Misery Does Not Love Company

Abstract

Conventional wisdom holds that friends protect against depression through the social support they provide; however, depression likely has a role in structuring friendship networks. In particular, we investigate friend selection mechanisms responsible for similarity in depression among friends (i.e., homophily). Preference is one explanation, yet several correlates of depression make homophilous selection among depressed individuals unlikely. We propose two alternative mechanisms—avoidance and withdrawal—that can produce depression homophily in the absence of preference. These alternative mechanisms create homophily indirectly by limiting friendship partners available to depressed individuals. We test the preference, avoidance, and withdrawal mechanisms using data from the National Longitudinal Study of Adolescent Health and a dynamic network model. Results provide support for the withdrawal mechanism. These findings help explain how depression affects friend selection and have broader implications for understanding selection mechanisms responsible for network patterns such as homophily.

Keywords

depression health homophily social networks adolescents

Research documents numerous pathways through which social networks affect individual well-being (Berkman and Glass 2000; Umberson, Crosnoe, and Reczek 2010). Networks offer support to buffer the impact of stress on health (House 1981), provide norms regarding health behavior (Berkman et al. 2000), and serve as the infrastructure for disease transmission (Morris 2004). The association between networks and health takes on added complexity when one considers that networks themselves are constituted through individuals’ actions. Health conditions that affect individual behavior can thus have broad implications for network structure. Scholars have long acknowledged important consequences of social structure for health (e.g., House, Umberson, and Landis 1988; Link et al. 1989). Only recently, however, has the link from health to social networks received concerted attention (Cornwell 2009; Crosnoe, Frank, and Mueller 2008; Haas, Schaefer, and Kornienko 2010).

The current study examines how health affects social networks by focusing on one of the most common mental health issues confronting adolescents—depression. Scholars estimate the prevalence of major depression among adolescents at 15 percent, with up to 40 percent experiencing depression at subclinical levels (Avenevoli et al. 2008; Compas, Ey, and Grant 1993). Subclinical depression (Cantwell and Baker 1991) is serious enough to contribute to negative outcomes.¹ Most prior research identifies aspects of social networks that serve as protective or risk factors for depression. For instance, support from friends often provides a protective function against depression (Herman-Stahl and Petersen 1996; Lin, Dean, and Ensel 1986); having more friends can thus improve mental health (Ueno 2005). Still, not all friends provide support (Wellman and Wortley 1990) and having too many friends may increase one’s risk for depression (Falci and McNeely 2009). Researchers have paid less attention to the converse link, that is, how depression may affect friendship network structure. Nevertheless, several correlates of depression affect the nature and quality of interpersonal interactions, and likely have implications for the friend selection process (van Zalk et al. 2010).

In examining how depression affects network structure, we are particularly interested in uncovering mechanisms responsible for similarity in depression among friends, that is, homophily (Hogue and Steinberg 1995; Rosenblatt and Greenberg 1991; Segrin 2004). Homophily is one of the strongest patterns of human association and is important to investigate because, in part, of how it restricts the flow of ideas and other resources across society (McPherson, Smith-Lovin, and Cook 2001). In the case of mental health, homophily suggests that individuals with mental health problems are reliant upon friends with similar problems, friends who may not have the capacity to offer needed assistance or support. For relatively healthy individuals, homophily shields them from others with mental health problems.

Homophily on a variable attribute, such as depression, could arise through processes of selection or socialization (Kandel 1978). Scholars have found some evidence of socialization on depression (Hogue and Steinberg 1995; Prinstein 2007; van Zalk et al. 2010), although others conclude that friend selection is most likely responsible for homophily (Segrin 2004). Our interest is in disentangling selection mechanisms responsible for depression homophily.² In particular, we focus on homophily that exceeds the level expected by chance based on structural factors that constrain interactions to be homogenous (e.g., population composition [McPherson et al. 2001]). Such homophily is often assumed to be preference-based, but we argue it can appear without preference for homophily.

Despite the adage that “misery loves company,” there are several reasons why depressed individuals may not prefer to associate with one another. We propose two alternative mechanisms that can lead to homophily. First is homophily through avoidance, which occurs when individuals are excluded by others because of a particular characteristic (e.g., depression). Exclusion restricts one’s friendship options to other excluded individuals, resulting in homophily. Second is homophily through withdrawal from social contact, brought on by a condition such as depression. When individuals withdraw from friendships, they become marginalized in the broader friendship network, which excludes them from the normative network processes that help build and sustain friendships (e.g., depressed adolescents may be less integrated into peer groups). Consequently, marginalized individuals develop friendships with others who are similarly marginalized (e.g., others who are depressed), resulting in homophily.

In the spirit of analytical sociology (Hedström and Bearman 2009), we identify explicit mechanisms linking individual behavior to patterns of social behavior (Hedström and Swedberg 1998). Although our empirical focus is on depression, the mechanisms we propose can be applied to other dimensions (e.g., behaviors or attributes) where network selection differs between actors with high versus low values on the dimension. This represents middle-range theory (Merton 1968) or a micro-macro model (Raub, Buskens, and van Assen 2011) that is general enough to explain homophily on many dimensions.

Mechanisms Of Selection Homophily

Two classes of theories explain homophily through selection. The first is rooted in features of social structure that create homogeneity in social interactions and thereby provide more opportunities for relationships among similar individuals (McPherson et al. 2001). The most basic structural factor is population distribution, which determines how much homophily would be expected by chance (i.e., baseline homophily). Another aspect of structure is the correlation of individual attributes in the population. Homophily can appear spuriously due to consolidation if the dimension in question is correlated with another dimension on which selection operates (Blau 1977). For example, because females experience depression at higher rates than do males (Lewinsohn et al. 1993), a preference for sex homophily could create depression homophily as a byproduct. Finally, substructures or foci, such as schools, workplaces, and churches, encourage homophilous associations by bringing similar sets of individuals into a common setting where friendships can develop (Feld 1981). The second explanation for homophily emphasizes factors that lead to a preference for friends similar to oneself. For instance, homophilous friendships facilitate communication (Rogers and Bhowmik 1970), promote identity reinforcement (Byrne 1971), and reduce uncertainty (Prisbell and Andersen 1980). Individuals may prefer homophilous friendships simply because they are easier to maintain and provide greater satisfaction.

We do not discount structural and preference-based explanations; rather, we suggest they are incomplete and may not account for selection homophily on all dimensions. Homo-phily on an attribute like depression requires that individuals at both ends of the continuum disproportionately associate with similar others (e.g., non-depressed individuals must have non-depressed associates and depressed individuals must have depressed associates). Oftentimes, however, individuals display different interpersonal behaviors or are perceived differently by others depending on their value on the attribute. In such cases, individuals situated at one end of the continuum may be more attractive in general than individuals at the other end. In the case of depression, there is strong theoretical justification to expect non-depressed individuals to prefer non-depressed friends; but, as we will explain, several aspects of depression raise doubts about whether depressed individuals truly prefer friends who are also depressed. We begin by reviewing evidence for why homophily on depression might arise through preferences. We then describe two alternative mechanisms that can help explain homophily among depressed individuals in the absence of preference.

Preference as a Homophily Mechanism

Scholars have advanced several theories to explain why depressed individuals would prefer homophilous friendships. Schachter (1959:24) proposed that individuals in stressful situations prefer the co-presence of others who are facing similar threats, concluding that “misery doesn’t love just any kind of company, it loves only miserable company.” One explanation for this finding is that others experiencing similar circumstances provide relevant information that helps alleviate uncertainty (Kulik and Mahler 2000). Another possibility is that homophilous relationships offer greater rewards with fewer emotional costs. For instance, depressed youth provide more compassion and understanding than do their non-depressed counterparts (Rook, Pietromonaco, and Lewis 1994). In terms of costs, depressed adolescents generally feel more anxious and hostile after interacting with non-depressed peers (Rosenblatt and Greenberg 1991). By contrast, depressed friends are less critical during interactions (Baker, Milich, and Manolis 1996) and less wary of the stigma associated with depression (Link et al. 1989). If depressed friends are better able to fulfill needs of depressed adolescents, with fewer costs, then homophily on depression may be preferred by both depressed and non-depressed individuals. Figure 1a depicts this preference-based mechanism.

Figure 1.

Mechanisms of Homophily

Alternative Homophily Mechanisms

Several behavioral correlates of depression may make depressed individuals less attractive as friends, even to one another. If this is the case, yet homophily on depression still exists, then a process other than preference must be responsible. We propose two such mechanisms: an alter-based mechanism, whereby others avoid depressed individuals, and an ego-based mechanism in which depressed individuals withdraw from friendships. The next sections explain how behaviors that accompany depression can affect network dynamics through avoidance and withdrawal and, in turn, lead to depression homophily.

Avoidance

Depressed individuals, for several reasons, might be avoided by their peers when it comes to forming new friendships or rejected by peers in established friendships. This ultimately leads to depression homophily through the process shown in Figure 1b. To begin, mental illness carries a stigma that can decrease one’s attractiveness as a friend (Link and Phelan 2001; Lucas and Phelan 2010; Martin et al. 2007). In addition, behaviors that accompany depression can produce negative reactions in peers (e.g., negative affect induction [Coyne 1976a, 1976b]). For example, depressed youth make more negative statements to friends (Segrin and Flora 2000), laugh less often (Rook et al. 1994), and exhibit poor interpersonal problem-solving skills, such as hostility in challenging peer situations (Rudolph, Hammen, and Burge 1994). Furthermore, depressed adolescents, especially females, are more likely to negatively dwell or ruminate about events (Jose and Brown 2008) and to seek excessive reassurance and negative feedback from others (Borelli and Prinstein 2006).

In turn, these behaviors produce negative responses in depressed youths’ friends and interaction partners. Interactions with a depressed partner can generate more negative emotions and less enjoyable experiences compared to interactions with non-depressed partners (Baker et al. 1996; Rockhill et al. 2007; Rudolph et al. 1994). In line with the greater demands placed on them, friends generally provide higher levels of support to their depressed peers (Daley and Hammen 2002). Yet depressed adolescents often believe they do not receive adequate support from friends (Rudolph, Flynn, and Abaied 2008). This discrepancy reflects the higher costs of maintaining a relationship with a depressed friend, which can ultimately threaten the relationship. Indeed, friendships with depressed youth are characterized by less stability (Chan and Poulin 2009) and lower quality in terms of companionship, support, and validation (Brendgen et al. 2002). When interactions with depressed youth are too emotionally intense or demand too much time and energy, potential partners may avoid depressed youth as friends.

Social exclusion leads to anxiety (Baumeister and Tice 1990) and discomfort resembling physical pain (Eisenberger, Lieberman, and Williams 2003). One response to the friendship deficit caused by exclusion is to adjust one’s friendship standards to be more inclusive. Potential partners who were previously deemed undesirable become more attractive and are treated more generously (Maner et al. 2007; Pennebaker et al. 1979). As individuals who others avoid lower their standards, they become less likely to avoid others, including others with that same unattractive attribute. In the case of depression, depressed individuals become more open to friendships with others who are also depressed. The result is homophily on depression among depressed individuals.

The avoidance mechanism differs from preference homophily in that similarity is not the motivation behind the friendship. Non-depressed individuals prefer non-depressed friends because they provide the greatest rewards (or fewest costs), not because such friends are similar to themselves. Among depressed individuals, friendships with non-depressed peers are also the most rewarding and hence most desirable. But because such peers reject them, depressed individuals lower their standards and accept friendships with others who are depressed. Depressed individuals’ partner selection behavior is thus at odds with their preferences (van Straaten et al. 2009).

Withdrawal

Depression homophily can also emerge due to the social withdrawal that accompanies depression, as shown in Figure 1c. Subclinical depression is characterized by somatic signs (e.g., fatigue, loss of energy and appetite, hypersomnia or insomnia, and somatic complaints), psychological symptoms (e.g., self-criticism, self-doubt, and a lack of interest in normative activities), and emotional dysregulation (e.g., sadness, guilt, irritability, and agitation). Each of these factors can reduce one’s investment in forming and maintaining friendships. Depressed individuals may also perceive a stigma associated with their mental state, in which case withdrawal can be a means to avoid negative interactions (e.g., devaluation and discrimination) and preserve their sense of self (Link et al. 1989). Finally, because depressed individuals tend to view themselves as less competent when providing emotional support to others (Rudolph, Hammen, and Burge 1997), they may withdraw from relationships rather than receive support and incur an obligation they cannot repay (Greenberg and Shapiro 1971).

Depression-related withdrawal does not by itself create depression homophily. Withdrawal reduces the number of friends, but it does not directly affect the distribution of depression among one’s friends (i.e., depressed individuals may have fewer depressed and non-depressed friends). Instead, withdrawal leads to homophily as depressed individuals come to occupy more marginalized network positions where they have less access to peers through the network processes that provide friendship opportunities and reinforcement. For instance, withdrawal makes one less likely to reciprocate others’ friendship gestures. Also, fewer friends generally means fewer opportunities for transitivity (i.e., friendships with a friend of a friend) because fewer peers will be accessible through one’s friends. Finally, individuals with fewer ties are less likely to have friendships through popularity, which occurs when individuals with many friends are more likely to receive future friendship gestures (i.e., the Matthew Effect). We refer to these three processes as normative because they are found in a wide range of friendship networks (Rivera, Soderstrom, and Uzzi 2010; Snijders, van de Bunt, and Steglich 2010).

Because depressed individuals are less embedded in the friendship network, they are more likely to have friendships outside of normative network processes. This leads to friendships with others who are also marginalized, including depressed peers. Although marginalized individuals may encounter one another less often than they encounter peers who are more prominent in the network, they find one another more amenable to friendship. Depressed individuals may prefer non-depressed friends, but such peers are inaccessible. Consequently, homophily emerges as an unintended consequence of the friend selection processes.

Current Study: Evaluating Homophily Mechanisms

Preference, avoidance, and withdrawal are each associated with a specific pattern of network change over time. Identifying these patterns requires differentiating changes in outgoing ties (i.e., friend selection) from incoming ties (i.e., being selected by others). Preference homophily is indicated by a tendency for individuals to select friends whose level of depression is similar to their own. Avoidance is indicated by depressed individuals having fewer incoming ties over time relative to non-depressed individuals. Withdrawal is indicated by depressed individuals having fewer outgoing ties over time relative to non-depressed individuals. Each of these patterns is represented by an effect in our statistical model. Concluding that one of these mechanisms is operating requires that the representative effect persist net of the other selection mechanisms and controls described below.

Any evaluation of selection mechanisms resulting in homophily must control for structural sources of homophily and other friend selection processes. To adjust for the distribution of depression in the network, we measure similarity relative to the average level of similarity across all possible dyads in the network (i.e., baseline homophily). We control for consolidation by including selection effects for several attributes related to friendship and depression (e.g., sex). Estimates of effects of depression are thus net of effects of characteristics associated with depression (e.g., we estimate preference for depression homophily net of preference for sex homophily). Our evaluation must also consider friend selection through normative network processes. Such processes must be present for the withdrawal mechanism to produce homophily as shown in Figure 1c. In addition, network processes can amplify effects of other selection processes (Wimmer and Lewis 2010). For example, a preference for homophily may lead one to select alters with similar levels of depression. Transitivity could then create a tie between those alters, resulting in a depression homophilous relationship without that being the alters’ intent. Controlling for effects of normative network processes ensures that other estimates are unbiased (Moody 2001; Mouw and Entwisle 2006).

Method

Data

We use data from the National Longitudinal Study of Adolescent Health (Add Health), a nationally representative sample of U.S. adolescents. Our analysis draws on in-home interviews conducted during the first two waves of data collection. We measure depressive symptoms and friendship networks during Wave I, and friendship networks again one year later in Wave II. To assess social network changes, it is critical to have high coverage of the target population, which in our case means all students within a school. Our sample thus consists of students who attended one of 16 saturated schools, where all students were targeted for network data collection during Waves I and II. We exclude one of the saturated schools from our analysis because of a low response rate (less than 75 percent).³ We exclude middle schools and high school seniors from the analysis due to their high rates of attrition. The sample consists of 1,820 9th to 11th grade students in nine high schools.

Measures

Individual attributes

Add Health measured depression using 19 items from the Center for Epidemiologic Studies Depression Scale (CES-D). Questions asked whether students had experienced particular depressive symptoms within the past week, with responses ranging from 0 to 3 (see Part A of the online supplement [http://asr.sagepub.com/supplemental]). We calculated a continuous measure of depression by reverse-coding the appropriate items and averaging scores (α = .86).⁴

To prevent spurious results due to consolidation, we control for the following individual characteristics that are correlated with depression and salient to the friend selection process: sex (0 = male, 1 = female), age, race/ethnicity, parent’s education, BMI, puberty, health, and same-sex attraction. We coded race/ethnicity as Hispanic, White, Black, Native American, Asian, and other. We calculated parent’s education as the highest level of education achieved by the more highly educated parent. Categories include 0 = never went to school, 1 = less than high school, 2 = high school graduate or completed GED, 3 = business/vocational/trade school graduate or some college, 4 = college graduate, and 5 = professional training beyond four-year degree. We calculated BMI the traditional way ([weight in pounds × 703] / height in inches squared) and then, for scaling purposes, divided the result by 10. We calculated puberty for females using three items based on breast development, curviness, and menstrual period. For males, three items inquired about underarm hair, facial hair, and lowered voice. To make the male and female items comparable, we transformed responses to range continuously from 0 to 1 and took their sum (see Part A of the online supplement for exact wording and transformations). The survey measured health by asking students to evaluate their health from “excellent” to “poor” on a five-point scale. Same-sex attraction was measured by asking respondents if they have ever had a romantic attraction to someone of the same sex (0 = no, 1 = yes). Table 1 describes the distribution of our sample on these measures for each school and overall.

Table 1.

Descriptive Statistics for Individuals and Schools, by School and Overall

	School
	1	2	3	4	5	6	7	8	9	Total
Individual^a
Female (%)	56.3	38.5	53.3	43.0	73.7	47.2	45.7	51.1	53.1	50.1
Age	15.4	15.4	15.4	15.8	15.3	16.5	15.3	17.0	15.8	15.4 (1.3)
Race/Ethnicity (%)
Hispanic	4.3	1.9	0	.8	5.3	.8	1.4	40.1	1.6	19.5
White	95.7	98.1	98.3	98.3	48.7	98.5	95.7	4.7	98.4	52.4
Black	0	0	0	.8	43.4	0	1.4	21.8	0	12.1
Asian	0	0	1.7	0	1.3	0	1.4	31.4	0	14.9
Native American	0	0	0	0	0	.4	0	.7	0	.4
Other	0	0	0	0	1.3	.2	0	1.3	0	.7
Parent’s Education	3.4	2.8	2.4	2.2	2.7	3.0	2.3	2.8	3.9	2.9 (1.2)
Puberty	1.9	2.0	1.9	2.0	2.1	2.2	1.9	2.0	2.0	2.0 (.6)
BMI/10	2.1	2.3	2.2	2.3	2.2	2.3	2.4	2.4	2.1	2.3 (.5)
Health	4.1	3.9	4.0	3.7	4.3	4.0	3.7	3.7	4.2	4.0 (.2)
Same-Sex Attraction (%)	2.1	5.8	6.7	2.5	1.3	5.8	2.9	5.8	4.7	4.2
Depression	.5	.5	.5	.5	.4	.6	.5	.7	.5	.6 (.4)
Wave I Degree^b	1.3	.5	3.2	1.7	1.4	3.1	1.7	1.5	1.3	1.9
Wave II Degree^b	1.8	.7	2.8	2.3	1.8	2.7	2.7	1.3	2.3	1.9
School
Type	Private	Private	Public	Public	Private	Public	Public	Public	Private	n/a
Location	Rural	Urban	Rural	Rural	Urban	Rural	Rural	Suburban	Urban	n/a
Region	West	South	Northeast	South	South	Midwest	Midwest	West	Midwest	n/a
Number of Students	48	52	60	121	76	479	70	850	64	1,820
Jaccard Index	.25	.26	.31	.22	.28	.26	.26	.23	.25	.25

Individual-level statistics are means and percentages. The total column includes standard deviations in parentheses.

Degree reflects the average number of outgoing and incoming ties, which by definition are equal.

Friendships

Friendships were measured at each wave by asking respondents to name their five best male and five best female friends.⁵ We required data from respondents and their friends and only include nominations of peers who attended the same school and completed the survey. For each possible dyad, we code the tie as present (1) or absent (0).⁶ Networks are directed, meaning a tie from i to j is measured separately from a tie from j to i. Across the nine schools, there are 988,142 possible directed friendship ties. The average number of friendships (i.e., degree) remains relatively constant across the two waves at 1.9 per student. The Jaccard index indicates that 25 percent of friendships that existed in either wave are present at both waves. None of the schools have a Jaccard index lower than .2, which would raise concerns about too much time between observations (Snijders et al. 2010).

Analytic Strategy

Our primary analysis uses a stochastic actor-based (SAB) model that predicts network change from Wave I to Wave II (Snijders 2001; Snijders et al. 2010). The model conditions on the Wave I network and tests several effects hypothesized to produce the network observed at Wave II. Although we observe the network at discrete time points, the model assumes that ties change on an underlying continuous time-scale that follows a Markov process. Specifically, network change occurs through a series of micro-steps in which actors are given the opportunity to modify their outgoing ties, either by forming new ties or dropping existing ties. Within each micro-step, a randomly selected actor evaluates all possible changes to outgoing ties based on effects included in the model. Changes are made in an effort to maximize the following objective function:

f_{i} (β, X) = \sum_{k} β_{k} s_{k i} (X)

where f_i (β, x) is the value of the function for actor (i) given the current set of parameter estimates (β) and state of the network (x). The k effects, represented as s_ki(x), may be based on network or individual attributes (v). After evaluating possible changes, an actor makes the change with the highest positive evaluation (although with a small amount of randomness introduced to the evaluation). For instance, when the reciprocity parameter is positive, changes that create a reciprocated tie or remove an unreciprocated tie are evaluated higher and are more likely to be made. If no changes have a positive evaluation, the actor maintains the existing set of outgoing ties. A rate effect regulates how often actors are given the opportunity to change their ties. Only one actor can change a tie at any given moment, preventing actors from coordinating relationships between themselves.

Individual and dyadic covariates related to depression at Wave I are the effects of greatest interest to us. We use the depression similarity effect to determine whether egos were more likely to select alters who were similar to themselves (i.e., the Attribute Similarity effect in Table 2). We center similarity scores using mean similarity across all possible dyads in the network $\bar{(s i m})$ . Centering within schools adjusts for different distributions of depression across schools to ensure that similarity is measured relative to the appropriate baseline. Dyads with positive similarity scores are more similar than expected by chance, whereas negative scores indicate less similarity than expected by chance. A positive similarity parameter in the model would be evidence that ties are more likely between adolescents with similar levels of depression. We evaluate avoidance using the alter depression effect (i.e., the Alter Attribute effect in Table 2). A negative parameter would indicate that adolescents with higher levels of depression are less likely to be selected as friends. Withdrawal is represented by the ego depression effect (i.e., the Ego Attribute effect in Table 2). Should the parameter estimate for this effect be negative, then higher levels of depression would lower i’s evaluation of a tie to each alter (j). This would indicate that adolescents with greater depression are less likely to add or retain friends over time.

Table 2.

Formulas for Selection Effects

Effect	Formula
Attribute Similarity^a	$\sum_{j} (x_{i j} s i m_{i j} - \bar{s i m})$
Alter Attribute	$\sum_{j} x_{i j} v_{j}$
Ego Attribute	$\sum_{j} x_{i j} v_{i}$
Attribute Same^b	$\sum_{j} x_{i j} I {v_{i} = v_{j}}$
Outdegree	$\sum_{j} x_{i j}$
Reciprocity	$\sum_{j} x_{i h} x_{i j} x_{j h}$
Transitive Triplets	$\sum_{j} x_{i j} \sqrt{x_{+ j}}$
Popularity^c	$\bar{(s i m})$

Note: All attributes were centered using the school mean prior to calculating the effect.

sim_ij = 1 - |v_i - v_j|/ Δ, with Δ = max_ij |v_i - v_j |.

I {v_i = v_j} equals 1 if v_i = v_j, and 0 otherwise.

X_+j is the number of incoming ties of alter j.

To further prevent spurious results due to consolidation, we include ego, alter, and similarity effects for sex, age, parent’s education, puberty, BMI, health, and same-sex attraction. For race/ethnicity, we use the Attribute Same effect (see Table 2), which indicates whether adolescents selected others with the same race/ethnicity. We code dyads in which adolescents were the same race/ethnicity as 1, and all others as 0. Should addition of any of these effects to the model reduce the magnitude of the depression effects, then selection on these attributes was creating the appearance of selection on depression though consolidation.

All models necessarily include a rate effect to control for opportunities to change ties and an outdegree effect to control for number of ties in the network. An outdegree parameter of 0 corresponds to a probability of .5 for each tie in the network. Because the probability of any single tie is generally quite small, estimated outdegree parameters are typically negative (Snijders et al. 2010). Measures of normative network processes include reciprocity, transitivity, and popularity (see Table 2 for formulas). A positive reciprocity parameter would indicate that adolescents were more likely to select peers who had previously selected them. We model transitive closure using the transitive triplets effect. A positive parameter would indicate adolescents were more likely to select a peer if the tie created a transitive structure (i.e., i → j, j → h, and i → h). The popularity effect reflects how the number of incoming ties (i.e., indegree) affected the future likelihood of being selected by others. The square root transformation reduces the weight given to actors with more incoming ties (Snijders et al. 2010). Thus, when an ego chooses among alters, differences between lower indegree alters matter more than differences between higher indegree alters. A positive popularity parameter would indicate that adolescents were more likely to select peers whom many other students had previously selected.

We conducted analyses using version 3.17w of the Simulation Investigation for Empirical Network Analysis (SIENA) software program (Snijders et al. 2009). We followed procedures outlined by Snijders for model fitting and testing for convergence (Snijders et al. 2010). Because each school represents a distinct network with no possibility for overlap, we estimated separate models by school. We then used a meta-analysis to combine model estimates and test for the significance of effects across schools (Snijders and Baerveldt 2003). Details of the meta-analysis are provided in Part B of the online supplement.

Results

Our analysis proceeds in three stages. First, to gain initial insight into the relation between depression and friendship networks, we examine the bivariate association between depression and network structure. Second, we use a series of SAB models to test the hypothesized homophily mechanisms. Third, given evidence that homophily is due to withdrawal, we decompose the friend selection process. For adolescents at different levels of depression, we calculate rates at which normative network processes were responsible for friendships and the extent to which homophilous friendships occurred outside of normative network processes.

Bivariate Analysis

We first consider the association between depression and adolescents’ number of friends. Results indicate that depression is significantly correlated with the number of incoming and outgoing friendship ties. Students with higher levels of depression were less likely to select others (r = –.06, p = .01) and to be selected by others (r = –.10, p < .001). These results are consistent with the withdrawal and avoidance mechanisms, respectively, but do not provide conclusive evidence for either (which requires the longitudinal analysis).

We next consider whether homophily on depression exists. We calculate odds of an ego selecting an alter with a similar level of depression, relative to odds of selecting an alter with a different level of depression. We define similarity as a depression score within .1 of one’s own score. The odds ratio of 1.22 is positive and significantly greater than one (95% CI: 1.13, 1.31), indicating that depression homophily occurred at a greater than expected rate. Odds of a friendship between two adolescents with similar levels of depression were 22 percent higher than the odds of a friendship between dissimilar adolescents.

To help convey the pattern of homophily, Figure 2 presents odds of friendship presence (versus absence) for each combination of ego and alter depression.⁷ The z-axis reflects odds of friendship; the x- and y-axes reflect depression levels of ego and alter, respectively. The saddle-shaped pattern provides clear evidence of depression homophily. Adolescents with low or high levels of depression were more likely to have homophilous friendships. Only adolescents with scores in the middle, around 1.0, were just as likely to have heterophilous as homophilous relationships.

Figure 2.

Odds of Friendship by Ego and Alter Depression

SAB Models

We now address mechanisms responsible for homophily on depression. We begin by testing whether homophily on depression exists over time. Model 1 assesses the total effect of depression similarity on friendship likelihood; it is not intended to evaluate mechanisms responsible for homophily, so it does not include controls for alternative selection processes. Meta-analysis results in Table 3 reveal a marginally significant tendency for homophilous relationships to form and persist (μ_θ = .52, p = .055). Recall that the SAB model predicts who selects whom as a friend. Adolescents who were more similar in their level of depressive symptoms were thus more likely to select one another as friends over time. Subsequent models determine whether this homophily was due to preference or other hypothesized mechanisms.

Table 3.

Stochastic Actor-Based Model: Unstandardized Friend Selection Coefficients and Standard Errors (Meta-Analysis of Nine Schools, 988,142 Dyads)

	Model 1	Model 2	Model 3	Model 4
Depression
Depression Similarity (Preference)	.52†	.29	.35	.19
	(.27)	(.27)	(.31)	(.29)
Alter Depression (Avoidance)		−.16***	−.18†	−.03
		(.05)	(.11)	(.05)
Ego Depression (Withdrawal)		−.18**	−.20	−.12*
		(.08)	(.13)	(.06)
Network Processes
Outdegree	−2.15***	−2.17***	−3.13***	−3.96***
	(.23)	(.23)	(.31)	(.28)
Reciprocity				2.08***
				(.18)
Transitive Triplets				.53***
				(.07)
Popularity (square-root)				.35***
				(.03)
Individual Attributes
Female Similarity			.37**	.33**
			(.14)	(.13)
Alter Female			−.06	−.12†
			(.09)	(.06)
Ego Female			.18	.15
			(.12)	(.14)
Age Similarity			4.35***	3.23***
			(.56)	(.43)
Alter Age			.00	.06**
			(.02)	(.02)
Ego Age			−.09**	−.06**
			(.03)	(.02)
Same Race			1.50***	1.21***
			(.12)	(.08)
Parent’s Education Similarity			.35	.25
			(.27)	(.29)
Alter Parent’s Education			.04	.03
			(.09)	(.08)
Ego Parent’s Education			.06	.03
			(.09)	(.09)
Puberty Similarity			.19†	.10
			(.10)	(.10)
Alter Puberty			.11**	.05
			(.04)	(.04)
Ego Puberty			.01	−.03
			(.09)	(.12)
BMI/10 Similarity			.46***	.23†
			(.13)	(.14)
Alter BMI/10			−.17***	−.10*
			(.05)	(.04)
Ego BMI/10			−.00	.10
			(.13)	(.14)

Note: Standard errors in parentheses.

†

p < .10; *p < .05; **p < .01; ***p < .001 (two-tailed tests).

In Model 2, we add effects for depression related to avoidance (alter depression) and withdrawal (ego depression). This model tests whether the marginal similarity effect observed in Model 1 is a consequence of depression-related avoidance or withdrawal. Results indicate that depression affected the friend selection behavior of oneself and one’s peers. The negative ego depression parameter suggests that more depressed adolescents selected fewer friends over time. The negative alter depression parameter suggests that adolescents with higher levels of depression were less likely to be selected as a friend by Wave II. Compared with Model 1, the magnitude of the depression similarity parameter drops by half and is no longer marginally significant. Consequently, there is no evidence of a preference for friends with similar levels of depressive symptoms. The tendency toward homophily found in Model 1 is an indirect product of withdrawal or avoidance.

Model 3 adds effects for consolidating factors that could create spurious depression homophily. As expected, most individual attributes play some role in the friend selection process.⁸ For sex, age, race, puberty, and BMI, friendships between similar adolescents were more likely than friendships between dissimilar adolescents. We observe a significant negative effect for ego age, suggesting that older adolescents were less likely to add or maintain friends. This is consistent with extant research documenting declines in the size of friendship networks during high school (Cairns, Xie, and Leung 1998). We also find a significant negative effect of alter BMI, meaning that as BMI increased, adolescents were less likely to be selected as a friend. This effect is also indicative of an avoidance process and is consistent with prior research (Crosnoe et al. 2008). The positive effect of alter puberty suggests that more developed students were more likely to be selected as friends.

Of particular interest was how effects of depression changed when adding possible consolidating factors. Magnitudes of ego and alter depression effects remain substantively unchanged, although their standard errors increase. These increases are driven by greater between-school variances, not larger standard errors for the school-level estimates (which would occur with consolidation). Thus, we conclude that consolidation is not responsible for observed effects of depression on network structure.

Model 4 adds effects for normative network processes in the form of reciprocity, transitivity, and popularity. This model tests whether depression-related withdrawal and avoidance operated net of the magnifying effects of these network processes. In addition, one or more of these processes must be present for the withdrawal mechanism to produce homophily. Results indicate that all three structural effects are significant and in the expected direction. Over time, students were more likely to select alters who had nominated them (reciprocity), friends of their current friends (transitive triplets), and peers whom many other students had previously selected (popularity).

Turning to effects related to depression, we see that they each drop in magnitude. The ego depression parameter remains negative and significant, which is evidence that depression led to withdrawal from social contact. By contrast, the alter depression parameter decreases and is nonsignificant, revealing that the avoidance of depressed peers seen in prior models is due to network processes introduced in Model 4. That is, depressed adolescents were less likely to be selected because they were less accessible through network processes, not because of their depression. For example, depressed adolescents who withdrew had fewer opportunities to be selected through reciprocity. The result is fewer incoming nominations, a pattern that could be amplified through the effect of popularity. In combination, these results suggest that withdrawal was ultimately responsible for the observed effect of homophily on depression.

To assess sensitivity of our results to specification of models and effects, we estimated a number of alternative models.⁹ First, the measure of depression similarity described earlier assumes that the effect of similarity is equal regardless of an individual’s level of depression. This may fail to capture homophily that is stronger for individuals at one or both ends of the depression scale. We thus tested an interaction between ego depression and depression similarity to determine whether similarity preferences are linearly related to one’s depression. We also specified depression similarity as the product of ego’s and alter’s centered depression scores, which gives greater weight to extreme scores (either high or low) and would capture a curvilinear effect of depression similarity. Second, although our interest is in friend selection related to depression, estimates of such effects could be biased if depression changes over time due to socialization or other forces. Consequently, we estimated models that account for changes in depression from Wave I to Wave II and include effects for socialization on depression. Finally, the right-skewed depression measure may contain outliers that could potentially influence results. As an alternative specification, we estimated models with a four-category measure of depression (0 to .4, .41 to .8, .81 to 1.2, and 1.2+), which minimizes the impact of outliers on results. None of these alternative specifications produce substantively different results for the friend selection process.

Depression and Network Processes

The pattern of results so far is consistent with the withdrawal mechanism, which proposes that depressed individuals occupy more marginalized network positions and are hence less reliant on normative network processes for friendship. In turn, homophily emerges because depressed individuals are more likely to encounter one another outside of normative network processes. However, results do not indicate how operation of normative network processes differed by depression level, nor that depressed adolescents were more likely to have homophilous ties outside of normative network processes. To support our inferences, we decompose the friend selection process for adolescents at different levels of depression. For each adolescent, we count the number of incoming and outgoing ties at Wave II that could be attributed to reciprocity, transitivity, or popularity.¹⁰ We also count how many ties cannot be attributed to any of the three processes. We divide these sums by the total number of ties at Wave II to obtain the proportion of ties attributable to each network process (or to none of them). For this analysis, we separate adolescents into four levels of depression and use a permutation test to assess differences in network processes across levels.¹¹ For each measure in Table 4, we regress the observed proportion on depression level. We compare the observed regression coefficient to a distribution of 1,000 regression coefficients obtained after randomly permuting depression scores and recalculating proportions. P-values indicate the proportion of permutations in which the measure and depression have the expected association with a magnitude as strong as that observed. We expect depression to be negatively associated with normative network processes and positively associated with their absence.

Table 4.

Proportion of Ties Created or Maintained Over Time by Network Process and Depression Level

	Depression Level
	0 to .4	.41 to .8	.81 to 1.2	1.2+	b ^a	p
Number of Students	598	734	337	151
Mean Number of Ties	4.45	3.67	3.36	2.59
Network Processes
Reciprocity	.37	.35	.33	.37	−.002	.658
Transitivity	.20	.17	.18	.15	−.014	.094
Popularity	.41	.44	.42	.38	−.012	.070
Unknown	.24	.25	.27	.31	.023	.042
Homophilous Ties Unattributable to Network Processes	.23	.26	.32	.36	.045	.045

Note: Columns for network processes do not sum to one because multiple processes may contribute to the presence of a tie.

We obtained beta coefficients (b) by regressing the proportion of ties attributable to each process on depression level (coded 1 to 4).

To begin, our results reveal that depression is negatively associated with number of friends over time. The most depressed adolescents had two fewer ties over time than did the least depressed adolescents (approximately one less incoming tie and one less outgoing tie). This complements the earlier finding that depression led to network marginalization. Turning to network processes, we observe that depressed adolescents were generally less likely to have friendships through normative network processes. Regression coefficients for each process are negative, indicating that adolescents with higher levels of depression were less likely to have ties through the process. Transitivity is behind 20 percent of the ties for adolescents in the lowest depression category, compared with only 15 percent of friendships for adolescents in the most depressed category. This pattern repeats for popularity, where the least depressed adolescents had 41 percent of their ties through popularity, compared with 38 percent for adolescents in the highest depression category. Reciprocity displays a decreasing trend over the three lower levels of depression but then increases for the most depressed adolescents.

Notably, the proportion of friendships that cannot be attributed to any of the normative network processes differs by depression level. As the Unknown row in Table 4 indicates, at higher levels of depression, normative network processes were significantly less likely to contribute to friendships. For the least depressed adolescents, 24 percent of their ties were outside of normative network processes, compared with 31 percent of ties for the most depressed adolescents. These results offer evidence that depressed adolescents’ greater marginalization left them with proportionally more friendships through other means.

Finally, we calculate the proportion of homophilous friendships unattributable to normative network processes. We define homophily as a friendship with someone in the same or an adjacent depression category. As Table 4 shows, the likelihood of homophilous friendships outside of normative network processes significantly increased with depression level. For the least depressed adolescents, 23 percent of homophilous friendships were outside of normative network processes. This percent increases across levels of depression, reaching 36 percent for the most depressed adolescents. This pattern exists because less depressed students were more embedded in the friendship network, which promoted contact with less depressed peers. By contrast, due to their marginalization, more depressed adolescents had friendships outside of normative network processes, where they encountered more depressed peers. In combination with results for the SAB models, these findings offer support for the withdrawal mechanism.

Discussion

Our goal in this study was to better understand the association between health behavior and social networks. We focused specifically on how depression among adolescents affects the friend selection process, resulting in depression homophily. We articulated three theoretical mechanisms through which individual action could lead to depression homophily. The simplest mechanism is a preference for homophily, in which all individuals (even those experiencing depression) prefer friends with similar levels of depressive symptoms. The second mechanism posits that the stigma and aversive behaviors associated with depression cause others to avoid depressed individuals (even others who are themselves depressed). To overcome the friendship deficit this creates, depressed individuals become more amenable to friendships with depressed peers. According to the third mechanism, depressed individuals withdraw from friendships, leaving them in more marginalized network positions where they have less access to others through common network processes. Consequently, depressed individuals develop friendships with other marginalized individuals, including others who are depressed.

Our analysis provides support for depression homophily through withdrawal. We found no evidence of preference for homophily or that depression encouraged avoidance by one’s peers. Instead, the process by which depression affected friend selection was more complex. Depressed adolescents withdrew from friendships over time, leaving them with fewer friends and in network positions that were less conducive to friendship. In turn, others were less likely to select depressed adolescents as friends because they occupied marginalized network positions, not because of their depression. Homophily on depression emerged indirectly as network processes helped support friendships among non-depressed adolescents, leaving depressed adolescents to find one another through other means.

The avoidance and withdrawal mechanisms that we proposed differ from prior theories for homophily that stress (1) structural forces exogenous to the network or (2) a preference for homophily (McPherson et al. 2001). By contrast, avoidance and withdrawal explain homophily through endogenous friend selection processes. Friendship outcomes depend not only on one’s own behavior but on selection behavior of other individuals in the network. Consequently, when a condition like depression affects something as basic as friendship quantity, the effect can reverberate through the network and culminate in homophily (even when no one prefers homophily). In the present study, withdrawal was responsible for homophily; however, other scholars suggest that avoidance can have similar consequences (Hektner, August, and Realmuto 2000; Hogue and Steinberg 1995), and this is supported by experimental evidence (Schaefer 2009). We thus conclude that homophily in excess of what can be accounted for by structural factors does not necessarily reflect a preference for homophily. Homophily can also be endogenously induced through processes of avoidance and withdrawal.¹²

These proposed mechanisms represent middle-range theory (Merton 1968) that can be extended to dimensions beyond depression, although the actual behaviors and perceptions serving as the basis for selection behavior will likely differ from those that accompany depression. An important next step is to clarify conditions under which these homophily selection mechanisms operate (Hedström and Swedberg 1998). Avoidance and withdrawal mechanisms require that actors at opposite ends of the continuum either behave differently or are perceived differently by others. Additional dimensions where this occurs include aggression, obesity, and physical disability, which are characterized by stigma or aversive interpersonal behavior (Crosnoe et al. 2008; Lucas and Phelan 2010; Sijtsema et al. 2010).

One must be careful to distinguish preference-based from endogenously induced homophily. These processes can produce the same cross-sectional association between homophily and friendship presence. Only by observing the pattern of friend selection over time is it possible to determine mechanisms responsible for homophily. Our research thus emphasizes the importance of using appropriate methods when investigating mechanisms behind patterns of social behavior such as friend selection. The ability to properly address questions such as ours is increasing as more longitudinal network data become available and dynamic network methods, such as the SAB model, are more widely adopted (Steglich, Snijders, and Pearson 2010).

The avoidance and withdrawal mechanisms have implications for outcomes beyond homophily. For instance, when aggressive children are excluded by peers and left to form homophilous friendships, their aggression may be reinforced (Sijtsema et al. 2010). Moreover, withdrawal and avoidance have independent negative consequences for individual well-being. Withdrawal can lead to difficulties with peers and school and engender mental and physical health problems (Baumeister and Leary 1995; Rubin, Coplan, and Bowker 2009). Subsequent social isolation stemming from these processes can limit access to social capital (Lin 2001) and increase mortality risk (House et al. 1988). The methods we used can inform investigations of influence, withdrawal, and avoidance processes, as well as help explain the emergence of homophily.

The conventional view of the link between social networks and health is unidirectional, with networks as exogenous (e.g., Umberson et al. 2010). Our results are consistent with prior research documenting a negative association between adolescent depression and social integration (Falci and McNeely 2009; Ueno 2005). However, the effect we observe is in the opposite direction: depression led to lower integration through withdrawal. Thus, the current findings serve as a challenge to current thinking about how health may affect friendship network structures. Our research not only demonstrates that individual characteristics affect network structure, but that a characteristic often considered to be a product of social network functioning also has a role in structuring the network. By treating network structure as the outcome, we were able to uncover processes through which patterns of relationships develop and individuals find themselves in their particular network positions. In conjunction with research on health (Haas et al. 2010) and personality (Kalish and Robins 2006), this expands our knowledge of how micro-level processes contribute to macro-level social structures.

We proposed multiple processes related to depression that could affect friend selection behavior, including reduced energy, fear of stigma, and disinterest in normative activities. Delineating which mechanisms are in operation is vital to appropriately target interventions to counter negative consequences of depression for social integration. For instance, psychotropic drugs are increasingly used to counteract depression in adolescents (Thomas et al. 2006). For such treatment strategies to positively affect social integration, they must alter the relevant causal pathway. In addition to the risks for suicide and suicidal ideation they carry (Vitiello and Swedo 2004), medication runs the risk of negative effects through alternative causal pathways: medication may carry its own stigma (Pescosolido et al. 2007), which can exacerbate avoidance and withdrawal, or the flat affect that psychotropic medications often produce can foster avoidance by peers. Moreover, medication alone does not compensate for sociobehavioral deficits that accompany depression. Even if antidepressants function as intended, they may leave adolescents with a desire for greater social integration but lacking the social skills to achieve such changes. Psychotherapeutic interventions (e.g., behavioral activation) that focus on reducing social withdrawal and inactivity by reorganizing adolescents’ social lives to increase levels of positive reinforcement may be more effective (Jacobson, Martell, and Dimidjian 2001).

While our research provides new insight into how depression affects network structure, there are several remaining questions. To begin, we were unable to distinguish between depression as a short-lived state versus a stable trait (Cole et al. 2006). Duration of depression likely has implications for others’ willingness to maintain relationships with depressed friends. On a related point, most depression in our sample was likely at the more prevalent subclinical level. The processes we investigated, such as avoidance, may be stronger among adolescents with clinical depression. For instance, peers who are clinically depressed may impose greater costs on friendship or carry greater stigma, which could lead others to directly avoid them (see Lucas and Phelan 2010). From a sampling perspective, we were only able to examine friendships among adolescents in the same school. Given that depressed students, especially girls, are more likely to nominate friends outside of school (Ueno 2005), it is important to investigate who these friends are and how their friendships differ from those within schools (van Zalk et al. 2010). It would also be useful to investigate the degree to which ties to non-school friends are a consequence of within-school dynamics and the experience of marginalization that we found. Finally, our sample included only nine high school networks. Relative to many network studies, nine is a lot. Still, testing these processes on a larger and more representative set of networks would provide greater precision.

This research investigated several processes through which social networks come into being. We identified effects of one dimension of friend selection—depression—yet much work remains in partialing effects of other salient dimensions of friendship, normative network processes, and the feedback of friendships on behavior. In moving forward, it is imperative that theory keep pace with methodological advances that allow for investigation of complex processes such as these.

Footnotes

Acknowledgements

We express our appreciation to Tom Snijders, jimi adams, Steven Haas, Sarah Hayford, Matthew Pittinsky, and Carey Cooper for their helpful comments on an earlier draft. This research uses data from Add Health, a project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P01-HD31921), with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to obtain the Add Health data files is available on the Add Health website ().

This research was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R21-HD060927).

Notes

References

Avenevoli

Shelli

Knight

Erin

Kessler

Ronald C.

Merikangas

Kathleen R.

2008. Epidemiology of Depression in Children and Adolescents. New York: Guilford Press.

Baker

Mary

Milich

Richard

Manolis

Mary B.

1996. “Peer Interactions of Dysphoric Adolescents.” Journal of Abnormal Child Psychology 24:241–55.

Baumeister

Roy F.

Leary

Mark R.

1995. “The Need to Belong: Desire for Interpersonal Attachments as a Fundamental Human Motivation.” Psychological Bulletin 117:497–529.

Baumeister

Roy F.

Tice

Dianne M.

1990. “Anxiety and Social Exclusion.” Journal of Social and Clinical Psychology 9:165–95.

Berkman

Lisa F.

Glass

Thomas

. 2000. “Social Integration, Social Networks, Social Support and Health.” Pp. 137–73 in Social Epidemiology, edited by Berkman

L. F.

Kawachi

New York: Oxford University Press.

Berkman

Lisa F.

Glass

Thomas

Brissette

Ian

Seeman

Teresa E.

2000. “From Social Integration to Health: Durkheim in the New Millennium.” Social Science & Medicine 51:843–57.

Blau

Peter M.

1977. Inequality and Heterogeneity. New York: Free Press.

Borelli

Jessica L.

Prinstein

Mitchell J.

2006. “Reciprocal, Longitudinal Associations among Adolescents’ Negative Feedback-Seeking, Depressive Symptoms, and Peer Relations.” Journal of Abnormal Child Psychology 34:159–69.

Brendgen

Mara

Vitaro

Frank

Turgeon

Lyse

Poulin

François

. 2002. “Assessing Aggressive and Depressed Children’s Social Relations with Classmates and Friends: A Matter of Perspective.” Journal of Abnormal Child Psychology 30:609–624.

10.

Byrne

Donn

. 1971. The Attraction Paradigm. New York: Academic Press.

11.

Cairns

Robert

Xie

Hongling

Leung

Man-Chi

. 1998. “The Popularity of Friendship and the Neglect of Social Networks: Toward a New Balance.” New Directions for Child and Adolescent Development 109:25–54.

12.

Cantwell

Dennis P.

Baker

Lorian

. 1991. “Manifestations of Depressive Affect in Adolescence.” Journal of Youth and Adolescence 20:121–33.

13.

Chan

Alessandra

Poulin

François

. 2009. “Monthly Instability in Early Adolescent Friendship Networks and Depressive Symptoms.” Social Development 18:1–23.

14.

Cole

David A.

Nolen-Hoeksema

Susan

Girgus

Joan

Paul

Gilda

. 2006. “Stress Exposure and Stress Generation in Child and Adolescent Depression: A Latent Trait-State-Error Approach to Longitudinal Analyses.” Journal of Abnormal Psychology 115:40–51.

15.

Compas

Bruce E.

Sydney

Grant

Kathryn E.

1993. “Taxonomy, Assessment, and Diagnosis of Depression during Adolescence.” Psychological Bulletin 114:323–44.

16.

Cornwell

Benjamin

. 2009. “Good Health and the Bridging of Structural Holes.” Social Networks 31:92–103.

17.

Coyne

James C.

1976a. “Depression and the Response of Others.” Journal of Abnormal Psychology 85:186–93.

18.

Coyne

James C.

1976b. “Toward an Interactional Description of Depression.” Psychiatry 39:28–40.

19.

Crosnoe

Robert

Frank

Kenneth

Mueller

Anna S.

2008. “Gender, Body Size and Social Relations in American High Schools.” Social Forces 86:1189–1216.

20.

Daley

Shanon E.

Hammen

Constance

. 2002. “Depressive Symptoms and Close Relationships during the Transition to Adulthood: Perspectives from Dysphoric Women, Their Best Friends, and Their Romantic Partners.” Journal of Consulting and Clinical Psychology 70:129–41.

21.

Eisenberger

Naomi I.

Lieberman

Matthew D.

Williams

Kipling D.

2003. “Does Rejection Hurt? An fMRI Study of Social Exclusion.” Science 302:290–92.

22.

Falci

Christina

McNeely

Clea

. 2009. “Too Many Friends: Social Integration, Network Cohesion and Adolescent Depressive Symptoms.” Social Forces 87:2031–62.

23.

Feld

Scott L.

1981. “The Focused Organization of Social Ties.” American Journal of Sociology 86:1015–35.

24.

Greenberg

Martin S.

Shapiro

Solomon P.

1971. “Indebtedness: An Adverse Aspect of Asking For and Receiving Help.” Sociometry 34:290–301.

25.

Haas

Steven A.

Schaefer

David R.

Kornienko

Olga

. 2010. “Health and the Structure of Adolescent Social Networks.” Journal of Health and Social Behavior 51:424–39.

26.

Hedström

Peter

Bearman

Peter

. 2009. “What is Analytical Sociology All About? An Introductory Essay.” Pp. 3–24 in The Oxford Handbook of Analytical Sociology, edited by Hedström

Bearman

Oxford: Oxford University Press.

27.

Hedström

Peter

Swedberg

Richard

. 1998. Social Mechanisms: An Analytical Approach to Social Theory. Cambridge: Cambridge University Press.

28.

Hektner

Joel M.

August

Gerald J.

Realmuto

George M.

2000. “Patterns and Temporal Changes in Peer Affiliation among Aggressive and Nonaggressive Children Participating in a Summer School Program.” Journal of Clinical Child & Adolescent Psychology 29:603–614.

29.

Herman-Stahl

Mindy

Petersen

Anne C.

1996. “The Protective Role of Coping and Social Resources for Depressive Symptoms among Young Adolescents.” Journal of Youth and Adolescence 25:733–53.

30.

Hogue

Aaron

Steinberg

Laurence

. 1995. “Homophily of Internalized Distress in Adolescent Peer Groups.” Developmental Psychology 31:897–906.

31.

House

James S.

1981. Work Stress and Social Support. Reading, MA: Addison-Wesley.

32.

House

James S.

Umberson

Debra

Landis

Karl R.

1988. “Structures and Processes of Social Support.” Annual Review of Sociology 14:293–318.

33.

Jacobson

Neil S.

Martell

Christopher R.

Dimidjian

Sona

. 2001. “Behavioral Activation Treatment for Depression: Returning to Contextual Roots.” Clinical Psychology: Science and Practice 8:255–70.

34.

Jose

Paul E.

Brown

Isobel

. 2008. “When Does the Gender Difference in Rumination Begin? Gender and Age Differences in the Use of Rumination by Adolescents.” Journal of Youth and Adolescence 37:180–92.

35.

Kalish

Yuval

Robins

Garry

. 2006. “Psychological Predispositions and Network Structure: The Relationship between Individual Predispositions, Structural Holes and Network Closure.” Social Networks 28:56–84.

36.

Kandel

Denise B.

1978. “Homophily, Selection, and Socialization in Adolescent Friendships.” American Journal of Sociology 84:427–36.

37.

Kulik

James A.

Mahler

Heike I. M.

2000. “Social Comparison, Affiliation, and Emotional Contagion under Threat.” Pp. 295–319 in Handbook of Social Comparison, edited by Suls

Wheeler

New York: Plenum.

38.

Lewinsohn

Peter M.

Hops

Hyman

Roberts

Robert E.

Seeley

John R.

Andrews

Judy A.

1993. “Adolescent Psychopathology: I. Prevalence and Incidence of Depression and Other DSM-III-R Disorders in High School Students.” Journal of Abnormal Psychology 102:133–44.

39.

Lin

Nan

. 2001. Social Capital: A Theory of Social Structure in Action. New York: Cambridge University Press.

40.

Lin

Nan

Dean

Alfred

Ensel

Walter M.

1986. Social Support, Life Events, and Depression. New York: Academic Press.

41.

Link

Bruce G.

Cullen

Francis T.

Struening

Elmer

Shrout

Patrick E.

Dohrenwend

Bruce P.

1989. “A Modified Labeling Theory Approach to Mental Disorders: An Empirical Assessment.” American Sociological Review 54:400–423.

42.

Link

Bruce G.

Phelan

Jo C.

2001. “Conceptualizing Stigma.” Annual Review of Sociology 27:363–85.

43.

Lucas

Jeffrey W.

Phelan

Jo C.

2010. “Stigma and Status.” Presented at the 2010 Group Processes Conference, Atlanta, GA.

44.

Maner

Jon K.

DeWall

C. Nathan

Baumeister

Roy F.

Schaller

Mark

. 2007. “Does Social Exclusion Motivate Interpersonal Reconnection? Resolving the ‘Porcupine Problem.’” Journal of Personality and Social Psychology 92:42–55.

45.

Martin

Jack K.

Pescosolido

Bernice A.

Olafsdottir

Sigrun

McLeod

Jane D.

2007. “The Construction of Fear: Americans’ Preferences for Social Distance from Children and Adolescents with Mental Health Problems.” Journal of Health and Social Behavior 48:50–67.

46.

McPherson

J. Miller

Smith-Lovin

Lynn

. 1987. “Homophily in Voluntary Organizations: Status Distance and the Composition of Face-to-Face Groups.” American Sociological Review 52:370–79.

47.

McPherson

Miller

Smith-Lovin

Lynn

Cook

James M.

2001. “Birds of a Feather: Homophily in Social Networks.” Annual Review of Sociology 27:415–44.

48.

Merton

Robert K.

1968. Social Theory and Social Structure. New York: Free Press.

49.

Moody

James

. 2001. “Race, School Integration, and Friendship Segregation in America.” American Journal of Sociology 107:679–716.

50.

Morris

Martina

. 2004. Network Epidemiology: A Handbook of Survey Design and Data Collection. London: Oxford University Press.

51.

Mouw

Ted

Entwisle

Barbara

. 2006. “Residential Segregation and Interracial Friendship in Schools.” American Journal of Sociology 112:394–441.

52.

Pennebaker

James W.

Dyer

Mary Anne

Caulkins

R. Scott

Litomtz

Debra L.

Ackreman

Phillip L.

Anderson

Douglas B.

McGraw

Kevin M.

1979. “Don’t the Girls Get Prettier at Closing Time: A Country and Western Application to Psychology.” Personality and Social Psychology Bulletin 5:122–25.

53.

Pescosolido

Bernice A.

Perry

Brea L.

Martin

Jack K.

McLeod

Jane D.

Jensen

Peter S.

2007. “Stigmatizing Attitudes and Beliefs about Treatment and Psychiatric Medications for Children with Mental Illness.” Psychiatric Services 58:613–18.

54.

Prinstein

Mitchell J.

2007. “Moderators of Peer Contagion: A Longitudinal Examination of Depression Socialization between Adolescents and Their Best Friends.” Journal of Clinical Child and Adolescent Psychology 36:159–70.

55.

Prisbell

Marshall

Andersen

Janis F.

1980. “The Importance of Perceived Homophily, Level of Uncertainty, Feeling Good, Safety, and Self-Disclosure in Interpersonal Relationships.” Communication Quarterly 28:22–33.

56.

Radloff

Lenore S.

1977. “The CES-D Scale: A Self-Report Depression Scale for Research in the General Population.” Applied Psychological Measurement 1:385–401.

57.

Raub

Werner

Buskens

Vincent

van Assen

Marcel A. L. M.

2011. “Micro-Macro Links and Microfoundations in Sociology.” Journal of Mathematical Sociology 35:1–25.

58.

Rivera

Mark T.

Soderstrom

Sara B.

Uzzi

Brian

. 2010. “Dynamics of Dyads in Social Networks: Assortative, Relational, and Proximity Mechanisms.” Annual Review of Sociology 36:91–115.

59.

Rockhill

Carol M.

Fan

Ming-Yu

Katon

Wayne J.

McCauley

Elizabeth

Crick

Nikki R.

Pleck

Joseph H.

2007. “Friendship Interactions in Children with and without Depressive Symptoms: Observation of Emotion during Game-Playing Interactions and Post-game Evaluations.” Journal of Abnormal Child Psychology 35:429–41.

60.

Rogers

Everett M.

Bhowmik

Phillip K.

1970. “Homophily-Heterophily: Relational Concepts for Communication Research.” Public Opinion Quarterly 34:523–38.

61.

Rook

Karen S.

Pietromonaco

Paula R.

Lewis

Megan A.

1994. “When Are Dysphoric Individuals Distressing to Others and Vice Versa? Effects of Friendship, Similarity, and Interaction Task.” Journal of Personality and Social Psychology 67:548-59.

62.

Rosenblatt

Abram

Greenberg

Jeff

. 1991. “Examining the World of the Depressed: Do Depressed People Prefer Others Who Are Depressed?” Journal of Personality and Social Psychology 60:620–29.

63.

Rubin

Kenneth H.

Coplan

Robert J.

Bowker

Julie C.

2009. “Social Withdrawal in Childhood.” Annual Review of Psychology 60:141–71.

64.

Rudolph

Karen D.

Flynn

Megan

Abaied

Jamie

. 2008. “A Developmental Perspective on Interpersonal Theories of Youth Depression.” Pp 79–103 in Handbook of Depression in Children and Adolescents edited by Abela

J. R. Z.

Hankin

B. L.

New York: Guilford Press.

65.

Rudolph

Karen D.

Hammen

Constance

Burge

Dorli

. 1994. “Interpersonal Functioning and Depressive Symptoms in Childhood: Addressing the Issues of Specificity and Comorbidity.” Journal of Abnormal Child Psychology 22:355–71.

66.

Rudolph

Karen D.

Hammen

Constance

Burge

Dorli

. 1997. “A Cognitive-Interpersonal Approach to Depressive Symptoms in Preadolescent Children.” Journal of Abnormal Child Psychology 24:33–54.

67.

Schachter

Stanley

. 1959. The Psychology of Affiliation: Experimental Studies of the Sources of Gregariousness. Stanford, CA: Stanford University Press.

68.

Schaefer

David R.

2009. “Towards a Theory of Network Formation in Groups.” Presented at the Annual Meeting of the American Sociological Association, San Francisco, CA.

69.

Segrin

Chris

. 2004. “Concordance on Negative Emotion in Close Relationships: Emotional Contagion or Assortative Mating?” Journal of Social and Clinical Psychology 23:815–35.

70.

Segrin

Chris

Flora

Jeanne

. 2000. “Poor Social Skills Are a Vulnerability Factor in the Development of Psychosocial Problems.” Human Communication Research 26:489–514.

71.

Sijtsema

Jelle J.

Ojanen

Tiina

Veenstra

René

Lindenberg

Siegwart

Hawley

Patricia H.

Little

Todd D.

2010. “Forms and Functions of Aggression in Adolescent Friendship Selection and Influence: A Longitudinal Social Network Analysis.” Social Development 19:515–34.

72.

Snijders

Tom A. B.

2001. “The Statistical Evaluation of Social Network Dynamics.” Pp. 361–95 in Sociological Methodology, edited by Sobel

M. E.

Becker

M. P.

Boston: Basil Blackwell.

73.

Snijders

Tom A. B.

Baerveldt

Chris

. 2003. “A Multilevel Network Study of the Effects of Delinquent Behavior on Friendship Evolution.” Journal of Mathematical Sociology 27:123–51.

74.

Snijders

Tom A. B.

Steglich

Christian E. G.

Schweinberger

Michael

Huisman

Mark

. 2009. Manual for SIENA version 3.3. Groningen, University of Groningen, ICS. Oxford: University of Oxford, Department of Statistics.

75.

Snijders

Tom A. B.

van de Bunt

Gerhard G.

Steglich

Christian E. G.

2010. “Introduction to Stochastic Actor-Based Models for Network Dynamics.” Social Networks 32:44–60.

76.

Steglich

Christian E. G.

Snijders

Tom A. B.

Pearson

Michael

. 2010. “Dynamic Networks and Behavior: Separating Selection from Influence.” Sociological Methodology 40:329–93.

77.

Thomas

Cindy P.

Conrad

Peter

Casler

Rosemary

Goodman

Elizabeth

. 2006. “Trends in the Use of Psychotropic Medications among Adolescents, 1994 to 2001.” Psychiatric Services 57:63–69.

78.

Ueno

Koji

. 2005. “The Effects of Friendship Networks on Adolescent Depressive Symptoms.” Social Science Research 34:484–510.

79.

Umberson

Debra

Crosnoe

Robert

Reczek

Corinne

. 2010. “Social Relationships and Health Behavior across the Life Course.” Annual Review of Sociology 36:139–57.

80.

van Straaten

Ischa Rutger

Engels

C.M.E.

Finenauer

Catrin

Holland

Rob W.

2009. “Meeting Your Match: How Attractiveness Similarity Affects Approach Behavior in Mixed-Sex Dyads.” Personality and Social Psychology Bulletin 35:685–97.

81.

van Zalk

Maarten Herman Walker

Kerr

Margaret

Branje

Susan J. T.

Stattin

Håkan

Meeus

Wim H. J.

2010. “It Takes Three: Selection, Influence, and De-selection Processes of Depression in Adolescent Friendship Networks.” Developmental Psychology 46:927–38.

82.

Vitiello

Benedetto

Swedo

Susan

. 2004. “Antidepressant Medications in Children.” New England Journal of Medicine 350:1489–91.

83.

Wellman

Barry

Wortley

Scot

. 1990. “Different Strokes from Different Folks: Community Ties and Social Support.” American Journal of Sociology 96:558–88.

84.

Wimmer

Andreas

Lewis

Kevin

. 2010. “Beyond and Below Racial Homophily: ERG Models of a Friendship Network Documented on Facebook.” American Journal of Sociology 116:583–642.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.02 MB

0.14 MB