Psychophysiological Evidence for Defensive Discourse in Attachment Interviews: Generalizability Across Sex and Ethnicity

Trained judges reliably apply a classification of “dismissing” to adults who appear to defensively distance themselves from the emotional content of the Adult Attachment Interview (AAI) by idealizing and/or downplaying the importance of their childhood relationships with parents. One way to confirm whether those who use such “deactivating” discourse strategies are actually avoiding discussion of malevolent early experiences, as is assumed by attachment researchers, is by seeking convergent evidence that such individuals show subtle signs of emotional distress when recounting childhood memories. This paper reports relevant data from a cross-cultural extension of research by Dozier and Kobak (1992) demonstrating a link between deactivating discourse in the AAI and electrodermal response, a specific physiological indicator of covert arousal. More specifically, results of physiological change-score and growth curve analyses in the current study suggest that deactivation in the AAI is discriminantly associated with concurrent electrodermal (but not cardiovascular) reactivity both across sex and within Chinese and European American samples. AAI Psychophysiology 3 Psychophysiological Evidence for Defensive Discourse in Attachment Interviews: Generalizability Across Sex and Ethnicity “The mind commands the body and the body obeys. The mind commands itself and finds resistance.” – St. Augustine (354-430) Although attachment theorists have long argued that individuals’ childhood experiences with primary caregivers form an important basis for their adult relationships (Bowlby, 1969/1982; Waters & Cummings, 2000), there are paradoxically few simple ways of testing this seemingly straightforward assertion. The most intuitive approach, of course, is to conduct longitudinal studies that follow children into adulthood (Masten, Hubbard, Gest, Tellegen, Garmezy, & Ramirez, 1999; Beckwith, Cohen, & Hamilton, 1999; Waters, Hamilton, & Weinfield, 2000; Roisman, Madsen, Hennighausen, Sroufe, & Collins, 2001). Unfortunately, such studies require time and labor that few researchers have. An alternative method involves asking individuals to describe their childhood experiences, analyzing the narratives produced in this context, and then relating such information to indicators of current functioning in salient adult relationships. While this method has the advantage of requiring fewer resources, one clear concern is whether accurate inferences can be drawn about individuals’ earlier experiences based on clinically informed analyses of their retrospective reports (Yarrow, Campbell, & Burton, 1970; Henry, Moffit, Caspi, Lanley, & Silva, 1994; Roisman, Padrón, Sroufe, & Egeland, in press). This paper, a cross-cultural extension of work by Dozier and Kobak (1992), was designed to provide a demonstration of how cross-sectional, psychophysiological data can be applied toward validating interview-based methods of assessing the nature of early experience. AAI Psychophysiology 4 One rather well-developed example of a retrospective yet developmentally informed strategy for studying childhood relationship experiences is available in the research tradition that has grown up around the Adult Attachment Interview (AAI), a semi-structured protocol about early experiences with parents (George, Kaplan, and Main, 1985; see Hesse, 1999, for an excellent review). Harnessing the theoretical insights of John Bowlby (1969/1982, 1973, 1980), the early methodological strides of Mary Ainsworth and her students (Ainsworth, Blehar, Waters, & Wall, 1978), as well as the accumulated longitudinal evidence available on the importance of early attachments (e.g., Sroufe, 1979, 1983; Carlson, 1998; Thompson, 2000), numerous developmental researchers are now actively exploring the antecedents, concomitants, and sequelae of individual differences in the ways in which adults have come to think about and emotionally process their childhood experiences with the AAI. Moreover, although the interview was originally developed with a targeted focus, that of predicting the quality of parent-child attachments in the next generation (see van IJzendoorn, 1995), its application has found farreaching utility in exploring diverse but related domains of childrens’ general adaptation (Crowell & Feldman, 1988; Cowan, Cohn, Cowan, & Pearson, 1996), adults’ romantic relationships (Cohn, Silver, Cowan, Cowan, & Pearson, 1992; Paley, Cox, Burchinal, & Payne, 1999; Roisman, et al., 2001), as well as the development of psychopathology (van IJzendoorn & Bakermans-Kranenburg, 1996; Main, 1996). All AAI studies rely critically on the theoretical rationale for the adult attachment classification system as developed by Main and Goldwyn (1998). Based on content analyses of individuals’ discourse, Main and her colleagues originally described (and made auxiliary inferences regarding) three principal ways in which emotions and cognitions are regulated with respect to parents and early experiences during the AAI (Main, Kaplan, & Cassidy, 1985). AAI Psychophysiology 5 Perhaps unsurprisingly, the majority of interviewees can be classified as secure/autonomous, telling a coherent narrative about their early experiences, whether described as good or ill. A growing body of literature has shown that individuals with autonomous states of mind are likely to parent securely attached infants (van IJzendoorn, 1995) and enjoy high quality romantic relationships with adult partners (e.g., Roisman, et al., 2001). Recent longitudinal evidence suggests also that such individuals are likely to have actually encountered a supportive relationship with a primary caregiver in their childhood, a finding critical to the notion that past experiences are represented veridically in the quality (though not necessarily content) of adults’ discourse (Beckwith, et al., 1999; Waters, et al., 2000; Roisman, et al., in press; but see Lewis, Feiring, & Rosenthal, 2000). The second classification described by Main & Goldwyn (1998) is described as preoccupied and is marked by discourse revealing either angry or passive enmeshment in past relationship experiences. Such individuals articulate answers to AAI probes that are characteristically hyperactivated, angry, and unrestrained or alternatively seem passive and confused. Although relatively little is known about its unique correlates, participants so classified self-report relatively high levels of psychopathological distress (e.g., on the Minnesota Multiphasic Personality Inventory; Pianta, Adam, & Egeland, 1996) and have been observed to parent their adolescent offspring in an anxious and intrusive manner (Kobak, Ferenz-Gillies, Everhart, & Seabrook, 1994). Although relevant evidence is limited, preoccupied states of mind also seem to be predictable outcomes of social experiences, foreshadowed by negative childhood life events, including parental divorce (Beckwith, et al., 1999) In striking contrast to the first two classifications, the last group of individuals described by Main and Goldwyn (1998) seem to defensively distance themselves from the emotional AAI Psychophysiology 6 content of the AAI, minimizing the importance of attachment relationships and/or presenting idealized depictions of their childhood experiences. These interviewees, categorized as dismissing, can be cool, removed, and sometimes hostile. In addition, they are oftentimes reluctant to elaborate upon early memories and, when they do, normalize seemingly harsh childhood experiences (Main & Goldwyn, 1998). Presaged by almost a century of psychoanalytic theorizing regarding such defensive presentations of interpersonal experience (Freud, 1936), it has been demonstrated convincingly in the last decade that this kind of repressive reporting style is by no means benign, being associated with emotionally explosive interactions with parents in adolescence (Kobak, Cole, Ferenz-Gillies, Fleming, & Gamble, 1993) and with less effective parenting in adulthood (Lutz & Hock, 1995). Again, key to the supposition that adult narratives can provide reasonably accurate windows on the true nature of past experiences, several recent longitudinal studies have now shown that dismissing states of mind in adulthood are anteceded both by insecurity in infancy (Hamilton, 2000; Waters, Merrick, Treboux, Crowell, & Albersheim, 2000) as well as chronically insensitive caregiving in the first year of life (Beckwith, et al., 1999). Obviously, careful appraisal of the soundness of the inferences made with respect to the measurement of AAI individual differences as explicated above is crucial for extracting meaningful conclusions from the adult attachment literature. While psychometric work on the protocol has indeed been rigorous with respect to its reliability, convergent, and discriminant validity (Beckwith et al., 1999; Bakermans-Kranenburg & van IJzendoorn, 1993; Crowell, Waters, Treboux, O’Connor, Colon-Downs, Golby, & Posada, 1996; De Haas, BakermansKranenburg, & van IJzendoorn, 1994; Sagi, van IJzendoorn, Scharf, Koren-Karie, Joels, & Mayseless, 1994), precious little empirical evidence testifies directly to the cross-cultural AAI Psychophysiology 7 generalizability and validity of the AAI’s classification system per se, especially as regards the heavily inference laden judgements required to classify individuals as dismissing. This paper seeks to help redress this error of emphasis, reporting data from a cross-cultural extension of a key study conducted by Mary Dozier and Roger Kobak (1992). Specifically, concurrent psychophysiological assessment was undertaken during the AAI in the present study to more broadly validate the dismissing classification’s use within and across Western and Eastern cultural contexts. The dismissing classification. Dozier and Kobak (1992) point out that perhaps the single greatest inferential leap in coding the AAI is the dismissing classification for narratives containing seemingly idealized descriptions of relationships with parents, limited recall of childhood experiences, and in which the significance of attachment relationships appears minimized. Following Main (1990), these researchers referred collectively to such discourse that seems to serve the purpose of limiting attention to attachment relevant topics as evidence of a deactivating strategy, a defensive posture putatively organized in response to the failure of species-normative attachment behaviors (e.g., distress-related vocalizations) to elicit sensitive, contingent, and timely caregiving in times of distress. Attachment researchers have long speculated that, over time, persistent dismissal of one’s tender needs by caregivers results in a blunting of one’s motivation and perhaps even ability to effectively signal distress in relationships (Main, 1981; Cassidy, 1994; Case, 1996; Cassidy & Kobak, 1998). Irrespective of the strength of the theoretical rationale that undergirds the dismissing classification, it must be emphasized that when trained AAI judges apply it to adult narratives (in the absence of confirmatory data regarding a client or participants’ actual developmental history) AAI Psychophysiology 8 they are making a clinical inference, not a direct assessment, that such deactivating discourse represents a strategy for minimizing attention to emotionally painful childhood experiences. Clearly, convergent evidence must be obtained that this kind of discourse is associated with (likely subtle) signs of emotional conflict or distress during the AAI. Psychophysiological rationale. To be sure, associations between adult attachment representations and individuals’ emotional experiences are hard to assess, however, because the dismissing classification is characterized by defensive processes resulting in the repressive reporting of affective states (Pianta, Egeland, & Adam, 1996). Fortunately, alternatives to self-reports exist that are far less biased by the subjective-interpretative and mood-related distortions endemic to such instruments. In particular, psychophysiological methods using noninvasive surface sensors have great promise in this regard due to the relatively objective and non-intrusive nature of their assessment. Surprisingly, only a single published report has exploited this promise, presenting data on the physiological correlates of individual differences inferred from the AAI (Dozier & Kobak, 1992; but see papers from Roisman, 2001). Specifically, Dozier and Kobak (1992) determined that deactivating discourse in the AAI, as assessed using a Q-sort approach to measuring attachment-related individual differences, was associated with rises in skin conductance from baseline, especially during AAI probes referencing separation, rejection, and other potentially threatening childhood experiences. Dozier and Kobak’s (1992) study was especially convincing to many in the field because it found that such increases in electrodermal activity were discriminantly associated with deactivation, being unrelated more broadly to insecurity generally (e.g., preoccupation). AAI Psychophysiology 9 Citing the theoretical work of Fowles (1980) and the empirical studies by Pennebaker and his colleagues on the physiological correlates of deception (e.g., Pennebaker & Chew, 1985), Dozier and Kobak (1992) interpreted their results as convergent evidence for the validity of the dismissing classification of the AAI in that deactivation was linked in their study to a physiological correlate believed to specifically index intrapsychic conflict and response inhibition. Framing this analysis is Gray’s widely cited two-process theory offering that electrodermal activity may be thought of as a reliable physiological marker of the effortful inhibition of behavior while cardiovascular measures in contrast more broadly tap behavioral activation (Gray, 1975). Using Gray’s (1975) theoretical framework as a point of departure, Dozier and Kobak (1992) went on to speculate (though they did not empirically test the hypothesis) that their electrodermal findings would not generalize to measures of cardiovascular arousal (e.g., heart rate). It should be emphasized that this discriminant prediction, if borne out, is important theoretically: it suggests that dismissing individuals may be partially able to suppress aspects of overt emotional distress (e.g., by modulating cardiovascular arousal) yet physiological signs remain that they have not fully resolved early negative attachment experiences (e.g., they experience electrodermal affective “leakage” when confronted by a task that unrelentingly directs attention toward potentially unpleasant memories). At least two issues remain unaddressed. First, Dozier and Kobak’s (1992) study was limited with regard to the demographics of its sample, which was predominantly Caucasian and female. It thus remains unknown if the deactivation-skin conductance linkage generalizes to other ethnic groups and to men. Second, Dozier and Kobak (1992) only measured participants’ electrodermal response during the administration of the AAI. Therefore, it remains quite AAI Psychophysiology 10 speculative whether deactivating discourse is uniquely associated with electrodermal reactivity, being unrelated more broadly to physiological measures of cardiovascular arousal, for example. In order to explicitly address both of the foregoing issues empirically, this paper reports data from a systematic replication and extension of Dozier and Kobak’s (1992) original study using a cross-cultural, gender balanced sample capable of identifying demographic variables that could theoretically moderate associations observed between deactivation in the AAI and measures of physiological change during the interview. In addition to the assessment of electrodermal activity (skin conductance levels), several indicators of cardiovascular arousal (cardiac inter-beat interval, finger pulse transmission time to the finger, and finger pulse transmission time to the ear) were also sampled second-by-second during the administration of the AAI to Chinese and Chinese American as well as European American college students. Cross-cultural rationale. Chinese and Chinese Americans were sampled because this group provides for a particularly severe test of the cross-cultural generalizability of the central hypothesis of this study that deactivating, defensively organized discourse during the AAI should be associated with electrodermal response. Chinese are often described in ethnographic reports as emphasizing social relationships and the maintenance of interpersonal connections (as contrasted with European Americans, who are depicted as stressing individual uniqueness and the differentiation of one’s self from others) (Markus & Kitayama, 1991). As a result, members of Asian cultures in general and Chinese in particular have been described as moderating and controlling of negative emotions that may disrupt interpersonal harmony (Tsai & Levenson, 1997). From an attachment perspective, these Chinese cultural imperatives may result in systematic differences in their articulation of early childhood experiences in a way superficially similar to those associated with AAI Psychophysiology 11 the dismissing classification, potentially rendering individual differences inferred from the AAI fundamentally confounded with aspects of cultural presentation. Attachment theory supports a view that attachment needs are universal, suggesting that their expression and psychological meaning are not restricted to any particular cultural context (van IJzendoorn & Sagi, 1999). Nonetheless, as stated earlier, relatively little is known about the cross-cultural validity of the AAI’s classification scheme, and this is especially true regarding its utility in Asian cultural contexts (see Rothbaum, Weisz, Pott, Miyake, & Morelli, 2000). Of note, one very recent study conducted in Japan found that individuals with secure states of mind regarding their past experiences are likely to parent securely attached infants, a finding consistent with results described earlier based on American and European samples (Kazui, Endo, Tanaka, Sakagami, & Suganuma, 2000; van IJzendoorn, 1995). Still, that deactivation in the AAI should be associated with covert signs of emotional reactivity across ethnicity remains a particularly important test of the cross-cultural comparability of the AAI experience because, as outlined earlier, it bears directly on the putative meaning of the individual differences inferred from the interview. More generally, this pattern of results would also suggest that clinically informed analyses of adults’ discourse can indeed provide a window on childhood experiences in a way that may transcend the methodological challenges associated with ethnicity and gender. Method Participants Sixty young adults (30 Chinese/Chinese American, 30 European American; 28 male) were recruited for this study from the undergraduate and graduate student population of a large Midwestern university. All participants completed a comprehensive assessment battery including (1) a mental health and cultural identity screener by phone, (2) a packet of developmentally AAI Psychophysiology 12 appropriate, self-report personality inventories, symptom measures, and demographic questionnaires completed before arriving at the lab, and (3) the Adult Attachment Interview while being videotaped and physiologically monitored. Inclusion criteria included an age restriction of 18-30 years. Due to the moderately stressful nature of the laboratory interview, participants were also screened for signs of depression, anxiety, and psychotic thinking. Participants who passed through the screening process and subsequently completed the entire protocol described above received a $20.00 honorarium. Cultural criteria for inclusion in the Chinese/Chinese American sample were as follows: (1) maternal and paternal grandparents were born in Mainland China, Taiwan, or Hong Kong, (2) parents were Chinese or first generation Americans of Chinese ancestry, (3) had Chinese/Chinese American friends in adolescence and/or childhood, (4) spoke a Chinese dialect (e.g., Mandarin, Cantonese) and/or grew up in a bilingual household, and (5) self-identified as Chinese/Chinese American. European American participants all had maternal and paternal grandparents from Europe or of European American ancestry and their parents were at least first generation Americans, born in the United States. In addition, all participants were required to be fluent in English as all interviews were conducted in this language. Few significant demographic differences were observed between the two cultural subsamples. As expected, however, the Chinese American group reported having spent less time in the United States (C/CA M = 10.6 years; EA M = 20.4 years) and indicated significantly less English fluency on a scale of one (not at all fluent) to five (extremely fluent); In addition, the European American group was marginally younger (C/CA M = 22; EA M = 21) and reported a marginally higher mean household income in childhood (also rated on a five point scale). AAI Psychophysiology 13 Apparatus Audiovisual. Remotely controlled, high-resolution color video cameras recorded the participants’ and interviewers’ facial behavior during the study. Cameras were hidden from participants’ view behind a darkened glass on a bookshelf. Lavaliere microphones clipped on participants’ clothing were used to record their verbal responses to the AAI, which were subsequently transcribed verbatim. Physiological. A system consisting of a Dell Pentium computer, HPVEE software, and Coulbourn Lab Link V bioamplifiers was used to obtain continuous recordings of participants’ physiological responses. Procedure Upon arrival to the laboratory, a female interviewer greeted each participant. Female interviewers were selected to help participants feel at ease during sensor attachment and the AAI. In addition, participants were matched with interviewers of the same ethnicity to increase their level of comfort in the research setting (Bradley, Snyder, & Katahan, 1972). Importantly, interviewers underwent extensive training (supervised by the author, a certified AAI coder) and followed a standardized, semi-structured interview script. Note that a total of four research assistants conducted all sessions. While no participant requested that the protocol be curtailed, interviewers were provided with a list of onand off-campus mental health resources should they be requested. In addition, all interviews were monitored closely either by the first author or a licensed clinical psychologist. Assessment of Physiology. Surface sensors measuring skin conductance levels and cardiovascular activity were adhered to participants’ lower ribs, ears, and fingers by research assistants and physiological channels were subsequently monitored second-by-second from an AAI Psychophysiology 14 adjoining room during the semi-structured attachment interviews as well as a three-minute rest period that preceded their administration. During this pre-interview period, participants were instructed to be silent and relax for three minutes to obtain baseline measures of physiological responding. An “X” was placed at eye level in front of the participants to facilitate relaxation during this rest period and participants were asked to look at this target and empty their minds of all thoughts, feelings, and memories. Interviewers were not present during this rest period. Next, participants’ physiological responses were continuously monitored during the AAI. Ultimately, changes in mean levels of physiology from baseline were calculated by interview question. Adult Attachment Interview. The AAI is a semi-clinical, semi-structured interview used to characterize individuals’ current state of mind with respect to past parent-child experiences (George, Kaplan & Main, 1985). This approximately hour-long protocol requires participants to describe their early relationships with their parents, revisit salient separation episodes, explore instances of perceived childhood rejection, recall encounters with loss, and describe aspects of their current relationship with parents, discussing salient changes that may have occurred from childhood to maturity (see Hesse, 1999). According to established protocol, AAI’s were transcribed verbatim and all personally identifying information was removed before transcripts were Q-sorted by pairs of three judges trained through and reliable with the lab of Dr. Mary Main (all coders had achieved ≥ .80 reliability with a training set of 32 cases on both three-way [secure, dismissing, preoccupied] as well as four-way [secure, dismissing, preoccupied, unresolved] AAI classification before data reduction commenced in the present study). Measures The Adult Attachment Interview Q-set. The Adult Attachment Interview Q-set (Kobak, 1993) consists of 100 descriptive cards that are sorted into a forced normal distribution across AAI Psychophysiology 15 nine piles from least to most characteristic (5, 8, 12, 16, 20, 16, 12, 8, and 5 cards per column, respectively). Sixty-six percent (40/60) of the AAI transcripts from this study were double-sorted and reliability of .6 or greater (Spearman-Brown prophecy formula) was achieved for 80% of these transcripts. A third coder rated transcripts for which initial coders were discrepant and sorts that were most highly correlated were ultimately averaged (reliabilities of composited sorts ranged from .73 .93, M = .82). In the final step of data reduction, Pearson correlations were computed between each of the composited sorts and both a prototypic “secure/insecure” and a “deactivation/hyperactivation” sort developed by Roger Kobak and his colleagues (see Kobak, Cole, Ferenz-Gillies, Fleming, & Gamble, 1993, for details). Based on this analysis, participants were assigned continuous scores ranging from -1.00 to 1.00 on each construct, with higher scores indicating greater resemblance to a prototypically secure and deactivating individual, respectively. Attachment scores were subsequently standardized for all correlational analyses. Physiology. Second-by-second measures of physiological responding were sampled from participants’ electrodermal and cardiovascular systems during the baseline period and AAI. Electrodermal response was measured by skin conductance level (SCL). A constant-voltage device was used to pass a small voltage between electrodes attached to the palmar surface of the middle phalanxes of the first and third fingers of the non-dominant hand. SCL was measured in microohms. The specific cardiovascular measures obtained were: (1) cardiac inter-beat interval (IBI). Beckman miniature electrodes with Redux paste were placed in a bipolar configuration on opposite sides of each participant’s chest. IBI was measured as time in milliseconds between successive R waves of the electrocardiogram (EKG), (2) pulse transmission time to the finger (FPT) was calculated by measuring time in milliseconds between the EKG R wave and the arrival of the pulse pressure at the finger, and (3) pulse transmission time to the ear (EPT). A AAI Psychophysiology 16 photoplethysmograph was attached to the ear lobe on the participant’s non-dominant side to measure the volume of blood in the ear. EPT was calculated in milliseconds by measuring time between the EKG R wave and the arrival of the pulse pressure at the ear. Results The principal analyses for this study focused on two multifaceted issues: (1) Is deactivating discourse discriminantly and uniquely associated with electrodermal response in the AAI? and, if so, (2) Does ethnicity or sex moderate this linkage? Given Dozier and Kobak’s (1992) hypothesis that deactivation should be unrelated to cardiovascular response, each dimension of cardiovascular change was examined separately, minimizing the potential for Type II error. In addition, as electrodermal (but not cardiovascular) measures proved to be significantly non-normally distributed (significant Kolmagorov-Smirnov tests obtained on these variables), nonparametric Spearman correlations are presented in all analyses focused on electrodermal response. Analysis began, however, by addressing several fundamental questions regarding potential cultural as well as sex differences on key constructs. Cultural and sex differences on key constructs Independent samples t-tests (see Table 1) were run to determine whether mean-level cultural or sex differences obtained for the Q-sort measures (security, deactivation). Although no sex difference was observed on either attachment variable, a cultural difference obtained for security (but not deactivation), with the Chinese/Chinese American group receiving a lower mean security score than did European Americans. This mean level difference should be understood with respect to absolute within-culture levels, however. More specifically, as the Chinese mean on security is consistent with previously published results from primarily Caucasian samples (e.g., Dozier & Kobak, 1992), an accurate interpretation of the cultural AAI Psychophysiology 17 difference is that security may be somewhat over-represented among European Americans in this sample. Is deactivating discourse discriminantly associated with electrodermal response in the AAI? Spearman correlations presented in Table 2 demonstrated striking consistency with Dozier and Kobak’s (1992) original findings. Specifically, deactivation was robustly related to rises in skin conductance from baseline to each question of the AAI. Moreover, as in Dozier and Kobak’s (1992) analysis, skin conductance was found to be discriminantly related to deactivation, as rises in skin conductance from baseline were not significantly associated with security/insecurity (see Table 2). Is deactivating discourse uniquely associated with electrodermal response in the AAI? Follow-up analyses revealed that deactivating discourse shared a unique association with electrodermal activity in that deactivation was unrelated to indices of cardiovascular activation during the AAI (see Table 3 for illustrative cardiac inter-beat interval data). Neither deactivation nor security/insecurity predicted changes in mean levels of cardiovascular response for any physiological channel for which relevant data were collected (cardiac inter-beat interval, finger pulse transmission time to the ear and finger). Does ethnicity or sex moderate associations observed between deactivation and electrodermal response? Follow-up regressions using ethnicity X deactivation and sex X deactivation interaction terms were used to test for moderating effects of culture and sex on the associations observed between deactivation and electrodermal response. In short, these analyses provided no evidence that ethnicity or sex conditioned the deactivation-electrodemal linkage (all p values were AAI Psychophysiology 18 nonsignificant, suggesting that effects were similar by culture and sex). The comparability of effects observed by ethnicity and sex is clarified in Table 4, which illustrates that findings replicated within culture and gender using split samples. Confirmatory analyses: Growth curve modeling Given longstanding concerns regarding the use of change scores in psychological research (Cronbach & Furby, 1970), supplementary growth curve analyses were conducted next to test whether convergent statistical evidence could be obtained that deactivation in the AAI is associated with electrodermal response. More specifically, a series of growth curve analyses were run using Bryk and Raudenbush’s (1992) HLM software. Essentially, HLM makes use of a two-step regression procedure: First, individual linear growth trajectories (or slopes) can be estimated in a set of Level I regressions. Next, a second (Level II) regression equation is estimated to explain individual differences in these slopes. In the current study, growth in mean levels of electrodermal response in each AAI question was modeled within participants at Level I to produce a set of slopes (representing linear electrodermal growth). Next, attachment-related individual differences (as well as demographic “dummy” codes) were entered as independent variables at Level II to predict variation in electrodermal growth. Findings using growth curve models were consistent with results from the change score analysis: controlling for sex (0 = Female, 1 = Male) and ethnicity (0 = Chinese/Chinese American, 1 = European American), deactivation predicted growth in electrodermal activity across the AAI (p < .05, see Table 5). Moreover, sex X deactivation and ethnicity X deactivation interactions were not significant in follow-up moderator analyses. As expected, results of additional analyses revealed that security/insecurity was not associated with growth in electrodermal activity across the AAI. AAI Psychophysiology 19