«Learning a Gendered Language: L2 Acquisition and Relative Processing Costs of Spanish Grammatical Gender and Number Alexander M. Hirsch Yale ...»
3A. Advanced proficiency vs. near-nativeness. The relative dearth of data on near-native L2 speakers has become more significant in recent years as experimentation has revealed important differences between advanced L2 speakers and these near-natives, who are sometimes said to have reached “ultimate attainment.” Furthermore, Carroll (2006) raises the important point that the lack of consistent terminology for proficiency levels makes interpretation of studies of L2 learners quite difficult. Standardization of terminology is especially important because, by both behavioral and neurological measures, near-native, or end-state, speakers pattern separately from advanced speakers who have not reached near-native levels of proficiency. Carroll also suggests that there may be a fundamental problem with comparing L2 learners to L1 monolinguals rather than to bilinguals (native speakers of more than one language); Gillon Dowens & Carreiras (2006) agree, suggesting that investigators should compare late L2 learners to early L2 learners 2 rather than comparing adult L2 processing to child L1 processing. In any event, the separation between advanced and near-native speakers is visible even when comparing only late L2 learners to one another.
This separation manifests itself in various ways. In a picture verification study by Sorace & Filiaci (2006) focusing on interpretation of intrasentential anaphora, near-native L2 Italian speakers employed many of the same parsing strategies as the native controls, as evidenced by similar patterns of performance on null-subject sentences. However, in contrast with the native controls, the experimental subjects tended to prefer to assign the antecedent of overt pronouns to the matrix clause rather than the relative clause. Sorace & That is, L2 learning begun after infancy but before the critical period. Such a situation often occurs in
Filiaci argue that while the near-native subjects have successfully acquired the underlying syntactic rules for such situations, some processing limitation prevents these speakers from accurately parsing exceptional sentences such as those containing overt pronominals.
Furthermore, Sorace & Filiaci contend that transfer does in fact play some role in speakers’ performance on such tasks—in this instance, certain aspects of pronominal interpretation strategies in English may exacerbate the existing processing difficulties that near-native Italian speakers were shown to have with the particular construction under examination.
Jegerski, Van Patten & Keating (2011) continue the line of inquiry begun by Sorace & Filiaci with a similar study of anaphora resolution in pro-drop and non-pro-drop languages, specifically Spanish and English. Jegerski, et al. performed four experiments, two of which established baseline strategies of anaphora resolution by English and Spanish natives, and two of which tested the same strategies in L1-English Spanish L2 learners. The L2 learners were divided into two groups for the two experiments, categorized as intermediate and advanced. It is important to note that the advanced L2 speakers were of relatively high proficiency but were crucially not near-native or end-state learners. From the results of the first two experiments, the researchers found that Spanish natives tend to resolve questions of anaphor antecedents using strategies predicted by Carminati’s (2002) Position of Antecedent (PA) hypothesis, the details of which are not relevant to this paper. English natives, on the other hand, tend to refer to the discourse structure of a two-clause sentence in order to decide the proper antecedent for a given anaphor. From these conclusions, the researchers predicted that the actions of the intermediate and advanced learners would be
Of the second-language learners, those at the intermediate level performed much like the native English speakers when parsing both English and Spanish sentences. That is, even when resolving anaphoric dependencies in sentences of their second language, they acted in certain respects as if these were sentences of English. Specifically, intermediate learners used the sentences’ discourse structures, not the more nativelike PA strategy, to parse such sentences; in fact, no significant PA-like effects were observed within the intermediate group. In terms of anaphor resolution, the intermediate learners seemed to simply transfer their L1 strategies to the L2, even if, in other respects, their abilities were more nativelike.
In parsing the same sentences, the advanced L2 learners also utilized the sentences’ discourse structure as part of their parsing routine; statistical analysis showed a borderline main effect for discourse structure. However, these speakers, in contrast to the intermediate learners, were observed to use a PA-like strategy as well. Though there was no statistically significant main effect for a general PA-like strategy, advanced learners did use this strategy in parsing sentences with coordination structures (as opposed to those with subordination structures, which were parsed randomly).
For the above results, the question arises as to how a single individual can utilize two processing strategies at once, given that they are often at odds and give conflicting results. According to Jegerski, et al., these speakers attempt to utilize both strategies in every situation. In certain situations, only one strategy is viable and produces a single, specific result. In other situations, the strategies conflict and the speaker chooses randomly between the two, as in the subordination example above. Furthermore, Jegerski, et al.
transfer-influenced parsing to more nativelike parsing of the kind seen in native speakers and perhaps near-native L2 speakers as well.
Alternatively, such hybrid behavior may suggest that what prevents these speakers from achieving nativelike performance is not a failure to acquire the correct strategies of the target L2 but rather a failure to discard the interfering strategies of the speakers’ L1.
Given that previous research has shown successfully nativelike performance in off-line tasks by L2 speakers (Clahsen & Felser 2006), the explanation above is certainly plausible.
Jegerski, et al. say as much, citing previous research by Rothman & Iverson (2007) and others showing that the relevant syntactic construction has already been acquired by the time speakers reach a proficiency level on par with Jegerski, et al.’s intermediate group.
Because the Jegerski, et al. study did not include end-state L2 learners, it is not possible to determine whether the advanced learners would have eventually shifted to or towards nativelike parsing strategies, given sufficient gains in proficiency.
3B. Attaining near-nativeness. Hopp (2006) aims to answer this question; namely, whether it is possible to attain nativelike proficiency. In a study of the parsing routines of English L1 and Dutch L1 speakers of German, compared to native German controls, Hopp seeks to separate L1 influence (i.e. transfer) from language proficiency levels of L2 speakers. To do so, Hopp separates 20 L1-English and 20 L1-Dutch speakers into two groups, advanced and near-native, based on the speakers’ results on a C-test, a general language proficiency test that is designed to attach an objective rating of language proficiency to each speaker. Even though these speakers spent similar amounts of time
able to separate these speakers into the two statistically distinct populations, the aforementioned advanced and near-native groups.
Using two experiments designed to examine the subjects’ syntactic reanalysis abilities—specifically their ability to resolve case and verb agreement mismatches—Hopp finds that not only do advanced and near-native speakers differ in their task success rate, they also differ drastically in their parsing and reanalysis strategy. As might be expected, the advanced group did not behave quite like native speakers when parsing the experimental sentences: they were not sensitive to (syntactically-licensed) subject-object scrambling in a self-paced reading task, and showed no difference in reaction time (RT) in a speeded acceptability judgment task (compared to native speakers’ increased RT).
Near-native speaker performance, on the other hand, was aligned with native speakers in almost every way measured. In the self-paced reading task, near-native speakers showed the same slowdown effects as a result of subject-object scrambling as native speakers, and likewise experienced similar RT increases to the native speakers in the speeded acceptability judgment task. Furthermore, near-native speakers outperformed advanced speakers in determining grammaticality of case marking, performing in a nativelike way even when under time pressure.
Although near-native speakers are quantitatively closer to native speakers than the advanced speakers, it is important to note that both L2 groups are proportionally worse at accurately judging case violations than verb agreement violations. Whether this shortcoming represents effects of L1 transfer or computational difficulties is not entirely clear. Because neither of the two L2-German groups—L1-English and L1-Dutch—speak
to determine from the results of the Hopp study alone whether the presence of robust case marking in the L1 would contribute to better performance compared to verb agreement.
Perhaps the most revelatory conclusion that Hopp makes is his contention that the results of the study do not support the broadest version of Clahsen & Felser’s Shallow Structure Hypothesis. While it is true that advanced L2-German speakers do not perform at a level comparable to the native speakers, as predicted by the SSH, the near-native group does not conform to the predictions of the SSH. Hopp’s results show that these near-natives process syntactic elements in a significantly deeper way than advanced speakers. They reliably utilize syntactic features during phrase structure reanalysis, and they are able to process these syntactic cues during serial parsing as well. Furthermore, the near-natives do not use heuristic processing routines, similar to those observed in some aphasics, to process sentences from a surface-relational perspective only (e.g. “a linear ‘subject-first’ preference,” Hopp 2006:391).
All of this is to say that, first, according to Hopp’s findings, the precise proficiency level of subjects can be very important when making conclusions on the nature of end-state acquisition or the possibility of true near-nativeness. Once proficiency is accurately quantified, it becomes clear that achieving native-level language proficiency in certain aspects of language is indeed possible for late L2 learners. However, this is not to say that all domains of language are fully attainable to late learners. According to some research, late L2 learners may not be able to fully acquire all facets of language. For instance, Hahne (2001) observes that even near-native speakers do not exhibit the same degree of automaticity as natives do in integrating syntactic category information into phrase
monitoring techniques, phrase-structure violations elicited a pronounced ERP (eventrelated potential) effect in native speakers but not near-natives, suggesting that the L2 speakers were not sensitive to these violations on the same level as the native speakers.
Although Hahne did not use the same kind of rigorous language-proficiency testing as Hopp, her results suggest that certain domains of language—specifically morphosyntactic features—cannot be acquired to the same extent as syntactic features like case marking.
Keating (2009) comes to a similar conclusion. In an eye-tracking study of beginning, intermediate, and advanced late L2-Spanish speakers 3, the investigator found that, although the advanced speakers are sensitive to noun-adjective gender agreement mismatches when such mismatches occur within a DP, their sensitivity drops off as the distance between the agreeing constituents increases. Native speakers, on the other hand, suffer no such drop-off in sensitivity. Keating concludes that, while gender agreement is in principle acquirable even after the critical period, late L2 learners do not seem to be able to achieve nativelike sensitivity to gender agreement even at endstate proficiency. However, he goes on to say that the broad trend of sentence-type-related reading times for advanced L2 learners tilts in the direction of nativelike performance, even on long-distance nounadjective dependencies, so it is possible that with further progress, the advanced speakers could exhibit more nativelike performance in this domain. To be sure, Hopp’s conclusions are at odds with those of Hahne and Keating in certain important ways, namely with regard to their observation of true nativelike processing in their subjects. Even so, all three researchers decide that, given the trajectories of their near-native subjects’ linguistic 3 Again, these L2 speakers were described as beginning, intermediate, or advanced because of their exposure to the language, not because of quantitative proficiency scores. In justifying the “advanced” label, Keating notes that all members of this group had at least a B.A. in Spanish and had experienced at least one period of complete Spanish immersion.
LEARNING A GENDERED LANGUAGE Hirsch 25 abilities, nativelike processing may be achievable at higher levels than those observed in their subjects, even if such proficiency cannot be attained on a broad scale. In the following sections, the possibility of nativelike processing by late L2 learners will be explored within
4. Descriptive Accounts of Grammatical Gender and Number Processing 4A. Behavioral data. In the section above, it was noted that Keating’s initial results on the acquisition of grammatical gender by L1-English speakers suggest that late learners can fully acquire a structure not present in their native language. As mentioned in section 1, above, the results of the behavioral study by Alarcon (2011) suggest that L2 learners can indeed acquire