In this study, we investigate the degree of grammaticalization of the griege-passive in Pennsylvania German as spoken by Anabaptists in Ohio. Parallel to the development of recipient passives in Continental German and English, Pennsylvania German is developing a passive construction using the verb griege ‘to get, receive’ as an auxiliary (Burridge 2006: 186-187). Examples taken from Burridge (2006: 186) are given in (1) and (2). 

(1) Ich hab e Buch gewwe griegt. 

I have a book given got 

‘I got given a book’ 


(2) Mir griege gesaagt. 

we get told 

‘We get told’ 

As illustrated in (1) and (2), the degree of grammaticalization varies. In (1), the sentence licenses the object Buch which is assigned the role of patient as would be expected with the lexical variant of griege and illustrated by the sentence Ich hab e Buch griegt ‘I got a book’, (cf. German Ich habe ein Buch geschenkt bekommen "I got given a book" versus Ich habe ein Buch bekommen "I got a book"). In contrast, (2) shows a greater degree of grammaticalization due to the absence of an overt object. Similarly, the lexical verb griege assigns the role of recipient to the subject, but this is not always the case for subjects in griege passive constructions as illustrated in (3) (example from Burridge 2006: 186): 

(3) Er hat sei Lewwe genumme griegt. 

he has his life taken got 

‘He got his life taken/He was killed.’ 

Once again, a parallel structure can be found in German in sentences like Sie haben den Ball weggenommen gekriegt ‘They got the ball taken away (from them)’ (Burridge 2006: 186). The forms in (2) and (3) can be taken as examples of semantic bleaching (desemanticization) and extension (i.e., the rise of new grammatical meanings via context-induced reinterpretation) that are hallmarks of grammaticalization (see discussion in Heine and Kuteva 2005). We note that in English it is possible for the subject of get-passive sentences to actually be assigned the role of patient whereas this is not grammatical in standard German, as shown in (4): 

(4) English: The ball got taken away. 

German: *Der Ball hat weggenommen gekriegt. 

The ball has away-taken gotten 

'The ball got taken away' 

The grammaticality of sentences such as (4) in Pennsylvania German has not been previously investigated to the best of our knowledge. In this study, we will systematically examine the grammatialization of griege-passives in Pennsylvania German along a cline illustrated by the sentences in (1) – (4). Reference Burridge, Kate. 2006. Language contact and convergence in Pennsylvania German. Grammars in contact, ed. by A. Y. Aihenvald and R.M.W. Dixon, 179-200. Oxford UP: Oxford and New York. Heine, Bernd and Tania Kuteva. 2005. Language contact and grammatical change. Cambridge UP: Cambridge.

Past research has shown that adult native speakers of Spanish immersed in an English-speaking environment adopt processing routines of their second language (L2) when processing their first language (L1). Here we ask whether changes in processing routines can be triggered by overexposing bilingual speakers to particular structures so that bilinguals who have undergone changes when processing in their L1 ‘move back.’ We hypothesized that if the parser’s configuration is related to language exposure (e.g., MacDonald & Seidenberg, 2006), bilinguals’ parsing preferences are expected to change as a function of the frequency with which the relevant structure appears in an experimental session. We investigated this in the context of temporarily ambiguous relative clauses as in Arrestaron a la hermana del hombre que estaba enferma (Someone arrested the sisterFEM of the man who was illFEM). Here, the relative clause que estaba enferma (who was ill) can attach to the higher noun (hermana/sister) or the lower noun (hombre/man) in the complex noun phrase. Because the adjective enferma (ill) is marked with feminine grammatical gender, the correct interpretation is one where the relative clause attaches to hermana (also marked with feminine gender). Spanish-English bilingual speakers living in an English-speaking (US bilinguals) or a Spanish-speaking (Spain bilinguals) environment were recruited. 

The study involved three phases. In phase 1, we carried out an eye-tracking study to determine the participants’ initial attachment preferences in their L1 (Spanish for both groups of speakers). In Phase 2, they participated in an ‘intervention’ study that exposed them to a biased sample of 120 relative clause constructions in their second language (English) over a 5-day period. Participants who showed initial attachment preferences in Spanish that favored low attachment received a biased sample of high attachment sentences; those whose initial attachment preferences favored high attachment received a low attachment treatment. In phase 3, eye-movement records were again collected to determine participants’ attachment preferences after the intervention study. The Spain bilinguals were tested two days after the intervention study had ended. The US bilinguals were, in addition, tested a week later to assess non-immediate effects of exposure. At pre-test, all participants were also administered the AX-CPT task (to measure sustained attention) as well as the Raven’s matrices and a standardized English language test (to match the US and Spain bilinguals). 

Preliminary results reveal the following: (1) Changes in attachment preference occurred in both groups of speakers but were modulated by performance in the AX-CPT task; (2) “High attachers” showed evidence of low attachment preferences at post-test 1 and at post-test 2 despite the fact that the input in the intervention study was in their L2, and even when they were immersed in an entirely Spanish-speaking environment. This provides evidence for a high level of permeability between the bilinguals’ two linguistic systems; and (3) “Low attachers” showed evidence of a high attachment preference only at post-test 2, suggesting that learning to attach high requires consolidation of information. The findings will be discussed in terms of the role of statistical learning in sentence comprehension processes (e.g., Gennari & MacDonald, 2009; Wells et al., 2009) and the implications of the results for theories of cognitive control.

A characteristic of bilingual speech is the occurrence of sentences that contain words of both languages. These codeswitches provide a window into how a bilingual’s two languages interact in conversation. The current project examines bilinguals’ online processing during comprehension of sentences that contain codeswitches. Previous psycholinguistic and neurocognitive research has shown that switching between isolated items incurs an asymmetrical processing cost where it is harder to switch into the dominant language than into the weaker language. This cost is thought to be related to the inhibition of the dominant language. In contrast, relatively little is known about the processing of codeswitches in a meaningful sentence context. In three experiments, sentential codeswitching was examined using a self-paced reading task and event-related potentials (ERPs) in two groups of Spanish-English bilinguals: those immersed in L2 English who codeswitch frequently in their daily life, and those immersed in L1 Spanish who do not codeswitch. Stimuli were 160 sentences that began in Spanish or English and either contained a codeswitch into the other language or not. Sentences that contained codeswitches were read more slowly than those without switches, but only when the sentences switched from the dominant to the weak language, not from the weak to the dominant language. Additionally, ERPs for both habitual and non-habitual codeswitchers showed a late positivity in response to codeswitched words as compared to non-codeswitched words, but again only when switching from the dominant to the weaker language. No effect of codeswitching occurred when switching into the dominant language. These results are in contrast to the language switching studies, and indicate that codeswitching in a meaningful sentential context requires different processing mechanisms. Given the appearance of the late positivity and consistency across habitual and non-habitual codeswitchers, it appears that the processing of sentential codeswitching may rely on fundamental sentence-level integration mechanisms related to activation of the weaker language.

Infants growing up in bilingual environments must build a language system that accommodates two languages. An important task for these infants is to discriminate and differentiate their languages. While it is easy to imagine an early bilingual environment that neatly packages the two languages in a way that facilitates this task (e.g. one-parent-one-language), most bilingual infants do not encounter their languages in this way. This talk will present evidence that bilingual infants typically hear their two languages in bilingual contexts: spoken by the same person, in the same situation, and/or within the same sentence. Experimental work is beginning to reveal how bilingual infants cope with the bilingual nature of their input, including language discrimination and the processing of code switched speech. These findings will be discussed in the context of the PRIMIR framework of infant speech perception and word learning.

Bilinguals activate both of their languages in parallel despite the intention to read or speak in a single language. Yet bilinguals are apparently able to use one of their two languages without overt interference from the unintended language. A fundamental question in recent psycholinguistic research has been to determine the mechanisms that allow for successful language selection. One potential linguistic mechanism is the language cue. Languages differ in many respects (e.g., orthography, phonology, and syntax) and environments differ in the requirement of which language(s) may be in use. These differences could theoretically allow bilinguals to access one language selectively. This talk will explore the role of syntactic differences between Spanish and English in negotiating parallel activation. The approach exploits evidence from two methodologies to examine the role structure during word recognition. First, word naming during sentence reading will be used to measure the degree of parallel activation within various syntactic structures (specifically: active, passive, and dative structures). If certain structures function as a language cue, evidence for parallel activation should be reduced or absent for words read within those structures. Second, cross-language syntactic priming will be used to confirm whether the chosen syntactic structures are in fact represented in a language independent manner.

Listeners utilize the structural properties of their language to help them interpret speech, leading to misperception of non-native sounds and sound sequences. For example, Spanish contains no words with initial sCV sequences, a phonotactic restriction well-known to influence cross-language borrowings. Words like snob are “repaired” by prepending an initial /e/ (esnob). Interestingly, similar repairs are apparent in on-line speech perception tasks. Native Spanish speakers confronted with a spoken token of snob report hearing esnob, even though the stimulus lacks the initial vowel, whereas speakers of languages that allow sCV sequences (e.g. French) do not (Cuetos et al., 2011; Hallé et al., 2008; to appear). 

How does bilingual perception relate to these monolingual extremes? Second language users can develop native-like perception of sCV sequences—even when this conflicts with their first language (Parlato-Oliveira et al., 2010). However, the perceptual consequences of possessing conflicting phonotactic constraints are unclear. Does acquiring a language that allows sCV structures eliminate misperceptions, even in the more phonotactically restrictive language? If not, how does bilinguals’ speech perception shift in response to changes in linguistic context? We investigated these possibilities in bilinguals fluent in both Spanish and English via a vowel detection task (Cuetos et al., 2011) and an AX discrimination task using a pretest-posttest design. The pretests were conducted in a monolingual Spanish experimental enviromnent, after which half the participants then performed an English picture-naming task before repeating the Spanish vowel task. The performance of this group was compared to a group that performed the picture-naming task in Spanish. A monolingual English control group was included for comparison. The bilinguals’ pretest performance reflected the influence of Spanish phonotactics, but this varied with language dominance in ways consistent with influence from English phonotactics. Intriguingly, changes in performance at posttest depended on the language of the intervening task, again modulated by language dominance. This suggests that bilinguals can deploy distinct phonotactic constraints dynamically depending on shifting linguistic context.

According to the Revised Hierarchical Model (Kroll & Stewart, 1994), second language (L2) learners initially access meaning of L2 words via the L1 whereas advanced learners access meaning directly. We tested this hypothesis with English learners of Spanish in a translation recognition task, in which participants were asked to judge whether English words were the correct translations of Spanish words. In each case, we gathered data on behavior and on the earliest time course of processing using ERPs. The critical conditions compared the ability of learners to reject distractors that were related to the translation in form or meaning when a long (750 ms) or short (300 ms) SOA separated the two words. For advanced learners, there were effects for semantic and form distractors in both measures at the long SOA, but at the short SOA, there were behavioral effects but only an N400 effect in the ERP record for semantic distractors. These results replicate Guo et al. (2012), suggesting that relatively proficient L2 speakers access the meaning of L2 words directly. For beginning learners, at the long SOA, there were semantic and form effects in both measures. At the short SOA, behavioral data were sensitive to distractor type, but the ERPs only revealed larger N400 and smaller LPC for translation distractors and no effect for semantic distractors. Overall, these data suggest that at early stages of L2 learning there is reliance on the L1 translation equivalent. Once proficient, they are able to retrieve the meanings of the words directly.

Past research shows that lexical access is non-selective with respect to language, allowing cross-languages interactions to occur in both comprehension and production (Dijkstra, 2005; Guo, Liu, Misra, & Kroll, 2011). A key question in bilingual research has been to understand the control mechanisms that allow bilinguals to select the language they intend to use. Language comprehension and production potentially differ in the way in which bilinguals achieve control of their two languages, e.g., in the time course of inhibition. 

In past research, we have shown that cross-language inhibition in comprehension seems to be relatively short-lived (Martín, Macizo, & Bajo, 2010). In contrast, studies of lexical production have shown that inhibition of the language not in use can be long lasting (e.g., Misra, Guo, Bobb, & Kroll, 2012), suggesting that there are multiple mechanisms of control. 

The present study explored the nature of the control mechanisms that underlie language selection in bilingual production and specifically whether there is evidence for both automatic and controlled selection processes. Relatively proficient Chinese-English bilinguals performed a picture naming task in language blocked or language mixed conditions (Guo et al., 2011). In one condition, they were instructed to name the picture, in another they also had to perform a concurrent updating task. 

Results showed that the concurrent task affected performance differentially in the blocked and mixed conditions. Under mixed conditions that included the demanding updating task, bilinguals who were strongly L1 dominant were as slow to speak the L1 as the L2. The updating task did not eliminate the inhibition of L1 under mixed conditions. In contrast, introducing the updating task in the blocked conditions appeared to eliminate the inhibitory effect of L1 when it followed L2. Findings will be discussed in the context of models of bilingual control.