«Learning a Gendered Language: L2 Acquisition and Relative Processing Costs of Spanish Grammatical Gender and Number Alexander M. Hirsch Yale ...»
In fact, Michael Ullman made a similar distinction over a decade ago: Ullman (2001) separates “words,” associative memories that are subserved by the declarative memory system based in the temporal lobe, and “rules,” combinatorial systems subserved by procedural memory and based in the frontal lobe. If this hypothesis is valid, such a division could be a motivation for dissociated transfer of rule application versus structure building, and such a division could serve as a neurological correlate of weak transfer hypotheses like
Empirical data is also available to support this hypothesis. In an fMRI study of syntactic structure building, Brennan, et al. (2012) found that structure building was strongly correlated with left anterior temporal lobe activity, while no correlation was found between structure building and left frontal lobe activation. Peele, Troiani & Grossman (2009), on the other hand, conducted a study of rule-based versus similarity-based judgments of a concept’s semantic features. The investigators found that the rule-based task provoked increased frontal lobe activity, while the similarity-based task resulted in higher temporal and parietal activation. The results of these two studies suggest that there may be a spatial division in the brain between rule application and structure building processes. Rüschemeyer, et al. (2005), add more support to the idea of physically separate language processing nodes with their fMRI study of L1 and L2 processing of specific (spoken) linguistic structures. The authors find that nonnative speakers’ activation patterns associated with processing semantic violations is more similar to natives than that of syntactic violations, which implies a cerebrally-based separation between these two linguistic aspects. In addition, the nonnative speakers showed activation in different cortical areas than natives, possibly implying a different overall processing routine altogether.
The “structure” versus “rule” distinction discussed above has implications on language processing both broadly—across the major linguistic divisions—and more specifically as well. In section 5 below, this distinction will be extended to the possible separation between processing of grammatical gender and number in Spanish, both as it
acquisitions hypotheses continues with descriptions and analyses of the more modern approaches to the subject.
2. Cognitive Contributions to Language Acquisition The Critical Period Hypothesis and its immediate successors seek primarily to determine the initial state of second language acquisition. These hypotheses take a theoretical approach, but they stop short of examining second language use and cognitive processing by proficient L2 speakers. Others in the field of language acquisition have gone further, examining second language processing—rather than simple behavior—and interactions with the native language. One of the major questions that these researchers attempt to answer is how native language processing differs from, and how it is similar to, processing of a second language. With the rise of detailed brain imaging and mapping techniques like PET and fMRI, this segment of the language acquisition field has grown immensely in the last two decades.
2A. The Shallow Structure Hypothesis. A leading theory in the field is the Shallow Structure Hypothesis, first proposed by Clahsen & Felser (2006). Although anecdotal evidence suggests that L2 speakers often suffer deficits in their language production and comprehension compared to native speakers, there was not much evidence of the specific character of this deficiency prior to Clahsen & Felser’s work. Clahsen & Felser’s paper begins by rejecting four main observations based on experimental evidence in studies conducted by various investigators and upon which previous theories of language acquisition were based.
First, language learners perceive and understand linguistic input in a fundamentally
sentences (e.g. “garden path” sentences) have found that native speakers utilize structural (e.g. syntactic) information much more than lexical-semantic and contextual cues, while L2 learners use the reverse tactics. Clahsen & Felser mention that this difference could be attributed either to differences in the processing system itself or to general cognitive limitations like lower working memory resources allocated to L2 comprehension. Second, time course-sensitive imaging methods such as ERP (event-related potential, a form of EEG monitoring) have revealed that L2 learners process language more slowly than native speakers, “possibly reflecting a lack of automaticity” (Clahsen & Felser 2006:4) compared to natives. Third, transfer from the native language may affect production and comprehension of the L2, as explored above; obviously, native speakers are not limited by previous linguistic knowledge, though they also do not receive any possible beneficial effects of transfer. Fourth, L2 learners may have wide-ranging access to lexical retrieval faculties but limited access to the procedural memory system associated with the rules necessary to parse natural language, whereas native speakers have comprehensive access to both systems. Clahsen & Felser describe each of these observations in detail and attempt to reconcile each with the contemporary acquisition and processing hypotheses. They ultimately offer a reasoned dismissal of each observation, with the rationale that none of the four observations adequately explains the gulf between native speakers and language learners.
Instead, the authors offer their Shallow Structure Hypothesis (SSH). In essence, this hypothesis is a modification of the fourth observation mentioned above. Clahsen & Felser begin with the notion, from the Weak Transfer hypotheses, among others, that L2 speakers
morphology-based tasks, on which L2 speakers perform roughly on par with native speakers. The differing success in sentence processing between L2 speakers and natives can be related to L2 speakers’ inability to utilize syntactic cues in parsing—or, at the very least, their overreliance on lexical cues at the expense of syntactic ones. Instead of building a series of associations based on the structure of a sentence as a whole to support effective and quick parsing, L2 speakers rely instead on short-distance lexical-semantic cues, pragmatic information, and relevant extralinguistic signals. In effect, Clahsen & Felser claim, L2 speakers initially parse sentences based on their knowledge of individual words and phrases, as well as the world around them, but not based on the relationships between phrases.
2B. Empirical details of the SSH. Consider the following sentence: “Someone saw the servant of the actress who was on the balcony” (Clahsen & Felser 2006:17). This sentence may be interpreted in one of two ways, based on whether the relative clause (RC) attaches to the matrix clause or the prepositional phrase: either the servant is on the balcony, or the actress is. RC attachment of this sort varies cross-linguistically and is generally systematic within a given language. English and Spanish differ in their preferred RC attachment, yet even highly proficient L2-learners fail to acquire the specific RC attachment. Rather than simply adhering to their native language’s RC attachment preference, however, these speakers tend instead not to prefer either attachment strategy and essentially choose randomly between the two available NPs. When a thematic preposition is present—that is, a preposition that explicitly assigns a thematic role to its complement—any randomness vanishes, and a clear preference for a specific RC
and similar results advocate generally for a departure from a transfer-based theory of L2 acquisition and specifically for a view of L2 processing based on word- and morphemelevel relations.
Though the abovementioned studies suggest a clear and persuasive distinction between natives and L2 learners, it is important to note that their results only surface in on-line tasks. When highly-proficient L2 learners are asked to parse the same sentences in an off-line setting, their judgments conform to those of native speakers (Clahsen & Felser 2006), indicating that the abovementioned effect cannot be linked strictly to grammatical deficits per se. A later literature review by the same authors, Felser & Clahsen (2009), similarly concludes that “even advanced L2 learners seem to rely more on lexical storage and semantic information than on grammatically-driven processing mechanisms” (316).
Other investigators have observed similar on-line deficits alongside accurate off-line performance as well (Montrul 2011). One explanation of this phenomenon could be that L1-L2 transfer only occurs as a result of automatic processing and not during deliberate, conscious (i.e. off-line) parsing. Perhaps L2 learners fall back on explicit language instruction or remembered knowledge of specific linguistic input while performing an offline task, but are not able to use such a strategy on-line because of time constraints.
Although this explanation is intuitively reasonable, empirical evidence for or against it is not yet available.
In essence, Clahsen & Felser’s Shallow Structure Hypothesis amounts to the above assertion that any L2-speaker deficits are related to a lack—or a weaker structuralization—of syntactic relations above the word level. Rather than building
speakers instead parse sentences heuristically, using their knowledge of words’ thematic structures (if not a priori then, at the very least, via predicate-argument relations), semantic and realistic meanings, and pragmatic relations to the discourse as a whole.
Rather than build a syntactic tree as the mental representation of a sentence, as it is assumed that native speakers do, Clahsen and Felser propose that L2 speakers construct a series of non-hierarchical relations from which a coherent whole emerges.
2C. The SSH and its relation to competing hypotheses. The SSH differs from other contemporary hypotheses in that it proposes that L2 speakers process language in a profoundly different way than natives (either mature speakers or young L1 acquirers). In this respect, the SSH directly supports the conclusions of the CPH for L2—namely, that L2 learners who begin a language after the critical period do not acquire that language to the same level of completion as native speakers. Furthermore, the SSH refines Penfield & Roberts and their adherents’ views of early language acquisition by proposing a precise modality of post-critical L2 deficits, one of the main holes in the original hypothesis.
In understanding Clahsen & Felser’s contribution to the CPH, it is important not to conflate their SSH with the weak transfer hypotheses of Vainikka & Young-Scholten and Eubank. Though both the SSH and the weak transfer hypotheses support the conclusions of the CPH, SSH puts aside transfer effects entirely in favor of a wholly organic view of L2 acquisition. Unlike weak transfer, the SSH implies that all L2 deficits are related to the act itself of learning a second language at an advanced age, not to interference from the L1.
Clahsen & Felser’s observations jibe with an account of the Critical Period based on the loss or decline of neurological plasticity and reorganization, as advanced by Lenneberg as early
relying instead on a more holistic approach that associates linguistic inequities with differing processing methods between natives and late L2 learners.
2D. Limitations of the SSH. Though Clahsen & Felser’s Shallow Structure Hypothesis does provide the CPH with much greater explanatory power, it is not without its limitations. First, their claim that L1 transfer plays no part in L2 acquisition may be too strong. Sorace (2006) notes that the limited processing abilities that Clahsen & Felser cite as the impetus for shallow structure-building may actually encourage L1 transfer over shallow L2 structures. That is, an L2 learner may revert to their L1 syntax when such a practice would be more “economical.” Furthermore, Sorace argues that it is likely that shallow processing is available to all speakers as a mechanism for parsing complicated structures, but that L2 speakers simply utilize this mode of processing more often. To this author’s knowledge, these questions have not been adequately investigated, but they do raise legitimate issues with Clahsen & Felser’s proposal.
Second, as Clahsen & Felser themselves note, the studies they cite focus almost exclusively on advanced language learners. There is not only a paucity of data on lessproficient learners (and, likewise, extremely proficient, “near-native” learners), but also on the changes in a speaker’s grammatical processing methods that take place as his or her proficiency increases. Such documentation is crucial to the understanding of L2 acquisition, since subsequent studies have shown that proficiency level is an important influence of processing ability (Foote 2009). Proficiency level will be explored in detail in the next section, especially with regard to the question of whether and to what extent nativelike proficiency in late L2 learners is possible.