discuss this in the case of the term incomplete, as we have in “Incomplete Grammars Exist in All Acquisition Contexts,” where it was argued that incomplete is more accurate than the alternatives proposed in the recent literature. However, it is then necessary to reflect further on other related terms, such as different, divergent, and defective, and consider whether the contexts in which they are applied typically meet the guideline for use that we propose in position 3.

CONCLUDING REMARKS

One question seems very relevant right now: How do we describe the process and outcome of heritage language acquisition? Do we need to revise the terminology we use in our field? Based on the evidence reviewed in this piece our answer is no, as the term incomplete acquisition can be legitimized by means of the theoretical model we assume. We have argued that describing some aspect of L1, L2, and bilingual grammars as incomplete or deficient is necessary to fully represent our understanding of the language acquisition process. The terminology used is not—nor can it be construed as being—the basis for any judgment on the speakers. Furthermore, replacing the existing terminology with imprecise or nebulous terms will not help us get close to characterizing how language acquisition takes place across different contexts and in varying conditions.

Because both complete acquisition based on attrited input and incomplete acquisition are possible for heritage speakers (scenarios 2 and 4 in “Heritage Language Grammars”), researchers would need to employ more sophisticated research methods to discern between these two possibilities. Research investigating the grammatical knowledge of heritage speakers would benefit from including at least two control groups, a group of monolingual speakers and a group of speakers representing the parental/community input that the heritage speakers receive in that context (see e.g., Montrul, Bhatt, & Girju, Reference Montrul, Bhatt and Girju2015). This could help clarify whether properties not instantiated in heritage speakers’ grammars were present in the input to start with.

Finally, our aim with this article is to contribute to an important discussion on the choice of some terminology in our field, which Kupisch and Rothman have laid out. We hope that our arguments stimulate more fruitful discussions on this very relevant issue in our field.