Age of arrival organizes immigrants' language learning environment Catherine L. Caldwell-Harris Boston University Marianna Staroselsky University of Chicago Svetlana Smashnaya Belarussian State Educational University Nadya Vasilyeva Northeastern University Keywords: Second language acquisition, bilingualism, age effects, immigrants, emotions, language usage, Russian speakers 25 double-spaced text pages, 5 tables, 6 Figures, 11,000 words total. Address for correspondence: Catherine L. Caldwell-Harris, Ph.D. email: charris@bu.edu, telephone: 617 353-2956 fax 617 353-6933 Department of Psychology, Boston University 64 Cummington St., Boston, MA 02215 USA Abstract For Russian speaking immigrants to the U.S., age of arrival (AoAr) strongly influenced frequency of using L1 vs. L2 with family and friends, perceived emotionality of the two languages, and self-perceived L1 and L2 proficiency. We argue that AoAr so strongly correlated with these variables because age of immigration is a potent organizing variable: early arrivals have an L1 that is not fully developed, and they encounter a usage environment that is richer in L1 and poorer in L2. Later arrivals encounter the opposite. The environmental variables accompanying early arrival set young immigrants on a path to switching their dominance to L2, putting these learners at risk for less than native speaker attainment of L1. The environmental variables that accompany later arrival facilitate maintaining L1 dominance, and can lead to lack of native speaker attainment of L2. Dominance that solidifies in the 23 years after immigration then further dictates usage patterns, with usage patterns then being the proximal mechanism for ultimate proficiency attainments. Immigrant SLA represents a complex dynamical system which should be studied from a multicausal perspective. 2 Age of arrival organizes immigrants' language learning environment Over the last two decades a central question in second language research has been understanding why early learners generally acquire a second language to higher proficiency than do older learners. This is frequently referred to as the question of how and why age of acquisition influences ultimate or long-term attainment (Birdsong, 1992, 2004; Bongaerts, Summeren, Planken & Schils, 1997; Jia, Aaronson & Wu, 2002; Lardiere, 2006). As often happens when a question is first investigated, early theorists emphasized single factor explanations, especially a critical period determined by maturation (Johnson & Newport, 1989; Patkowski, 1980, 1990). A great deal of research on age effects has appeared since these early papers, with authors noting logical and empirical failings of the single-factor maturational accounts (Bialystok & Hakuta, 1998; Kellerman, 1995; Schumann, 1978; Snow, 1983; Moyer, 2004). In the last decade, researchers have increasingly proposed that multiple factors are involved (Bialystok & Miller, 1999; Birdsong, 2004; Birdsong & Molis, 2001; Flege, Yeni-Komshian & Liu, 1999; MacWhinney, 2005, and others). In a comprehensive study of immigrants to the U.S., Jia, Aaronson, and Wu (2002) argued that adults' proficiency in L2 (English) depended not just on age of acquisition, but on similarity between the first and second language, cultural and usage factors such as home use of the L2 and frequency of using two languages with peers. While there is increasing acceptance of the view that multiple factors determine ultimate attainment of native-like proficiency, an important role for age remains compelling, especially in studies of longterm attainment in immigrants. For immigrants, age of arrival (AoAr) to the host country is frequently the strongest single factor correlating with outcome variables (such as proficiency; Jia et al., 2003). When AoAr emerges in multiple regression as the variable able to uniquely explain all or most of the variance in proficiency, researchers conclude that AoAr is either the single causal factor (Newport & Johnson, 1989) or the strongest causal factor (Birdsong & Molis, 2001; McDonald, 2000). Authors who propose alternatives to a strong role for maturation argue that AoAr is confounded with other factors, such as length of stay, years of education in the host country, social/motivational factors, and family members' proficiency (Jia & Aaronson, 2003). For example, Flege et al. (1999) reported that AoAr is not the strongest predictor of ultimate attainment when years of education in the U.S. is taken into account, and thus argued that AoAr may not be the causal factor. We are sympathetic with these proposals but feel they do not go far enough because they do not explain why AoAr has such high correlation with outcome measures. We propose that AoAr be regarded as a powerful "organizing variable." We define "organizing variable" as one that causes, creates, organizes, or sets into motion a large set of factors which jointly contribute to an outcome. Age influences myriad environmental factors including schooling and treatment by peers. Perhaps more importantly, age also influences 3 cognitive and social abilities, including learners' ability and motivation to seek out L1 vs. L2 speaking friends. A particular factor may exert only weak effects on its own, but aggregated together with other age-related factors the cumulative impact can be large (see related ideas discussed by Moyer, 2004). Age of acquisition is likely to be an especially potent organizing variable in the case of immigrants. The experience of immigration at a young age means being thrust into schools and among L2-speaking peers before L1 acquisition is complete. Jia and Aaronson (2003) found that within a year of arrival young immigrants (age 9 and younger) had switched their preferred language to L2, while older immigrants maintained L1 dominance. Young age is a powerful organizing variable because of cognitive development and social skills. Young children have less ability and opportunity to choose their reading material and their friends. In Jia and Aaronson's (2003) study, older immigrants had the skills and maturity to seek out, even at long distance, L1-speaking peers, while young immigrants ended up socializing with the L2 children in the immediate neighborhood. Older immigrants likely have both more motivation and greater ability to locate and maintain friendships with L1 speakers, and the older they are at time of immigration, the greater their motivation and friend-finding resources become (MacWhinney, 2005). Jia et al. (2003) documented substantial L1 language development through late childhood. At immigration, teenage and young adults will usually have entrenched and highly proficient L1 abilities; indeed, the existing neurological and cognitive commitments to a first language has long been proposed as a key mechanism underlying age effects (MacWhinney, 2005). Their awareness of their poor L2 speech can be the motivation to seek out L1speaking friends. Their greater experience of life in the country of origin creates nostalgia for L1 culture and themes, which is further motivation to seek out L1-speaking friends. In the current paper, we investigate adults who emigrated to the U.S from Russian speaking countries (or who were the children of Russian immigrants), along with a comparison group of nonimmigrant Americans who studied Russian as a foreign language. We seek to support and extend the findings of Jia and Aaronson (2003) that the cognitive, social and linguistic immaturity of early arrivals reduces the richness of the L1 environment and increases the amount of L2 encountered postimmigration. In contrast, older arrivals have high social agency and entrenched L1. This allows them to find L1-speakers to maintain their socio-emotional and linguistic connections to their country of origin. Researchers who advocate brain maturation (e.g., Johnson and Newport, 1989) as the cause of age effects have focused on young age as the factor that allows native-like attainment of L2. A less well studied outcome of immigration is failure to maintain native-speaker levels in L1 for those immigrating in early and middle-childhood (Jia & Aaronson, 2003). Maturational factors can not explain why 4 proficiency is often at risk for early arrivals and thus instead L2 interference combined with cognitive, social and motivational factors must be posited (Pallier, 2007). A goal of the current paper is to investigate age effects in both L1 and L2, and to investigate not just how proficiency changes with AoAr, but to compare AoAr effects on proficiency with AoAr effects on the aspects of environmental language that are likely to vary with AoAr, such as family vs. friend usage of L1 and L2, and emotional resonances of the two languages. Emotional resonances of a language The "Emotional resonances" of a language are the perceived strength of taboo words and phrases like "I love you" (Dewaele, 2004, 2008; Pavlenko, 2005). Bilinguals frequently report experiencing reduced emotion when using their second language (Altarriba & Santiago-Rivera, 1994; Bond & Lai, 1986; Gonzalez-Reigosa, 1976; Marian & Kaushanskaya, 2004; Schrauf, 2000). Self-reports have been confirmed by electrodermal monitoring, revealing that skin conductance amplitudes are reduced when bilinguals read or hear emotional words and phrases in their second language (Harris, Aycicegi, & Gleason, 2003; Harris, 2004). The first language may generally be experienced as more emotional than the second language because emotional regulation systems co-evolve with language (Bloom & Beckwith, 1989), or it may be that the contexts of early childhood is generally more emotional than the contexts of later childhood, because of bonding with caregivers (Harris, Gleason & Aycicegi, 2006). The first language is not always the most emotional (Aycicegi-Dinn & Caldwell-Harris, 2009; Dewaele, 2006; Schrauf and Durazo-Arvizu, 2006). In Dewaele and Pavlenko's (2001) web questionnaire, respondents downgraded the illocutionary strength of L1 taboo/swear words, anger and the expression "I love you" if L1 was no longer the dominant language (Dewaele 2004, 2006, 2008). For each of these types of emotional expressions, multiple factors influenced how multilinguals evaluated the emotional force of these expressions in their different languages. The most consistently relevant were, in order of strength of effect: language dominance, proficiency, usage frequency, age of acquisition and acquisition context (naturalistic vs. instructed). Harris et al. (2006) named their framework for predicting which language is more emotional the "emotional contexts of learning hypothesis." Drawing on past research in developmental psychology and language acquisition, this theoretical framework proposes that language comes to be experienced as emotional to the extent that it is learned and used in emotional contexts. Finding how perceived emotionality of L1 vs. L2 varies with AoAr, proficiency and usage factors will allow us to support or quality this hypothesis. 5 In the current paper, we investigated how judgments of emotional resonances correlated with AoAr, family/friends usage, and self-reported proficiency. In particular, enjoyment of using the native language with friends could contribute to seeking out L1-speaking friends, which then helps maintain both L1 proficiency and the strong emotional resonances in L1. Our goal was thus to identify interacting variables that mutually support each other and lead to acquisition of L2 and/or maintenance of L1. Studying Russian-English bilinguals The current research builds heavily on the questionnaire-based studies of Dewaele (2004, 2006, 2008) and Pavlenko (2005). Their sample reflects the European experience of bi- and multi-lingualism, including respondents who ranged from having only classroom exposure, to those who lived in or moved between polyglot communities and countries with several national languages. Their web-based questionnaire resulted in a large numbers of respondents, which conferred considerable statistical power, which was useful in discerning broad patterns in their heterogeneous sample. The current study focuses on bilingualism in the American context of immigration. We studied only Russian-English bilinguals to benefit from cultural and linguistic homogeneity in our sample. Russian immigrants to the U.S. and children of immigrants present a useful addition to the literature on SLA in U.S. immigrants for several reasons. Many of the existing studies either focus on an L1 that is very dissimilar to English, such as Mandarin (Jia et al., 2002; Jia & Aaronson, 2003), Korean (Johnson & Newport, 1989), or Vietnamese (McDonald, 2000), or very similar, such as Spanish (Birdsong & Molis, 2001) or Dutch (Kellerman, 1995). Russian is between these extremes in its linguistic similarity to English, having grammatical structures and vocabulary which are less similar to English than Dutch and Spanish, but more similar than Asian languages. It also has a writing system which has both shared and different letters from English. The Russian immigrants we studied lived in the Boston area where they are highly economically successful and well integrated into the dominant culture, while still numerous enough for clusters of visibly Russian-oriented shops and restaurants. However, we will be able to draw correlations between AoAr and self-reported proficiency to the Russian-English bilinguals studied by McElree, Jia and Litvak (200), whose sample was broadly similar to ours. We expected young arrivals to mostly have L2-English as their dominant language, consistent with SLA studies of young immigrants (Birdsong & Molis, 2001; Jia & Aaronson, 2003). Many of them would likely identify themselves as bi-cultural, and we wanted to be able to determine where they more resembled their monocultural, monolingual English-speaking peers, and where they resembled the Russian-oriented older arrivals. We included respondents who had been born in the U.S. to immigrant 6 parents if Russian had been spoken in the home while growing up. A monolingual English comparison group could not be asked whether English or Russian felt more emotional. We thus included native English speakers who studied Russian as adults so that they would be able to answer the same questions as our bilinguals about whether English or Russian felt more emotional, and to obtain a different perspective on Russian-English bilingualism, Method Participants and procedure Recruitment. Russian-English bilingual speakers were recruited from the Boston University campus and the greater Boston area communities via flyers, word of mouth, and internet postings on Craigslist and Facebook. We desired a naturalistic sample and thus included everyone who asked to be in the study who identified themselves as fitting either of the following two descriptions: (1) Fluent in both Russian and English with Russian learned in childhood, or (2) studied Russian as a foreign language as a teenager or later for at least 2 years and judged themselves to have reasonable listening comprehension of basic Russian. This second group was also recruited by asking Russian language instructors at several colleges to identify their best students. Demographic and learning history variables for the three groups in the sample (Table 1) reflect the naturally occurring characteristics of persons in the greater Boston area, mainly associated with a college campus, who had self-identified as having at least good comprehension and who wanted to be in a study of Russian-English bilinguals. Constructing two categories of native Russian speakers. Those who had learned Russian in childhood were categorized according to age of acquisition of English, defined as age of immigration to the U.S., following Johnson & Newport (1989) and Birdsong and Molis (2001). We labeled those who had immigrated before age 10 as "Early bilinguals" (N=23) and those who arrived at 10 or older "Late arrival in the U.S." (N=21). This cutoff is essentially arbitrary and using a slightly earlier or later cutoff does not change the pattern of results significantly. Although other researchers favor puberty (roughly age 13) as the preferred cutoff because of earlier arguments for a critical period in second language acquisition (Lenneberg, 1967), there is no reliable evidence of a brain-based changed in language acquisition occurring with puberty (Elman, et al., 1996). We adopted a cutoff which emphasized the critical socializing role of education in the host country. Russian speakers who immigrate at age 10-13 have had a substantial Russian educational experience and exposure to life in a Russian speaking country which makes them more similar to those who immigrate in the teen years. Choosing age 13 as a cutoff means that these individuals would be classified with bilinguals who were born in the U.S. In contrast, choosing age 7 as a cutoff would err in the opposite direction, because it would categorize 7 7 and 8 year-olds with those who immigrate in adulthood. Persons who immigrate at age 8-9 will have almost all of their schooling and substantial peer socialization in the U.S., making them a poor fit with those who immigrate at older ages. Our cutoff of age 9 for the "early learning" group is consistent with the finding of Jia and Aaronson (2003) that immigrants who arrived at age 9 or younger switched their language preference to English, while those who arrived at age 10 or later maintained their L1 as their preferred language. Eight respondents grew up speaking a language in addition to Russian, having immigrated from the former Soviet republics outside of Russia. These languages were: Belorussian (3), Ukrainian, Kazakh, Mongolian, Azerbaijan or Farsi (see preliminary analyses for the differences between the bilinguals and trilinguals). The third group to be studied consisted of American individuals who learned Russian as a foreign language in a classroom setting. Many had additional sources of expertise beyond classes, such as BA’s or masters in Russian (4), travel to Russia (5), Peace Corps in Russia (1), certified interpreter (1) and Russian- speaking wife (1). No participants reported a hearing, language, or learning disability. Procedure. Respondents completed the questionnaire sitting in a laboratory room with a research assistant sitting near by to clarify questions. The questionnaire was followed by an interview/debriefing with one of the authors or an assistant. Most respondents than participated in a study of electrodermal responsiveness to emotional phrases, described elsewhere (Staroselsky, Caldwell-Harris & Vasilyeva in preparation). Respondents volunteered, received course credit for Introductory Psychology, or received $20. Materials Questions were designed to assess three broad domains (see summary and item descriptions in Table 2). Overall learning history. Respondents answered questions about languages spoken, perceived dominance, ages of first exposure to each language, immigration dates and overseas travel and other demographic and language learning questions similar to those in Dewaele and Pavlenko (2001) and Marian et al. (2007). Proficiency in both English and Russian was an important question for us, but it is difficult to find objective tests in both these languages such that scores on the two tests can be compared. Self-report measures are known to correlate highly with linguistic measures of proficiency (Kroll et al., 2002; Jia et al., 2003; Marian et al., 2007) and have been used successfully in prior studies (Dewaele, 2006; Bialystok & Hakuta, 1998; Chiswick & Miller, 2008). It is indeed unclear whether objective tests are necessarily superior to self-report, since objective tests have the disadvantage of individual differences test-taking ability, text anxiety, and frequently display considerable measurement 8 variance. In Jia et al's study of European immigrants, there was no correlation between AoAr and grammar test scores. In contrast, in the McElree et al study of Russian-English bilinguals, self-report of both English and Russian proficiency showed strong age effects. We thus decided to rely on individuals’ verbal SAT scores (which were available for 47 participants), and self-reported proficiency. A 1-7 scale was employed (7=native speaker, 6=very good, 5=good, 4=fair, 3=adequate, 2=limited, 1=little, 0=none), with separate measures for listening comprehension, speaking, reading and writing. We then used the mean of these four values as our overall measure of proficiency, following Jia et al (2003). A reason to take the mean of the four ratings is that the mean of 4 separate introspective measures is likely to be more reliable than a single introspective measure, because the respondent is making 4 self-assessments of their proficiency. Bilinguals who are at the edge of native-speaker proficiency may feel confident about giving themselves an assessment of "7" for listening comprehension, but will select a 6 or 5 for reading and writing. The average of the 4 measures is thus likely to be a nuanced reflection of respondents' perception of their overall language proficiency. Emotional intensity. Questions asked about respondents' experiences with the three domains of anger expressions, positive emotional expressions and lying, following the content of items used in Dewaele and Pavlenko (2001). Relative emotional intensity of English vs. Russian verbal expresses was assessed with questions such as: "When you express anger, which language is stronger for you in meaning?" and "I feel emotionally uncomfortable when I am telling a lie in..." (response options ranged form Only English to Only Russian, see Table 2). These same response options were used to measure inner speech, with questions like "When I have angry thoughts they are in..." Usage preference was measured with questions such as: "What language do you prefer for expressing anger?" Here, response was with a checklist, scored as 1 for Russian, 0 for English, .5 for "No difference." As these examples show, we did not use Likert scales. While Likert scales have many advantages, one disadvantage is individual differences in response bias: some individuals may have a tendency to respond overall strongly or weakly. Our questions explicitly required respondents to compare the relative emotional force of their two primary languages. This means that we will not be able to identify whether some individuals felt emotion equally strongly in their two languages vs. equally weakly. We will simply know that a mid-point value on the English<->Russian scale indicates that the two languages were perceived as having similar emotional resonances. Because perception of relative emotional strength was our goal, obtaining a single scale was appropriate for the current project, and resulted in a relatively short questionnaire and streamlined statistical analysis. Frequency and context of use. Respondents were asked to estimate the percentage of time using each of their languages with family members, friends and in personal/intimate relations, both currently 9 and in the past, following Marian et al. (2007). For the bilingual speakers, the estimates for Russian and English summed to 100%, but because the questions were simple percentage estimates, their sum was sometimes below 100% for the small number of trilingual speakers, especially for estimates concerning use with family. Results Standardization of responses and preliminary analysis. Responses were scored from 0 (indicating Always/Only English) to 1 (Always/Only Russian) (see Table 2), to allow percentage scales checklist scales to yield means between 0 and 1. In a preliminary analysis, participants responded similarly to all anger items and all positive emotion items, regardless of whether the question concerned emotional intensity, usage preference or inner speech. This was confirmed by high Cronbach alpha scores when these different items were treated as comprising a scale (see alpha values listed in Table 2). For this reason, the averages for responses about anger, positive emotional expressions and lying were treated as uni-construct scales, meaning that emotional intensity, usage preference and inner speech jointly contributed to these scales. For convenience we will refer to these as the emotional intensity scales. Note that emotional intensity and usage preference do not invariably co-occur. In Dewaele's (2004) questionnaire of multilinguals' swear word usage, taboo phrases and curses were judged to be stronger in the more proficient language, but respondents sometimes named a different language as the preferred language for cursing. Dewaele (2004) pointed out that the speakers' usage preference depended on whether they wanted to feel the full illocutionary force of the swear word. No gender differences were obtained on any of the measures (all p > .2), including basic demographic factors as well as the usage and emotionality scales. Gender will thus not be discussed further. T-tests were conducted to determine if any of the measures varied by bilingual/trilingual status for the Late English group (N=7 trilingual, N=14 bilingual). There was no overall difference in the age of arrival but the trilinguals had immigrated more recently, having a length of stay of 3.7 years compared to Bilinguals' 11.2 years); t(19)=3.3, p < .01. Trilinguals reported less use of Russian with family (.65 vs. .95), t(19)=2.7, p < .02, because they used their third language with family. Because the bilinguals and trilinguals were very similar on the remaining measures (all p > .2), we generally grouped trilingual and bilingual speakers together in subsequent analyses except analyses which related to family usage. 10 Differences in the three learner groups One-way Analysis of Variance (ANOVA) revealed significant main effects of group for each of the ten scales. F(1,61) values ranged from 4.3 to 64.1, indicating substantial statistical power. Compared to other groups, for late arrivals, emotional expressions felt stronger in Russian, and they reported higher frequency of use for Russian and valuing of Russian culture compared to the other two learner groups. The striking pattern in all cases is that the early English bilinguals were numerically intermediate between the late English bilinguals and Russian foreign language (RFL) learners. In addition, the groups differed reliably in verbal SAT scores (highest score=800): Russian foreign language learners, 718, early English bilinguals, 675, late English bilinguals, 585; all pair wise comparisons different, p < .05. The graphs in Figures 1-3 were designed to depict the "in-between" status of the early arrivals Fishers PLSD was used to identify whether the intermediate group (early arrivals) differed (p < .05) from either or both the other two groups, and these are noted in the titles of each panel in the figure. The scale for "Dating Russian partner" was not graphed, but the early and late arrivals patterned together and differed from the Russian Foreign Language learners. Figure 1 graphs the three emotional intensity scales. Scores for the late arrivals are between .5 and .6 (values of 0 indicating English is always more emotional, 1 indicates Russian always more emotional), indicating that respondents judged English and Russian to have roughly equal emotional intensity, with a slight tilt toward Russian for anger expressions. Early bilinguals patterned with the late arrivals, and indeed were not significantly different from them, but did differ from the Russian Foreign Language Learners. Figure 2 graphs the three frequencies of use scales. The early bilinguals resembled the late arrivals in frequently using Russian with family. Not surprisingly, the RFL learners reported minimal use of Russian with family. For use of Russian with friends, the early arrivals were intermediate between the other two groups, indicating low use of Russian with peers. They had slightly higher use of Russian in personal relations (.30), similar to the rate of the Russian Foreign Language Learners (.22). The RFL's report of use of Russian in personal relationships may seem higher than one would expect given the very low report of use of Russian with friends (compare the middle and right-hand panels in Figure 2). The questions underlying this scale asked about percent of time Russian is used in current personal relationships; a separate question asked about use of Russian in past relationships. Several of the RFL learners reported having dated a Russian speaker in the recent past (and one had a Russian-speaking spouse). Two respondents who had previously dated Russian speakers gave values of 11 25% for frequency of past use of Russian in relationships, while the respondent with a Russian-speaking spouse listed that Russian was used in 80% of his intimate relations. An additional question in the scale was: "When I have warm thoughts they are in.." (response format was the Russian-English scale described in Table 2). Two respondents choose "usually Russian" and one chose "Only Russian" (this was the certified Russian interpreter). These reports indicate that use of a foreign language for intimate relations can have a different status than use with friends, with substantial inner speech even in the absence of frequent in-person conversations. The possibility of using the foreign language for intimatepartner talk can bolster the personal significance of the language and contribute to maintenance, as theorists have noted (Pavlenko, 2005). While the habits of foreign language learners has inherent interest, our primary point here is simply that the early arrivals are sufficiently integrated into a nonRussian- speaking social life that their use of Russian in personal relationships resembles that of Russian foreign language learners and contrasts with late arrivals. Figure 3 graphs two of the three scales about language/cultural values. The early arrivals resembled the late arrivals in valuing Russian language culture equally to English and American culture, as indicated by the mean scale values of .5. In contrast, the early arrivals patterned with the Russian foreign language learners by identifying with American culture (values of .35 and .27, with 0 being "strongly American" and 1 indicating "strongly Russian"). Note that even the late arrivals were intermediate on the scale (.47) meaning they identified with both cultures. We also plotted an additional usage scale, the use of each language for thinking and problem soling. Here the early arrivals differed significantly from both of the other two groups, using English for problem solving more than late bilingual but less than Russian Foreign Language Learners. Graphing age effects Visual depiction of how proficiency, usage and emotionality vary as a function of age of arrival can provide powerful clues to life-span changes in bilingual speakers, even in a cross-sectional sample such as the one in the current study. Figure 4 shows that proficiency in L1-Russian rose sharply with increasing AoA, but by arrival age of 14 and later, respondents’ judged themselves to have Russian at native-speaker levels. Proficiency in L2 English was at native speaker levels for AoAr until age 8, and was reduced for AoArs 10-15, with no further decline for later arrivals, who judged their English to be close to "very good." This pattern of very good long term attainment for AoArs of older than 15 is consistent with results from grammatical judgment tests for European learners of L2-English (Birdsong & Molis, 2001; Jia et al., 2002). Age-trajectories for the two languages cross at age 11, the AoAr for respondents who perceived their languages to be roughly equal in proficiency. 12 Figure 5 depicts frequency of use of Russian with friends and with family. Use with family rises steeply as age-of-arrival varies from 0 to age 9, with a peak at "Always Russian" from age AoAr of age 9 and later. Reported frequency of use of Russian with friends is low for respondents who arrived in childhood and rises linearly, reaching high levels for the latest arrivals, consistent with findings about greater L1 use with peers for older immigrants in Jia and Aaronson (2003). Because the trilingual speakers (those who immigrated from the former Soviet Republics) reported lower use of Russian with family, their data is plotted separately (see open circles and squares). Respondents generally judged anger and positive expression to be similar in emotional strength, as shown by values in the midrange of the scale from 0 (English stronger) to 1 (Russian stronger). Perceived emotionality of anger in Russian increased slightly with AoArs ranging from birth to age 11, but thereafter was flat, while ratings for positive emotional expressions increased steadily across the entire range of AoAr in our sample. Respondents with younger AoArs judged anger expressions to feel stronger in Russian than English, but positive emotional expressions were judged to feel stronger in English. This pattern was absent for those who arrived later. Visually comparing age effects in Figure 4 (AoAr and proficiency) to Figure 6 (AoAr and emotionality) powerfully suggests the greater age-dependence of proficiency compared to emotionality. To show how closely emotionality tracks proficiency, Figure 6 also plots “Russian dominance”, which indicates Russian proficiency relative to English proficiency. This is highly correlated with AoAr (r=.78). In the next section, we report correlations and statistical significance for the variables plotted in Figures 4-6. Correlations between scales Table 3 presents correlations between proficiency variables, the family/friends frequency of use scales, emotional intensity scales, and learning history variables (age of arrival in the U.S. and length of stay; excluding Russian Foreign Language Learners). Valuing Russian language/culture, self-identity and dating scales had weaker correlations than these and are thus excluded for succinctness. The table is organized so that correlations with Russian proficiency can be read down the first column, and correlations with a second key variable, age of arrival in the U.S. (AoAr), can be read across the bottom row. Only significant correlations (p < .05) are displayed, and strong correlations (greater than r=.55) are outlined with boxes. The strongest correlations occurred between age of arrival and the two proficiency variables (r=.73 for Russian and r=-.78 for English). These correlations are similar in magnitude to those reported by 13 McElree et al. (2000) in their study of 26 native-Russian teens and young adults who immigrated to the U.S. between ages 7 to 22 (their correlations were r=.68 and r=-.70 for Russian and English respectively). This is consistent with prior findings that AoAr plays a strong role in ultimate attainment (Flege, Yeni-Komshian & Liu, 1999; Birdsong and Molis, 2001). Verbal SAT scores correlated with AoA at only r= -.35, with the relatively weak correlation likely due to the fact that SAT scores were not available for those older than age 20. Length of stay (LoS) (which is current age minus AoAr) was not as highly correlated with other variables as was age of arrival. The only case where LoS had a correlation as strong as AoAr was frequency of family usage of Russian. The later the age of immigration, and the shorter the length of stay, the more respondents used Russian with family. The other case of decoupled LoS and AoAr concerns the emotional intensity of lying in Russian. Perceived strength of lies in Russian was weaker with longer LoS, but did not vary with AoAr. But the "Expressing lies" measure was an exceptional one in our data set, because it was not strongly correlated with any of our variables. It was only weakly correlated with Russian proficiency, family Russian usage and length of stay. Later AoAr was associated with feeling emotions more strongly in Russian for positive emotions (r=.56) and anger (r=.42), but not for lying. These are stronger associations with age than were found in Dewaele's (2008) “I love you” study, presumably because age of acquisition is a stronger organizing factor for the immigrants studied in the current paper, compared to the heterogeneous, predominantly European and multilingual learners studied by Deawele (2008). Expressing positive emotions in Russian was more highly correlated with frequency of usage with friends (r=.61) than with family (r=.50), while expressing angry emotions is more highly correlated with usage with family (r=.58) than with friends (r=.50). The association of anger with family may be the reason that expressions of anger are not as tightly linked to AoAr (r=.42), probably because Russian usage with family is not tightly linked to AoA. In contrast, the associations of positive emotions with friends means that positive emotions are strongly associated with AoAr (r=.56). Russian immigrants to the U.S. who continue to use Russian with friends are those who report experiencing positive emotional expressions as feeling strong in Russian. In an exploratory analysis, we categorized all bilingual participants according to whether they had a Russian-speaking romantic partner or spouse (8 reported "yes", 36 "no"). Those who reported "yes" were older (mean age 30 vs. 22), had greater AoAr (18.7 vs. 9.7 years), reported lower English proficiency (5.8 vs. 6.5), and valued dating a Russian speaker (.75 vs. .48). They reported greater use of Russian in personal relations (.61 vs. .37), with friends (.62 vs. .35), and had higher scores on the positive emotional expressions scale (.63 vs. .40), all ps < .05. It is likely that having a Russian14 speaking partner or spouse plays a substantial role in the strong association between AoAr and judging positive emotional expressions to feel stronger in Russian than English. Multiple regression analyses Multiple regression to predict Russian and English proficiency was performed with predictors AoAr and frequency of using Russian with friends, the two highest correlates in Table 3. Because these two variables are highly correlated (r=.87), they can be understood as jointly explaining variance in proficiency. When entered into multiple regression, the variable "use with friends" emerged as the sole significant predictor for Russian proficiency, t(61)=3.5, p < .001, accounting for 60% of the variance. The sole significant factor of English proficiency was age of immigration to the U.S., t(61)=3.4, p < .005, accounting for 53% of the variance. Table 4 lists the correlations between each of the three emotionality variables and the other predictors separately for the early English and late English bilingual speakers. AoAr was not a significant correlate, presumably because of the restricted range that results by subdividing data according to A0Ar groups. Russian proficiency was the strongest predictor of strength of angry emotions in Russian for early arrivals, while low English proficiency was the strongest predictor for emotions feeling strong in Russian. To extract further generalizations about emotionality and our measures, factor analysis was employed to reduce scales to a single factor which could then be used in multiple regression. The lefthand panel of Table 5 lists the factor reductions for the predictor variables (e.g., "Usage" is the reduction from family usage and friends usage; "Proficiency" is the reduction from the English and Russian proficiency measures). The 3 emotionality scales (anger, positive emotion, and expressing lies) were reduced to a single factor, "emotionality," which accounted for 45% of the variance in the original 3 scales for the Late English bilinguals and 52% for the Early English bilinguals. A high percentage of variance captured by a single factor indicates that factor-reduction was performed on a set of correlated variables. For each of the two bilingual learner groups, stepwise multiple regression was run with the three predictors and the dependent variable of emotionality. As shown in the right-hand panel of Table 5, usage was the single best predictor of emotionality for the Late English group, while proficiency was the best predictor for the Early English group. 15 Discussion Our main question concerned how age of arrival influenced frequency of using L1-Russian with friends vs. family, relative strength of emotional expressiveness, and perceived proficiency in the two languages. Self-reported proficiency correlated strongly with age of arrival in the U.S. (AoAr), r=.73 for L1Russian, and r=-.78 for L2-English. Early arrivals (before age 10) rated themselves as having native speaker ability in L2-English, while later arrivals (after age 15) rated themselves as having native speaker abilities in L1-Russian; those with AoAr between 10 and 15 reported mixed dominance. Classic studies of age effects in SLA for immigrants have focused on how late immigration reduces long-term attainment of L2 (e.g., Birdsong & Molis, 2001; Newport & Johnson, 1989). A different result was found in the current study: early immigration reduced long-term proficiency in L1. The late arrivals maintained native-speaking abilities in L1-Russian, but also reported high ability in L2-English, although less than native-speaker ability. This pattern resembles prior research using grammaticality judgments to measure both L1 and L2 proficiency for immigrants to the U.S. from European countries, although the correlations using self-report measures are higher than has been found with grammaticality judgments (e.g., (Jia et al., 2002; McElree et al., 2000), Compared to perceived proficiency, emotionality was less strongly related to AoAr, although correlations were still substantial, r=.42 for anger and r=.56 for positive emotions. The greater agedependence of proficiency is apparent by comparing Figure 4 (AoAr and proficiency) to Figure 6 (AoAr and emotionality) reveals. Figure 6 also reveals that relative proficiency (i.e., Russian judged better than English) was highly correlated with AoAr. The judgment that emotional expressions felt stronger in Russian closely tracked the judgment that Russian proficiency was better than English proficiency. We thus conclude that judging emotions to feel stronger in L1 than L2 is tightly linked to the relative proficiency of the two languages. Interesting patterns emerged in the association of anger with family usage (r=.58), positive emotions with friend usage (r=.61), and how usage frequencies were linked to arrival age. To explain this, we will first discuss usage frequency patterns. As shown in Figure 3 and Table 3, frequency of Russian usage with friends increased with arrival age (r=.87) more strongly than did frequency of Russian usage with family (r=.61) Most respondents generally reported high uses of Russian with family (usually meaning their parents), but only late arrivals sad they frequency used Russian with friends. Family usage was generally high across all but the youngest arrivals, presumably because it is learners' parents who control family Russian usage. Most of the respondents indicated that their parents preferred using L1 with their 16 children. We infer that parents' preference established this as the household custom, regardless of whether children were age 8 or age 20. The exception to high Russian use with family occurred for respondents who were born in the U.S. or arrived before age 6 (see Figure 5). The average frequency of Russian use for these young arrivers was only 61%, compared to 86% for AoAr age 6 and older. Children born in the U.S. or arriving early are introduced into the American school system and grow up with English as their dominant language. They may thus Americanize the discourse in their households, leading to mixed language use within the home. An alternative explanation is that more English was spoken in the home because the parents of early arrivals were themselves younger and thus open to pursuing high L2 proficiency. Regardless of the explanation, the reported usage patterns support Jia and Aaronson's (2003) conclusion that early arrivers grow up in a richer L2 and poorer L1 environment than do later arrivers. In contrast to the pattern of consistently high family usage for arrivals greater than age 10, frequency of using L1-Russian with friends increased linearly with AoAr (see Figures 2 and 5). The striking correlation between friends usage and AoAr (r=.87) has several provocative implications. Using Russian with friends correlated with feeling that positive emotions were strong in Russian, which also correlated with AoAr. For later arrivals, high use of L1 with friends could maintain high proficiency, protecting against the L1 proficiency decline observed for early arrivals. Feeling that positive emotions are strong in L2 may then be a motivation to seek out L2-speaking friends, which then increases proficiency, resulting in a nexus of reinforcing factors to maintain high L2 proficiency. Confirming the proposal that friend-usage is an L1 proficiency protector for immigrants awaits future research. However, the current data does allow us to extend the findings of Jia and Aaronson (2003) to an older age group: older immigrants have both more motivation and greater ability to locate and maintain friendships with L1 speakers, and the older they are at time of immigration, the greater their motivation and friend-finding resources become. One of the reasons for this is that at immigration teenagers and young adults will usually have entrenched and highly proficient L1 abilities (see discussion in MacWhinney, 2005). Their awareness of their poor L2 speech can be the motivation to seek out L1-speaking friends. Their greater experience of life in the country of origin creates nostalgia for L1 culture and themes, which is further motivation to seek out L1-speaking friends, with friends then influencing judgments that positive emotional expressions feel strong in Russian. Support for this in the current data was early arrivers' greater identification with American culture compared to later arrivers. 17 How arrival age can dictate L1 proficiency and from there, usage patterns We argue that AoAr reliably influences L2 proficiency because of the way it determines L1 proficiency, and L1 proficiency determines family and friend usage patterns, as explained above. We propose the following broad (and approximate) phases for immigrants to the U.S. AoAr Early childhood (birth to age 5): These very early arrivals rapidly become dominant in L2English and this alters household dynamics, leading parents and siblings to use more L2, resulting in a richer L2 and poorer L1 environment in the home. (Many early arrivals said parents only spoke Russian in the home but that they responded to parents’ Russian with English and use English with siblings.) Lacking an environment featuring exclusive L1 use, and experiencing rapid cognitive and social development in an L2-English world, these individuals become English native speakers and their ability in their parents' language ranges from limited to a maximum ability of only "good." AoAr Middle childhood (age 6 to 9): L1 is firmly the language of the home, but L2-English is the language used with friends. L2 becomes the dominant language and because of the restrictive context of use of L1, L1 is usually not attained to native-speaker levels AoAr late childhood (age 10-15). For this group, L1 is not just the language of home, but also a language that can be used with L1-speaking friends in the new country and for maintaining connection to the country of origin. But L2 will be the language of schooling in at least some of the teen years, and will also rapidly become the main language of peer socialization. These individuals have the best opportunity to feel they are native speakers in both of their languages. In adulthood they will have high proficiency in both of their languages and will regard both languages as having strong emotional resonances, because both languages have an emotional context of use: the emotional context of use with family for L1, and the emotional context of peer friendships for L2. AoAr young adulthood (age 16-30): L1 will remain the language used with family members and may remain the dominant language for at least a decade post-immigration, or even for life (see Bahrick, 1992). It will also frequently be the language with the greatest emotional resonances. Individuals will seek out chances to socialize with other L1 speakers, and the number of L2-speaking friend will diminish with greater age of arrival. Depending on idiosyncratic personal factors, such as an L2 or L1 speaking spouse/partner and proportion of L1 speaking friends, these individuals will attain greater or lesser proficiency in L2. 1 1 Of course, these idiosyncratic factors could exist in all 3 groups. Younger immigrants may marry an L1 speaking spouse or may immigrate to an L1 rich community, 18 Factors influencing perceived emotional resonances The thesis that the first language is experienced as more emotional is powerful, because it fits with contemporary neuroscience ideas about the primacy of events from early childhood and the crucial role of brain development. In the current study, perceived emotionality did vary consistently by AoAr, but, we argue, was not caused or influenced by brain maturation. Instead, emotionality tracked usage patterns and proficiency, which are themselves powerfully sensitive to AoAr. The current study is thus consistent with findings of Dewaele (2004, 2006, 2008) and Pavlenko (2005) which found that usage factors, proficiency and contexts of acquisition were more important than age of acquisition for judgments of strength of emotional expressiveness. This is also consistent with the "emotional context of learning" hypothesis, which proposes that language comes to be experienced as emotional when it is used and learned in emotional contexts (Harris et al., 2006). Perceived emotionality differed from relative proficiency in the case of judgments about angry vs. positive expressions. Angry expressions in L1-Russian felt stronger than positive expressions for young arrivals, but these judgments reversed for older arrivals. The following three broad age periods are offered as a summary of the patterns in Figure 6. AoAr Childhood (birth to 9): Positive emotional expressions were perceived to be weaker in L1Russian than L2-English. Angry expressions felt stronger in L1 than did the weakly felt positive expressions. AoAr Late childhood and teen years (10-20): Tracking relative proficiency, these respondents experienced Russian angry and positive expressions to be more powerful than did the younger arrivals. Like the younger arrivals, angry expressions felt stronger in Russian than did positive expressions. AoAr 22 years and older: Positive emotional expressions felt substantially stronger in Russian than English, such that positive expressions were felt by this group to feel as strong in Russian as angry expressions. Why would family L1 language usage be correlated with experiencing angry expressions as feeling strong in L1? We infer that using L1-Russian in the limited context of family promotes easy retrieval of emotionally charged angry expressions but not positive expressions. In contrast, using Russian extensively with friends promotes easy retrieval of emotionally charged positive expressions. Social interactions among family vs. friends differs: the former is marked by hierarchical, authority-based relations, and the latter with egalitarian, voluntary relationships. Hierarchal relations with parents include the usage of childhood reprimands, found to elicit high skin conductance responses in a first language but not a second language (Harris et al, 2003, 2006). Skin conductance responses to positive 19 expressions were found to be equally strong in a first and second language (Staroselsky et al., in preparation). Different factors influenced emotionality for early vs. later arrivals. As revealed by multiple regression following factor reduction, for later arrivals, usage patterns most strongly influenced the judgment that Russian felt more emotional than English. For early arrivals, the key variable was proficiency. Both effects may reflect the necessity of having good proficiency in order to experience strong emotions in a language. The early arrivals had highly variable proficiency in L1-Russian, and the ones who had good proficiency reported feeling that Russian was more emotional than English. For the older arrivals, their proficiency in both languages was generally high, meaning restricted range for the proficiency variable. Once proficiency is high, then frequency of use is what determines perceived emotionality. Age of acquisition as an organizing variable In Johnson & Newport's (1989) seminal study of age effects in immigrants' second language acquisition, the authors were careful to document that it was age of arrival in the host country rather than length of stay which was most predictive of ultimate proficiency. Subsequent publications demonstrated the irrelevance of length of stay for proficiency, thus appearing to bolster maturational accounts, including the idea of a critical/sensitive period for language acquisition. Twenty years later, we can speculate about what kind of ideas about language learning would result in this inference. It suggests binary thinking of the following sort: LOS as the strongest predictor of proficiency is consistent with language learning occurring via immersion in a speech community. Since LOS isn't the strongest predictor, then language learning doesn't simply happen as an automatic result of exposure. If exposure isn’t sufficient, this means that language development is critically dependent on a special maturational period in development. The two alternate concepts of language learning hinted at above resemble the innate language acquisition device developed by Chomsky (1965) and the behaviorist view of gradual accruement with practice (Skinner, 1957). If these are the only two possibilities, then the either-or reasoning italicized above is logical. However, from a contemporary perspective, learning a second language after early childhood is understood to be influenced by environmental demands and individual motivation (Schumann, 1978, 1997). Contemporary theorists now make different inferences when confronted with empirical data that LoS is mostly irrelevant and that AoAr is crucial. Instead of concluding that age 20 effects must result from brain-based maturational changes, researchers now see that LoS is mostly irrelevant precisely because we understand that passive exposure is irrelevant to language learning. Contemporary researchers can understand that AoAr is so potent a predictor of ultimate proficiency because environment L1/L2 richness and language-learning motivation are tightly determined by agerelated cognition and social skills (Moyer, 2004; Jia et al., 2002) The current data demonstrate that frequency and contexts of L1 and L2 use are important determinants of long term proficiency. We argued that AoAr has its strong effects on ultimate outcome because age of immigration powerfully organizes many variables related to frequency and contexts of use, from the fragility of the L1 for young arrivals to the need to maintain self-identity via socialization with L1 speakers among older arrivals. Theorists have long discounted frequency of use explanations because of the phenomena of fossilization, meaning lack of improvement despite years of exposure (see chapters in Han & Odlin, 2005). But adults whose L2 fossilizes are generally also older arrivals who use L2 in restricted contexts, continuing to use L1 for their emotionally meaningful conversations with friends and family. Future research The bicultural, in-between status of early L2 learners As part of examining age effects on usage patterns, emotionality and proficiency, we assembled a profile of early vs. late learners by asking how early arrivals to the U.S. resembled late arrivals and how they resembled their American peers who grew up speaking only English (the Russian foreign language learners). Early arrivals resembled late arrivals in judging both English and Russian to elicit strong emotions. Early arrivals also resembled later arrivals in their frequent use of Russian with family, and with the extent to which they valued Russian language and culture. However, early arrivals used Russian less with friends than did later arrivals. Early arrivals differed from later arrivals in the degree to which they identified with American culture, and on this variable did not differ significantly from nonimmigrant Americans (Russian foreign language learners). What is useful in this portrayal of the "in-between" state of the early L2 learners is that they are not identically in-between on every measure. For issues related to family, they most resembled the later arrivals. For issues outside of the family, such as dating and cultural identification, they most resembled their non-immigrant English speaking peers. We suspect that the details of this pattern may be specific to the population we sampled: Russian speaking immigrants, mostly college students or university affiliated. It would be useful to compare the observed "in-between" state of our early arrivals with those 21 in other immigrant groups who differ in variables such as integration with the dominant culture and emphasis on L1 in the home. Perceived emotional strength of lying Perceived strength of lying in Russian correlated with frequency of using Russian with family (r=.48), Russian proficiency (r=.34) and length of stay in the U.S. (r=.39). The correlation with Russian family usage is consistent with our analysis of why strength of anger judgments correlated with Russian family usage: lying, like expressing anger, may bring to mind autobiographical examples of lying to parents in childhood and may thus be related to parental authority relationships. In designing the current questionnaire, we assumed that perceived strength of lying in L1-Russian would resemble our other two categories of emotional language. However, lying correlated with fewer learning history variables than did anger and positive emotions. This suggests that judgments about lying may reflect individual opinions about when an dhow lying is most emotionally distressing. For many respondents, this may be the context of lying to family members, as mentioned above. For others, lying to peers or romantic partners may be the most distressing. Judgments about lying may also differ from other emotional expressions because the emotion activated by an angry statement is determined by statement content, while the emotion associated with lying may come less from statement content than from personal discomfort about lying. During interviews, Spanish-English bilinguals in the U.S. and Turks learning English as a foreign language in Turkey discussed whether they preferred to lie in L1 or L2 (Caldwell-Harris, Sanchez, Ventura, Angun, & Aycicegi-Dinn, 2007; Caldwell-Harris & Aycicegi-Dinn, 2008). The majority of respondents preferred L1, their more proficient language, in order to avoid the dual-task load of monitoring their L2 production while navigating their lie. A significant minority preferred their less proficient language, saying that they wouldn't feel the lie as strongly and thus their lie wouldn't be as discernable. If participants had nearly equal proficiency in both languages, then their responses were more predictable: they preferred to lie in the language they experienced as less emotional. Given the variability in language preferences for lying, variability in judgments of emotional strength of lying is less surprising. Future work on judgments of the emotionality of lying (and indeed, all emotional expressions) could benefit from specifying details of the situation and the conversation partners involved. 22 References Altarriba, J. and Santiago-Rivera, A.L. (1994). Current perspectives on using linguistic and cultural factors in counseling the Hispanic client. Professional Psychology: Research and Practice, 25, 388397. Aycicegi-Dinn, A., & Caldwell-Harris, C.L. (2009). Emotion memory effects in bilingual speakers: A levels-of-processing approach. Bilingualism: Language and Cognition. Bahrick, H.P., Hall, L.E., Goggin, J.P., Bahrick, L.E., and Berger, S.A. (1994). Fifty years of language maintenance and language dominance in bilingual Hispanic immigrants. Journal of Experimental Psychology: General, 123, 264-283. Bialystok, E. & Miller, B. (1999). The problem of age in second language acquisition: Influences from language, task, and structure. Bilingualism: Language and Cognition, 2, 127-145. Bialystok, E., & Hakuta, K. (1998). Confounded age: Linguistic and cognitive factors in age differences for second language acquisition. In D. P. Birdsong (Ed.), New perspectives on the critical period hypothesis for second language acquisition. Hillsdale, NJ: Erlbaum. Birdsong, D. (2004). Second language acquisition and ultimate attainment. In A. Davies and C. Elder (Eds.), The handbook of applied linguistics. Malden, MA: Blackwell. Birdsong, D. (1992). Ultimate attainment in second language acquisition. Language, 68, 706-755. Birdsong, D. and Molis, M. (2001). On the evidence for maturational constraints in second language acquisition. Journal of Memory and Language, 44, 235-249. Bloom, L., & Beckwith, R. (1989). Talking with feeling: Integrating affective and linguistic expression in early language development. Cognition and Emotion, 3, 315-342. Bond, M.H. and Lai, T. (1986). Embarrassment and code-switching into a second language. Journal of Social Psychology, 126, 179-186. Bongaerts, T., Van Summeren, C., Planken, B., & Schils, E. (1997). Age and ultimate attainment in the pronunciation of a foreign language. Studies in Second Language Acquisition, 19, 447-465. Caldwell-Harris, C.L., & Aycicegi-Dinn, A. (2008). Emotion and lying in a non-native language. International Journal of Psychophysiology. 23 Caldwell-Harris, C.L., Sanchez, N., Ventura, B., Angun, C., Aycicegi-Dinn, A. (2007). Preferring to lie in L1 vs. L2: Is emotionality or proficiency more important? Annual Meeting of the Association for Psychological Science, Washington, D.C. Chiswick, B. & Miller, P. (2008). A test of the critical period hypothesis in language learning. Journal of Multilingual and Multicultural Development, 29, 16-29. Chiswick, B., Lee, Y.L., & Miller, P. (2005). Family matters: The role of the family in immigrant's destination language skills. Journal of Population Economics, 18, 631-647. Chomsky, Noam (1965). Aspects of the Theory of Syntax. MIT Press. Dewaele, J-M. (2004). The emotional force of swearwords and taboo words in the speech of multilinguals. Journal of Multilingual and Multicultural Development, 25, 204-222. Dewaele, J-M. (2008). The emotional weight of I love you in multilinguals' languages. Journal of Pragmatics, 40, 1753-1780. Dewaele, J-M., Pavlenko, Aneta, (2001). Web questionnaire Bilingualism and Emotions. University of London. Dewaele, Jean-Marc, (2006). Expressing anger in multiple languages. In A. Pavlenko (Ed.), Bilingual minds: Emotional experience, expression, and representation, (pp. 118-151). Clevedon: Multilingual Matters. Elman, J., Bates, E., Johnson, M., Karmiloff-Smith, A., Parisi, D. and Plunkett, K. (1996). Rethinking Innateness: A Connectionist Perspective on Development. Cambridge, Mass.: M.I.T. Press. Flege, J. E., Yeni-Komshian, G. H., & Liu, S. (1999). Age constraints on second-language acquisition. Journal of Memory and Language, 41, 78-104. Gonzalez-Reigosa, F. (1976). The anxiety arousing effect of taboo words in bilinguals. In C.D. Spielberger and R. Diaz-Guerrero (Eds.), Cross-cultural Anxiety, (pp. 89-105). Washington, DC: Hemisphere. Hakuta, K., Bialystok, E., & Wiley, E. (2003). Critical evidence: A test of the critical-period hypothesis for second-language acquisition. Psychological Science, 14, 31-38. Han, Z., & Odlin, T. (2005, Eds.), Perspectives on Fossilization. Clevedon, UK: Multilingual Matters. Harris, C.L. (2004). Bilingual speakers in the lab: Psychophysiological measures of emotional reactivity. Journal of Multilingual and Multicultural Development, 25, 223-247. 24 Harris, C.L., Aycicegi, A. and Gleason, J.B. (2003). Taboo words and reprimands elicit greater autonomic reactivity in a first than in a second language. Applied Psycholinguistics, 4, 561-578. Harris, C.L., Gleason, J.B.,& Aycicegi, A. (2006). When is a first language more emotional? Psychophysiological evidence from bilingual speakers. In A. Pavlenko (Ed.), Bilingual minds: Emotional experience, expression, and representation. Clevedon, United Kingdom: Multilingual Matters. Hernandez, A.E., & Li, P. (2007). Age of acquisition: Its neural and computational mechanisms. Psychological Bulletin, 133, 638-650. Jia, G., & Aaronson, D. (2003). A longitudinal study of Chinese children and adolescents learning English in the U.S. Applied Psycholinguistics, 24, 131-161. Jia, G., Aaronson, D., Wu, Y. (2002). Long-term language attainment of bilingual immigrants: Predictive variables and language group differences. Applied Psycholinguistics, 23, 599-621. Johnson, J.S. & Newport, E.L. (1989). Critical period effects in second language learning: The influence of maturational state on the acquisition or English as a second language. Cognitive Psychology, 21, 60- 99. Kellerman, E. (1995). Age before beauty: Johnson and Newport revisited. In L. Eubank., L. Selinker and M.S. Smith, (Eds.), The Current State of Interlanguage: Studies in Honor of William E. Rutherford. Amsterdam: John Benjamins. Kroll, J., Michael, E., Tokowicz, N. and Dufour R. (2002). The development of lexical fluency in a second language. Second Language Research, 18, 141-175. Lardiere, D. (2006). Ultimate Attainment in Second Language Acquisition. Mahwah, NJ: Erlbaum. Lenneberg, E. H. (1967). Biological Foundations of Language. Wiley. MacWhinney, B. (2005). Emergent Fossilization. In Z. Han and T. Odlin (Eds.), Perspectives on Fossilization. Clevedon, UK: Multilingual Matters. Marian, V., & Kaushanskaya, M. (2004). Language-mediated self-construal and emotion in bicultural bilinguals. Journal of Memory and Language, 51, 190-201. Marian, V., Blumenfeld, H.K., & Kaushanskaya, M. (2007). The Language Experience and Proficiency Questionnaire (LEAP-Q): Assessing language profiles in bilinguals and multilinguals. Journal of Speech Language and Hearing Research, 50, 940-967. 25 McDonald, J. L. (2000). Grammaticality judgments in a second language: Influences of age of acquisition and native language. Applied Psycholinguistics, 21, 395-423. McElree, B., Jia, G. X., & Litvak, A. (2000). The time-course of conceptual processing in three bilingual populations. Journal of Memory and Language, 42, 229-254. Moyer, A. (2004). Age, accent and experience in second language aquisition: An integrated approach to critical period inquiry. Clevedon, United Kingdom: Multilingual Matters. Pallier, C. (2007). Critical periods in language acquisition and language attrition. In Barbara K√∂pke, Monika S. Schmid, Merel Keijzer, and Susan Doste (Eds.), Language Attrition: Theoretical perspectives. Amsterdam: John Benjamins. Patkowski, M. (1980). The sensitive period for the acquisition of syntax in a second language. Language Learning, 30, 449-472. Patkowski, M. S. (1990). Age and accent in a second lan- guage: A reply to James Emil Flege. Applied Linguistics, 11, 73-89. Pavlenko, A. (2002). Bilingualism and emotions. Multilingua, 21, 45-78. Pavlenko, A. (2005). Emotions and Multilingualism. Cambridge, MA: Cambridge University Press. Pavlenko, A. (1998). Second language learning by adults: Testimonies of bilingual writers. Issues in Applied Linguistics, 9, 319. Schrauf, R.W. & Rubin, D.C. (1998). Bilingual autobiographical memory in older adult immigrants: A test of cognitive explanations of the reminiscence bump and the linguistic encoding of memories. Journal of Memory and Language, 39, 437-457. Schrauf, R.W. (2000). Bilingual autobiographical memory: Experimental studies and clinical cases. Culture and Psychology, 6, 387-417. Schrauf, R.W., Durazo-Arvizu, R. (2006). Bilingual autobiographical memory and emotion: theory and methods. In A. Pavlenko (Ed.), Bilingual minds: Emotional experience, expression, and representation, (pp. 284-311). Clevedon: Multilingual Matters. Schumann, J. (1978). Affective factors and the problem of age in second language acquisition. Language Learning, 25, 209-235. Schumann, J.H. (1997). The neurobiology of affect in language. Malden, MA: Blackwell. Skinner, B.F. (1957). Verbal Behavior. Acton, MA: Copley Publishing Group. 26 Smith, L.B., Thelen, E. (2003). Development as a dynamic system. Trends in Cognitive Sciences, 7, 343-348. Snow, C. E. (1983). Age differences in second language acquisition: Research findings and folk psychology. In K. M. Bailey, M. H. Long, & S. Peck (Eds.), Second language acquisition studies, (pp. 141-150). Rowley, MA: Newbury House. Staroselsky, M., Caldwell-Harris, C.L., & Vasilyeva, N. (In preparation). Psychophysiological studies of emotional arousal to bilingual speakers' first and second languages. 27 Table 1. Demographic and learning history of the three learner groups Later Arrival to the U.S. (N=21) Gender Percent born in Russia or other USSR republics Dominant language (%) Age (range) Age at arrival in the US (range) Length of stay (range) Age of Exposure to Rus (range) Russian Prof. (1-7 scale)b (sd) English Prof. (1-7 scale) (sd) 66% 60% Russia 45% Other Rep. 10 % English 40% Both 50% Russian 27.7 (18-56) 19.4 (10-40) 8.8 (1-26) .5 (0-6)a 6.7 (.4) 6.1 (.6) Early Bilinguals (N=23) 56% 42% Russia 39% Other Rep. 46% English 54% Both 0% Russian 20.0 (17-24) 4.1 (0-9) 14.2 (9-21) 0 5.0 (1.3) 6.6 (.7) Russian L2 in Adulthood (N=20) 45% 0% 100% English 27.5 (18-64) 0 Not applicable 20.9 (14-59) 4.4 (1.0) 7.0 (.2) Tables Notes. aTwo participants learned a first language (Mongolian or Azerbaijan) before learning Russian at age 6. bAverage of ratings for proficiency in speaking, listening comprehension, reading and writing. 28 Table 2. Scales in the three questionnaire domains with sample items and response formats Subscale names N Sample Items Cronbach Alphaa Emotional intensity (dependent variables) Angry Emotions When I have angry thoughts they are in.... (Rus-Eng scale) Expressing Lies I feel guilty when I am telling a lie in … (Rus-Eng scale) Positive Emotions I experience strong emotion when expressing intimacies in... 6 3 4 .86 .90 .83 4 4 4 1 1 .94 .80 .80 --- 7 .84 5 3 .75 .76 Frequency and context of use (predictor variables) Use with family Use with friends Personal relations Personal Problems Thinking Percentage of time Russianb spoken with family (%) Percentage of time Englishb spoken with friends (%) Percentage of time Russianb spoken in relationships (%) When analyzing a personal problem I use...(Rus-Eng scale) When thinking about an academic topic I use...(Rus-Eng scale) (Problem Solving and Thinking combined into one scale) Self-identification and valuing Russian language/culture Self-Identification The culture I feel most in tune with is... (Rus-Am scale) Valuing Russian It’s important to me that my partner appreciates the Russian Lang, Culture culture…(5 point scale, Strongly disagree to Strongly agree) Dating Russian I prefer to date speakers of… (Rus-Eng scale) Partner Format of response scales Russian-English (Rus-Eng) scale Only Eng. _ Usually Eng._ Both languages equally_ Usually Russian_ Only Russian Russian-American (Rus-Am) scale Strongly Am. _ Mainly Am._ Both cultural equally _ Usually Russian_ Only Russian _ Scoring: 0 .25 .5 .75 1 Table notes. aChronbach alpha calculated on data from all three learner groups. bSame question asked about English, but English percentages were subtracted from 100 so these items could be averaged with items asking about Russian. 29 Table 3: Correlations between measures Proficiency Usage RusProf EngProf Comf Family Friend Emotionality Problem solving Anger Positive Lies Length of Stay EngProf -0.58 * Comfort 0.51 -0.41 * Family 0.59 -0.47 0.38 * Friend 0.78 -0.74 0.43 0.68 * Problem 0.61 -0.67 0.45 0.51 0.65 * Anger 0.51 -0.34 0.27 0.58 0.50 0.40 * Positive 0.49 -0.66 0.38 0.50 0.61 0.63 0.49 * Lies 0.34 -0.12 0.20 0.48 0.23 0.27 0.29 0.23 * LoS -0.55 0.47 -0.17 -0.63 -0.56 -0.48 -0.47 -0.28 -0.39 * AoA 0.73 -0.78 0.40 0.61 0.87 0.63 0.42 0.56 0.21 Family Friend Problem solving Anger Positive Lies -0.55 Length of Stay RusProf EngProf Comf Table Note. Correlations performed after excluding Russian Foreign language learners 30 Table 4: Anger, positive emotions and lying for the two bilingual groups Expressing Lies Positive emotions Late English Bilinguals Angry emotions Expressing Lies Positive emotions Angry emotions Early English Bilinguals Russian Proficiency 0.61 0.43 0.19 0.52 0.12 0.18 Eng Proficiency -0.44 -0.52 0.03 0.23 -0.62 -0.07 Comfort in Russian 0.45 0.15 0.33 0.05 0.43 0.32* Family usage freq 0.42 0.24 0.38 0.42 0.33 -0.01 Friend usage freq 0.29 0.2 0.27 0.22 0.54 -0.16 Problem solving 0.47 0.5 -0.28 -0.15 0.49 0.18 Length of Stay -0.28 -0.16 -0.41 0.12 0.11 0.03 Age of Arrival 0.25 0.11 0.37* -0.08 0.39 -0.19 Table notes. *n.s. 31 Table 5: Factor reductions and multiple regression Variance obtained following reduction to 1 factor Dependent Predictors Late English Early English Multiple Regression Results Usage AoA, LoS Proficiency Emotionality Signif. Predictor 81% 59% 51% 45% Usage 52% Profici -ency 64% 93% 63% R2 Beta F 0.29 0.57 F(1,19)=9.0 0.47 0.70 F(1,21)=20.0 32 Figure 1. Relative strength of emotional expressions in L1-Russian vs. L2-English are plotted for the three groups of learners. Strength is measured from 0 (English stronger) to 1 (Russian stronger) with .5 indicating that emotional expressions feel equally strong in the two languages. 33 Figure 2. Frequency of using L1-Russian vs. L2-English is plotted for the three learner groups. Usage frequency is measured as percent of total language use, such that 0.5 indicates that the two language are used equally in the three situations of family, friends and personal/intimate relationships. 34 Figure 3. The three learner groups are plotted according to how they valued Russian language, use of Russian vs. English for problem solving, and extent of identification with American culture. 35 Figure 4. Self-rated proficiency for Russian and English is plotted as a function of age of arrival in the U.S. Each depicted data point is the average for 3-4 respondents. Trilinguals are not plotted separately because their proficiency ratings did not differ from bilinguals. See Table 3 for correlation coefficients. Proficency (7=Native Speaker) 7.0 6.0 5.0 4.0 Russian Proficiency English Proficiency 3.0 0 5 10 15 20 25 30 Age of Arrival in the U.S. 36 Figure 5. Frequency of using Russian with family (circles) and friends (squares) is plotted separately for bilinguals (filled circles and squares) and trilinguals (open circles and squares; each trilingual data point is the average of 1-2 respondents). Trilinguals immigrated between 1999-2006 from former Soviet republics outside of Russia and thus have shorter length of stay in the U.S. and reported less use of Russian with family. See Table 3 for correlation coefficients. 100% Use of Russian with…(%) 90% 80% 70% 60% 50% 40% 30% Use Use Use Use 20% 10% with Family with Friends Family (Tri) Friends (Tri) 0% 0 5 10 15 20 25 30 Age of Arrival in the U.S. 37 Figure 6. Perceived emotional intensity of Russian compared to English is plotted on a scale ranging from 0 (English stronger) to 1 (Russian stronger). Each bilingual data point is the average for 3-4 respondents. Trilinguals are separately plotted with each data point corresponding to 1-2 participants. Trend line is plotted through “Russian dominance values” indicated by . These values were obtained by dividing Russian proficiency by the sum of English and Russian proficiency; equal proficiencies thus have a relative proficiency rating of .5, and ratings of higher Russian than English veer towards 1, while higher English than Russian veers towards 0. See Table 3 for correlation coefficients. Strength of emotions in Russian 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 5 10 15 20 Age of Arrival in the U.S. 25 30 38 Acknowledgements The authors thank Vicky Rukovets, Elizabeth LoTempio, Victoria Choate, Deborah Williams, Noreen Cipriano and Ebi K. Poweigha for research assistance. 39