Session 3D

A fundamental inquiry within bilingual processing research addresses the underlying mechanisms of lexical access (e.g., Costa et al., 2017; Dijkstra & Van Heuven, 2002, Jiang, 2019). Research involving bilinguals of orthographically-similar scripts has revealed that cross-language activation is non-selective (Colomé, 2001; van Heuven, Dijkstra & Grainger, 1998), which causes the bilingual brain to regularly manage the activation of two languages (for a discussion, see Kroll, Dussias, Bogulski, & Valdes Kroff, 2012). Such continuous management of two languages has led some researchers to argue that the bilingual experience contributes to enhanced executive control (e.g., Bialystok et al., 2004, 2005; cf. Paap & Greenberg, 2013; Paap & Sawi, 2016). However, research on selectivity in lexical access has overwhelmingly involved bilingual speakers of orthographically-similar scripts (e.g., Degani, 2011; Hermans, Bongaerts, De Bot, & Schreuder, 1998), with minimal examination of bilingual speakers of orthographically-distinct scripts. Additionally, while active management of both languages is expected for bilinguals, little is known as to whether language selectivity is related to individual variation in executive control. Rather, research investigating executive functions (EFs) in relation to bilingual processes has largely been conducted within the context of switch costs (Meuter & Allport, 1999; Festman, Rodriguez-Fornells, & Münte, 2010), which has been associated with methodological issues (for a review, see Bobb & Wodniecka, 2013 and Declerck & Philipp, 2015). Additionally, switch costs have mainly been used to reflect language control rather than degree of cross-language activation. Language control is typically examined in relation to inhibition, which has recently been questioned as a valid psychometric construct (e.g., Rey-Mermet, Gade, & Oberauer, 2017). In light of the issues outlined above, the current study investigated selectivity in lexical access among bilinguals of orthographically-distinct scripts and explored the relationship between the degree of selectivity and top-down goal maintenance, an EF that was theoretically posited by the Adaptive Control Hypothesis (Green & Abutalebi, 2013). By manipulating degree of language task demands (comprehension and production), the study adopted an individual differences approach to lexical access and employed alternative non-switch tasks to investigate the relationships between EF relevant to lexical access and cross-language activation. 140 Arabic-English bilinguals participated in monolingual phoneme monitoring and masked primed lexical decision tasks, which assessed language production and comprehension, respectively. To assess the targeted EF, participants also completed the AX-CPT task. The data were fitted into Linear Mixed-Effects Models (lme4 package, version 1.1.20) and Multiple Linear Regression Models in R (version 3.5.2). While the analyses revealed non-selective lexical access in language production (p < 0.0001), they were inconclusive for language comprehension (p > 0.05). In language comprehension, bilingual participants varied in the degree of selectivity. When examining selectivity in language comprehension in relation to EF, top-down goal maintenance accounted for 10.9% of variance (R2 = .11), and was marginally significant for language production (R2= 0.067) The study revealed that selectivity is partially influenced by task-dependent variables as well as individual differences in executive functions. For the former, unlike language production tasks, comprehension tasks implicitly elicit the non-target language, which may cause a less pronounced activation of the non-target language. For the latter, higher cross-language activation is associated with higher reliance on proactive inhibitory control. Rather than focusing on demonstrating cognitive advantages to bilingualism, the present study explored the factors that contribute to the differences in executive processes underlying these supposed EF advantages that appear in some bilinguals but not others (for a review, see Dong & Li, 2015). The study contributes to the field’s understanding of language and cognition that research limited to inhibition, switch costs, and languages of orthographically similar scripts alone has not yet achieved.

This study builds on research investigating how adult learners break into a novel acoustic signal to transform it from a stream of incomprehensible noise into a sequence of recognizable forms through a cross-linguistic study. Previous studies have examined to what extent specific input properties, namely frequency and transparency, play an important role in input processing. Some studies claim an important role of frequency in word recognition (Braine et al. 1990; Slobin 1985; Ellis 2002; Gullberg et al. 2010; Rott 1999), whereas others maintain that this frequency effect may not be immediate (Slobin 1985; Rast 2008), or may well be very limited at the early stages (Carroll 2012). With respect to transparency, research shows that transparent words (better known as ‘cognates’) facilitate segmentation (Carroll 1992; Kellerman 1983; Rast 2008; Singleton 1993-94), with words in middle position being more difficult to extract than those on the edges of utterances (Shoemaker & Rast 2013; VanPatten 1996). Research also demonstrates that the learner’s first language (L1) provides a linguistic and conceptual ‘frame’ that allows the learner to deduce hypotheses based on the target language input (Giacobbe 1992). The current study seeks to better understand the interactions between frequency and transparency on one hand, and linguistic features of the target language and the L1 on the other when exposed to a novel target language. Data were collected by means of a Word Recognition task from two distinct groups of French native speakers attending a beginning language course designed to control and document target language exposure throughout the instruction and experimental period. One of the groups studied Polish (n=17); the other studied Arabic (n=11). Participants were tested on their ability to extract words from Polish or Arabic sentences at two time intervals (T1: before language instruction began; T2: after 6.5 hours of language instruction for the Arabic group, and 7 hours of instruction for the Polish group). During the task, participants heard a sentence, followed by a word. They were then asked to indicate whether the word was in the sentence or not. Two input variables were investigated: 1) the effect of frequency of forms in the input (absent vs. frequent, the latter being 20+ tokens); 2) the effect of phonological transparency (transparent vs. opaque), i.e. phonological similarity between the lexical forms in the L2 and the L1 (e.g. ‘transparent’ item - Polish profesor vs. French professeur and – Arabic italiyy vs. French italien; or ‘opaque’ item - Polish t?umacz vs. French traducteur and – Arabic yunaniyy vs. French grec Greek). Results show that the interaction between certain properties of the target language input (frequency and transparency, in this case), linguistic features of L1 (French), and linguistic features of the target languages, either Polish or Arabic, affected the progression of word recognition differently. At test time 1, transparency played a strong role in participants’ ability to recognize words in both Polish and Arabic. Opaque words, however, were recognized better in Polish than in Arabic. At test time 2, all items were recognized at the same accuracy rate across the two French native speaker groups, including opaque/absent words. Interestingly, at the second test time, the learners exposed to Arabic performed as well as those exposed to Polish, suggesting that the additional input exposure was particularly beneficial to those learning Arabic. We discuss the results in light of phonotactic features of the test items and general prosodic characteristics of Polish and Arabic.

A combined look at development of L2 processing and a common L2 experience, studying abroad, reveal and interesting combined trend: an intermediate stage where a decline in performance is observed. Previous research in L2 processing of gender agreement shows that early learners have been found to process like native speakers (e.g., Dekydspotter & Renaud, 2009; Renaud, 2011), intermediate learners to differ from native speakers (e.g., Scherag, Demuth, Rösler, Neville, & Röder, 2004), and advanced learners to eventually process like native speakers (e.g., Dussias, Valdés Kroff, Guzzardo Tamargo, & Gerfen, 2013). L2 learners abroad have been found to perform worse after a short stay abroad (e.g., Cubillos, Chieffo, & Fan, 2008; Isabelli-García, 2010; Schwieter & Klassen, 2016). These results have been explained as shift towards a communication-based strategy, in which gender agreement is not essential for communication. Twenty-six (L1 English) L2 Spanish learners were tested before and after a short-term immersive study abroad (SA) program. In a grammaticality judgement task, the learners rated sentences which had matched or mismatched gender for definite determiner-noun pairs. Results showed that the participants rated gender matched sentences higher than gender mismatched sentences both pre- and post-SA (? = -0.87, t(0.15, 15.31) = -5.754, p = 3.52e-05). In the self-paced reading task, the learners read gender matched or mismatched determiner-noun pairs. Results revealed that participants read gender-matched pairs significantly faster than mismatched pairs in the pre-SA session (d=164ms), but not in the post-SA session (d=21ms; ? = 132.14, t(59.03, 511.60) = 2.238, p = 0.0256). These results indicate that the learners were sensitive to gender mismatches prior to their SA experience but lost this sensitivity by the end of the experience. The third experiment was a picture selection task in which participants saw two images that differed in gender of the item and/or colour. The learners heard an auditory stimulus consisting of a definite determiner and an adjective but no noun (e.g., Elija la roja ‘Choose the.fem red.fem’). When the two images had the same colour but different gender, the participants selected the target image slower than when the colour was different, whether they had the same or different gender, in both the pre- (d=351ms) and post-SA sessions (d=185ms; ? = 146.587, t(50.086, 943.30) = 2.927, p = 0.00351). The results indicate that learners ignored the gender agreement morphology unless required to attend to it in both pre- and post-SA sessions. The implications from these experiments suggest that early-intermediate learners rely less on morpho-syntactic cues in processing over time spent abroad. This is in line with previous explanations for decline in SA contexts which posit that the prioritization of communication causes learners to stop attending to morpho-syntactic cues which do not lead to a breakdown in communication (e.g., Collentine, 2009; Schwieter & Klassen, 2016).

It has been postulated that the age at which individual speakers become bilingual may affect the weighting of lexical vs. grammatical cues: Bilinguals who were younger when they were first exposed to an L2 rely more on rules and grammatical cues, while older learners may tend to prioritize lexical information and non-grammatical cues (e.g. Cunnings, 2017). To what extent a similar, reversed effect may for L1 attrition has not, to date, been comprehensively explored. In a comparative study on L1 attriters and L2 learners, we test the relative weighting of cues for a grammatical feature which combines both item-based and rule-based characteristics. Grammatical gender in German is such a feature: while gender assignment to nouns is typically taken to be arbitrary and opaque, there are some phonological characteristics which are stochastically (although not deterministically) associated with some genders. On the other hand, gender agreement is systematic, but due to the interaction with other grammatical features (case, number, definiteness) the inflectional system contains a great deal of syncretism and the surface realisation of gender is therefore often opaque. We present the findings from a pilot study that will serve to underpin a more comprehensive investigation in L1 attrition and L2 acquisition, treating bilingualism not as a dichotomous variable but as a continuum and studying both gender assignment and gender agreement processes. In order to gain insight into potentially differential processes subserving gender assignment, this pilot study assesses two populations of bilinguals: late immersed German-English expats (L1 attrition group) and late English-German L2 learners (L2 group), as well as a monolingual control group (all groups n=30). In a reaction-time study, they are presented with German words and pseudo-words of all three genders. The gender of half of these is predictable based on phonological properties identified e.g. by Köpcke & Zubin (1984), while the other half is unpredictable. Half of the real-word items are high-frequency and the other low-frequency. The task is a simple article-assignment task, and we measure accuracy (for both words and non-words) and reaction time across all populations. We assess the impact of linguistic properties (predictability, gender), item frequency and of extralinguistic factors relating to bilingualism (AoA, proficiency level, cognitive capacity) on accuracy and reaction times and discuss the results in the light of different theoretical approaches to bilingual development. References Cunnings, I. (2017). Interference in native and non-native sentence processing. Bilingualism: Language and Cognition, 20(4), 712-721. Köpcke, K. M., & Zubin, D. A. (1984). Sechs Prinzipien für die Genuszuweisung im Deutschen: ein Beitrag zur natürlichen Klassifikation (pp. 26-50). na.