ABSTRACT. The majority of the world population is bilingual or multilingual. In this talk, I will argue that accommodating multiple communicative codes is the prototypical state of the human brain, and will do so by introducing a new book that considers the power of language from the unique vantage point of bilingualism and multilingualism. I will present eye-tracking, mouse-tracking, EEG, and fMRI evidence that the continuous co-activation and interaction of multiple languages shapes the human linguistic, cognitive, and neural function. Being bilingual or multilingual results in profound changes not only to how we use language, but also how we perceive the world, what we remember, how we learn, our creativity, emotions, decision making, and identity at both the individual and society levels. I will conclude with a call for placing the study of language-mind interaction and multilingualism among the core areas of scientific investigation in the next decades if we are to gain an accurate understanding of humanity’s potential.

ABSTRACT. The processing of linguistic dependencies has informed debate about native (L1) and non-native (L2) sentence processing [1,2,3,4]. Some posit L1/L2 differences based on L2ers' difficulty in computing syntactic structure [1], while others describe L1/L2 differences in terms of working memory demands [2,3,4]. Whether L2ers violate constraints on linguistic dependencies during processing has been important in assessing these theories, but existing research has revealed mixed findings [e.g. 5.6]. We contribute to this debate by examining subject-verb agreement and reflexives. We are unaware of any existing study that has examined these two dependencies in the same L2 group. Two factors were investigated in each dependency; sentence grammaticality and ‘interference’. For agreement, number match between the subject and the verb was manipulated to construct grammatical (1a-b) and ungrammatical conditions (1c-d). Interference was tested by manipulating the number properties of a distractor, which cannot grammatically enter into an agreement dependency with the verb. That is, (1c) is ungrammatical but the similarity between the distractor “patients” and the verb in number may attenuate the ungrammaticality effect, resulting in interference manifested as reduced reading times in (1c) relative to (1d). The same factors were examined in reflexives, but with a gender manipulation (2a-d).

In study 1, 176 L1 English speakers and 176 intermediate-advanced Arabic L2 English speakers completed a word-by-word web-based self-paced reading (SPR) experiment. 24 items were included for each dependency. Results (Fig. 1-2) revealed main effects of grammaticality and significant grammaticality by group interactions in both dependencies, with larger grammaticality effects for L2ers in reflexives but for L1ers in agreement. There was also a main effect of distractor in reflexives, with shorter reading times in conditions with matching distractors. We did not find any significant interaction with distractor and group in both dependencies.

In study 2 (conducted 6-11 months after study 1), number was manipulated in reflexives, as in (3). As L2ers in study 1 showed smaller grammaticality effects in subject-verb agreement compared to reflexives, study 2 examined if this is due to the number feature manipulation itself, or whether it is dependent on dependency type. 164 L2ers, who had previously participated in study 1, and 164 L1ers completed a web-based SPR experiment with conditions like (3a-d). Results (Fig. 3) showed a main effect of grammaticality,which did not significantly interact with group. This suggests that the smaller grammaticality effect in subject-verb agreement for L2ers in Study 1 may be due to the dependency rather than number per se.

These findings suggest both groups applied similar parsing mechanisms during processing. Although the grammaticality effect was smaller in L2ers for subject-verb agreement, an additional analysis with proficiency indicated clearer grammaticality effects for higher proficiency learners. Contrary to some previous L1 studies, especially on subject-verb agreement [7,8], no consistent evidence of interference was detected during processing. This should be taken with caution however, as we found a significant effect of distractor in study 1 in reflexives. Together, the results imply that L2ers compute syntactic dependencies similarly to L1ers, and potential differences might be driven by individual differences.

ABSTRACT. Previous studies on L1 morphosyntactic attrition have investigated immersed late bilinguals with extensive L2 exposure (e.g., Chamorro et al., 2016; Tsimpli et al., 2004). However, little is known about the potential changes experienced in the L1 of immersed bilinguals during the early years of L2 exposure. Moreover, the L1 (in)stability of instructed late bilinguals has also been overlooked. Thus, our study will fill these gaps by exploring how the L1 of L1 Spanish-L2 English immersed and instructed bilinguals might change in the processing and comprehension of null/overt subject pronouns, a domain shown to be vulnerable for attriters (Chamorro & Sorace, 2019; Gürel, 2019). 207 L1 Spanish-L2 English participants (33 beginners, 80 advanced instructed bilinguals in Spain, and 94 immersed bilinguals in the UK) completed a picture selection task (PST) and a self-paced reading (SPR) task testing the predictions from the Position of Antecedent Strategy (Carminati, 2002), i.e., whether null pronouns are coreferential with subject antecedents and overt pronouns with object antecedents. Data were also collected from a background questionnaire (Bilingual Language Profile, Birdsong et al., 2012) providing a continuous dominance score. Firstly, the PST (adapted from Tsimpli et al., 2004) contained 20 experimental items where a main clause with two potential same-gender antecedents in subject and object position respectively was followed by a subordinate clause with a null or an overt pronoun. Participants selected which picture (subject vs. object bias) best matched their interpretation of the sentence. Secondly, the experimental stimuli for the SPR task were the same as the ones used in the PST. The main experimental manipulation in this task was the presentation of each sentence with a matching or unmatching picture which would bias towards either a subject or an object interpretation for the null and overt pronouns in the subordinate clause. The data from the PST and SPR tasks were analysed using (generalised, for PST) linear mixed-effect models in R (Bates et al., 2015), and the final model included Pronoun (overt/null), the BLP score and Group (beginners/instructed-advanced/immersed-advanced) and their interactions as fixed effects, and the maximal converging random-effect structure allowed by the design (Barr et al., 2013). The results from the SPR task show that none of the two advanced bilingual groups showed sensitivity to pronoun mismatches in online processing, whereas the beginner group took significantly more time to read overt pronoun sentences which were forced to bias towards a subject interpretation (Figure 1), in line with Chamorro et al. (2016). Additionally, the results from the PST indicate that the two advanced bilingual groups differed from beginners only in the overt pronoun condition, and that the more dominant instructed and immersed participants were in the L2 (Figure 2), the more flexible their interpretation of the overt pronoun was, biasing more towards the subject (Chamorro et al., 2016). These results shed light on the unexplored initial stages of L1 morphosyntactic attrition in immersed bilinguals, provide new evidence of L1 instability in advanced instructed bilinguals and call for the use of continuous measures to investigate gradience in bilingualism.

ABSTRACT. Second-language (L2) learners can experience difficulties in the production of L2 sounds. These difficulties can be remediated by training, where participants receive visual feedback on their production, based on the acoustic analysis of their speech. As little as 3 days of training can improve L2 learners’ production by 20% 1,2. However, little is known about the transfer of articulatory training-related gains in production to perception and the mechanisms underlying improvements in L2 production. This study aimed to investigate whether training-related gains in the production of non-native speech sounds transfer to perception. Furthermore, it explored whether articulatory training alone could lead to neural plastic changes as indexed by the mismatch negativity (MMN) signal. Previous research suggested that phonemic learning is accompanied by an increase in the MMN magnitude 3. 48 native Spanish speakers, unfamiliar with French, were trained to produce a French vowel contrast /e/ - /ԑ/, which is difficult for native Spanish speakers, as both vowels are assimilated to the single native Spanish /e/ 4. The experiment spanned over 5 days; day 1 was used to establish a baseline for the neural response to the non-native vowel contrast and day 5 was to measure the impact of training on the MMN signal. On both days, we recorded participants’ processing of the nonnative contrast using electroencephalography (EEG). Over days 2-4 the articulatory visual feedback training took place. During training, on each trial, participants heard the target non-native sound and had to repeat it; immediately after that, participants received visual feedback on the screen displaying the position of the participants’ vowel, in the acoustic space, along with the target vowel space, derived from the productions of native French speakers. There were 1000 training trials per vowel, in total. On day 2 – before the training – and day 4 – after the training – a vowel repetition task and an AXB discrimination task were used to measure participants’ production and perception of the trained sounds, respectively. The results indicated that articulatory training led to improvements in the production of non-native vowels (X2 = 43.62, p < .0001), though this improvement did not transfer to the perceptual discrimination of these vowels F (1, 39) = .08, p = 3.83. In regards to the neural plasticity, the results revealed that there was no statistically significant change in the MMN mean amplitude from before to after articulatory training (p > .1), suggesting that articulatory training alone did not lead to neural plastic changes, as indexed by the MMN signal. Combined with previous research, these results suggest that L2 production and L2 perception are weakly related in novice learners but might align with experience 5 and indicate that training-related improvements in L2 vowel production might be purely articulatory at the onset of learning but can transfer to perception with more experience.

ABSTRACT. Trilinguals’ L1 and L2 phonetic systems influence L3 phonetics resulting in either hybrid L1/L2 transfer (Llama & Cardoso, 2018; Wrembel, 2011) or primarily L2 transfer (Llama et al., 2010; Wrembel, 2010). Regressive transfer – the influence of L3 phonetics and phonology on either L1 or L2 – has only recently started attracting attention and renders inconclusive findings: Two studies found no L3-to-L2 influence (Cabrelli Amaro, 2013; Gut, 2010), while another two studies reported stronger L3-to-L2 than L3-to-L1 influence (Cabrelli Amaro, 2017; Cabrelli Amaro & Rothman, 2010). The latter two studies led to the Phonological Permeability Hypothesis, which states that an L2 acquired post-puberty is more vulnerable to phonological L3 influence than the L1.

The present study investigates phonetic and lexical regressive transfer in highly-proficient Spanish-Basque early-bilingual adults with L3 English, who acquired Spanish from birth, Basque at around two years of age and English in later childhood at school. These speakers are immersed in a Spanish-Basque bilingual society where both Spanish and Basque are supported in the educational system. Participants performed a speeded trilingual switching task, in which they had to name pictures either in Spanish, Basque, or English (as indicated by a flag) within 900 ms. We analysed non-switch trials in L1-Spanish and L2-Basque, that is, those trials that were preceded by a trial in the same language. We measured English lexical intrusions in 180 Spanish and 180 Basque non-switch trials per participant. In a subset of 108 of these 180 trials per language, we measured participants’ VOT production in voiceless plosives, which take short lag VOT in Spanish and Basque but long lag VOT in English. In a control picture naming task blocked by language, participants produced VOT of 26 ms in Spanish, 27 ms in Basque, and 39 ms in English.

According to the Phonological Permeability Hypothesis, we hypothesized that L3-English has a stronger impact on L2-Basque than on L1-Spanish, resulting in longer VOT in Basque than in Spanish in the speeded trilingual switching task. If the L3 generally transfers more strongly into the L2 than into the L1, this should also result in more L3-English lexical intrusions in L2-Basque than in L1-Spanish. Results of the first 24 participants (final N=50, data collection expected to finish in April 2022) support both predictions. Participants’ VOT in the speeded trilingual switching task was longer in Basque (M=33 ms) than in Spanish (M=24 ms; p<.001), which was not the case in the control picture naming task. Participants also produced more L3-English lexical intrusions in L2-Basque (M=7.53) than in L1-Spanish (M=4.53; p=.008). This supports the Phonological Permeability Hypothesis and suggests that this account can be extended to trilinguals with a very early-acquired and highly-proficient L2. Moreover, this study shows that the L3 lexicon transfers more strongly into the L2 than into the L1, suggesting that the stronger connection between the L3 and the L2 is not restricted to the sound-level.

  References Cabrelli Amaro, J. (2013). The Phonological Permeability Hypothesis: Measuring regressive L3 influence to test L2 phonological representations [Unpublished doctoral dissertation]. University of Florida. Cabrelli Amaro, J. (2017). Testing the Phonological Permeability Hypothesis: L3 phonological effects on L1 versus L2 systems. International Journal of Bilingualism, 21, 698−717. Cabrelli Amaro, J. & Rothman, J. (2010). On L3 acquisition and phonological permeability: A new test case for debates on the mental representation of non-native phonological systems. International Review of Applied Linguistics in Language Teaching, 48, 275–296. Gut, U. (2010). Cross-linguistic influence in L3 phonological acquisition. International Journal of Multilingualism, 7, 19−38. Llama, R., & Cardoso, W. (2018). Revisiting (non-)native influence in VOT production: Insights from advanced L3 Spanish. Languages, 3, Article 30. Llama, R., Cardoso, W., Collins, L. (2010). The influence of language distance and language status on the acquisition of L3 phonology. International Journal of Multilingualism, 7, 39−57. Wrembel, M. (2010). L2-accented speech in L3 production. International Journal of Multilingualism, 7, 75−90. Wrembel, M. (2011). Cross-linguistic influence in third language acquisition of voice onset time. In W.-S. Lee & E. Zee (Eds.), Proceedings of the 17th International Congress of Phonetic Sciences (pp. 2157–2160). City University of Hong Kong.

ABSTRACT. Whether speakers use a noun phrase (NP; e.g., the woman) or a pronoun (e.g., she) may depend, among other things, on the needs of their listeners (e.g., Arnold & Griffin, 2007). For listeners, an unambiguous way to signal their communicative needs is through overt feedback, for example, asking “Do you mean this one?” in case of referential ambiguity. But will receiving such feedback from a particular listener lead speakers to increase their referential specificity in other situations? Here, we study the aftereffects of receiving overt feedback on bilingual pronoun production. Spanish-English bilinguals in their second language (L2) use under-specified references (i.e., pronouns instead of NPs) more often than monolinguals, even when same-gender referents create ambiguity (e.g., Mary met Jane when she was young Contemori & Dussias, 2016). We test the hypothesis that interlocutor feedback in the form of clarification requests may increase bilinguals’ explicit references in a different situation, presumably by making them more aware of their under-specificity. In an on-going experiment*, fourteen Spanish-English bilinguals used their L2 (English) to complete an oral picture-description task, consisting of a Pretest, Feedback and Posttest phases. In the Pretest phase, participants saw two pictures of one or two characters (with same or different gender) performing different actions, read two context sentences introducing the referents, and had to complete a third sentence (e.g., “A girl was arguing with a boy. The boy got really annoyed.” Subsequently…). Only participants who produced 60% or more pronouns in this phase continued the experiment. In the following Feedback phase, participants viewed pictures, read a lead-in sentence (e.g., “Diego and Juan are at the office”), selected the correct continuation among two choices (Figure 1), and read the lead-in and continuation to the confederate, who presumably used the information in a picture-matching task. For experimental items, the continuation contained a pronoun, ambiguous for half of the items, e.g., “He is eating a banana”. To these, the confederate provided two types of feedback (varied between participants): 1) informative about the pronoun ambiguity (e.g, Did you mean Diego?), or 2) uninformative about the pronoun ambiguity (e.g., Did you mean he is eating a yogurt?). An additional control group always received NP continuations and uninformative feedback. Fillers for all groups contained NP-continuations. After the feedback phase, bilinguals participated in an immediate posttest similar to the pretest (without a listener). We performed LMER analyses on the rate of explicit (NPs) references produced in the different character conditions (1 or 2 characters; similar or different gender) in the pretest and posttest. These preliminary analyses did not include feedback condition due to a lack of power. Bilinguals produced significantly more NPs in all character conditions in the posttest than in the pretest (p<.005, see Table 1); numerically, the effect was similar across all feedback groups. These preliminary results suggest not only that speakers are sensitive to the communicative requirements of the context, but that this sensitivity leads them to adapt their referential choice in a different situation (and even when their initial speech is not underspecific).

ABSTRACT. Learners of a second language will at some point encounter new vocabulary that will strongly remind them of words from their mother tongue. English-German examples for this are nose and Nase, gold and Gold, or shoe and Schuh. Such word pairs, which display a considerable overlap in form (orthography and/or phonology) and in meaning, are called cognates.

Up to date, numerous studies have shown that cognates can facilitate language processing: shorter RTs for cognates have been observed in picture-naming (e.g., Costa et al., 2000), in lexical decision (e.g., Dijkstra, 2005), and in translation (e.g., Friel and Kennison, 2001). There is also evidence, cognates may trigger the switching of languages (Broersma, 2011). These ob-servations have led many researchers to argue that cognates take on a ‘special’ place in the bi-lingual mental lexicon. Several approaches were developed to describe the uniqueness of cog-nates (see, e.g., Sánchez-Casas and García-Albea, 2005), nevertheless, there has not yet been one exact, encompassing explanation for cognates in the lexicon and the connections between them. Against this background, an intriguing question is whether cognates are subsumed into one lexical entry and therefore produce facilitation, or if they constitute two separate entries with strong links.

In order to shed on this issue, a study was conducted to examine the lexical environment of cognates, making use of a monolingual and a parallel corpus. First, the German corpus DeReKo (from IDS Mannheim) was searched for English forms of English-German cognates. The re-search question here was whether the English forms of cognates appear in German texts. Hypo-thetically, using the English form of a cognate in German requires justification; it needs to be licensed. This may be accomplished by use of quotes, attached translations or explanatory con-structions. In the study, the presence of quotes was used as dependent variable. The statistical analysis revealed that cognates indeed appear significantly more often in quotes when in their English form.

In a second step, the same set of cognates was searched for in the parallel corpus Europarl (accessed via SketchEngine). The research question in this part of the study was if cognates are translated with their equivalent from the other language. With the cognate facilitation effects in mind, we would assume this to be the case. However, the analysis does not support this as-sumption. In fact, cognates are oftentimes part of phrasemes. In such cases, translators opt for formulations which do not include the cognate – even if the same phraseme exists in the other language. For example, both English and German make use of the phraseme where the shoe pinches (in German respectively wo der Schuh drückt), yet translations into the other language show a variety of options apart from the equivalent.

The findings of the study will be presented against the background of current research on the bilingual mental lexicon. We will especially discuss to what extent the results are compatible with our knowledge about lexical access and entries in the lexicon.

References: Broersma, M. (2011). Triggered code-switching: Evidence from picture naming. In M. S. Schmid & W. Lowie (Eds.), Modeling bilingualism (pp. 37-58). John Benjamins. Costa, A.,Caramazza, A., & Sebastian-Galles, N. (2000). The cognate facilitation effect: Impli-cations for models of lexical access. Journal of Experimental Psychology: Learning, Memory and Cognition 26(5), 1283-96. Dijkstra, T. (2005). Bilingual visual word recognition and lexical access. In J. F. Kroll & A. M. B. de Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches (pp. 179-201). Oxford University Press. Friel, B.M., & Kennison, S.M. (2001). Identifying German-English cognates, false cognates, and non-cognates: Methodological issues and descriptive norms. Bilingualism: Language and Cognition 4(3), 249-74. IDS. (2020). Deutsches Referenzkorpus / Archiv der Korpora geschriebener Gegenwartsspra-che 2020-I. (Release of 21.01.2020). Koehn, P. (2005). Europarl: A Parallel Corpus for Statistical Machine Translation. MT Sum-mit. Sánchez-Casas, R., & García-Albea, J. (2005). The representation of cognate and noncognate words in bilingual memory. In J.F. Kroll & A. M. B. de Groot (Eds.), Handbook of bilin-gualism: Psycholinguistic approaches (pp. 226-250). Oxford University Press.

ABSTRACT. Children who can speak more than one language show advantages in Executive Functions (EFs; Bialystok & Viswanathan, 2009; Carlson & Meltzoff, 2008; Martin-Rhee & Bialystok, 2008) and perspective-taking (Fan et al., 2015; Goetz, 2003; Kovács, 2009; Rubio-Fernández & Glucksberg, 2012) as compared to monolinguals. Bilinguals’ superior perspective-taking abilities have been explained by enhanced inhibitory control (Goetz, 2003; Kovács, 2009). However, it has been suggested that bilingual enhanced perspective-taking ability is social in origin and not related to superior EFs (Fan et al., 2015; Liberman et al., 2017). We explored the inhibitory control and the social account for young multilinguals’ perspective-taking advantage in referential communication. We recruited two groups of 4- to 7-year-olds, 40 monolinguals and 42 multilinguals. Groups were matched by SES, extracurricular activities, verbal ability, and non-verbal reasoning skills. We compared the groups in two conditions, a social and a non-social condition, of a referential perspective-taking task. The social condition (Director Condition) was a version of the classic Director Task by Keysar et al. (2000). The non-social condition (Camera Condition) was based on a variant by Santiesteban et al. (2015) where the Director is replaced by a camera. We hypothesized that, if multilinguals’ advantage in referential perspective-taking was social in origin, they would perform better than monolinguals in the social, but not in the non-social condition. On the contrary, if multilinguals’ superiority were underpinned by inhibitory control, then multilinguals would outperform monolinguals in both conditions, but would perform comparably in the two versions of the task. For corroborating evidence for either a social or EFs account, we tested children in three different measures of inhibitory control and in a test of emotion understanding with low demands on EFs, the Test of Emotion Comprehension (TEC; Pons et al., 2004). Multilingual children outperformed monolingual children in both the Director and the Camera condition of the communication experiment. All children, both monolinguals and multilinguals, achieved overall higher scores in the social condition than they did in the non-social variant. Multilinguals also demonstrated superior inhibitory control in the inhibition tasks. No difference was recorded between multilinguals and monolinguals in the TEC. Although inhibitory control and emotion understanding were positively related with performances in the communication tasks, neither inhibition nor emotion comprehension accounted for multilinguals’ superiority in the Director or the Camera tasks. We speculated that a third factor, metalinguistic awareness, might be crucial for multilinguals’ superiority in referential perspective-taking.

ABSTRACT. While research investigating bilingual language processing has become more prominent, there are still very few studies that systematically investigate classic oculomotor behavior patterns in L2 speakers, at least not to the extent that is seen in L1 speakers (e.g., skipping behavior and parafoveal processing). A recent study[1] investigated parafoveal processing while reading in English, with L1 English and L1 German/L2 English readers using the gaze contingent boundary paradigm[2]. Unexpectedly, L2 readers derived a reading time interference from a non-cognate translation parafoveal mask (arrow vs. pfeil (arrow in German)), but derived a reading time benefit from a German orthographic parafoveal mask (arrow vs. pfexk) when reading in English. The authors argued that bilingual readers incurred a switching cost from the complete German parafoveal word, and derived a benefit by keeping both lexicons active from the partial German parafoveal word. In this registered report (in-principle accepted1), we further test this finding with L1 German/L2 English participants using improved items in German. Our main research question is whether we see the same pattern of results when reading in German (i.e., interference from Pfeil and facilitation from Pfexk). In addition, we included new comparisons and mask types (see Table 1 & 2). We included an English pseudo-word (Clain) to test whether previous facilitation from the orthographic mask was due to the overlap between the translation word, or from the fact that orthographic mask was “word-like” (given that parafoveal inhibition increases as a mask become less “word-like”[3]). An English word condition was added (Array) to test whether the shared semantic information between the German word and English translation (Pfeil and Arrow) leads to the parafoveal inhibition (due to the language switch) or whether it is the non-identical word is in the parafovea that leads to inhibition. If there is facilitation (in terms of shorter reading times from Arrow vs Array), this would suggest that there is a language switch occurring, and that Array takes longer to read because it shares no semantic information with Pfeil. Finally, we included a comparison between the English translation and the two English non-words (Arrow vs. Arrzm + Clain) to test whether the two non-word masks (Arrzm and Clain) cause less of an inhibition than a real L2 word (Arrow), this would be in contrast with what has been found with L1 speakers (e.g., [3]). If there is facilitation (in terms of a shorter reading time for English translation relative the two non-words) this would provide further evidence of semantic facilitation from the English translation word relative to the other uninformative masks. This comparison is particularly interesting because it could provide evidence that unlike L1 readers, L2 readers show an parafoveal inhibition for L2 words and facilitation from “word-like” non-words. We believe that this study will have important theoretical impact for bilingual word recognition and for eye-movement-while-reading models. To the knowledge of the authors, this type of full-word parafoveal inhibition and partial-word parafoveal facilitation has not been explored in L2 speakers.

ABSTRACT. Our study investigates lexical production and innovation of adult and child heritage Russian speakers dominant in Hebrew and American English (n = 144). While studies on heritage language (HL) production and creativity exist (Rakhilina et al, 2016), this is the first to assess results across both age groups and societal languages (SLs) in the lexical domain. Lexical errors can reveal the strategies heritage speakers use to cope with competing language systems in their mental lexicon (Altman et al, 2017). Investigating these patterns across age groups allows us to track the HL trajectory over time, and considering distinct SLs can distinguish universal HL trends from language-specific ones. Furthermore, when studying naming vocabulary, it is important to consider both nouns and verbs, as verbs are acquired at a later age (Hamann et al, 2017) and therefore pose a greater challenge for bilinguals, further elucidating strategies of lexical processing (Faroqi-Shah et al, 2021).

In our study, we asked: How do heritage Russian speakers from different age groups and SLs compare on production accuracy of nouns and verbs? What error patterns do they exhibit at the group level? Thus, we conduct both a general quantitative and an in-depth qualitative analysis of responses. Quantitatively, we hypothesized that Hebrew-dominant speakers will outperform English-dominant speakers, and adults will outperform children, due to increased HL exposure for both of the former groups. Qualitatively, we predicted similar patterns of non-target responses between adults and children (Polinsky, 2006) and potential distinctions between the two SL groups.

We tested our participants on the production accuracy of 51 nouns and 51 verbs, taken from the “Verb and action: stimuli database” and “Noun and object: stimuli database” (Akinina et al, 2015). Stimuli were coded alongside supplementary information including word frequency and age of acquisition (AoA). Results were coded in layers. First, we used a binary system (0,1) to indicate whether the participant produced the target response. Next, we classified the non-target responses into categories: Semantic error (eg. strekoza ‘dragonfly’- komar ‘mosquito’), Phonological error (eg. skrepka ‘clip’- skripka ‘violin’), Both (eg. kopilka ‘piggy bank’- kopeika ‘penny’ ), Code-switching (eg. rul’ ‘steering wheel’- koleso ‘wheel’), Innovation (eg. vymja ‘udder’- kran korovy ‘cow faucet’), Unrelated (eg. kastrjulja ‘pot’- koza ‘goat’) or no response. We added further subdivisions to drill down into specific categories (ie. for code-switching, we differentiate between calques and borrowing), and noted if the response did not fit the assigned noun/verb category (eg. chokat’sa ‘to clink glasses’- tost ‘a toast’).

A binomial mixed-effects regression analysis found no effect of SL or age group on noun production accuracy. By contrast, effects of both these factors were found on verb production accuracy, supporting the initial hypothesis. Word frequency and AoA predicted performance on both. Preliminary qualitative results found that, on both tasks, the vast majority of errors were semantic. Verbs saw over three times the innovations and non-category responses as nouns, with nouns having almost twice as many cases of code-switching. In our talk we will present a detailed breakdown of the 6,000+ non-target responses.

ABSTRACT. The non-selective nature of bilingual lexical access often results in cross-linguistic influence (CLI) in lexical processing; however, such effects are not as evident at the syntactic level. While research on L2 sentence production suggests a high degree of L1-L2 interactivity also at a structural level (e.g. Hartsuiker & Bernolet, 2017), work on L2 sentence comprehension often reports uniform processing patterns despite L1 differences. These discrepancies in CLI have led to interest in the interaction between lexical and syntactic processing in bilinguals, with available evidence showing that interactive lexical processing can impede target syntactic processing (e.g. Hopp & Lemmerth, 2018). Such findings have led to the Lexical Bottleneck Hypothesis (Hopp, 2018), according to which incomplete parsing partially arises from slowdowns and CLI in lexical processing in bilinguals.

This empirical study extends previous work on lexical-syntactic interactions in L2 sentence processing by examining the detection of ungrammaticality and type of L1-L2 grammatical gender mapping in Spanish-German Interlanguage. Adult L1 Spanish speakers (n=61) with intermediate-advanced L2 German proficiency and L1 German speakers (n=36) completed a 80-item German self-paced reading task focusing on the interaction between noun gender congruency and word order with attributive adjectives. Target nouns were either congruent (same gender in the two languages), incongruent (masculine-feminine mismatches) or L2 neuter (Spanish masculine/feminine). These nouns were paired with attributive adjectives appearing either prenominally (grammatical in German, ungrammatical in Spanish; see (1a)) or postnominally (ungrammatical in German, grammatical in Spanish; see (1b)).

A three-way interaction of Group*Grammaticality*Congruency in reading times in Region 4 (see Figure 1) showed that the detection of syntactic ungrammaticality among L2 speakers was influenced by L1-L2 lexical gender congruency. L2 adults slowed down in reading for ungrammatical word orders only for lexically gender-congruent NPs in Region 4. These findings reveal earlier and stronger sensitivity to ungrammatical syntax when the gender of the noun is the same in Spanish and German (either feminine or masculine). For these gender-incongruent nouns, interactions with proficiency suggest that only high-proficiency L2 adults detected ungrammaticality in the L2.

This study shows cross-linguistic lexical gender effects in the detection of syntactic ungrammaticality, even when the latter is independent of gender information. In this regard, it highlights the important role of the specific type of cross-linguistic lexical overlap in L2 sentence processing, illustrating that contrasts in syntactic L1-L2 features, such as those determining word order differences, go beyond a mere binary same/different distinction and are sensitive to cross-linguistic properties in the lexicon. These findings highlight that there is a need for an integrated and more nuanced study of the bilingual language system, encompassing both the bilingual mental lexicon and the L2 parser.

ABSTRACT. Processing English non-canonical object-initial sentences has been found to be challenging for both native and non-native speakers. In particular, English object-relative clauses (ORCs) are processed less accurately and slower than subject-relative clauses (SRCs) (Hopp, 2016; Lim & Christianson, 2013; Street, 2017). Besides the general structural complexity of ORCs and their less-frequent appearance compared to SRCs, for L1 German speakers, there is another contributing factor to this processing difficulty: The L1 German parse of ORCs is ambiguous between an object and a subject-reading. Hence, L2 learners will have to use their L2 syntactical knowledge to parse correctly. Additionally, implausible sentences (i.e., inconsistent with our world knowledge) cause more processing difficulties than plausible ones and are thus more likely to be misinterpreted (Ferreira & Patson, 2007; Lim & Christianson, 2013). Finally, lexical effects may also modulate (L2) sentence processing, both within and across languages (Hopp, 2016). Cross-linguistic lexical effects could be induced by, for example, cognate words (i.e., form-similar words across languages that share the same meaning) that have been shown to activate cross-linguistic transfer in L2 sentence comprehension (Hopp, 2017). The current project looks at the processing of plausible and implausible SRCs and ORCs in German L2 learners of English (mean age = 22.5; SD = 1.9) and investigates whether potential lexical co-activation of the L1 through cognates induces either (A) L1 syntactic co-activation in L2 sentence comprehension, which would further impede the processing of English ORCs in German learners, or (B) a facilitation of L2 syntactic processing, in line with the lexical bottleneck hypothesis (Hopp, 2018), or (C) no effect whatsoever. In a self-paced reading experiment, 50 advanced learners of English read plausible and implausible English SRCs and ORCs and judged their plausibility. The critical relative clause contained either German-English cognate NPs and verbs (1-4), or noncognate control words (5-8). Prior to the reading task, participants performed an English lexical decision task on the cognate and noncognate stimuli in isolation. Generalized linear mixed-effect models (GLMERs) for plausibility judgment accuracy revealed an interaction between sentence type and plausibility with a processing disadvantage for plausible ORCs. For reading times (RTs) of the critical phrase, linear mixed-effects models (LMERs) replicated overall faster processing of SRCs than ORCs as well as faster processing of plausible compared to implausible sentences. LMERs for the RTs in the post-critical phrase replicated the accuracy results by returning an interaction between sentence type and plausibility with a delay in RTs for plausible ORCs. However, contrary to initial predictions, no effects of cognate status were found for either accuracy or reading time data in both the lexical decision as well as the self-paced reading task. In our presentation, we will discuss possible reasons for the absence of cognate effects and will suggest that the difficulty in judging plausible ORCs may stem from the particular difficulty of revising an initial negative plausibility judgment.

ABSTRACT. Listeners can retune their native (L1) phonetic categories to adapt to anomalous productions when exposed to acoustically ambiguous sounds in biasing lexical contexts [1]. This is known as perceptual learning. The same learning has also been observed with late second-language (L2) learners under some circumstances [2]. Native perceptual learning only takes place when exposure entails perceptual ambiguity but not when anomalous but perceptually unambiguous productions are presented. Whether this is also true for L2 learners remains an open question. The case of L2 learners is particularly interesting because their L2 lexical representations are ‘fuzzy’ [3] and, on top of that, they have difficulties teasing some non-native phonological contrasts apart [4]. The three experiments here investigated whether German learners of English retune their phonetic categories after hearing English words containing vowel substitutions for sounds in an easy (/i/-/ɪ/) vs. a difficult L2 contrast (/ɛ/-/æ/). In addition, the generalizability of such retuning across talkers was assessed.

Eighty-nine advanced German learners of English residing in Germany (NExp1 = 29; NExp2 = 29; NExp3 = 31) underwent a three-part online adaptation of the perceptual learning paradigm which combined exposure phases where listeners heard lists of English words and test phases where they categorized sheep-ship and bet-bat continua. In parts 1 and 3, all exposure words were pronounced canonically, whereas in part 2, the vowels /æ/ and /ɪ/ were produced as [ɛ] and [i], respectively (e.g., *dr[ɛ]gon, *w[i]zard). The main question was whether the vowel substitutions in part 2 would lead to perceptual shifts in the following categorization phases relative to part 1. Crucially, the same male speaker of Standard Southern British English was used as exposure and test talker in Exp. 1, while the test talker was a different male with similar vowel acoustics in Exp. 2 and a female talker in Exp. 3. The results showed significant shifts in the expected direction for /i/-/ɪ/ (i.e., more ship responses after exposure to *w[i]zard-like words) in Exp. 1 and 2 but not Exp. 3. For the difficult /ɛ/-/æ/ distinction, by contrast, a shift was observed but in the opposite direction, as exposure to *dr[ɛ]gon-like words led to fewer bat responses, and only in Exp. 1.

These results offer interesting insights on the interplay between lexical representations and phonetic categories in late L2 learners, as well as on the effects that non-native-like productions have on the perception of difficult L2 contrasts. The findings for /i/-/ɪ/ suggest that L2 learners may be more likely to learn from (and thus adapt to) unambiguously anomalous pronunciations than native speakers (see [5]) because of their comparatively less robust lexical representations, and that this learning is constrained by acoustic similarity [6]. Regarding /ɛ/-/æ/, the pattern in Exp. 1 hints at the possibility that exposure to L1-accented-like word productions may have direct detrimental effects on the perception of challenging L2 phonological distinctions (and not only on lexical representations [7]), but the fact that the effect was restricted to the same-talker scenario calls for caution and further research on the topic.

ABSTRACT. Introduction: It has been proposed that bilinguals’ language use patterns are differentially associated with executive control (e.g., Green & Abutalebi, 2013). To further examine this, the present study relates the social diversity of bilingual language use, as measured by language entropy, to performance on a color-shape switching task (CSST) in a group of bilingual university students with diverse linguistic backgrounds. Language entropy reflects a spectrum of language use in a variety of social contexts, ranging from compartmentalized use to fully integrated use (Gullifer & Titone, 2020). Methods: Language entropy for university entropy and non-university contexts were calculated from questionnaire data on language use from 44 participants (33 women; age M = 22.75, SD = 2.78). Reaction times (RTs) on the CSST were measured to calculate global RT (overall RT in mixed blocks), switching costs (RT difference between switch and non-switch trials), and mixing costs (RT difference between non-switch and single trials) on the CSST, representing conflict monitoring, mental set shifting, and goal maintenance, respectively. In addition, this study innovatively recorded a potentially more sensitive measure of set shifting abilities, namely pupil size during task performance. Results: Higher university entropy scores were related to slower global RT. Mixing costs in RTs were reduced for bilinguals with higher diversity of language use in non-university contexts. No such effects were found for university entropy. Neither university entropy nor non-university entropy scores were associated with switching costs as manifested in RTs. Bilinguals with more compartmentalized language use in non-university contexts showed a larger difference in pupil dilation for switch trials in comparison to non-switch trials. Discussion: These results point to the social diversity of bilinguals’ language use as being associated with executive control, but the direction of the effects may depend on social context (university vs. non-university). Importantly, the results also suggest that some of these effects may only be detected by using more sensitive measures, such as pupil dilation. The presentation discusses theoretical and practical implications regarding the language entropy measure and the cognitive effects of bilingual experiences more generally, as well as how methodological choices can advance our understanding of these effects.

ABSTRACT. Background. Greek, Russian and Turkish (GRT) belong to different types of null subject languages (strict vs. non-strict): Greek is a consistent (strict) pro-drop language, i.e., null subjects appear regardless of grammatical features and syntactic environments [1]. Russian is claimed to be either a partial pro-drop language, or a non-pro-drop language with abundant subject ellipsis [2]. Turkish is a topic (pro-)drop language [3] where in highly unspecified contexts overt subjects are necessary. Despite the differences in the language systems, heritage-varieties of these languages show similarities in pro-drop realization. Overtly realized pronominal subjects are more frequent in heritage-varieties (mostly in contact with Germanic languages) than in monolingual ones (cf. [4, 5] for heritage Greek, [6, 7] for heritage Russian, [8] for heritage Turkish, but see [9]). Based on these observations, we derive the following research questions and hypotheses for a corpus study on RUEG data [10]:

• RQ1 Does GRT heritage speakers’ expression of (pro)nominal reference align with monolingual speakers’ productions? H1 Heritage speakers will show transfer effects from the majority languages German and English by increase in overt realization of pro [11]. • RQ2 How do heritage GRT diverge and converge in their realization of (pro)nominal reference in our corpus? H2 Subject pro-drop is linked to person, gender and number agreement on the verb. If the morphological inflection on the verb sufficiently denotes the subject, it can and often must be phonologically not realized. Heritage speakers, particularly those with restricted language exposure, might not be able to sufficiently mark agreement in all instances which is why they would have to produce more overtly realized subjects. • RQ3 Are there effects of register (formal / informal) and modality (written / spoken) on the production of (pro)nominal reference? H3 Heritage speakers sometimes lack features that are transmitted via formal education such as register. We assume that pro-drop is sensitive to register and modality differentiation which heritage speakers cannot fully exhibit. Thus, we expect a weakening in heritage speakers’ pro-drop production due to register-levelling and overexplicitness [12].

Methodology. We will retrieve our data from the last version of the RUEG corpus [10] (manual pro-drop annotation currently in progress) which consists of semi-spontaneous narrations from monolinguals and heritage speakers (total N=548) (see Table 1). Discussion. So far, studies in the field of heritage languages have rarely used a unified methodology which questions the generalizability of those results. Our study offers cross-linguistically comparable data that can be generalized as the outcome rises from distinct heritage communities residing in the same countries, selected according to the same criteria. Combining different types of pro-drop languages (GRT) with two non-pro-drop languages (German or English) in a bilingual’s mind gives us insights into diverse outcomes. We can provide evidence of heritage languages’ divergence from the monolingual variety and exclude some impossible scenarios in language, for example, it is likely not the case that a pro-drop language abolishes this feature completely due to bilingualism. Rather, there are slight adjustments but not a complete alternation of the system.

References 1. Philippaki-Warburton, I. The theory of empty categories and the pro-drop parameter in Modern Greek. Journal of Linguistics 23, 289–318 (1987). 2. Shushurin, P. in Pronouns in embedded contexts at the syntax-semantics interface 145–169 (Springer, 2018). 3. Öztürk, B. in Linguistik Aktuell/Linguistics Today (ed Taylan, E. E.) 239– 259 (John Benjamins Publishing Company, Amsterdam, 2001). isbn: 978-90-272-2765-2. https://benjamins.com/catalog/la.44.10ozt (2022). 4. Argyri, E. & Sorace, A. Crosslinguistic influence and language dominance in older bilingual children. Bilingualism: Language and Cognition 10, 79– 99 (2007). 5. Tsimpli, I., Sorace, A., Heycock, C. & Filiaci, F. First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8, 257–277 (2004). 6. Gagarina, N. Discourse cohesion in the elicited narratives of early Russian- German sequential bilinguals. Multilingual individuals and multilingual societies, 101–120 (2012). 7. Ivanova-Sullivan, T. Theoretical and experimental aspects of syntax-discourse interface in heritage grammars (Brill, 2015). 8. Haznedar, B. Transfer at the syntax-pragmatics interface: Pronominal subjects in bilingual Turkish. Second Language Research 26, 355–378 (2010). 9. Azar, Z., Özyürek, A. & Backus, A. Turkish-Dutch bilinguals maintain language-specific reference tracking strategies in elicited narratives. International Journal of Bilingualism 24, 376–409 (2020). 10. Wiese, H. et al. RUEG Corpus 2021. https://zenodo.org/record/ 5808870. 11. Tracy, R. Spracherwerb und Mehrsprachigkeit. R. Braches-Chyrek, C. Röhner & Sünker, Heinz: Hopf, Michaela (Hrsg.), Handbuch zur frühen Kindheit, 185–197 (2014). 12. Pashkova, T., Murphy, M. & Allen, S. in, BUCLD 45 (BUCLD 45, Oct. 2020).

ABSTRACT. It is well attested that, when learning another language, speakers struggle in obtaining a native-like pronunciation of L2, resulting in speech often characterized by a “foreign accent”. Patterns of prosody and phonology, in fact, have been described as among the hardest to get to proficiency, though it may depend on several factors, such as age of acquisition and the amount of time spent in the L2 country (Major 2001 among others). According to Flege’s Speech Learning Model the sounds of a L1 and a L2 share a common phonological space (Flege, 1995). As a consequence, the influence of these patterns is bidirectional, leading to potential effects also of a L2 towards the native language. There is evidence of this phenomenon in speakers of minority or heritage languages. Despite practicing both languages daily, at a certain stage the input these speakers receive from the dominant language is typically greater than the one offered by the native one: one of the consequences might be attrition of their native language, also resulting in pronouncing it with a foreign accent (namely, the prosody of the dominant language) (Stangen et al., 2015). The aim of our study is to investigate the extent of prosodic influence on the native language in terms of transferring L2's segmental and suprasegmental features to speakers' L1. We compare different groups of speakers: minority speakers of Slovenian living in the border between Italy and Slovenia (MB) (European Charter for Regional or Minority Languages), heritage speakers of Slovenian living in an Italian territory (HS), Italian L2 learners of Slovenian (L2), and two control groups of monolinguals from both languages (SB and IB). We will focus on specific segmental and suprasegmental patterns typical of Italian and we will check whether they are transferred in Slovenian. In particular, at the segmental level we will test: i) intervowel voicing; ii) stress retraction; iii) final word lengthening. At the suprasegmental level we are interested on: i) the directions of rise on declarative sentences and on yes/no questions; ii) the pre/post pausal shortening and lengthening (Nespor & Vogel, 1986). These patterns will be elicited with a semi-spontaneous speech trough the Map Task paradigm (Anderson et al., 1989). Within this repertoire of features, we expect a tendency for the greater degree of transfer in terms of a greater number of transferred features as a function of exposure to Slovenian, in particular: MB (the least transfer, the most exposed to Slovenian)  HS  L2 (the most transfer, the least exposed to Slovenian). In addition, we predict this might be more evident at the suprasegmental level on the basis of the prosodic hierarchy framework, according to which, when prosodic features are acquired, also phonological interferences disappear (Missaglia 1999). By targeting different types of bilingualism, this study will offer new insights into the phonological processing of the bilingual mind and contribute to its theoretical framework.

ABSTRACT. Language switching is believed to involve suppression of the non-target language, suggesting a strong need for cognitive control, especially inhibition (Green, 1998; Green & Abutalebi, 2013). We investigated these control processes at the switch initiation (i.e. at pre-target cue = proactive control) and during target presentation (reactive control), measuring induced and evoked EEG activity. 89 healthy right-handed bilinguals aged 18-33, all highly proficient in a Romance language and English, took part in a cued picture-naming language switching task. Language (L1 vs. L2, defined by proficiency level) and type of trial (switch vs. non-switch) was modulated. In line with previous results, we found switch trials to be less accurate (z = -5.995, p < .001) and slower (t = 19.741, p < .001) than non-switch trials. In addition, L1 trials were slower than L2 trials (t = -4.491, p < .001), and a trials type x language interaction (t = -2.435, p = .015) indicated a larger switch cost for L1. Time-locked to the cue, switching (compared to not switching) was associated with a larger occipital positivity (290-670 ms), accompanied by anterior negativity. This pattern was the case for both languages. In line with previous task-switching findings (Elchlepp et al., 2012; Kopp et al., 2020; Rushworth et al., 2005), this suggests proactive control that is independent of language. The induced cue-locked activity demonstrated significant differences in alpha (8-12 Hz) for switch versus non-switch trials for L1, but not for L2. There was more alpha for switch trials at 200-550 ms and for non-switch trials at 650-1300 ms. As alpha has been related to inhibition of irrelevant information (Gladwin & de Jong, 2005), the earlier effect suggests inhibition for switch initiation into L1. The later effect is in line with effects for non-verbal switching paradigms (Sauseng et al., 2006) and likely reflects the avoidance of intrusion of the irrelevant language. At target, non-switch trials versus switch trials elicited a significantly higher central negativity (730-1150 ms) for L2, but not L1. This possibly is a late sustained negativity related to motor-response preparation and task set maintenance - usually higher for repeat trials (Gajewski et al., 2018). L2 also showed significantly higher parietal positivity (880-1690 ms) than L1, reflecting higher inhibition needed to suppress L1 during L2 speech production (Timmer et al., 2018). These findings correspond well with the higher RT costs for L1. To conclude, our group of highly proficient bilinguals demonstrated a previously reported asymmetric behavioural switching cost, suggesting more control needed for switching into L1 than L2. At the cue level, both evoked and induced EEG responses suggested higher proactive cognitive control demand for switching that was independent of language. But there was also evidence for higher proactive inhibition for switching into the more proficient L1. At the target level, a late positive component suggested inhibition of L1 during L2 production. Interestingly, the higher behavioural L1 switch cost was not reflected in a larger neural effect at target level, but in an absence of a neural effect.

ABSTRACT. The corpus callosum is the largest white matter structure of the human brain, which enables interhemispheric transfer of information. Research has shown that bilinguals have larger anterior regions of the corpus callosum (Felton et al., 2017; Coggins et al., 2004) when compared to monolinguals, but little is known about the behavioural consequences. Besides, as a potential effect of age-related corpus callosum thinning, interhemispheric transfer seems to be less efficient in older adults (Delvenne & Castronovo, 2018) . The current study aimed to investigate the effect of bilingualism and age on the bilateral processing advantage – the phenomenon that easier tasks are processed faster by one hemisphere, while processing by both hemispheres is more efficient for more complex tasks. In this word-matching study, four groups of younger (18-25 years) and older (65-85 years) monolinguals and bilinguals were tested with three difficulty levels, where they had to decide whether two words were semantically related or not. The difficulty level corresponded to the association strength between the words, which were displayed simultaneously, either both on one side of the screen (unilaterally) or one on each side of the screen (bilaterally). As expected, all groups showed a significant effect of difficulty, with slower reaction times in more difficult trials, and an effect of presentation, with slower reaction times in unilateral trials. However, this bilateral processing advantage was consistently found in younger adults, but it disappeared for older adults when task difficulty increased. No difference between monolinguals and bilinguals was found in this study, except for slightly slower reaction times for bilinguals among the younger adults. These results could indicate that, although bilingualism seems to alter white matter integrity and corpus callosum volume, it does not necessarily have an impact on interhemispheric communication (Woelfle & Grahn, 2013). However, ageing seems to slow down interhemispheric communication, especially for tasks that are complex and require recruitment of more cognitive resources.

Coggins, P. E., Kennedy, T. J., & Armstrong, T. A. (2004). Bilingual corpus callosum variability. Brain and Language, 89, 69-75.

Delvenne, J.-F., & Castronovo, J. (2018). Reduced inter-hemispheric interference in ageing: Evidence from a divided field Stroop paradigm. Brain and Cognition, 122, 26-33.

Felton, A., Vazquez, D., Ramos-Nunez, A. I., Greene, M. R., Macbeth, A., Hernandez, A. E., & Chiarello, C. (2017). Bilingualism influences structural indices of interhemispheric organization. Journal of Neurolinguistics, 42, 1-11.

Woelfle, R., & Grahn, J. A. (2013). Auditory and visual interhemispheric communication in musicians and non-musicians. PLOS ONE, 8(12), 1-6.

ABSTRACT. The language control processes used by multilinguals to monitor and control their language use are commonly investigated in the lab using picture naming paradigms in which participants are instructed when to switch between languages. However, language switching in daily life often occurs freely when interlocutors share multiple languages, a process which is rarely studied using neuroimaging. This study investigates the electrophysiological manifestation of switching effects in voluntary compared to cued language switching.

Cued language-switching tasks usually show slower responses on trials in which participants switch to a different language (switch trials) compared to no-switch trials (repeat trials) (i.e. switching cost; e.g. Meuter & Allport, 1999). Previous behavioural studies found this switching cost could be the same (de Bruin et al., 2018) or smaller (Jevtović et al., 2020) in voluntary compared to cued switching, or even absent (Blanco-Elorrieta & Pylkkänen, 2017). It could be that less top-down control of the non-target language is required when speakers are free to choose their language and can simply use whichever words come most easily (e.g. Green, 1998).

In the present study, adult Dutch-English late bilinguals perform two picture-naming tasks in which they are required to switch between Dutch and English, while EEG is recorded. In one task, they are requested to use both languages, but can decide the language for each trial voluntarily, while in the other, language is cued by the background colour of the image. For control purposes, the voluntary task alternates between two backgrounds as well. This results in a 2*2 design with the variables task (voluntary/cued) and trial (switch/repeat).

The analyses in this study focus on the N2 ERP component and mid-frontal theta oscillations, two common electrophysiological markers of cognitive control in task switching. Earlier cued language switching studies have found an N2 switch effect (e.g. Jackson et al., 2001) and a mid-frontal theta switch effect (unpublished analysis of data from Zheng et al., 2020). The properties or size of these top-down control markers could differ between repeat and switch trials, as well as between voluntary and cued switching. Previous research in the speech production and task-switching literature has found N2 effects in relation to various types of top-down control and other non-linguistic studies show higher mid-frontal theta power for trials that require more cognitive control (e.g. Eisma et al., 2021). Therefore, it is plausible that the electrophysiological switching cost (the difference between the N2 amplitude or mid-frontal theta power in switch versus repeat trials) is bigger in the cued task than the voluntary task.

This study was pre-registered. Currently, half of the 24 intended datasets have been collected. The study’s results and conclusions will be presented, providing new insights into the neural processes of language control behind voluntary switching in multilinguals. Our results will also aid neuroimaging research on multilingual processing to better reflect the variety in real-life multilingual experiences.

ABSTRACT. Performance on traditional cognitive control tasks, such as Flanker, Simon, and Stroop, has been shown to improve following conflicting or incongruent trials. This heightened performance (faster response time (RT) and higher accuracy) is called the congruency sequence effect (CSE), and has been ascribed to upregulation of cognitive control. Some studies have found CSE effects to be affected by bilingualism, with bilinguals exhibiting a weaker CSE effect than monolinguals following incongruent trials, although this has been controversial (Grundy et al., 2017; Paap et al., 2019). However, CSE effects in general have come under scrutiny, with some critics arguing that heightened performance can be due to feature and response repetition confounds rather than upregulation of high-level cognitive control (e.g., Weissman, 2014).

The current study employs a lesser-used task called the Partial Repetition Cost (PRC) task (Huffman, 2020) to probe bilingual vs. monolingual performance to series of fully, partially, or non-repetitive stimuli. In other words, the current study makes direct use of the proposed cognitive-control task confounds to investigate the contentious bilingual advantage. Participants in this study see a sequence of three stimuli: a set of arrows (response cue stimulus, henceforth RC), followed by a colored square (cued response stimulus, henceforth CRS), followed by another colored square (discrimination response stimulus, henceforth DRS). The arrows may face left (<<<<<) or right (>>>>>) and the squares may be green or blue.

Regardless of its color, participants are expected to press left to the CRS if the arrows in the RC faced left, and to press right if the RC arrows faced right. In response to the DRS, participants are expected to press left if its color is green, and right if it is blue. The squares’ spatial locations vary, which participants are instructed to ignore. The stimulus sequence fully repeats if the CRS and DRS share all features and responses (figure 1). The stimulus sequence partially repeats if the CRS and DRS share a feature or response but differ in another feature or response (figure 2). The stimulus sequence is non-repeating if none of the features or responses are shared between the CRS and DRS (figure 3).

Data collection is currently in progress, using the on-line platforms Finding Five and Prolific. Replicating other studies, we expect participants to perform better on fully or non-repeating trials than on partially repeating trials, which require inhibition of prior feature-response associations. If bilinguals in this study outperform monolinguals on the partial-repetition trials, this suggest that bilinguals are better at dealing with changing lower-level associations. This in turn would suggest that effects seen in higher-level cognitive control tasks could be explained, in part, by stimulus feature and response repetition confounds, which draw on low-level processes, as opposed to upregulated cognitive control, to which the bilingual advantage is commonly attributed.

ABSTRACT. We present an artificial language learning experiment that contributes to novel insight about the relative influence of linguistic similarity between pre-existing grammars and a third language (L3). Using an artificial language allows us to explore the very beginning of the acquisition process and to have full control over the participants’ exposure to the L3. This is important, as we tested a linguistic structure to which the participants had not been exposed, in order to avoid the confound of learning.

Our participants were Norwegian–English sequential bilinguals (N = 120). They were randomly assigned to one of four different L3s (Inputs A–D). Importantly, the L3s differed in terms of morphosyntactic and lexico-phonotactic similarities to Norwegian and English, as follows: • Input A: Norwegian-based lexico-phonotactics and neutral morphosyntax. • Input B: English-based lexico-phonotactics and neutral morphosyntax. • Input C: Norwegian-based lexico-phonotactics and English-based morphosyntax. • Input D: English-based lexico-phonotactics and Norwegian-based morphosyntax. The neutral morphosyntax in Inputs A/B refers to Subject-Verb-Object (SVO) word order, which is found in both Norwegian and English. Crucially, this cue does not contribute to the establishment of one of the previously acquired languages as more similar to the L3 than the other. Example (1) shows an artificial SVO sentence. In Input C, we also exposed the participants to do-support (example (2))—a feature that exists in English, but not in Norwegian. This means that the learners of input C were exposed to incongruent morphosyntactic and lexico-phonotactic cues, with the morphosyntax being more similar to English and the lexico-phonotactics to Norwegian. In input D, the additional morphosyntactic cue to SVO word order was post-nominal possessives (example (3)). Norwegian accepts both pre- and post-nominal possessives, but English only accepts pre-nominal possessives, i.e., there was a morphosyntactic mismatch between the L3 and English. Again, this shows incongruency between the lexico-phonotactic and morphosyntactic cues, as the former was based on English and the latter on Norwegian. After exposure to the artificial L3, we collected forced-choice acceptability judgements of non-subject-initial declarative clauses that varied in verb placement (example (4)). The word order in (4a) is shared with Norwegian and the word order in (4b) with English. Examining the participants’ acceptability judgements should reveal the preferred source of crosslinguistic influence and cannot be a result of learning. There were 18 sentence pairs in total: 12 fillers and six non-subject-initial declaratives.

The participants’ choices are shown in Figure 1. We fitted a binomial mixed-effects model to the data and found significant main effects of lexico-phonotactic and morphosyntactic similarities (Figures 3 and 4, respectively). This suggests that bilinguals are sensitive to these cues in the input after minimal exposure to the target language. The effect of lexico-phonotactic similarity was particularly strong. This may be attributed to the early access to information about (pseudo)cognates and similar sounds in a new language. The results support similarity-driven models of L3 acquisition that argue for linguistic similarity between the L3 input and pre-existing grammars as the main driving force behind crosslinguistic influence.

ABSTRACT. For this priming study, 185 subjects (sex at birth: 102 f., 83 m.) were tested and recruited locally, and on the platform Prolific (https://www.prolific.co). The mean age was 22.21 years (range = 18-31; SD = 2.44) and participants represented the entire gender identity spectrum (96 cis-gender; 31 non-binary/genderqueer, 26 transgender, 32 both non-binary/genderqueer, and transgender). All subjects were multilingual speakers of different languages, but all with advanced English knowledge: 39 bilinguals from birth, 39 early bilinguals (before the age of 6), and 107 sequential bilinguals. The subjects performed an online priming study, with a stereotypical role noun as prime (e.g., the nurse), followed by either the personal pronoun he or she. Subjects had to indicate by button press (‘q’ for he and ‘p’ for she) which pronoun they saw as target. The hypothesis is that subjects are slower and less accurate for pronouns that don’t match the prime. This study aimed to gain a comprehensive picture of gender stereotyping by investigating the effect of variables such as language and social background, political views, parental upbringing, own gender identity, sexist attitudes, personality traits, social dominance orientation, and views on gender roles. Bilingual children have been reported to show less racial bias than monolingual peers (Singh et al., 2019; 2020) and we aimed at examining whether this effect may expand to gender stereotyping in young adults. Accordingly, we expected simultaneous bilinguals to show reduced stereotyping due to their upbringing in comparison to sequential bilinguals, who mostly learned their L2s in a classroom setting. Gender identity was included to explore whether this might be a better predictor than sex at birth. A main goal was to look at bilingualism and gender identity as expanding on a spectrum rather than being dichotomous categories (monolingual vs. bilingual; male vs. female). We think that a critical reconsideration of these variables could contribute to the solving of the replication crisis in the field of psycholinguistics. The cis-gender subjects were in general faster than the transgender and/or non-binary subjects and different variables seem to have influenced the gender stereotyping of both groups. Therefore, separate explorative linear mixed effects models for these two groups are constructed to assess the effect of the various background variables. First analyses revealed interesting patterns: cis-gender subjects tended to be slower for gender stereotype incongruent in comparison to congruent trials. This pattern was reversed for the transgender and/or non-binary subjects: they were generally slower in the gender stereotype congruent in comparison to the incongruent trials. These findings merit further investigation and discussion as they also suggest that including the entire gender identity spectrum is relevant and can offer new insights on psycholinguistic processing.

ABSTRACT. Numeral classifiers are a grammatical category that serves an important role in categorising nouns into units (Adams, 1989). Speakers of classifier languages are well-reported to show classifier-based categorization in cognitive processes (Saalbach & Imai, 2007, 2012; Zhang & Schmitt, 1998). Speakers of classifier languages tended to judge the objects with the same classifier to be perceptually more similar than objects with different classifiers. Most of the previous studies have compared speakers of classifier languages and non-classifier languages. No studies have yet attempted to investigate the classifier effect through the lens of bilingualism. The current study takes the first step to investigate how knowledge of classifiers in the first language (L1 Chinese) and second language (L2 Japanese) affect Chinese-Japanese bilinguals’ perceived object similarity. Bilinguals with short-term vs. long-term residence in Japan were compared to monolinguals to reveal how bilinguals’ object perception might differ from monolinguals and whether length of residence (LoR) influenced bilingual perception.

161 adults participated in the study. These include 36 Chinese monolinguals, 42 Japanese monolinguals, 56 bilingual long-term stayers (averaged LoR 5.2 years) and 27 bilingual short-term stayers (averaged LoR 2 years) in Japan. All bilinguals reported having reached advanced Japanese proficiency. An elicited imitation test (EIT) (Ortega et al., 2002) was used to measure Japanese proficiency. Japanese monolinguals reached 99.4% accuracy in the EIT; the long-term stayers scored 86.5% and the short-term stayers scored 80.2%, both of which were significantly below the monolingual benchmark with no statistical difference between the two LoR groups.

A similarity judgment task (Saalbach & Imai, 2012) was used to measure participants’ object perception. The task adopted a 2 Chinese classifier \times 2 Japanese classifier design and asked participants to rate on a Likert scale the perceptual similarity of object pairs that either share (+) or do not share (-) the same classifier in the target languages. The 2 \times 2 factorial design yielded four types of object pairs (Table 1). The task stimuli were validated for classifier usage in a norming study on Chinese and Japanese monolinguals. Ratings were z-transformed for a linear mixed-models analysis in R.

Results showed that all the four groups’ similarity judgments were influenced by the manipulation of same vs. different classifiers in both Japanese and Chinese, but to a different extent in different groups (Figure 1). Chinese monolinguals’ judgment was more affected by Chinese classifiers than by Japanese classifiers, whereas Japanese monolingual judgment was more influenced by Japanese classifiers than by Chinese classifiers. Bilinguals demonstrated a profile distinct from both monolingual groups. Their judgment was less affected by Japanese classifiers compared to Japanese monolinguals, indicating that bilinguals had not achieved nativelike conceptual categorisation based on the target classifier system. Meanwhile, the Chinese-classifier effect was weaker among bilinguals compared to Chinese monolinguals, suggesting bilinguals’ conceptual deviation from L1-based categorisation patterns. LoR did not yield a difference in classifier effects. The current findings imply that bilinguals experienced conceptual restructuring which could lead to a merged bilingual conceptual domain distinct from both L1 and L2 (Aveledo & Athanasopoulos, 2016).

ABSTRACT. The bulk of studies on the impact of bi-/multilingualism on L1 language abilities have focused on lexical-level/single-word domains and measures such as vocabulary size, lexical retrieval speed, and verbal fluency. Most studies reported a negative impact of bilingualism on the above domains/measures (Bialystok, 2009). It is unknown if the negative impact of bi-/multilingualism on L1 language abilities extends to the morphosyntactic level. Given the co-activation of lexical representations encoding specific morphosyntactic features in bi-/multilinguals, one could presume that amount of use and proficiency level in their L2/foreign languages (FLs) might modulate the activation level of these lexical representations and the strength of the competition between them in morphosyntactic tasks. The number of languages one knows could also matter, as proficiency in several languages should increase lexical competition due to the larger number of activated lexical representations. This study investigates the effect of sequential bi-/multilingualism on L1 verb-related morphosyntactic production. A sentence completion task tapping into production of past and future reference in L1 Norwegian was auditorily administered to 75 neurotypical middle-aged and older university professors differing in degree of bi-/multilingualism. Language background and demographic information was obtained using an adapted version of the Language and Social Background Questionnaire (Anderson et al., 2018). Participants were auditorily presented with a source sentence (SS) and the beginning of a target sentence (TS), and were required to orally complete the TS producing the missing fragment, which included the verb appearing in the SS but in a different form (Table 1). Reaction times (RTs) were calculated by measuring the duration from the cue onset to the response onset. To identify these two time points, audio recordings’ spectrograms and waveforms were analyzed using Praat (Figure 1). Bi-/multilingualism was treated both as a categorical and as a continuous variable. Separate scores were computed for speaking, writing, listening, and reading. We considered 23 bi-/multilingualism-related variables. To avoid harmful multicollinearity, we excluded variables with unacceptable Variance Inflation Factors (VIFs) and fitted generalized linear mixed-effect models including bi-/multilingualism-related variables with acceptable VIFs as predictors and RT as the dependent variable. These models included eight predictors: cumulative proficiency in FLs in speaking, number of FLs known, ±active bilingualism, cumulative amount of use of FLs in speaking, amount of L2 use in writing, average proficiency in FLs in speaking, L2 proficiency in speaking, and age of L2 acquisition. Since a preliminary analysis revealed that the future reference condition elicited longer RTs than the past reference condition, the two models were fitted not only to the whole dataset, but also to its two subsets. None of the eight bi-/multilingualism-related variables above showed significant main effects. This was also the case when focusing on the production of future reference, the most demanding morphosyntactic condition. Results suggest that bi-/multilingualism does not affect time reference production in L1. The results will be discussed in light of influential bilingual production models (e.g., De Bot, 2007) and hypotheses about the effects of bi-/multilingualism on language production in L1.

ABSTRACT. At least under specific conditions of individual-level experience/engagement with bilingual language exposure and use, managing multiple languages can lead to structural and functional adaptations in the brain. However, the effects of bilingualism on the neural underpinnings of executive function (EF) remain understudied. Research using time-frequency representations (TFRs) has shown that EF tasks (e.g. Flanker task) modulate power within theta- and alpha frequency bands. These power modulations have been linked to a greater engagement of the executive control system (Cavanagh & Frank, 2014; Suzuki et al., 2018). Herein, we use EEG with a Flanker task to investigate how individual differences in language experience may modulate neurocognitive outcomes (specifically oscillatory dynamics). EEG and behavioral data were collected from 60 bilinguals (28 early bilinguals; 32 late-acquired L2 learners). Participants also completed the Language and Social Background Questionnaire (LSBQ; Anderson et al., 2018). TFRs were computed for both the incongruent- and congruent trials, and the difference between the two (Flanker effect vis-à-vis cognitive interference) were then 1) compared between the early- and later acquired bilinguals (via cluster-based permutations analysis) and 2) modeled as a function of individual differences in language experience using continuous measures of bilingualism derived from the LSBQ. At the group level, we predicted greater brain engagement in early compared to late bilinguals, specifically increased theta activation followed by alpha suppression for the Flanker effect. Furthermore, we hypothesized degree of active bilingualism would predict changes in alpha and beta bands in both early and late bilinguals. Finally, we predicted a correlation between reaction times (RTs) and power modulation within the alpha- and theta bands. Incongruent trials incurred significantly slower RTs than congruent trials, higher theta power in central electrodes 300-600ms post-stimulus onset, and decreased alpha power in centro-parietal electrodes 600-950ms post-stimulus onset. No significant differences were observed between groups for either behavioural or neural responses. However, individual differences analyses revealed significant correlations between age, age of acquisition, and usage of the non-societal language at home with alpha and beta band activity for late bilinguals, whereas only age effects were found in early bilinguals. Furthermore, when correlating alpha power with RTs, early bilinguals showed a negative correlation while later bilinguals show a positive correlation. Taken together, the results indicate adaptations towards differential brain recruitment to deal with the cognitive demands associated with variation in language experience.

ABSTRACT. The present study investigates the production of rhetorical questions (RQs) by adult heritage speakers (HSs) of Italian. It is commonly assumed that the grammar of adult HSs may differ from that of monolinguals, and that phenomena that combine different domains of language can be particularly challenging for HSs (e.g., Kupisch & Rothman, 2016; Polinsky & Scontras, 2019). RQs are an interface phenomenon par excellence, understudied in HSs and language acquisition more generally. RQs are syntactically interrogatives but their pragmatic function is similar to an assertion, which is used to signal the speaker’s attitude rather than requesting information (Biezma & Rawlins, 2017; Sorianello, 2019). RQs can be distinguished from information-seeking questions (ISQs) through a variety of linguistic cues, including discourse particles, morphosyntax and prosody. These cues vary across languages. The study examines the strategies that German-dominant HSs of Italian use to mark questions as rhetorical, focusing on prosody. We address the following questions: 1. Which prosodic cues do HSs use in Italian? 2. Do HSs use different cues compared to Italian monolinguals? 3. When compared to the cues that are used in German, can cross-linguistic influence be detected? In a production task, participants were given RQ- and ISQ-eliciting contexts, together with a recorded model sentence, as in (1). HSs were tested in both languages. (1) Ma chi lo mangia il melone? (RQ prosody) But who CL.it eat-3SG the melon ‘But who eats melon?!’ 40 HSs (27 f:13 m, age=18-45, mean=27.7) and 41 monolingual Italians (26 f: 15m, age=18-37, mean=25.9) took part in the study. Results showed that RQs were marked by a slower speech rate (longer duration) (p<.001) and a smaller pitch range (flatter contour) (p<.0001) than ISQs. Additionally, RQs had predominantly a low final contour (L%), while ISQs were mostly high or rising (H%, LH%) (see Figure 1). For these parameters, the groups behaved alike: although HSs had an overall slower speech rate, they exploited the aforementioned phonetic and phonological cues in the same way as monolinguals. Like monolinguals, HSs placed the nuclear pitch accent in ISQs either on the verb (mostly falling tones, H*+L and H+L*) or on the object (mostly rising tones, L*+H and L+H*). In RQs, instead, we find some differences (Figure 2). Both groups, when placing the nuclear accent on the verb, used either a falling accent (H+L*, also commonly found in declaratives, or H*+L), or a rising accent (L+H*, typically used to mark contrastive focus). However, monolinguals never placed the nuclear accent on the object, whereas HSs did so. A qualitative analysis shows that HSs used a configuration that is not found in monolingual data: a rising pitch accent on the final object noun followed by a fall in the boundary tone (L+H* L%). This configuration recalls the typical German “rhetorical contour” (Braun et al., 2019) and thus appears a sign of transfer. In conclusion, our study supports the hypothesis that HSs have a good knowledge of the prosodic properties of their minority language in their interface with pragmatics (Lleó, 2016), while also showing some differences due to transfer from the majority language.

ABSTRACT. Grammatical gender in Welsh is a persistent area of difficulty in child bilingual acquisition but not in adult control data (Binks & Thomas, 2019; Sharp, 2012). Typically gender production has been investigated through its use in combination with the complex morphophonological mutation system, which are consonant-initial alternations, e.g., the Welsh determiner ‘y’ triggering soft mutation on feminine nouns, cath > y gath (the cat) but not on masculine nouns, ci > y ci (the dog). This study extends these findings to disambiguate the representation and processing of Welsh grammatical gender, by establishing whether gender is as robust when encoded through or independent of mutations. Our research questions are:

1. Do Welsh-English adult bilinguals make use of grammatical gender in production and reading? 2. What affect does the mutation system have on the representation and processing of gender?

The first experiment of two included the BLP (Gertken et al., 2014), elicited imitation task and two cloze tests (English and Welsh). Data were collected from 40 self-reported Welsh-English bilingual adults (19 females; age: 19-64). The EI task contained four contexts to disentangle gender from mutations:

• Mutation independent of gender using pre-nominal adjectives triggering SM on nouns irrespective of gender: ceffyl [m] > hen geffyl (old horse) • Gender independent of mutation using the gendered numeral four: pluen [f] > pedair pluen (four feather) • Gender encoded locally using the gendered numeral two triggering SM: merch [f] > dwy ferch (two girl) • Gender encoded in distant contexts using 3rd person singular pronouns as distant references: troed [f] > ar ei droed (on his foot)

The EI task results show that participants performed best when gender was independent of mutations (M=94%, SD=8.1%), producing nouns in their bare form. Performance was next best when mutation was separate from gender (M=91%, SD=6.5%). However, accuracy decreases when the systems operate together, when gender was encoded through mutations locally (M=76%, SD=20.7%) and in distance contexts (M=82%, SD=13.8%). These results indicate a stronger representation of the gender system when used separately from mutations, suggesting that speakers encounter difficulties when the systems are used in conjunction with one another. It may be possible to account for the results under the MISH (Prévost & White, 2000), arguing that the speakers have underlying syntactic representations of gender but have trouble spelling out gender-marked forms when involved with the intricate mutation system during production.

A follow up experiment is currently underway, collecting data from at least 20 participants from experiment one. Battery items include a self-paced reading task to extend the findings from representation to processing. The results will indicate the moment where processing difficulty and/or facilitation arises regarding gender (and mutation) in Welsh, focusing on whether gender is more robust when independent rather than encoded through mutations. We also consider the effects of memory on the processing of gender, namely, working, procedural and declarative memory as measured by the Operation Span task, Tower of Hanoi and Continuous Visual Memory Test.

ABSTRACT. In this symposium contribution, we present a study that investigates the effects of third-age language learning on cognition. With the average life expectancy increasing, ageing is considered one of the greatest social and economic challenges of the 21st century (He, Goodkind, & Kowal, 2016) and healthy ageing is put high on the agenda. Recently, studies have looked at potential cognitive benefits of late onset language training studies for senior monolinguals (Bak, Long, Vega-Mendoza, & Sorace, 2016; Pfenninger & Polz, 2018; Ramos, García, Antón, Casaponsa, & Dunabeitia, 2017). However, these studies have found mixed results, partly because seniors’ language learning needs remain unclear (Ramírez-Gómez, 2016), and partly because training is often short (Del Maschio, Fedeli, & Abutalebi, 2018; Singleton & Pfenninger, 2019). We studied 19 senior citizens (65+) learning English longitudinally for a period of three months. During these three months they were tested three times (pre-test, post-test and retention test) on linguistic, cognitive, and socio-affective measures. Due to COVID the language course took place online via Zoom. As previously mentioned, part of the reason why mixed results have been found is because language learning needs are not clear. Therefore, we ascertain language learning needs by offering two types of teaching: with or without explicit grammar instruction. Concretely, this means we offer two conditions of the same English course, one where the classroom materials and homework include explicit explanations and mentions of grammatical constructs, and one where these explanations and mentions are replaced with extra input of the target structure. Importantly, we also investigate the durability of these outcomes by including a retention test three months after the course is finished. Our study, therefore, assesses what the cognitive effects of learning English through different methods are on seniors. Currently, the final wave of data collection is in progress, but will be finished by April. The preliminary data of 10 participants show a similar developmental pattern for both groups in some cognitive tests, and opposite patterns of development for the groups in other cognitive tests.

References Bak, T. H., Long, M. R., Vega-Mendoza, M., & Sorace, A. (2016). Novelty, challenge, and practice: The impact of intensive language learning on attentional functions. PloS One, 11(4), e0153485. Del Maschio, N., Fedeli, D., & Abutalebi, J. (2018). Bilingualism and aging: Why research should continue. Linguistic Approaches to Bilingualism, 8, 2-19. He, W., Goodkind, D., & Kowal, P. (2016). An aging world: 2015. Washington, D.C.: U.S. Government Publishing Office. Pfenninger, S. E., & Polz, S. (2018). Foreign language learning in the third age: A pilot feasibility study on cognitive, socio-affective and linguistic drivers and benefits in relation to previous bilingualism of the learner. Journal of the European Second Language Association, 2(1), 1-13. Ramírez-Gómez, D. (2016). Critical geragogy and foreign language learning: An exploratory application. Educational Gerontology, 42(2), 136-143. Ramos, S., García, Y. F., Antón, E., Casaponsa, A., & Dunabeitia, J. A. (2017). Does learning a language in the elderly enhance switching ability? Journal of Neurolinguistics, 43, 39-48. Singleton, D., & Pfenninger, S. E. (2019). Exploring a poorly understood variable: An agenda for classroom research on the age factor. Language Teaching, 52(1), 111-127.

ABSTRACT. Despite its low phoneme-to-grapheme correspondence, the English spelling system is more systematic than often assumed: It is ‘morpho-graphic’ (Berg & Aronoff, 2017) in that affixes are spelt in a consistent way (e.g., ‹ous› for word-final /əs/ in adjectives such as "famous") that distinguishes them from homophonous word endings without grammatical function (e.g., ‹us› for /əs/ in nouns such as "bonus"). Importantly, these regularities are not usually taught explicitly (in the L1 or L2 classroom) and spellers have to induce these regularities from the input. Previous research has shown that both native and non-native English spellers are aware of these implicit spelling regularities and apply them to novel words (Ulicheva et al., 2020; Heyer, 2021). However, the investigated non-native speakers had an L1 with the same script as their L2 English, which may allow for the transfer of established grapheme-to-phoneme correspondences as a starting point for the induction of spelling regularities. In order to test whether spelling regularities in an L2 are solely induced from the L2 input or interact with established L1 grapheme-to-phoneme correspondences, this study compares the application of morpho-graphic spelling regularities in English novel word spelling by spellers with different native scripts.

Thirty-six same-script (L1: German) and 35 cross-script (L1s: Chinese or Korean) learners of English spelt novel words ending in /əs/ presented orally in either an adjective context (i.e., biasing towards the affix spelling ‹ous›, see (1)) or a noun context (i.e., biasing towards an alternative spelling such as ‹us›, see (2)).

Results show that the groups differed with respect to the variety of spellings produced and the modulation by context: First, while the same-script group produced one of four majority spellings in 72% of trials, the cross-script group used these in only 58% of trials and showed more variation, producing 52 spellings (e.g., ‹orse›, ‹ouse›, ‹use›) unique to this group. Second, generalised linear mixed-effects models revealed that, for the same-script group, the proportion of ‹ous› spellings was significantly higher in the adjective (M=30%) than the noun (M=17%) context (p<.001), whereas different-script learners produced equal amounts of ‹ous› spellings in both contexts (M=16% and 14% respectively, p=.210). This indicates that same-script spellers detect the implicit spelling regularities in the input and use the appropriate spelling to mark a novel word’s lexical category (adjective vs. noun), while cross-script spellers are aware that ‹ous› is a common spelling for word-final /əs/ but do not connect this to adjectives.

In conclusion, the induction and application of morpho-graphic spelling regularities in an L2 without explicit instruction is influenced by learners’ native script. Despite acquisition of basic phoneme-to-grapheme mappings in the L2, learners with a different native script do not fully master phoneme-grapheme correspondences and are thus delayed in the detection of implicit regularities. In addition to the emergence of morpho-graphic spelling skills as an interaction of script effects and rule induction from the input, the present study will discuss the effect of the presence of word-class marking affixes in alphabetic vs. logographic languages as further influencing factor.

ABSTRACT. Background: With an estimated half of the world using more than one language every day, bilingualism is a norm rather than an exception. However, researchers in the field of language acquisition and processing traditionally focused on monolinguals and viewed bilingual speakers simply as “two monolinguals in one person”. Research into differences between monolingual and bilingual performance and processing is not only relatively new but has also been producing conflicting findings, which fuels the existing social ambivalence relating to the acquisition of two languages in childhood. However, studying heritage speakers (here defined as early bilinguals of a minority language (Montrul, 2006:161) is important not only from the point of educational policymaking but it is also central to our understanding of the architecture of language as it can offer a window into bilingual minds. Therefore, the population, which was first studied mainly by applied linguists, is now becoming of interest to theoretical linguists investigating the role of input and maturational factors in language acquisition. What makes heritage speakers an interesting group is that they acquire their HL as their first language, yet their linguistic competence is often different from that of monolingual native speakers, which has led to the formulation of the Incomplete Acquisition Hypothesis (Polinsky, 2006; Montrul, 2008). According to it, certain patterns are not fully acquired in HS’s minds. It is this divergent competence that is of interest in this paper and 3 aspects of proficiency are discussed in the context of speech samples: fluency, syntactic complexity and lexical diversity. Additionally, receptive grammar is also tapped into as there is a documented imbalance between HS’ receptive and productive skills with the former usually being much stronger.

The study: focuses on 7-9-year-old preadolescents growing up in Polish families living in the UK, whose speech samples have been collected, transcribed and coded for lexical and syntactic complexity, and fluency, and subsequently analysed. The results have been compared to monolingual control groups through multiple regression modelling.

Results: Although bilinguals are slightly less fluent, they generally fall within the norms for monolingual speakers for most variables. The most interesting finding, though, is that their language is more complex syntactically than that of monolinguals in both languages. This provides further evidence that HS’ language is not incomplete but that the outcome is divergent from that of monolinguals. It could also shed light on the role of creativity and imitation in language acquisition and cultural transmission, and provide further evidence that children imitate selectively when they have a better understanding of the function of a given construction or linguistic element. When the function is not fully understood, they imitate more faithfully (Klinger, Mayor and Bannard, 2016).

ABSTRACT. Bilingualism has been shown to contribute to more successful cognitive aging (CA) across healthy and clinical populations (Bialystok, 2021; Gallo et al., 2022). However, most research on bilingualism and aging to date has examined older individuals (typically defined as aged 65+ years old), with middle-aged populations being an under-researched age group. To better understand age-related changes in brain and cognition, it is important to study an earlier “preclinical” period as this provides a critical window for early detection of cognitive change during aging (Sperling et al., 2011). Previous research on bilingualism and CA has also largely treated bilingualism as a binary variable (that is, one is bilingual or not), as opposed to examining individual bilingual experiences that can lead to distinct adaptations in the mind and brain (DeLuca et al., 2019, 2020). Lastly, bilingualism is not a sole contributor to individual variability in cognitive aging trajectories. Other life experiences and lifestyle factors, most notably education, and occupational attainment (Darwish et al., 2018), sustained engagement in physical activity (Young et al., 2015), healthy dietary patterns (Clare et al., 2017) and cognitively active lifestyle (Wilson et al., 2021) are also suggested to contribute to the building up of reserves against symptoms of CA (Stern et al., 2020) and correlate with improved brain- and cognitive outcomes in the older age, but are seldom accounted for in detail in the bilingualism literature. Considering the above, the present study had the following aims: (1) to tease apart factors within bilingualism that contribute to neurocognitive outcomes; (2) to investigate bilingual experiences within a wider context of other lifestyle factors; and (3) to study older middle-aged adults (aged 45+ years old) to capture any potential early bilingualism-related neurocognitive changes. Data collection for this study is presently ongoing. We are gathering an exhaustive dataset on bilingual individuals' language experiences and use patterns, participant background, and lifestyle. This linguistic and demographic information is coupled to behavioral data tapping into aspects of executive control, episodic memory, and working memory. Electroencephalography (EEG) is measured both at rest and during the tasks to obtain recordings of brain activity, which can provide an insight into brain health or a “cognitive reserve signature” (Fleck et al., 2017, 2019). By combining all these measures, we will be able to build a comprehensive linguistic and neurocognitive profile and reveal predictive correlations with key experience-based variables. Initial results are discussed during the symposium. The outcomes of this study have the potential to elucidate and pinpoint the effects of bilingual experience over and above other, potentially confounding factors as the aging brain and individual variability in bilingual and life experiences are viewed in a more holistic manner.

ABSTRACT. In a recent study utilizing EEG neural oscillations, researchers found that when individuals process linguistic information under taxing conditions (e.g. in their second language (L2)) alpha and beta oscillations decline earlier in time (Rossi & Prystauka, 2020). Similar studies have also found changes in alpha/band as an effect of working memory demands, suggesting that changes in alpha/beta bands may be linked to neural processes underlying working memory and/or the recruitment of working memory resources (Wianda & Ross, 2019). To investigate the role of working memory in sentential language processing, we tested 29 monolingual native English speakers during a sentence comprehension task while EEG was recorded. Experimental items included correct sentences, sentences containing a lexico-semantic violation targeted to elicit an N400 component, and syntactic violations to elicit a P600 component. Critically, we explicitly manipulated working memory load between subjects, such that half of the participants (n=15) completed the task with a concurrent memory condition, and a second group completed the same task without the memory load condition (n=14). We predict that if working memory resources are taxed or recruited, we should observe the same neurophysiological patterns (in ERPs and oscillations) as seen in Rossi & Prystauka’s (2020) L2 group during the sentence processing task. Preliminary data demonstrate reductions in both the N400 and P600 components for trials that included semantic and syntactic errors, respectively, only in the memory taxed group. No changes were observed for the non-memory taxed group. These results suggest that the neural signatures of the memory load group reflect an impact on language processing while working memory is taxed. These preliminary data support our predictions in line with Rossi & Prystauka (2020) and support the basis of the possible relationship between working memory and language processing and potentially bilingual language processing. Currently, this project is finalizing data collection and the oscillatory analysis is underway.

ABSTRACT. It is controversial whether adult L2 learners below native-like proficiency levels are capable of target-like processing of L2 grammar. In this self-paced reading study, we explore whether highly proficient adult L2 learners whose L1 is Croatian converge with native speakers in their online comprehension of null and overt subject pronouns in Italian. We base our predictions on the Position of Antecedent Strategy (PAS), according to which the null pronoun is biased towards the subject antecedent and the overt pronoun towards a non-subject antecedent in intra-sentential contexts in Italian (Carminati, 2002). Croatian null and overt subject pronouns seem to exhibit the same antecedent biases as the Italian ones (Kraš, 2008).

Two groups of native speakers (n=48) and L2 learners (n=44) took part in two experiments in which they read 36 complex bi-clausal sentences with null or overt pronouns in the subordinate clause introduced by the adverbial quando (‘when’). The pronoun matched either the subject or the object of the main clause, depending on its gender. In one experiment the subordinate clause preceded the main clause giving rise to backward anaphora (e.g. Quando lei/lui/Ø è entrata/-o in ufficio dopo pranzo, Adriana ha salutato Roberto con un grande sorriso, ‘When she/he/Ø entered[fem/masc] the office after lunch, Adriana greeted Roberto with a big smile’) and in the other it followed the main clause giving rise to forward anaphora (e.g. Roberto/Adriana ha salutato Adriana/Roberto con un grande sorriso quando lui/Ø è entrato in ufficio dopo pranzo, ‘Roberto/Adriana greeted Adriana/Roberto with a big smile when he/null entered[masc] the office after lunch’). All sentences were followed by a comprehension question that targeted the pronoun antecedent. Additional 84 sentences were included as fillers.

In both experiments, L2 learners converged with native speakers in their online (and offline) comprehension of the pronouns. More precisely, both groups of speakers had a subject preference for the null pronoun and an object preference for the overt pronoun in backward anaphora, as predicted by the PAS. In forward anaphora, they had an object preference for the overt pronoun (in line with the PAS) and no detectable antecedent preference for the null pronoun (not in line with the PAS, but in line with the results of some previous offline studies). We compare these results with our previous closely related work with English-speaking L2 learners of Italian (whose L1 has only overt pronouns), who linked subject pronouns to the linearly closest antecedent in both types of anaphora; the native speakers showed antecedent preferences consistent with the PAS (Kraš, Sturt, & Sorace, 2014, 2015). Based on this, we conclude that L2 learners are capable of target-like processing of L2 grammar, at least when their L1 does not differ from their L2 as far as the relevant properties are concerned.

References

Carminati, M. N. (2002). The processing of Italian subject pronouns. Unpublished PhD thesis, University of Massachusetts at Amherst. 

Kraš, T. (2008). Anaphora resolution in Croatian: Psycholinguistic evidence from native speakers. In M. Tadić, M. Dimitrova-Vulchanova, & S. Koeva (Eds.), Proceedings of the Sixth International Conference “Formal Approaches to South Slavic and Balkan Languages” (pp. 67–72). Zagreb: Croatian Language Technologies Society – Faculty of Humanities and Social Sciences.

Kraš, T., Sturt, P., & Sorace, A. (2014). Native and non-native processing of Italian subject pronouns: Evidence from eye-tracking. Paper presented at the AMLaP XX Conference, Edinburgh, 3-6 September.

Kraš, T., Sturt, P., & Sorace, A. (2015). Processing forward anaphora in native and non-native Italian: An eye-tracking study. Paper presented at the 28th CUNY Conference on Human Sentence Processing, Los Angeles, 19-21 March.

ABSTRACT. Making active predictions about upcoming information is a crucial aspect of language comprehension. Second language users (L2ers) have been found to predict, although in a less stable way than first language users (L1ers) do. Various factors affect prediction, such as type of cue and L1-L2 similarities. In L2 Dutch for instance, van Bergen & Flecken (2017) found some evidence for prediction based on the semantic contrast “leggen” (put.LIE) and “zetten” (put.STAND) only among L2ers who have this semantic contrast in their L1. Hopp (2015) investigated the relative use of morphosyntactic and semantic cues in L2 prediction. While L1ers used both types of cues to predict, L2ers used only semantic cues but not morphosyntactic ones (case marking). However, in Hopp’s study, case was absent from participants’ L1 while semantic cues were available in their L1.

This study is part of an ongoing project about prediction in online and offline processing. In this first study, we investigate whether German L2ers predict based on semantic cues (contrast between put.LIE “stellen” and put.STAND “legen”) and/or morphosyntactic cues (gender) if their L1 (French) use gender marking but does not use this semantic contrast. 14 L1ers and 25 advanced L2ers of German (with French as L1) completed an online reaction time experiment implementing the visual world paradigm. Two similar sets of 32 stimuli were developed. Each stimulus consists of a German sentence played auditorily while four cliparts of objects placed on a table or on a chair (reproduced from van Bergen & Flecken, 2017) are presented visually, with three seconds’ preview. The displays in the “gender+semantic combination” condition contain twice the same object in either a lying or a standing position (one of them being the target) + a gender competitor + a distractor, while the auditory sentences contain a semantic cue in the verb (absent in participants’ L1) and a gender cue in the article. The combination of both cues is needed to predict the verb complement (see figure 1). In the “semantic-only” condition, the semantic cue in the verb (which direct translation is available in French) can immediately disambiguate which of the four objects will be spoken about. In the “no prediction” and “no disambiguation” conditions, sentences do not contain any cue. Additionally, in the “no disambiguation” condition, two alternatives (same object in both positions) could still be chosen even upon hearing the verb complement.

As visualised in Figure 2, analyses suggest a main effect of language, with L1ers being particularly faster at predicting based on semantic cues. Neither L1ers nor L2ers used gender to predict in this offline decision-making task, contradicting previous findings about L1 prediction based on morphosyntactic cues in online processing. L1ers may generally use morphosyntactic/gender cues but may have avoided the effort of integrating the semantic and the gender cues to make predictions. Instead, they may have waited for the complement to be expressed before making their conscious choice as there was no benefit for them in completing the task faster, which may point to the role of utility and goals in prediction (Kuperberg & Jaeger, 2016).

More data are currently collected to support these findings. A second, eye-tracking study with the same stimuli will examine whether L1ers and/or L2ers integrate gender with semantic cues to predict in online speech processing without decision making.

ABSTRACT. Heritage language (HL) bilinguals often exhibit more variable morphosyntactic comprehension and production patterns than monolinguals (Scontras & Polinsky, 2019), likely due to exposure to more variable input, more restricted language experience, and processing pressures. To contribute to a more comprehensive picture of individual variation in HL representation and processing, we examine the effects of individual language experience on the processing of pronominal clitic order in Bosnian heritage speakers (HSs) in Norway. Bosnian pronominal clitics cluster after the first word/phrase in a clause in the so-called Wackernagel clitic-second position, with a potential for a long distance from verbs to which they serve as arguments in terms of phrase/word number. This movement might represent a processing challenge and/or could be under the influence of societal language (SL) with post-verbal object pronoun placement. Indeed, bilinguals and potential attriters of HLs with the same or related clitic placement (Serbian & Bulgarian) in contact with English produce (Dimitrijević-Savić, 2008) or do not disprefer examples of post-verbal clitics in the marked non-target-like third position (Ivanova-Sullivan et al., 2022). In an online self-paced listening task, Bosnian-Norwegian HSs will listen to sentences with fragments containing clitics in unmarked and marked positions. They will also judge each sentence after its presentation on its well-formedness. In the unmarked target-like condition, the clitic is the second position in the unit, whereas in the marked non-target-like condition it is in the third position, with either a verb or adverb occupying the clitic position (Example 1). In a Gorilla-run online self-paced listening pilot study, native Bosnian-dominant participants in Bosnia and Herzegovina (n = 39) showed sensitivity to the clitic placement. Target fragments with the clitic in the marked third position were processed significantly slower. Sentences containing such fragments were judged as sounding worse than the sentences with clitics in the unmarked position. This slow-down did not continue into the post-target regions.

Example 1: Preamble: Kad se zaliva ovaj cvijet? When do you water this floweri? Target: unmarked, clitic pre-verb Tetka ga zaliva ujutru Aunt it waters in.the.morning marked, clitic post-verb Tetka zaliva ga ujutru unmarked, clitic pre-adverb Tetka ga ujutru zaliva marked, clitic post-adverb Tetka ujutru gai zaliva

Post-target 1: ali ja sam čitala but I read Post-target 2: da treba uvječe. that you should do it in the evening.

Before or after the task (order counterbalanced), HS participants will complete language proficiency tests and a newly developed comprehensive HL background questionnaire. We predict that (i) the sensitivity to the clitic placement will likely be modulated by different factors such as HL proficiency, quantity and quality of exposure to HL, positive HL attitudes, literacy training, the relationship between which will be first determined with Principal Component Analysis, and (ii) SL influence will manifest through increased preference for post-verb clitic placement, the extent of which will be modulated by participants' SL experiences. The results of the study will contribute to the understanding of the nature of HL representation and processing to account for the full spectrum of HL variability.

References

Dimitrijević-Savić, J. (2008). Convergence and Attrition: Serbian in Contact with English in Australia. Journal of Slavic Linguistics, 16(1), 57–90. http://www.jstor.org/stable/24599635 Ivanova-Sullivan, T., Sekerina, I. A., Tofighi, D., & Polinsky, M. (2022). Language-Internal Reanalysis of Clitic Placement in Heritage Grammars Reduces the Cost of Computation: Evidence from Bulgarian. Languages, 7(1), 24. https://doi.org/10.3390/languages7010024 Polinsky, M., & Scontras, G. (2019). Understanding heritage languages. Bilingualism: Language and Cognition, 22(5), 1–17. https://doi.org/10.1017/S1366728919000245

ABSTRACT. Children simultaneously acquiring a minoritized home language alongside a socially dominant language represent a large and growing group in the United States and beyond. Emergent bilinguals (EBs) remain underrepresented in language development research and their language skills and practices are often compared (unfavorably) to monolingual norms. Few studies have used longitudinal, naturalistic language samples to describe changes in practices unique to bilingual development. One such practice is code-switching, or the use of multiple languages within a single conversational exchange (Yow et al., 2018). Previous work suggests that while younger EBs (age 3-4) tend to code-switch to fill lexical gaps, the code-switching patterns of older EBs increasingly adhere to the sociolinguistic norms of the school environment and are less correlated with language proficiency (Halpin & Melzi, 2018; Montanari et al., 2019). By drawing on three waves of observations (age 3-6), we examine children’s code-switching and lexical diversity in speech during parent-child interactions. We also explore the relations between changes in code-switching and changes in lexical diversity. We aim to characterize this EB sample’s continuity and individual rank-order stability as they transition to school in early childhood (Bornstein et al., 2017).

Spanish-speaking parents with a 3 or 4-year-old child in the northeastern U.S. were visited in the summer before preschool entry (Wave 1, n = 35) nine months later (Wave 2, n = 32), and again 16 months later (Wave 3, n = 24). At Waves 1 and 2, parents were observed interacting with their child using a wordless picture book and two sets of age-appropriate toys provided by the researcher. At Wave 3, parents shared a wordless picture book and engaged in talk about one past event and one future event. Interactions were transcribed in CLAN (MacWhinney, 2000) and coded for children’s initiated codeswitches (utterances that contained at least one English word, following an utterance that contained only Spanish words). Lexical diversity was operationalized as word types per minute for Spanish, English, and overall.

As W3 transcription is underway, we report preliminary results from Waves 1 and 2. On average, children code-switched to English slightly more at Wave 2, with marginal significance (p = 0.052). Children’s individual rank-order within group was highly stable, such that children who code-switched more at W1 also code-switched more at W2 (r(31) = 0.89***). We observed significant increases in children’s lexical diversity in Spanish and English, with greater rank-order stability in English (see Fig. 1). Finally, changes in code-switching from Wave 1 to Wave 2 were significantly positively correlated with changes in English lexical diversity and overall lexical diversity, but were not correlated with Spanish lexical diversity, suggesting that code-switching may add to children’s linguistic repertoires without necessarily indicating declines in Spanish use (see Table 1). For this presentation, this analysis will be extended to include changes from W2 to W3. Based on the literature suggesting that older children code-switch for socio-pragmatic purposes more than to fill gaps in linguistic knowledge, we hypothesize that correlations between code-switching and lexical diversity will weaken from W2 to W3.

ABSTRACT. Bilinguals constantly needed to control language co-activation and to switch between their languages (Hoshino & Thierry, 2011). As a result, a large body of studies showed a better adjustment of cognitive control strategies in bilinguals (Morales et al., 2013; Morales et al., 2015). More importantly, recent research has shown that brief training in language switching facilitates bilingual’s cognitive control strategies (Liu et al., 2019). Our memory for future intentions, this is, prospective memory (PM), is also highly dependent on how we manage our cognitive control strategies. Thus, it can be suggested that the use and frequency of switching between languages could be a factor that modulates cognitive abilities in bilingual people. In fact, López-Rojas et al. (2022) has shown that the bilingual experience modulates the recall of future intentions. Specifically, they showed that early bilinguals differed from late bilinguals and monolinguals performing a prospective memory (PM) task. In a typical PM task, participants are asked to carry out an ongoing activity (ON), while also remembering to perform a prospective intention when they encounter a specific cue (PM cue) previously encoded. This study aimed to explore if practicing switching between languages can modulate bilinguals’ performance in a PM task. We evaluated 56 Spanish-English bilinguals that were asked to perform a language switching task (switching group) or between languages or not (passive control group) before performing a PM task where we manipulated the monitoring demands of the PM cue (focal/non-focal cues). Results showed faster response times for the switching group in the PM trials regardless of the focality of the cue. These results suggested that switching between languages modulates the prospective processes engaged in the recall of future intentions. In line with previous studies, these data pointed to the benefit of language switching over cognitive processes involved in prospective memory (Zhang et al., 2015). These findings remark the need to continue exploring the effects of different factors related to bilingual experience in memory processes.

ABSTRACT. Despite the growing body of research on third-age language learning, second language (L2) fluency remains an understudied topic. It is important to address this topic, as L2 fluency is known to facilitate communication (Rossiter, 2009). Whereas L2 fluency has been quite frequently investigated in the context of younger learners, it has, to our knowledge, not yet been investigated in depth in seniors. This exploratory case-study investigates L2 fluency in seniors learning an L2 during an online course. In addition to that, we investigate if and how motivation impacts fluency development in this age group. To summarize, we aim to examine 1) the development of L2 fluency over time, 2) motivation pre-, during and post-course and 3) the extent to which these two measures are related. To do so, we take Complex Dynamic Systems Theory (CDST) approach to second language development (De Bot et al., 2007). Two Dutch seniors (76-years-old and 80 years-old) followed a 12-week English as an L2 course in an online setting. The course consisted of online classes and homework sessions accompanied by short diaries in which the participants scored their daily motivation and perceived learning outcomes. Before the course started and after the course ended, the seniors completed the AMTB and an IELTS-inspired speaking task. To examine general L2 fluency gain, the seniors’ fluency performance on the IELTS-inspired speaking tasks before and after the course was rated. To examine L2 fluency development, speaking exercises completed during the homework sessions were transcribed and analyzed on six different fluency measures (i.e., speech rate, mean number of words per minute (pm), number of silent pauses pm, mean length of silent pauses, number of filled pauses pm, number of repairs/repetitions pm). To examine motivation development, the AMTB and the daily homework diaries were analyzed. This study showed that 1) seniors develop their L2 fluency even after 12-weeks of English instruction and show general trends in L2 fluency development, 2) no direct relationships between motivation and fluency were found, 3) fluency and motivation show different developmental trajectories for both seniors, suggesting that seniors’ L2 learning trajectories are heterogenous.

References De Bot, K., Lowie, W., & Verspoor, M. (2007). A dynamic systems theory approach to second language acquisition. Bilingualism, 10(1), 7–21. https://doi.org/10.1017/S1366728906002732 Rossiter, M. J. (2009). Perceptions of L2 fluency by native and non-native speakers of english. Canadian Modern Language Review, 65(3), 395–412. https://doi.org/10.3138/cmlr.65.3.395

ABSTRACT. The resolution of third-person pronouns has attracted substantial interest in psycholinguistics in recent decades. The most influential claim is that pronoun resolution is guided by the saliency of syntactic roles. In pro-drop languages such as Italian and Croatian, the null pronoun often refers to the most salient subject antecedent, and the overt pronoun to the less salient object antecedent (Carminati, 2002; Kraš, 2008). In other pro-drop languages such as Chinese, however, both overt pronouns and null pronouns refer to the subject (Li, Mak, & Sanders, 2016). Moreover, pronoun resolution might be subject to the influence of implicit causality (IC) bias (Fedele & Kaiser, 2015). For a sentence with an IC verb (e.g., John angered/praised Peter because...), the semantic bias of the verb indicates which referent (NP1 or NP2) is likely to be the cause and therefore is the preferred agent of the second clause.

The aim of this study is to investigate how implicit causality interacts with pronoun-type bias in two typologically distant pro-drop languages, Croatian and Chinese. We hypothesised that null pronouns are more acceptable in NP1-biased sentences than in NP2-biased sentences in Croatian, whereas overt pronouns will be more acceptable in NP2-biased sentences. In contrast, both null pronouns and overt pronouns should be better accepted in NP1-biased sentences in Chinese. The participants were adult monolingual speakers of Croatian (n=76) and Chinese (n=67). They completed an acceptability judgement task in an online questionnaire. Twenty sentences in 4 conditions were constructed with a similar structure in both languages (see Figure 1 for examples in Croatian and Chinese). We manipulated the direction of the verb bias (IC-NP1 vs IC-NP2) and the pronoun type (null vs overt). The strength of the verb bias was held constant, with minimum bias values of 70%. Participants were instructed to rate the naturalness of the sentences on a Likert scale ranging from 1 (completely unacceptable) to 6 (completely acceptable). Data was analysed using mixed-effect linear regression models.

Results presented in Table 1 show that null pronouns are in general more acceptable than overt pronouns in Croatian, while the opposite was observed in Chinese. The effect of verb bias was similar, however: for overt pronoun sentences in both languages, NP2 biased sentences were more acceptable than NP1 ones. Unexpectedly, for sentences with null pronouns, NP1 and NP2 sentences were of the same acceptability.

There is a cross-linguistic difference between Croatian and Chinese in pronoun-type bias. Croatian is more dependent on syntactic features, with the null pronoun sentences being more natural to native speakers. Chinese pro-drop has been considered more discourse dependent, and thus, sentences with null pronouns are in general less acceptable than those with overt pronouns. For null pronoun sentences, the implicit causality verb bias is strong enough to overwrite the null pronoun bias. For sentences with overt pronouns, the verb bias does not play a similar role. We will discuss the implications of these results for future online processing experiments.

ABSTRACT. Simultaneous interpreting involves language processing unparalleled in any other linguistic task, as interpreters have to manage and coordinate source language comprehension and target language production at the same time, often with no clue where the speaker is heading. This produces significant cognitive load, which, however, normally fluctuates across an interpreting task. Research has revealed several symptoms of increased cognitive load, detectable either in the interpretation or in the interpreter’s physical reactions (Chen 2017). The former include disfluencies such as filled and silent pauses, vowel lengthening and false starts (Plevoets & Defrancq, 2018; Collard & Defrancq, 2020; Gumul, 2021) that lie in the focus of our interest. Directionality means interpreting either from one’s foreign language (L2 or L3) into the native language (L1) or vice versa. International organizations such as the UN or the EU favour interpreting into L1 due to the widespread belief in the superiority of this direction – it tends to be seen as both easier for the interpreter and conducive to better quality (for a discussion of the so-called directionality debate, see, e.g., Bartłomiejczyk 2015). However, L1-L2 interpreting often fulfils a genuine market need, and this is definitely the case when the source language belongs to a number of EU languages of lower diffusion (i.e., rarely learned as L2), including Polish. A plethora of empirical studies shows directionality effects for various language combinations (e.g, Godijns & Hinderdael 2005; Yinyin & Posen 2018). In this vein, we would like to explore disfluencies in L2-L1 vs. L1-L2 interpreting by comparing output in each of the interpreting directions (English-Polish and vice versa) by twelve interpreters with Polish as L1. We additionally compare their interpretations from Polish into English with those delivered by three native speakers of English (who do not work in the opposite direction). Our quantitative analysis focuses on three types of disfluencies: anomalous pauses exceeding three seconds, hesitation markers (i.e., non-lexical fillers that assume the form of meaningless strings of sounds, often prolonged), and false starts (i.e., cases of word truncation, both retraced and non-retraced). The material under analysis comes from the extensive EP-Poland corpus (Bartłomiejczyk et al. forthcoming) containing material from eleven plenary debates of the European Parliament. The twelve interpreters with Polish as L1 have jointly contributed over five hours of output, and their three colleagues with English as L1 over three hours. If, in accordance with the common belief, cognitive load is lower for interpreting into L1, Polish-English interpretations by Poles should exhibit significantly more disfluencies of various types than both English-Polish interpretations by the same individuals and Polish-English interpretations by Brits.

  References

Bartłomiejczyk, M. (2015). Directionality. In F. Pöchhacker (Ed.), Routledge encyclopedia of interpreting studies (pp. 108–110). Routledge. Bartłomiejczyk, M., Gumul, E., & Koržinek, D. (forthcoming). EP-Poland: Building a bilingual parallel corpus for interpreting research. GEMA Online Journal of Language Studies. Chen, S. (2017). The construct of cognitive load in interpreting and its measurement. Perspectives, 25(4), 640–657. Collard, C., & Defrancq, B. (2020). Disfluencies in simultaneous interpreting: A corpus-based study with special reference to sex. In L. Vandevoorde, J. Daems, & B. Defrancq (Eds.), New empirical perspectives on translation and interpreting (pp. 264–299). Routledge. Godijns, R. & Hinderdael, M. (Eds.) (2005). Directionality in interpreting: The "retour" or the native? Communication & Cognition. Gumul, E. (2021). Explicitation and cognitive load in simultaneous interpreting: Product- and process-oriented analysis of trainee interpreters’ outputs. Interpreting 23 (1), 45–75. Plevoets, K. & Defrancq, B. (2018). The cognitive load of interpreters in the European Parliament: A corpus-based study of predictors for the disfluency uh(m). Interpreting, 20(1), 1–32. Yinyin, W. & Posen, L. (2018). Re-conceptualising interpreting strategies for teaching interpretation into a B language. The Interpreter and Translator Trainer 12 (2), 188–206.

ABSTRACT. Adult research has shown that structural priming effects become stronger in both L1 and L2 speakers with increasing statistical input (i.e., cumulative priming; Jaeger & Snider 2008; Kaan & Chun 2018; Kaschak et al., 2011) and, that this effect mirrors implicit adaptation or learning. However, the literature on cumulative priming effects in children is limited to some monolingual studies using blocked designs, often accompanied by post-tests (Gámez & Shimpi, 2016; Huttenlocher et al. 2004; Kidd, 2012a, 2012b) where children are primed with one structural option throughout one session, which rather reflects blocked learning. Because interaction is not only one structure, we aimed to investigate cumulative effects across alternating trials preferably with intervening fillers, resembling more variation in the structural input, following the design of adult studies. Furthermore, there is no research in bilingual children on cumulative priming effects and no child studies looking at development across several age groups. We conducted a video description priming study in German with alternating prepositional object (PO) ‘Dora bringt den Hasen zu Boots’ (Dora brings the rabbit to Boots) and double object (DO) structures ‘Dora bringt Boots den Hasen’ (Dora brings Boots the rabbit). This resulted in 16 prime-target trials interspersed with two intransitive filler-filler trials ‘Wendy schaukelt” (Wendy is swinging). We recruited 167 L2 German-speaking children and adults (L1 other than German, taking into account the dative alternation of their L1) aged 3-4 (N=41), 5-6 (N=52), 7-8 (N=35) and adults (N=39).We also recruited 193 native German-speaking controls aged 3-4 (N=60), 5-6 (N=51), 7-8 (N=45) and adults (N=37). Importantly, German is a DO biased language and leads to the question: to extent will both bilingual and monolingual children adapt the PO across alternating trials? The results of the logit mixed effect model revealed a main effect of trial suggesting that the PO production increases as a function of growing trial number in all groups. There was a main effect of language type indicating an overall larger PO production in the bilingual groups compared to their monolinguals peers. Furthermore, a two-way interaction of trial and the 3-4 year olds implied that the youngest children showed the strongest adaptation to the PO structure across trials compared to all older groups. Further post-hoc analyses revealed that susceptibility to adaptation weakens with growing age (see Fig.1). To our knowledge, this is the first demonstration of a developmental decline of implicit structural adjustments to infrequent structural exposure with growing age in both monolingual and bilingual children which is in line with findings showing that learning rate decreases with age (McClelland et al., 1995). Our data supports claims that at some point in time our accumulated experiences stabilize in our internal representations up to an extent, when our internal system doesn’t experience the necessity to be overwritten (Branigan & McLean, 2016; Chang et al., 2006).

ABSTRACT. Heritage children are bilingual children whose home/community language (heritage language) is not the societal dominant language (Montrul & Polinsky, 2021). Among others, structures with non-canonical word order, e.g., passives, have been shown to cause significant difficulties for heritage children to produce and/or to comprehend (e.g., Chondrogianni & Schwartz, 2020; Hao & Chondrogianni, 2021). Recent studies suggest that linguistic factors, e.g., the presence or absence (Hao & Chondrogianni, 2021) or the transparency of the morphosyntactic cues within the structures (Chondrogianni & Schwartz, 2020) and cross-linguistic influence from the societal dominant language, which further interacts with word order (Hao & Chondrogianni, 2021), modulate heritage children’s acquisition of non-canonical structures. Additionally, child-level factors, e.g., chronological age (Hao & Chondrogianni, 2021), and HL input quantity (Daskalaki et al., 2019), may also play a role. However, the empirical results are mixed concerning the role of these factors, partially because of the high heterogeneity of samples in previous studies. Additionally, it remains to be understood if heritage children process non-canonical structures in a monolingual-like way in real-time.

To examine how Mandarin-English heritage children process non-canonical structures in real-time, and to illustrate the effect of different linguistic and child-level factors in modulating HL development and processing with a more homogenous group of heritage children, we tested 32 Mandarin-English second-generation heritage children and 32 age-matched monolingual children on their production and comprehension (offline and online). We targeted three Mandarin non-canonical structures: BA-constructions (1a), BEI-constructions (1b) and OSV-constructions (1c), with differing word order (agent-initial, BA-constructions, versus patient-initial, BEI- and OSV-constructions) and the presence or absence of morphosyntactic cues (ba, bei vs. null). For the production task, comprehension-to-production priming was adopted. The three non-canonical structures all served as primes, and there were no lexical overlaps in any prime-target pair. For comprehension, we adopted a self-paced listening task with picture verification. In the task, the visual stimuli (Figure 1) either matched or mismatched the sentences, which were divided into five segments (Table 1).

Heritage children showed overall worse performance in production (Figure 2) and offline comprehension (Figure 3) across structures compared to age-matched controls. As for online processing (Figure 4), heritage children showed sensitivity to the cues in non-canonical structures in real-time as monolingual children did, but took longer to revise their initial misinterpretations. For the role of linguistic factors, the presence of morphosyntactic cues assisted performance for both heritage and monolingual children, such that the structure with a morphosyntactic cue, i.e., BEI-constructions, induced better performance relative to the one without, i.e., OSV-constructions, across tasks. Unique to heritage children, we observed that they performed better on BA-constructions, which share the agent-patient configuration with the structure overlaps between Mandarin and English, i.e., SVO-constructions, relative to BEI- and OSV-constructions, in production and offline comprehension. Importantly, heritage children overproduced SVO-constructions. These findings suggest an effect of cross-linguistic influence. Lastly, for the role of child-level factors, neither input quantity nor age predicted heritage children’s online processing. Both input quantity and age modulated their production, but only input quantity predicted their offline comprehension.

ABSTRACT. Bilingual experience has been argued to relate to domain-general cognitive control adaptations. However, as evidence is inconsistent, this relationship remains poorly understood. Some contributing factors to this inconsistency are likely the heterogeneity of language use patterns and experiences among bilingual participants (Luk & Bialystok, 2013; Leivada et al., 2020), and a relative lack of research into the functional brain adaptations which result from these experiences (e.g., Gullifer et al., 2018; Kalamala et al., 2021). The key aim of this study was thus to investigate how bilingual experience, and crucially individual differences thereof, affect domain-general cognitive control and its neural underpinnings.

A cohort of 223 healthy young adults (153 bilingual and 70 monolingual) completed a Simon and Flanker task, while their neural activity was measured with electroencephalography (EEG). Importantly, individual differences in bilingual experience-based factors (e.g. language diversity, intensity of use, duration of second language use, language switching and proficiency) were obtained via a confirmatory factor analysis run on the output of a battery of language demographic and fluency measures including the the Language and Social Background Questionnaire, the Bilingual Switching Questionnaire, Verbal Fluency tasks, Oxford Quick Placement Test and the National Adult Reading Test. Both event-related potentials (ERP) and time frequency representations (TFR) were calculated for the EEG data.

Both tasks showed reliable (behavioral) congruency effects. The ERP analysis revealed robust N2 and P3 components for the flanker and simon task, and the TFR analysis showed robust theta band activation over mid-frontal electrodes followed by alpha band suppression, primarily over posterior/occipital sites for both tasks. Results from between-group (monolingual vs bilingual) analyses showed behavioral and neurophysiological group differences. Monolinguals exhibited faster overall RTs but no group difference was found in the congruency effect for either task. The Flanker task also showed group differences in the neurophysiological correlates of the congruency effect, with P3 latency and alpha responses differing between groups. The Simon task showed a between-group difference for the Simon effect in theta responses, which was larger for bilinguals. These results indicate varying functional recruitment patterns for cognitive control demands as a function of bilingual experience. Regarding individual differences, results from structural equation modeling (SEM) showed specific experience-based factors to predict neural recruitment patterns which in turn predicted task performance, particularly for the flanker task (Fig 1). Specifically, increased extent of engagement in bilingual experience negatively correlated with theta power and P3 amplitude for the flanker effect, which in turn (theta power) positively correlated to size of the flanker effect (reaction time). This indicates an adaptation towards greater efficacy in handling cognitive control demands as a function of more intensive engagement in bilingual experience (DeLuca et al., 2020). Taken together, these data highlight the utility of an individual differences approach and use of multiple measures in examining bilingual effects on neurocognition.

ABSTRACT. Bilinguals make wrong-language intrusions extremely rarely in situations when their other language will not be understood. In the most established theory of this phenomenon, bilinguals inhibit the non-target language to prevent interference during target-language production (Inhibitory Control Model, Green, 1998). Such inhibition can act at the level of individual lexical representations (local inhibition) or at the level of the whole language (global inhibition). The most robust behavioral index of inhibitory control is a naming delay of previously inhibited words from the non-target language when this language becomes target, attributed to recovery from inhibition (and such recovery may last for at least ten minutes: Christoffels et al., 2016). This effect is more pronounced or only present for bilinguals’ dominant language, consistent with the Inhibitory Control Model’s feature that inhibition – and hence recovery from it – is proportional to the strength of the language it acts on (Calabria et al., 2012; Meuter & Allport, 1999). Such a slow-down in dominant-after-non-dominant picture naming is extremely robust, but it is unknown how bilingual inhibitory control dynamics affect connected speech. In connected speech, lexical retrieval delays (assumed to reflect retrieval difficulties) should be manifest in a reduced speech rate, more filled (uhs and uhms) and unfilled pauses (Hartsuiker & Notebaert, 2009), fewer words overall, and/or an increased use of cognates (words with the same meaning and a similar form across two languages), which may be less affected by inhibition. More speculatively, a greater use of easier-to-retrieve words such as higher-frequency and more generic words (expected in the face of lexical retrieval difficulties, e.g. in AD: Ostrand & Gunstad, 2020) would be inconsistent with the implication of the Inhibitory Control Model that more robust representations are more strongly inhibited. Method (Fig. 1). Eighty-six English-dominant Spanish-English bilinguals viewed two 8-min. videos (Tom-and-Jerry-type cartoons with no language) and after each viewing orally explained the video contents. Participants in the Changed-language group explained the first video in Spanish and participants in the Same-language group explained it in English (Phase 1). All participants explained the second video in English (Phase 2). Of interest was how the speech rate, fluency and quality during dominant English production in Phase 2 would be affected in the Changed-language group relative to the Same-language group. Also, half of the participants in each group explained the same two videos in Phases 1 and 2 (to target local inhibition), while the other half explained different videos (to target global inhibition). Bilinguals’ English and Spanish proficiency (Table 1) was assessed with tests of productive vocabulary (MINT, Gollan et al., 2012) and grammar knowledge (MELICET and DELE ), and a language history questionnaire. The data were analyzed with 2 (Phase 1 language) x 2 (Video Identity) ANOVAs. Contrary to the Inhibitory Control Model predictions, the Phase 2 English speech of the Changed-language group showed no significant differences from that of the Same-language group in speech rate, unfilled pauses and filled pauses. However, the Changed-language group produced fewer words overall (p = .04), fewer unique content words (p = .04), and words of higher overall frequency (p = .04) than the Same-language group (Figs 2-4). Video identity across phases had no effects except for unique content word frequency (Fig. 5). The remaining analyses will target a continuous measure of cognate status and, more exploratory, mean utterance length and number of clauses. In conclusion, connected speech in bilinguals’ dominant language showed clear effects of language control induced by previously speaking the non-dominant language. However, these effects were only partially consistent with strong predictions of the Inhibitory Control Model, and there was little support for a division of inhibition into local and global. Instead, our results may suggest that bilinguals possess compensatory measures to recover from adverse language-control effects on the dominant language to maintain speech fluency and quality – instead of being more disfluent or speaking more slowly, they used fewer and easier words.

ABSTRACT. Societies are progressively multilingual and mobile, with most of the world’s population learning and using more than one language across the life-span, either as a result of living in a highly bilingual environment or by acquiring the second language (L2) as classroom learners. In addition, increasing global migration calls for migrants to quickly acquire the new host country language, making bilingual language experience increasingly multifaceted, as exemplified by Heritage Speakers. The relatively recent recognition that bilingualism is more dynamic than previously thought has also started to be confirmed by nuanced effects on neural measures of language and cognitive processing. This body of work has started to reveal a fine-tuned system for language processing and cognition that is afforded by the variable experience that is bilingualism. In line with recent holistic models of bilingual language experience (e.g., Systems Framework of Bilingualism, Titone & Tiv, 2022) I will propose that second language (L2) and bilingualism research need to integrate the complex sources of variation at the individual level (e.g., linguistic, neurocognitive), with indices of language use and its variability measurable beyond the individual self, including sociolinguistic variation  that finally influences people’s language use, development, and neurocognition. In support of this idea, I will present data from multimodal approaches to illustrate how variability in individual measures of bilingual language use (such as Age of Acquisition, proficiency, bilingual language engagement, and code-switching) shape the signatures of bilingual language processing in young adults. Towards this goal, I will present recent neuroimaging data including EEG data (ERPs, Resting-State EEG), and structural/functional MRI data in young adults demonstrating that tracking individual variability in bilingual language use is key to understanding linguistic and neural outcomes. In addition,  to investigate how variability in social interactions at the macro-level (beyond the individual self) can inform linguistic and cognitive outcomes in the bilingual speaker, I will present data from a new line of research that uses personal social network and psychometric network modeling. The main goal of this new line of research is to start describing how the compositional and structural features of a speaker’s social network can shape the individual’s language and cognitive experience across different stages of life, and conversely, how a multifaceted bilingual experience can modulate one’s social network. Altogether, I hope to illustrate the benefits of integrating cross-disciplinary, multi-level data towards a better understanding of bilingualism.