What can the "attrition via acquisition" model predict?

Abstract

In their contribution, Hicks and Domínguez (2020, henceforth H&D), develop a unified formal explanation for language development that accounts for both language acquisition and language attrition. Indeed, they characterize language attrition as a potential outcome of the very same language acquisition process. This approach could seem too ambitious, but it is just logical if one wants to account for the dynamism of bilingualism. The idea that some sorts of language development – if not all – share a commonality is not new (Meisel, 2011; Sánchez, 2017; Schmid and Köpke, 2017), as the resemblances between language acquisition and language attrition have already been noticed, particularly in near-native contexts (Sorace, 1999, 2000a, 2000b, 2003). Drawing from already existing theoretical proposals such as the difference between input and intake (see, amongst others, Carroll, 2001; Corder, 2009; Lidz and Gagliardi, 2015; Putnam and Sánchez, 2013), the Feature Reassembly Hypothesis (Lardiere, 2008, 2009), and assuming an enduringly active Language Acquisition Device, H&D propose that a speaker of any state grammar may continue processing input for the purposes of acquisition as long as the speaker receives a new type of input. This conceptualization demotes the idea of an apparently stable native grammar in adult speakers –the traditional view within generative linguistics – which was already challenged in recent proposals such as those of Schmid and Köpke (2017), or Putnam and Sánchez (2013). The originality of H&D resides in their limiting of the malleability of adult native grammars, in an apt attempt to, first, acknowledge and, then, account for the paradox in first language (L1) attrition, i.e.: the low levels of real L1 attrition evidenced, notwithstanding the active mechanisms for language reformulation. In this sense, I consider this delimitation of the effects of L1 attrition an improvement with respect to the position defended by Schmid and Köpke (2017). H&D explain this paradox by proposing that speakers only re-evaluate their grammar in cases in which they receive and process input with new information, i.e.: structurally different from the L1, that generates some sort of language acquisition resulting in grammar restructuring. They further propose that this restructuring is in all instances overviewed by Universal Grammar, which is coherent if we assume a Full Access mode for L2 or Ln acquisition. In that case, we need to pose the same type of questions that generative second language acquisition research did in the 1980s and 1990s, namely: Is this type of acquisition that results in attrition a different type of L1 acquisition? If so, is it fundamentally similar or fundamentally different to typical L1 acquisition? Does the initial L1 have a special status with respect to the other later-acquired languages or varieties? Also, what type of evidence do we have or can we have in order to support the Full Access to UG? How can we know if the attrited speaker has gone beyond the L1 or L2 input and accessed UG in order to restructure the native grammar? Do we have any evidence of cases in which the resulting attrited variety does not come from the L1, the L2, or a mix of the two? Can this type of attrition via acquisition occur in a poverty of stimulus situation? Even though these seem to be issues already dealt with in the past, they are still relevant and worth considering when we propose a new type of grammar acquisition. Another issue that is worthwhile here concerns the applicability of this model to other, more established, long-term bilingualism or language contact situations. H&D claim that their model does not resort to attrition-specific mechanisms and that it is in principle extendable to ‘all language acquisition contexts’. Many times, the resulting process of language contact is the creation of hybrid grammars (Aboh, 2015), or grammars with constant, probably stable optionality, that oscillate between the parametric values of the L1, the L2, a mix of the two, or a resulting restructuring in both languages (Perpiñán, 2018; Sánchez, 2004). How exactly would the ‘attrition via acquisition’ account for these new language contact varieties? And, in particular, how does this model formally account for optionality, widely attested in L1 attrition and other near-native contexts (Sorace, 2000a, 2000b, 2004)? Do they propose unspecified features? Perhaps features that can be set in the same speaker in their positive and negative value? Maybe a fluctuation between two features or parameter settings (Ionin et al., 2008)? Accounting for variation or optionality poses difficulties to most formal approaches, particularly to categorical ones, so this problem is not unique to this model. Nonetheless, we cannot overlook it. In this article, the authors propose a formal model for language attrition based on a feature reassembly approach (Lardiere, 2008, 2009) and allege to exemplify their model with a linguistic phenomenon that they have studied extensively: the expression of null/overt pronominals. However, in no section of the article have they laid out a single formal matrix of features from the initial L1, the new input (the L2 or L1-variety in contact with the initial L1), or the resulting reassembled attrited L1. This is a crucial matter as the predictability of this model directly hinges on the formalization it proposes. H&D put forward that the continuing acquisition through exposure to new input may result in a ‘rearrangement of feature assemblies in the [L1] functional lexicon [that would] occur principally for aspects of the L1 grammar which share featural properties to a significant degree but where differences nevertheless obtain’ (H&D: 160). That is, the authors identify as the locus of L1 attrition the partial ‘alignment’ of features between the L1 and the L2 so that almost identical bundles of features that slightly differ in their assembly are susceptible to attrite, but the creation or disappearance of complete new categories do not seem to be predicted. This assumption appears to have implicitly adopted the overlap condition (Hulk and Müller, 2000; Müller and Hulk, 2001) and the structural overlap constraint resulting in syntactic optionality (Sorace, 2000a, 2000b). Part of the uniqueness that the H&D’s model claims is precisely the formalization of this overlap constraint and syntactic optionality. And even though this theoretical model is in principle testable, this article does not really show how to operationalize the model in practical terms and, as a consequence, does not really prove whether the model fully works. It would have been very helpful to see the precise applicability of the model with the data the authors discuss. I can only add that, based on my own research on Catalan–Spanish bilingualism, I see how the model would successfully explain the new distribution of DOM in bilingual Catalan (Perpiñán, 2018), since the two DOM systems overlap partially, or the acceptance of definite DPs in existential sentences (Perpiñán, 2015), but I cannot see how this model would provide an explanation for why some functional categories with morphological correspondence in one language but not in the other appear to be more vulnerable than others in the contact variety (Perpiñán, 2017). That is, this model seems to account well for the reconfiguration of overlapping features, but not for the creation or disappearance of categories.