In longitudinal ERGM, several purely structural effects were found that were common to all four types of friendship networks. Overall, the signs of all parameters were identical for the structural effects across the four types of networks. The reciprocity effect was consistently observed. In a similar vein to previous findings (Hachen et al., 2009), the structure of friendship networks might be rooted in a dyadic and mutual exchange of activity as a basic social function in both face-to-face and text message- mediated communication. Also, the combination of the positive transitive triad and negative 2-path parameters was found regardless of network type.
The differences in parameter significance across the networks also suggest that the purely structural processes played out in different ways in each type of friendship network. The obtained patterns for triad parameters were similar in face-to-face superficial networks, face-to-face selfdisclosing networks, and text message-mediated superficial networks, suggesting that friendship networks evolve through transitive clustering and cyclic-based generalized exchange. This finding is contrary to the
Granovetter’s (1973) notion that weak ties show less transitivity than strong ties. In terms of the social network measure used in this study, it might be possible that the characteristics of superficial (i.e., greeting) and self-disclosing networks that were determined a priori by the researcher did not perfectly reflect the nature of nonintimate and intimate relationships. Thus, first-year undergraduates might include both nonintimate and intimate relationships when answering their greeting networks. Future research should consider using alternative social network questions that can clearly differentiate nonintimate from intimate relationships.
However, text message-mediated self-disclosing networks indicated a unique process of network evolution: first-year undergraduates developed self-disclosing networks via text messages through the formation of transitive clusters but not cyclic clusters. As shown in the reciprocity effect, people tend to disclose themselves to intimate friends at a dyadic level and in a triadic hierarchical manner.
Concerning actor-relation effects of a visible attribute, no systematic effect of gender on the social selection process was observed in face- to-face friendship networks. In other words, the evolution of face-to- face friendship selection was not particularly gender specific. In previous research, gender-based friendship selection emerged only in the initial stage of relationship processes (van Duijn et al., 2003). As suggested in Newcomb’s (1961) seminal work on friendship selection processes, people tend to form connections with accessible others in their immediate environment. In face-to-face settings, male or female activity might be a trigger to make their relationships more proximate, but from a year-long perspective, the purely structural processes and other invisible individual dispositions play a more important role on friendship formation.
In contrast, significant gender effects were observed in text message- mediated friendship networks. In text message-mediated superficial networks, males tended to be more popular than females, suggesting that text messages might promote gender intermixing in terms of the formation of casual and unidirectional relationships from females to males. In self-disclosing friendship networks, the homophily effect of gender was exhibited in the process of network evolution. Also, compared to females, males were more likely to send and receive self-disclosing text messages to and from others. In summary, both males and females prefer empathic bonds with the same gender via text messages, but at the same time, males took positive actions to nominate, and were also nominated by females, through text-based intimate communication. These findings might be explained by the evidence that males are more likely to disclose themselves to strangers and acquaintances, whereas females are more likely to disclose to intimate friends (Stokes, Fuehrer, & Childs,
1980). Although these results are contrary to the findings by Igarashi et al. (2005), this inconsistency can be interpreted according to the significant differences between the current research and Igarashi et al. (2005) with regard to the sample (first-year undergraduates only vs. undergraduates of several grades) and the study period (1 year vs. one semester). In addition, Igarashi et al. used a dyad independent method without containing other structural parameters, a method that may underestimate (or even ignore) regularity in self-organizing features of friendship networks. Moreover, longitudinal ERGM allows us to test social selection processes of gender homophily (male-male, female-female) and heterophily (male- female) in tie formation and dissolution over time when controlling for purely structural effects. In sum, the current findings revealed that forming and maintaining intimate relationships linked via text messages are explained by the social selection process based on gender.
In terms of actor-relation effects of invisible individual dispositions, IDgroup had a significant impact on relationship formation in the process of network evolution. As suggested in the previous literature (Lawler & Yoon, 1996; Paxton & Moody, 2003), social identity is rooted in self-esteem and emotional bonding with others, and would serve as an adaptive motive to form friendship networks in a novice setting. First- year undergraduates identifying with their university immediately after matriculation might behave actively toward others, resulting in the nomination of a greater number of ties through the evolution of friendship networks. However, it is not clear why the significant sender effect of IDgroUP was observed only in face-to-face self-disclosing networks and text message-mediated superficial networks, each located at the opposite end of friendship intimacy. This nonsystematic pattern of findings may be partly due to a lack of salient cues regarding the social identity of others. Although there is evidence that invisible attributes have significant impacts on friendship network development (van Duijn et al., 2003), the application of salient cues for social identity each first-year undergraduate manifests, such as the ownership of university merchandise (e.g., T-shirts, mugs), may provide a more consistent effect of social identity on friendship selection across the four types of networks.
In conclusion, the current longitudinal research provides fruitful evidence of the social selection and purely structural processes through the evolution of friendship networks, in that both structural and individual factors have a great impact on people’s friendship choices in novel social environments. Future research should focus more on event-related effects on friendship network formation and maintenance over time. Also, the examination of other types of micro-macro dynamics, such as group norms and individual knowledge of groups, would expand the validity of the findings reported here. To further understand the dynamics of network evolution with individual factors, it is also intriguing to consider analyzing actual text message records and modeling the multiplex coevolution of different types of friendship networks. For these purposes, stochastic actor-oriented models implemented in SIENA software (Snijders, van de Bunt, & Steglich, 2010; Steglich, Snijders, & West, 2006; van Duijn et al., 2003) would also draw insightful findings.