A same-system view of L2 processing: Evidence from long-distance syntactic dependencies in L2 Spanish

BUCLD 38 Proceedings To be published in 2014 by Cascadilla Press Rights forms signed by all authors

A same-system view of L2 processing: Evidence from long-distance syntactic dependencies in L2 Spanish

Tania Leal Méndez,a Thomas A. Farmer,b and Roumyana Slabakovac

1. Introduction*

It is generally agreed upon that the outcomes of first (L1) versus second

(L2) language acquisition are remarkably different, and that these differences must be accounted for by any theory of second language acquisition (SLA). Historically, differences in the linguistic knowledge possessed by L1 vs. L2 speakers have been assessed with off-line tasks (e.g. grammaticality judgments), providing evidence for “lack of convergence” across multiple linguistic domains including phonology (e.g. Munro, Flege & MacKay, 1996) and morpho-syntax (e.g. Hyltestam & Abrahamsson, 2003). One influential account, the Fundamental Difference Hypothesis (Bley-Vroman, 1990), proposes that these differences are due to core differences in the grammatical representations of L1 and L2 speakers. The Fundamental Difference Hypothesis has earned support from many empirical studies (e.g. Hawkins & Hattori, 2006; Tsimpli & Dimitrakopoulou, 2006), although across-group differences are not always insurmountable (Leal Méndez & Slabakova, 2012; Iverson, Rothman & Kempchinsky, 2008; Montrul & Slabakova, 2003).

More recently, explanations for native vs. learner ultimate-attainment differences have focused more prominently on processing-based differences, producing a spate of demonstrations that the moment-to-moment processing of an unfolding linguistic signal differs between the two groups (e.g. Dussias, 2001; Frenk-Mestre, 2002). Although these processing-centric accounts vary from one to another across a wide variety of dimensions, they can be roughly categorized into two groups, based upon the degree to which they propose that the processes driving on-line L2 comprehension are the same as, rather than different from, those responsible for L1 comprehension.

“Different-system” accounts of L2 processing propose (obviously, based on the name that we’ve given them) that L2 processing does not rely on the same processing systems that are responsible for L1 comprehension (i.e. L1 and L2 processing are qualitatively different). One of the most well articulated examples of a different-system account is the Shallow Structure Hypothesis (SSH; Clahsen & Felser, 2006; but see for example Ullman, 2001, for another

                                                                                                               *a Tania Leal Méndez, University of Iowa, [email protected]; bThomas A. Farmer, University of Iowa, [email protected]; cRoumyana Slabakova, University of Iowa and University of Southampton, [email protected]

instantiation of a different-system account of L2 processing). The SSH proposes a dual-route processing model with two distinct parsing paths. The first is a full syntactic path (Full Parsing), which is based on a wholly specified syntactic description. The second is a “shallow” path, which is based on lexical, thematic, pragmatic, and contextual information (to the exclusion of syntactic representations). Shallow representations are argued to display both deficient hierarchical structure and a lack of “configurationally-determined elements,” (e.g. movement traces), such that this shallow parsing route is necessarily confined to local domains (e.g. morphosyntactic agreement of adjacent dependencies, word segmentation). The main claim of the SSH is that L2 learners are restricted to the shallow route. The SSH explicitly proposes that the fundamental differences between native and non-native processing are due not to more general processing deficits, per se, but rather to differences in the underlying syntactic representations available to L2ers, which are hypothesized to be anomalous. Based on experiments that fail to detect relationships between more general processing abilities, such as working memory, and patterns of on-line processing for L2ers (Juffs, 2004), the inference was made that differences in L2 processing could be isolated to L2 reliance on shallow structure.

On the other hand, approaches that could be roughly termed “same-system” approaches assert that the systems responsible for processing an L2 are the same systems that drive processing in L1 (e.g. Dekydtspotter, Schwartz & Sprouse, 2006). Instead of proposing different processing systems for natives vs. learners in order to account for observed across-group differences in processing, these differences are typically explained by factors that are unrelated to the systems and mechanisms responsible for grammatical processing, such as non-target prosody, differences in lexical access routines, heteromorphy of semantic fields, and differences in the “timing” of computational moments (Dekydtspotter et al., 2006). In general, “same-system” accounts propose that the same systems, processes, and mechanisms involved in on-line L1 comprehension also serve the fundamental basis for L2 processing, although more subtle differences do exist among these accounts with regard to how those systems are engaged between native and non-native groups during processing. Evidence in support of these approaches comes in the form of similar processing patterns observed across both natives and learners on linguistic tasks (e.g. Hopp, 2006; Kim & Christianson, 2013), although there is often more variability in the processing patterns of learners as a group (e.g. Trenkic, 2007). To explain this variability in learner processing, it is often argued that learners may rely on, activate, or otherwise engage the proposed process differently due to differences in the learning histories of members in the learner group (e.g. McDonald, 2006). One compelling result tends to emerge in support of this claim: the more proficient the learner, the more native-like the processing (e.g. Hopp, 2006; Hoover & Dwivedi, 1998; Jackson, 2008; Lim & Christianson, 2013).

This positive relationship between proficiency and the degree of native-like grammatical processing suggests a degree of malleability in the processing and representation of L2 grammar, and may be potentially informative with respect

to illuminating the nature of the systems responsible for learning in L2. However, the relationship is difficult to interpret because “proficiency” is difficult to define. L2 learners differ substantially from native speakers on a wide array of dimensions that, in combination, produce the construct of “proficiency” (e.g. Cummins, 1980; Hulstijn, 2011). To date, there is no absolute consensus in the literature on what these dimensions are (e.g. Benati, 2009; Verhoeven & De Jong, 1992). Proficiency is graded, in the sense that learners are unlikely to fully acquire some grammatical construction at some discrete point during the learning process (DeKeyser, 2005; Ellis, 2002; MacWhinney, 2004). Instead, each dimension of “proficiency” is likely to emerge in a progressive manner over the course of learning. Here, we argue that proficiency, although complex and multidimensional, is likely to be driven by experience with the input, and thus serves as a coarse-grained “proxy estimate” of the amount of exposure an individual has had to the L2 input. Unfortunately, empirical work on proficiency has historically focused on identifying the components of the construct, instead of what factors drive a change in proficiency. However, the notion that native language knowledge and processing ability improve as a function of experience over time is well supported from work in multiple fields (e.g. Christiansen & Chater, 2001; Elman, 1993; Kemmer & Barlow, 2000; Saffran, Aslin, & Newport, 1996). That said, we believe that accrued experience with the input is quite likely to be a driving force in the increase of L2 proficiency over time.

Coupling observations from research interpreted as providing support for same-system approaches with the grounded assumption that linguistic experience drives increases in L2 proficiency, we test the following claim regarding the oft-observed positive relationship between proficiency and the degree of native-like grammatical processing:

Through learning about the regularities inherent to L2 input over time, as a function of experience, learners acquire progressively more robust knowledge about the structure of the L2 input and the L2 grammatical system, thus driving their grammatical processing ability toward a progressively more native-like state.

2. The Present Study

Inherent to the claim that we seek to address is the notion that experience drives learning about the structure of a language and thus the building of mental representations. But, through what mechanisms is that learning achieved? There is evidence that learners can acquire knowledge about grammatical dependencies inherent to the input of either an L1 or an L2 using statistical probabilities (e.g. Gómez, 2002; Newport & Aslin, 2004). Furthermore, there is also evidence (e.g. Ellis, 2002) that L2 learners are unconsciously attuned to statistical regularities of the input (e.g. frequency). The learning mechanisms that facilitate the acquisition of grammatical dependencies over learning provide one plausible explanation for how exposure to the L2 input facilitates the

progressive, incremental acquisition of knowledge about structure over time. In the experiment reported here, we examine the processing patterns of native-English speaking learners of Spanish on a complex syntactic structure that contains a long-distance dependency.

Long-distance dependencies, which are a fixture in natural languages, have been argued to pose a significant challenge to language processing because a dislocated element must be held in short-term memory until the dependency can be completed later on (Felser & Roberts, 2007). In fact, some accounts of L2 processing (Clahsen & Felser, 2006) predict special difficulties with regard to long-distance dependencies. The Shallow Structure Hypothesis (as discussed above) proposes that L2 processing is impaired due to a lack of detailed grammatical representations in L2 learners. Approaches such as the SSH propose that L1 processing and L2 processing are substantially different. With respect to long-distance dependencies, the SSH predicts that the processing of the non-adjacent elements (e.g. filler-gap dependencies) will not be guided by structural principles. Therefore, the SSH predicts that L2 learners will not show evidence of native-like processing patterns when processing long-distance dependencies, and it is not immediately apparent exactly how different-system models such as SSH would be able to account for the aforementioned positive relationship between proficiency and the degree of native-like grammatical processing on structures containing long-distance dependencies.

In Romance languages, Clitic Left Dislocation (CLLD) is a frequently employed marking of topicalization, where a phrase that has been previously mentioned (i.e. a topic) is re-introduced into the discourse by virtue of being set apart (i.e. left-dislocated). CLLD is an example of a natural long-distance syntactic dependency (a topicalized phrase is connected to the comment via a clitic.)1 For the long-distance dependency to be completed, the clitic and the left-dislocated phrase must agree in phi features (person, number, gender), as in (1). (1) CONTEXT: Are you not going to the party with Elena?

Pues a Elena sí la invitaron, pero a mí no. well to Elena indeed cl.ACC invited.PAST.3rd.PL but to me neg ‘Well, Elena was indeed invited, but I wasn’t.’

Once the topicalized phrase (a Elena) has been introduced, an agreeing clitic (la: feminine, singular) must subsequently appear in order to complete the sentence felicitously. Romance clitics appear before finite verbs (Rizzi, 1982), such that comprehenders would expect the clitic to appear pre-verbally.

Natural long-distance dependencies can be especially useful in examining the development of linguistic processing in language acquisition because they are highly predictable. In the case of CLLD, Spanish speakers have knowledge of this dependency (dislocated phrase ! clitic) and thus can “predict” that a clitic must appear downstream after they have encountered a topicalized phrase. In fact, in the L1 processing literature there is increasing evidence that

                                                                                                               1 We are referring here to pronominal clitics. These are elements that phonologically depend on a host and are argued to have special syntactic properties (Riemsdijk, 1999).

predictive/anticipatory processes aid in the ability of speakers to quickly and accurately interpret an incoming linguistic signal (Altmann & Kamide, 1999; DeLong, Urbach & Kutas, 2005; Dikker, Rabagliati, Farmer & Pylkkanen, 2010; Farmer, Brown & Tanenhaus, 2013; Fine, Jaeger, Farmer & Qian, 2013).

However, there is little work to date concerning whether L2 learners actively predict upcoming linguistic material. In this study, we examine the learning/processing of a long-distance dependency in L2 Spanish with a special focus on whether learners develop an emerging sensitivity to the violation of expectations (predictions) during on-line processing. Data from native Spanish speakers demonstrate when the long-distance dependency is not completed where expected (predicted), processing time increases, thus providing an index of how sensitive native speakers are to the expectation violation. Additionally, by sampling learners from broad swaths of the proficiency spectrum, we demonstrate progressive sensitivity to the violation of expectations (predictions). This sensitivity is progressive in the sense that it can be expressed as a function of increasing proficiency, driven, we argue, by the amount of experience an individual learner has with the regularities inherent to the input. Thus, once again, we see evidence of the oft-reported relationship between proficiency and degree of native-like performance: the more proficient the learner, the more native-like the processing.

3. Method 3.1. Participants

We tested a total of 120 native speakers (NSs) of English who were learners of Spanish (85 women; M = 26.4 years, SD = 9.48 years) and 36 (monolingual) NSs of Spanish (M = 43.8, SD = 11.71). L2 participants were recruited at two U.S. universities. At the time of testing, they were either enrolled in Spanish courses or teaching Spanish courses at the high school or college level. L2 participants were exposed to Spanish after their first language had already been learned (mean age of exposure = 14.4). Two learners from the advanced group were excluded because they took Spanish in Elementary school. The participants in the Spanish NSs group were tested in their native country (Mexico). These participants were included in order to provide a baseline index of native-processing ability against which to compare learners. None of the Spanish NS participants reported fluency in any language other than Spanish. Because previous research has shown evidence that level of education and other variables such as socioeconomic background are closely related to language outcomes (Pakulak & Neville, 2010; Street & Dabrowska, 2010), NS participants were recruited from a middle class local arts college and a state hospital. In order to seek a close match in educational backgrounds, all participants had to have completed at least a year of college in order to be included in the study. 3.2. Experimental materials

3.2.1 Proficiency Test As an index of L2 Spanish proficiency, we included an independent

proficiency test. This test includes a shortened version of a standardized test used for official Spanish language accreditation in Spain (Diploma de Español como Lengua Extranjera). It also includes a section from the reading and vocabulary sections of the Modern Language Association Cooperative Foreign Language Test (Educational Testing Service). This test has been used successfully as a discriminator of proficiency in prior L2 Spanish acquisition research (e.g., White, Valenzuela, Kozlowska-MacGregor & Leung, 2004). The test consisted of 50 multiple-choice items focusing on vocabulary and grammar. The first 30 items were randomized per participant. The last 20 were part of a coherent paragraph and, consequently, they were presented in the same order. Most participants completed this task within 35 minutes. Partly based on previous literature (e.g. Slabakova, Rothman & Kempchinsky, 2011), accuracy range cutoffs were set at 40-50 for the advanced group, 25-39 for the intermediate group, and 10-24 for beginners. Because the self-paced reading task was rather difficult on both syntactic and lexical counts, participants needed to have knowledge of basic lexical items. Thus, this study includes only the data of participants who scored at least 50% on the test, and thus met the prescribed criteria for being classified as of at least “intermediate” proficiency (n=91). 3.2.2. Self-paced reading task

We included a self-paced reading task in order to investigate whether participants were able to predict downstream linguistic material based on previously appearing syntactic information. Stimuli were presented using a self-paced word-by-word moving window display (Just, Carpenter & Wooley, 1982). Target sentences were preceded by a context (necessary in order to make the topicalization natural) and followed by a yes/no comprehension question. Stimuli included dislocated sentences in two conditions (early/late): one where a sentential-conferred prediction was violated (dependency completed late), relative to another where the prediction was met (dependency completed early).

(2) CONTEXT (In Spanish): (Several students visited the department where they had sent their registration.) Target: A aquellas estudiantes la linda secretaria felizmente (les) contó To those students the lovely secretary happily (CL) told.3rd.sg que probablemente las admitirán en el programa. that probably (clitic) admit.pl.3rd.FUT in the program

‘The lovely secretary happily told the students that they would probably be admitted to the program.’

These materials were previously designed by Pablos (2006) to test long-

distance dependencies in a group of Peninsular Spanish NSs. In order to verify NSs clitic preferences, Pablos used an off-line grammaticality-rating task to verify that NSs preferred clitic over clitic-less sentence continuations after a

dislocated phrase was introduced. The sentences used in this study were only slightly simplified from Pablos (2006) by removing a level of embedding; some lexical items were also replaced with more standard counterparts in order to avoid any differences due to dialectal varieties. Example (2) is a sample token.

In this manipulation, the sentential-conferred expectation after encountering a left-dislocated phrase (A aquellas estudiantes) is that a clitic will appear later on. Because Spanish clitics appear before finite verbs (Rizzi, 1982), the clitic would be anticipated before the matrix verb (contó). If speakers had predicted an upcoming clitic after encountering a topicalized phrase, their reaction times (RTs) should be slower at the main verb if the clitic was not present. Pablos (2006) chose matrix verbs that were optionally ditransitive (e.g. ‘to tell’), which increases the chances of an equal bias for the presence/absence of the clitic. We constructed two lists with the 48 experimental sentences so that each participant would only see one version of each sentence (24 experimental sentences per list). An additional 48 sentences, which did not include any dislocated items but were also preceded by a context, were included as fillers. Items were randomized per participant. The experiment lasted approximately 45 minutes.

3.2.3. Language Background Questionnaire

In this questionnaire, we collected demographic information, as well as language background (profile on language use) and language learning history. 4. Results

Figure 1 depicts RTs in the early and late conditions across all words of the target sentences for NSs (panel a), Advanced learners (b), and Intermediate learners (c). As noted in the discussion above, proficiency is a graded construct, one that is best expressed continuously rather than through discrete (and admittedly arbitrary) groupings. However, given space considerations and in order to present a clear picture of how processing unfolded across the entire sentences while maintaining some information about proficiency level, we graphically depict the data based on proficiency-score groupings that have been utilized in the SLA literature.

As illustrated in Fig. 1, there exist surprisingly few differences in mean RT values between the clitic-early and clitic-late conditions across most words of the target sentences. The only statistically reliable differences between the early versus late conditions were located in two regions: 1) the main verb, which occurred directly after the expected clitic did not appear (late condition) versus the main verb that appeared directly following the /presence (early condition), and 2) on the word following the main verb (the complementizer que), providing evidence for a spill-over effect associated with the violation of expectation.

We calculated a difference score that captured, on a subject-by-subject basis, the average difference in RTs produced when the expectation was violated (clitic-late condition) relative to when it was met (clitic-early condition). This difference score provides a by-subject index of the magnitude of the expectation violation effect produced by each learner. Higher difference scores indicate a

larger difference between RTs on the manipulation, and thus, a larger expectation violation effect. There is a significant positive correlation between proficiency scores and matrix-verb RTs when the clitic was absent relative to when it was present (r = .361, n = 91, p < .0005). In other words, with increasing proficiency, non-native RTs appeared progressively more sensitive to the expectation violation, approximating native-like levels of sensitivity at the high ends of the proficiency spectrum.

Fig. 1. RTs per region: A1 aquellas2 estudiantes3 la4 linda5 secretaria6 felizmente7 les8

contó9 que10 probablemente11 las12 admitirán13 en14 el_programa15 In an additional analysis that was designed to assess the degree to which

experience with the L2 was related to proficiency, we coded learner responses to the background questionnaire by calculating the number of months that a learner

reported exposure to or other experience with Spanish, and the number of months that each learner spent in study abroad (thus, living in countries where Spanish was the dominant language). Proficiency scores correlated significantly with both overall exposure to Spanish, r = .472, n = 91, p < .0005, and with study abroad, r = .508, n = 91, p < .0005). These correlations, considered “moderate” in magnitude, provide evidence that proficiency is driven, at least to some degree, by the amount of exposure learner has had to the language.

In a final set of analyses, we assessed the degree to which our experience measures were able to account for variability in the magnitude of the expectation violation. Because the number of months spent learning Spanish is logically confounded with the number of months spent studying abroad, we calculated two partial-correlations. There was a significant positive correlation between the magnitude of the effect elicited by the violation of expectations and number of months of study abroad, after statistically controlling for the total number of months spent learning Spanish overall (r = .249, n = 91, p = .017). However, after controlling for the number of months of studying abroad, there was no significant relationship between the number of months spent learning Spanish and the magnitude of the expectation violation (r = .116, n = 91, p = .272). 5. Discussion and Conclusions

Our study investigated the on-line processing of a long-distance dependency (CLLD) by L2 Spanish learners with levels of proficiency that span a broad swath of the proficiency spectrum. In the paradigm developed here, on-line processing performance has been used to provide a window into L2 the developmental time-course of L2 learning. As proficiency increased, non-native RTs appeared progressively similar to NSs’ in terms of the magnitude of the expectation violation (late condition RTs on the main verb minus early condition RTs at the main verb). We argue that increased RTs on the matrix verb when the clitic was late (expectation violated), relative to when it was early (expectation met), provide an index of the degree to which the subjects were able to use context to generate an expectation for the clitic, and thus, of the degree to which a given participant had learned about the non-adjacent dependency. The more proficient the learner, the more native-like the processing.

We believe that focusing on the development of the ability to generate predictions, and thus on long-distance dependency learning, has direct ramifications for theoretical debates in the L2 learning and processing literatures (Clahsen & Felser, 2006; Dussias, 2001; Frenck-Mestre, 2002) as well as for debates about ultimate levels of L2 attainment (Birdsong, 2006; Long, 2005). In the context of the same-systems versus different-systems debate addressed in the introduction, the results presented here seem most consistent with a same-system perspective. The systems, processes, and mechanisms that drive L1 learning and processing, such as implicit statistical learning and predictive processing, appear to be utilized during L2 learning and processing.

In principle, the data presented here could be deemed problematic for different-systems approaches to SLA, such as the SSH. According to the SSH (Clahsen & Felser 2006), adult L2 learners can only process language “shallowly” because they are argued to lack detailed L2 grammatical representations. With regard to our study, the SSH predicts that learners will only have access to pragmatic and semantic information when processing long-distance dependencies in their L2. Thus, they were not expected to show evidence of native-like language processing. However, our results showed evidence of native-like processing of CLLD by advanced learners.

In reality, however, we haven’t really demonstrated that the exact same systems that are responsible for L1 and L2 learning and processing produced the patterns of RTs across learners and natives. In fact, it’s quite difficult to determine how one could feasibly test the strongest version of the same-systems approach, in which exactly the same L1 systems are engaged in exactly the same way during L2 processing in order to give rise to exactly the same processing behavior with exactly the same grammatical representations. And, although RT patterns appear to be quite similar between the really advanced learners and the natives, that doesn’t mean that the grammatical representation and processes possessed and utilized by the advanced learners were truly nativelike. Indeed, many parameters inherent to the SSH could be modified to afford it the ability to capture the data presented here.

So, then why are these data interesting? We deem them interesting because they are consistent with recent developments in the field of L1 sentence processing that place a premium emphasis on the role of prediction during processing, and on mechanisms and frameworks that acknowledge the inseparability of learning and processing. They also highlight the utility of examining or otherwise addressing the developmental time-course of learning and processing in adult SLA. Crucially, they suggest the necessity of linguistic experience, coupled with basic learning mechanisms, for fostering the protracted, progressive, and incremental L2 acquisition over time.

The “experience is important” point should be obvious to any reader, but the data presented here do provide one additional insight that may be underappreciated in certain SLA frameworks. The numerical outcomes of our language proficiency measure were correlated to language history metrics such as exposure to Spanish (in months) and study abroad (in months), suggesting that L2 proficiency is at least partially contingent on the amount of accrued experience with the L2. Interestingly, only one of these language metrics (study abroad) was correlated with the magnitude of the effect elicited by the violation of expectations. We believe that this result is due to the distribution of CLLD in the input. Although, to our knowledge, there are no large-scale corpora studies on Spanish CLLD (but see Slabakova, 2013), there are many authors noting that CLLD is frequent in spoken Spanish (López, 2009; Quesada, 1997).

In order to examine whether less proficient Spanish learners would have exposure to CLLD, we performed a cursory search of Spanish first- and second-year textbooks and found that CLLD was not included in any of the grammatical

explanations. Because the learners in our study were, at least initially, classroom learners, it makes sense that they had not developed a sensitivity to clitic presence early on in their development. On the other hand, those learners with more study abroad experience would have been exposed to CLLD more often in a naturalistic environment. More broadly, this result suggests that it’s not just the overall amount of exposure to L2 input that drives L2 learning in adults, but also the overall richness of the experience afforded by the environment in which the learning takes pace (see also Pliatsikas & Marinis, 2012).

Learning, exposure through experience, and the richness of the learning environment clearly influence the slope of the L2 learning curve, and we see the compelling potential for advancement in the field of SLA by considering how the effects of these variables can be situated within the theoretical landscape of the field. References Altmann, Gerry, & Kamide, Yuki. (1999). Incremental interpretation at verbs: Restricting

the domain of subsequent reference. Cognition, 73(3), 247-264. Bar, Moshe. (2007). The proactive brain: Using analogies and associations to generate

predictions. Trends in Cognitive Sciences, 11 (7), 280–289. Benati, Alessandro G. (Ed.). (2009). Issues in second language proficiency. London:

Continuum International Publishing Group. Bialystok, Ellen. (1994). Representation and ways of knowing: Three issues of second

language acquisition. In Nick Ellis (Ed.). Implicit and explicit learning of languages (pp. 549-69). London: Academic Press.

Birdsong, David. (2006). Age and second language acquisition and processing: A selective overview. Language Learning, 56(1), 9-49.

Bley-Vroman, Robert. (1990). The logical problem of foreign language learning. Linguistic Analysis, 20, 3–49.

Christiansen, Morten H., & Chater, Nick. (2001). Connectionist psycholinguistics: Capturing the empirical data. Trends in Cognitive Sciences, 5, 82-88.

Clahsen, Harald, & Felser, Claudia. (2006). Grammatical processing in language learners. Applied Psycholinguistics, 27(1), 3-42.

Cummins, James P. (1980). The construct of language proficiency in bilingual education. Current issues in bilingual education, 81-103.

Dekydtspotter, Laurent, Schwartz, Bonnie, & Sprouse, Rex. (2006). The comparative fallacy in L2 processing research. In Proceedings of GASLA 2006 (pp. 33-40).

DeLong, Katherine A., Urbach, Thomas P., & Kutas, Marta. (2005). Probabilistic word pre-activation during language comprehension inferred from electrical brain activity. Nature neuroscience, 8(8), 1117-1121.

Dikker, Suzanne, Rabagliati, Hugh, Farmer, Thomas A., & Pylkkänen, Liina. (2010). Early occipital sensitivity to syntactic category is based on form typicality. Psychological Science, 21(5), 629-634.

Dussias, Paola E. (2001). Sentence parsing in fluent Spanish–English bilinguals. In J. Nicol (Ed.), One mind, two languages: Bilingual language processing (pp. 159–176). Malden, MA: Blackwell.

Ellis, Nick. (2002). Frequency effects in language processing. Studies in Second Language Acquisition, 24(2), 143-188.

Elman, Jeffrey. (1993). Learning and development in neural networks: The importance of starting small. Cognition, 48(1), 71-99.

Farmer, Thomas. A., Brown, M., & Tanenhaus, Michael K. (2013). Prediction, explanation, and the role of generative models in language processing. Behavioral and Brain Sciences, 36, 211-212.

Fine, Alex, Jaeger, Florian, Farmer, Thomas. A., & Qian, Ting. (2013). Rapid expectation adaptation during syntactic comprehension. PLoS ONE, 8: e77661.

Frenck-Mestre, Cheryl. (2002). An on-line look at sentence processing in the second language. In Roberto Heredia & Jeannette Altarriba (Eds.), Bilingual sentence processing (pp. 218– 236). Amsterdam: Elsevier.

García Mayo, María del Pilar, & García Lecumberri, María Luisa. (Eds.). (2003). Age and the Acquisition of English as a Foreign Language. Multilingual matters.

Gómez, Rebecca. (2002). Variability and detection of invariant structure. Psychological Science, 13(5), 431-436.

Hawkins, Roger, & Hattori, Hajime. (2006). Interpretation of English multiple wh-questions by Japanese speakers: a missing uninterpretable feature account. Second Language Research 22, 269-301.

Han, Yongju & Ellis, Rod. (1998). Implicit knowledge, explicit knowledge and general language proficiency. Language Teaching Research, 2, 1-23.

Hopp, Holger. (2006). Syntactic features and reanalysis in near-native processing. Second Language Research, 22, 369–397.

Hoover, Michael L., & Dwivedi, Veena D. (1998). Syntactic processing in skilled bilinguals. Language Learning, 48, 1–29.

Hulstijn, Jan H. (2011). Language proficiency in native and nonnative speakers: An agenda for research and suggestions for second-language assessment. Language Assessment Quarterly, 8(3), 229-249.

Hyltenstam, Kenneth & Abrahamsson, Niclas. (2003). Maturational constraints in SLA. In Catherine Doughty & Michael Long, (Eds.), The handbook of second language acquisition (pp. 539–88). Blackwell.

Iverson, Michael, Kempchinsky, Paula, & Rothman, Jason. (2008). Interface vulnerability and knowledge of the subjunctive/indicative distinction with negated epistemic predicates in L2 Spanish. Eurosla Yearbook, 8(1), 135-163.

Just, Marcel. A., Carpenter, Patricia A., & Woolley, Jacqueline D. (1982). Paradigms and processes in reading comprehension. Journal of Experimental Psychology: General, 111(2), 228.

Juffs, Allan. (1998). Main verb versus reduced relative clause ambiguity resolution in second language sentence processing. Language Learning, 48, 107–147.

Juffs, Allan. (2004). Representation, processing and working memory in a second language. Transactions of the Philological Society, 102(2), 199-225.

Kemmer, Suzanne, & Barlow, Michael. (2000). Introduction: A usage-based conception of language. In Michael Barlow & Suzanne Kemmer (Eds.), Usage based models of language (pp. vii-xxviii). Stanford, CA: CSLI.

Lardiere, Donna. (1998). Case and tense in the ‘fossilized’ steady-state. Second Language Research, 14, 1–26.

Leal Méndez, Tania, & Slabakova, Roumyana. (2012). The Interpretability Hypothesis again: A partial replication of Tsimpli and Dimitrakopoulou (2007). International Journal of Bilingualism, doi: 10.1177/1367006912448125.

Lim, Jung Hyun & Christianson, Kiel. (2013). Second language sentence processing in reading for comprehension and translation. Bilingualism: Language and Cognition. Doi:10.1017/S1366728912000351.

Long, Michael. (2005). Problems with supposed counter-evidence to the Critical Period Hypothesis. IRAL in language teaching, 43(4), 287-317.

López, Luis. (2009). A derivational syntax for information structure Oxford: U Press. McDonald, Janet L. (2006). Beyond the critical period: Processing-based explanations for

poor grammaticality judgment performance by late second language learners. Journal of Memory and Language, 55(3), 381-401.

MacWhinney, Brian. (2004). A multiple process solution to the logical problem of language acquisition. Journal of Child Language, 31(04), 883-914.

Munro, Murray J., Flege, James & MacKay, Ian. (1996) The effects of age of second language learning on the production of English vowels. Applied Psycholinguistics 17, 313-334.

Newport, Elissa L., & Aslin, Richard N. (2004). Learning at a distance I. Statistical learning of non-adjacent dependencies. Cognitive psychology, 48(2), 127-162.

Pablos, Leticia. (2006). Pre-verbal structure building in Romance languages and Basque. Ph.D. Dissertation, University of Maryland.

Pakulak, Eric, & Neville, Hellen J. (2010). Proficiency differences in syntactic processing of monolingual native speakers indexed by event-related potentials. Journal of Cognitive Neuroscience, 22(12), 2728-2744.

Pliatsikas, Christos, & Marinis, Theodoros. (2012). Processing empty categories in a second language: When naturalistic exposure fills the (intermediate) gap. Bilingualism: Language and Cognition, 1(1), 1-16.

Quesada, J Diego. (1997). Obituary: Adios to passive in Spanish. Linguistique, 33, 41-62. Rizzi, Luigi. (1982). Issues in Italian syntax. Amsterdam: Walter de Gruyter. Saffran, Jenny, Aslin, Richard, & Newport, Elissa. (1996). Statistical learning by 8-

month old infants. Science, 274, 1926-1928. Slabakova, Roumyana, & Montrul, Silvina. (2003). Genericity and aspect in L2

acquisition. Language Acquisition, 11(3), 165-196. Slabakova, Roumyana, Rothman, Jason, & Kempchinsky, Paula. (2011). Gradient

competence at the Syntax-Discourse Interface. EuroSLA, 11(1), 218-243. Street, James A., & Dąbrowska, Ewa. (2010). More individual differences in language

attainment: How much do adult native speakers of English know about passives and quantifiers?. Lingua, 120(8), 2080-2094.

Trenkic, Danijela. (2007). Variability in second language article production: beyond the representational deficit vs. processing constraints debate. Second Language Research, 23(3), 289-327.

Tsimpli, Ianthi & Dimitrakopoulou, Maria. (2007). The interpretability hypothesis: Evidence from wh-interrogatives in second language acquisition. Second Language Research 23, 215-242.

Ullman, Michael. (2001). The neural basis of lexicon and grammar in first and second language: The declarative/procedural model. Bilingualism: Language and Cognition, 4, 105-122.

Verhoeven, Ludo & De Jong, John. (1992). Construct of language proficiency: Applications of psychological models to assessment. Amsterdam: Benjamins.

White, Lydia, & Genesee, Frank. (1996). How native is near-native? The issue of ultimate attainment in adult SLA. Second Language Research, 12, 233–365.

White, Lydia, Valenzuela, Elena, Kozlowska-MacGregor, Martyna & Leung, Yan-Kit Ingrid. (2004). Gender and number agreement in nonnative Spanish. Applied Psycholinguistics 25, 105-133.