Development over Time of L2 Neurocognitive Processing

Many of the conflicting results of previous L2 ERP research are likely due to the cross-sectional nature of the work, and one of the improvements in the field over the last five years is the implementation of longitudinal research designs that allow researchers to better understand the development of L2 neurocognitive processing while controlling factors that cannot be controlled well with cross-sectional designs. One of the primary differences among L2 ERP studies is the proficiency of L2 learners, which is often only self-reported and, if measured objectively, is generally measured with different instruments across studies. Research has shown that groups of learners at different levels of proficiency have evidenced different ERP signatures to the same stimuli (e.g., Bowden et al., Reference Bowden, Steinhauer, Sanz and Ullman2013; Hahne, Reference Hahne2001; Hahne & Friederici, Reference Hahne and Friederici2001), which led to the proposal that ERP effects change qualitatively over time as proficiency and experience increase (Steinhauer et al., Reference Steinhauer, White and Drury2009). Longitudinal research studies are essential for testing this prediction. Fortunately a body of longitudinal L2 ERP research, which began with Osterhout et al. (Reference Osterhout, McLaughlin, Pitkänen, Frenck-Mestre and Molinaro2006) and has become more common over the last five years, has shown that nativelike ERP responses can change qualitatively over time (Davidson & Indefrey, Reference Davidson and Indefrey2009; McLaughlin et al., Reference McLaughlin, Tanner, Pitkänen, Frenck-Mestre, Inoue, Valentine and Osterhout2010; Morgan-Short, Finger, Grey, & Ullman, Reference Morgan-Short, Finger, Grey and Ullman2012; Morgan-Short, Sanz, Steinhauer, & Ullman, Reference Morgan-Short, Sanz, Steinhauer and Ullman2010; Morgan-Short, Steinhauer, Sanz, & Ullman, Reference Morgan-Short, Finger, Grey and Ullman2012; White, Genesee, & Steinhauer, Reference White, Genesee and Steinhauer2012). Davidson and Indefrey (Reference Davidson and Indefrey2009) demonstrated that L1 Dutch–L2 German speakers who had not evidenced any ERP response to German case declension violations (a form unique to their L2) during pretesting, developed a P600 response to the form during training that was also elicited at posttesting. (Note, however, that they did not develop such a P600 to gender agreement violations—a form present in their L1 but with a different structure. See the “L1-L2 Similarity in L2 Neurocognitive Processing” section for further discussion about similar patterns of results.) Similarly, White et al. (Reference White, Genesee and Steinhauer2012) provided evidence that native speakers of Korean and Chinese, who did not show any ERP effect to L2 English past-tense violations at the beginning of a 9-week intensive English course, developed a P600 to such violations by the end of the course.

Qualitative changes in the ERP signature are not limited to changes from null effects to P600s—other patterns of qualitative change have also been observed. McLaughlin et al. (Reference McLaughlin, Tanner, Pitkänen, Frenck-Mestre, Inoue, Valentine and Osterhout2010) report a series of longitudinal studies that examine the development of neurocognitive processing of different L2 morphosyntactic and morphological structures over the course of the first-year, university-level language classes. In each study, learners evidence a qualitative shift from N400 effects after a few weeks of exposure to P600 effects after an extended amount of exposure, suggesting discrete stages of grammatical processing with increasing exposure and proficiency. A similar pattern of findings was reported in work by Morgan-Short and colleagues (Morgan-Short et al., Reference Morgan-Short, Sanz, Steinhauer and Ullman2010; Morgan-Short, Finger, et al., Reference Morgan-Short, Finger, Grey and Ullman2012; Morgan-Short, Steinhauer, et al., Reference Morgan-Short, Finger, Grey and Ullman2012). In this study, a group of learners who were trained on an artificial language demonstrated a N400 response at lower levels of proficiency to both morphosyntactic and syntactic violations. At higher levels of proficiency, these same learners showed a P600 response to morphosyntactic violations and as well as an AN and a P600 response to syntactic violations (see “Effects of Learning Contexts on L2 Neurocognitive Processing” section for more details).

Taken together, the results of longitudinal studies are largely consistent with the pattern of findings from previous cross-sectional L2 ERP research and lend support to the claims made by Steinhauer et al. (Reference Steinhauer, White and Drury2009), which state that for grammatical processing, L2 learners may initially evidence no ERPs or N400s at low levels of proficiency and exposure, P600s at more intermediate stages, and finally ANs and P600s at the highest levels of proficiency and exposure (although the elicitation of ANs may not be expected to be reliable as they are not consistently elicited in L1). Furthermore, this body of longitudinal research has established that this progression is not an artifact of cross-sectional research designs but rather that it manifests itself in individuals as they gain proficiency and experience with their L2. This pattern of findings, however, does not necessarily mean that every learner will experience this shift for every (morpho)syntactic structure. Development may be affected by multiple factors, which are considered in the following section.

L1-L2 Similarity in L2 Neurocognitive Processing

One major area of interest in L2 ERP research has been whether L1-L2 similarity plays a role in the development of nativelike neurocognitive processes. Indeed, more ERP studies have been conducted to examine this issue than any other issue, and evidence is available to be able to arrive at some synthesis of the results for L2 learners at low and intermediate levels of proficiency (Bond, Gabriele, Fiorentino, & Alemán Bañón, Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011; Gabriele, Fiorentino, & Alemán Bañón, Reference Gabriele, Fiorentino and Alemán Bañón2013; White et al., Reference White, Genesee and Steinhauer2012) as well as for L2 learners at higher levels of proficiency (Foucart & Frenck-Mestre, Reference Foucart and Frenck-Mestre2011, Reference Foucart and Frenck-Mestre2012; Frenck-Mestre, Foucart, Carrasco, & Herschensohn, Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009; Gabriele et al., Reference Gabriele, Fiorentino and Alemán Bañón2013; Gillon Dowens, Guo, Guo, Barber, & Carreiras, Reference Gillon Dowens, Guo, Guo, Barber and Carreiras2011; Gillon Dowens, Vergara, Barber, & Carreiras, Reference Gillon Dowens, Vergara, Barber and Carreiras2010).

In regard to L2 learners at lower levels of proficiency, some studies have found nativelike P600s for linguistic features that are present in the L1 as well as for structures that are unique to the L2 (Bond et al., Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011; White et al., Reference White, Genesee and Steinhauer2012). In White et al. (Reference White, Genesee and Steinhauer2012), L1 Korean-L2 English speakers and L1 Chinese-L2 English speakers were presented with sentences with either correct or incorrect usages of the English regular past tense. For the L1 Korean speakers, this feature was present in their L1, although it was instantiated in a different form. For the L1 Chinese speakers, this feature was present only in the L2. Neither group evidenced any ERP effect when they were initially tested, but both groups evidenced P600 effects to the stimuli after a 9-week intensive English course that resulted in L2 developmental gains, although the P600 in the L2 Chinese speakers showed a delayed latency. A similar set of findings was found by Bond et al. (Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011). In this study, low proficiency L1 English–L2 Spanish speakers, tested in an L1 context, showed P600 effects to agreement structures that were (a) similar to their L1 (i.e., noun-verb number agreement), (b) different from their L1 (i.e., noun-adjective number agreement), and (c) unique to the L2 (i.e., noun-adjective gender agreement), although the P600 to this structure was smaller than that evidenced for noun-adjective number agreement. Gabriele et al. (Reference Gabriele, Fiorentino and Alemán Bañón2013) examined L1 English–L2 Spanish speakers processing L2 number and gender agreement within and across phrases across learners at different proficiency levels. The number feature is present in English, but the gender feature is not. Although the low proficiency L2 speakers did not evidence P600s, which were observed in Spanish native speakers, the L2 learners at intermediate and high proficiency levels did—for both number and gender agreement and both within and across phrases, although the P600 effect was only marginal for gender agreement in intermediate learners. The set of results from these studies are generally consistent with the findings reported in the McLaughlin et al. (Reference McLaughlin, Tanner, Pitkänen, Frenck-Mestre, Inoue, Valentine and Osterhout2010) study, in which first-year L2 learners evidenced P600 effects for structures that were both present in their L1 and unique to their L2. In all, it appears that at low levels of proficiency, nativelike ERP effects can be found for linguistic structures that are present in learners’ L1 or unique to their L2 (Bond et al., Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011). These effects are even more likely to be evidenced when learner proficiency increases (Gabriele et al., Reference Gabriele, Fiorentino and Alemán Bañón2013; McLaughlin et al., Reference McLaughlin, Tanner, Pitkänen, Frenck-Mestre, Inoue, Valentine and Osterhout2010; White et al., Reference White, Genesee and Steinhauer2012). In some cases, nativelike effects can also be found for structures that are different from learners’ L1, even at low levels of proficiency (Bond et al., Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011).

Nativelike processing of L2 features has been evidenced more consistently in learners at high levels of proficiency. Gillon Dowens et al. (Reference Gillon Dowens, Vergara, Barber and Carreiras2010) examined number and gender agreement processing in advanced L1 English-L2 Spanish speakers. Although Spanish instantiates both number and gender features, English instantiates number features only. Thus if L1 similarity plays a role in the development of L2 processing, one might expect qualitatively different patterns of results between number and gender agreement in the L2 speakers. The L2 speakers in this study began to learn Spanish after the age of 20 years, had at least 12 years of immersion experience, reported using their L2 daily, and were tested in an immersion context. The native-Spanish speakers in the study evidenced LANs, P600s, and late ANs to both number and gender agreement between determiner and nouns and between nouns and adjectives. The L2 speakers showed nativelike LAN and P600 effects to both number and gender agreement violations between determiners and nouns as well as an additional late AN for number agreement. Thus they evidenced qualitatively similar effects compared to native speakers regardless of whether the L2 feature was found in their L1. There was, however, a quantitative difference between the size of the P600 effect for number and gender agreement, which was larger for number agreement than for gender agreement. Interestingly, the same learners only evidenced a P600, and not a LAN, for number and gender agreement violations between nouns and adjectives. Again the P600 effect was larger in response to number violations as compared to gender violations. In all, the results from this study suggests that L2 learners can evidence nativelike processing, including a LAN effect for a grammatical feature that is not instantiated in their L1, although the elicitation of such an effect may depend on the linguistic structure on which it is expressed, and its magnitude may be affected by L1-L2 similarity.

Gillon Dowens et al. (Reference Gillon Dowens, Guo, Guo, Barber and Carreiras2011) extended this research with L1 Chinese–L2 Spanish speakers who do not have number or gender features in their native language. Participants in this study had begun to study Spanish after the age of 18 years, were at the end of their third or fourth year of a university degree in Spanish studies, had demonstrated a high level of proficiency on a standard Spanish test, and reported using Spanish daily, but were tested in a L1 context. They were tested with the same materials used in Gillon Dowens et al. (Reference Gillon Dowens, Vergara, Barber and Carreiras2010). For these learners, neither number nor gender features were available for transfer from their L1, yet the learners evidenced P600 effects for number and gender agreement between determiners and nouns as well as between nouns and adjectives. The learners also evidenced a late AN to both types of number agreement violations and a trend to a late AN to both types of gender agreement violations. Contrary to the L1 English–L2 Spanish learners, the L1 Chinese–L2 Spanish learners did not evidence a LAN effect for any structure, which the authors interpret as evidence that an earlier “less controlled stage of morphosyntactic processing of violations is not available to the Chinese late learners” (Gillon Dowens et al., Reference Gillon Dowens, Guo, Guo, Barber and Carreiras2011, p. 1658). However, they also point out that other factors, such as immersion experience and/or proficiency, could account for the difference in the elicitation of the LAN effect across the two studies. In general though, the results from these studies suggest that nativelike P600 effects can be elicited in L2 even for grammatical features that are not present in learners’ L1, and that nativelike ANs can be elicited for unique L2 features, although this seems to depend on the typological similarity of the L1 and L2.

Additional fine-grained examination of the role of L1-L2 similarity on L2 neurocognitive processing has been provided through a series of studies conducted by Foucart, Frenck-Mestre, and colleagues (Foucart & Frenck-Mestre, Reference Foucart and Frenck-Mestre2011, Reference Foucart and Frenck-Mestre2012; Frenck-Mestre et al., Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009). In these studies, gender agreement in L2 French speakers has been examined across speakers from different L1s, for different structures, and for the presence or absence of overt phonological gender realization. In three studies, L2 learners at advanced levels of proficiency in French were tested in a French-speaking environment. In each study, different groups of native speakers of French exhibited P600 effects in response to all gender agreement violations.

Frenck-Mestre et al. (Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009) first examined the influence of L1 on L2 processing across different learner groups. L1 English-L2 French learners and L1 German-L2 French learners were tested on gender agreement between plural nouns and postposed adjectives. Gender agreement is a feature that is not present in English. In German, gender agreement is present but does not express itself on postposed adjectives that modify plurals nouns. Results from the study showed P600 effects for L1 English speakers but not for L1 German speakers. The study also examined whether overt expression of gender agreement was necessary for L2 learners to be able to process gender in their L2. L1 Spanish–L2 French speakers were exposed to gender agreement between singular nouns and postposed adjectives. Gender agreement is present in Spanish and expresses itself overtly on all singular postposed adjectives. In French, gender agreement expresses itself orthographically but not phonologically on some adjectives (phonologically invariable adjectives) and both orthographically and phonologically on other adjectives (phonologically variable adjectives). L1 Spanish–L2 French speakers demonstrated the same pattern as that of native-speakers of French: P600s were found for both invariable and variable adjectives, although the effects were larger for variable adjectives in both groups. The only difference found between the speaker groups was that the P600 response in the L2 groups was somewhat delayed compared to the native speakers. In all, the results from this study suggest that L2 learners can process features that are unique to their L2 or present in a similar form in their L1, but have more difficulty processing features that are present but expressed in a different form in their L1.

In order to further tease apart how an L2 feature that is present in L1 but expressed differently is processed, Foucart and Frenck-Mestre (Reference Foucart and Frenck-Mestre2011) continued to examine L2 French gender processing in L1 German speakers for multiple forms: preposed and postposed adjectives and determiners. In this study, the results indicated that the learners could process L2 gender agreement between determiners and nouns, although only a subgroup showed the P600 when the lexical gender of the noun in French did not match the lexical gender in their L1. No evidence was found for a P600 for L2 speakers for noun-adjective gender agreement, even when the surface structure of the noun phrase matched that of their L1 (det-adjective-noun). Note a couple of differences between the determiner-noun and the noun-adjective conditions that might have contributed to the differing results: For the determiner-noun stimuli, all stimuli were singular and participants were asked to consider whether the sentences were semantically acceptable or not. For the noun-adjective conditions, all the stimuli were plural and the participants were asked to judge whether the sentence was correct or not (syntactically and/or semantically). The differing results between the linguistic structures could have also been due to potential proficiency differences for the particular structures. This possibility, however, is difficult to evaluate because the behavioral results are not reported in such a way that it is possible to interpret the results across conditions, although all participant groups seemly came from a similar population (but this was not statistically established). In sum, processing differences were found among the determiner-noun and the noun-adjective conditions, but it is not clear whether these differences were due to the forms themselves or to the task or proficiency differences.

Finally, Foucart and Frenck-Mestre (Reference Foucart and Frenck-Mestre2012) examined the processing of features unique to a learner's L2. L1 English-L2 French speakers showed a P600, which was more frontally distributed than in native-speakers, for noun-adjective gender agreement when the adjectives followed the nouns, which is the most frequent placement for adjectives in French. The L2 speakers, however, evidenced a N400 rather than a P600 for gender agreement on preposed adjectives, which occur less frequently in French. Thus there were processing differences tied to the frequency of the form even though the L2 speakers rejected a similar proportion of violation sentences for the two conditions. For gender agreement on predicative adjectives, native-speakers showed a P600, but L2 speakers did not. The researchers noted, however, that individual differences were apparent with some learners showing a sustained positivity. In a final eye tracking study, the L2 learners showed online sensitivity, which was not distinguishable from L1 speakers, to gender agreement on predicative adjectives, suggesting that L2ers can process grammatical gender over hierarchical distances when they are able to read the stimuli at their own pace.

The results from this body of research reflect a converging but complex picture regarding the influence of L1 features on L2 processing. It appears that L2 learners can evidence nativelike processing of L2 features when they are unique to the L2 (Bond et al., Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011; Foucart & Frenck-Mestre, Reference Foucart and Frenck-Mestre2012; Frenck-Mestre et al., Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009; Gillon Dowens et al., Reference Gillon Dowens, Vergara, Barber and Carreiras2010; Gillon Dowens et al., Reference Gillon Dowens, Guo, Guo, Barber and Carreiras2011; McLaughlin et al., Reference McLaughlin, Tanner, Pitkänen, Frenck-Mestre, Inoue, Valentine and Osterhout2010; White et al., Reference White, Genesee and Steinhauer2012), although quantitative differences can be evident between L2 and L1 speakers (Frenck-Mestre et al., Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009) and within learners for features that are either present or absent in L1 (Gillon Dowens et al., Reference Gillon Dowens, Vergara, Barber and Carreiras2010). Nativelike processing patterns also appear to be evidenced when the feature is expressed similarly in L1 and L2 (Bond et al., Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011; Foucart & Frenck-Mestre, Reference Foucart and Frenck-Mestre2011; Frenck-Mestre et al., Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009; McLaughlin et al., Reference McLaughlin, Tanner, Pitkänen, Frenck-Mestre, Inoue, Valentine and Osterhout2010; White et al., Reference White, Genesee and Steinhauer2012). Interestingly, it seems that the most difficulties for processing L2 features come about when the feature is present in the L1 but is expressed according to a different set of rules in the L2 (Foucart & Frenck-Mestre, Reference Foucart and Frenck-Mestre2011; Frenck-Mestre et al., Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009). This pattern of findings is also consistent with findings from recent longitudinal work (Davidson & Indefrey, Reference Davidson and Indefrey2009) as well as with findings from earlier L2 ERP research (Sabourin & Haverkort, Reference Sabourin, Haverkort, van Hout, Hulk, Kuiken and Towell2003; Sabourin & Stowe, Reference Sabourin and Stowe2008; Tokowicz & MacWhinney, Reference Tokowicz and MacWhinney2005). In addition to providing converging evidence about the general effects of L1-L2 similarity, this body of recent research had shown that nativelike processing patterns can be evidenced even when the feature is only orthographically (and not phonologically realized; Frenck-Mestre et al., Reference Frenck-Mestre, Foucart, Carrasco and Herschensohn2009) and when the lexical specification is not shared between the two languages (Foucart & Frenck-Mestre, Reference Foucart and Frenck-Mestre2011). An important outstanding issue to be addressed by future research seems to be what specifically interferes with L2 processing when a feature is present in both L1 and L2 but is expressed differently.

Individual Differences in L2 Neurocognitive Processing

Longitudinal L2 ERP research has established that the ERP response elicited by grammatical processing varies within learners as L2 proficiency develops. One might also expect quite a bit of variability between learners as they progress along a developmental continuum. In other words, if we were to examine a set of learners’ ERP responses at a given stage of L2 development, we might expect to see both quantitative and qualitative variability among learners. Given that ERP variability has been evidenced even in native speakers (Newman, Tremblay, Nichols, Neville, & Ullman, Reference Newman, Tremblay, Nichols, Neville and Ullman2012; Osterhout, Reference Osterhout1997; Pakulak & Neville, Reference Pakulak and Neville2010) and that variability in L1 and in L2 may be obscured by examining only the grand mean effects in ERP data (Tanner, McLaughlin, Herschensohn, & Osterhout, Reference Tanner, McLaughlin, Herschensohn and Osterhout2013), several researchers have begun to employ correlational and regression-based analyses to explore what types of individual differences exist in L2 neurocognitive processing and what factors may account for such individual differences.

Emerging evidence has shown that ERP effects are subject to individual differences in L2 proficiency, verbal aptitude, and other cognitive factors (Batterink & Neville, Reference Batterink and Neville2013; Bond et al., Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011; Tanner, Inoue, & Osterhout, Reference Tanner, Inoue and Osterhout2014; Tanner et al., Reference Tanner, McLaughlin, Herschensohn and Osterhout2013; White et al., Reference White, Genesee and Steinhauer2012). In regard to L2 proficiency, research has generally confirmed that increasing proficiency is related to stronger P600s (Batterink & Neville, Reference Batterink and Neville2013; Tanner et al., Reference Tanner, McLaughlin, Herschensohn and Osterhout2013; Tanner et al., Reference Tanner, Inoue and Osterhout2014; White et al., Reference White, Genesee and Steinhauer2012). Tanner and colleagues (Reference Tanner, McLaughlin, Herschensohn and Osterhout2013) examined individual differences in ERP responses across native-speakers of English enrolled in first-year and 300-level German classes (more advanced courses generally taken during the third year of language study) at a U.S. university and found that scores on a judgment task of subject-verb agreement were positively related to the magnitude of the P600 amplitude. In addition, a regression analysis showed that scores on the judgment task but not hours of instruction predicted the magnitude of the P600 effect, suggesting that proficiency rather than amount of exposure was the driving factor in the development of this nativelike processing effect. The positive relationship between the magnitude of the P600 effect and scores on judgment tasks has been replicated cross-linguistically for different grammatical forms, specifically for the English past tense (White et al., Reference White, Genesee and Steinhauer2012) and for determiner-noun number agreement and word order, but not for subject-verb agreement, in “miniature” French, which is comprised of a subset of natural words and grammatical rules from French (Batterink & Neville, Reference Batterink and Neville2013, p. 939). Thus, it seems reasonable to predict that future research will also evidence such a relationship between the size of learners’ P600 and their L2 proficiency, at least as reflected by scores on experimental judgment tasks.

The relationship between the amplitude of an ERP response and proficiency likely extends beyond the specific relationship between the P600 amplitude and judgment scores. Judgment scores have been shown to have a negative relationship with the amplitude of the N400 effect (Tanner et al., Reference Tanner, McLaughlin, Herschensohn and Osterhout2013). In addition, Tanner et al. (Reference Tanner, Inoue and Osterhout2014) reported a positive relationship between native Spanish speaker scores on a standardized English proficiency test, which did not specifically test knowledge of a target form, and the overall magnitude of learners’ ERP response, regardless of whether it was a N400 or P600 effect, when processing English subject-verb agreement. Note, however, that there are slight differences between the relationship between ERP effects and proficiency as measured either by an experimental, judgment task or by an independent, objective measure of proficiency more generally. On the one hand, proficiency assessed by a judgment task that targets a particular L2 form was related to a smaller N400 effect and a larger P600 effect—a finding consistent with the model that L2 learners are likely to show no ERP response or N400s to grammatical processing at low levels of proficiency and P600s at intermediate stages and more advanced stages of proficiency (Steinhauer et al., Reference Steinhauer, White and Drury2009). On the other hand, proficiency assessed by an objective test of general L2 knowledge was related to larger ERP effects regardless of whether they are N400 or P600 effects. Future research is needed to flesh out the relationship between ERP responses and more constrained or more general measures of proficiency.

Individual differences in ERP effects appear to extend beyond L2 proficiency. A study of native English speaker processing of three types of L2 morphosyntactic forms in Spanish (i.e., subject-verb agreement, noun-adjective number agreement, and noun-adjective gender agreement) examined how ERP effects were related to verbal and nonverbal aptitude (Bond et al., Reference Bond, Gabriele, Fiorentino, Alemán Bañón, Herschensohn and Tanner2011). Verbal aptitude was assessed using the short version of the Modern Language Aptitude Test (MLAT; Carroll & Sapon, Reference Carroll and Sapon1959), including the MLAT3 Spelling Clues, which assesses vocabulary knowledge of English as well as sound-symbol association ability, MLAT4 Words in Sentences, which assesses grammatical sensitivity, and MLAT5 Paired Associates, which assesses rote learning ability. Nonverbal aptitude was assessed with Raven's Advanced Progressive Matrices (Raven's APM), which is a nonverbal assessment of general intelligence (Raven, Reference Raven1965). Results showed a positive relationship between the P600 elicited by noun-adjective number agreement and both the MLAT4 and the composite MLAT scores. This relationship, however, may be limited to particular forms, given that no relationship was found between the MLAT and ERP responses to noun-adjective gender agreement or to subject-verb agreement. In addition, no evidence for a relationship between nonverbal aptitude and L2 processing emerged as no correlation existed between Raven's APM and processing of any of the grammatical forms. In all, these findings suggest that some aspects of verbal aptitude, grammatical sensitivity in particular, may be related to the magnitude of learners’ ERP signatures for some target forms but do not provide evidence of relationship with nonverbal aptitude.

Behavioral L2 acquisition research has explored a multitude of individual differences that affect L2 development beyond aptitude (e.g., Dornyei & Skehan, Reference Dornyei, Skehan, Doughty and Long2003; Ehrman, Leaver, & Oxford, Reference Ehrman, Leaver and Oxford2003), and ERP L2 research is beginning to explore whether several of these factors are related to neurocognitive processing. Beyond considering the relationship between the overall magnitude of ERP responses and proficiency, Tanner et al. (Reference Tanner, Inoue and Osterhout2014) also examined whether age of arrival in an L2-speaking environment, total length of residence, frequency of use, and motivation to speak like a native speaker were related to individuals’ ERP responses to subject-verb agreement processing. None of these factors predicted the magnitude of the overall ERP response (as proficiency did), but age of arrival and motivation both predicted whether L2 speakers evidenced N400 or P600 responses. Specifically, the longer an L2 speaker had resided in an L2-speaking environment and the higher their motivation, the more likely they were to evidence a P600 effect rather than a N400 effect. These findings, taken together with other individual difference research, suggest that L2 neurocognitive processing is related to various linguistic and nonlinguistic factors, including specific and general L2 proficiency, verbal aptitude, length of residence in an L2-speaking environment, and motivation. Further research that uses measures of individual ERP responses rather than only grand average measures across learners will surely prove powerful in helping us to not only describe L2 neurocognitive processing but to also explain how and why such processing develops.

Effects of Learning Contexts on L2 Neurocognitive Processing

A final trend that has emerged in L2 ERP research in recent years is how neurocognitive processing may be affected by the context of learning. One question that has been posed by researchers interested in the context of learning is whether L2 learners can converge on nativelike processing responses through mere exposure to the L2. Mueller, Oberecker, and Friederici (Reference Mueller, Oberecker and Friederici2009) exposed German native speakers, who had no knowledge of Italian, to a nonadjacent dependency structure in L2 Italian by asking them to simply listen to short, correct sentences that contained the L2 structure. Exposure to the target structure alternated with testing blocks, where correct and incorrect sentences were presented auditorily and learners were asked to give a grammaticality judgment regarding each sentence. Native speakers of Italian also completed these tasks and evidenced both N400 and P600 effects to violation stimuli. The learner group, however, only evidenced a late N400 along with an anterior positivity (possibly a P3a). The researchers concluded that learners had only been able to employ lexical processing and general attentional processes, whereas the native speakers relied on grammatical processing. Citron, Oberecker, Friederici, and Mueller (Reference Citron, Oberecker, Friederici and Mueller2011) modified the research design of Mueller et al. (Reference Mueller, Oberecker and Friederici2009) to explore whether the alternation of the testing with the learning had impeded the development of the P600 effect in L2 learners. One group of learners received the same alternating design as in the original study and evidenced a N400 effect only. A second group of learners listened to the complete set of the learning stimuli before being asked to judge the grammaticality of the sentences. These learners evidenced a N400 but also evidenced a later positivity that was interpreted either as an attentional-based ERP component (P3a) or as a P600. Although all groups of learners in these studies evidenced behavioral gains and emerging ERP effects, the ERP effects did not clearly converge with that seen in native speakers. This finding could have been due in part to the fact that the native speakers presumably knew all the words and parsed the sentences for meaning whereas the learners did not know any of the vocabulary and must have only been able to parse the sound patterns presented to them. Thus the question about whether L2 can be learned through mere exposure remains open and should be addressed by future research that exposes learners to a novel L2 in a meaningful way.

A second question about whether context may influence how L2s are processed has focused on comparing the effects of different types of contexts. In behavioral research, different contexts, such as learning in a classroom as compared to learning in an immersion environment, have been shown to effect L2 development (Collentine & Freed, Reference Collentine and Freed2004; Freed, Segalowitz, & Dewey, Reference Freed, Segalowitz and Dewey2004; Segalowitz & Freed, Reference Segalowitz and Freed2004). Similarly, a large body of research has examined whether L2 development is affected by different L2 training conditions, particularly looking at differential effects of explicit training, where rule explanation or direction to search for rules is provided, as compared to implicit training, where exposure to the L2 is provided in absence of rules or directions to search for rules (for reviews and meta-analyses of this research, see Norris & Ortega, Reference Norris and Ortega2000; Sanz & Morgan-Short, Reference Sanz, Morgan-Short and Sanz2005; Spada & Tomita, Reference Spada and Tomita2010). Although explicit training often leads to advantages in L2 development, the explicit biases of many tests, among other limitations common to the research, limits the robustness of this conclusion. Indeed, when implicit training is provided in conjunction with task-essential practice, the advantages for explicit training can disappear (Sanz & Morgan-Short, Reference Sanz and Morgan-Short2004, Reference Sanz, Morgan-Short and Sanz2005). Research on the neurocognition of L2 grammatical processing is only beginning to broach the topic of whether different learning contexts or types of training can affect not only L2 development but also the processing patterns that underlie such development (Batterink & Neville, Reference Batterink and Neville2013; Morgan-Short et al., in press; Morgan-Short et al., Reference Morgan-Short, Sanz, Steinhauer and Ullman2010; Morgan-Short, Finger, et al., Reference Morgan-Short, Finger, Grey and Ullman2012; Morgan-Short, Steinhauer, et al., Reference Morgan-Short, Finger, Grey and Ullman2012).

In order to elucidate whether L2 neurocognitive processing can be affected by training context, Morgan-Short and colleagues (Morgan-Short et al., Reference Morgan-Short, Sanz, Steinhauer and Ullman2010; Morgan-Short, Finger, et al., Reference Morgan-Short, Finger, Grey and Ullman2012; Morgan-Short, Steinhauer, et al., Reference Morgan-Short, Finger, Grey and Ullman2012) used an artificial language paradigm. The artificial language, Brocanto2—a modified version of Brocanto (Friederici et al., Reference Friederici, Steinhauer and Pfeifer2002), is a meaningful, productive artificial language whose syntactic rules conform to natural-language universals. Participants in these studies learned Brocanto2 in order to play a computer-based board game. After learning the rules of the game, participants received training on the artificial language itself under one of two conditions, each of which lasted about 13 minutes. Participants received either explicit training, which provided metalinguistic explanations and meaningful examples (phrases and sentences along with their corresponding game constellations and moves), or implicit training, which provided only meaningful examples of sentence-game constellation pairs. After training, participants engaged in practice with the language, which was identical for all learners and was comprised of both comprehension practice (listening to an aural Brocanto2 statement that referred to a move on the game board and then making that move) and production practice (seeing a move on the game board and then describing that move out loud). ERP data was collected at two points in the study—immediately after participants showed above chance performance on practice and again after all practice items (n = 880) had been completed. Participants engaged in a grammaticality judgment task of Brocanto2 sentences that contained either syntactic or morphosyntactic violations as the ERP data was collected.

Behavioral results on the judgment task showed no differences between the groups at low proficiency or at end-of-practice for morphosyntactic or for syntactic processing, although both groups were more accurate on judging syntactic violations as compared to morphosyntactic violations. There also were no group differences for the ERP responses to morphosyntactic stimuli at end-of-practice—both groups showed a N400 to gender agreement between nouns and adjectives and a P600 to gender agreement between determiners and nouns. There were, however, clear ERP differences in all other cases. For morphosyntactic processing at low proficiency, the implicit group showed a N400 for noun-adjective and determiner-noun gender agreement whereas the explicit group did not (Morgan-Short et al., Reference Morgan-Short, Sanz, Steinhauer and Ullman2010). For syntactic processing, the explicit group did not show any ERP effect at low proficiency but showed an anterior positivity (which is not a language-related component) followed by a P600 at end-of-practice. The implicit group showed a N400 at low proficiency and a bilateral AN followed by a P600 at end-of-practice, which is consistent with Steinhauer et al.'s (Reference Steinhauer, White and Drury2009) predicted pattern of development. Morgan-Short, Steinhauer, et al. (Reference Morgan-Short, Finger, Grey and Ullman2012) point out that only the implicitly trained learners evidenced the full biphasic ERP pattern often seen in native (morpho)syntactic processing. Interestingly, when tested 3–6 months later after no exposure to the artificial language, both groups retained their proficiency with the artificial language and showed more nativelike ERP responses after this amount of time of no exposure to the artificial language in that the explicit group lost the nonlanguage related anterior positivity and the implicit group's AN effect became left-lateralized (Morgan-Short, Finger, et al., Reference Morgan-Short, Finger, Grey and Ullman2012). The overall pattern of the results from these studies suggests that neurocognitive processing of L2 (morpho)syntax can be affected by the type of training a learner receives (although this is not necessarily so), even when behavioral differences are not evident, and that implicit training might be more effective in eliciting biphasic nativelike processing patterns.

Batterink and Neville (Reference Batterink and Neville2013) also explored whether explicit or implicit training differentially affected neurocognitive processing for L2 (morpho)syntactic structures. Learners were exposed to a miniature version of French, which contained subject-verb agreement, number agreement between determiners and nouns, subject-verb agreement, and a fixed word order. Learners assigned to the explicit group were provided with a list and a description of the grammatical rules of mini-French and were informed that they would be tested on these rules at the end of training. Learners assigned to the implicit group were not provided with any rule explanation and were not told of the grammar test. All learners were asked to read 18 narratives in mini-French that were comprised of sentences paired with pictures and to responded to comprehension questions after each narrative. A total of 357 sentences were presented. After the training phrase, ERP data was collected as learners judged the grammaticality of mini-French sentences. ERP data was also collected for native speakers who evidenced P600 effects for each of the linguistic conditions as well as an AN for the subject-verb agreement condition. Explicitly and implicitly trained L2 learners showed behavioral differences on the judgment task with an advantage for the explicit group. This group also evidenced a P600 for each condition along with an additional posterior negativity for the subject-verb agreement. The implicit group was divided into two groups, one of which reached similarly high levels of proficiency and also showed P600 effects for all conditions, and another which showed intermediate levels of proficiency but did not evidence any significant ERP effects. Batterink and Neville concluded that the type of training affected learners’ final level of proficiency but did not affect neurocognitive processing, at least when comparing learners at similar levels of proficiency. This conclusion was further supported by regression analyses that showed that training predicted proficiency but not ERP effects, but that proficiency predicted ERP effects.

These initial studies of the effect of type of training on the neurocognition of L2 processing have yielded both converging and diverging results. Both studies found that at higher levels of proficiency, both explicit and implicit training can lead to P600 effects: P600s for syntactic violations and (morpho)syntactic violations (i.e., article-noun gender agreement of Brocanto2; Morgan-Short et al., Reference Morgan-Short, Sanz, Steinhauer and Ullman2010) as well as P600s for syntactic and (morpho)syntactic violations in mini-French (Batterink & Neville, Reference Batterink and Neville2013). The results of the studies differed, however, in that Batterink and Neville (Reference Batterink and Neville2013) did not find ERP effects for a group of implicitly trained learners who did not reach high levels of proficiency whereas the implicitly trained learners in the Morgan-Short et al. studies evidenced N400 effects at low proficiency. In addition, Morgan-Short and colleagues found an AN effect in implicitly but not explicitly trained learners, suggesting that they could uniquely arrive at a biphasic AN and P600 effect that has been evidenced in native speakers of natural languages. Several methodological differences between the studies could account for the differences in the findings. For example, the amount of exposure to the L2 through practice was greater in the Morgan-Short et al. studies as compared to the Batterink and Neville study. If additional practice had been provided in Batterink and Neville, more implicit learners might have achieved high levels of proficiency and evidenced P600 effects. Another important difference is that learners in the Batterink and Neville study only engaged in comprehension practice with the L2 whereas learners in Morgan-Short et al. studies engaged in both comprehension and production practice. The production practice in these studies might have engaged the learners more actively in using the grammatical structures for a meaningful purpose (Swain, Reference Swain1993; Swain & Lapkin, Reference Swain and Lapkin1995), which might have led to the N400 effect at low proficiency and the AN at high proficiency. Many other differences between the studies exist (e.g., aural versus written presentation of the L2), so it is not possible to determine what was responsible for the differing results. The results, however, clearly warrant further investigation into the effects of context and training on L2 neurocognitive processing.