ABSTRACT. Codeswitching (CS) is a powerful phenomenon to explore how the properties of the two language systems interact in the mind of the bilingual (e.g. Jorschick et al. 2010, Liceras et al. 2016, Burkholder 2018). This study focuses on CS by analyzing eye-tracking data from a group of L1 Spanish-L2 English bilinguals. Given that Spanish-English bilingual communities have been shown to exhibit an overwhelming tendency to produce determiner–noun CS (1) (e.g. Liceras et al. 2008, Valdés Kroff 2016, Fernández Fuertes & Liceras 2018), we formally explore the directionality of the switch (2) and the type of implicit gender agreement mechanism in the case of Spanish determiner CS (3). (1) El hombre ha apagado el fire very quickly (2) a. la window (Spanish Det – English N) b. the ventana (English Det – Spanish N) (3) a. laF windowFeminine in SP / elM bookMasculine in SP (gender congruent) b. elM windowFeminine in SP / laF bookMasculine in SP (gender non-congruent)

We measured eye-movement data of 25 L2 bilinguals while reading code-switched structures (48 experimental sentences, 54 fillers and 54 distractors). Gaze duration, regression path duration and total duration were calculated for two target regions: determiner and noun. Our results show that Spanish determiner switches take longer to process than English determiner switches (2a vs. 2b), and that processing costs are higher in the case of gender non-congruent switches (3b). Results are linked to the L2 status English has for these bilinguals and to the strength of gender features in Spanish, their L1.

ABSTRACT. The reported study tested 80 non-native speakers of English and Russian in a self-paced reading task which used a cross-linguistic variation to attach the RC in (1) high in Russian and low in English (Fodor 2002).

(1) Bill saw / arrested the mother of the woman [RC that was speaking about herself / her in the yard]. This person was speaking about: (a) the mother (b) the woman

The stimuli manipulated perception and non-perception matrix verbs vs. the type of anaphora. Anaphora resolution is as a proxy for a language-specific preference in relative clause (RC) resolution. In Russian, the nearest c-commanding element to the anaphora is the NP [the mother], in English-like attachment – [the woman]. If binding principles are observed, there is a pattern ‘herself = the mother // her = the woman’ in Russian and ‘herself = the woman // her = the mother’ in English. This pattern means bottom-up sentence parsing. A perception verb (saw) in the matrix clause of a restrictive RC triggers an eventive interpretation: Bill saw (what?) [the mother of the woman’s talking in the yard], leaving only one grammatical doer of talking – [the mother]. In top-down parsing, a potentially possible eventive complement favors a switch to HA of the RC after a perception verb. The results show first signs of a combination of top-down and bottom-up strategies in L1 and L2 parsing. L2-parsing is L1-governed but there are first signs of the speakers’ sensitivity to language specific cues in the L2.

ABSTRACT. What is more relevant for L2 processing: Universal structuring principles or language-specific distributions? Sub-lexical frequencies play a crucial role in pre-lexical processing, providing a processing advantage for words of high phonotactic probability (Luce & Large 2001), and guide L2 processing as well (Lentz 2011). However, another factor for ease of non-native speech perception is sonority sequencing (Berent et al. 2007). We investigated the relative influences of target-language phonotactic probability and sonority sequencing on L1 and L2 perception. Our hypothesis, grounded in usage-based linguistics, was that target-language phonotactics have a greater effect than sonority. Moreover, we assumed that L2 listeners will also be influenced by their native phonotactics. In an identification-in-noise experiment, 22 Australian learners of German (mean age: 34.32) and 35 native speakers of German (mean age: 24.06) listened to German-sounding pseudowords (embedded in multi-talker babble) beginning with consonant clusters of varying frequency in German and varying sonority sequencing. Listeners' transcriptions were then analysed for accuracy of the initial consonant cluster. There was a significant facilitative effect of German, but not English, cluster frequencies in both listener groups, which was slightly stronger for the non-native listeners. Both groups showed better perception of sonority-violating clusters, conflicting with phonological predictions. The results indicate that speech perception is shaped by previous experience with phonotactic distributions and that this experience is categorized language-specifically. Optimal sonority sequencing does not facilitate perception. This supports the primacy of usage-based principles over phonological ones in speech processing.

ABSTRACT. Second language (L2) speakers experience syntactic priming, as they tend to re-use the syntax of recently experienced sentences to formulate their own sentences, and such priming can lead to L2 learning. While there is large within-group variation in L2 syntactic priming and the resulting learning, our understanding of why this is the case remains limited. In this study, we examined the role of task and individual characteristics. We compared priming on two syntactic structures and in conditions with and without lexical overlap. We also examined the effect of individual differences in attention, motivation and proficiency. Using picture description tasks, we investigated English L2 French learners’ primed production of passive and fronted temporal adverbial phrases. For both alternations, we manipulated between-subjects whether there was lexical overlap between prime and target sentences. We measured immediate priming (repeating a syntactic structure after a prime) and long-term priming (increase in target structures production in post-tests without primes relative to pre-tests). We assessed attention, motivation and proficiency with questionnaires. Without overlap, the learners experienced long-term effects of priming on their production of passive and fronted sentences, but only showed significant immediate priming for fronted sentences. Lexical overlap boosted immediate priming of both structures. Surprisingly, it also increased long-term priming of passives. The degree of priming varied with individual differences in attention but not motivation and proficiency. These results have important implications for models of priming.

ABSTRACT. In their native language, listeners use allophonic cues to facilitate word recognition. For example, vowels are nasalized before nasal consonants in English, and upon hearing a nasal vowel, English listeners predict an upcoming nasal consonant (Lahiri & Marslen-Wilson, 1991). A similar sensitivity to allophonic cues has been found in late L2 learners, suggesting successful learning of allophonic contrasts (Ito & Strange, 2009; Shea & Curtin, 2010). However, in both cases, these contrasts played no role in the listeners’ native language. Here, we focus on a case in which the L2 allophonic contrast maps onto a native phonemic one: in French, vowel nasality is phonemic and hence not predictive of a following segment (e.g., paix [pɛ] ‘peace’ - pain [pɛ̃] ‘bread’); do French learners of English nonetheless exploit allophonic vowel nasality to restrain lexical access in English? We examine this question using an offline gating paradigm with a 2AFC task and an online eye-tracking visual world paradigm task, testing both French learners of English and control native English listeners. Target stimuli are 20 pairs of English nouns matched in syllable structure and frequency. In each pair, words are phonemically identical until the end of the first syllable’s nucleus, which is an oral vowel in one word and its nasal correspondent in the other (e.g., butter - bunny). Gating task results suggest that the French group uses allophonic nasality to facilitate recognition of English words. Data collection for the eye-tracking task is ongoing, and we expect to have preliminary results in May.

References

Ito, K. & Strange, W. (2009). Perception of allophonic cues to English word boundaries by Japanese second language learners of English. Journal of the Acoustical Society of America, 125(4), 2348-2360.

Lahiri, A. & Marslen-Wilson, W. (1991). The mental representation of lexical form: A phonological approach to the recognition lexicon. Cognition, 38, 245-294.

Shea, C. E. & Curtin, S. (2010). Discovering the relationship between context and allophones in a second language: Evidence for distribution-based learning. Studies in Second Language Acquisition, 32, 581-606.

ABSTRACT. A major open question in sentence processing is how L2 learners combine multiple sources of information on-line [1]. Previous studies on adult L2 learners show that various types of information, namely lexical and discourse information, individually facilitate recovery from garden-path effects [2], but how different information types interact in L2 garden-path recovery is still unclear. Some approaches to L2 processing claim that discourse information is prioritized in L2 processing [3], while others identify the integration of multiple information types as compromised [4].

This study investigates if adult L2 learners integrate lexical and discourse information in recovery from temporary object-subject ambiguities (e.g., While the mother photographed/dressed the/a baby cried very softly). We tested 36 L1-German and 40 L1-Turkish learners of L2-English and 40 English natives in a self-paced reading task. Building on [5], we manipulated the embedded-clause verb to be either optionally-transitive (OPT, photographed) or reflexive-absolute-transitive (RAT, dressed) as well as the ambiguous noun phrase’s definiteness and givenness properties. In sentences with OPTs, natives reanalysed subject-like (i.e., definite/given) noun phrases more easily than object-like (i.e., indefinite/new) noun phrases. In sentences with RATs, however, they did not show any effects of discourse information due to the reflexive nature of RATs. German and Turkish learners also displayed sensitivity to verb subcategorization information, yet they did not demonstrate any modulating effects of discourse information. Rather than prioritizing discourse information, then, L2 learners, unlike natives, fail to incorporate multiple sources of information in on-line comprehension. We discuss implications for current approaches to L2 processing.

ABSTRACT. Several accounts propose that syntactic constraints are differentially used in native and non-native comprehension [1,2]. We tested this claim in the context of gender predictions using a visual-world paradigm. 70 native and 117 non-native German speakers (54 Spanish and 63 English) clicked on an object after hearing an instruction containing a possessive pronoun: “Click on his/her blue button”. German possessives have a complex pattern of gender agreement: their stem encodes possessor agreement (masculine vs. feminine antecedent) but their suffix encodes possessee agreement (masculine vs. feminine object). Crucially, only possessee agreement is relevant to predict an upcoming object. Our research question was whether the gender of the possessor would interfere with predictions, despite not being syntactically relevant. We manipulated the gender of the possessor, such that it matched or mismatched the gender of the target object. Using a bootstrapping procedure to identify the onset of predictions [3], we found that non-native speakers were slower to predict objects than native speakers. Further, both groups showed an interference effect, taking longer to start looking at the object in possessor mismatch configurations. This finding suggests that interference in the use of syntactic gender constraints presents a general challenge in predictive processing, even for native speakers. Among non-native speakers, the magnitude of the interference effect was comparable, regardless of whether their L1 encoded grammatical gender (Spanish vs. English), consistent with a general learner effect [4].

ABSTRACT. While crosslinguistic influence is widespread in bilingual speech production, the nature of representation in a multilingual system remains unclear. Prior work typically examines phonetically distinct yet phonologically similar sounds, for which phonetic convergence provides evidence for crosslinguistic links (Chang, 2015). Convergence is more challenging to assess when sounds are already phonetically similar, as with English and Cantonese initial long-lag stops. Here, we leverage the articulatory uniformity framework (Chodroff & Wilson, 2017; Faytak, 2018) to assess whether bilinguals share an underlying laryngeal feature across languages. Using a Cantonese-English bilingual speech corpus (n=34; Johnson et al., 2020), we identified prevocalic word-initial /ptk/ from force-aligned transcripts refined with AutoVOT (Sonderegger & Keshet, 2012). After accounting for speech rate, there were significant correlations for mean VOT values in English (3/3 pairs: r > 0.57), Cantonese (2/3 pairs: r > 0.54), and to some extent across languages (3/9 pairs: r > 0.55). These moderate correlations suggest some level of uniformity but are less compelling than prior findings. Additionally, there was low adherence to the expected ordinal relationship between /ptk/ means and within-talker inconsistencies across languages. A linear mixed-effects model highlights clear VOT differences across languages but also demonstrates that talker intercepts account for substantially more VOT variation than by-talker random slopes for place of articulation or language. This analysis indicates a role for both language and individual-specific factors in accounting for the structure of VOT variation (similar to speech rate findings: Bradlow, Kim, & Blasingame, 2017) and highlights the utility of the uniformity framework.

ABSTRACT. Previous research suggests that native listeners do not process typical syntactic errors as grammatical violations when they are produced in a foreign accent [1]. This implicit tolerance or “reduced sensitivity” has not yet been examined among non-native listeners, who may be privy to an interlanguage speech benefit [2]. That is, non-native listeners may be able to detect all violations produced by this accented speaker, regardless of error typicality. As such, the current EEG study examined the processing of foreign-accented syntactic errors by non-native listeners.

Thirty L1-English (L2-Spanish) participants listened to syntactically correct and incorrect Spanish sentences produced by a native Spanish speaker and a native English speaker. The violation in the incorrect sentences was caused by errors that are typical for L1 English speakers (i.e., gender errors; *la color ‘theFEM colorMASC’) or errors that are atypical for these speakers (i.e., number errors; *los color ‘thePLUR colorSING’).

Repeated-measures ANOVAs yielded a late P600 effect (900-1200 ms) for syntactic errors (main effect of Correctness: F=4.268, p=.019), with no differences according to error typicality (all ts<.5, ps>.68). Crucially, there were no interactions involving Accent, indicating that incorrect sentences were processed as violations in foreign-accented speech and native-accented speech alike. A re-analysis involving the L2-Spanish participants from the current study and L1-Spanish participants from previous research [1] showed that while non-native listeners are not sensitive to error typicality, native listeners’ processing is indeed modulated by typicality (p<.05). These findings support the existence of a syntactic interlanguage speech benefit for non-native listeners.

ABSTRACT. Previous research ([e.g., 1,3,4]) has demonstrated that highly proficient second language (L2) speakers of a gendered language, whose first language (L1) does not contain gender, use feminine and masculine gender cues to facilitate processing of upcoming nouns ([e.g., 2]). However, studies have not specifically focused on the impact of Manner of Acquisition (MoA, the way a language is learned). Canada is home to a wide variety of school-based French language programs, where a large portion of the population learns and acquires French as an L2. Yet, the way in which these speakers process their L2 when it is learned in various bilingual environments remains unclear. Seventeen native French speakers and 16 L1 English-L2 French bilinguals completed a lexical decision task using a masked priming paradigm in order to examine how a gender marked determiner (i.e., la) facilitates access to its corresponding noun (i.e., maison). L2 speakers were further grouped based on MoA (i.e., according to whether they acquired their L2 in a more naturalistic manner through all French schooling (FF, n=6) or in a more formalistic manner through an Early French Immersion program (EFI, n=10)). Native and FF speakers were found to pattern similarly in that they did not use gender information from masculine or feminine determiners to facilitate access to an upcoming noun of the same gender. Contrastively, EFI participants utilized the feminine gender cues facilitatively, but not masculine cues [p=0.043]. This suggests that more explicit learning of an L2 influences the use of gender cues during online processing.

ABSTRACT. Bilingual children regularly produce utterances in one language clearly influenced by the other. For example, bilingual children acquiring a Romance and a Germanic language have been found to produce ungrammatical N(oun)-Adj(ective) structures in their Germanic language (e.g., “apple green”) under influence from their Romance language (e.g., “pomme verte”) (Nicoladis, 2006). Such cross-linguistic influence (CLI) has recently been conceptualized as cross-language structural priming (e.g., Serratrice, 2016). We test this proposal by investigating (i) whether ungrammatical N-Adj orders observed in previous research are replicable using structural priming (Study 1&2), (ii) to what extent this cross-language priming occurs bi-directionally (Study 2), and (ii) whether its magnitude is modulated by language proficiency and use (Study 1&2).

Participants in Study 1 were 29 Spanish-Dutch bilingual 5-to-7-year-old children. In a within-language (Dutch-to-Dutch) priming task, children never produced ungrammatical N-Adj orders (e.g., “een dokter groen”), whereas after hearing the Spanish N-Adj prime (e.g., “un astronauta roja”) in the cross-language (Spanish-to-Dutch) condition, they did so regularly (129/696). Cross-language priming was thus found (Hervé et al., 2016), even in the children’s dominant language (cf. Hsin et al., 2013). Further analyses will determine which factors explain the considerable individual variation. In study 2 we examine whether – within the same group of French-Dutch bilingual children (n=29; age 4 to 8) – such cross-language priming effects are bidirectional.

The two studies in this paper extend the empirical basis of research on structural priming in bilinguals in two ways, namely by including (i) children, and (ii) individual differences variables (following Kidd, 2012).

ABSTRACT. Researchers are increasingly using eye-tracking to investigate second language processing (Felser and Cunnings, 2012; Jegerski and Sekerina, 2019). Few studies, however, have focused on how secondary tasks influence processing. Existing, albeit limited, research suggests that secondary tasks (e.g. comprehension questions and judgment tasks) modulate L2ers’ processing of morphosyntactic violations during self-paced reading (Leeser, Brandl and Weissglass, 2011).

The current study looks how secondary tasks influence the L2 processing of morphosyntactic violations using eye-movement data and mood-modality mismatches as a test case. Participants included 45 upper-intermediate-to-advanced English-speaking L2 French and 30 L1 French speakers. In Experiment 1, we recorded participants’ eye movements as they read 72 sentences, including 36 test items split across two conditions: subjunctive (grammatical) and indicative (ungrammatical). After 36 trials, participants answered a content-focused comprehension question, whereas in Experiment 2, assessed the acceptability of each sentence.

Mixed-effects models revealed that L1 speakers exhibited longer reading times at the critical region for indicative sentences (versus subjunctive sentences) for all four eye-tracking measures (first fixation duration, gaze duration, go-past time and total reading time), regardless of task type. L2ers, however, only exhibited longer reading times for indicative sentences (vs. subjunctive sentences) for gaze duration (a measure sensitive to semantic and syntactic anomalies, Rayner et al., 2004) in the judgment task, as revealed by a significant interaction between condition and task (t=-2.58, p>0.05).

These findings suggest that secondary tasks modulate L2ers’ processing of morphosyntactic violations using eye-movement data and that both tasks appear to activate different

ABSTRACT. Several models have been put forward to account for lexical access in monolingual speakers. While some consider that lexical items are processed as whole units (Butterworth, 1983), others claim that they are fully decomposed into morphological units (Taft, 2004), or that both routes are possible, depending on word familiarity (Burani & Laudanna, 1992). However, it is not clear whether those models can also account for bilingual morphological processing, and whether bilinguals modify their processing strategies as proficiency increases. Thirty Spanish monolinguals and 78 English L2 learners of Spanish (30 beginners, 22 intermediate, 26 advanced) completed a lexical decision task in which they indicated whether the word that was presented was a real Spanish word. Experimental words varied in number of morphemes (two morpheme words, e.g. abarcar and three morpheme words, e.g. achicar) and in surface frequency. Statistical analyses indicated that beginner and intermediate L2 learners presented similar reaction times in simpler words compared to those with three morphemes. Advanced L2 learners and Spanish monolinguals, on the other hand, were significantly slower at recognizing more complex words, compared to the morphologically simpler condition. Finally, word surface frequency did not seem to modulate reaction times in either group. Results suggest that in the early stages of L2 acquisition, bilinguals might process words in the L2 as wholes. However, bilinguals seem to modify their cognitive processing strategies as they further advance on the acquisition of a L2. More specifically, advanced L2 learners behave more native-like, as they tend to decompose morphologically complex words.

ABSTRACT. Language perception—particularly visual perception—studies on bilinguals often show that cognates (words that share form and meaning across languages) are easier to process than non-cognates (words sharing only meaning), which is referred to as the cognate effect. Although this effect has been demonstrated repeatedly for the visual modality (defining cognates through orthographic similarity), there is little evidence for it in the auditory modality. In this study, late highly proficient Spanish-English bilinguals carried out a lexical decision task in their second language—both visually and auditorily. Words were either phonological cognates (PC) or non-cognates and orthographic cognates (OC) or non-cognates (fully crossed). We also included perfect orthographic cognates (POC)—i.e., words that are visually identical in both languages. In the auditory modality, signal detection improved with PCs, but got worse with OCs. In the visual modality, OCs improved signal detection. Although OC and POCs had very little difference, POCs showed disproportionately large effects on response time. For the visual modality, this was a substantial decrease in response time and for the auditory modality, this was a substantial increase. This suggests that processing cognates is categorically different by modality, with OC in the auditory domain making word identification more difficult, as well as providing support for the idea that POCs are a special category within cognates.

ABSTRACT. Research in bilingual sentence processing has focused on the similarities and differences between native and non-native, late sequential bilinguals (L2 learners)1,2. However, differences between early bilinguals in adulthood to both adult L2 learners and monolinguals remain largely unexplored. Accounts emphasising age of acquisition as the differentiating factor1,3 predict similarities between early bilinguals and monolinguals. However, this issue hasn't been systematically explored in sentence processing. This study addresses this gap by investigating processing of which-questions using the visual world paradigm4. 107 adult participants (43 monolinguals, 35 early bilinguals, 29 L2 learners) heard which-questions whilst looking at two pictures depicting the different interpretations of the action conducted by the agents and themes, answering the question by selecting a picture. We manipulated question type (subject vs object which-question) and whether the two NPs matched or mismatched in number. Example sentences: (1) Subjects (number match/mismatch): Which donkey is carrying the zebra/zebras? (2) Objects (number match/mismatch): Which donkey is the zebra/are the zebras carrying? For comprehension accuracy, the early bilinguals and monolinguals were equally more accurate than the L2 learners. In the gaze data, the early bilinguals patterned with the L2 learners; both groups reoriented their gaze towards the correct image in the display slower than the monolinguals. This reflects slower processing which is not predicted under accounts postulating age of acquisition as the differentiating factor. All groups showed similar processing patterns as they had greater difficulty with object questions and benefitted from a mismatch in number between the two NPS in the object condition.

ABSTRACT. Comprehenders use real-world knowledge to consider possible messages a speaker might try to convey and linguistic knowledge of how phonetic, lexical, and morphosyntactic information constrain message realization. Native listeners (L1) integrate both types of knowledge in processes ranging from the resolution of lexical and syntactic ambiguity, to the anticipation of upcoming speech. This study uses eye-tracking to explore similarities and differences between first and second language (L2) listeners’ integrative and predictive sentence processing. L1 and L2 Mandarin Chinese speakers heard sentences that began with a specific location (e.g., “Look in the closet.”), followed by simultaneous presentation of four images and a spoken classifier-noun phrase (e.g., the classifier for long, thin objects and the noun ‘scarf’). Participants’ eyes were recorded as they searched for the target (scarf) among a classifier competitor that shared class membership and class-specific semantic properties (snake), a location competitor with location-specific semantic relevance (hat, which uses the headwear classifier), and a distractor (watermelon). The context and target either matched (closet-scarf) or mismatched (closet-snake) listeners’ location expectations. In match trials, L1 and L2 listeners looked to the classifier competitor (snake) and the location competitor (hat) only at chance levels, favoring the semantically-relevant and classifier-consistent target (scarf). In mismatch trials, however, L1 and L2 participants differed in their looks to the classifier competitor (scarf) and location competitor (hat) suggesting real-world and linguistic knowledge is used differently by L1 and L2 listeners and that such differences condition the time courses of recovery from incorrect anticipatory looks.

ABSTRACT. Working memory (WM) positively correlates with prediction of upcoming words in monolingual processing [1]. However, it is unclear whether WM also mediates prediction of word endings and whether WM modulates L2 prediction. Interpreting models include prediction and WM as key components for efficient language and processing control during interpreting [2]. Thus, interpreters could shed light on how language experience drives adaptation of language processing. We investigate the role of WM in the use of lexical stress and syllabic structure to predict verbal suffixes in Spanish. Spanish monolinguals, non-interpreter L2 learners of Spanish, and interpreter L2 learners of Spanish completed a visual-world eye-tracking task and a WM test. Participants listened to a sentence while seeing two words on the screen and selected the word contained in the sentence by pressing a button. For WM, participants listened to numbers and letters and organized first numbers in numerical order and then letters in alphabetical order. We controlled for two variables in target words: stress (initial stressed/unstressed syllable, BEbe/beBIÓ) and syllabic structure (CV/CVC, BEbe/CAMbia). A Growth Curve Analysis evaluating looks towards target words over the time course revealed that monolinguals and non-interpreters with higher WM span predicted better in the stressed condition, whereas those with lower WM span predicted better in the unstressed condition. Interestingly, interpreters with higher WM were conservative making predictions in the unstressed conditions. These findings indicate that language experience via interpreting adapts the allocation of cognitive resources during language processing [3] and inform bilingual control and prediction models.

ABSTRACT. Research on second-language (L2) acquisition has demonstrated that speech patterns of L2 production are influenced by speakers’ L1 phonological categories and that L2 speech performance is highly correlated with target language experience and phonetic context. Given that previous studies in this field often examined L2 production with isolated words in which it fails to reflect the real-time speech phenomenon, the present study explores the production of Mandarin high vowels in both word and sentence modes by beginning and fluent L2 learners and to evaluate the effect of consonantal context. To better understand the crosslinguistic influence in their speech productions, native Taiwan-Mandarin speakers were included. The production experiment was implemented through a repetition task in which participants were instructed to repeat target vowels in disyllabic Chinese phrases and in short simple sentences. Results of the study showed that, in terms of production pattern, L2 speakers displayed a robust influence from the L1, particularly in sentence utterances, whereas native speakers made distinct differences among each vowel category regardless of the speech style. As for the comparison of each vowel category across speaker group, a significant difference of /i/ productions was found between L1 and L2 groups, which contrasts with previous studies suggesting a very similar /i/ category for both languages. Additionally, even the fluent L2 Chinese speakers could hardly avoid producing /u/ with a fronted feature in sentence utterances. Through an examination of L2 production in different speech styles, this study will provide a further understanding of L1-L2 vowel interactions across languages.

ABSTRACT. Much of the work on how morphological constructions are processed shows that L1 speakers are sensitive to subtle information, such as the frequency of the relationship between the head and the modifier in a compound. For example, L1 speakers take more time to process mountain magazine (magazine ABOUT mountains) than mountain stream (stream IN a mountain) because the location relation (IN) is more common for the modifier (mountain) than the ABOUT relation (Gagné & Shoben, 1997). Are L2 speakers also sensitive to this kind of information? Several existing frameworks predict that this sensitivity should be absent in L2 processing, either because it cannot generate expectations based on the modifier (RAGE hypothesis; Grüter et al., 2014), because the relation is not processed (SSH; Clahsen & Felser, 2006) or because L2 processing is too demanding (e.g. Hopp, 2010). We replicated experiment 1 of Gagné and Shoben (1997) with L2 English speakers whose L1 is Mandarin Chinese. Participants (N=39) performed a sense/nonsense judgment task (48 practice items; 57 critical items; 57 nonsense fillers). Fillers were nonsense items; critical items made sense and were divided into three conditions (HH, HL and LH) depending on whether the relation was used frequently (H) or infrequently (L) with either the modifier or the noun. Reaction times and accuracy were analysed. In our preliminary analysis, we found no effect for accuracy and a main effect of condition for reaction times (p=0.027). Consistent with the original experiment results, this suggests that L2 processing is sensitive to relation frequency information.

ABSTRACT. English articles are known to be problematic for L2 users of English and their use has been associated with L1 transfer.L1 transfer in the article system in L2 English is generally more prominent in bilingual adults than in children. However, little research has been done on the performance on Arabic-English bilingual children acquisition of English article. The current study examined similarities and differences in the acquisition of the English article system by L1 Arabic- L2 English bilingual adults and children (6-10) and the effect of learning English articles in the L1. Specifically, we focused on analysing patterns of article use in the L1 and L2, as well as the processes of L1 transfer on L2 (forward transfer) and backward transfer (L2 to L1 transfer) of English. The study also examined how factors such as age of acquisition, age of onset of L2, duration of exposure to L1 versus L2 and L1, and L2 proficiency affects these patterns. In the study, different tasks were employed that included mainly the use of Arabic/English narratives and Arabic/English repetition tasks. For comparison purposes, two control groups were added: monolingual Arabic speakers and monolingual English speakers groups. Preliminary results from the narrative task, it was revealed that bilingual adult’s use of English articles exhibited L1- Arabic transfer which was apparent in the significant amount of omission the indefinite article a/an. Children’s use, however, exhibited a general developmental pattern, often found as an intermediate stage among L1/L2 children, in overusing the English article the.

ABSTRACT. Processing Instruction (PI) is an effective input-oriented and meaning-based pedagogical intervention to grammar instruction with consistent, durable and transferable effect (Lee & Benati, 2009). Ito & Wong (2019) showed that both auditory and written PI are effective, but auditory PI does not immediately enhance learners’ sensitivity to grammatical cues. We know much less about whether PI delivered in written input modality only would be equally as effective for third language (L3) learners of English as for L2 learners of English (both school-age learners). Hence, we aim to find out a) whether there are any differences on how L2 English learners with L1 German and bilingual Albanian-Macedonian learners of L3 English process the third person present simple singular -s (primary) and the past simple -ed (secondary), b) whether any significant instructional effects of the written PI intervention will be attested for both auditory and written interpretation assessment tasks and c) whether the effects can be retained over time. Our results demonstrate the effectiveness of the PI intervention for both features (primary and secondary) and for both experimental groups (L2 and L3) with a stronger intervention improvement effect for the L2 group regarding the secondary feature. There was no intervention effect for the control groups.

ABSTRACT. Foreign-accented speech of second language learners is often difficult to understand for native listeners of that language. Part of this difficulty has been hypothesized to be due to increased within-category phonetic variability in non-native (L2) speakers’ productions. However, until recently, there have been few direct tests for this hypothesis [1-2]. We compare the realization of vowels and word-final stops in productions of native-English L1 speakers and native-Mandarin speakers of L2 English. With the largest sample size to date (10 speakers per accent group; 7404 vowel tokens and 4176 stop tokens), we show that at least proficient non-native speakers exhibit little or no difference in category variability, compared to native speakers. Our results relativize a common (and a priori plausible) assumption that competition between L1 and L2 representations necessarily lead to increased variability—or, equivalently, decreased precision, compactness, consistency, and stability—of non-native speech. L1-L2 competition without doubt affects various aspects of L2 production, but it does not seem to reduce articulatory precision. Instead the effects of non-nativeness on category variability are category- and cue-specific. Together with other recent work [1-2], this casts doubt on the assumption that non-native speakers are inherently more variable in production. Intriguingly, the same speakers show substantial deviations from native speech in category means, as well as the correlations among cues they produce. They also consistently show less separability between neighboring categories (e.g., /i/-/ɪ/ or /d/-/t/). We discuss the potential consequences of these characteristics of L2 speech on the development of L2 representations and L2 listening.

ABSTRACT. The retrieval of pronoun antecedents plays a central role in language processing. Reflexive anaphors, according to the Condition A of Chomsky’s (1981) Binding Theory, must be linked to a local c-commanding NP that matches the reflexive in gender, person and number. L2 acquisition studies try to establish whether non-native speakers can acquire these requirements in the target language. Offline tasks mostly reveal native-like responses; however, online tasks have shown that non-native speakers initially also take into account candidate antecedents that are not allowed by Condition A, but have a prominent structural and discourse role (e.g. eye-tracking studies on L2 English, for German- and Japanese-speaking learners; Felser & Cunnings, 2012; Felser, Sato, & Bertenshaw, 2009). Looking at an additional language, we studied online reflexive anaphora resolution in Italian. A group of 96 native speakers and 96 L1 Croatian/Bosnian/Serbian highly proficient non-native speakers participated in a self-paced reading task. The participants read 36 test sentences, which included two candidate antecedents for the reflexive pronoun se stesso/stessa ‘himself/herself; however, only the second antecedent was locally bound, and it was thus the only choice allowed. All sentences were followed by a comprehension question that targeted the pronoun antecedent. Non-native speakers’ reading was slower overall, but their offline interpretation of reflexive pronouns was native-like. In both groups, the interpretation was slower for structurally prominent (c-commanding) candidate antecedents. This possibly indicates that both native and non-native speakers initially consider inappropriate (non-local) antecedents. Such results point to non-native speakers’ native-like sensitivity to structural cues during online processing.

ABSTRACT. Code-switching is fairly frequent in bilingual discourse. Despite its frequency, psycholinguists have found that its processing is costly. Sociolinguistics, on the other hand, has uncovered numerous functions for CS, suggesting that code-switching provides benefits to bilinguals which override the costs. Informed by the patterns found by sociolinguists, such that taboo words often follow a code-switch, we test whether code-switch processing can reduce the emotionality effects of taboo words.

We use eye-tracking while reading to compare the reading times of taboo vs. neutral words in three conditions: 1) Spanish unilingual sentences, 2) English unilingual sentences, and 3) after a code-switch in Spanish-English code-switched sentences. Data from Spanish-English bilinguals with predominantly reversed dominance (n = 30) suggest that they show a significant emotionality effect in English, their dominant language, in early (Gaze Duration) and late measures (Total Duration). Nevertheless, they exhibit a significant reduction of the emotionality effect in both Spanish and Code-switched sentences to a similar degree, despite the critical words being in English, their dominant language, in the Code-switched condition. We thus believe that processing a code-switch reduces the emotionality to the same level or beyond the use of the less dominant language.

The emotionality reduction for subsequent taboo words represents a strong motivation for the use of CS, despite the costs of its immediate processing. We advocate for shifting the focus from code-switch costs to the benefits it may provide downstream. This novel approach stands to provide a fresh, more complete picture of code-switch processing.

ABSTRACT. Research findings on L2 morphological processing vary widely: Some studies suggest that L2 speakers decompose morphologically complex words in the same way as native speakers do (Diependaele et al., 2011); others attribute masked priming found for derived (but not inflected) words to activation between related lexical entries (Kɪrkɪcɪ & Clahsen, 2013); and yet others suggest that L2 learners rely on orthographic overlap rather than morphological relatedness (Heyer & Clahsen, 2015). The present study teases apart the contributions of morphology and orthography to masked priming effects in a novel way, using hyphens to emphasize shared letters in morphologically or orthographically related words visually (e.g., ask-ed vs. ask-ew). In a masked priming experiment, 50 L2 speakers of English were asked to make lexical decisions on verbs (ASK) preceded by both morphologically and orthographically related primes, each spelt normally or with a hyphen: (1) inflected normal (asked), (2) inflected hyphenated (ask-ed), (3) orthographic normal (askew), (4) orthographic hyphenated (ask-ew) and (5) unrelated (scala). LMER analyses revealed priming following all four prime types compared to unrelated primes; t>2.12). Crucially, neither inflected hyphenated (t=1.31) nor orthographic normal (t=1.63) primes led to significantly more or less priming than inflected normal ones, while a hyphen in orthographic primes significantly reduced facilitation (t=2.75). These results indicate that visual accentuation of the shared letters in morphologically and orthographically related words leads to facilitation, yet only when hyphenation indexes morphological decomposition. We conclude that L2 processing is not fully dependent on orthography and relies on morphological units, which contribute meaning.

ABSTRACT. The cognate facilitation effect causes bilinguals to process cognates faster and more accurately compared to non-cognates [1,2]. Degree of orthographic and phonological overlap have been shown to modulate the cognate facilitation effect [3,4]. Our investigation used a fine-tuned 3-level cline of morphosyntactic and phonological similarities of Spanish-English cognate verbs:

(i) true cognates have the closest phonology and syntax. Representar/represent have a Levenshtein distance of 0 (ignoring infinitival morphemes) and share both transitivity and argument structure (measured by verb frames). (ii) lexemic cognates have close phonology with differing syntax. Condescender/condescend have a Levenshtein distance of 0, but differing transitivity in English and Spanish. (iii) lemma cognates have close syntax with differing phonology. Estipular/stipulate have a Levenshtein distance of 4 but the same transitivity and argument structure.

We investigated degree of cognate facilitation using self-paced listening. Ten simultaneous and late bilingual participants heard Spanish and English sentences blocks with cognate and non-cognate verbs. Sentences presented 4-5 segments each, and RTs were collected from button presses to hear proceeding segments. RM ANOVAs showed a significant interaction of cognate status, language and L1 (p<.05), and a significant interaction of cognate type, language and L1 (p<.05). Post-hoc analyses indicated results were driven by differences between true and lemma cognates for L1 Spanish-speakers and L1 English-speakers (Spanish & English), true and lexemic cognates by L1 Spanish-speakers and L1 English-speakers (Spanish), and lemma and lexemic cognates by L1 Spanish-speakers and L1 English-speakers (English) and L1 Spanish-speakers (Spanish). Results suggest morphosyntax with L1 modulates the cognate facilitation effect.

ABSTRACT. Number transcoding is the cognitive task of converting numbers between analogic, verbal and visual symbolic codes. Visual to verbal symbolic transcoding, i.e. number reading strongly relies on language proficiency. How do bilinguals transcode two-digit numbers? We evaluated four age groups: 5th, 8th, 11th graders and adults of German-French bilinguals from Luxembourg transcoding in German and French. In the Luxembourgish educational system, children learn mathematics in German (LL1math) until the 7th grade, and then the language of mathematic learning (LL2math) switches to French. German two-digit number words follow a unit-decade order while French has a decade unit order. Moreover, French number words above 60 are in base 20, which is not the case in German. Transcoding was evaluated with a symbolic visual to verbal production task and a verbal to symbolic visual recognition task. The results for the production task show for all age groups a consistent cost for numbers above 60 in French and a cognitive cost in LL2math transcoding. Recognition also reveals a cognitive cost for numbers above 60 in French for all age groups, but the cognitive cost in LL2math only manifests for 5th and 8th class. In conclusion, we replicate a language-effect for numbers above 60 related to the French base 20 system. Moreover, transcoding is generally less well mastered in LL2math. This is particularly persistent over time during the production task, while performance in both languages become similar in the recognition task. This study supports the link between numbers and language, especially during number production.

ABSTRACT. Quebec francophones substitute /td/ for English /θð/ (think that → [tɪŋk dæt]), presumably because perceptual assimilation generates non-targetlike underlying representations (Best, 1994). Nonetheless, francophones sometimes produce targetlike output, which is hard to reconcile with inaccurate representations. Possibly, variation derives from dual underlying representations (URs): when learners overcome perceptual assimilation, they revise lexical entries with targetlike representations without overwriting initial URs, both remaining available for output (John & Cardoso, 2017).

For our study, 50 francophones performed a reading-aloud task to test for variable /θð/ production. Variable producers, being beyond perceptual assimilation, should show accurate /θð/ perception, but also lexical confusion for tasks involving items with dual URs.

To test these predictions, participants completed two auditory tasks. First, 450 trials in an oddball paradigm, with easy (fan-pan) and difficult (thank-tank) contrasts to verify perceptual assimilation. Second, 150 sentences with easy/difficult substitutions. We expected variable producers to encounter greater difficulty with incongruous sentences containing difficult substitutions they themselves make (θ→t This movie is better than I taught) than the reverse (t→θ The soccer game ended in a thigh), only the former being confounded by dual URs.

ANOVA results revealed significant differences (p<.001) between easy/difficult contrasts on both auditory tasks, including for variable producers, a finding not consistent with their having overcome perceptual assimilation. Nonetheless, significant differences between /θ/ conditions in the second task support the dual UR proposal. Unexpectedly, both tasks showed considerable range in accuracy (8-100%). To account for these discrepancies, we suggest learners develop alternate strategies for L2 perception and production.

ABSTRACT. An early start into second language learning is often considered important for a future mastery of the target language. But it has also other advantages. For instance, the presence of two languages in the mind has been shown to be beneficial to general cognitive skills. Bilingual children demonstrate superior abilities in control over attention and switching between rules. Young bilinguals also show a less egocentric approach in solving perspective-taking problems, regardless of their competence in the L2. We recruited three groups of monolingual children aged 4-5 from reception classes in Primary schools in England. One group (N=32) attended bilingual schools, another group (N=29) had weekly L2 classes and a third control group (N=38) did not have any L2 education. Groups were matched by SES, extra-curricular activities, verbal ability and non-verbal reasoning skills. We tested children in tasks of switching and perspective-taking in the first few months of the school year and 24 weeks later. The two groups of L2 learners outscored the control group both in switching and perspective-taking at the second (but not first) testing point. Pupils attending bilingual schools showed an additional enhanced progress in their switching abilities. These results suggest that even a limited weekly exposure to a L2 has an impact on children’s socio-cognitive development, but the amount of language exposure increases cognitive flexibility.

ABSTRACT. Code-switching is highly socially constrained and is only felicitous when those present in the context are proficient in both languages. Bilinguals therefore need to dynamically adjust their language control and expectation of code-switching to the social situation. The aim of the present EEG study was to investigate how and when language control in the comprehension of code-switches is affected by the language knowledge of others present, established via a referential communication task with a confederate. Spanish-English bilinguals read sentences with and without code-switches together with another Spanish-English bilingual or with an English monolingual. Switches elicited an early fronto-central positivity, but only when a monolingual was present at the start of the study. In addition, switches elicited a Late Positive Complex (LPC) relative to control words. This LPC switch effect was smaller when a bilingual vs. a monolingual was present for those participants who indicated that their monolingual partner did not understand the code-switched sentences. Time-frequency analyses showed a gamma-band power decrease for switches and an overall increase in alpha-band power when a bilingual was present, likely reflecting socio-pragmatic meaning integration and dynamic language control, respectively. These findings suggest that the bilinguals in our study expected and activated both languages when initially paired with a bilingual and that they more easily accommodated code-switches in the presence of a bilingual versus a monolingual.  Our results support the view that language control is modulated by the perceived language knowledge of others present and support a dynamic control model for bilingual language comprehension.

ABSTRACT. Asymmetric switch cost typically refers to when code-switches in the dominant L1 incur a greater processing cost than L2 code-switches. This asymmetry has been found in the auditory comprehension of Spanish and English code-switches (Olson, 2017). Mandarin and English code-switches have not yet been compared in auditory comprehension, and are likely to exhibit an asymmetric switch cost due to differences in the frequency of their use in speech: Inserting Mandarin code-switches into English sentences is infrequent, but inserting English code-switches into Mandarin sentences is frequent (e.g., Lu, 1991). The current study investigates the roles of dominant language and frequency in a potential asymmetric switch cost in the auditory comprehension of Mandarin and English code-switches. Forty-eight Mandarin-English bilingual listeners who self-reported as habitual code-switchers participated in an eye-tracking task that auditorily presented Mandarin and English sentences with either Mandarin or English target nouns (i.e., code-switched or unilingual stimuli). Results using growth curve analysis show an asymmetric switch cost associated with frequency: While the infrequent Mandarin code-switched words incurred a switch cost relative to unilingual Mandarin words (Est = -.08, SE = .02, p < .001), the more frequent English code-switched words were actually easier to recognize than unilingual English words (Est = .06, SE = .01, p < .001). Participants’ language dominance had no significant effect on switch cost asymmetry, but did affect processing ease. These results are discussed in the context of the experiment environment – an English-speaking university – and different ways of operationalizing language dominance.

ABSTRACT. The goal of the study is to investigate the role of sub-lexical units in cross-language and cross-modal lexical access. We ran two experiments on parallel activation in spoken Spanish and in Spanish Sign Language (LSE) in 56 hearing bimodal bilinguals using the visual world paradigm. Experiment 1 investigated parallel phonological competition in Spanish from words sharing onset or rhyme while seeing LSE signs. Experiment 2 investigated parallel competition in LSE from signs sharing handshape or location while hearing Spanish words. The results showed co-activation of the spoken language in a signed context, and vice versa, co-activation of the signed language in a spoken context. In Experiment 1, this was shown through word onset competition but no rhyme competition. In Experiment 2, location and handshape both showed competition, with location competition preceding handshape competition. These findings demonstrate that bimodal bilinguals experience bidirectional co-activation between languages that do not overlap phonologically. Furthermore, the findings suggest an important role for onsets in spoken word coactivation and location in sign co-activation, possibly reflecting the temporal dynamics of the linguistic signal (i.e., incremental processing of words) and perceptual salience (location is more salient than handshape). This suggests important parallels between within-language and cross-language lexical access.

ABSTRACT. The last four decades of research on bilingual lexical-semantic representation and processing point towards an integrated bilingual lexicon with nonselective lexical access. However, the interplay between speaker- and word-level factors and their role in shaping the lexicon and its functioning is less well understood. The most prominent theoretical models differ in their conceptualization of bilingual semantic representations (localist vs distributed models). This project investigates bilingual lexical-semantic representation and processing employing cross- language visual priming. In particular, we focus on the role of word frequency and language use as continuous proxies of subjective frequency, and explore potential modulations of bilingual lexical processing as a function of relative semantic overlap. Furthermore, we novelly investigate the impact of executive functioning on cross-language priming effects.

We tested 200 highly-proficiency Spanish-English bilinguals with different degrees of L1/L2 use in two cross-language priming lexical decision tasks with more than 400 word pairs, as well as a dimensional change card sort task. The large number of observations, together with a conservative analysis (α = .01 for main effects) with mixed-effects models (Baayen, 2008), yield a robust data set to explore several speaker- and word-level variables simultaneously.

We found significant effects of prime frequency and executive functioning in both translation directions. Significant effects of concreteness and language use were only present in the L1-L2 data. We discuss these results in light of theoretical models of bilingual lexical-semantic representation and processing, and suggest new directions for future research.

ABSTRACT. This study examined how relative language proficiency and exposure influence the development of executive function (EF) in 7-12 year-old bilingual ‘returnee’ children. Returnees are children of immigrant families who were immersed in an environment where their second language (L2; English) was the majority societal language and returned to their native language (L1; Japanese) environment after the period of prolonged, naturalistic L2 exposure. Targeting this population allows us to address the question of how the loss of opportunities to engage in bilingual activities may longitudinally affect EF development. We administered EF inhibition (Simon) and monitoring/updating skills (N-back) tasks shortly after the children’s return to their L1 environment and again one year later. The results from the background questionnaire (BiLEC) showed that the average L2 English exposure decreased by 42.3% (from 46.8% to 4.5%) after a year of re-immersion in the L1 Japanese environment (Table1). Critically, the findings from the mixed effect model showed that the amount of reduction in L2 exposure (i.e., the difference in L2 exposure when they lived in an L2 majority language environment vs. back in the L1 environment) affected children’s monitoring and updating abilities. The greater reduction the children experienced in L2 exposure, the smaller their improvement was on the updating task in the second interval (Figure 1). However, relative proficiency at the onset of return did not influence children’s development in inhibition ability. The finding suggests that losing access to one’s L2, that is, less active bilingualism is associated with attenuated effects in EF development.

ABSTRACT. The language sciences as a whole and bilingualism as part of them are in need of ashared and quantified theoretical framework that gives researchers the freedom toformulate their own questions, but at the same time furthers cross-disciplinaryintegration. After discussing general requirements for such an approach, I will showhow the computational model Multilink+ implements these for monolingual andbilingual word processing with respect to different stimulus materials, tasks, andparticipant populations. I will discuss the potential of the model’s lexicon (recentlyextended in size and to six languages), the necessity of lateral inhibition, and theusefulness of various types of semantic representations. Next, I will apply the modelto several new empirical studies. These include an orthographic and semanticpriming study involving cognates, translation equivalents, orthographic neighbors,and unrelated control words (Dijkstra et al., submitted), and word translationproduction studies with cognates (Pruijn et al., in prep.) and interlingual homographs (Goertz et al., in prep.). I will conclude the theoretical lecture with my DIY list,including wishes and concerns with respect to, for instance, the (dis)advantages of asymbolic approach and the model’s potential generalization to languages with nonalphabeticscripts and to second language acquisition. If time allows, a brief practicaldemonstration of Multilink+ will be given, in the hope of igniting a lively discussion.