Abstract
Nonnative English speakers are a growing population in the United States and United Kingdom, yet nothing is known about how second-language interviewing affects the assessment (or classification) of internal working models (IWM) of attachment. We suggest that these assessments may be subject to deleterious effects when conducted in an individual’s second language, because they (and the IWM structure itself) are linguistically bound, emotionally laden, and complicated by cultural loading. In this review, research regarding how the language of assessment broadly affects language production, interview content, emotional expression, and interviewee behavior is reviewed. We close with areas for future research and a table of existing attachment assessments by language in an effort to spur cross-cultural and bilingual research in this area. This review can therefore serve as a resource for clinicians wishing to assess attachment with linguistically diverse clients and researchers wishing to address the basic question of this review—how is the assessment of IWMs affected when conducted in an individual’s nonnative tongue?
The importance of assessing adolescents’ and adults’ internal working models (IWMs) of attachment has been clearly reviewed in other papers tying secure models to reduced psychopathology, improved social relatedness, and improved well-being and quality of life (adolescent, for example, Brown & Wright, 2003; Lacasa, Mitjavila, Ochoa, & Balluerka, 2015; Miljkovitch, Pierrehumbert, Karmaniola, Bader, & Halfon, 2005; Venta, Shmueli-Goetz, & Sharp, 2014; adult, for example, Barone, 2003; Berlin, Cassidy, & Appleyard, 2008; Deklyen & Greenberg, 2008; Dykas & Cassidy, 2011; Fournier, Brassard, & Shaver, 2011; Mikulincer & Shaver, 2007; Schindler et al., 2005; Shi, Bureau, Easterbrooks, Zhao, & Lyons-Ruth, 2012). Critically, no empirical evaluation has examined how the assessment of IWMs, which is typically undertaken through semistructured interviews (e.g., Adult Attachment Interview [AAI], George, Kaplan, & Main, 1985; Child Attachment Interview [CAI], Target, Fonagy, Shmueli-Goetz, Data, & Schneider, 2007) or script- and story-based tasks (see Dykas, Woodhouse, Cassidy, & Waters, 2006; Waters & Waters, 2006) developed and conducted in English is impacted when conducted in an individual’s nonnative language. This ought to be a growing concern for clinicians and researchers, in light of census data showing large numbers of nonnative English speakers in the United States, where approximately 21% of people 5 years old and older speak a language other than English at home (U.S. Census Bureau, 2015), and the United Kingdom, where 8% endorse a main language other than English (U.K. Office for National Statistics, 2011). Moreover, individuals for whom English is a second language may have had early (or continued) exposure to a different culture (Helms, 1992), adding complexity to the process of identifying adequate assessment tools (Dana, 2000). Finally, nonnative English speakers (including ethnic minorities) are overrepresented in clinical settings, where the assessment of attachment may be most relevant.
The aim of this review is to present empirical research and theoretical models with the overarching goal of suggesting that the assessment of IWMs of attachment via interview and script- and story-based methods, dominant practices with adolescents and adults, may be particularly subject to deleterious effects if conducted in an individual’s nonnative tongue because both the IWM and measurement are linguistically bound, emotionally laden, and complicated by cultural loading. In light of the vastly limited empirical research conducted on attachment in bilingual individuals, we close with several identified areas for future research as well as a table that presents existing assessments in various languages, in an effort to support future research and clinical application of attachment measures to nonnative English speakers. Throughout this review, we seek to discuss the conceptual issues of assessing attachment in a second language, moving beyond basic psychometric and pragmatic issues related to utilizing appropriately translated and normed measures (e.g., Malley-Morrison, Soon You, & Mills, 2000; Wang & Mallinckrodt, 2006).
The present review does not discuss the assessment of IWMs of attachment during infancy or childhood, when assessment is primarily based on observable behavior rather than language (Solomon & George, 2008). Indeed, the most widely studied IWM assessments for use during infancy and childhood include coding observations from separation and reunion paradigms (e.g., Strange Situation Paradigm, Cassidy-Marvin Assessment of Attachment in Preschoolers, Main-Cassidy Attachment Classification for Kindergarten-Age Children) or naturalistic behavior in the child’s home (Solomon & George, 2008). These observational tools have been successfully used across cultural and linguistic groups (see van Ijzendoorn & Sagi-Schwartz, 2008) and that research is outside the scope of the present review. A smaller literature assesses IWMs in children using projective and semiprojective techniques including picture responses, doll play, and family drawings. Validation of these methods is incomplete (Solomon & George, 2008) and, thus, will not be reviewed herein. As that literature emerges, however, research should investigate the effects of assessment in a child’s nonnative tongue, as the coding of projective assessments may be subject to linguistic and cultural bias (Vega, Lasser, & Afifi, 2016).
IWMs of Attachment are Linguistically Bound
At present, our understanding of IWMs of attachment is linguistically bound for two reasons. First, attachment interviews, the gold-standard methodology for assessing IWMs of attachment after early childhood, are linguistically complex and demanding. Likewise, script- and story-based procedures, which have received increasing attention over the last decade, are inherently linguistic. Second, IWMs themselves emerge from language, and attachment classifications have demonstrated associations to linguistic abilities. This section will focus on reviewing literature relevant to these two domains. A final section will describe the possible effects of IWM assessments conducted in a nonnative language.
Attachment Interviews are Linguistically Complex and Demanding
Attachment interviews are linguistically burdensome by design. Indeed, in creating the AAI, Main, Kaplan, and Cassidy (1985) deliberately moved away from the behaviorally based assessment of attachment that dominates infancy and early childhood to the “level of representation.” The linguistic demand of this interview is evident simply in its structure and length; the respondent is asked to provide five adjectives and examples that characterize his or her early relationships with caregivers as well as describe times that they were upset and needed help. This occurs during an entirely spoken interview that ranges in length between 45 min and 1½ hr (Main et al., 1985). In addition to the amount of verbal content requested, linguistic and cognitive complexity is sought through these prompts (among others) and forms the basis of the AAI’s coding. Indeed, the coder focuses on how individuals talk about their attachment experiences, using the clarity, coherence, and consistency of the narrative as clues about the nature of the IWM (Bretherton, 1995; Main et al., 1985). Thus, much of the assignment of attachment classifications requires the use of discourse analyses: an autonomous individual is able to talk about caregiving experiences in a way that is balanced, coherent, consistent, and open; a dismissing adult will struggle to provide words and examples that adequately describe their attachment world; and a preoccupied adult will produce an interview that is dominated by excessive language in the form of superfluous examples and detail.
A downward extension of the AAI, the CAI (Target et al., 2007) has been developed and used with adolescents (Venta et al., 2014) as well as with school-age children (Shmueli-Goetz, Target, Fonagy, & Datta, 2008), with a continued focus on “representation” in assigning attachment styles. Although less demanding, the CAI similarly uses verbal prompts to elicit descriptors and examples relating to attachment figures and experiences. Much like the AAI, the CAI takes a discourse analysis approach in assigning attachment classifications, paying particular attention to coherence, consistency, and use of examples as providing insight to the youth’s IWM (Shmueli-Goetz et al., 2008; Venta et al., 2014). Across both the CAI and the AAI, transcripts and recordings of the interview are required for coding—further highlighting the importance of the participant’s produced language in coding their IWM. Indeed, word count–based analyses alone have identified meaningful differences between classifications on the AAI (Borelli et al., 2013).
Script- and Story-Based Assessments are Inherently Linguistic
Based on the heading alone, it should be clear that another category of IWM assessments—those based on script and story prompts—is inherently linguistic as well. This category of assessments has been steadily increasing in popularity over the last decade, due in part to the costly nature of administration and coding training for instruments like the AAI. Indeed, this was the focus of a special issue of Attachment and Human Development (Vaughn et al., 2006), which broadly supported the use of these procedures for measuring IWMs in adults.
Bretherton (1991) was one of the first proponents of examining IWMs of attachment by tapping mental scripts—the mental sequences outlining attachment-related causes and effects like “When I am hurt, my mother soothes me” (Dykas et al., 2006, p. 222). To tap these mental scripts, researchers administer a prompt such as the beginning of a story or a set of word prompts and ask participants to complete the story. Although these sorts of tasks were initially developed for use with children, with some tasks using dolls and props in the procedure (see Bretherton, Ridgeway, & Cassidy, 1990), the last decade has seen an increase in the application and standardization of story- and script-based tasks with adults. Indeed, the validity of this method has been demonstrated, in part, through construct validity studies with the AAI (Waters & Rodrigues-Doolabh, 2001).
Despite variations in prompts and coding across studies, the content and, to some extent, the language of the completed story are used to glean information about the accessibility of a secure script and, thus, the participant’s IWM. Generally, a script conveys security if it includes the participant and an attachment figure engaged in an interaction, distressed by some difficulty, making a bid for help, receiving help, accepting the help, making use of the help to overcome the difficulty, and down-regulating affect such that they can return to the previous interaction (Waters & Waters, 2006). Certainly, the capacity to even tell such a story is inextricably linked to linguistic ability. Like the AAI and CAI, a story is judged partially on its linguistic characteristics; attachment security should be reflected in a script that is “complete, [and] well consolidated” (Waters & Waters, 2006, p. 188). Moreover, the narrative must stay on topic and have sufficient content and elaboration to be scored (Waters & Waters, 2006). Although linguistic markers that are important in the scoring of the AAI and CAI—like passivity, tense, and errors—are not factored into the scoring of story- and script-based tasks, authors have nonetheless acknowledged the “highly verbal nature” of these measures (Dykas et al., 2006, p. 229). For instance, very lengthy or very brief stories receive low scores or are deemed unusable, indicating that, certainly at the extremes, linguistic characteristics affect secure base script scores. Moreover, significant, albeit moderate, correlations between story- and script-based tasks and verbal skills have been reported (Dykas et al., 2006).
In addition to being affected by the language generated by the participant, story- and script-based tasks are also affected by the language used to prompt the participant. In the first application of this research to adolescents, for instance, Dykas et al. (2006) asked adolescent participants to complete stories with themes like acne, getting a haircut, a tennis match, and studying for an exam. Notably, the authors asked girls and boys to complete different stories based on pilot testing that had suggested differential gender-based sensitivity to these themes. Moreover, the researchers identified differences in the secure base script scores for stories related to mothers and fathers and discussed the possibility that the prompting language generated differences in the scripts produced. The concern that the prompting language and content may affect the script produced has also been echoed in adult research. Although cross-cultural research using story- and script-based tasks is limited, existing research indicates that prompts must be culturally modified (Rodrigues-Doolabh, Wais, Zevallos, & Rodrigues, 2001; Rodrigues-Doolabh, Zevallos, Turan, & Green, 2003) to produce script scores that are comparable to U.S. samples.
IWMs Emerge in Linguistic Contexts
This statement is twofold: first, IWMs emerge in childhood through linguistic interaction and across development, and second, IWMs emerge in attachment interviews through their unique linguistic properties. Bowlby (1982) described IWMs as arising from the communication between a caregiver and a child. Theoretically, the caregiver–child interactions during infancy and early childhood produce nonverbal/emotional structures that, with development, evolve into a verbal IWM. That is, as the child acquires linguistic capabilities, the internalized self-other interaction patterns increase in complexity and the basic, nonverbal model is transformed into a mature, linguistic IWM (Appelman, 2000). Put succinctly, “Language becomes the dynamic tool through which the individual negotiates meaning and transforms the organized, nonverbal pattern of relatedness (Fogel, 1993) into narrative” (Appelman, 2000, p. 198). It is the narrative-bound, mature IWM that is of interest for clinicians and researchers studying attachment beyond early childhood. Indeed, early conceptualizations of the IWM of attachment made so clear the impossibility of separating attachment representation from their linguistic caregiving context that it seemed “possible to question whether individual differences in security of attachment might instead reflect individual differences in some underlying verbal competence” (Main et al., 1985, p. 92).
Although existing empirical data have conclusively answered Main and colleagues’ (1985) question—the answer is no, attachment (at least when measured through interview-based or story- and script-based measures) is not simply verbal competence—the question itself highlights the fundamentality of the linguistic context for the development of the IWM. Across age groups and languages, early attachment classifications have been linked to linguistic (Bacro, 2011; Costantini, Cassibba, Coppola, & Castoro, 2012; van IJzendoorn, Dijkstra, & Bus, 1995) and cognitive abilities (Egeland & Farber, 1984; van Bakel, & Riksen-Walraven, 2004; van IJzendoorn & van Vliet-Visser, 1988), but these links are viewed as reciprocal (wherein secure attachment enhances ability just as ability enhances security; Stievenart, Roskam, Meunier, & van de Moortele, 2011). Moreover, no such link between verbal ability and attachment classification exists when attachment interviews are used with adults (Bakermans-Kranenburg & van IJzendoorn, 1993; Sagi et al., 1994) or children (Shmueli-Goetz et al., 2008). Likewise, after controlling for verbal ability, script-based assessments remain significantly associated with other attachment tasks (Dykas et al., 2006). These results suggest (reassuringly) that attachment interviews and story- and script-based tasks do not mistakenly ascribe security where only verbal ability exists. However, the fact remains that, at least in early childhood, attachment relationships, attachment measures, and linguistic ability are correlated.
In the AAI, recent studies based on Linguistic Inquiry Word Count (Pennebaker, Booth, & Francis, 2007)—narrative discourse analyses method—suggest that the IWM emerges in unique and replicable ways in the linguistic context of the interview. For example, Waters, Steele, Roisman, Haydon, and Booth-LaForce (2016) analyzed more than 1,600 AAI narratives using this method and noted that the transcripts of individuals with a dismissing IWM were consistently shorter and minimized attachment relationships whereas the preoccupied counterparts were longer and contained increased references to conflict and anger. Together, linguistic variables accounted for remarkable portions of the variance in the coder-derived AAI classifications—36% to 54%. These findings build upon the earlier work of Cassidy, Sherman, and Jones (2012), who, using a very similar method, noted consistent linguistic differences related to the attachment classification and, moreover, reported agreement between classification based on AAI coding and based on linguistic analyses in 71% of cases. Together, this emerging body of research illustrates the way that IWMs are both linguistically created and linguistically bound—at least when conceptualized within semistructured attachment interviews.
Language of Assessment Affects Linguistic Production and Understanding
Together, the reality that attachment interviews are linguistically demanding and that IWMs of attachment emerge in a linguistic context suggests that optimal assessment should address the individual’s native language—coincidentally referred to as the mother tongue. Indeed, aspects of language that are critical for the discourse analyses of the AAI and CAI, including emotional openness, coherence (including reflection about own and others’ mental states), and clarity of language, could be negatively impacted by limitations in fluency. Although story- and script-based tasks de-emphasize discourse analysis, the coherence and length of story production remains relevant to IWM assessment. In the next section, we will describe how assessment in a second language might have particular detrimental impacts upon the respondent’s production of emotional language. In addition, reduced verbal fluency, including longer and more frequency pauses as well as more repetitions, is associated with second-language discourse with large effect sizes as compared with first-language discourse (De Jong, Groenhout, Schoonen, & Hulstijn, 2015), suggesting that dysfluency of discourse, paucity of elaboration, unlicensed pauses, or unintended inconsistencies within the narrative might be misinterpreted as indicators of attachment insecurity in the AAI and CAI while really reflecting reduced fluency. These limitations would compound limitations in expressive vocabulary as well as issues of misunderstanding associated with incomplete receptive vocabulary in a second language (Cameron, 2002; Hutchinson, Whiteley, Smith, & Connors, 2003). Finally, fatigue, emotional expression during the interview, and controlling/derogatory behavior toward the interview or interviewer could be spurred by the interview’s increased difficulty when conducted in a second language or misunderstanding related to the aim of the interview and interviewer prompts.
Several empirical studies from the field of linguistics illustrate the possible hurdles of assessing an IWM in someone’s nonnative language. At the most basic level, research demonstrates that completing psychological tasks using a nonnative language can be detrimental for someone’s performance (Kisser, Wendell, Spencer, & Waldstein, 2012; Sotelo-Dynega, 2013). In addition, difficulties communicating complex concepts, like metaphors, have been associated with assessment in a second language. For instance, Hmong adolescent immigrants were found to have different understandings of the meanings and situational uses of metaphors for emotions and attitudes when compared with English speaking adolescents, and when expressing themselves in their primary (Hmong) versus their second language (English; Dunningan, McNall, & Mortime, 1993). This metaphorical nonequivalence resulted in poor reliability and validity coefficients on standardized measures (Dunningan et al., 1993). Interviewees assessed in their second language also report feeling that there were things they wanted to convey that were untranslatable, such as metaphors, jokes, and idioms (Homayounpoour & Movahedi, 2012). Likewise, clinicians report difficulty capturing the meaning of sarcasm, and face difficulties when dealing with interruptions during interviews, when there was a mismatch between the client and clinician’s native languages (McNamara, 1997).
Moreover, assessment in a second language can influence the content that is disclosed as well as the interviewee’s emotional reaction to the interview—an important consideration in coding the AAI and CAI. Specifically, memories may be better and more accurately recalled, involuntarily accessed, and discussed in the client’s native language rather than in a second language (Kinginger, 2004; Pavlenko, 2006). Furthermore, a qualitative study found that clients interviewed in a language other than their native language tend to report feeling misunderstood, unsatisfied, and alienated (Homayounpoour & Movahedi, 2012)—that may exacerbate the already difficult emotions that may be elicited by attachment interviews. Finally, they believed that conducting the interview in their second language facilitated their wish to hide information that they believed would not be able to hide from a clinician who spoke their mother tongue (Homayounpoour & Movahedi, 2012). To the same end, bilingual people have reported perceiving themselves as different people when using different languages (Pavlenko, 2006). Linguistic studies note that second-language processing is “qualitatively” different from first-language processing in that it consistently reflects shallower and less detailed processing (see review by Clahsen & Felser, 2006), which critically persists even when the individual is proficient in the second language. Together, these studies suggest that above and beyond detrimental effects on performance, assessing IWMs in a second language may inadvertently influence the content disclosed, the interviewees’ emotional response to the task, and the degree to which the interviewee can obscure aspects of the self without the interviewer’s knowledge.
Conclusion
To our knowledge, there are no studies addressing how the spoken language specifically affects the report of early childhood experiences in attachment classification assessments among nonnative English speakers. The research reviewed in this section suggests that the native language is critically linked to the variables assessed by attachment interviews and stories—recollection of memories, coherent discourse, and emotionality within the task. Indeed, the assessment occurring in the interviewee’s second language may be sufficient to increase expressions of distance, superficiality, and detachment that are subsequently interpreted as a dismissing stance.
IWMs of Attachment are Emotionally Laden
We have already discussed the reality that interviews designed to assess IWMs, like the AAI and the CAI, rely heavily on the respondent’s use of labeled and elaborated emotion. Further evidence bolstering the notion that IWMs of attachment are emotionally laden comes from literature bases conceptually defining the IWM as inextricably linked with emotion and describing differences in emotional functioning based on attachment classifications. This section will review those literatures first; then, because empirical research has not evaluated the degree to which these aspects of IWMs are affected by assessment in a second language, the remainder of this section will review extant literature regarding what happens to the assessment and expression of emotion, broadly, in a second language. The effects of caregiving on emotional language learning will not be discussed, though those are thoroughly reviewed by authors who, importantly, note that the attachment context is where a first language, particularly the emotional content of that first language, is learned (Harris, Gleason, & Ayçiçegi, 2006).
Emotion Is a Structural Component of IWMs
Bowlby’s initial conceptualization of IWMs was steeped in emotion—leading to the view that IWMs are “hot structures” created out of and driving emotion (Pietromonaco & Barrett, 2000). Indeed, in the landmark study moving the conceptualization of attachment “to the level of representation,” Main and colleagues described the IWM as emerging from the emotional and behavioral outcomes of attachment-related events and subsequently directing future emotional experience (Main et al., 1985). Moreover, both emotional reactivity and emotion regulation have been tied to the theoretical structure of IWMs across models (see reviews by Fraley & Shaver, 2000; Pietromonaco & Barrett, 2000), with affect being generally accepted as the “organizer” of interpersonal relationships (Fonagy & Target, 2007) since infancy, and, therefore, a structural foundation of IWMs of attachment. In this sense, the IWM is more a pattern of felt emotion than a pattern of behavior or social expectation—indeed secure attachment, for example, becomes a felt “expectation of distress being met by comforting” (Fonagy & Target, 2007, p. 422).
Bolstering the link between IWMs and emotion are numerous studies identifying individual differences with regard to emotion regulation based on attachment classifications. Indeed, in a 2010 review, Allen and Miga (2010) asserted that attachment interviews may be interpreted as “primarily assessing” emotion regulation capacities, perhaps explaining high correlations between attachment security and emotion regulation in adolescents (e.g., Venta, Sharp, & Newlin, 2015). While not all authors would agree with this proposition, there is an uncontested link between attachment and emotional understanding and regulation (Mikulincer, Shaver, & Pereg, 2003). Indeed, emotion regulation and understanding have been conceptualized as emerging from attachment relationships (Calkins, 2004), via what Chen and colleagues (2004) call the co-regulation of affect, and explaining individual differences in affect regulation strategies (Mikulincer et al., 2003).
Language of Assessment Affects Emotional Content
To date, no study has examined the extent to which emotional expression during the AAI or CAI differs in response to the language of assessment. Still, a broader literature developed within the last century has consistently demonstrated that the emotional valence of words affects how words are processed (Altarriba & Bauer, 2004; Gray, Hughes, & Schneider, 1982; McGinnies, 1949) and consistently demonstrates that emotion is processed more deeply in languages that (a) are acquired early in life; (b) are demonstrated to be, or are self-reported by the participant to be, the more dominant/proficient language; (c) were acquired in a naturalistic setting; or (d) are the more heavily socialized languages (Dewaele, 2004, 2008). These factors have demonstrated import on a hierarchal scale, with age of acquisition being the least influential factor when dominance/proficiency, context, and socialization are taken into account. The most variance in emotional processing is explained by context of acquisition and socialization (Dewaele, 2008; Harris et al., 2006)). Most research, however, has been done on bilinguals who acquired their second language later in life, whose more dominant/proficient language is their first language, who acquired their second language in an instructional setting, and who are more socialized in their first language. In the subsequent sections, we will describe existing empirical research regarding the effects of second language use on memory and communication.
Memory
Research on word types among monolinguals has revealed that, when comparing concrete with abstract with emotion words, emotion words are more easily remembered than concrete words, which are more easily remembered than abstract words (Altarriba & Bauer, 2004). Similar research on bilinguals, though, reveals that emotion words are only more easily remembered than neutral words in the first language, suggesting that late learners of a second language do not automatically acquire the emotional valence of words along with the semantic significance (Anooshian & Hertel, 1994). Although this can change when the second language becomes socialized and the second-language learner incorporates the emotional scripts of the second language into their language repertoire (Pavlenko, 2002), bilingual individuals may not display the same tendency as monolingual individuals to recall emotion words.
Fiacconi, Dekraker, and Köhler (2015) demonstrated that it is the physiological reaction that accompanies emotion that helps to more deeply encode emotion words. Buchanan (2007) demonstrated through brain imaging that it is not solely that the emotion faced in the initial experience and encoding of the memory that affects recall, but also that recall is improved when an individual experiences the same emotion during retrieval of the memory. Foa and Kozak (1986) took this idea one step further to claim that the more relevant information that is present during retrieval, the more accurate the memory will be. Therefore, processing and retrieval will more fully and accurately take place if the language of retrieval matches the language of encoding (Altarriba, 2003; Marmolejo, Diliberto-Macaluso, & Altarriba, 2009; Schrauf & Rubin, 1998).
Studies on free recall have been the most instrumental in exhibiting the relationship between emotion and language. One such study asked participants to speak about a personal life experience in both the language in which the experience occurred and in the participants’ other language. This study found that recollection in which the language used was both the language of encoding and retrieval was more emotionally rich, with more expression, imagery, and dynamic tone as compared to a version of the same memory in which the language of retrieval was dissimilar to the language of the experience (Javier, Barroso, & Muñoz, 1993). Such an effect has also been demonstrated for recall in participants’ first versus second language. Seminal research by Bond and Lai (1986) demonstrated that Chinese–English bilinguals avoided speaking to great lengths about embarrassing moments in their native language, Chinese, and instead preferred talking about such events in their second language, English, because of the intense emotion experienced when recounting the embarrassing event in their first language. These data suggest that the emotional experience of recounting a memory is strongly influenced by language—a reality that has not yet influenced changes in the assessment of IWMs.
The heightened effect of aversive words, such as taboo words and childhood reprimands, on memory has likewise been shown in memory word tasks. Although participants were able to recall more adverse, emotionally charged words in their second language (Ayçiçegi & Harris, 2004), lab measures such as skin conductance responses (Harris, 2004; Harris, Ayçiçegi, & Gleason, 2003) show that participants have a stronger physiological reaction to aversive words in their native language, suggesting that the relatively heightened memory effect in the second language is experienced due to the participants’ desire to avoid experiencing the deep emotion attached to the aversive native language words, resulting in their superficial encoding (Ayçiçegi & Harris, 2004; Wu & Thierry, 2012). Likewise, a recent study by Ivaz, Costa, and Duñabeitia (2016) revealed that participants were able to more accurately and more quickly match self-relevant stimuli that had been paired with arbitrary shapes as compared with stimuli that were not relevant to the self. This “self-bias” was much larger in the participants’ native, more dominant language, than in their second language, suggesting that emotional reactivity to self-relevant stimuli is stifled in the context of a second language. Although some of these biases, like the aforementioned differences in skin conductance responses, are eliminated for early second-language learners who were more proficient in their second language (Harris, 2004), studies like these demonstrate the complex interaction of language and emotion on memory.
Communication
When learning a second language, many individuals will fail to acquire the pragmatics or the social and emotional effect of the language unless acquisition takes place in a naturalistic setting, or they are immersed into the culture of the language (Dewaele, 2008). As more language interactions take place in more diverse settings, more memory traces will be created, affording the individual the ability to more readily understand the second language at a deeper, more emotional level. This concept has been termed the “emotional contexts of learning theory” (Harris et al., 2006, p. 272). This process is made easier when the emotional concepts and semantic aspects of the native and the second language are similar, and made more difficult when the two languages are dissimilar. “Cariño” is an emotional concept in Spanish, for example, that has no exact equivalent in English (Altarriba, 2003), similarly “tú” and “usted” are semantic aspects of the Spanish language that have no equivalent in English (Marcos, 1976). According to the “emotional contexts of learning theory,” the emotional significance behind these concepts will be harder to acquire and communicate based on language dissimilarity.
Until an individual experiences a sufficient number of contexts, an individual’s first language will be more emotionally meaningful than their second. This has been demonstrated in several studies on oral communication, such as Dewaele’s (2008) work on The Emotional Weight of “I love you” which demonstrated that the emotional weight of “I love you” depended on the frequency and quality of contexts in which the phrase was uttered. Indeed, the weight of “I love you” in a second language surpassed the weight in a first only when the feeling of love was more often experienced in the second language. Likewise, Dewaele’s (2004) work on The Emotional Force of Swearwords and Taboo Words exhibited that swearwords and taboo words experienced more often and in more naturalistic settings were rated by participants as being more profound, even if not in the first language. These studies may have direct relevance to the assessment of IWMs, with questions on both the AAI and the CAI assessing feelings of love in attachment relationships as well as coding strategies that explicitly examine the presence of re-experienced anger (which may well be communicated through swearwords or taboo words during the interview). The emotional experience associated with the production and comprehension of speech is important because affect may not match production. This may be the case, for example, with an interviewee who describes with flat affect and rudimentary adjectives in their second language the love that they feel for their caregivers. Yet if they are given the opportunity to elaborate in their first language, the deep emotion felt by the client is eloquently communicated (Altarriba & Santiago-Rivera, 1994).
While less relevant for the assessment of IWMs, which is typically undertaken via oral interviews, limited language access to emotion is seen in written communication as well. Lab-based research using event-related potentials has demonstrated that there is a discrepancy in language access when reading and comprehending emotionally charged text. Data reveal that emotional valence influences processing speed in the second language causing the second language to be accessed more slowly than the first language, thus diminishing the effect of the emotion (Opitz & Degner, 2012). This may help explain why participants who read emotional passages interpret them with more emotion in their first language than in their second as exhibited by functional magnetic resonance imaging (fMRI) scans (Hsu, Jacobs, & Conrad, 2015).
Conclusion
It is evident that many factors influence the ability to properly answer the following question: “what happens to emotion when assessed in a second language?” Studies like those presented throughout this section lend support to the call for appropriate language measures in assessing emotion. This is not to say that there is any one appropriate language measure, but rather that the assessment should be specific to the individual based on factors such as (a) age of language acquisition, (b) language dominance/proficiency, (c) context of acquisition, and (d) degree of socialization. This section touched on the “whats” and possible “hows” of two basic concepts, memory and communication, central not only to the assessment of emotion, broadly, but also to the assessment of IWMs of attachment specifically.
Conclusions and Future Directions
The literature reviewed thus far supports two broad arguments: IWMs are linguistically bound and emotionally laden. Although no research to date has specifically examined the effect of nonnative language assessment on IWMs of attachment, related research demonstrates that the assessment may be subject to deleterious effects related to the production of language, miscommunication, increased distancing from the content, difficulty generating memories, and emotional language, among many other basic challenges related to task demands and fatigue. The broad aim of this review is to suggest that these artifacts of second-language assessment may be misconstrued as attachment insecurity, thereby negatively impacting the validity of attachment interviews. A summary of translated attachment interviews and questionnaire-based measures is provided in Table 1. The aim of summarizing existing attachment assessments for non-English speakers is to spur research and clinical practice that assesses attachment in the individual’s native language and examines the central questions of this review empirically. Indeed, an important avenue for future research—capable of explicitly addressing some of the questions posed in this review—is the random assignment of bilingual participants to completion of verbal attachment assessments in either their native or second language and comparing data across groups. The wide array of languages represented in Table 1 suggests that this task may be undertaken with participants from many linguistic backgrounds.
Overview of Attachment Measures Translated for Use With Non-English Speakers.
An additional use of the data summarized in Table 1 would be assessment of changes in IWMs across the life span, as fluency and dominance of a second language increases. Indeed, attachment security that emerges in adolescence or adulthood may occur within the context of relationships navigated in an individual’s second language. For instance, an individual may transition from attachment insecurity with their parent to attachment security with their romantic partner, potentially transitioning from a native language to a second language as the dominant language of attachment through the natural course of development. In that case, the assessment of their IWM in the second language could facilitate production of a coherent, emotionally laden narrative associated with the security felt toward their partner, whereas assessment in their native language could generate a narrative characterized by greater insecurity. This possibility remains a completely open field of experimental study—as do related lines of inquiry regarding shifting IWMs and temporally similar changes in linguistic dominance.
It would be neglectful to omit mentioning that many nonnative English speakers will also identify with a minority culture. Although the effect of culture on attachment interviews is outside the scope of this review and has been well-reviewed elsewhere, it bears mentioning that there have long been concerns that the measurement of attachment is biased against non-Western cultures and that measures developed in the United States and Western Europe would overclassify other ethnic groups as insecurely attached. Some studies have demonstrated the overrepresentation of insecurity among children from parts of Africa, Japan, Indonesia, and Israel (see Miyake, Chen, & Campos, 1985; van, IJzendoorn, & Bakermans-Kranenburg, 2010; van Ijzendoorn & Sagi-Schwartz, 2008) as feared. However, most evidence suggests that secure attachments are readily detected across varied cultures (van Ijzendoorn & Sagi-Schwartz, 2008), that intracultural differences are larger than intercultural differences (van Ijzendoorn & Sagi-Schwartz, 2008), and that attachment styles are quite similar across studies conducted with many cultures (van IJzendoorn & Bakermans-Kranenburg, 2010) as long as sensitivity to cultural manifestations of parenting and care are considered. Similar conclusions have been drawn by a small body of researchers examining story- and script-based assessments of IWMs across cultures (Rodrigues-Doolabh et al., 2001; Rodrigues-Doolabh et al., 2003). Still cultural differences may not be at the forefront of an interviewer or researcher’s mind, as he or she endeavors to conduct an AAI or CAI or develop a story- or script-based prompt. Moreover, subtle changes in behavior, body language, and affective and cognitive processing capacities during the task (McNamara, 1997; Pavlenko, 2006), while possibly enhancing the quality of the clinical interaction (Appelman, 2000; Tausczik & Pennebaker, 2010), may pose an additional barrier when an interviewer fails to consider the relevance of the interviewee’s culture. The effects of culture may be particularly profound during attachment tasks. Indeed, it is well noted within cognitive and achievement testing that when a psychological assessment possesses a high degree of both linguistic demand and cultural loading, it is most subject to validity issues (e.g., Rhodes, Ochoa, & Ortiz, 2005). Although IWM interviews have not been placed within this Culture-Language Interpretive Matrix before, the reality that these interviews and tasks have both high linguistic demand and cultural loading suggests that interviewee performance may be most negatively impacted.
To our knowledge, no study has examined the effect of nonnative language assessment on the CAI, AAI, or story- and script-based tasks. In our view, these are critical tasks for future research, in light of research suggesting that the linguistic demand, emotional content, and cultural loading of constructs affects their validity when administered in a second language. In identifying areas for future research, we wish to highlight the work of Mallinckrodt and Wang (2004) who endeavored to translate the Experiences in Close Relationships scale into Chinese, assess the semantic equivalence of items when administered in English versus in Chinese to bilingual speakers, and evaluate the psychometric properties of each version of the scale in a bilingual sample. This study reported promising results for a questionnaire-based measure of attachment, but no such study has been conducted with interview-, story-, or script-based measures, which we suggest will be more affected by second-language assessment. Still, the guidelines of Mallinckrodt and Wang (2004) for measure translation and assessment may serve as a model for researchers seeking to address the fundamental question of this review empirically—how is the assessment of IWMs affected when conducted in an individual’s nonnative tongue?
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.