ABSTRACT. Multilingualism is an interactional experience between an individual and her environment through multiple languages. For children and adolescents, language use in families and schools represent two of their language environments. In this presentation, I will address how the multilingual social contexts are associated with language and learning outcomes in children and adolescents using multi-modal data. First, I will show that linguistic diversity marked in the U.S. census was associated with perceived value of multilingualism and parental engagement of home language. Then, using structural and functional neuroimaging data, I will show that more extensive functional connectivity was observed between bilateral brain regions among multilingual adolescents when learning new information in their second language compared to monolingual adolescents learning in their only language. Finally, I will share how multilingual experiences among children can be conceptualized as an interaction of development and learning mechanisms. These studies converge to show that developmental outcomes are: (1) relevant to multilinguals’ proximal and distal language contexts; (2) more effortful when new knowledge is acquired through second language; and (3) multilingualism is both a developmental and learning experience interacting with her social and linguistic environment. Though this program of research was situated in the U.S., there are opportunities and challenges of transposing research from the English-dominant U.S. to Canada, specifically in French-dominant Québec. I will end the presentation with pilot work conducted in Canada to illustrate the diverse nature of multilingualism in the two North American countries and why social contexts matter in understanding multilingualism in development.

ABSTRACT. Bilinguals activate both of their languages when processing speech, resulting in phonological competition within and between languages (Canseco-Gonzalez et al., 2010; Marian & Spivey, 2003). While linguistic co-activation is known to influence bilinguals’ speech processing, the long-term consequences of phonological competition are understudied, and it remains unclear how co-activated labels impact subsequent cognitive processes like memory. The current study therefore examined how phonological competition and bilingual experience influence memory for visual items. English-speaking monolinguals and Spanish-English bilinguals completed a visual search task in which they had to identify an auditory target in English (e.g., candle) in a four-item search display. Phonological competition was manipulated by having critical distractor items overlap phonologically with the target item either within-language (e.g., candle-candy), between-language (e.g., candle-candado, lock in Spanish), or non-overlapping in either language (e.g., candle-wing). Participants later completed a task probing the recognition memory of critical distractor items (Within-language Competitors, Between-language Competitors, Control Items), as well as a battery of cognitive and linguistic tests. The effects of bilingualism and phonological competition on item memory were examined with generalized linear mixed-effects models using recognition accuracy as a binomial outcome variable. The effects of bilingualism were analyzed both as a continuous variable (Spanish to English dominance) and as a categorical variable with three levels (Monolinguals, English-dominant Bilinguals, Spanish-dominant Bilinguals). Across all participants, recognition memory for previously-seen within-language English competitors (M = 30.8%, SD = 18.6) was significantly better than for control items (M = 22.2%, SD = 16.7, p = .006), suggesting that within-language phonological overlap with targets facilitated subsequent recognition memory of competitor items. While the effect of within-language competition did not vary by language background, the between-language competition from Spanish was moderated by language group, as well as by individual differences in verbal working memory. Recognition memory for between-language Spanish competitors was significantly better than for control items in Spanish-dominant bilinguals (p = .003), especially those with higher verbal working memory. Memory for between-language Spanish competitors and controls did not differ for English-dominant bilinguals or monolinguals (p > .05), likely due to lower levels of Spanish-language activation. To summarize, within-language competition during encoding facilitated recognition memory in both monolinguals and bilinguals, while between-language competition facilitated item recognition in Spanish-dominant bilinguals. Our findings reveal that both phonological competition and bilingualism can impact memory for visual items, and that this effect is moderated by individuals’ linguistic and cognitive abilities. These results suggest that cross-linguistic co-activation impacts higher order cognitive processes such as memory, and have implications for applied and clinical settings. For example, it may be possible to leverage the effects of phonological competition to develop strategies for improving memory, such as grouping similarly sounding objects together to facilitate later retrieval. We conclude that linguistic labels of objects play an important role in episodic memory, and that both within- and between-language phonological competition should be considered when designing memory-related research and interventions.

ABSTRACT. Word segmentation is a core step in L2 learning that allows people to understand L2 speech. Previous studies have shown that the differences in how segmentation cues are used across languages can significantly impact L2 segmentation. Mattys and Bortfeld (2016) argue that cues involving lexical knowledge may be more critical for segmentation in L1, while prosodic cues such as stress may be more important for L2 segmentation. Studies conducted by Tremblay et al. (2012, 2018 and 2021) show that adapting to the use of prosodic patterns for L2 segmentation depends on how similar those prosodic cues are used in L1 versus L2. Tremblay et al. (2016) proposed that even phonetic differences in the realization of F0 rise and peak may cause difficulties in L2 segmentation. In this study, we look at the overall effect of F0 as a stress correlate to L1 and L2 Spanish listeners and the effects of detailed F0 peak manipulations. Our experiment comprises a mouse-tracking listening task that presents a target and competitor word each at an upper corner of the screen. Participants listen to temporarily ambiguous phrases of the type "palo marron" vs. "paloma roja" and use their mouse to choose the correct word. The position of lexical stress on the target word presents an early cue to find the correct parsing (PAlo vs. paLOma). For each stress placement, we manipulated F0 to create 3 conditions: neutral (no F0 rise/peak), L*H (common in Spanish), and H* (common in English). Peak manipulations were based on Ortega-Llebaria (2013). We recruited 25 L1 speakers and 15 L2 speakers (L1 English) for online testing. Our results showed significant effects within and between L1 and L2 groups. For within-group results, we found a significant effect of segmental information in both groups, reflected in their higher response accuracy for 3-syllable over 2-syllable words. This indicates speakers may give more importance to the 3rd syllable that biases “paloma'' over “palo” than to the initial stress pattern; note, this effect was not as big for L1 listeners. Similarly, the F0 neutral trial responses for both speaker groups were significantly slower than all trials that included F0 peaks. Such observations indicate that the presence of F0 peak is enough to make the stimulus more informative about word boundaries for L1 and L2 listeners. For between-group differences, L1 speakers seemed to find L*H patterns easier to process than H* ones, while L2 speakers’ results indicated a much more pervasive effect of segmental information; this implies that for most 2 syllable word trials, the 3rd syllable carried enough weight to overlook stress patterns in the final decision. Such an effect is still modulated by F0 where a peak presence makes the stimuli more informative; this is reflected in higher accuracy and faster RTs. In comparison, the effect of F0 peak configuration in L2 listeners is less clear. We discuss the possible causes and consequences of the segmental and F0 effects on L2 listeners.

ABSTRACT. The present study is a replication and extension of previous findings showing a disparity in one’s willingness to update political and non-political beliefs in the face of counterevidence (Kaplan, Gimbel, & Harris, 2016). We extended Kaplan et al.’s methodology to a population of bilinguals (n=30) in order to examine the role of the Foreign Language Effect (FLE) on belief maintenance. The FLE is the finding that bilinguals make decisions differently in their non-native language than their native one in a variety of domains. Within a dual-processing approach, FLE results are largely attributed to an attenuated reaction from the fast and automatic processing of System 1 (Alter & Oppenheimer, 2007; Greene & Haidt, 2002; Kahneman, 2011; Kahneman, Lovallo, & Sibony, 2011; Frankish, 2010), such that inherently emotional topics, tasks, or questions in the first language are less so in the second. Participants filled out a pre-screening survey on which they responded to 14 non-political questions (e.g., Thomas Edison invented the lightbulb) and 9 political questions dealing with traditionally liberal stances (e.g., Same sex marriage should be legal). Responses were given on a 1-7 likert scale where 1 was “Strongly disagree” and 7 was “Strongly agree.” Only participants who gave ratings of 6 or higher on at least 8 political and 8 non-political questions were asked to complete the experimental task. In the experimental task, participants saw the same 8 political and non-political statements to which they had ratings of 6 or higher on the screener. After each statement, however, five challenge statements appeared in an effort to measure change the from their original rating. Finally, participants saw the main statement once again and were asked to rate it on the same likert scale. The difference between the screener ratings and the post-challenge ratings was measured. Data from 30 English-Spanish bilinguals having a strong alignment with liberal politics showed that, like English monolinguals (n=28), non-political beliefs underwent greater change post-challenge than political statements (p < .001 for both groups), a finding shown in Kaplan et al’s original study. Also similar to Kaplan et al., both groups in our study showed a high degree of correlation between change on non-political and political trials, indicating that the more malleable on one, the more malleable on the other (English: p = .01, r = .632; Spanish: p < .001, r = .59). However, bilinguals’ post-challenge ratings for political beliefs were significantly lower than the monolinguals’ ratings in the same condition (p < .01). Bilinguals also had slower reading times for political challenge statements than monolinguals, perhaps indicating greater deliberation. Together, results indicate that confronting counterevidence to deeply held political views in a non-native language can reduce emotional reactions that might otherwise hinder cognitive and emotional flexibility in a native language context.

ABSTRACT. This paper investigates how speakers of Dutch who were born in Korea and adopted into Dutch families during childhood, recognize verbally expressed emotions later in life, both in their adopted language Dutch and in their birth language Korean, compared to non-adopted Dutch-born peers. In vocal emotion recognition, members of the same cultural-linguistic group recognize each other’s emotions more accurately than members of different groups (Laukka & Elfenbein, 2021). Internationally adopted children typically rapidly catch up in the adopted language with non-adopted age-matched children (Glennen, 2016). While they do not keep any conscious birth language memories, abstract phonological knowledge remains intact, and can be accessed later in life upon re-exposure (*****, 2017a,b; Pierce et al., 2014; Singh et al., 2011; Zhou et al., 2021). This study investigates if vocal emotion recognition follows similar patterns as linguistic processing in international adoptees. Testing the exact same participants as ***** (2017a,b), it addresses two research questions: RQ1. Do the adoptees exhibit a Current In-group Advantage, recognizing emotions expressed in the language of their current environment, Dutch, more accurately than in Korean, and is this advantage similar in size to the non-adopted controls’ in-group advantage in Dutch? RQ2. Do the international adoptees exhibit a Discontinued In-group Advantage, recognizing emotions expressed in Korean relatively accurately compared to the controls? Participants. 29 adult Korean adoptees in the Netherlands (mean age 32), adopted between 3 months and 7 years old, and a matched control group of 29 native Dutch speakers, were tested. Materials. We used a corpus specifically developed for the purpose of Dutch-Korean cross-linguistic comparison (*****, 2010a,b), including eight emotions balanced in valence (positive-negative), arousal (active-passive), and basicness (basic/non-basic), recorded by eight Dutch and eight Korean professional actors, always in the same nonsense phrase that was phonologically legal in both languages. Procedure. Participants heard 128 Dutch and 128 Korean stimuli (blocked by language), and indicated which emotion it expressed by clicking on one of the eight emotions or “neutral” on an emotion wheel (Fig.1). They did a pre-test and a post-test separated by 11 days during which they received training on Korean phonology, and further exposure to Korean within the context of diverse experiments (*****, 2017a,b). Results. Significant effects of valence, arousal, basicness, and test (pre/post) attest to the experiment’s validity. As Fig.2 and mixed effects models show, there were no differences between adoptees and controls in emotion recognition accuracy. Both groups recognized emotions more accurately when produced by Dutch than Korean speakers. The adoptees’ Current In-group Advantage was similar in size to the controls’ In-group Advantage. This is in accord with findings that adoptees’ linguistic functioning in the language of their current environment becomes similar to that of non-adopted peers soon after adoption (Glennen, 2016). Adoptees did not exhibit a Discontinued In-group Advantage; they recognized emotions in Korean similarly accurately as the controls. This differs from findings that adoptees retain phonological memories of the birth language (*****, 2017a,b). It suggests that, unlike phonological skills, vocal emotion recognition is not affected by early linguistic experience, and aligns with the adoptees’ present cultural-linguistic group membership.

ABSTRACT. We examine native English speakers’ ability to distinguish German vowels both in clear speech and in noise (energetic masking at two signal-to-noise ratios: 8dB and 0dB). In particular, we focus on the front rounded vowels /y/ and /ø/ in relation to the acoustically similar ones of the same vowel height (/i/ & /u/ and /e/ & /o/, respectively), as English speakers have previously been reported to map these onto the category of a back rounded vowel (e.g. Strange et al. 2005). All vowels are further contrasted with /a/ in a control condition. The aim of the study is to discover if a difficulty in perceiving a non-native vowel contrast is equally severe at each noise level or if there might be a perceptual shift so that the distinction between the front vowels that are acoustically more similar to each other (/i/ & /y/ and /e/ & /ø/) than to the back vowels of the same height (/u/ and /o/, respectively) becomes more difficult as the noise level increases. To this end, participants are presented with German-sounding pseudowords with varying vowels in an oddity discrimination task similar to the one in Darcy & Krüger (2012). The pseudowords adhered to both English and German phonotactics. In DIFFERENT trials, one of the three speakers produces the word with a different vowel than the other two (e.g. /piːm/ versus /pyːm/). In SAME trials, all speakers say the same word. Participants are then asked to indicate which speaker said a different word or if there was no difference. Both native English speakers with no prior exposure to German and a control group consisting of native German speakers are being tested. Current results support the previous finding that native English speakers tend to assimilate both round vowels of the same height to the same phoneme category in clear speech (/y/ & /u/, /ø/ & /o/) as they fail to hear a difference between them. As the noise level increases, however, the two acoustically closer front vowels of each height category (/i/ & /y/, /e/ & /ø/) pattern together and become less distinctive, regardless of rounding, but discriminating vowels of the same height becomes more difficult overall. In the noise conditions, the German control group shows a similar perceptual shift as the English speakers, though much less strongly. However, the contrast with /a/ remained relatively stable for both groups. In line with Cutler et al. (2008), current results suggest that L1 speakers can tolerate more noise distortion than L2 speakers. Further, performance of English speakers that had learned French and thus had previous exposure to the vowels /y/ and /ø/ was better overall than of those English speakers who had not learned French, particularly at the crucial contrasts of /y/ & /u/ and /ø/ & /o/. This suggests that previous exposure to non-native phonemes is an advantage that even holds in adverse listening conditions.

References:

Cutler, Anne, Maria Luisa Garcia Lecumberri & Martin Cooke (2008). Consonant identification in noise by native and non-native listeners: Evidence of local context. Journal of the Acoustical Society of America 124, 1264-1268. Darcy, Isabelle & Krüger, Franziska (2012). Vowel perception and production in Turkish children acquiring L2 German. Journal of Phonetics 40, 568-581. Strange, Winifred, Bohn, Ocke-Schwen, Nishi, Kanae & Trent, Sonja A. (2005). Contextual variation in the acoustic and perceptual similarity of North German and American English vowels. Journal of the Acoustical Society of America 118, 1751-1762.

ABSTRACT. During emotion perception, individuals from Western cultures are more distracted by information from the visual modality (i.e., facial expressions), whereas individuals from Eastern cultures are more distracted by information from the auditory modality (i.e., tone of voice). It is unclear, however, how multisensory emotion perception is influenced by the amount of daily exposure to a new culture and the length of immersion due to migration to a new country with distinct social norms. In the present study, Chinese-English bilinguals who immigrated to the United States from China viewed facial expressions and heard emotional but meaningless speech from their previous Eastern culture (i.e., Asian face-Mandarin speech) and from their new Western culture (i.e., Caucasian face-English speech). Facial expressions and emotional speech were presented either simultaneously or separately. When both modalities were presented simultaneously, participants were asked to judge either the emotion of the facial expression or the emotion of the voice and to ignore the other modality. Analyses of daily cultural exposure revealed that bilinguals with low daily exposure to the U.S. culture experienced greater interference from the auditory modality, whereas bilinguals with high daily exposure to the U.S. culture experienced greater interference from the visual modality. These results demonstrate that everyday exposure to new cultural norms increases the likelihood of showing a modality interference pattern that is more common in the new culture. Analyses of immersion duration revealed that bilinguals who spent more time in the United States were equally distracted by faces and voices, whereas bilinguals who spent less time in the United States experienced greater visual interference for emotional information from the West, potentially due to over-compensation when evaluating emotional information from the less familiar culture. While increased daily exposure to the new culture aids with the adaptation to new cultural norms, increased length of cultural immersion leads to similar patterns in modality interference between the old and new cultures. We conclude that cultural experience shapes the way bilingual immigrants perceive and evaluate the emotions of others.

ABSTRACT. In comparison to speech directed to adults, speakers modify their speech to infants (e.g. Fernald, 1992) and these modifications are thought to support infants’ first language acquisition, facilitating the extraction or segmentation of word forms from continuous speech (Thiessen, Hill, & Saffran, 2005) and word learning (Ma, Golinkoff, Houston, & Hirsh-Pasek, 2011). This advantage has also been found in adult ab initio learners’ word learning (Golinkoff & Alioto, 1995). The current study investigated whether the speech modifications present in infant-directed speech (IDS) also support adult ab initio learners’ initial speech segmentation, as segmentation is one of the first tasks facing foreign language learners.

German target (n = 24) and filler (n = 72) words were selected and sentences containing one target word were constructed such that half of the target words appeared frequently across sentences (8 times), while the other half appeared infrequently (2 times). A female, first language speaker of German produced the training sentences using both infant-directed (IDS) and adult-directed speech (ADS) and the target and filler words in isolation.

In Experiment 1, participants were first familiarized with the sentences and then in a test phase indicated whether they had heard the word (target or filler) during familiarization (see Gullberg, Roberts, & Dimroth, 2012). The target responses of adult English (IDS: n = 16; ADS: n = 16) and German speakers (IDS: n = 16) were evaluated using a mixed-effect logistic regression model including fixed effects for participant group and frequency and individual participant and item intercepts. Accuracy was greater for German compared to English speakers trained with IDS (β = 1.64, SE = 0.41, p < 0.0001) and the interaction with frequency was also significant (β = -1.36, SE = 0.31, p < 0.001; see Figure 1). Accuracy was only above chance for the group German IDS for frequent words (z = 6.63, p < .0001). These results suggest that the ab initio learners were unable to segment and subsequently recognize words, regardless of the modifications present in the input.

German speakers also struggled to recognize infrequent words in Experiment 1, suggesting high task demands. In Experiment 2, English speakers (n = 63) were tested in a simplified paradigm (Shoemaker & Rast, 2013), hearing first one of the sentences from Experiment 1 and then indicating whether they had heard a subsequently presented word (target or filler) in that sentence. A mixed-effect logistic regression model including fixed effects for register (IDS, ADS) and individual participant and item intercepts revealed a significant effect of register (β = -1.54, SE = 0.08, p < 0.0001; see Figure 2). Although target accuracy was above change for both IDS (z = 10.0, p < 0001) and ADS (z = 7.32, p < 0001), participants were more accurate at recognizing words that they segmented from IDS. Taken together, the results show that when task demands are lowered, the speech modifications present in IDS can help support ab initio learners in segmenting speech from a new language.

ABSTRACT. Even highly-proficient bilinguals may over-use pronominal forms in a non-null subject second language (L2) such as English (Contemori & Dussias, 2016) or a null subject L2 like Italian (e.g., Belletti, et al. 2007). Such residual indeterminacy in L2 referential choice could be partially due to an increased need for cognitive resources when computing interface structures between syntax and pragmatics such as referential expressions (e.g., Sorace, 2011). The present research tests this hypothesis by examining (written) pronoun production in nonbalanced Spanish-English bilinguals under time-pressure, a type of cognitive load. The hypothesis would be supported if bilinguals of lower proficiency (for whom L2 production is more effortful) produce disproportionally more pronouns under time pressure. Fifty-two Spanish-English bilinguals participated in a written picture-description task in English (bilinguals’ L2). Participants saw two pictures of one or two characters (with same or different gender) performing different actions. They first read a two-sentence description of the first picture (e.g., 2-characters/different-gender: A girl was arguing with a boy. The boy got really annoyed. Subsequently…), and then produced written descriptions of the second picture to continue the narrative. Participants completed two blocks, counterbalanced across-participants: one including time-pressure (15-seconds to complete the description) and one without time-pressure (30-seconds to complete the description). Participants’ English proficiency was assessed with the Michigan English Language Institute College Entrance Test (MELICET) and ranged between 22 and 45 (out of 50; Mean=37, SD=5.4). As expected, bilinguals produced most pronouns with only one referent and fewest pronouns with two referents of the same gender (all ps<.001). Bilinguals also produced more pronouns in the 15-seconds condition compared to the 30-seconds condition (p=.02), and bilinguals with lower English proficiency produced more pronouns than bilinguals with higher proficiency (p<.04). However, of most relevance, there was no evidence that proficiency level modulated the timing effect (p=.9). Finally, there was a marginal tendency to persist with the behavior established in the first block, despite the change in timing in the second block: Beginning under time pressure led to slightly more pronouns throughout, while beginning under no pressure led to slightly fewer pronouns throughout. We tentatively attribute this tendency to self-priming. In sum, bilinguals of varying proficiency levels were sensitive to the demands of speeded production: They tended to be less explicit in their choice of referring expressions under time-pressure than under no pressure. The current study also shows a stronger effect of proficiency on overall referential choice than prior research (e.g., Contemori & Ivanova, 2020), evidencing an L2 developmental pattern. However, the fact that lower-proficiency bilinguals were not more strongly impacted by the time pressure than higher-proficiency bilinguals is inconsistent with the hypothesis that indeterminacy in L2 referential choice may be due to an increased need for cognitive resources. This result replicates and extends previous findings that bilinguals were not disproportionally affected by other forms of cognitive load (perspective taking in privileged ground or keeping verbal or visual information in working memory) relative to monolinguals (Contemori & Ivanova, 2020).

ABSTRACT. Previous experimental work shows that negation processing can be direct in bipolar contexts where positive/negative states of affairs can be expressed by available lexical opposites (‘remember’/ ‘forget’) in monolingual speakers. However, in a unipolar context where such opposites are not available (‘sing’/‘not sing’), the processing first proceeds through the positive and only then the negative state of affairs (Orenes et al., 2014). We test this claim in two experiments with bilinguals to answer two questions. To what extent do a) the processing routes and b) the conceptual representation of the negated statement differ in bipolar/unipolar contexts when bilinguals process negation in their L1 and L2?

In Expt.1, Chinese-English bilinguals (N=40) were tested in a Negative Stroop Task. The participants were instructed to verify whether the positive/negative English/Chinese colour expressions match the colour they are printed on, either in bipolar (Figure 1) or unipolar contexts (Figure 2). The results showed that adding the parameter Language as a fixed factor significantly improved the fit of mixed-effects regression models in Expt. 1 (χ2(8) = 45.15, p< .001.), confirming that language drives changes in negation processing. Significant interactions were found between Language, Polarity and Context, indicating that negation processing routes were similar in L1 and L2 in the unipolar context; however, it was more difficult for the bilingual participants to process negation in L2 than in L1 for the bipolar context. Our explanation is that the translation equivalents of negators in English (i.e., not) and Chinese (bu ‘not’) could be processed in language-specific manner in bipolar contexts.

We also zoomed in on the conceptual representations of the negated statement and tested another 40 Chinese-English bilinguals in an Orientation Task (Expt.2). Participants compared positive/negative descriptions against pictures, with the location of a star manipulated in either bipolar (Figure 3) or unipolar contexts (Figure 4). Unlike Exp.1, Expt.2 would not start timing until a participant’s button press when they finished reading the sentence. That is, Expt.2 taps more directly into the conceptual representation of the negated statement since the RTs do not include the reading time or negation integration time. The results showed that adding the parameter Language as a fixed factor significantly improved the fit of mixed-effects regression models in Expt.2 (χ2(8)=16.08, p=.041), indicating that bilinguals responded faster to English than Chinese stimuli. The response speed advantage was found across both the negative and the positive conditions, and there was no interaction between Language and any other fixed factors, which we interpret as a signal that negation processing in the L2 is not necessarily more demanding than in the L1 (Ćoso & Bogunović, 2019). Our result aligns with recent research showing that compared with a cognitively more complex negation in L1, processing in L2 can be even easier (Zhang & Vanek, 2021). In sum, the results of the two experiments show that variations in the bipolar and unipolar contexts (Orenes et al., 2014) also hold for bilinguals, and that language of operation is an additional significant factor that drives processing variation.

ABSTRACT. How native (L1) and non-native (L2) speakers form linguistic dependencies between non-adjacent elements during sentence processing has been widely debated. Some argue that L1 and L2 processing are different, such that L2ers have difficulty using syntactic structures to guide dependency resolution (Clahsen & Felser, 2006), while others argue for L1/L2 similarity, with differences being explained by factors such as working memory and lexical processing (e.g., Cunnings, 2017; Hopp, 2018). To examine these issues, studies have examined whether L1 and L2 readers violate constraints on linguistic dependencies during processing, with mixed results (Felser & Cunnings, 2012; Lago & Felser, 2018). We contribute to this debate by examining subject-verb dependencies between dislocated subjects and verbs in two pre-registered self-paced reading experiments.

In each experiment, we recruited 96 L1 English speakers and 96 German L2 English speakers. The L2 participants’ mean age of onset of English learning was 9 (range 5–14) for both Experiments 1/2, and their mean Quick Placement Test scores were 53 (34–60) for Experiment 1 and 51 (30–60) for Experiment 2. The experiments used sentences like (1/2). In these sentences, we used a plausibility diagnostic, which has recently been used in direct-object filler-gap dependencies in L1 and L2 processing (Cunnings & Sturt, 2018; Fujita & Cunnings, in press), and extended it to subject-verb dependencies. Namely, we manipulated the animacy of the subject NP (“The thief”/“The table”) and a structurally unlicensed ‘distractor’ NP (“the robber”/“the locker”). That is, “the robber” and “the locker” in (1/2) are distractor NPs because forming a dependency between them and “stole” is structurally prohibited.

We predicted longer reading times at “stole” in (1c/d) than (1a/b) because the inanimate subject NP is an implausible actor of the critical verb (e.g., The table…stole) and thus should cause reading disruption. We also predicted that the distractor’s (in)animacy may affect dependency formation, particularly for L2ers if their processing is susceptible to structurally unlicensed elements. Across the two experiments, we manipulated whether the distractor was in a subject position (Experiment 1) or an object position (Experiment 2). These manipulations were to explore potential effects of the distractor’s saliency on dependency formation, and to generalise the findings using different manipulations.

Pre-registered analyses using mixed-effects models showed significant implausibility effects at the spillover region (ps < .05) in Experiments 1/2, with longer reading times for implausible sentences. A significant main effect of distractor (p < .05) was observed in the pre-registered analysis of Experiment 2, and an additional (non-pre-registered) analysis of the sentence final region in Experiment 1, indicating shorter reading times in the animate-distractor than inanimate-distractor conditions.

Implausibility effects indicate that the participants retrieved the subject NP after encountering the critical verb. Distractor effects suggest that both L1 and L2 participants’ processing were influenced by the distractor’s (in)animacy. Crucially, we did not find any significant evidence of L1/L2 differences in these distractor effects in both Experiments 1 and 2. These findings suggest that L1/L2ers form subject-verb dependencies in a similar manner.

ABSTRACT. CLI takes place when the languages of a bilingual exert influence on one another; vulnerable when the two languages have (i) surface structure overlap, and (ii) at the interface between two modules of the grammar (i.e syntax & pragmatics) (Hulk & Müller, 2001). Previous research has found that the overlapping variant is usually produced more frequently in the language with two variants, when compared to monolingual peers (Bernardini, 2003; Kupisch, 2014; Rizzi et al., 2013; Westergaard & Anderssen, 2015). The current study explores the outcomes of CLI when both languages have two syntactic variants, but with opposite pragmatic implications: possessive structures in Norwegian-Italian bilingual children. Both languages have the prenominal and postnominal possessive, and their use is context dependent. In Norwegian the postnominal possessive is the unmarked variant used for neutral contexts whereas the prenominal possessive is marked and signals contrast or emphasis, while the opposite is true for Italian (table 1). This combination of factors has good grounds for CLI to occur, but the direction of CLI and which factors play a role is currently theoretically unexplored. We designed an elicitation task that tested both neutral contexts (characters interacting with their own objects) and contrastive contexts (characters interacting with objects belonging to other characters). Thirty-one Norwegian-Italian bilingual children (15 female) aged 4-10 (mean=6;3) were tested in both languages. Most of the participants were residing in Norway (n=28). Children were tested in both languages. Our generalized linear model found (i) more postnominals in the contrast condition in Italian (p<0.05) indicating some intuition on the pragmatic use of the variants, (ii) more marked forms in the neutral context in Norwegian (p<0.001), (iii) a strong interaction of condition and language (p<0.001) signaling a higher usage of marked forms in the contrast conditions in Norwegian, thus being more target-like. To this model, we then added the effects of dominance (Italian-dominant, Balanced, Norwegian-dominant) obtained based on the data of a preliminary task. Dominance had an effect only on Norwegian, since the Italian system was too simplified. We found (i) a marginal significance (p<0.1) between balanced and Italian-dominant participants (ii) more post-nominal structures in neutral conditions in Norwegian-dominant than the balanced participants (p<0.05). Thus, the children are more target-like in Norwegian as their Norwegian dominance/proficiency increases (fig. 1). Since the responses in the Italian task were almost exclusively prenominal (fig.2). We will thus argue for a simplification of the Italian system to the unmarked and more frequent variant, similar to what the literature on heritage languages reports (Montrul, 2010). This cannot be attributed to CLI from Norwegian as we would expect the exposure to Norwegian to enhance the use of the postnominal variant. Nevertheless, the simplified Italian system may still influence the use of Norwegian variants: 11 participants were target-like in the contrast condition but overused the prenominal in the neutral condition. This is pragmatically infelicitous, but linguistically in line with the expected effect of Italian on Norwegian. Dominance has an effect, but it cannot influence an already simplified system, such as the heritage language.

ABSTRACT. This study examines cross-linguistic influence (CLI) by comparing Ln speakers of Norwegian with structurally different L1s: Spanish, Italian, English, Dutch, Russian, German, Polish and Turkish. We focus on possessive pronominal agreement in Norwegian. Norwegian possessive pronouns come in two sets. In local binding, the pronouns sin (masculine) si (feminine), and sitt (neuter) are used, which agree in gender with the possessee noun (example 1). In non-local binding, the pronouns hans (masculine) and hennes (feminine) are used, which agree in gender with the possessor (example 2). The L1s in our sample differ with respect to their pronominal agreement systems. While Romance possessive pronouns agree only with the possessee (example 3), English and Dutch pronouns agree only with the possessor (example 4). We therefore expect different patterns of CLI for the different L1 groups. Previous research has shown that L2 speakers of English whose L1 is a Romance language often produce errors with possessive pronouns (e.g. Collins et al. 2009, Anton Méndez 2011), especially when there is a mismatch between the gender of the possessor and the possessee (e.g. *John kisses her mother), where the gender of the possessee noun impedes the selection of the correct gender of the pronoun at the stage of sentence planning. Relatively less attention has been paid to the comprehension of pronouns, a notable exception being Lago et al. (2019), who tested the processing of possessor agreement in L3 German by Spanish-English bilinguals in a self-paced reading task. Spanish-dominant speakers were less sensitive to possessor agreement errors than English-dominant ones, suggesting CLI. However, contrary to the evidence from production studies, no support for a mismatch effect was found, possibly because in comprehension the pronoun is encountered before the noun, and therefore no interference from the gender of the noun can occur. The present study uses a similar design as Lago et al. (2019), but crucially, Norwegian pronouns are by default postnominal, which means that the gender of the noun can indeed be expected to influence the processing of the gender of the pronoun. Our study will therefore be able to answer the thus far unanswered question as to whether mismatch effects between the gender of the possessor and the possessee also apply in the domain of comprehension.

Our participants, who will be matched in terms of age of onset, length of exposure, and proficiency in Norwegian and English, carry out a timed judgment task and a self-paced reading task including grammatical sentences in which the gender of the possessor and possessee either match or mismatch, as well as infelicitous sentences containing gender errors.

A pilot study with L1 Norwegian speakers (N = 11) shows that the control group behaves in the expected manner: there is slow down after a gender violation in all conditions, but, not surprisingly, they do not respond to the mismatch condition (figure 1). The L2 data is currently being collected on the online platform Gorilla. The results of the study will have important implications for models of cross-linguistic influence and multilingual language processing.

Examples 1. Arthur møter ikke mye folk, men han i møter mora si i hver mandag Arthur meets not many people, but he meets mother POS.F each Monday “Arthur doesn’t meet a lot of people, but he meets his (own) mother every Monday.” 2. Mia møter ikke Lucas i så ofte, men hun møter mora hans i hver mandag Mia meets not Lucas so often, but she meets mother POS.M each Monday “Mia doesn’t meet Lukas so often, but she meets his mother every Monday.” 3. Gianni bacia sua madre John kisses POS.F mother “John kisses his mother.” 4. Jan kust zijn moeder John kisses his.M mother “John kisses his mother.”

Figure 1: Reaction time per region per condition in native speakers of Norwegian

References Antón-Méndez, I. (2011). Whose? L2-English speakers' possessive pronoun gender errors. Bilingualism: Language and Cognition, 14(3), 318-331. Collins, L., Trofimovich, P., White, J., Cardoso, W., &Horst, M. (2009). Some input on he easy/difficult grammar question: An empirical study. The Modern Language Journal, 93, 336–353. Lago, S., Stutter Garcia, A., & Felser, C. (2019). The role of native and non-native grammars in the comprehension of possessive pronouns. second language research, 35(3), 319-349. Sorace, A., & Serratrice, L. (2009). Internal and external interfaces in bilingual language development: Beyond structural overlap. International Journal of Bilingualism, 13(2), 195-210.

ABSTRACT. In this semantic priming study, we compared how bilingual and monolingual listeners recruit top-down semantics to process word pairs when both acoustic and visual speech cues from a face-masked speaker are impaired. In typical conditions, listeners negotiate between top-down sources of language input — such as previous knowledge, meaning, context, and inferencing — and bottom-up sources of input — such as visual speech cues and the acoustic information in the speech signal — to comprehend an utterance efficiently (Field, 2004). Bilinguals may show a greater reliance on bottom-up sources compared to monolinguals demonstrated by better recall for details and poorer recall for big-picture ideas and inferences, even in typical listening conditions (Field, 2004). In suboptimal conditions such as listening in noise, monolinguals adapt by recruiting meaning-related information and lexical information. However, bilinguals exhibit slower response times and are less accurate in SPIN tasks (Shi, 2012; Rogers et al., 2006) even when matching monolingual scores in ideal listening conditions. Because bilinguals have demonstrated a lesser ability to compensate for a loss of bottom-up input with top-down information compared to monolinguals, this study seeks to quantify to what degree the loss of visual and acoustic cues resulting from mask-wearing impacts bilingual speech perception. Mask wearing obscures both acoustic and visual cues used for speech perception; if bilinguals are less able to recruit top-down semantic information to aid processing, they experience more difficulty listening to masked speakers than monolinguals. To test whether listeners recruit top-down semantic cues, we conducted a semantic priming task with 100 videos of spoken word pairs presented sequentially to bilinguals and monolinguals similarly to Golestani et al. (2009). Word pairs were either strongly semantically related (e.g., "bee"/"honey"), weakly semantically related (e.g., "coffee"/"cream"), or unrelated (e.g., "desk"/"plane") with half of the primes appearing in all lists (e.g., weak: "bee"/"flower" and unrelated: "bee"/"chair") and the other half appearing in two lists. All pairs came from lists normed for syllable length and forward strength of association (Golestani, et al. 2009). All 100 word pairs were either spoken by an unmasked speaker or a masked speaker to determine the holistic effects of reducing both visual and audio cues on perception compared to typical perception. Participants viewed all masking conditions of all word pairs, while accuracy and reaction time of relatedness judgments were measured and compared across all stimuli conditions. Participants include 25 high-proficiency simultaneous bilinguals of English and Spanish (AOA: 0-5;0) and 30 English monolinguals. Although data collection will be completed in March 2022, preliminary results confirm bilinguals are slower in processing unrelated word pairs compared to monolinguals in masked conditions. In semantic-rich conditions (i.e., highly-related pairs), bilinguals are only slightly slower. This indicates bilingual listeners of masked speakers may struggle in processing when little semantic information is available, but are otherwise monolingual-like. Impacts are discussed in terms of the compounded impact of mask-wearing on bilingual perception.

ABSTRACT. Rhetorical questions (RQs) like in (1), are syntactically interrogatives with the pragmatic function of an assertion signalling the speaker’s attitude (Biezma & Rawlins, 2017). In German, RQs can be distinguished from information-seeking questions (ISQs) through phonetic (e.g., duration, voice quality) and phonological (e.g., pitch accents, boundary tones) cues (Braun et al., 2019) as well as lexical-syntactic ones, such as discourse particles (DiPs; Bayer & Struckmeier, 2017; Biezma & Rawlins, 2017). Research on the acquisition of RQs in general and their acquisition in second language learners (L2ers) in particular is scarce. Previous research on ISQs has shown that L2ers differ from monolinguals (L1ers) on a phonological (e.g., Arvaniti et al., 2006) and a phonetic level (Gut, 2009; Kainada & Lengeris, 2015). In addition, L2ers have difficulties conveying emotions or attitudes through prosody (e.g., Chen et al., 2015). This study investigates (i) which prosodic and lexical-syntactic cues Italian L2ers of German use when producing RQs in German and (ii) whether they produce RQs in a monolingual-like manner. Therefore, we conducted an elicitation task that provided RQ- and ISQ-eliciting contexts, together with a recorded model sentence, as in (1). A total of 40 L2ers (mean age = 33, range 24-45) who acquired German as an L2 after the age of six participated in the study, as well as 40 German monolinguals (mean age = 26, range = 19-44). (1) Wer mag denn schon Bananen?! (RQ prosody) ‘Who likes DiP bananas?!’ Phonologically, wh-RQs as in (1) are typically realized with a low rising pitch accent (L*+H) associated with the object noun (Braun et al., 2019). Preliminary results of the produced accent types are summarized in Figure 1. Our results show that both groups use this accent more frequently in RQs than in ISQs. However, L2ers produce this accent type significantly less often than L1ers (β = 1.01, SE = . 42, z = 4.16, p < .001) and use other accent types instead (e.g., H+L*: β = 2.52, SE = . 56, z = -4.5, p < .001). The falling accent (H+L*) is an accent pattern which is typically observed in Italian questions and declaratives and its overuse could be a result of transfer. In addition, L2ers produce significantly more pitch accents on the verb (32%) than the L1ers (2%; β = 3.75, SE = .52, z = 7.2, p < .001), which could also be an influence from Italian where pitch accents in wh-questions are typically associated with the verb (Bocci et al., 2021). Overall, L2ers seem to perform similar to L1ers, however quantitative differences can be observed. In the talk, we will provide an extensive prosodic analysis including, among other things, boundary tones, voice quality and duration as well as a lexical-syntactic analysis. Additional tasks and a background questionnaire control for factors such as length of exposure, language use, proficiency and irony comprehension. This will allow to explore the differences between L2ers and L1ers further and to provide insights into L2 acquisition of prosody in general and RQs in particular.

ABSTRACT. Cross-linguistic influence is a phenomenon in multilingual speakers in which Language A can influence language B for a specific property. According to Hulk and Müller (2001) CLI is likely to happen when there is (i) surface structure overlap, and (ii) interface between two modules of the grammar (i.e syntax & pragmatics) (Hulk & Müller, 2001). CLI is widely studied in bilinguals’ production (Anderssen & Bentzen, 2018; Kupisch, 2014; Nicoladis, 2006; Westergaard & Anderssen, 2015), but it can also influence acceptability in a way that they accept a structure more than their monolingual peers (Sorace et al., 2009). In the current paper we investigate CLI in possessive structures in Norwegian-Italian bilingual children. Both languages have two possessive variants: one prenominal and one postnominal, but these have opposite distributions in terms of contextual use and frequency (table 1). In each language the marked variant signals change of possessor when the context allows for it (1,2). We designed a forced-choice acceptability judgment task in OpenSesame Web. The task consisted of short animations in which a character either interacted with their own object (neutral condition) or with someone else’s object (contrast condition); two additional characters then described the scene, each using a different possessive structure, and the participants had to choose who described it better. Thirty-one Norwegian-Italian bilingual children (mean age=6;5) completed the task. The preliminary results for the responses (table 2) show that in Italian marginally more marked variants (postnominal) are used in the contrast condition when compared to the neutral condition (p<0.1) which is in line with the pragmatics; marginally more marked variants (prenominal) used in neutral condition in Norwegian than in Italian, which may indicate CLI from Italian to Norwegian. Statistically, there is no indication of CLI from Norwegian to Italian, but the raw numbers in table 2 suggest this: more postnominal possessives used in the Italian neutral condition than in the (appropriate) contrast condition. There was no significant effect in the reaction times (RTs) (table 3). Overall, the children were faster in the Norwegian task- here also the RTs reflected the contextual use as the fastest responses were for prenominals in contrast and postnominals in neutral condition. In Italian the fastest mean RTs is for the prenominal in neutral condition, but the children were slowest in choosing the postnominals in contrast condition, indicating that they still struggle with this choice. We conclude that there is bidirectional CLI for possessives in Norwegian-Italian children. Nevertheless, the results from the responses and RTs incompatible: CLI from Italian to Norwegian is more pronounced, but the responses for Italian are slower and indicate that the choice of variant in the contrastive condition may still be problematic. CLI for acceptability is complex, but these results show an interesting interplay between response and processing. In future analyses we will integrate dominance and age, and we plan to compare this with production data from the same group.

ABSTRACT. The medial temporal lobe (MTL) and the prefrontal cortex (PFC) are critical for episodic memory function, as well as other cognitive functions such as language learning (Nilsson et al., 2021). Episodic memory decline in healthy aging is argued to be due to a decrease in efficacy in PFC-MTL interactions (Nyberg, 2017). At the same time, MTL atrophy is associated with a higher risk of progression from healthy aging to cognitive impairment (Fotuhi et al., 2012) and from the latter to Alzheimer’s dementia (Apostolova et al., 2006). However, crucially, the MTL and the PFC are highly malleable brain structures that can change in volume and function as a result of behavioral interventions in older adulthood (Li et al., 2014; Lövdén et al., 2012). These brain changes are presumed to transfer to behavioral indices of episodic memory. As such, it is pivotal to investigate ways of promoting cognitive resilience in older adults with and without cognitive impairment. Language learning has great potential to boost episodic memory in these populations. This partially stems from the involvement of the MTL and PFC in this cognitive exercise (Antoniou et al., 2013). Foreign language learning has previously been associated with increases in hippocampal volume and the left PFC (Mårtensson et al., 2012) and enhanced associative memory in younger adults (Mårtensson & Lövdén, 2011), but much less is known about the impact of introducing a bilingual experience later in life on episodic memory in older adults. This poster will present preliminary results of an intervention study investigating the cognitive effects of language learning in older adulthood. Functionally monolingual Dutch older adults with and without cognitive impairment participated in a three-month online English course consisting of intensive self-study and online bi-weekly English lessons. Participants were classified as having cognitive impairment when they reported a diagnosis of Mild Cognitive Impairment or if they reported subjective cognitive decline and scored below the traditional Montreal Cognitive Assessment cut-off point of 26/27 (Nasreddine et al., 2005). Episodic memory was assessed using the Visual Association Test-Extended (VAT-E; Meyer et al., 2017), a test based on paired association. Performance on the VAT-E before and after the language course will be compared for both groups. Considering how the MTL and PFC are receptive to behavioral interventions, we expect improvements in episodic memory performance to ensue from language learning, with greater increases in the older adults with cognitive impairment relative to their healthy peers. If language learning is proven successful in enhancing episodic memory performance in older adults, it could be a promising healthy aging tool to prevent and/or treat late-life memory disorders.

ABSTRACT. Heritage speaker (HS) bilinguals are native speakers of their heritage language (cf. Rothman & Treffers-Daller, 2014). Like homeland speakers of the same language, HSs acquire their heritage language (HL) early and naturalistically. Differently, however, they often do so in a context of significantly reduced input and/or opportunities (over the lifespan) to use the language in a comparatively similar way. It is no surprise then that a substantial amount of research over the past decades has documented significant differences between HS and homeland native speakers (Montrul, 2016, Polinsky, 2018; Polinsky & Scontras, 2020) across a wide array of grammatical domains. Whether or not this is equally true in linguistic processing is not as clear given the disproportionate number of studies utilizing offline behavioral methods.  For example, behavioral HS studies would lead us to the conclusion that grammatical gender is vulnerable in Spanish as an HL in the context of the United States (US) (e.g. Montrul, Foote & Perpiñan, 2008).  However, will neural signatures during grammatical processing confirm or problematize such a conclusion (see Bayram et al. 2022)? Herein, we seek to begin to fill the gap in neurolinguistic methods applied to HS linguistic processing in an otherwise well-studied domain of grammar. To do this, in this study we focused on grammatical gender processing in Spanish, which has been well researched at both the behavioral and neural (ERP) levels across homeland native Spanish speakers and even L2 learners, but only behaviorally in HSs (no neuroimaging studies exist).  Given the significant numbers of homeland native EEG studies on grammatical gender, which converge on at least robust P600 (sometimes accompanied by eLAN), for gender agreement anomalies, we have a solid basis of comparison for the current study.

EEG/ERP data was collected from 30 Spanish-English heritage speakers in the US with diverse bilingual experiences. Participants were asked to read a total of 360 sentences in Spanish presented word-by-word, including number and gender agreement violations (concord on the adjective), while their brain activity was being recorded using EEG/ERPs. Preliminary analyses of the neural data reveal clear P600 effects, the brain response generally associated with morphosyntactic error processing (see Morgan-Short, 2014), for both the number and gender agreement conditions (see Figure 1). These preliminary results reveal both systematicity and target processing of these domains of grammar among HSs of Spanish in the US and provide strong support for the argued need and importance of incorporating online neuroimaging measures of grammatical processing in heritage language bilingual empiricism, to complement offline behavioral measures and, especially, as a checks and balance to them (Bayram et al., 2022).  Whereas behavioral studies on grammatical gender in HSs of Spanish in the US often conclude that gender is a particularly vulnerable domain, the present EEG data strongly suggest that gender is robustly represented and deployed for real-time processing qualitatively the same way as in homeland baseline speakers.

ABSTRACT. It is inherently human to want to belong to a social group. Social groups are often based on distinctive commonalities, such as religion, ethnicity, or supporting the same sports team. The social groups we belong to determine how we see ourselves, but also how we see groups we are not a part of. Simply put, we consider those who are similar to ourselves ‘us’ (the in-group). Anyone who differs from us is considered ‘them’ (the out-group). We are often quick to like and trust fellow in-group members, while we are negatively biased towards those in the out-group (Tajfel, 1974). Interestingly, a stronger sense of in-group identity is associated with more negative attitudes towards the out-group (e.g., Mummendey et al., 2001). The languages we speak also affect group-belonging and favoritism within the in-group. For instance, it was found that speakers of West-Frisian (spoken in the province of Fryslân in the Netherlands) experience a high sense of general well-being, despite scoring low on measures like education or income (Fries Sociaal Planbureau, 2020). The strong sense of belonging built through Frisians’ shared language and culture was argued to be one cause for this increased happiness. At the same time, exactly this strong Frisian in-group identity could mean that non-Frisian speakers, who make up a quarter of the Frisian population, are excluded. As a consequence, non-Frisians speakers may experience higher degrees of loneliness, or lower overall well-being. This has not yet been studied in detail, however.. Therefore, the present study investigates if non-Frisian speakers in Fryslân report lower social well-being (e.g., loneliness and self-reported amount of affection) than their Frisian-speaking counterparts. Language background data was collected from participants (around 13,000) who completed a questionnaire as part of the Lifelines project (Scholtens et al., 2015), a longitudinal cohort study that collects health data from 10% of the population of the Northern Netherlands. Data analyses are currently being conducted, and preliminary results suggest that Frisian speakers score better on some measures of well-being (e.g., they display a lower degree of loneliness) than non-Frisian speakers. The results of this work have important implications for future researchers and policy makers in the development of more targeted strategies to counteract social exclusion.

ABSTRACT. Does experience with two languages facilitate the pronunciation of difficult words in a foreign language? Previous studies have found that acquiring another, phonotactically less restrictive, language improves the production of phonotactically complex structures (Jabbari & Pourmajnoun, 2016; Masuda & Arai, 2010). The standard explanation is that positive transfer of structures from the less restrictive language helps improve the production. We wanted to know if there is a general effect of bilingualism, which is independent of the restrictiveness of the languages involved and thus goes beyond transfer of structures. It has been shown that bilinguals are better at word learning (Kaushanskaya & Marian, 2009; Singh et al., 2018), simple phonetic learning tasks (Antoniou et al., 2015), and L3 pronunciation (Kieseier, 2021). This suggests a general bilingual advantage in various linguistic tasks.

To test this hypothesis, we analysed productions of phonotactically complex Slovak words by adult Spanish monolinguals and Spanish-Basque simultaneous or sequential bilinguals (total n = 227). While both Spanish and Basque are phonotactically restrictive, the Slovak test words contained complex clusters of up to five consonants (e.g., šmrnc [ʃmr̩ n͡ts] ‘elegance’). We analysed how age of acquisition, type of bilingualism (monolingual, dominant bilingual, balanced bilingual), skill level in the L2, and percentage of L2 used in everyday life affected production accuracy of the Slovak words. Production accuracy was operationalised as feature-weighted linguistic distance between the target and the production and calculated with the R package alineR (Downey et al., 2017).

Preliminary results revealed that while there was a slight trend for Slovak production accuracy to improve with higher self-reported L2 skill levels, none of the predictors showed statistically significant effects.

This can be taken as an indication that experience with several languages per se does not boost production of phonotactically complex words in a third language, contrary to the enhanced skills in word learning and phonetic learning. In a study similar to the present one, Hanulíková et al. (2012) did not find a relationship between L1 vocabulary skills and production accuracy in complex Slovak words. It seems therefore that experience specifically with phonotactically complex languages is necessary to develop this skill. Neither L1 skills nor general experience with two languages, both of which have previously been shown to facilitate aspects of L2 acquisition, showed an effect on production accuracy in complex Slovak words. However, more language pairs need to be tested in order to reach reliable conclusions.

ABSTRACT. Much neurophysiological (EEG) research has been devoted to examining the interplay between affective word processing and bilingualism (e.g., Velez-Uribe, 2021; Jończyk, 2016). However, previous studies have mostly tested affective processing using visual stimuli (i.e., sentences presented in a written form), and have not been devoted to investigating the processing of audio materials of a different affective value. Since in widely bilingual society where intercultural communication is omnipresent, taboo words (including swear words) appear more frequently in an audio form than in writing, testing auditory stimuli would increase the ecological validity of the investigation of how taboo words are processed in the native (L1) and non-native (L2) language. Thus, the aim of our event-related potential (ERP) study is to test the processing of auditory stimuli containing taboo and neutral words in the bilingual context. .

The study, which is currently being conducted, includes testing 30 native speakers of Polish who are proficient learners of English as their second language. They are all aged 21–29, and are students of the Faculty of English (Adam Mickiewicz University, Poznan). Each participant is presented with two blocks of words (one in each language) presented auditorily. Each block contains180 words matched on valence, arousal, concreteness, number of syllables, and frequency. (Imbir, 2016; Warriner, 2013; Brysbaert et al., 2014; van Heuven, 2014, Mandera, 2015). Additionally, they were all controlled for their volume and pitch. The words included taboo (e.g., dickhead, handjob) and neutral words (e.g., surname, shoelace). EEG data are being collected using a 64-channel EEG system, with ERP analyses focused on the N400 response as an index of lexico-semantic processing. The presentation will be devoted to presenting the results of the study, thus providing insights into how bilingual speakers process taboo words presented in the auditory form, in L1 and L2.

ABSTRACT. Russian differs from many other languages in that it has two basic color terms for blue: goluboj ‚light blue‘ and sinij ‚dark blue‘ (cf. Corbett & Morgan, 1988; Morgan & Corbett 1989). German, on the other hand, possesses only one basic term blau. It has been shown by previous research that sinij and goluboj represent basic color terms for monolingual speakers of Russian, but not for American-Russian bilinguals. While late bilinguals who emigrated as adults did not differ from the monolingual controls, heritage speakers exhibited a semantic shift. For some of them, sinij became the sole basic term for ‘blue’. For others, both terms functioned as subsets of a single category blue, just as ‘dark blue’ and ‘light blue’ in English (Andrews, 1994, for a similar finding see Pavlenko et al., 2017). Our own study revisits this categorical distinction on the basis of data obtained from different groups of Russian-German bilinguals. In contrast to the previous studies, we applied a more rigorous methodology to shed a clearer light on the acquisition and loss of these categorical distinctions in Russian by Russian-German bilinguals. We report on data from an experiment that targeted the granularity of the lexical distinction between sinij and goluboj. To explore the category boundaries of these two terms, we created twelve color chips, ranging from light blue/goluboj to dark blue/sinij. These chips were used to color different objects depicted on a single sheet of paper. In a first part of the experiment, participants were exposed to sheets that showed six different objects each, all of them depicted in a different basic color, with one object depicted either in a prototypical sinij or prototypical goluboj shade. As a second step, participants were shown two sheets with six objects on each sheet, all of them depicted in different shades of sinij and goluboj (representing the twelve color chips selected for the study). In a pilot study we tested (1) eleven heritage speakers born in Germany (aged 10-22 years), (2) ten early bilinguals (aged 24-45 years) who came to Germany around puberty and (3) nine late bilinguals (aged 50-68) who came as adults (between the ages of 28 and 43). Currently we are expanding our data base for all three groups and we will also add age-matched monolingual control groups from Russia. Our preliminary results reveal that the term sinij is used in all three groups to denote objects showing the respective color (cf. Fig. 1). Goluboj, however, is often replaced by sinij and/or secondary terms (e.g. birjuzovyj ‘turquoise’). The semantic shift is especially obvious for the heritage speaker group where 1/3 does not use the term goluboj for light-blue objects altogether. Only the late bilinguals use goluboj consistently. The early bilingual group occupies an intermediate position, as almost 1/3 replace goluboj with sinij in the first part of the experiment. In the second part they more often resort to goluboj, but goluboj clearly starts to function as the bright variation of the single category ‘blue’, denoted by the term sinij.

ABSTRACT. Speakers of languages with strict Subject-Verb-Object word order may use verb-semantics (eat) to predict an upcoming object (cake; e.g., Altmann & Kamide, 1999), whereas speakers of languages with flexible word order (e.g., Turkish) may also rely on case-marking cues (e.g., Özge et al., 2019). Language experience has been argued to modulate prediction skills of monolingual speakers and late bilinguals (e.g., Karaca et al., 2021). However, little is known about whether early bilingual adults use similar predictive processing strategies as monolinguals, and whether language experience modulates their prediction skills. The current study therefore investigated to what extent (1) early bilingual adults, like monolinguals, are able to use case-marking on the first noun phrase (NP1) to predict the second noun phrase (NP2), and (2) language experience modulates predictive processing.

In a visual world eye-tracking experiment, 24 Turkish monolingual adults (Mage=26.33, SD=6.25) and 25 Turkish-Dutch bilingual adults (Mage=26.68, SD=5.21) listened to Turkish sentences in which case-marking on NP1 (accusative/nominative) and verb position (sentence-medial/sentence-final, counterbalanced) were manipulated (Table 1), while looking at a visual scene with three pictures (cf. Özge et al., 2019). The pictures represented the NP1 (e.g., rabbit), a plausible patient in a context where NP1 was agent (e.g., carrot), and a plausible agent in a context where NP1 was patient (e.g., fox). In verb-final sentences, case-marking was the only predictive cue, whereas in verb-medial sentences, verb-semantics was also available. All bilingual participants were exposed to Dutch before the age of 4. Language experience, measured as relative amount of language exposure, use and literacy activities in Turkish compared to Dutch, was assessed using a language background questionnaire.

Fixations to agent versus patient pictures were analyzed using mixed effect logistic regression during the time window between NP1 and NP2. The results showed a significant interaction between Time, Condition (accusative vs. nominative), and Group (monolingual vs. bilingual) in verb-final (β=-0.452, SE=0.05, z-value=-9.981, p<.001) and verb-medial sentences (β=-0.302, SE=0.04, z-value=-7.842, p<.001). For verb-final sentences, a significant interaction between Time and Condition demonstrated a prediction effect for monolinguals (β=0.257, SE=0.03, z-value=8.678, p<.001), but not for bilinguals (β=-0.193, SE=0.03, z-value=-5.612, p<.001). For verb-medial sentences, a significant interaction between Time and Condition revealed a prediction effect for monolinguals (β=0.411, SE=0.02, z-value=17.245, p<.001) and bilinguals (β=0.119, SE=0.03, z-value=3.880, p<.001), albeit smaller in magnitude for bilinguals (Figure 1). For bilinguals, the predictive patterns in the verb-medial sentences were significantly modulated by relative exposure to Turkish (β=0.082, SE=0.037, z-value=2.209, p=0.027) and relative literacy activities in Turkish (β=0.284, SE=0.112, z-value=2.544, p=0.011), but not by relative Turkish use (β=0.051, SE=0.068, z-value=0.746, p=0.456).

Our results suggest that early Turkish-Dutch bilingual adults are able to use case-marking cues predictively when integrated with verb-semantics (i.e., verb-medial sentences), but not when presented alone (i.e., verb-final sentences). Moreover, the more they were exposed to Turkish and the more they engaged in literacy activities in Turkish, the better their prediction skills. These findings provide evidence for the use of morphosyntactic cues in predictive processing of early bilinguals whose languages do not use the same type of cues, and the prominent role of language experience in predictive processing.

ABSTRACT. Eye tracking can be used to measure language processing in real-time and without a need for an overt response. This is particularly advantageous for studies with participants with aphasia (PWA) – a language impairment resulting from brain lesion – who have language production deficits. Previous eye tracking studies show that neurologically healthy bilingual speakers predict upcoming nouns (e.g., “bread”) while hearing constraining verbs (e.g., “eat”) in sentences (Dijkgraaf, Hartsuiker & Duyck, 2017). An eye tracking study with PWA demonstrated that the participants showed preferential looking patterns to the target noun at the offset of the constraining verb (Mack, Ji, & Thompson, 2013). Prediction processes in bilingual people with and without language impairment are not well understood. The purpose of our study is to investigate whether bilingual adults with or without aphasia predict upcoming nouns in sentences with constraining verbs, and whether language proficiency, working memory performance, and processing speed mediate prediction ability (Huettig, 2015). Forty neurologically healthy Norwegian-English bilingual adults (age 20-80 years) and four Norwegian-English speaking PWA were enrolled thus far. Verb-mediated prediction ability was measured via a visual world paradigm in Norwegian and in English. The experiment included 36 sentences with constraining verbs (e.g., “Mary eats Anna’s pizza”) and 36 sentences with non-constraining verbs (e.g., “Mary looks at Anna’s pizza”). Language proficiency was assessed via a word translation task, verbal-fluency tasks, and self-report. Working memory was measured with digit span and Corsi block tasks, and processing speed through reaction times on the Flanker and Corsi block tasks. The PWA were also tested with the Comprehensive Aphasia Test (CAT) in both Norwegian and English. Preliminary analysis of the neurologically healthy participants demonstrated that the participants predicted upcoming nouns in both languages, but they were faster at predicting in L1 (Norwegian) than in L2 (English). For PWA, results from the CAT revealed varied patterns of strengths and deficits in the two languages. Three of them were most proficient in Norwegian, while the fourth one was equally proficient in Norwegian and English. The eye tracking results for the PWA showed great interindividual variability but suggest that PWA can predict upcoming information in both languages. We also found correlations between processing speed and prediction ability for all participants. We are in the process of collecting data from additional participants and examining the predictive value of language proficiency in the eye tracking results. Our findings will elucidate predicting patterns in typical and atypical language processing in L1 and L2.

ABSTRACT. Visual colour perception is a complex process, common across humans, triggered by photoreceptors’ detection of light. Linguistic colour categorization, in contrast, is language specific. Languages differ in the number of basic terms to describe colours. Previous studies reported that speakers of languages that have different words for light and dark blue (e.g., Russian siniy and goluboy) display a colour category effect: Russian speakers discriminate colour chips sampled from these two linguistic categories faster than speakers of languages that use one basic colour term for blue (e.g., English ‘blue’). In Winawer et al. (2007) the colour category effect was reported to be disrupted when participants engaged in a verbal interference task, suggesting that active language use can modulate the linguistic category effect. The current study focused on how momentary language use affects colour perception in Lithuanian-Norwegian bilinguals performing a speeded blue colour discrimination task, with and without verbal interference in their L1-Lithuanian and in their L2-Norwegian, and a colour identification task to determine the location of colour boundaries, individually, for each participant. The analyses of the colour discrimination tasks were performed for each individual and aligned to their specific category boundaries to measure their within- and between-category colour perception. Two control groups, native monolingual Lithuanians living in Lithuania (LL) and native Norwegians in Norway (NN), performed the same colour discrimination tasks without interference and with the verbal interference in their native languages. Participants’ colour discrimination was tested following the design of Winawer et al. (2007) where, on each trial, participants saw three colour chips: one colour chip on the top of the screen and two colour chips on the bottom of the screen, and had to choose, which of the two bottom chips matched the top chip best. In the verbal interference blocks, we compared Lithuanian-Norwegian bilinguals’ between-category and within-category colour discrimination while retaining 8-digit combinations in Lithuanian and in Norwegian. The current study demonstrates that colour perception momentarily affected bilinguals’ language activation mode. First, in bilingual LN speakers, the colour category effect - faster between-category than within-category colour discrimination - was observed when the colour discrimination task was performed under verbal interference in Lithuanian - that has two linguistic categories for blue - but not in Norwegian, indicating that the active language modulates colour perception. Second, under verbal interference, native monolingual (LL) speakers of Lithuanian showed the colour category effect, while native speakers of Norwegian, that has one basic colour term for the same part of the colour continuum, did not. Our results support the weak linguistic relativist approach and highlight how two domains that are seemingly so different - language and perception - actually interact dynamically within the same individuals.

ABSTRACT. Introduction Research with bilingual speakers has demonstrated that mandatory or cued switching between languages is effortful (e.g., Meuter & Allport, 1999). More recently, studies found that voluntary switching may eliminate mixing costs, while switch costs remain (e.g., De Bruin et al., 2018; Gollan & Ferreira, 2009). These studies also showed that language proficiency guides voluntary switching behavior, as more balanced bilinguals switched more frequently, and ease of lexical access predicted language choice. The present study aimed to further explore the differences between free and cued language switching. In particular, we directly investigated the influence of second language proficiency and more general switching abilities on free and cued language switching.

Methods Native speakers of Dutch (N = 40) with English as a second language took part in an online experiment at home. Participants were recruited to be part of low or high proficiency groups. All participants completed a questionnaire about the relevant demographic and language variables and performed the LexTALE as a measure of English proficiency (Lemhöfer & Broersma, 2012). The experiment consisted of four tasks: (1) Single language: picture naming in Dutch or English blocks. (2) Free switching: picture naming in the language that first came to mind. (3) Cued switching: picture naming in English or Dutch, depending on a cue. (4) Within-language switching: picture naming in Dutch with an adjective constituent indicating color (blue or red) or size (big or small) of the picture, depending on a cue. Linear mixed effects models were used to analyze the reaction times and switching data.

Results & Conclusions The results of our initial analyses showed that participants experience switch costs in both cued (β = 0.06, SE = 0.01, p < 0.001) and free switching (β = 0.03, SE = 0.01, p = 0.015). However, language had opposite effects on switch costs in these tasks (β = 0.05, SE = 0.02, p = 0.022): Switching into English was more effortful in the cued condition, while switching into Dutch was more effortful in the free condition. Additionally, a significant four-way interaction between task, sequence, language and proficiency (β = -0.005, SE = 0.002, p = 0.009), indicates that English proficiency influenced task performance to different degrees (Figure 1). Overall, participants became faster with increasing English proficiency, except for the repeated Dutch trials in the free condition. This suggests that higher proficiency involves better control over languages in contexts where languages compete. Self-rated proficiency, but not LexTALE scores, significantly influenced switching behavior in the free task, as participants who rated their English proficiency as higher were more likely to switch (β = 0.03, SE = 0.01, p = 0.002). Finally, we failed to find significant effects of the switch costs in the within-language switching task on performance in the cued and free conditions. To conclude, higher proficiency implies better control over languages, which is reflected in differences between voluntary and mandatory switching between languages.

[Figure 1]

References De Bruin, A., Samuel, A. G., & Duñabeitia, J. A. (2018). Voluntary language switching: When and why do bilinguals switch between their languages? Journal of Memory and Language, 103, 28–43. https://doi.org/10.1016/j.jml.2018.07.005 Gollan, T. H., & Ferreira, V. S. (2009). Should I stay or should I switch? A cost–benefit analysis of voluntary language switching in young and aging bilinguals. Journal of Experimental Psychology. Learning, Memory, and Cognition, 35(3), 640–665. https://doi.org/10.1037/a0014981 Lemhöfer, K., & Broersma, M. (2012). Introducing LexTALE: A quick and valid lexical test for advanced learners of English. Behavior Research Methods, 44(2), 325–343. https://doi.org/10.3758/s13428-011-0146-0 Meuter, R. F. I., & Allport, A. (1999). Bilingual language switching in naming: Asymmetrical costs of language selection. Journal of Memory and Language, 40(1), 25–40. https://doi.org/10.1006/jmla.1998.2602

ABSTRACT. Cross-linguistic form similarity of translation equivalents (cognates) results in these words being processed faster and more accurately compared to non-cognates, known as cognate (facilitation) effects.1 Degree of orthographic and/or phonological form overlap has been shown to modulate the cognate facilitation effect.2,3 However, Degree of morpho-syntactic overlap has not previously been investigated. We investigated cognate effects from (morpho)syntactic overlap (and phonological form overlap) in cognate verbs. Spanish-English cognate verbs were categorized by degree of (mis)match in verb argument structure and phonological form overlap, which resulted in a fine-tuned 3-level cline of morphosyntactic and phonological similarities of cognate verbs:

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

20 bilingual participants heard the same cognate verbs in 2 tasks: a language identification task (LID) wherein participants had to mark words heard as English or Spanish,4 and a self-paced listening task (SPL; based on self-paced reading paradigms) wherein participants pressed a button to hear sentences presented in 4-5 segments, with Spanish and English in different blocks; target (non)cognate verbs were presented as the only word in middle segments. RTs in both tasks were collected from button presses to hear following items.

Across tasks, RM ANOVAs revealed that cognates were processed faster than non-cognates (both, p<.05). Additional RM ANOVAs within each task discriminated 4 lexical categories (3 types of cognates plus non-cognates). In the LID task, differences between true and lemma cognates (p=.032), and true cognates and non-cognates (p=.004) were significant. Thus, LID results indicated that only degree of phonological form overlap mediated cognate effects. In the SPL task, three-way differences between cognate types were significant, but not equivalent within each language (p<.001). In English differences between true and lemma cognates (p<.001), lexemic and lemma cognates (p<.001), and non-cognates and true (p<.001) as well as lexemic cognates (p<.001) were significant; in Spanish, differences between true and lexemic cognates (p<.001) were significant. Thus, SPL results indicated that both degree of phonological form overlap and (morpho)syntactic overlap mediated cognate effects.

These results across tasks follow prior research finding effects from phonological overlap on cognate effects. Further, these results suggest that (morpho)syntactic overlap may only modulate cognate effects in sentential contexts. These results may explain why prior studies using verb cognates have found less robust effects compared to noun cognates, or have found no effect of cognate status in verbs.5–9 In addition, these results may indicate that context (or lack of one) differentially activates the lemma stratum. Though access to the lemma stratum is available, these results suggest that listeners do not always activate a word’s grammatical content when it is irrelevant.

ABSTRACT. The ‘price’ of language switching depends on who you ask: On the one hand there are many carefully controlled laboratory studies revealing the cost of language switching, e.g., bilingual participants name pictures more slowly when asked to switch from L2 to L1 compared to naming just in L2 (Christoffels, Firk, & Schiller, 2007). On the other hand, there is evidence demonstrating frequent instances of spontaneous language switching (e.g., codeswitching) in many bi-/multilingual communities, suggesting that for these individuals, the costs appear to be acceptable (enough) to switch languages anyway. These apparent contradictory findings could be partially explained by a difference in ecological validity, since language switching costs have been shown to be reduced under more naturalistic circumstances (e.g., Blanco-Elorrieta & Pylkkänen, 2016). This study aimed to examine language switching in a more ecologically valid task in which 29 Dutch-English bilingual participants listened to naturally spoken sentences that were either not switched (unilingual) or contained intrasentential codeswitches. Moreover, to examine the possibility of ‘acceptable-enough costs’ we contrast language switching with within-language lexical switching (Moreno, Federmeier, & Kutas, 2002), by presenting participants with Expected and Unexpected sentence endings. Specifically, we compare processing of lexically-unexpected Dutch words (e.g., “kapper” (‘hairdresser’)) to processing of lexically-expected English words (e.g. “glasses”, Table 1). If language switching costs are comparable to within-language lexical switch costs, then processing of a lexically-expected codeswitched word should be as difficult as processing a non-switched unexpected word, the latter listeners typically successfully do (e.g., Frade, Pinheiro, Santi, & Raposo, 2021). Our findings indicate that codeswitched sentences elicited an increased N400 response (F(1,26) = 8.79, p = .002) and a Late Positive Component (F(1,26) = 7.96, p = .008) relative to non-switched sentences. Therefore, language switching appears to incur a cost when directly comparing codeswitched to non-switched sentences. However, when contrasting codeswitched sentences (ending expectedly) with within-language lexical switches (non-switched sentences ending unexpectedly), we found no difference in terms of N400 amplitude (t(1, 28) = -.97, p = .342), but the conditions still differed significantly in terms of LPC (t(1, 28) = 4.15, p = <.001). Thus, processing lexically expected codeswitched words is similar to processing within-language lexically unexpected words regarding the N400, but not the LPC. These results extend our knowledge on language switching costs and can potentially unite the previously reported apparent contradictory findings on the price of language switching. Although our results indicate that it is more costly to process codeswitched sentences compared to non-switched sentences, these costs are comparable to those found when processing a within-language lexical switch as indexed by the N400. Tentatively interpreting these components: If the N400 indicates semantic integration difficulty, then processing the English “dentist” seems to be as challenging for our participants as processing the Dutch “kapper” (‘hairdresser’; Table 1). However, albeit equally challenging in terms of semantic integration, the language switch is still registered as expressed by the difference in LPC revealing some processing cost for the language switch, possibly indicating sentence-level reanalysis for codeswitched, but not for non-switched sentence endings.

ABSTRACT. Research shows that L2 prediction is less stable than L1 prediction (Ito & Pickering, 2021). Yet, the sources of variability in L2 prediction are uncertain. L2 learners rely more strongly on lexical cues versus morphological cues (Armstrong, Bulkes, & Tanner, 2018). However, it is unclear whether morphological cues hinder L2 prediction. Prosody is a cue for the prediction of suffixes. Swedish tones trigger prediction of tense and number morphology in native speakers and in some L2 learners (Roll, 2015; Schremm, Söderström, Horne, & Roll, 2016). Similarly, Spanish lexical stress triggers prediction of tense morphology in native speakers and to a lesser extent in L2 learners (Sagarra & Casillas, 2018). Interestingly, lexical stress also triggers prediction of non-morphological word endings in both native speakers and L2 learners similarly (Lozano-Argüelles, Sagarra, & Casillas, 2021). We investigate whether morphology or size of semantic neighborhood hinders prediction in interpreter L2 learners of Spanish. We chose professional interpreters because of their high knowledge and use of the L2, and their extensive use of prediction both in comprehension and production (Zhao, Chen, & Cai, 2022).

Two groups of participants (Spanish monolinguals, L2 adult learners of Spanish with extensive interpreting experience) completed an L2 proficiency test, language history questionnaire, a working memory task, and three eye-tracking experiments. The proficiency test was an abbreviated version of the DELE (focused on grammar and vocabulary) and the language history questionnaire included items related to age and context of L2 acquisition, L2 use, and interpreting experience. The working memory test was the letter-number sequencing test. In all eye-tracking experiments, participants listened to a sentence in Spanish while seeing two words on the screen and were asked to select which of the two words appeared in the sentence. Both groups were comparable in working memory span. Experiment 1 investigated whether lexical stress triggered prediction of verbal morphology (LAva-laVÓ, ‘(s)he washes-washed’). Experiment 2 researched whether lexical triggered prediction of non-morphological noun endings in words semantically unrelated (BAla-baLÓN, ‘bullet-ball’). Experiment 3 focused on the use of lexical stress to predict morphological noun endings in words semantically related (TApa-taPÓN, ‘lid-cap’). Target and distractor words of all three experiments had an identical initial syllable that could only be distinguished by lexical stress. The critical variables for all three experiments were lexical stress (stressed-unstressed), syllabic structure (CV-CVC), and group (monolinguals-interpreters).

We conducted three Growth Curve Analyses. For experiments 1 and 3 (morphological endings), results showed that monolinguals made prediction under all four conditions, while interpreters could not predict when the initial syllable was CV-stressed. However, in experiment 2 (non-morphological endings), monolinguals and interpreters predicted under all four conditions alike. These findings indicate that weighing morphological cues is particularly challenging for L2 learners, even when they have extensive L2 knowledge, use, and practice with L2 prediction. Our results are in line with research showing that processing morphology is cognitively taxing for L2 learners (Sagarra & Ellis, 2013).

ABSTRACT. The aim of the study is to explore the function of the mental lexicon in bilingual developing grammars. Specifically, we examine the effect of sentential context on lexical ambiguity resolution in Greek speaking monolingual children and Albanian-Greek speaking bilingual children. Context and word frequency are factors that can contribute diversely to lexical processing depending on the child’s age (Booth et al., 2006; Khanna & Boland, 2010), however, the role of them has not been examined in L1 and L2 Greek. To this aim, we systematically assessed sentence context effects in homonym meaning activation in 40 monolingual and bilingual children aged 10 to 12 years old (M: 10;8, SD: 0;5) using a cross-modal priming paradigm. Primes were sentences biasing towards the first, second, or neither meaning of a sentence-final homonym and visual targets were related to either the first or the second meaning of the homonym. The homonyms, control items and visual targets were selected with the use of three preliminary tests; one to identify a list of 30 homonyms with a high frequency first meaning (M: 74,8%) and a low frequency second meaning (M: 38,4%), the second test to select the most frequently associated target words to each meaning and the third task to verify that homonyms and control words were of similar frequency. We additionally administered (a) the Raven's Progressive Matrices Test as a measure of basic cognitive functioning for non-verbal intelligence, (b) the Digit Span Backwards Recall of the Wechsler Adult Intelligence Scale (WAIS) as a measure of verbal working memory, and (c) the Nonverbal Stroop Card Sorting Test as an inhibitory control measure. Verbal working memory and inhibitory control were employed to examine whether the use of sentential context for lexical ambiguity resolution relates to bilingualism and/or cognitive processing capacity. For the linguistic profiling of our bilingual participants, we also administered a background questionnaire adapted from Mattheoudakis, et al. (2016) to quantify language input indexes for the L1(s) and L2 along with a standardized test for productive vocabulary skills in Greek, namely the Crichton Vocabulary Scales. The data analysis showed (a) processing differences due to bilingualism, (b) children’s processing times affected by cognitive skills, and (c) visual word recognition skills of bilinguals related to their linguistic profiling.

References Booth, J. R., Harasaki, Y., & Burman, D. D. (2006). Development of lexical and sentence level context effects for dominant and subordinate word meanings of homonyms. Journal of Psycholinguistic Research, 35, 531–554. Khanna, M. M., & Boland, J. E. (2010). Children's use of language context in lexical ambiguity resolution. The Quarterly Journal of Experimental Psychology, 63(1), 160-193. Mattheoudakis M., Chatzidaki A., Maligkoudi C. & Agathopoulou E. (2016). Family and school language input: Their role in bilingual children’s vocabulary development. Journal of Applied Linguistics, 31, 49-69.

ABSTRACT. Language can impact event memory (e.g., Loftus & Palmer, 1974) through the mental models of events, where lexical-semantics specifies the content, and aspect (perfective: washed, imperfective: was washing), the viewpoint (Langacker, 2008) and level of event construal during memory retrieval (Trope & Liberman, 2010). Lower construal level corresponds to greater focus on event detail, consistent with imperfective. Higher construal level presents an event holistically, consistent with perfective. Russian has a highly grammaticalized aspectual system with obligatory aspectual marking, whereas English does not. We asked whether obligatory marking of aspectual oppositions impacts memory for event endings in L1 and L2 English and Russian. During encoding, participants (51 L1 English, 48 L1 Russian, 28 L2 English, 26 L2 Russian) watched 48 event videos (24 completed, 24 incomplete) and read descriptions (half perfective, half imperfective) in English (L1 and L2 English participants) or Russian (L1 and L2 Russian participants). L2 Russian and L2 English participants were native speakers of English and Russian, respectively. To control for potential effects of event semantics, half of the event videos were resultative (e.g., peeling an apple) and half were non-resultative (e.g., stirring soup). After watching all videos, there was a surprise memory test. Participants viewed still frames from the previously watched videos and judged whether they corresponded to the last frames of the videos they saw. Signal Detection Theory measure d’ was calculated. Sum-coded data was analysed using linear effects modelling with Event Stage (complete, incomplete), Aspect (perfective, imperfective), and Event Type (resultative, non-resultative) as within-subjects predictors and Task Language (English, Russian) and Language Background (L1, L2) as between-subjects predictors. In English native speakers, aspect affected memory, but only when the task was performed in L1. In the L1 English group, imperfective descriptions improved memory for ceased events (M = 0.25, SD = 0.77), and made memory worse for in-progress events (M = –0.03, SD = 0.79). This suggested more focus on detail promoted by imperfective resulting in better encoding of ceased event videos, consistent with a level of construal effect. It also distorted memory for in-progress events by inference: in-progress events were more frequently misremembered as completed. In Russian native speakers, memory was affected by Event Stage, specifically, completed events were remembered best (completed: M = 0.33, SD = 0.83; incomplete: M = –0.06, SD = 0.73) regardless of aspect or task language. Event Stage did not affect memory in L2 Russian speakers. Better memory for completed events in native Russian speakers is consistent with strong focus on completion of Russian perfective. To conclude, aspect seems to affect memory through different mechanisms in English and Russian: through the level of event construal in English and through habitual event attention patterns in Russian. Memory in L2 was not affected by task language. L2 participants’ performance was consistent with their respective L1s, suggesting L1-entrenched preferences in event processing (e.g., Filipović, 2018; Flecken et al., 2015). Current effort is focused on collecting more bilingual data.

ABSTRACT. While late second language (L2) acquisition is largely explicit, there is some evidence that adults are able to acquire the vocabulary and the grammar of novel languages under incidental learning conditions (Rebuschat et al., 2021; Ruiz et al., 2018). However, it remains unclear which aspects of language can be/are learned implicitly and under what conditions. Here, we revisit the question of whether adults can learn grammar incidentally and investigate whether word order and morphology are susceptible to implicit learning to the same degree. Additionally, we ask what cognitive abilities support early L2 learning and whether their roles change as a function of time.

Forty-one English monolinguals, aged 18-35, were exposed to Kepidalo (1), an artificial language that had case marking and variable word order (SOV and OSV). The study included five online sessions. First, participants were explicitly trained on the nouns of the language. This was followed by a two-alternative forced-choice task (2AFCT) consisting of two blocks. In block 1 (270 trials), participants received vocabulary training, while being implicitly exposed to grammar. Two videos, each showing two aliens performing an action, were presented whilst a sentence was played (Figure 1). The videos differed in 1) one of the aliens, or 2) the color of one alien or, 3) the action performed. Participants had to select the video that matched the sentence and were given feedback on their responses. Block 2 served as a grammatical comprehension test. Here, the two videos differed in that the agent/patient roles were reversed. Participants completed 90 trials without any feedback. The 2AFCT was repeated in the first four sessions. In session 5, a grammaticality judgment task (GJT) including both word order and case marking violations assessed participants' grammatical knowledge. During the study, participants completed a series of cognitive abilities tasks testing explicit learning (EL), implicit statistical learning (ISL), sustained attention, grammatical sensitivity, and speed of automatization.

We found that, although vocabulary learning increased significantly across sessions (Figure 2), grammatical comprehension showed little improvement over time, and improvement was limited to SOV sentences only (Figure 3). Moreover, in the GJT, participants performed better on grammatical than ungrammatical sentences and were better at detecting word order violations than case marking errors. Vocabulary and grammar learning as well as accuracy in the GJT were found to be modulated by EL. Sustained attention and accuracy in the initial noun training additionally predicted vocabulary learning, while, in the GJT, the positive EL effect appeared to be stronger for participants with higher sustained attention scores. Interestingly, for both grammar and vocabulary learning, the ISL effects were most pronounced early on, whereas the EL effects increased with time. However, the interaction between session and ISL was statistically significant only for vocabulary learning. Taken together, our results underscore the role of EL and ISL in early L2 acquisition. Furthermore, learners’ difficulty with case marking confirms the presence of a threshold in incidental L2 acquisition which is tightly linked to the first language experience (Ellis & Wulff, 2020).

ABSTRACT. Grammatical gender (hereafter gender) – especially in systems (like Romance languages) that typically have a relatively transparent system – is acquired early by monolingual children (e.g., Kupisch, Müller & Cantone, 2002). Yet gender shows variability in (some) heritage speaker bilinguals (HSs). In a HS context, it is vulnerable for low proficiency speakers generally and especially when the majority language lacks gender (e.g., Polinsky, 2008). Conversely, gender seems to be on target when acquired in HS individuals with high proficiency, especially when the majority language has gender (e.g., Bianchi, 2013). Herein, we examined sources of potential morphological variability in Italian HSs living in Germany (a language pairing where both have gender, albeit with important differences), with a focus on morphological markedness (masculine as the default) and task type (explicit vs. implicit knowledge).

Fifty-four adult Italian HSs living in Germany and 40 native speakers of Italian living in Italy participated in two experiments. Experiment 1 examined gender agreement in an offline grammaticality judgement task (GJT) involving sentences with grammatical and ungrammatical noun-adjective sequences with masculine (unmarked) and feminine (marked) nouns in the singular (unmarked) and plural (marked). Gender violations were realised on the adjective. This probed for HSs’ potential overreliance on unmarked forms or “defaults” (masculine/singular). Experiment 2 examined the same conditions in an online self-paced reading (SPR) task (tapping more implicit knowledge) to address how HSs process gender agreement violations. In addition, to address possible relationships between gender agreement, language exposure/use and language proficiency, all participants filled in a detailed language and social background questionnaire adapted from Lloyd-Smith, Einfeldt and Kupisch (2019) as well as the Language and Social Background Questionnaire (LSBQ) (Anderson et al., 2018) and to gauge proficiency, participants completed the DIALANG vocabulary test (Alderson, 2005).

In both linguistic tasks, native-dominant Italian controls performed at ceiling and showed faster reading times (RTs) compared to HSs (see Figures 1 and 2). In the GJT, HSs showed high accuracy and were more sensitive in detecting violations realised on marked adjectives; they showed an effect of markedness for gender (more accurate with masculine vs. feminine) and for number (more accurate with singular vs. plural). In the SPR, both groups showed sensitivity to violations realised on marked vs. unmarked adjectives but only in the masculine (longer RTs in the critical region ‘antica’ in Figure 2). Moreover, proficiency in Italian was significant for HSs in the GJT, but not in the SPR.

Notwithstanding some differences, HSs’ performance on both offline and online comprehension was qualitatively similar to that of the Italian controls, suggesting that these HSs of Italian process gender in a native-like manner, albeit modulated differentially by markedness and proficiency at the individual level.

ABSTRACT. I. Research Questions Children and adults tend to code-switch (CS) within specific conversational contexts which allow them to use all the linguistic features available in order to communicate. This is a sophisticated skill that requires the person to be highly proficient in both languages in order to achieve intra-sentential code-switching.1 This skill is not well-understood in children,2, especially among those with disabilities.3 Recent studies4-6 of code-switching structure in children have focused on monolingual contexts. Therefore, our goal was to describe the structures that children use within a CS context, the child characteristics associated with these patterns, and if they exhibit syntactic priming in following the structure of their adult partner’s CS. The following questions guided the research: 1. What child characteristics are associated with the types of intra-sentential CS that children produce in a bilingual context? 2. To what extent do children follow the type of CS used by their conversation partner?

II. Methodology To elicit the production of sentences in a CS context, we used a scripted confederate dialogue paradigm7 in which children took turns with a video partner who used English-Spanish CS to describe images and to prompt the child (e.g., Cuéntame de tu dibujo, and I’ll find it). The conversation partner produced alternational and insertional CS while describing pictures (see Table 1 for examples) in 20 pseudo-randomized trials. Children completed a similar task in each language with monolingual partners. In addition, children completed the English and Spanish morphosyntax subtests of the Bilingual English Spanish Assessment (BESA) as a measure of their language skills. Furthermore, parents described their child’s daily exposure to each language and to CS. Data collection is in progress over Zoom with a target sample of 30 children, ages 4;0-7;2, including both neurotypical children and those with developmental language disorder (DLD) or autism (ASD). Inclusion criteria include Spanish exposure from birth and the ability to produce simple phrases in both Spanish and English.

III. Analysis and Results Children’s picture descriptions will be coded for the presence and type of CS following Muysken’s typology (i.e., insertions, alternations, congruent lexicalization). Preliminary data from six children (four neurotypical, one with ASD, one with DLD) reveal a variety of response patterns. Some children responded entirely in Spanish, while others engaged in code-switching across various typology categories (see Table 2 for definitions and examples). Further analyses with the full sample will examine child characteristics (e.g., language exposure, proficiency, relative dominance, diagnosis) as predictors of children’s use of alternations, insertions, and congruent lexicalization. In addition, the typology of children’s code-switched productions will be compared to the structure of the immediately preceding picture description from their partner.

IV. Implications This study provides an opportunity to examine Muysken’s typology in children, including those with disabilities. The findings can inform clinicians about sources of variability in the structure of children’s code-switching in a context where the use of both languages is encouraged. This information may assist in interpreting assessment results and planning bilingual therapy.

Table 1 Sample Picture Descriptions from Code-switching Adult Partner

Type Example Insertion La niña está escondiendo el toy detrás de la silla. Alternation El señor está looking at the butterfly on the tree.

Table 2 Sample Child Utterances by Type of Code-switching

Type and Definition

Insertion: Using an L2 content word in a sentence mostly composed of L1 words. (e.g., Mi dibujo es un niño que tiene milk.)

Alternation: Alternation between longer L1 and L2 segments. (e.g., La niña está poniendo (la) the rabbit in the box.)

Congruent lexicalization: Using a variety of components from L1 and L2 within the sentence. (e.g., Mi picture es una niña limpiando the window.)

Note: Definitions are based on the discussion of Muysken’s typology in Treffers-Daller (2022).

References 1. Kroff, J. V., & Fernandez-Duque, M. (2017). Experimentally inducing Spanish-English code-switching: A new conversation paradigm. In K. Bellamy, M. W. Child, P. González, A. Muntendam, & M. C. P. Couto (Eds.), Multidisciplinary Approaches to Bilingualism in the Hispanic and Lusophone World (pp. 211–233). John Benjamins.

2. Treffers-Daller, J. (2022). Code-switching in bilinguals and trilinguals. To appear in Stavans, E. & Jessner, U. (Eds). The Cambridge Handbook of Child Multilingualism. Cambridge University Press.

3. Yu, B. (2016). Code-Switching as a Communicative Resource Within Routine, Bilingual Family Interactions for a Child on the Autism Spectrum. Perspectives of the ASHA Special Interest Groups, 1(Part 1), 17–28.

4. Gutierrez-Clellen, V. F., Simon-Cereijido, G., & Leone, A. E. (2009). Code-switching in bilingual children with specific language impairment. International Journal of Bilingualism, 13(1), 91–109.

5. Kapantzoglou, M., Brown, J. E., Cycyk, L. M., & Fergadiotis, G. (2021). Code-switching and language proficiency in bilingual children with and without Developmental Language Disorder. Journal of Speech, Language, and Hearing Research, 64(5), 1605–1620.

6. Smolak, E., de Anda, S., Enriquez, B., Poulin-Dubois, D., & Friend, M. (2019). Code-switching in young bilingual toddlers: A longitudinal, cross-language investigation. Bilingualism.

7. Kootstra, G. J., van Hell, J. G., & Dijkstra, T. (2010). Syntactic alignment and shared word order in code-switched sentence production: Evidence from bilingual monologue and dialogue. Journal of Memory and Language, 63(2), 210–231.

ABSTRACT. Studies that are recently conducted on bilingual gender processing show that gender is processed predictively, and that the prediction efficiency varies across bilingual groups (see Hopp, 2016; Hopp and Lemmerth, 2018). Late bilinguals do not usually acquire the prediction skills, contrary to simultaneous bilinguals that are as efficient as monolinguals in terms of using predictive gender processing. However, almost all previous studies that painstakingly investigated L2 predictive gender processing focused on languages that have salient prenominal grammatical gender materials (gender cues). In this regard, little attention has been given to languages that have postnominal grammatical gender materials. The present study aims to fill this lucuma. It investigates late L1Oromo (postnominal) learners of Amharic (prenominal) and addresses three concerns: (1) determines whether late L2 learners acquire predictive gender processing skills; (2) illustrates to what extent L1 that has postnominal grammatical gender material facilitates late L2 predictive gender processing, and (3) shows to what extent dialect variability in L1 grammatical gender influences L2 predictive gender processing. The study investigated 100 L1 Oromo learners of Amharic, along with L1 Amharic monolingual control groups. The study employed visual word paradigm in a webcam eye-tracking experiment. Provisional results obtained from the eye-tracking experiment indicate that late L2 learners use prediction while processing L2 grammatical gender. However, they manifest endured latency of eye-movement, compared to the monolingual counterparts. By condition comparison across various grammatical conditions has also shown a significant facilitative effect of L1 grammatical gender on L2 predictive grammatical gender processing. Furthermore, grammatical gender features that are consistently used in dialects of L1 have shown a stronger facilitative influence, compared to grammatical gender features that are less consistently used. The reported results have significant contributions to the current knowledge of L2 gender processing. They also contribute to the knowledge of crosslinguistic influence in L2 grammatical gender processing. Finally, the study has an ecological importance as it provides empirical data from understudied languages.

ABSTRACT. Language outcomes depend on biological, cognitive, and environmental factors. How do these factors impact development in bilingual language acquisition? This is especially relevant for sequential bilingual children whose proficiency varies widely due to differences in these factors (Granena, 2014). While factors like age, memory and linguistic input can all impact language proficiency in sequential bilinguals, results are mixed for particular language features (e.g., Armon-Lotem et al., 2011; Chondrogianni & Marinis, 2011; De Cat, 2020; Paradis, 2011; Paradis et al., 2017; Rothman et al., 2018; Unsworth, 2016). Our research questions are: (1) How do biological, cognitive, and environmental factors affect proficiency in sequential bilinguals? (2) Do the effects vary across language features?

As biological and cognitive factors, we investigated the effects of chronological age and short-term memory, while environmental factors included length of L2 exposure, richness of the L2 environment (i.e., play and social activities in the L2) and socioeconomic status (Table 1). Language features included vocabulary, morphology (third person singular and past tense) and syntax (active and passive voice, and subject and object relative clause sentences) (Table 2).

Short-term memory was tested with a non-word repetition task (Wagner et al., 1999), while environmental factors were measured with a parent questionnaire which was adapted from the Alberta Language Environment Questionnaire (Paradis, 2011) and translated to Arabic. Language features were assessed using standardised tests (vocabulary and morphology) and an experimental task (syntax) (Table 2).

Participants were 40 Arabic-speaking children acquiring English aged 5;7-12;2 (M=8;4) with length of L2 exposure ranging from 0;7 to 10;6 (M=4;1).

Generalised linear mixed effects modelling revealed the following: while vocabulary and morphology were both predicted by longer L2 exposure (both p’s <.001, Figures 1-2), accuracy on each of the syntactic structures was predicted by richness of the L2 environment (p=<.001, Figure 3). However, for the more complex syntactic structures (passive voice and object relative clauses), this effect of richness of the L2 environment was driven by older children, with a significant interaction observed for the complex structures between richness of the L2 environment and chronological age (p=<.001, Figure 4): while higher accuracy was observed with a richer L2 environment for older children, L2 richness did not predict accuracy for younger children. Short-term memory and socioeconomic status did not predict any language features.

These results suggest that in child sequential bilingualism, lexical and morphological development depend on overall language experience. Input also impacts syntax; however, it is a specific type of input namely, L2 exposure through activity- or play-based situations, which drives proficiency of these structures. Moreover, for complex syntax, the effect of this richer input is observed only in older children. This suggests that for younger sequential bilinguals, complex syntax is less input-dependent. Results also suggest that short-term memory and socioeconomic status do not predict language measures when other factors, including fine-grained input, are accounted for.

ABSTRACT. Cognitive accounts of individual variation have long been of interest to SLA researchers. With the current advances in neuroimaging and psycholinguistic techniques, we have been able to address individual variation with increasing precision. However, a place remains for offline behavioural tests that tap into the same phenomena, but which are freely available to researchers. One such cognitive area investigated since the early prognostics tests (Henmon, 1934) is language learning aptitude. Recent studies have investigated how the LLAMA aptitude tests (Meara, 2005) relate to cortical morphometry and white matter microstructure (Novén et al., 2021). While language learning aptitude remains undefined at a theoretical level (cf. Grigorenko et al., 2015; Robinson, 2001; Skehan, 2016), researchers have argued that the aptitude tests are really tests of working memory (Wen, 2016) or long-term memory (Buffington & Morgan-Short, 2019).

The issue is that much of this research relies on the freely available LLAMA tests (Meara, 2005), which have been subject to a number of criticisms, including their reliability (Bokander & Bylund, 2020) and construct validity. These criticisms relate to the original LLAMA tests but in (2019), Meara & Rogers released a revised version (v.3) for three of the four tests (LLAMA B (vocabulary), LLAMA E (sound-symbol correspondence) and LLAMA F (grammatical inferencing). LLAMA D (sound recognition) was not substantively changed. This paper seeks to empirically investigate:

1. Are the LLAMA v.3 tests (more) reliable? 2. Do the LLAMA v.3 tests measure the same thing as common WM and LTM tests?

The four LLAMA tests were re-programmed into gorilla.sc for online administration. The test battery also included a Flanker test and auditory digits forwards and backwards tests (various WM components), Tower of Hanoi (procedural memory) and CVMT (declarative memory). 210 participants took the 8 tasks as well as a background questionnaire (F=145, Mean age = 23, (range 18-84)).

Data analysis is ongoing. Preliminary analysis for RQ1 shows that the Cronbach’s alpha scores for the three revised tests are all over .80 suggesting they are an improvement on the previous version in terms of internal consistency. Table 1 compares the Cronbach alpha scores from this study with the new LLAMA tests and those found in Bokander & Bylund’s (2020, table 4) paper with the original tests. The LLAMA D scores are comparable (n=136 due to a computer error).

In terms of RQ2: a principal components analysis was carried out to establish if the Flanker, CVMT and LLAMA B, E & F tests were measuring the same construct. A chi-squared test showed the model was significant (p<.001) and two components were produced (see Table 2).

The results suggest that the declarative memory measure loads on the same component as the three LLAMA tests whereas the executive function test measures something else contra Wen (2016). Additional analysis will include the remaining tests and the reaction time measures collected.

ABSTRACT. A developmental approach to syntactic representations requires an understanding of how bilingual children process their two languages. For adult bilingual speakers, significant cross-linguistic priming effects provide evidence arguing for shared syntactic representations1,2. Evidence for cross-linguistic structural priming in bilingual children is still sparse, but studies have shown that children can also be primed across languages3,4,5. However, not all studies have investigated bi-directional priming, and, so far, the syntactic constructions used (passives in Spanish and English, ditransitive constructions in Norwegian and English) have been syntactically identical across languages. In two cross-linguistic structural priming studies, we set out to investigate the following questions: 1. Do priming effects vary as a function of structural overlap? 2. Are there effects of directionality on priming? 3. Does language proficiency predict priming? In Study 1 we primed attributive constructions (Adj+N (AN) vs. Relative Clauses (RC)); Study 2 targeted possessive constructions (Possessor First (Poss1) vs. Possessor Second (Poss2); Figure 1). The syntactic structures in Study 1 are identical across Polish and English, while in Study 2 they are not: both languages have Poss1 and Poss2 constructions, but while English marks possession using the clitic genitive marker “‘s” in Poss1 constructions, and the case marking preposition “of” in Poss2 constructions, Polish always marks the possessor for genitive case inflectionally and the order is simply reversed. Each study included 48 experimental items, with no lexical or semantic overlap between primes and targets. 48 Polish-English bilingual children living in the UK (24 girls; Mage 92mth, SD=16.65) took part in both studies. In a game of ‘Snap!’, the experimenter provided a prime in one language and the child described the target in the other (e.g. Polish prime>English target); across two sessions, children used both languages. We also administered expressive vocabulary and sentence repetition tasks in each language. In Study 1, children used very few RCs in either language – we found no significant effect of priming (Z=1.02, p>.05) in either direction (Z=.05, p>.05; Figure 2). Weak syntactic representations for RCs may have affected the parsing of the prime and/or the computation of the target. In Study 2, children produced more Poss2 in Polish than English (Z=7.05, p<.001) but there was significant priming (Z=5.05, p<.001) which did not differ by direction (Z=-0.71, p>.05; see Figure 3). This effect of priming suggests that what is being primed is not exactly a syntactic structure, but the word order of semantic role of possessor. Language proficiency did not significantly contribute to either model. Thus, where constituent structure was not the same, priming of possessor-possessee word order nonetheless occurred where the two languages allowed the same alternation between the order of roles. However, we found no evidence that, in this sample, representations are shared when constituent structure was the same, i.e. Study 1. Children rarely produced non-standard syntactic forms suggesting that syntactic representations were not transferred incorrectly. Follow-on studies are investigating whether Polish-English bilingual speakers have abstract representations for these constructions and whether they do ultimately acquire shared representations.

ABSTRACT. Acquiring non-canonical word orders, e.g. object questions and relative clauses, in an L2 implicates successful reanalysis abilities. Previous research on relatively highly proficient adult L2 learners reports that reanalysis abilities in an L2 are first and foremost affected by general L2 experience and proficiency as well as L1-L2 overlap in the morphosyntactic marking of non-canonical word orders, e.g. case marking (e.g. Hopp; 2010), as well as cognitive capacity, i.e. working memory (Havik et al., 2009). Evidence from L1 and L2 adults indicates that reanalysis in temporarily ambiguous sentences is affected by executive function, in particular inhibitory control (e.g. Hsu et al., 2021; Hussey et al., 2017). These findings suggest that reanalysis ability is affected by both linguistic and cognitive variables. However, little is known about the interplay of these factors and how they affect reanalysis among lower-proficiency learners at earlier stages in L2 development. This study investigates (cross-)linguistic and cognitive predictors of successful reanalysis in the L2 comprehension of object wh-questions and relative clauses among low-intermediate L1 German adolescent learners of L2 English. Specifically, we test the degree to which reanalysis in an L2 is affected by L2 proficiency, reanalysis ability in the L1, cognitive control and cognitive capacity. In a timed picture-selection task (Figure 1), 110 German-speaking L2 learners of English (age range: 12-14) selected target pictures for auditorily presented wh-questions and relative clauses in (1-4) in two separate comparable experiments (with 40 experimental items each), first in L2 English (1a-4a) then in L1 German (1b-4b). We assessed learners‘ cognitive control (Flanker and Stroop), working memory (forward and backward digit span) as well as proficiency in English and German (Cross-Linguistic Lexical Task). Results showed a strong subject preference for wh-questions and relative clauses in comprehension accuracy and RTs (see Table 1). Logistic mixed-effects models for accuracy returned several interactions with word order: Both learners’ L2 proficiency and their relative judgement accuracy on object orders in the L1 (i.e. L1 reanalysis ability) predicted accuracy on object questions and object relative clauses in the L2. For object relative clauses, forward digit span additionally predicted accuracy. For reaction times, LMERS showed faster RTs on subject questions and main effects of forward digit span. As for interactions with word order, L2 proficiency predicted RTs on object questions, and both L2 proficiency and L1 reanalysis ability for German relative clauses predicted RTs on object relative clauses. In all, this study finds that L2 knowledge and L1 comprehension ability of non-canonical orders affect the L2 acquisition of non-canonical orders, with cognitive capacity (forward digit span) affecting reanalysis accuracy only in more complex sentences. There is no evidence that the measures of cognitive control used in the study affect comprehension accuracy or reaction times for non-canonical orders in low-intermediate L2 learners. In all, these findings suggest that linguistic abilities outweigh cognitive abilities in sentence revision in low-intermediate L2 acquisition. Specifically, linguistic reanalysis ability in an L1 appears to transfer to L2 comprehension, suggesting that the parser can cross-linguistically recruit processing routines honed in other languages.

ABSTRACT. Our paper deals with crosslinguistic influence in a specific type of bilingual speech production, namely, simultaneous interpreting (SI), that inherently involves concurrent processing of L1 and L2/L3. SI combines source language comprehension and target language production under extreme time pressure. We focus on the phonetic layer of retour (i.e., from L1 into L2) interpretations. Apart from the regular impact exerted by the phonetic system of the speaker’s native language, retour interpreters are additionaly facing constant interference from the source text delivered in their L1. Retour interpreting is often seen as a necessary evil rather than a quality service in its own right (e.g., Déjéan le Feal, 2003). However, in the EU institutions, it is a widespread practice for some of the newer EU languages, including Polish, to be interpreted into English by native speakers of the source language (e.g., Graves et al. 2022). We therefore present an exploratory qualitative study into retour interpreting in the European Parliament for the language pair Polish/English. We analyse material extracted from the EP-Poland, a large Polish/English parallel corpus (Bartłomiejczyk et al., forthcoming), to assess the phonetic-acoustic properties of speech in English interpretations delivered by native speakers of Polish. We use the methodology from L2 speech research, which relies on identifying segmental and suprasegmental departures from native norms. The speech samples from eight interpreters were inspected aurally and visually by means of spectrogram and waveform. We consider the following questions: 1. Is the pronunciation characterised by strong interference from Polish? Is it fully intelligible? 2. Are there marked differences in that respect among the individual interpreters? 3. How consistent is the interpreters’ performance? The third question is related to our hypothesis that for interpreting into a foreign language, deteriorating pronunciation in some speeches or speech fragments might be linked with increased cognitive load – along other, already established symptoms, detectable either in the interpretation or in the interpreter’s physical reactions (Chen 2017). The detected departures from native norms, common for all the interpreters, emerge mostly from the differences between Polish and English sound systems. Although some of the interpreters may have more native-like accent than others, all the analysed interpretations are phonetically proficient and fully intelligible. What appears to impact the general impression and comprehensibility more negatively than the interpreter’s slightly non-native pronunciation are disfluencies and at times too rapid articulation rates; problems not directly linked with L2 language production.

In three of the interpreters, interesting regularities concerning pronunciation inconsistencies were discovered. The hypothesized pronunciation deterioration related to increased cognitive load, however, was revealed only in one interpreter, from whom, incidentally, we have a relatively large sample. Another regularity, an unexpected one, was revealed in two interpreters: a deterioration was observed for fragments in which they attempted to imitate emotional speech. It may be connected with the impact of an emotional load on L2 speech in that emotional speech leads to resorting to default L1 articulatory habits.

  References

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. Déjéan le Feal, K. (2003). Impact of the international status of the interpreting student’s mother tongues on training. Forum, 1(1), 53–76. Graves, A., Pascual Olaguíbel, M., & Pearson, C. (2022). Conference interpreting in the European Union institutions. In M. Albl-Mikasa, & E. Tiselius (Eds.), The Routledge handbook of conference interpreting (pp. 104–114). Routledge.

ABSTRACT. Background: Recent work shows that heritage speakers (HSs) of Spanish can use grammatical gender cues on articles (el/la) to facilitate word recognition of the subsequent noun during real-time processing in their heritage language [1], in a way that is qualitatively similar to monolingual adult [2,3] and child [2,4,5] speakers of languages with gender. However, given that articles frequently co-occur with nouns in Spanish, the findings for HSs are so far consistent with both a grammatical account of facilitation (wherein participants access an abstract gender feature on the inflected prenominal element and use it to narrow the search for the subsequent noun in the lexicon) and a probabilistic account (wherein facilitation relies on transition probabilities between frequently co-occurring elements) [4,5]. The present eyetracking study tests these accounts by investigating facilitative use of gender in HSs of Polish, which has three genders (Ex.1) but does not have articles, and in which gender cues are located on prenominal adjectives. Participants: The study tested 23 HSs, who reported Polish either as their sole first language or as acquired simultaneously with English and who had lived in Poland 8 years or less (mean=0.8yrs, sd=1.0). The study also tested 18 adult speakers of Polish immigrated to US after 18+ years of living in Poland (mean=25.0yrs, sd=8.7). Participants filled out a background questionnaire and completed an oral production task to determine for which nouns they knew the expected gender. Method: In an eyetracking study in the Visual World Paradigm, participants saw 108 visual displays with two images equidistant from a fixation cross. For each gender there was a match condition and two mismatch conditions, depending on the gender of the distractor (Table 1). Participants were prompted to look at the target noun by the carrier phrase Gdzie jest [COLOR]-{y/a/e} [NOUN]? “Where is [color]-{M/F/N} [NOUN]?” (Table 2). Participants’ eye movements were recorded. Results: Reaction times were calculated as the first fixation on the target item following 200ms after the onset of the gender-marked adjectival suffix. To control for knowledge of gender, only trials for which participants knew the gender for both items in the display were included in the analysis. Linear mixed effects models were fitted to the data and show that both groups were faster to fixate on target items in mismatch than match conditions in each target gender (Fig. 1). For conditions with a masculine or neuter target item, reaction times were also modulated by the gender of the distractor: for example, looks to a masculine target were slowed by neuter distractors but not feminine ones (Fig. 1b & 1c). Discussion: While the modulation effect is beyond the scope of this talk, the main effect of facilitation (a) is consistent with findings for HSs of Spanish [1] and (b) provides evidence for a grammatical account of facilitative use of grammatical gender by HSs. Since adjectives are optional and infrequent, the findings suggest that, when knowledge of gender is controlled for, HSs can access and deploy abstract gender agreement features during real-time processing in a target-line manner.

ABSTRACT. Successful second language (L2) comprehension requires the ability to integrate information in real time from multiple cues in a given context, which may be difficult if the L2 learner’s native language (L1) does not draw on the same morphosyntactic cues, or when under greater cognitive load. For example, when comprehending “Yesterday the girl paint(-ed) sunflowers in the park’”, L1 Mandarin learners of English (whose L1 does not use inflectional morphology) may assimilate temporal information only from temporal adverbials during comprehension, and consequently fail to access relevant grammatical knowledge and detect potential morphosyntactic violations (e.g. inflectional omission). Such failures might be exacerbated in auditory (requiring additional phonological processing) compared to visual comprehension (Johnson, 1992; Murphy, 1997).

Two self-paced comprehension experiments investigated whether L1 Mandarin learners of English could integrate information from temporal adverbials and inflection morphology during real-time sentence processing, and whether this integration was affected by comprehension modality (auditory or visual). Advanced L1 Mandarin learners of English (L1M; AoA > 5 yrs) and L1 English controls (L1E) either listened to (Experiment 1) or read (Experiment 2) English sentences in a self-paced moving-window paradigm (Table 1; Ferreira et al., 1996; Just et al., 1982). We manipulated Temporal Context (T.C., Present Habitual, Past) via different temporal adverbials (e.g. Every day, Last week) and Grammaticality via omitting obligatory inflections (3rd person singular -s (3SG -s), past -ed). If L1 Mandarin learners of English can consistently integrate information from temporal adverbials and morphosyntactic cues, they should show processing difficulties (i.e., longer RTs) where obligatory inflectional morphology was omitted. Additionally, if such integration is affected by comprehension modality, they should show differential sensitivity to inflectional omissions in the auditory and visual modalities.

Our data showed that consistently across auditory and visual modalities, L1M showed non-native-like sensitivity to inflectional omissions (Group x Grammaticality; Exp 1: p<.001; Exp 2: p<.001). Specifically, in Exp 1, L1M were selectively sensitive to inflectional omissions, exhibiting significantly longer RTs for ungrammatical than for grammatical verb segments in the auditory modality (Figure 1a; Grammaticality p=.001; d=0.17; T.C. x Grammaticality p=.002), but less so than L1E (Figure 1b; Grammaticality: p<.001; d=0.55; T.C. x Grammaticality: p=.042). However, Experiment 2 failed to replicate this finding in the visual modality. After accounting for visual word length, L1M did not exhibit significantly longer RTs for ungrammatical than for grammatical verb segments for neither inflections (Figure 1c; p=.034; d=-0.10), but rather showed a spill-over grammaticality effect for 3SG -s on the object segment (Grammaticality: p=.016; d=0.16), unlike L1E who showed an overall grammaticality effect on the verb segment for both inflections (Figure 1d; p<.001; d=0.24).

Our findings suggest that L2 learners can integrate temporal information with L1-absent morphosyntactic cues during comprehension, though grammatical sensitivity can vary across different inflections. Our results also suggest that auditory comprehension is not invariably more difficult for L2 learners; rather, sensitivity to grammatical violations can in some contexts be facilitated by perceptually salient auditory cues.

ABSTRACT. L2 listeners successfully use semantic cues on verbs to predict upcoming direct objects in simple (e.g., [1]) and complex (e.g. [2]) SVO clauses. However, no studies have investigated whether L2 speakers use semantic cues predictively when embedded in a word order that poses difficulties in L2 production (e.g., [3]). In two visual-world experiments, we investigate how word order differences modulate semantic prediction, especially when L1 and L2 word orders differ, to investigate whether and how syntax constrains L2 versus L1 semantic prediction. In Experiment 1, L1 (N=32) and L2 (N=32, L1 English) German speakers listened to subject-first (SVO) and adverb-first (AdvVS) sentences. For subject-initial sentences, English and German share SVO surface order (1), while non-subject initial sentences have V3 order in English (AdvSV), but V2 order in German (2). We tracked participants’ eye-movements to image displays (Fig1a) to measure whether they used semantic information from the lexical verb to anticipate the upcoming noun (constraining-verb; 1a/2a). Target looks in sentences using modal verbs (neutral-verb; 1b/2b), where the lexical verb appears sentence-finally, served as a baseline.

(1a) SimoneSUB füttertV täglich [den Hund]OBJ im Garten. (SVO; constraining-verb) Simone feeds daily the dog in the garden (1b) SimoneSUB sollVmod täglich [den Hund]OBJ im Garten fütternV. (SVO; neutral-verb) Simone should daily the dog in the garden feed “Simone feeds/should feed the dog daily in the garden.” (2a) Im Sommer springtV täglich [der Frosch]SUB ins Wasser. (AdvVS; constraining-verb) In summer jumps daily the frog into the water (2b) Im Sommer wirdVmod täglich [der Frosch]SUB ins Wasser springenV. (AdvVS; neutral-vb) In summer will daily the frog to the water jump “In summer, the frog will jump/jumps into the water daily.”

Divergence point analyses [4] revealed more looks to the target in constraining- versus neutral-verb conditions prior to noun onset for SVO and AdvVS sentences among both L1 and L2 speakers, though effects were delayed for L2 speakers. Divergence points in SVO and AdvVS sentences were similar, with overlapping CIs (Fig2), in both groups; yet increases in looks to the target were attenuated in AdvVS sentences among L2 listeners. Thus, adult L2 speakers engage in semantic prediction across syntactic contexts, including contexts absent in the L1, suggesting that modulations in L2 semantic prediction based on syntax may be quantitative, not qualitative, in nature [5]. To further explore whether anticipatory looks to post-verbal subjects in Experiment 1 were driven primarily by semantic association ('bag-of-arguments', [6]) rather than V2 syntax, Experiment 2, currently underway, includes AdvVS V2-sentences with transitive verbs (e.g., In der Nacht erschießtV plötzlich [der Jäger]SUB [einen Tiger]OBJ im Dschungel. “At night the hunter suddenly shoots a tiger in the jungle”), and visual scenes including both the subject and object (Fig1b). Reliance on semantic associations should result in increased looks to the subject and object following constraining-verb sentences, whereas reliance on syntax should lead to more looks to the subject only. Divergence point analyses will be employed again to probe for potentially different eye-gaze patterns among L1 and L2 German speakers.

ABSTRACT. Being able to see a speaker's mouth benefits native and non-native speech perception (Navarra & Soto-Faraco, 2007; Sumby & Pollack, 1954), and native and non-native listeners increase their attention to a speaker's mouth under adverse listening conditions, such as noise (Drijvers et al., 2019; Vatikiotis-Bateson et al., 1998). Assuming listening to an L2 is more challenging than listening to an L1, Birulés et al. (2020) hypothesized attention to the mouth would be greater among L2 than L1 listeners (H1) and modulated by L2 proficiency (H2). In an eye-tracking study with L1 and L2 English (L1 Spanish/Catalan) listeners, they found support for H1 but, surprisingly, not for H2. The present study examines the reproducibility and generalizability of these findings through an extended replication of Birulés et al. (2020, Exp.2; henceforth 'B').

Method: Participants watched a 60-second video of a native English-speaker talking (Fig.1) while their eye gaze was recorded, followed by nine multiple-choice comprehension questions, closely matching B's original materials and procedure. Unlike B's materials, which were constructed solely for the purpose of the experiment, we adapted a passage from an existing listening comprehension test (Papageorgiou et al., 2012) in an attempt to increase validity. We also included a second video (Fig.2), always shown after the first, featuring a less prototypical (self-identified) native speaker of English (non-white, multilingual) as an exploratory step to probe the generalizability of the hypothesized effects across more diverse talkers. As in B, participants completed the 25-item Cambridge General English Test (www.cambridgeenglish.org/test-your-english/general-english/). To obtain more comprehensive measures of proficiency, participants also completed the LexTALE English test (Lemhöfer & Broersma, 2012). Participants: A power analysis based on available data from B indicated samples of 30-35 L1 and L2 speakers are needed for power of .8 to detect between-group effects of the size reported in B. Data collection is on-going and expected to be completed by June. Here we report preliminary data from 10 L1 and 11 L2 participants (various L1s; MLexTale = 59.9, MCambTest = 15.8). (No inferential statistics due to insufficient sample sizes.) Preliminary Results: Figures 3 and 4 illustrate proportion of total looking time (PTLT) to eyes and mouth by group for videos 1 and 2, respectively. While both groups primarily attend to the eyes, we see increased and more variable attention to the mouth among L2 listeners, consistent with H1 and B's findings. However, while L2 participants' LexTale and Cambridge scores correlated strongly (rs = .88), there is no indication so far that relative attention to the mouth (PTLTMouth – PTLTEyes) decreases with increasing proficiency (|rs| < .1), also consistent with B.

Findings from the full sample will contribute towards more replication in L2 research (Marsden et al., 2018), as well as to our understanding of how relative reliance on different speech and co-speech cues during real-time listening may be modulated by nativeness and/or proficiency of the listener, and consequently, how different listeners may be differentially impacted by the absence of certain cues in the signal (e.g., by face masks obscuring the mouth).

ABSTRACT. Incremental processing of speech leads listeners to build expectations and to make predictions about upcoming sentence information. Prediction has, however, been argued to be compromised in L2 listeners (e.g., [1, 2), in particular for morphosyntactic information (e.g. [3]) and for discourse-related information [1]. Such findings have been related to generally lower reliance of L2 learners on syntactic or discourse cues (e.g., [4-6]). This eye-tracking study explores the time course of L2 sentence processing and the (anticipatory) integration of different types of linguistic information in different types of noise. We investigate (a) whether L2 listeners continue to predict during sentence comprehension even when noisy speech causes phonetic unreliability or difficulties in information integration, and we examine (b) the degree to which different linguistic cues are affected by different types of noise in L2 listeners. This way, we aim to investigate which cues are particularly vulnerable in L2 predictive processing and why. According to the Shallow Structure Hypothesis [4] and [5], morphosyntactic cues are less robust in L2 processing than discourse cues; according to the Interface Hypothesis [6], discourse cues are less robust than inflectional cues. Using a visual world paradigm, we tested predictive processing among 72 German advanced L2 listeners of English in three linguistic conditions (discourse, morpho-syntax, lexical semantics; see examples 1-3 in Fig.1), and across three acoustic conditions (quiet, stationary noise, multitalker babble noise; see visualizations in Fig. 2), which were presented to three subgroups (=noise groups). Noise can impact information integration in differential ways. Whereas stationary speech-shaped noise serves as an energetic masker, lowering the salience of bottom-up information (e.g. inflectional information, function words), multi-talker babble combines energetic and informational masking, thus adding cognitive load [7]. Language proficiency is known to modulate effects of noise on speech perception and processing [7], raising the question which linguistic cues are particularly vulnerable in L2 predictive processing, and how different noise types may affect the processing of these cues. Results from cluster-based permutation analyses and lmer suggest overall effects of noise group and of linguistic condition. As expected, the use of lexical semantic information remains predictive across noise groups. Listening to speech in stationary noise (cf. Fig. 2) leads to general delays in prediction. We also observed some interactions of noise group x linguistic condition. Comparing prediction in default vs. marked structures, our L2 listeners presented with sentences in stationary noise (compared to quiet) did not show any predictive use of inflection in (2b), suggesting processing delays based on perceptual difficulties with low-salient information. Listeners presented with sentences in multi-talker babble, by contrast, failed to predict based on discourse-related cues, suggesting that the use of discourse cues in an L2 is subject to greater difficulties when the integration of information is compromised, in line with [6] and [7]. In all, these findings suggest that linguistic and acoustic aspects interact in predictive processing in the L2. We discuss potential implications for models of L2 language processing and models.

ABSTRACT. Although research on bilingualism has attracted great scientific interest in recent decades, we still do not fully understand how bilinguals’ language experience impact language access and cognitive functioning. I will discuss a series of studies that demonstrate that being exposed to one language, even for a short time, can influence the ability to use the other language and it also affects how efficiently we process non-linguistic information. I will focus on two types of effects related to the prior language experience of bilinguals: (1) the impact of exposure to the second language on native language processing and (2) the impact of bilingual's patterns of language use on cognitive control. As we will see, the available behavioral and neuropsychological evidence demonstrates that both language and cognitive systems are subject to rapid changes, however, understanding the underlying mechanism of the effects is far from obvious. We will discuss the possible convergence of mechanism related to short and long term language experience, as well as sketch out exciting research avenues for the future.