Experiences with Warmth in Middle Childhood Predict Features of Text-Message Communication in Early Adolescence

Acknowledgement: This research is funded by grants awarded to Marion K. Underwood from the NIMH (R03 MH52110, R29 MH55992, R01 MH63076, K02 MH073616 and R56 MH63076) and the NICHD (R01 HD060995 and R21 HD072165). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies. Portions of this research were presented at the Society for Research on Child Development, Philadelphia, Pennsylvania (March, 2015).

Relationships in middle childhood provide important skills for developing and maintaining peer relationships in early adolescence. Experiences with warmth (i.e., friendly, affectionate behaviors toward others) in particular seem to facilitate the successful navigation of later interpersonal relationships (e.g., Bryant & Conger, 2002). As children reach adolescence, relationships with friends become increasingly close, and romantic relationships begin to form (Laursen & Williams, 1997). Interpersonal behaviors that facilitate the development of intimacy (e.g., via increased displays of positive affect) rather than inhibit it (e.g., via increased duplicity) should therefore be critical to the success of these relationships.

One feature that sets the current generation of adolescents apart from previous generations is their wide access to digital communication. Unlike face-to-face communication, digital communication can occur anywhere that adolescents have access to the Internet or smartphones. Adolescents heavily use text messaging to communicate with peers, and their text messaging communication is rich with emotional tone conveyed in relatively few words (Underwood, Ehrenreich, More, Solis, & Brinkley, 2015). Texting appeals to adolescents because it is discreet, outside the purview of adults, and provides a private forum in which they are able to quickly exchange information (Ehrenreich, Underwood, & Ackerman, 2014; Ling, 2005). However, text messaging also poses risks; harassment via text messaging is the most common and distressing form of cyberbullying (Fenaughty & Harré, 2013). Being the victim of even a single episode of cyber harassment can be devastating for adolescents (Underwood & Ehrenreich, 2017), perhaps in part because they care so deeply about their online social lives (boyd, 2014). Understanding precursors of text messaging communication is therefore important given that adolescents rely heavily on this medium to stay connected to peers (Lenhart, 2015).

The present research investigates links between experiences with warmth in middle childhood and affective states and disagreeable interactions within text-based digital communication in early adolescence. Below we discuss the important role that digital communication plays in early adolescence and describe theoretical frameworks to help explain the transmission of warmth in middle childhood to later text-messaging communication. Because parents and peers likely both play a role in socialization, we introduce a statistical model that helps to disentangle these different sources of behavior and thus provide a more nuanced picture of how warmth in middle childhood comes to be related to text-message communication.

Common forms of digital communication include short message service (aka text-messaging) via mobile phone, instant-messaging (IM) using services such as Snapchat and Facebook messenger, and posts or announcements (aka microblogging) using social networking websites such as Twitter, Instagram, and Facebook. These digital communication forms differ with regards to the typical content exchanged (e.g., text for Twitter, photos for Instagram) and the size of the audience (e.g., whereas Twitter can reach a large audience, IM is generally restricted to one person at a time). Unlike social networking website communication, text-messaging via mobile phone generally involves dyadic exchanges of written content.

Adolescents report preferring text messaging as a way of communicating with friends, even more than face-to-face interactions (Lenhart, Ling, Campbell, & Purcell, 2010). Almost 75% of youth ages 12–17 have access to cell phones (Lenhart, 2015). Whereas only 25% of teens spend time daily with friends outside of school, 55% of teens report spending time every day texting with friends (Lenhart, 2015). Moreover, youth ages 12–17 report sending an average of 60 text messages per day (an increase from an average of 50 in 2009; Lenhart, 2012).

Both the frequency and content of adolescents’ text messaging relate to psychological adjustment. Self-report and observational studies of text messaging show that frequency of texting relates to externalizing symptoms (Ling, 2005), social anxiety (Pierce, 2009), sleep difficulties (Murdock, 2013), and somatic complaints (Underwood et al., 2015). Specific types of text messaging content may also be associated with adjustment. For a sample of 14-year-olds in which the content of text messages was captured and coded, frequency of text messages including negative talk about others predicted overall internalizing problems and anxious depression (Underwood et al., 2015). In this same study, texting about sex correlated with overall internalizing symptoms and somatic complaints for girls (Underwood et al., 2015).

Although negative text messaging content may be associated with maladjustment, frequency of text messaging may also relate positively to the quality of adolescents’ friendships. For a sample of adolescents in the Netherlands, self-reported online communication (e.g., IM, e-mail) was related to perceived closeness with friends (Valkenburg & Peter, 2007). Moreover, Canadian adolescents reported that they relied more on digital communication for contact with close friends than those with whom they had weaker ties (Desjarlais & Willoughby, 2010).

We define warmth in the present research as interpersonal behavior characterized by positive valence and benevolence. Whereas we consider higher levels of warmth to involve nurturing and agreeable behavior (e.g., friendliness, sympathy, and cooperativeness; Moskowitz, 1994; Wiggins, 1979), we consider lower levels of warmth to entail cold and quarrelsome behavior (e.g., cold-heartedness, unresponsiveness, and criticalness; Moskowitz, 1994; Wiggins, 1979; see Wiggins, Trapnell, & Phillips, 1988, for a similar conceptualization of warmth). Our conceptualization of warmth therefore combines features of both the interpersonal circumplex dimension of warmth and the five-factor model (FFM) trait of agreeableness.

Expressions of warmth and quarrelsomeness typically invite similar behaviors (warmth and quarrelsomeness, respectively) from others (see, e.g., Markey, Funder, & Ozer, 2003; Markey & Kurtz, 2006; Markey, Lowmaster, & Eichler, 2010; Sadler, Ethier, Gunn, Duong, & Woody, 2009). Whereas engaging in agreeable behavior predicts greater levels of pleasant affect, engaging in quarrelsome behavior predicts greater levels of unpleasant affect (Moskowitz & Cote, 1995). Furthermore, warmth leads to liking (Fiske, Cuddy, Glick, & Xu, 2002) and acceptance from others (Stinson, Cameron, Wood, Gaucher, & Holmes, 2009), and is thus likely to discourage disagreeable behaviors when interacting with others.

Much of the available empirical evidence on links between offline and online behavior supports a perspective called co-construction theory, that “adolescents are psychologically connected to their online worlds similarly to their offline worlds” (Wright & Li, 2011, p. 1959). Just as they do in face-to-face conversations, adolescents explore issues of identity and sexuality in their chatroom communications (Subrahmanyam, Smahel, & Greenfield, 2006). This utilization of online venues to explore the same developmental issues encountered offline fosters consistency in offline and online behavior. For instance, adolescents’ aggressiveness in offline interactions is positively correlated with their cyber aggression (see Kowalski, Giumetti, Schroeder, & Lattanner, 2014, for a meta-analytic review). Adolescents’ prosocial behavior offline is also correlated with their prosocial communication online (Wright & Li, 2011).

Co-construction theory therefore suggests that warmth in text messaging might follow from warmth in earlier, offline relationships. Figure 1 depicts the conceptual model guiding our research. As can be seen, we expect that experiences of warmth in offline relationships in middle childhood will predict participants’ dispositional expressions of warmth within the platform of text-messaging (as we discuss in more detail below, we plan to partition the variance observed in offline warmth into dispositional and relationship-specific components).

Figure 1 further underscores that adolescents’ dispositional tendencies to express warmth in digital communication will be observed in more positive affect, less negative affect, and less disagreeable interactions within their text-messages. Guided by the work of Watson, Clark, and Tellegen (1988), we define positive affect as feelings of excitement, interest, and enthusiasm that manifest in text-messaging behaviors such as exaggerated punctuation and increased positive hyperbole; we likewise define negative affect as feelings of distress, anger, and disdain that manifest in text-messaging behaviors such as criticalness toward the interaction partner and a tendency to bring up negative events. Last, we define disagreeable interactions as text-message communication characterized by social aggression (i.e., attempts to damage the self-esteem or social status of people outside the conversation that manifest in text-messaging behaviors such as social exclusion, gossip, and negative talk about others), urgency/demandingness (i.e., the need for immediate feedback and continued responses from interaction partners that manifest in text-messaging behaviors such as demands for prompt responses), and duplicity (i.e., lying or deceitful behavior that manifests in intentional attempts to mislead others via text).

The development of early adult romantic relationships (DEARR) model (cf. Bryant & Conger, 2002) posits that individuals’ experiences interacting with their family members early in life should impact characteristics of their early adult romantic relationships by shaping their dispositional behavioral tendencies. In line with the DEARR model, we propose two mechanisms that may account for the transmission of warmth in middle childhood to early adolescence: behavioral continuity and socialization. The behavioral continuity hypothesis posits that individuals’ interactional styles should be maintained over time due to stability in particular traits (Bryant & Conger, 2002; see Roberts & DelVecchio, 2000). The socialization hypothesis predicts that individuals’ interactional styles should be learned via actual experiences with the socializing agents (Bryant & Conger, 2002; see Cui, Conger, Bryant, & Elder, 2002; Kuttler & La Greca, 2004; Ladd, 1999). For individuals in middle childhood, both parents and peers serve as important socialization agents (Vandell, 2000). Satisfactorily addressing the behavioral continuity and socialization hypotheses therefore requires research designs in which target children interact with both parents and peers. By collecting data on both of these relationship types, researchers can investigate how much of the behavior can be attributed to dispositional features of the target children (aka behavioral continuity) and how much is due to the specific relationships that the target children have with their parents and peers (aka socialization).

One useful design that can be used to assess multiple relationships is the one-with-many (OWM) design. In this design, a focal individual is tied to several partners and those several partners are tied only to the focal individual. When working with the OWM, a distinction is made between the “one” (i.e., the focal person who has ties to several partners) and the “many” (i.e., the many partners who have ties to the focal person). Moreover, data can be gathered from the one (e.g., the focal person’s expressions of warmth toward all the partners), the many (e.g., the partners’ expressions of warmth toward the focal person), or both (in which case the design is said to be reciprocal). It is also possible that the “many” can be meaningfully distinguished (e.g., by role such as parent vs. friend). The OWM design differs from the more widely known round-robin designs because the partners (or the “many”) do not interact with one another. Nevertheless, the underlying model of interpersonal behavior is the same.

Kenny, Kashy, and Cook (2006) showed how the social relations model (SRM) can be used to understand the structure of nonindependence generated by an OWM design. Consider a scenario in which a focal child (e.g., Chloe) has ties to her mother and a friend (e.g., Clark). Suppose further that Chloe’s mother and Clark do not interact with one another. The SRM stipulates that dyadic behavior (e.g., Chloe’s warmth toward Clark) is a function of a group mean effect, actor effect, partner effect, and relationship effect. The group mean effect refers to the mean level of behavior present in a group; it captures the existence of a tone or climate (e.g., Chloe and Clark may come from a group for which expressions of high warmth are the norm). The actor effect captures the consistency with which the behavior is expressed across partners (e.g., Chloe may possess a dispositional tendency to be warm with others). The partner effect captures the consistency with which the behavior is received from partners (e.g., Clark may possess a dispositional tendency to elicit warmth from others). Finally, the relationship effect captures the unique adjustment that one person makes to the other after accounting for the group mean and actor and partner effects (e.g., Chloe may be especially warm toward Clark, over and above Chloe’s dispositional tendency to be warm with others and Clark’s dispositional tendency to elicit warmth from others). SRM analyses estimate the variance in these effects across groups, and the relative sizes of these variances speak to the importance of the respective SRM components in driving the behavior.

When data are collected from a reciprocal OWM design with distinguishable members—as would be the case when target children interact with their parents and peers, but the parents and peers do not interact with each other—variance partitioning analyses based on the SRM can illuminate the extent to which target children’s expressions of warmth toward their parents and friends are due to their dispositional tendencies to express warmth toward others (i.e., the target children’s actor effects) or the unique relationships they have with their parents and friends (i.e., the child-to-parent and child-to-friend relationship effects). These analyses can further determine the extent to which parents’ and friends’ expressions of warmth toward the target children are due to the target children’s dispositional tendencies to elicit warmth from others (i.e., the target children’s partner effects) or the unique adjustments that parents and friends make to the target children (i.e., the parent-to-child and friend-to-child relationship effects). However, because there are only three people and the parents and peers do not interact with each other, there are not enough data to estimate the group mean effect or the actor and partner effects for the parents and peers. This approach nevertheless enables a reasonably sophisticated understanding of how experiences with parents and peers contribute to later outcomes.

In addition, OWM analyses can provide insight into whether target children’s dispositional tendencies to express and receive warmth are connected (i.e., generalized reciprocity), as well as whether unique expressions of warmth are reciprocated within the parent–child and friend–child relationships (i.e., parent–child and friend–child dyadic reciprocity). Because research on affective constructs generally finds stronger reciprocity in within-generational (e.g., siblings) than between-generational relationships (e.g., parent–child; Eichelsheim, Dekovic, Buist, & Cook, 2009; Rasbash, Jenkins, O’Connor, Tackett, & Reiss, 2011), we expect to find stronger dyadic reciprocity correlations within friend–child pairs than parent–child pairs in the current research.

The current research uses a prospective multimethod design across 8 years to investigate linkages between actual behavior in middle childhood and early adolescence. Unlike other studies, we include both parents and peers, which permits us to better understand how unique features of these relationships contribute to the manifestation of warmth over and above children’s dispositional tendencies to express and receive warmth. Moreover, by including both the child’s parent and friend in the same study, we are able to evaluate whether connections between observed warmth in middle childhood and digital communication in adolescence are better explained by behavioral continuity or socialization mechanisms.

With this in mind, our research is guided by the following sets of research questions:

• Research Question Set 1. What is the nature of warmth in parent–child and friend–child interactions in middle childhood? Are expressions of warmth primarily driven by the dispositional features of the target child or by the unique features of the target child’s relationships with the parents and peers? Moreover, are expressions of warmth reciprocated?
• Research Question Set 2. Are expressions of warmth in middle childhood connected to increased positive affect, decreased negative affect, and decreased disagreeable interactions in digital communication? If such links exist, are those links better explained by behavioral continuity or socialization mechanisms?

Procedures were approved for the project, Social Aggression: Precursors and Outcomes, by the Institutional Review Board at the University of Texas at Dallas (IRB# 07–36). Participants included 218 typically developing children (and their parents and peers) who were involved in an ongoing longitudinal study examining the antecedents and consequences of social aggression (Underwood, Rosen, More, Ehrenreich, & Gentsch, 2012). This was a community sample wherein the target children were recruited from suburban public schools surrounding a large city. Approximately 50% of the target children were girls, and consistent with the demographics of the population from which they were sampled, most children were White (53.7%), Hispanic (19.3%), and Black (18.8%). At the first wave of the study discussed here, target children were on average 10.04-years-old (SD = 0.43; range from 9 to 12), and their parents reported the following for annual income: 14.2% earned less than $25,000; 22.2% earned between $26,000 and $50,000; 21.7% earned between $51,000 and $75,000; 26.9% earned between $76,000 and $100,000; and 7.1% earned greater than $100,000 (8% did not report income). The median reported income range ($51,000–$75,000) was higher than the median family income for the county where participants were recruited ($48,000; U.S. Census Bureau, n.d.).

Target children were asked to bring their parent and a same-gender, similarly aged friend to the laboratory in the fourth (n = 205), fifth (n = 195), sixth (n = 188), and seventh (n = 173) grade. Parents primarily consisted of target children’s mothers (fourth: 94.3%; fifth: 94%; sixth: 94.7%; seventh: 94.3%). To ensure that latent variables capturing features of the parent–child relationship were comparable across waves, we excluded those participants who did not bring the same parent to the lab visits. Although target children always brought a parent to the visit, they did not always bring a friend (6.9% of participants did not bring a friend in fourth grade; 20.2% in fifth grade; 29.8% in sixth grade; and 38.2% in seventh grade). The majority of participants (72.9%) did not bring the same friend to two or more of the visits over the course of the study. While at the lab, target children interacted with their parents and peers separately across six tasks (for a list of the tasks, see Appendix A on the Open Science Framework [OSF] page for this project: https://osf.io/v4nb3/). Interactions across these tasks were videotaped and later coded for expressions of warmth.

During each of the summers prior to entering Grades 9, 10, and 11, participants completed sessions either in their home or in the laboratory (ninth, n = 172; 10th, n = 180; and 11th, n = 178 grades). At the end of these sessions, participants were provided with BlackBerries with paid service plans, configured to capture all incoming and outgoing text-message communication exchanged on the phone. Participants received a new BlackBerry at the conclusion of each visit, and were compensated $50 for their time. Participants used the phones frequently and heavily, exchanging an average of 151.48 texts during each day of coded content (SD = 175.16). Moreover, the overwhelming majority of text-message communication coded for this research was exchanged with their peers (91.84%). Although participants were not prevented from using other devices, we believe that they used these BlackBerries as their primary phones throughout the study given that the average number of messages exchanged per day was slightly higher than self-reported texting rates identified in other studies (Lenhart, 2012, 2015) and participants reported using the BlackBerry device for texting between “most of the time” and “always” when they were asked (Underwood et al., 2012).

Attrition analyses revealed that 36.9% of participants from the laboratory visit in fourth grade (Time 1) did not have data for the digital communication variables in the 11th grade (Time 7). Although participants’ attrition status was unrelated to their gender and their scores on the digital communication variables, of the 20 tests that investigated links between attrition and warmth, six emerged as statistically significant. Further, the two more strongly powered tests of these six effects suggest that participants who remained in the study from Times 1 to 7 expressed and received less warmth from their friends (though these differences were relatively small; d′s = −.30 and −.33). In sum, there is not much evidence that attrition status was systematically related to the study variables (see Appendix B on the OSF page for this project for more details).

Coding for the behaviors described below was performed by teams of graduate and undergraduate research assistants in the last author’s (MKU) laboratory (32 assistants for warmth and 24 assistants for digital communication). Before coders could provide ratings, they were required to complete weeks-long training (10 weeks for warmth and 8 weeks for digital communication) that involved reviewing the corresponding coding manual and coding practice segments or transcripts until reliability was achieved. Although most segments (80%) were coded by one rater, 20% of transcripts were double coded to assess interrater reliability; given the smaller sample size available to compute interrater reliability, all estimates of interrater reliability that we report below were computed using the entire sample (i.e., the sample includes data from target children who brought different parents to the study across waves).

Observed warmth in middle childhood

Trained coders rated both the target child and her/his interaction partner on warmth for each task at each visit. Coders rated the prominence of positivity/negativity in participants’ communication using a 7-point scale (1 = completely positive, 2 = very positive, 3 = mostly positive, 4 = moderate, 5 = mostly negative, 6 = very negative, 7 = completely negative). They were instructed to consider communication that conveyed friendliness, pleasantness, and kindness as positive, and that which conveyed quarrelsomeness, criticalness, and unresponsiveness as negative. These two forms of communication seem to align closely with items from the “warm-agreeable” (e.g., soft-hearted, accommodating, and kind) and “cold-hearted” (e.g., cruel, unsympathetic, and cold-hearted) subscales from the Interpersonal Adjectives Scales-Revised (Wiggins et al., 1988), respectively (see Appendix C on the OSF project page for additional construct validity evidence). Scores were reverse-coded to represent warmth prior to scale construction and analyses.

Interrater reliability was acceptable across the six tasks at each wave (target children-to-parents: α ranged from .70 to .89, Median = .83; target children-to-friends: α ranged from .85 to .94, Median = .92; parents-to-target children: α ranged from .72 to .91, Median = .84; friends-to-target children: α ranged from .88 to .93, Median = .92). Each task showed low-to-moderate consistency across all four visits (target children-to-parents: α ranged from .45 to .61, Median = .54); target children-to-friends: α ranged from .13 to .33, Median = .24); parents-to-target children: α ranged from .52 to .71, Median = .62); friends-to-target children: α ranged from .18 to .37, Median = .21). Because expressions of warmth were very consistent for participants across tasks (target children-to-parents: α ranged from .84 to .91, Median = .88; target children-to-friends: α ranged from .87 to .93, Median = .90; parents-to-target children: α ranged from .85 to .91, Median = .87; friends-to-target children: α ranged from .87 to .92, Median = .91), we computed averages for warmth across the tasks and used these composites for the analyses.

Digital communication in early adolescence

Daily transcripts from 2-day samples of the target children’s text-message communication in the Fall and Spring semesters of ninth, 10th, and 11th grades were selected for coding. To ensure that sufficient levels of social interaction would be observed, we initially sampled text-message communication from the target children on days near the homecoming football game and Valentine’s Day for the Fall and Spring semesters, respectively. If we could not find text-message communication for participants near these days, the search parameters were expanded until content was found. This resulted in up to 4 days per year, for a total of 12 possible days of coded text content (M = 9.2 days, SD = 2.92 days, range = 1–12 days). Participants who did not have transcripts during a given day, or sent fewer than 10 text messages within the day, were not rated on the scales described below.

Coding focused on messages sent by the participant, but received messages were used for additional context. Transcripts were coded for: positive affect (i.e., cheerful and caring communication; interrater reliability for each day of messages ranged from α = .46 to .79; Median = .68); negative affect (i.e., cold and quarrelsome behaviors; interrater reliability ranged from α = .32 to .85; Median = .63); social aggression (i.e., behaviors expressed to damage another person’s social relationships; interrater reliability ranged from α = .36 to .83; Median = .72); urgency/demandingness (i.e., behaviors indicating the requirement of immediate feedback; interrater reliability ranged from α = .42 to .92; Median = .71); and duplicity (i.e., deceptive behaviors; interrater reliability ranged from α = .59 to .92; Median = .81).

Prior to conducting these analyses, we used the Monte Carlo simulation capabilities of Mplus Version 6.11 (Muthén & Muthén, 1998–2017) to estimate the level of power that we should have to detect small, medium, and large effect sizes with a sample size of 200 for a representative set of models. Power analyses for the OWM model for distinguishable dyads suggest that we should have power of approximately .89 to detect reciprocity correlations ≥ .50. Moreover, power analyses for a model similar to our longitudinal variant of the OWM (namely a model involving longitudinal relations between the OWM for distinguishable dyads in middle childhood and growth-curve parameters for variables in adolescence) suggest that we should have power of approximately .70 to .80 to detect medium-sized relations (e.g., r = .30). Taken together, these analyses suggest that we should have sufficient power to detect medium and large-sized effects.

Table 1 provides the descriptive statistics and zero-order correlations between the ratings of warmth across middle childhood, and Figure 2 illustrates the OWM model for expressions of warmth in a three-person group involving the target child (TC), the target child’s parent (PA), and the target child’s friend (FR). Because the parents and friends did not interact, there are four variables to denote each possible directed dyadic interaction.

As Figure 2 shows, the two dyadic interaction variables in which the target child is emitting the behavior are indicators of the target child’s actor effect latent variable. Moreover, the two dyadic interaction variables in which the target child is receiving the behavior are indicators of the target child’s partner effect latent variable. The covariance between the actor and partner variables estimates generalized reciprocity for the target child. Figure 2 additionally shows that there is unexplained variation in each directed dyadic expression of behavior after the actor and partner effects are accounted for. This unexplained variation represents the relationship effect (plus error) variance. In this model, covariances are freely estimated between the directed relationship effects involving the same two people to capture dyadic reciprocity.

We used structural equation modeling with maximum likelihood estimation within Mplus 7.4 to fit the cross-sectional reciprocal OWM model with distinguishable members to observed warmth at each of the four time points in middle childhood. Table 2 provides (a) the percentage of variance accounted for by the respective OWM components (e.g., how much of the variance in warmth from the target child-to-parent is due to the target child’s actor effect and the target child-to-parent relationship effect); (b) the total variance connected with each variable (e.g., observed warmth from target child-to-parent in the fourth grade); c) the OWM reciprocity coefficients (i.e., generalized reciprocity and the two dyadic reciprocity correlations); and (d) the fit statistics connected with each model. Because RMSEA for small degrees of freedom models may lead to falsely rejecting a model with good fit (Kenny, Kaniskan, & McCoach, 2014), we instead focused on other measures of fit, such as CFI, TLI, and SRMR. With the exception of the SRMR value for Grade 6, all models demonstrated excellent fit based on these other measures.

As can be seen, there was little evidence that target children’s experiences with warmth when interacting with their parents and peers were driven by their dispositional tendencies to express or elicit warmth. Instead, it appears that observed warmth was primarily relationship-specific; put another way, the amount of warmth expressed within the parent–child and friend–child pairings had more to do with the unique relationships the target children had with their partners in these observed interactions than their general tendencies to provide or receive warmth. In addition, the dyadic reciprocity correlation coefficients suggest that these relationship-specific expressions of warmth within the parent–child and friend–child interactions were strongly reciprocated. To evaluate whether the dyadic reciprocity correlations significantly differed between friends and parents, we imposed equality constraints on the parameters and performed chi-square difference tests. These findings revealed that the reciprocal exchange of warmth was stronger in the friend–child relationships than the parent–child relationships in the sixth, Δχ2[1] = 7.76, p = .005; and seventh, Δχ2[1] = 13.25, p < .001 grades.

Longitudinal OWM model of warmth

Given our interest in understanding how general qualities of warmth in middle childhood relate to early adolescent digital communication, we sought to construct a longitudinal OWM model that specified the time-specific OWM parameters as indicators of general OWM parameters (see Figure 3). To accomplish this, the parent–child and friend–child dyadic reciprocity coefficients at each wave were modeled with latent variables that treated the directed relationship effects comprising them as indicators. Second, stable OWM parameters for the target child’s actor effect, the target child’s partner effect, and the dyadic reciprocity between parents and the target children were modeled by treating the time-specific versions of these parameters as indicators.

Because target children typically brought different friends to the study at each wave, indices of friend–child dyadic reciprocity across time were not strongly related; consequently, instead of treating time-specific versions of this variable as indicators of a broader dispositional variable, we freely estimated correlations between the time-specific indices of friend–child dyadic reciprocity across waves. Last, to account for stability in the directed relationship effects, we treated time-specific versions of the relationship effects as indicators of broader dispositional variables (e.g., the stable target child-to-parent relationship effect was indicated by the time-specific versions of this variable at the fourth, fifth, sixth, and seventh grades; note that stable relationship effects are omitted in Figure 3 to prevent clutter). For the sake of parsimony, we chose to impose unit loadings for all indicators of broader variables.

We used insights from the cross-sectional OWM analyses to build up the longitudinal OWM model (e.g., fixing the partner effect variances for Grades 6 and 7 to 0). To be consistent with future analyses using skewed outcomes (e.g., digital social aggression), we used robust maximum likelihood (MLR) estimation for this and all subsequent analyses. The initial model did not run. We decided to fix all small and nonsignificant variances that were being estimated as negative in the current model to 0 (namely the time-specific actor effect for the sixth grade and the stable friend-to-target child relationship effect). This modified model ran successfully and demonstrated excellent fit, χ2(n = 218, df = 95) = 135.68, p = .004, CFI = 0.98, TLI = 0.97, RMSEA = .04, 90% CI [.03, .06], SRMR = 0.07.

Regressing digital communication on OWM components

We examined several features of digital communication across early adolescence (i.e., positive affect, negative affect, social aggression, urgency/demandingness, and duplicity). All features were coded for ninth, 10th, and 11th grade, except for social aggression, which was only coded in 10th and 11th grade. Figure 4 displays the longitudinal measurement model that was used to capture dispositional expressions of these variables across early adolescence, and Figure 5 presents the structural portion of the model connecting warmth in middle childhood to digital communication in early adolescence (indicator variables are omitted to prevent clutter). Given the robust evidence for cross-sectional dyadic reciprocity within the parent–child and friend–child relationships (and the corresponding lack of evidence for cross-sectional actor and partner effects), we chose to use the latent variables for stable parent–child and friend–child dyadic reciprocity as predictors of early adolescent social behavior. Because we assume that the latent variables for friend–child dyadic reciprocity at each wave are tapping into the same process and are thus interchangeable, we imposed equality constraints on the relations between each of these variables and digital communication (see Figure 5). With the exception of urgency/demandingness, Δχ2(3) = 8.28, p = .04, imposing these equality constraints did not significantly worsen model fit (negative affect: Δχ2(3) = 4.64, p = .20; positive affect: Δχ2(3) = 6.85, p = .08; social aggression: Δχ2(3) = 6.31, p = .10; duplicity: Δχ2(3) = 0.63, p = .89).

Table 3 shows the results for analyses wherein the digital outcomes were regressed on the latent dyadic reciprocity variables. Reciprocal exchanges of warmth were unrelated to observed positive affect and social aggression in text-messages. In contrast, reciprocal exchanges of warmth within the parent–child relationship predicted lower levels of urgency/demandingness. Moreover, reciprocal exchanges of warmth within the friend–child relationship predicted lower levels of negative affect and duplicity. (At the request of a reviewer, we conducted supplementary analyses with offline measures of friendship quality in high school and found that, among other results, reciprocal exchanges of warmth predicted increased intimacy and decreased relational aggression; see Appendix D on the OSF project page for more details).

We next evaluated whether experiences with parents and peers had differential effects on the digital outcomes by imposing equality constraints on all of the paths in Figure 5. Using the appropriate tests for MLR (Satorra & Bentler, 1994), chi-square difference tests revealed that with the exception of urgency/demandingness, Δχ2(1) = 6.38, p = .01; there were no such differences in the effects for duplicity, Δχ2(1) = 2.21, p = .14, negative affect, Δχ2(1) = 3.57, p = .06, positive affect, Δχ2(1) = 1.55, p = .21, or social aggression, Δχ2(1) = 0.60, p = .44. As such, a more simplified model indicated that greater reciprocal exchanges of warmth within the parent–child or friend–child relationships in middle childhood are related to fewer observed expressions of negative affect (b = −0.10, SE = .05, p = .037; β’s ranging from −.09 to −.21) and duplicity (b = −0.15, SE = .05, p = .006; β’s ranging from −.12 to −.29) within text-messaging communication in adolescence. Last, we explored possible gender differences in these socialization effects, and we did not find any evidence of such moderation.

Simplified longitudinal analyses

One of the benefits of the longitudinal analyses presented above is that they provide insight into the general relations between reciprocal exchanges of warmth in parent–child and friend–child interactions in middle childhood and different features of digital communication in adolescence. That being said, we acknowledge that our sample size may not be ideal for estimating a model of this complexity. To ensure that our results were not simply an artifact of this more complex longitudinal latent variable model, we specified additional simplified longitudinal models that used the dyadic reciprocity variables from the cross-sectional OWM models at each time point to predict more simplified forms of each of the latent variables for digital communication (for more details, see Appendix E on the OSF project page). With the exception of failing to find evidence that the effects of dyadic reciprocity on urgency/demandingness significantly differed for parent–child and friend–child dyads, the results from these simplified longitudinal analyses showed a similar substantive pattern of results and thus bolster our original set of conclusions.

The present research explored whether experiences with warmth in middle childhood with parents and peers are linked to more positive affect, less negative affect, and less disagreeable interactions in text-message communication in adolescence. The results suggest that observed expressions of warmth in middle childhood are primarily driven by unique features of the parent–child and friend–child relationships. Further, consistent with co-construction theory and the mechanism of socialization, greater exchanges of warmth within these relationships predicted lower levels of negative affect and duplicity within digital communication.

Warmth, constituting one of the two fundamental axes in the interpersonal circumplex (Wiggins, 1979), captures the trait manifestation of communion. People with higher warmth express greater sympathy, listen attentively, and compliment others (Moskowitz, 1994). Warmth has ties to the popular FFM of personality (cf. McCrae & Costa, 1989), and can be conceptualized as a blend of extraversion (i.e., the gregarious features) and agreeableness (i.e., the compassionate features; see DeYoung, Weisberg, Quilty, & Peterson, 2013).

Research investigating the test–retest reliability of the FFM domains and facets found that the facet of warmth displayed one of the highest test–retest reliability coefficients (McCrae, Kurtz, Yamagata, & Terracciano, 2011). Experience-sampling studies have also shown relatively high temporal stability for agreeableness (Brown & Moskowitz, 1998; Moskowitz, 1994) and high cross-situational consistency for agreeableness in communal situations with close friends and acquaintances (Moskowitz, 1994). Research applying the SRM to similar constructs also suggests that warmth is consistent across interaction partners (Eichelsheim et al., 2009; Rasbash et al., 2011). In light of this, we found it surprising that we did not find consistent evidence that observed expressions of warmth have a dispositional basis.

This discrepancy may be partially attributed to differences in the design. Whereas past dyadic research has generally used round-robin SRM designs with families to draw their conclusions, the current research used an OWM design with parents and peers. It could be that family members are more likely to be treated in a similar manner (e.g., each person representing a communal situation; Moskowitz, 1994). Indeed, Moskowitz (1994) observed less cross-situational generality when situations differed in the power and status of interactional partners.

Another potential reason that significant actor and partner effect variances were not found in the present research is that the number of partners was only two. This small number of partners makes it more difficult to detect systematic consistency (see Kenny et al., 2006, p. 248). It is possible that adding more peers and/or family members would have generated more stable actor and partner effects, thus making them easier to detect. Future research should investigate this possibility and evaluate whether behavioral continuity still constitutes a plausible mechanism for the transmission of offline warmth to online venues.

This study suggests that observed expressions of warmth in middle childhood between children and their parents and peers are primarily relationship-specific. Indeed, rather than being driven by dispositional features of the target child, expressions of warmth were primarily a function of the unique interpersonal situations encountered in the parent–child and friend–child relationships. This evidence regarding relatively large relationship effect variances is consistent with work by Eichelsheim et al. (2009) and Rasbash et al. (2011) who have shown that a large amount of the variance in affective constructs is relationship-specific (see also Kenny, 1994).

Relevant to this, the present research also showed that warmth is strongly reciprocated within the parent–child and friend–child relationships. This is consistent with the interpersonal theory principle of complementarity (Markey et al., 2003, 2010; Markey & Kurtz, 2006; Sadler et al., 2009), which posits that warmth invites warmth and quarrelsomeness invites quarrelsomeness. Of note, the strong dyadic reciprocity correlations hint at the existence of potentially influential socialization contexts. Indeed, target children are ostensibly modeling their behavior in response to two key figures in their lives, and the learned responses in these relationships appear to have implications for their later adolescent relationships.

Given past research illuminating potential differences in horizontal versus vertical relationships (e.g., Rasbash et al., 2011), we investigated whether the exchange of warmth within friend–child relationships was stronger than that within parent–child relationships. Consistent with past work, the dyadic reciprocity with friends was stronger than the dyadic reciprocity with parents in the later years of middle childhood. This fits with the notion that reciprocity is especially salient in these egalitarian relationships (cf. Hartup, 1989).

Co-construction theory posits that, “. . . users . . . [create] their [digital] contexts in conjunction with other users, thus influencing and being influenced by the very online culture that they are helping to create” (Subrahmanyam & Smahel, 2011, p. 34). Certain developmental tasks, such as establishing intimate relationships, are carried over from offline to online venues (Reich, Subrahmanyam, & Espinoza, 2012). As such, relevant offline behaviors pertaining to intimacy development should play out similarly online. Supporting the notion that adolescents’ online communication reflects important qualities of their offline social lives, we found that observed qualities of relationships several years earlier predicted adolescents’ text-messaging.

In particular, the results indicate that reciprocal exchanges of warmth within parent–child and friend–child relationships predict less negative affect and duplicity. These findings corroborate research on offline behavior showing that increased expressions of warmth are linked to decreased negative affect (Moskowitz & Cote, 1995). They are also consistent with work showing links between parental warmth and children’s social competence (e.g., MacDonald & Parke, 1984), and research showing that adolescents who have more positive interactions in their friendships have more positive interactions in their dating relationships (Kuttler & La Greca, 2004). Broadly speaking, it seems that reciprocating warmth in middle childhood predisposes individuals to less quarrelsome modes of digital communication.

We also found that reciprocal exchanges of warmth within the parent–child relationship had unique implications for children’s observed urgency/demandingness. Relevant to this, greater parental warmth predicted adolescents’ increased levels of effortful control 2 years later (Eisenberg et al., 2005). Warm relationships with an authority figure may foster emotion regulation by teaching children that they can express and manage emotions without fear of punishment and trust that the other person will be available and supportive. Reciprocal warmth with friends may have been less predictive of urgency/demandingness because those relationships are less enduring. That being said, we must advise caution with over interpreting these results given that we did not obtain similar findings with simplified longitudinal analyses.

In contrast to previous theoretical (cf. MacDonald, 1992) and empirical work (Moskowitz & Cote, 1995; Tracey, 2004), we did not find that reciprocal exchanges of warmth predicted increased positive affect. It is possible that the unique affordances provided by the medium of text-messaging (e.g., the lack of social cues in the platform) altered the manifestation of warmth and its consequences for positive affect. It may also be that warmth has stronger implications for the dampening of negative affect and quarrelsome behavior than it does for the generation of positive affect. Such possibilities should be explored in future research.

Our results also shed some light on the plausibility of behavioral continuity and parental and peer socialization as mechanisms linking offline behavior to online behavior longitudinally. Simply put, the behavioral continuity mechanism did not receive much support. Because the results of the OWM analyses indicated that observed expressions of warmth did not have a strong dispositional basis, behavioral continuity could not constitute a tenable mechanism given that it presupposes a certain degree of consistency across interaction partners that is carried forward across time. As noted previously, however, it is possible that the current OWM design prevented us from detecting significant actor effect variance. Future research with round-robin SRM designs will be needed to determine whether behavioral continuity helps to explain the transmission of warmth in offline relationships to online relationships.

On the other hand, the OWM analyses indicated that observed expressions of warmth have a strong relational basis. The results further suggest that expressions of warmth were reinforced within these relationships and thus that parental and peer socialization constituted plausible mechanisms for the transmission of warmth. When we investigated linkages, we found that experiences with warmth within these relationships related to decreased negative affect and duplicity within text-message communication. During middle childhood, people further develop their relational schemas (Baldwin, 1992) for how to interact with other people. For instance, they are likely to obtain procedural knowledge of modes of relating to others via their interactions with their parents and peers. They are also likely to learn scripts (i.e., “. . . schemas representing situationally appropriate sequences of events,” Baldwin, 1992, p. 463) and apply them to future interactions. Although we believe that these social–cognitive mechanisms help to explain our findings, future research will be needed to evaluate them as possible mediators.

The present research had a number of strengths. Because we had observational measures of warmth and digital communication, we were able to eliminate shared-method variance as a plausible explanation for our results. Moreover, we were able to observe warmth in two different key relationships—parent–child and friend–child—both of which have clear implications for later development (e.g., Hartup, 1989; Vandell, 2000). The longitudinal design enabled us to estimate prospective relations between warmth in middle childhood and digital communication in adolescence. Further, by coding the actual content of adolescents’ text messages, we could obtain less biased assessments of socially undesirable behaviors (e.g., duplicity, social aggression). Finally, use of the one-with-many design permitted us to gain a more detailed picture of the links between earlier relationships and adolescent behavior.

Our research also had several limitations. To begin, our operationalization of warmth combined features of both warmth from the interpersonal circumplex and the agreeableness dimension of the FFM. As such, higher levels of warmth reflected both more nurturing and agreeable behavior, and lower levels reflected both colder and more quarrelsome behavior. However, researchers have argued that the opposite of warmth may be better construed as aloofness or indifference rather than quarrelsomeness or hostility (e.g., Horowitz et al., 2006). Although there is a strong correspondence between warmth and agreeableness (Trapnell & Wiggins, 1990), future research would benefit from measuring them separately and evaluating whether they are differentially related to the outcomes studied in this research.

In addition, our use of a nonexperimental panel design prevents us from being able to make causal inferences. Although we were able to demonstrate prospective relations between the variables, future research will be needed to evaluate whether reciprocating warmth in key relationships actually reduces the degree to which individuals engage in quarrelsome modes of digital communication. Another limitation is that we did not actually assess warmth in digital communication, which would have permitted us to directly evaluate co-construction theory.

Furthermore, by using the OWM design, we could not disentangle all possible sources of variance. Although nonindependence in the “warm” behaviors is modeled in the analyses via the OWM, it does not necessarily account for the fact that the entire network may be similar in terms of the behavior. To our minds, this is what the “group mean” effect is supposed to capture. Unlike the SRM for a four-person group, however, there are not enough data to estimate all relevant parameters for the SRM; in such situations, it is common to omit the group effect variance (Kenny et al., 2006), which we did in the present case. Of note, removing the group mean effect may result in larger estimates for the actor and partner effect variances (cf. Kenny et al., 2006). Perhaps more importantly, because we could not estimate the group mean, we could not evaluate whether general exposure to warmth statistically predicted different features of digital communication (aka observational learning). In addition, because there were no dyadic interactions involving the parents with peers, the directed relationship effects involving the target child as the originator of the behavior are confounded with the partner effect for the other person in addition to error. Similarly, the directed relationship effects involving the target child as the receiver of the behavior are confounded with the actor effect for the other person in addition to error. Dyadic reciprocity correlations are thus confounded with generalized reciprocity of the person who is not the target child and possibly attenuated (cf. Marcus, Kashy, & Baldwin, 2009).

Given the rising availability of online IM applications such as Snapchat and Facebook messenger, it is reasonable to wonder whether adolescents and young adults continue to use text messaging as frequently (see Gerpott, 2015). The most recent survey data available from the PEW Research Center suggest that among smartphone owners ages 18–29 in the United States, 100% use text messaging (Pew Research Center, 2015) whereas only 49% of smartphone owners in this age group use other messaging apps such as WhatsApp or iMessage, and only 17% use applications that automatically delete sent messages such as Snapchat (Duggan, 2015). Furthermore, a study of college students in the United States found that this group uses text messaging more frequently than messaging applications, but views texting as highly similar to messaging applications such as Facebook Messenger and Snapchat (Bailey, Schroeder, Whitmer, & Sims, 2016). Of the participants in this study, 93% reported having unlimited text messaging as part of their cell phone service plans, suggesting that cost may not be a factor leading young people to choose other messaging applications over texting. We therefore believe that text-messaging will continue to be a prominent feature of digital communication in the future.

Notwithstanding these limitations, this research represents the first attempt to link observational qualities of parent–child and peer–child relationships in middle childhood to the actual content of future text-message communication. Using the OWM design, we found that observed expressions of warmth in middle childhood were primarily relationship-specific. In addition, expressions of warmth within the parent–child and friend–child relationships were linked to less quarrelsome modes of digital communication. Taken together, these findings suggest that children make unique adjustments to their parents and friends in terms of warmth, and these unique adjustments are related to how they use text-message communication.

Bailey, S. K., Schroeder, B. L., Whitmer, D. E., & Sims, V. K. (2016, September). Perceptions of mobile instant messaging apps are comparable to texting for young adults in the United States. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 60, 1235–1239. 10.1177/1541931213601288

Baldwin, M. W. (1992). Relational schemas and the processing of social information. Psychological Bulletin, 112, 461–484. 10.1037/0033-2909.112.3.461

boyd, D. (2014). It’s complicated: The social lives of networked teens. New Haven, CT: Yale University Press.

Brown, K. W., & Moskowitz, D. S. (1998). Dynamic stability of behavior: The rhythms of our interpersonal lives. Journal of Personality, 66, 105–134. 10.1111/1467-6494.00005

Bryant, C. M., & Conger, R. D. (2002). An intergenerational model of romantic relationship development. In A. L.Vangelisti, H. T.Reiss, & M. A.Fitzpatrick (Eds.), Stability and change in relationships (pp. 57–82). New York, NY: Cambridge University Press.

Cui, M., Conger, R. D., Bryant, C. M., & Elder, G. H. (2002). Parental behavior and the quality of adolescent friendships: A social contextual framework. Journal of Marriage and Family, 64, 676–689. 10.1111/j.1741-3737.2002.00676.x

Desjarlais, M., & Willoughby, T. (2010). A longitudinal study of the relation between adolescent boys’ and girls’ computer use with friends and friendship quality: Support for the social compensation or rich-get-richer hypothesis?Computers in Human Behavior, 26, 896–905. 10.1016/j.chb.2010.02.004

DeYoung, C. G., Weisberg, Y. J., Quilty, L. C., & Peterson, J. B. (2013). Unifying the aspects of the Big Five, the interpersonal circumplex, and trait affiliation. Journal of Personality, 81, 465–475. 10.1111/jopy.12020

Duggan, M. (2015). Mobile messaging and social media – 2015. Pew Research Center. Retrieved from http://www.pewinternet.org/2015/08/19/mobile-messaging-and-social-media-2015/

Ehrenreich, S. E., Underwood, M. K., & Ackerman, R. A. (2014). Adolescents’ text message communication and growth in antisocial behavior across the first year of high school. Journal of Abnormal Child Psychology, 42, 251–264. 10.1007/s10802-013-9783-3

Eichelsheim, V. I., Dekovic, M., Buist, K. L., & Cook, W. L. (2009). The social relations model in family studies: A systematic review. Journal of Marriage and Family, 71, 1052–1069. 10.1111/j.1741-3737.2009.00652.x

Eisenberg, N., Zhou, Q., Spinrad, T. L., Valiente, C., Fabes, R. A., & Liew, J. (2005). Relations among positive parenting, children’s effortful control, and externalizing problems: A three-wave longitudinal study. Child Development, 76, 1055–1071. 10.1111/j.1467-8624.2005.00897.x

Fenaughty, J., & Harré, N. (2013). Factors associated with distressing electronic harassment and cyberbullying. Computers in Human Behavior, 29, 803–811. 10.1016/j.chb.2012.11.008

Fiske, S. T., Cuddy, A. J. C., Glick, P., & Xu, J. (2002). A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology, 82, 878–902. 10.1037/0022-3514.82.6.878

Gerpott, T. J. (2015). SMS use intensity changes in the age of ubiquitous mobile Internet access—A two-level investigation of residential mobile communications customers in Germany. Telematics and Informatics, 32, 809–822. 10.1016/j.tele.2015.03.005

Grotpeter, J. K., & Crick, N. R. (1996). Relational aggression, overt aggression, and friendship. Child Development, 67, 2328–2338. 10.2307/1131626

Hartup, W. W. (1989). Social relationships and their developmental significance. American Psychologist, 44, 120–126. 10.1037/0003-066X.44.2.120

Horowitz, L. M., Wilson, K. R., Turan, B., Zolotsev, P., Constantino, M. J., & Henderson, L. (2006). How interpersonal motives clarify the meaning of interpersonal behavior: A revised circumplex model. Personality and Social Psychology Review, 10, 67–86. 10.1207/s15327957pspr1001_4

Kenny, D. A. (1994). Interpersonal perception: A social relations analysis. New York, NY: Guilford Press.

Kenny, D. A., Kaniskan, B., & McCoach, D. B. (2014). The performance of RMSEA in models with small degrees of freedom. Sociological Methods & Research, 4, 486–507.

Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New York, NY: Guilford Press.

Kowalski, R. M., Giumetti, G. W., Schroeder, A. N., & Lattanner, M. R. (2014). Bullying in the digital age: A critical review and meta-analysis of cyberbullying research among youth. Psychological Bulletin, 140, 1073–1137. 10.1037/a0035618

Kuttler, A. F., & La Greca, A. M. (2004). Linkages among adolescent girls’ romantic relationships, best friendships, and peer networks. Journal of Adolescence, 27, 395–414. 10.1016/j.adolescence.2004.05.002

Ladd, G. W. (1999). Peer relationships and social competence during early and middle childhood. Annual Review of Psychology, 50, 333–359. 10.1146/annurev.psych.50.1.333

Laursen, B., & Williams, V. A. (1997). Perceptions of interdependence and closeness in family and peer relationships among adolescents with and without romantic partners. In S.Shulman & W. A.Collins (Eds.), New directions for child development: Romantic relationships in adolescence: Developmental perspectives (No. 78, pp. 3–20). San Francisco, CA: Jossey-Bass. 10.1002/cd.23219977803

Lenhart, A. (2012). Teens, smartphones, & texting. Retrieved from http://www.pewinternet.org/wp-content/uploads/sites/9/media/Files/Reports/2012/PIP_Teens_Smartphones_and_Texting.pdf

Lenhart, A. (2015). Teen, social media, and technology overview. Retrieved from http://www.pewresearch.org/wp-content/uploads/sites/9/2015/04/PI_TeensandTech_Update2015_0409151.pdf

Lenhart, A., Ling, R., Campbell, S., & Purcell, K. (2010). Teens and mobile phones. Retrieved from http://www.pewinternet.org/2010/04/20/teens-and-mobile-phones/

Ling, R. (2005). Mobile communications vis-à-vis teen emancipation, peer group integration, and deviance. In R.Harper, L.Palen, & A.Taylor (Eds.), The inside text: Social, cultural, and design perspectives on SMS (pp. 175–193). Amsterdam, the Netherlands: Springer. 10.1007/1-4020-3060-6_10

MacDonald, K. B. (1992). Warmth as a developmental construct: An evolutionary analysis. Child Development, 63, 753–773. 10.2307/1131231

MacDonald, K., & Parke, R. D. (1984). Bridging the gap: Parent-child play interaction and peer interactive competence. Developmental Psychology, 55, 1265–1277. 10.2307/1129996

Marcus, D. K., Kashy, D. A., & Baldwin, S. A. (2009). Studying psychotherapy using the one-with-many design: The therapeutic alliance as an exemplar. Journal of Counseling Psychology, 56, 537–548. 10.1037/a0017291

Markey, P. M., Funder, D. C., & Ozer, D. J. (2003). Complementarity of interpersonal behaviors in dyadic interactions. Personality and Social Psychology Bulletin, 29, 1082–1090. 10.1177/0146167203253474

Markey, P. M., & Kurtz, J. E. (2006). Increasing acquaintanceship and complementarity of behavioral styles and personality traits among college roommates. Personality and Social Psychology Bulletin, 32, 907–916. 10.1177/0146167206287129

Markey, P. M., Lowmaster, S., & Eichler, W. (2010). A real-time assessment of interpersonal complementarity. Personal Relationships, 17, 13–25. 10.1111/j.1475-6811.2010.01249.x

McCrae, R. R., & Costa, P. T., Jr. (1989). The structure of interpersonal traits: Wiggins’s circumplex and the five-factor model. Journal of Personality and Social Psychology, 56, 586–595. 10.1037/0022-3514.56.4.586

McCrae, R. R., Kurtz, J. E., Yamagata, S., & Terracciano, A. (2011). Internal consistency, retest reliability, and their implications for personality scale validity. Personality and Social Psychology Review, 15, 28–50. 10.1177/1088868310366253

Moskowitz, D. S. (1994). Cross-situational generality and the interpersonal circumplex. Journal of Personality and Social Psychology, 66, 921–933. 10.1037/0022-3514.66.5.921

Moskowitz, D. S., & Cote, S. (1995). Do interpersonal traits predict affect? A comparison of three models. Journal of Personality and Social Psychology, 69, 915–924. 10.1037/0022-3514.69.5.915

Murdock, K. K. (2013). Texting while stressed: Implications for students’ burnout, sleep, and well-being. Psychology of Popular Media Culture, 2, 207–221. 10.1037/ppm0000012

Muthén, L. K., & Muthén, B. O. (1998–2017). Mplus user’s guide (6th edition). Los Angeles, CA: Author.

Pew Research Center. (2015). The smartphone difference. Retrieved from http://www.pewinternet.org/2015/04/01/us-smartphone-use-in-2015/

Pierce, T. (2009). Social anxiety and technology: Face-to-face communication versus technological communication among teens. Computers in Human Behavior, 25, 1367–1372. 10.1016/j.chb.2009.06.003

Rasbash, J., Jenkins, J., O’Connor, T. G., Tackett, J., & Reiss, D. (2011). A social relations model of observed family negativity and positivity using a genetically informative sample. Journal of Personality and Social Psychology, 100, 474–491. 10.1037/a0020931

Reich, S. M., Subrahmanyam, K., & Espinoza, G. (2012). Friending, IMing, and hanging out face-to-face: Overlap in adolescents’ online and offline social networks. Developmental Psychology, 48, 356–368. 10.1037/a0026980

Roberts, B. W., & DelVecchio, W. F. (2000). The rank-order consistency of personality traits from childhood to old age: A quantitative review of longitudinal studies. Psychological Bulletin, 126, 3–25. 10.1037/0033-2909.126.1.3

Sadler, P., Ethier, N., Gunn, G. R., Duong, D., & Woody, E. (2009). Are we on the same wavelength? Interpersonal complementarity as shared cyclical patterns during interactions. Journal of Personality and Social Psychology, 97, 1005–1020. 10.1037/a0016232

Satorra, A., & Bentler, E. M. (1994). Corrections to test statistics and standard errors in covariance structure analysis. In A.von Eye & C. C.Clogg (Eds.), Latent variables analysis: Applications for developmental research (pp. 399–419). Thousand Oaks, CA: Sage.

Stinson, D. A., Cameron, J. J., Wood, J. V., Gaucher, D., & Holmes, J. G. (2009). Deconstructing the “reign of error:” Interpersonal warmth explains the self-fulfilling prophecy of anticipated acceptance. Personality and Social Psychology Bulletin, 35, 1165–1178. 10.1177/0146167209338629

Subrahmanyam, K., & Smahel, D. (2011). Digital youth: The role of media in development. New York, NY: Springer. 10.1007/978-1-4419-6278-2

Subrahmanyam, K., Smahel, D., & Greenfield, P. (2006). Connecting developmental constructions to the internet: Identity presentation and sexual exploration in online teen chat rooms. Developmental Psychology, 42, 395–406. 10.1037/0012-1649.42.3.395

Tracey, T. J. G. (2004). Levels of interpersonal complementarity: A simplex representation. Personality and Social Psychology Bulletin, 30, 1211–1225. 10.1177/0146167204264075

Trapnell, P. D., & Wiggins, J. S. (1990). Extension of the Interpersonal Adjective Scales to include the Big Five dimensions of personality. Journal of Personality and Social Psychology, 59, 781–790. 10.1037/0022-3514.59.4.781

Underwood, M. K., & Ehrenreich, S. E. (2017). The power and the pain of adolescents’ digital communication: Cyber victimization and the perils of lurking. American Psychologist, 72, 144–158. 10.1037/a0040429

Underwood, M. K., Ehrenreich, S. E., More, D., Solis, J. S., & Brinkley, D. Y. (2015). The BlackBerry Project: The hidden world of adolescents’ text messaging and relations with internalizing symptoms. Journal of Research on Adolescence, 25, 101–117. 10.1111/jora.12101

Underwood, M. K., Rosen, L. H., More, D., Ehrenreich, S. E., & Gentsch, J. K. (2012). The BlackBerry project: Capturing the content of adolescents’ text messaging. Developmental Psychology, 48, 295–302. 10.1037/a0025914

U.S. Census Bureau. (n.d.). Small area income and poverty estimates. Retrieved from https://www.census.gov/programs-surveys/saipe.html

Valkenburg, P. M., & Peter, J. (2007). Preadolescents’ and adolescents’ online communication and their closeness to friends. Developmental Psychology, 43, 267–277. 10.1037/0012-1649.43.2.267

Vandell, D. L. (2000). Parents, peer groups, and other socializing influences. Developmental Psychology, 36, 699–710. 10.1037/0012-1649.36.6.699

Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. 10.1037/0022-3514.54.6.1063

Wiggins, J. S. (1979). A psychological taxonomy of trait-descriptive terms: The interpersonal domain. Journal of Personality and Social Psychology, 37, 395–412. 10.1037/0022-3514.37.3.395

Wiggins, J. S., Trapnell, P., & Phillips, N. (1988). Psychometric and geometric characteristics of the Revised Interpersonal Adjective Scales (IAS-R). Multivariate Behavioral Research, 23, 517–530. 10.1207/s15327906mbr2304_8

Wright, M. F., & Li, Y. (2011). The associations between young adults’ face-to-face prosocial behaviors and their online prosocial behaviors. Computers in Human Behavior, 27, 1959–1962. 10.1016/j.chb.2011.04.019

Submitted: August 7, 2017 Revised: August 22, 2018 Accepted: August 23, 2018