Attachment-Related Differences in Perceptions of an Initial Peer Interaction Emerge Over Time: Evidence of Reconstructive Memory Processes in Adolescents

Abstract

This study examined whether attachment, assessed using the Adult Attachment Interview (AAI; George, Kaplan, & Main, 1996) was linked to how adolescents reconstructed their memory for an initial interaction with an unfamiliar peer. Adolescents (N = 189, 62% female) completed a 10-min laboratory task with a student whom they did not know. Immediately following this task, adolescents rated their perceptions of the interaction. Adolescents completed the same perception measure 2 weeks later. Although adolescents classified as secure and insecure on the AAI did not differ in how they perceived unfamiliar peers initially, attachment-related differences emerged over time. Insecure adolescents remembered the interactions as less positive and more negative and also reported being treated with greater hostility than they had initially reported 2 weeks earlier. In contrast, secure adolescents’ memories for the negative aspects of the interaction and for hostile treatment remained stable, although, like insecure adolescents, they remembered the conflicts as being less positive than initially reported.

Interactions with peers have considerable developmental significance, and the social-cognitive features of such interactions have been the focus of much theoretical and empirical inquiry. In childhood and adolescence, for example, a wealth of data indicates that successes and failures with familiar peers (e.g., classmates) might be associated with underlying adaptive versus maladaptive information processing patterns (e.g., Crick & Dodge, 1994). Much less is known, however, about the social-cognitive phenomena surrounding initial peer interactions, especially concerning the cognitive aftereffects that ensue from meeting a peer for the first time. Investigating these cognitive aftereffects is important because every peer relationship begins as an initial interaction with an unfamiliar person, and such aftereffects could influence how an individual subsequently perceives and behaves toward the once-unfamiliar peer, thereby contributing to future relationship quality.

From an information processing perspective, following an initial interaction with an unfamiliar peer, an individual stores information about that interaction in memory (Crick & Dodge, 1994). A period of time then elapses before a second possible interaction occurs, and it is likely that the initial peer-related perceptions degrade and become distorted images of the actual event that occurred (see Baldwin, 1995; Fiske & Taylor, 1991). Thus, as these perceptions degrade, individuals would need to “fill in the gaps” or reconstruct them. Moreover, given that social information processing is typically subjective and based on stable underlying cognitive structures (see Pasupathi, 2001; Tversky & Marsh, 2000), these reconstructive memory processes are also likely biased.

If individuals reconstruct their memories for initial interactions with unfamiliar peers—and do so in a biased way—then we would expect these reconstructive memory processes to differ as a function of individuals’ underlying cognitive structures known as internal working models of attachment. The basis for the proposed link between attachment and individuals’ reconstructive memory for interactions with unfamiliar peers rests on both attachment theory (Bowlby, 1969/1982, 1973) and related research on attachment-memory linkages (see Dykas & Cassidy, 2011, for a review). According to attachment theory, individuals develop relatively stable cognitive–affective mental representations of their attachment experiences over the course of repeated daily experiences with caregivers (see Bretherton & Munholland, 2008, for an account of this process). Many developmentalists consider attachment experiences, and the internal working models forged from them, to be the basis from which much social experience is processed (Bowlby, 1973, 1988; Dykas & Cassidy, 2011; Pietromonaco & Barrett, 2000; Thompson, 2006). For example, although these cognitive models initially store knowledge to help predict the behavior of attachment figures (in the service of calibrating the attachment behavioral system; Bowlby, 1973), these models later generalize to assist the individual in processing a wide array of social information (Bowlby, 1973; see also Dykas & Cassidy, 2011). Internal working models perform this function to expedite the processing of social information and to enhance the person’s ability to predict behavioral and relational outcomes (Bretherton & Munholland, 2008). Several prior studies, for example, have linked attachment to children’s memory for nonpersonal social events (Belsky, Spritz, & Crnic, 1996; Kirsh & Cassidy, 1997), as well as to adults’ memory for emotional stimuli (e.g., Edelstein, 2006; Fraley & Brumbaugh, 2007; Zeijlmans van Emmichoven, van IJzendoorn, de Ruiter, & Brosschot, 2003), childhood events (e.g., Mikulincer & Orbach, 1995), and current interpersonal experiences (e.g., Mikulincer & Horesh, 1999; Pietromonaco & Barrett, 1997).

The principal aim of this investigation was to examine whether and how individuals reconstructed their memories for an initial interaction with an unfamiliar peer as a function of attachment. We focused on adolescents’ interactions with unfamiliar peers, given the considerable importance of peer relationships during this period of life-span development (see Collins & Steinberg, 2006). In adolescence, the quality of individuals’ internal working models of attachment is most often operationalized (through the use of the Adult Attachment Interview; AAI; George, Kaplan, & Main, 1996) in terms of adolescents’ state of mind with respect to attachment (Main & Goldwyn, 1998). Individuals are considered to possess a secure state of mind when they value attachment and think about and discuss coherently and openly attachment-related memories; for example, these individuals objectively appraise positive and negative attachment-related experiences and the effect these experiences have had on their social and emotional well-being. In contrast, individuals are considered to possess an insecure state of mind when they are unable to participate, within the interview, in coherent attachment-related discourse (see Hesse, 2008).

Considerable data already suggest that attachment is linked to the dynamics of adolescents’ peer relationships, including some aspects of social information processing (see Allen, 2008; Berlin, Cassidy, & Appleyard, 2008; Schneider, Atkinson, & Tardif, 2001). In adolescence, for example, connections exist between attachment and adolescents’ peer-related expectations (Zimmermann, 1999, 2004), mental representations (Furman, Simon, Shaffer, & Bouchey, 2002) and self-referential information seeking (e.g., Cassidy, Ziv, Mehta, & Feeney, 2003). Moreover, attachment security has been linked to children’s, adolescents,’ and emerging adults’ behavior toward strangers (Easterbrooks & Lamb, 1979; Feeney, Cassidy, & Ramose-Marcuse, 2008; Roisman, 2006). The present study is the first to explore attachment in relation to the social-cognitive aftereffects of initial interactions with unfamiliar persons. Indeed, how adolescents reconstruct their memories with unfamiliar peers could contribute to the ways in which adolescents subsequently relate (or do not relate) to the unfamiliar peer, which would be an important example of how individuals are active participants in their own development outside the context of family and close relationships (cf. Scarr & McCartney, 1983).

In terms of perceptions immediately following the initial interaction, we did not set a specific hypothesis about links with attachment. On the one hand, attachment is thought to influence individuals’ attention, expectations, perceptions, and interpretations of social experiences (Bowlby, 1973; Crick & Dodge, 1994), all of which could shape initial perceptions about the interaction. On the other hand, given that the peer interaction was immediately prior to the initial report, adolescents may not need to draw on their internal working models of attachment to mentally reconstruct the interaction but may instead be more guided by the immediate (nonattachment-related) circumstances of the interaction. Moreover, although Bowlby’s original thinking was that attachment representations would indeed be used in new relationships, he focused on close relationships, particularly with romantic partners and children. Indeed, findings that attachment links reliably to perceptions of fleeting encounters such as the brief laboratory interaction with a stranger examined here would have implications for the important debate within developmental psychology about the “narrow” versus “broad” ways in which attachment might contribute to socioemotional development (Belsky & Cassidy, 1994; see also Goldberg, Grusec, & Jenkins, 1999).

In relation to reconstructed memory for the initial interaction assessed at Time 2 (T2), we did formulate a theoretically and empirically based hypothesis: Adolescents’ perceptions of this interaction changes over an extended period of time as a function of attachment, such that secure adolescents reconstruct their memory for this type of interaction more favorably over time than do insecure adolescents. This hypothesis is based on the idea that a meeting with an unfamiliar peer is, by definition, a novel experience, and adolescents will reconstruct their memory for it schematically over time (i.e., in a way consistent with the knowledge contained in their internal working models of attachment; Dykas & Cassidy, 2011). This hypothesis is also supported by recent observations of attachment-related reconstructive memory processes observed in close relationships, such that secure individuals remembered interactions with parents, children, or romantic partners more favorably than did insecure individuals (Dykas, Woodhouse, Ehrlich, & Cassidy, 2010; Simpson, Rholes, & Winterheld, 2010; see also Feeney & Cassidy, 2003), as well as by data indicating that attachment is linked to individuals’ reconstructive memory processes about emotional reactions to life events (e.g., Gentzler & Kerns, 2006; for a related study, see Pietromonaco & Barrett, 1997). Our study, however, represents a conceptually significant extension beyond this previous work in that we are examining attachment-related reconstructive memory processes outside the context of close personal relationships.

Method

Participants

Participants were 189 adolescents (mean age 16.5 years) who, after participating in a large data collection session at their high schools, expressed interest in participating in laboratory-based data collection sessions with their families and peers (see Woodhouse, Dykas, & Cassidy, in press, for details regarding the larger study). Approximately 20% of students who expressed interest in the laboratory data collection sessions and who met the eligibility requirements (i.e., living in two-parent, English-speaking families), served as participants in the present study). Participants identified themselves as White/Caucasian (73%), Black/African American (14%), Asian (10%), or Hispanic (3%). Participants were paid $125 for their participation in the laboratory sessions.

Measures

Adult Attachment Interview (AAI; George, Kaplan, & Main, 1996).

This widely used semistructured interview assesses the quality of attachment (state of mind with respect to attachment) through a series of questions focused principally on memories of attachment-related experiences during childhood (see Hesse, 2008, for a detailed description of the AAI protocol and a summary of its psychometric properties). We made minor modifications to this interview to make some questions more appropriate for an adolescent population (Ward & Carlson, 1995). Interviews lasted approximately 1 hr and were audiotaped for later verbatim transcription. Four coders (trained by Mary Main and Eric Hesse and unaware of additional participant information) used Main and Goldwyn’s (1998) classification system to rate adolescents’ probable attachment-related experiences and current state of mind with respect to attachment on a series of 9-point scales. Based on an integrated consideration of the transcript’s scale scores, coders assigned one of three principal classifications to the transcript: secure/autonomous, insecure/dismissing, or insecure/preoccupied. In addition, coders classified adolescents as insecure/unresolved if they showed evidence of lack of resolution of the loss or trauma of an attachment figure. When transcripts could not be placed into any other category, they were labeled insecure/cannot classify. Interrater reliability was assessed continuously; a randomly selected 29% of cases (n = 55) were coded by at least two coders (78% agreement, κ = .61). All disagreements on the AAI classification data were resolved by a fifth independent coder who coded no additional data.

Emotional Response to Conflict Scale (ERCS; Cassidy, 1998).

We used this 31-item self-report instrument to assess individuals’ immediate (T1) and follow-up (T2) perceptions (on eight subscales using Likert-type scales with appropriate anchors) of a 10-min laboratory conflict discussion task. Following Feeney and Cassidy’s (2003) analytic procedures, we calculated two summary scores by averaging the Positive Discussion, Own Positive Emotions, Other Positive Emotions, and Positive Treatment subscales (yielding a positive interaction summary score; 16 items; α = .83 for T1, α = .87 for T2) and the Negative Discussion, Own Negative Emotions, and Other Negative Emotions subscales (yielding a negative interaction summary score; 11 items; α = .71 for T1, α = .79 for T2). We also calculated a hostile treatment summary score containing the four items assessing hostile treatment (α = .58 for T1, α = .75 for T2). All summary scores (across both time points) consisted of items that loaded onto a single factor (eigenvalues ranging from 1.80 to 2.93) that accounted for between 45% and 73% of the variance in these scores. The validity of the ERCS has been established in studies reporting theoretically consistent links between attachment representations and reconstructive memory in close relationships (Dykas et al., 2010; Feeney & Cassidy, 2003).

Procedure

As part of their enrollment in a larger study, participants visited our laboratory during the summer following 11th grade. First, they completed the AAI. Next, participants were paired with a same-sex, same-race, unfamiliar peer who attended a different school. Approximately half of the participants were paired with a peer who was also enrolled in the study, and the other half—because of their scheduling constraints—were paired with an unfamiliar peer from another school who was not enrolled in the current study; these nonenrolled unfamiliar peers (n = 55) were drawn from the same pool as the participants to serve solely as interaction partners (paid $25 for their assistance), and they provided no data for this report. No adolescent participated more than once. At the beginning of this visit, the adolescent and unfamiliar peer independently completed a survey in which they rated how much they agreed with statements about controversial topics (e.g., illicit drug use, gun control). An experimenter later compared their responses and identified three topics of disagreement. The adolescent and unfamiliar peer were then introduced and instructed to “present your positions on the issue and then discuss your differences” for 10 min. After completing this task, participants completed the ERCS (T1 ERCS). Two weeks later, participants received a follow-up ERCS in the mail and mailed it back to our laboratory (T2 ERCS). Despite implementing these procedures for collecting data from all participants, sample size reduction resulted from missing data. Data were missing for 1 AAI, for 2 T1 ERCS, for 30 T2 ERCS, and for 4 additional participants, both T1 and T2 ERCS, with a final analysis sample size of n = 153; of note, compared with adolescents who completed the ERCS at T2, adolescents who did not complete the ERCS at T2 did not differ as function of adolescent gender, χ²(1, 188) = .01, p = .93 or attachment, χ²(1, 188) = .13, p = .72. We used maximum-likelihood estimation of available data (see Verbeke & Molenberghs, 2000) to conduct our principal analyses (described below).

Results

Descriptive Statistics

AAI classifications.

The distribution of the adolescent AAI classifications was as follows: 126 secure/autonomous (67%), 44 insecure/dismissing (23%), 10 insecure/preoccupied (5%), 6 insecure/unresolved (3%), and 2 insecure/cannot classify (1%). Given the theoretical rationale of the present study (and because there were few adolescents in the insecure subgroups), we combined these subgroups into one insecure group and examined secure versus insecure AAI group differences.

Principal Analyses

Using SAS/STAT® 9.1 software and its PROC Mixed procedure, we conducted three repeated-measures analyses of variance to examine whether attachment was linked to both attachment-related differences in adolescents’ T1 ERCS summary scores and change in adolescents’ ERCS summary scores obtained over the 2-week period. In each analysis, adolescents’ T1 and T2 ERCS summary scores (which we normalized using a square-root transformation to correct for skewness) served as the repeated factor. The fixed factors were the time at which participants completed the ERCS (i.e., ERCS Time; 2 levels: T1 vs. T2) and Adolescent Attachment (2 levels: secure vs. insecure); we also included a covariate—Adolescent Gender (2 levels: boy vs. girl)—as a fixed factor. Furthermore, we modeled the ERCS Time × Adolescent Attachment first-order interaction to examine our expectation that adolescents’ perceptions of an interaction with an unfamiliar peer would change over time as a function of attachment. When significant interaction effects emerged, we plotted adolescents’ ERCS scores and conducted accompanying post hoc t tests to test our hypothesis about how, on average, adolescents’ perceptions of conflict changed over a span of 2 weeks as a function of adolescent attachment. Given McClelland and Judd’s (1993) assertion that statistical difficulties can hinder the detection of interaction effects, we followed their suggestion and enhanced power by raising the p value when testing these effects (to p < .10). We used the conventional p < .05 for all other analyses not involving interaction effects. We report the measure of effect size as Hedge’s g (Cooper & Hedges, 1994).

In Table 1, we present the means and standard errors for adolescents’ ERCS summary scores as a function of adolescent attachment classifications. With regard to the examination of attachment-related differences in adolescents’ T1 ERCS summary scores, secure and insecure adolescents did not differ for their perceptions of positive interaction, t(151) = 1.04, p = .30; negative interaction, t(151) = 0.67, p = .50; or hostile treatment, t(151) = 0.51, p = .61. However, with regard to change in adolescents’ ERCS summary scores, the expected significant ERCS Time × Adolescent Attachment interaction effects emerged for adolescents’ positive interaction ERCS summary scores, F(1, 151) = 5.40, p = .02; negative interaction ERCS summary scores, F(1, 151) = 2.60, p = .10; and hostile treatment ERCS summary scores, F(1, 148) = 4.51, p = .03. These findings corroborated our expectation that adolescents’ perceptions of an interaction with an unfamiliar peer would change over time as a function of attachment. More precisely, both secure, M_{Trans T1-T2} = −.03, SE = 0.01, t_T2-T1(151) = −2.34, p = .02, g = .20 and insecure, M_{Trans T1-T2} = −.09, SE = 0.02, t_T2-T1(151) = −4.53, p < .001, g = .36 adolescents remembered the conflict as less positive during the 2-week period, although the effect was larger for insecure individuals (see Figure 1). As a result, insecure adolescents’ T2 ERCS positive interaction summary scores, M_Trans = 1.80, SE = 0.02, were significantly lower than secure adolescents’ summary scores, M_Trans = 1.88, SD = 0.02, t(151) = −2.99, p = .003. Moreover, insecure adolescents showed an increase in their propensity to remember the conflict as negative, M_{Trans T1-T2} = .06, SE = 0.02 t_T2-T1(151) = 3.12, p = .002, g = .26, whereas secure adolescents’ memory for the negative features of the conflict remained stable during the 2-week period, M_{Trans T1-T2} = .02, SE = 0.01, t_T2-T1(151) = 1.60, p = .11, g = .10 (see Figure 2). Despite this change in insecure adolescents’ memory over time, insecure and secure adolescents’ T2 ERCS negative summary scores did not differ significantly, t(151) = −0.88, p = .38. Finally, insecure adolescents showed an increase in their propensity to remember being treated with hostility by the unfamiliar peer, M_{Trans T1-T2} = .06, SE = 0.02, t_T2-T1(148) = 3.39, p < .001, g = .25, whereas secure adolescents’ memory for being treated with hostility remained stable during the 2-week period, M_{Trans T1-T2} = .01, SE = 0.01, t_T2-T1(148) = 1.13, p = .26, g = .11 (see Figure 3). Insecure adolescents’ T2 ERCS hostile treatment summary scores were not significantly greater than secure adolescents’ summary scores, t(148) = 1.89, p = .06.

Table 1.

Adolescents’ Mean Emotional Response to Conflict Scale Summary Scores as a Function of Attachment

	Summary score/time
	Positive interaction		Negative interaction		Hostile treatment
Attachment	T1	T2	T1	T2	T1	T2
Insecure	1.89 (.02)	1.79 (.02)	1.15 (.01)	1.21 (.02)	1.03 (.01)	1.09 (.01)
Secure	1.92 (.02)	1.87 (.03)	1.16 (.02)	1.19 (.02)	1.04 (.02)	1.05 (.02)

Note. Transformed means are reported. Standard errors are reported in parentheses.

Figure 1.

Attachment-related changes in adolescents’ positive interaction summary scores as measured by the Emotional Response to Conflict Scale (ERCS; Cassidy, 1998).

Figure 2.

Attachment-related changes in adolescents’ negative interaction summary scores as measured by the Emotional Response to Conflict Scale (ERCS; Cassidy, 1998).

Figure 3.

Attachment-related changes in adolescents’ hostile treatment summary scores as measured by the Emotional Response to Conflict Scale (ERCS; Cassidy, 1998).

Discussion

Study findings reveal attachment-related, short-term longitudinal differences in adolescents’ memories for interactions with unfamiliar peers. Although no attachment-related differences existed for adolescents’ initial perceptions of the interactions (suggesting that such perceptions may be guided initially by immediate, nonattachment-related circumstances of the interaction), the two attachment groups differed in the nature of their change across time for all three dimensions, despite the fact that, to our knowledge, no subsequent social interactions occurred between the two adolescents (who had not known each other previously and who attended different schools). All adolescents remembered the interaction as less positive over time, yet the decline for insecure adolescents was steeper than it was for secure adolescents, resulting in a significant T2 difference between attachment groups for the positive dimension. Interestingly, only insecure adolescents remembered the interaction as characterized by greater negativity and hostility than they had initially perceived.

This pattern of findings lends support to Bowlby’s (1973) proposition that individuals’ internal working models serve to influence information processing about specific social experiences (see also Dykas & Cassidy, 2011). Although these findings converge with those from previous studies reporting links between attachment and memory, most notably those showing that insecure attachment is linked to the negative recall of attachment-relevant social information (e.g., Mikulincer & Horesh, 1999; Rom & Mikulincer, 2003; see Dykas & Cassidy, 2011, for a review of these studies), our study is novel in that no study has yet examined attachment-related differences in memory for initial social encounters with unknown individuals. Indeed, because there were no further interactions between the adolescents, results are consistent with the possibility that after initial perceptions of an interaction with an unfamiliar peer faded, adolescents relied on their internal working models to reconstruct their memories of this interaction. More precisely, insecure adolescents, who are thought to possess negative experience-based internal working models about the self and others (Bowlby, 1973; Mikulincer & Shaver, 2007), could have relied on these negative models to fill in their memory gaps for the interaction in a negatively biased way.

Although our findings are consistent with previous studies showing basic links between attachment and memory in childhood and adulthood (see Dykas & Cassidy, 2011, for a review), they address a question left unanswered in this previous work: Do findings that insecure adults, compared with secure adults, show more negative memories of the past reflect the fact that the past events were originally perceived as more negative, or a reconstructive processes whereby insecure individuals alter their initial perceptions? Moreover, these findings offer a new perspective on previous thinking about peer relations by suggesting that the attachment-related reconstructed memories for past events (rather than simply initial perceptions) may be important to consider when studying adolescents’ interactions with peers. This perspective is a further specification of the thinking of several researchers who have proposed that past social events contribute to future behaviors with peers (e.g., Crick & Dodge, 1994). As such, it is possible that these negatively biased memory reconstructions of insecure adolescents may have long-term consequences for adolescents’ interactions with peers. For example, if insecure adolescents are biased to remember their interactions with peers as more negative and hostile over time, then they may be likely to behave in ways that are less prosocial or more conflictual relative to other adolescents (see Allen, Moore, Kuperminc, & Bell, 1998, and Dykas, Ziv, & Cassidy, 2008, for attachment-related differences in the quality of adolescents’ peer relationships). Future research should examine whether long-lasting effects of attachment-related memory biases contribute to changes in the quality of adolescents’ peer relationships and social status.

Although this study’s results converge considerably with previous research on attachment and reconstructive memory for interactions with parents (e.g., secure adolescents’ reconstructive memories with others are favorable, and insecure adolescents’ memories are less favorable; Dykas et al., 2010), a notable difference also emerged. Dykas et al. (2010) speculated that secure adolescents showed a pattern of “adaptive forgetting” of negativity in their interactions with mothers and fathers, yet no such adaptive forgetting appeared to emerge for secure adolescents with regard to interactions with unfamiliar peers. This apparent lack of adaptive forgetting may occur for at least three reasons. First, there may not have been enough time for such a process to emerge (only 2 weeks passed between the initial unfamiliar peer interaction and the memory recall about the peer experience, whereas 6 weeks passed between the interactions with parents and the later recall). Second, the finding of secure adolescents’ adaptive forgetting about negative experiences with parents may have emerged because of the importance of perceiving attachment figures with whom one lives as positive and supportive; individuals may be less motivated to perceive strangers with such a positive bias (see Flannagan, Marsh, & Fuhrman, 2005). Third, the memories for parent–adolescent interactions may have been associated with subsequent interactions that adolescents had with their parents during the 6-week interval; to our knowledge, adolescents did not interact with their unfamiliar peer partners beyond their unique laboratory experience. Given that memories for past events are affected by subsequent events (Loftus, 2004), it is possible that subsequent parent–adolescent interactions interfered with adolescents’ memories for their perceptions of the earlier laboratory task.

This study is the first to report attachment-related reconstructive memory biases outside the context of significant interpersonal relationships. Adolescents’ interview-based narratives that focused on one domain (i.e., attachment relationships with parents) were linked to reconstructive memory processes related to another domain (i.e., an interaction with an unfamiliar peer). Future studies could examine whether this link can be replicated in adolescents with different characteristics than those in our sample (e.g., in adolescents from single parent families). Researchers could also examine how AAI attachment classifications are linked to perceptions of nonsocial events, compared with social events (e.g., perceptions of events involving solitary activities versus activities with peers; see Gentzler & Kerns, 2006). It will also be important to examine whether these attachment-related reconstructive memory patterns emerge in more “real-life” contexts. In particular, most initial adolescent interactions are unlikely to involve conflict of the sort we attempted to elicit in this laboratory paradigm; we note, however, that there was, in fact, very little actual conflict during this task, with most adolescents instead trying to minimize disagreement, listen to the other perspective, and find common ground. It will be important for future studies to examine more naturalistic controlled contexts as well as contexts outside of the laboratory. If similar reconstructive processes were to emerge from interactions in naturalistic contexts, then attachment-related reconstructive memories could be a factor in new relationship formation, with implications for long-term peer relationship functioning and clinical health outcomes. For example, adolescents who consistently remember peer interactions with a negative bias may be more likely to experience an increase in depressive symptoms. Future studies could also test whether reconstructive memory biases emerge for different insecure classifications (i.e., dismissing and preoccupied adolescents), as well as how personality factors (e.g., extroversion, which meta-analytic data indicate is not linked to attachments; see Roisman et al., 2007) or “social distance” (i.e., how much an adolescent wants to interact with a once-unfamiliar person in the future; see Trope & Liberman, 2010) may influence the links identified in this report. Finally, future researchers could examine how adolescents’ emerging attachments to peers may contribute to reconstructive memory (see Allen, 2008), as well as how attachment-related reconstructive memory processes (and associated mechanisms) predict behavior in future interactions with the same or another peer, and whether ongoing interactions with familiar peers (i.e., events occurring between T1 and T2) could shape memories for the initial experience with an unfamiliar peer.