Discrepancies About Adolescent Relationships as a Function of Informant Attachment and Depressive Symptoms

Abstract

Developmental scientists studying interpersonal relationships often find that informants disagree in their reports, and two theoretical perspectives suggest that these discrepancies may be predictable. In two studies of adolescents’ interpersonal relationships, we examined two factors that may predict the absolute magnitude and the direction of the discrepancies in reports about several types of relationships. Specifically, we examined informants’ depressive symptoms and attachment as predictors of absolute and directional discrepancies in reports about (a) adolescents’ relationships with peers, (b) parent-adolescent relationships, and (c) adolescents’ friendships. Findings revealed that informant depressive symptoms sometimes were associated with discrepancies. In contrast, informant attachment more consistently accounted for absolute and directional discrepancies.

Adolescence is a time of increased attention to social relationships, including relationships with parents, peers, and friends (see Smetana, Campione-Barr, & Metzger, 2006, for a review). These relationships are thought to be interconnected in meaningful ways, and the quality of these relationships has important implications for adolescents’ functioning (Grotevant, 1998). Researchers have given special attention to adolescents’ relationships, which is perhaps not surprising given that interpersonal relationships have far-reaching associations to such topics as immune functioning (Kiecolt-Glaser et al., 2005), workplace productivity (Rayner & Hoel, 1997), and neural threat responses (Coan, Schaefer, & Davidson, 2006). Indeed, findings from several longitudinal studies indicate that a lack of high-quality social relationships ranks among the top risk factors for heart attacks and suicides (e.g., House, Landis, & Umberson, 1988). Thus, studying the correlates of individual differences in adolescent relationships is the focus of considerable research.

Researchers who study relationships are inevitably faced with the issue of how to assess the quality of a relationship and must make crucial decisions. With the hope that obtaining multiple assessments will best capture the underlying relationship, researchers often incorporate multiple informants into their research designs. These multiple assessments, however, often produce discrepant reports, and researchers are then faced with the decision of how to handle informants’ divergent views (e.g., Holmbeck, Li, Schurman, Friedman, & Coakley, 2002). As such, it is important for psychologists to attempt to gain insight into why these discrepant perceptions about relationships exist. Moreover, in light of evidence that the presence of discrepancies is a risk factor for problematic development (e.g., Lippold, Greenberg, & Feinberg, 2011; Reynolds, MacPherson, Matusiewicz, Schreiber, & Lejuez, 2011), it is important to understand what contributes to discrepant reports.

Previous examinations of informant discrepancies have occurred largely in the context of discrepant reporting about individuals’ psychological functioning. Evidence has emerged in support of the depression distortion hypothesis (Richters, 1992), which predicts that symptomatic informants will report with a negative bias (see De Los Reyes & Kazdin, 2005). Informant (and in most cases, parent) depressive symptoms have been the focus of much of this research. Although a number of studies have identified links between informant depressive symptoms and discrepancies in reports of children’s psychological functioning (e.g., Treutler & Epkins, 2003; Youngstrom, Loeber, & Stouthamer-Loeber, 2000), other studies suggest that parents’ depressive symptoms might not always result in greater discrepancies in reports (e.g., Conrad & Hammen, 1989; Weissman et al., 1987). This research has provided some explanation for the presence of informant discrepancies, but it is likely that additional factors contribute to discrepant reports, particularly when the focus of the reports is something other than individuals’ psychological functioning.

Investigations of informant discrepancies have examined both absolute and directional discrepancies in reports, two scores that provide unique insight into the pattern of differences in informants’ reports. Absolute discrepancy scores indicate the magnitude of distance between two reports, in contrast to directional discrepancy scores, which indicate whether one reporter consistently over- or under-reports compared to another reporter. Informant discrepancies can exist with or without one reporter’s systematic over- or under-reporting relative to another reporter. Discrepancies that are predictable in absolute magnitude but not direction may signal inconsistent behaviors across contexts or a lack of communication between the two reporters. For example, Ehrlich, Cassidy, and Dykas (2011) found predictable absolute but not directional discrepancies in mother-adolescent reports about adolescents’ depressive symptoms. In contrast, discrepancies that are predictable in direction but not absolute magnitude suggests that one informant has a bias to report higher or lower scores than the other reporter (e.g., Berger, Jodl, Allen, McElhaney, & Kuperminc, 2005). Thus, absolute and directional discrepancies can provide meaningful information about the nature of how informants disagree in their reports. Moreover, in clinical practice, the type of discrepancy between informants’ reports could facilitate decisions about appropriate interventions, depending on whether discrepant reports signaled a lack of communication versus a directional bias in reporting.

When informants are reporting about relationships, one particularly useful framework for considering explanations of informant discrepancies is Bowlby’s (1969/1982) attachment theory, with evidence from two studies demonstrating attachment (assessed with the Adult Attachment Interview [AAI]; George, Kaplan, & Main, 1984, 1985, 1996) to be a predictor of informant discrepancies (Berger et al., 2005; Ehrlich et al., 2011). Attachment may contribute to informant discrepancies in several ways. First, attachment insecurity is associated with a lack of open communication about feelings (Bretherton, 1987); such restriction could increase the likelihood that the reports that insecure individuals provide will be discrepant in absolute magnitude from those of another reporter (e.g., dyadic reports about a relationship that differ dramatically, with the two relationship partners perceiving the relationship in very different ways). Second, attachment insecurity is associated with biases in cognitive representations of the self and others (Bowlby, 1973). For example, insecure individuals’ negative cognitions about the self in social contexts (Mikulincer & Shaver, 2007) may lead them to systematically view themselves within social relationships more negatively than others view them. Yet some insecure individuals (i.e., insecure-avoidant individuals) show evidence of idealization of relationship partners (Cassidy & Kobak, 1988), and such idealization could contribute to reports that are in fact more positive than partners’ reports. Thus, attachment insecurity could be associated with a systematic directional bias in reporting (i.e., over- or under-reporting, relative to the relationship partner). Third, the directionality of discrepancies also may be influenced by attachment-specific emotion regulation strategies. One strategy, avoidance, is characterized by a suppression of distressing emotions, whereas a second strategy, anxiety, is characterized by heightened positive and negative emotional expressiveness. Attachment avoidance and attachment anxiety are considered “minimizing” and “maximizing” strategies, respectively (Cassidy, 1994), and it may be that these approaches to relationships are associated with under- or over-reporting in comparison to another reporter when asked about relationship qualities.

To date, three studies have tested the hypothesis that attachment insecurity would be associated with discrepancies in informant reports (Barker, Bornstein, Putnick, Hendricks, & Suwalsky, 2007; Berger et al., 2005; Ehrlich et al., 2011). Berger et al. (2005) found that adolescent attachment (assessed with the AAI) predicted discrepancies in parent and adolescent reports about adolescents’ symptoms. Ehrlich et al. (2011), who also used the AAI, found that adolescent attachment was a reliable predictor of absolute discrepancies in reports about adolescent depressive symptoms, parent-adolescent conflict, and adolescents’ externalizing behavior. Only one study has incorporated the use of a self-report assessment of attachment in the prediction of informant discrepancies. In this study, Barker et al. (2007) found that adolescent self-reports on the Security Scale (Kerns, Klepac, & Cole, 1996) did not uniquely predict discrepancies in mother-adolescent reports of adolescents’ internalizing and externalizing symptoms when tested with five related predictors of discrepancies (e.g., adolescent-reported negative communication with mothers, mother-reported adolescent effortful control – a measure of adolescents’ emotion regulation abilities). It is possible that self-reported attachment was unrelated to informant discrepancies in the Barker et al. study because the other predictors in the model tapped components of attachment that would theoretically contribute to informant discrepancies, such as open communication and emotion regulation.

An alternative possibility is that self-reports of attachment do not reliably predict informant discrepancies, and indeed, there is evidence that the AAI and self-report attachment measures are typically not related to each other and may therefore assess different aspects of attachment (e.g., Roisman et al., 2007). However, it is possible that the common elements of attachment that both the AAI and self-report methods tap are those that would contribute to informant discrepancies (e.g., lack of open communication, suppression vs. heightening of emotions; for a review of data supporting this proposition, see Mikulincer & Shaver, 2007). The present study builds on the findings of Barker et al. (2007) by testing whether self-reports of attachment and depressive symptoms, in the absence of highly related constructs (e.g., open communication, emotion regulation), can predict discrepancies in reports about relationships (rather than reports about symptoms).

We present data from two studies in which multiple informants’ reports about a variety of adolescent relationships were available (i.e., adolescent–peer reports of social acceptance [Study 1], adolescent–parent reports of parental behavior [Study 2], and adolescent–friend reports of friendship [Study 2]). Our goal was to mesh with the extensive literature examining depressive symptoms as a predictor of informant discrepancies, and to extend this research by examining the role of informant attachment as an additional predictor of discrepancies. Attachment security and depressive symptoms are often moderately correlated in adolescence (Allen, Porter, McFarland, McElhaney, & Marsh, 2007), and so an important question becomes whether informant depressive symptoms and attachment avoidance and anxiety can uniquely predict absolute and directional discrepancy scores. This examination differs from and builds on the extensive research on the ways in which attachment influences perceptions of the self and others (see Cassidy & Shaver, 2008, for a review) by examining how attachment predicts the distance between two informants’ reports.

Given previous research on the depression-distortion bias, we predicted that informant depressive symptoms would be associated with both absolute and directional discrepancies, such that increases in symptoms would be associated with more discrepant and negatively biased perceptions. In relation to attachment avoidance and anxiety, we predicted that each would be associated with increases in absolute discrepancies because of a lack of open communication associated with attachment insecurity. Predictions about attachment and directional discrepancies differed across studies. For Study 1, we predicted that higher attachment avoidance and attachment anxiety would be related to lower reporting of acceptance relative to peers because both individuals high on attachment avoidance and on attachment anxiety hold negative views of themselves in social relationships (Mikulincer & Shaver, 2007). For Study 2, our predictions differed for avoidant and anxious individuals as a function of the emotion regulation patterns characteristic of these individuals (Mikulincer & Shaver, 2007, 2008). Specifically, we predicted that higher attachment avoidance would be associated with lower reports of negative relationship characteristics compared to other reporters, whereas higher attachment anxiety would be associated with higher reports of both positive and negative relationship characteristics compared to other reporters.

Study 1

Method

Participants and procedure.

Adolescents in 11^th grade (N = 1881; 60% girls; 65% White) from seven high schools participated in a school-based assessment of social relationships. Approximately 1400 of these students received attachment questionnaires; missing data yielded a final sample of 1156 students (62% girls) for this report. Adolescents completed all questionnaires during two 50-minute classroom sessions in the spring of their 11^th grade year.

Measures

Experiences in Close Relationships Scale (ECR; Brennan, Clark, & Shaver, 1998).

Adolescent attachment avoidance and anxiety were measured using the ECR, a 36-item scale that taps secure versus insecure dimensions of attachment styles. The avoidance subscale (18 items; α = .90) measures the extent to which a person is uncomfortable with closeness, intimacy, and dependence on close relationship partners (e.g., “I prefer not to show close relationship partners how I feel deep down”). The anxiety subscale (18 items; α = .90) measures the extent to which a person worries about being rejected, abandoned, or unloved (e.g., “I worry a fair amount about losing my close relationship partners”). For each item, participants rated their degree of avoidance or anxiety using a 7-point Likert-type scale, with responses ranging from 1 (disagree strongly) to 7 (agree strongly). This questionnaire has demonstrated excellent psychometric properties, including test-retest reliability and construct validity (Brennan et al., 1998).

Children’s Depression Inventory (CDI; Kovacs, 1985).

Adolescents completed the 27-item CDI, which taps self-reported symptoms of depression (e.g., disturbed mood, negative self-evaluative thoughts). School administrators requested that we drop the item related to suicidal ideation, leaving 26 items (α = .85). For each item, adolescents chose one out of three responses that best described their feelings over the past two weeks (e.g., “I have fun in many things,” “I have fun in some things,” and “nothing is fun at all”). Each item is scored from 0 – 2, with higher scores indicating higher depressive symptoms. Approximately 7% of the sample had depressive symptom scores of 20 and greater, which are considered to be in the range of clinical depression (Kovacs, 1985).

Self-Perception Profile (Harter, 1982).

Self-reports of social acceptance were measured with the widely used five-item peer acceptance scale from the Self-Perception Profile (α = .80). Using the “some kids, other kids” format (constituting a 4-point scale), the questions on this scale tap adolescents’ perceptions of how well they are liked by their peers (e.g., “Some teenagers feel that they are socially accepted”). Harter (1982) demonstrated extensive reliability and validity data for this scale.

Peer-Reported Social Acceptance (Parkhurst & Asher, 1992).

Peer perceptions of adolescents’ acceptance were calculated using a modified version of a sociometric procedure (Parkhurst & Asher, 1992). Acceptance scores were derived from responses provided by adolescents’ classmates, who used randomly generated rosters of names of 75 students to identify the extent to which they endorsed, for each student, the statement “I like to be involved in activities with this person.” Responses could range from 1 (not at all) to 5 (very much), and respondents also had the option to identify students on their rosters whom they did not know. Each adolescent participating in the study was listed on at least 75 rosters and thus had ample opportunity to be nominated by peers. We created an average peer-reported social acceptance score from these responses, which was then standardized by school. This measure of social acceptance is negatively correlated with other peer-reported measures of social behavior, such as aggression, victimization, and shyness (Dykas, Ziv, & Cassidy, 2008).

Absolute and directional discrepancy score calculation.

When identical measures across informants are unavailable, comparable measures can be substituted, using standardized z scores to create equal variances across the measures (De Los Reyes & Kazdin, 2004; Jensen, Traylor, Xenakis, & Davis, 1988). As recommended by De Los Reyes and Kazdin (2004), we calculated discrepancy scores by taking the difference between two informants’ standardized scores. Directional discrepancy scores were computed by subtracting adolescents’ scores from the other informants’ scores (i.e., peer). As such, positive directional discrepancy scores indicate that adolescents under-reported, and negative scores indicate that adolescents over-reported, relative to other informants.

Results and Discussion

Means, standard deviations, and intercorrelations among study variables can be found in Table 1. We used multiple regression (controlling for gender and race) to examine the role of adolescent depressive symptoms and attachment avoidance and attachment anxiety as unique predictors of absolute and directional discrepancies in adolescent and peer reports of the adolescent’s acceptance.

Table 1.

Study 1: Descriptive Statistics and Intercorrelations Among Principal Variables

Variable	M	SD	1	2	3	4	5	6	7
1. Adolescent Gender	.39	.49	-	.01	−.05	−.03	−.11***	−.05†	−.07**
2. Adolescent Race	.68	.47		-	−.13***	−.01	−.08**	.08**	−.03
3. AR ECR Avoidance	3.1	1.0			-	−.10***	.21***	−.22***	−.02
4. AR ECR Anxiety	3.5	1.1				-	.44***	−.23***	−.07*
5. AR Depressive Symptoms	10.2	6.5					-	−.40***	−.13***
6. AR Social Acceptance	15.5	3.0						-	.29***
7. PR Social Acceptance	.03	1.0							-

Note. AR = Adolescent-reported. PR = Peer-reported. Adolescent gender was coded as 0 = female, 1 = male. Adolescent race was coded as 0 = non-White, 1 = White.

^†p = .055.

^*p < .05.

^**p < .01.

^***p < .001.

Missing data.

Analyses using Little’s MCAR test indicated that data were “missing completely at random,” χ²(40) = 45.08, p = .27. As such, we used listwise deletion to handle missing data, which generates unbiased estimates when data are MCAR (Hayes, Slater, & Snyder, 2008).

Inclusion of covariates.

Because child demographic characteristics can be predictive of discrepancies (De Los Reyes & Kazdin, 2005), we included adolescent gender and race as covariates in the following analyses. We did not include adolescent age in analyses because all adolescents were in their junior year of high school (all but a few adolescents were 16–17 years old).

Preliminary analyses.

A number of gender and race differences emerged in adolescent- and peer-reported levels of social acceptance. First, compared to boys, girls reported that they were more socially accepted, t(1670) = 1.92, p = .055, d = .10. Similarly, peers rated girls as more socially accepted than boys, t(1523) = 2.82, p = .004, d = .15. In addition, minority adolescents self-reported lower social acceptance than did White adolescents, t(1633) = −3.34, p = .001, d = .18; peer-reports, however, revealed no race-related differences in social acceptance, t(1495) = 1.32, p = .19, d = .07. Examination of covariates included in the regression analyses revealed that relative to White adolescents, minority adolescents had greater absolute and directional discrepancies in reports of social acceptance, such that their reports were both more discrepant from peer reports, and they reported having lower social acceptance than their peers reported about them (see Table 2).

Table 2.

Study 1: Absolute and Directional Discrepancies in Adolescent and Peer Reports of Social Acceptance

	Social Acceptance
	Absolute			Directional
	β	R2	sr2	β	R2	sr2
Step 1		.01**			.01**
Adolescent Gender	.00		.00	.01		.00
Adolescent Race	.09**		.01	−.11**		.01
Step 2		.02**			.08***
AR ECR Avoidance	.02		.00	.12***		.01
AR ECR Anxiety	.05		.00	.03		.00
AR Depressive Sx	.06**		.00	.20***		.03

Note. AR = Adolescent-reported.

^*p = .05.

^**p < .01.

^***p < .001.

Principal analyses predicting discrepancies.

Regression analyses indicated that increases in adolescents’ depressive symptoms were associated with increases in absolute and directional discrepancies in social acceptance reports. Specifically, greater depressive symptoms were associated with under-reporting social acceptance relative to peer reports. Adolescent attachment avoidance predicted directional discrepancies in reports of adolescent social acceptance, such that greater adolescent avoidance scores were associated with under-reporting social acceptance relative to peer reports (see Table 2). Attachment anxiety was unrelated to discrepancies.

These findings provide support for the depression distortion hypothesis, such that adolescents with greater depressive symptoms perceived their peer acceptance in a less positive light than their peers did. The finding that adolescents’ attachment avoidance scores were associated with adolescents’ under-reporting of social acceptance relative to peer reports is consistent with with past research showing that individuals high in attachment avoidance hold negative views of themselves within social relationships (Mikulincer & Shaver, 2007).

Study 2

Method

Participants and procedure.

One-hundred and sixty-one adolescents (age: 14–18; M = 16.1, SD = .99; 56% girls; 27% White) and their parents (145 mothers, 14 fathers, 2 other) participated in a larger study about adolescent risky behavior. Target adolescents and parents completed questionnaires in the laboratory. Technical problems resulting in missing data yielded a sample of 119. During the laboratory session, adolescents nominated their same-sex best friend, who was later invited to complete questionnaires that were sent through the mail. We received complete data from 49 friends of the target adolescents (60% girls), 80% of whom reported that the target adolescent was also their best friend. A Pearson chi-square test indicated that we received packets from the friends of White adolescents more frequently than from minority adolescents, χ²(1, N = 153) = 10.25, p =.001. No other differences emerged. Means, standard deviations, and intercorrelations among study variables can be found in Tables 3 and 4.

Table 3.

Study 2: Descriptive Statistics and Intercorrelations Among Parent and Adolescent Variables

Variable	M	SD	1	2	3	4	5	6	7	8	9	10
1. AR ECR Avoidance	3.1	.97	-	.23**	.22**	.03	.10	.00	.20*	−.06	.14	.00
2. AR ECR Anxiety	3.2	1.2		-	.21**	−.03	.04	.02	.35***	−.01	−.01	−.06
3. AR Depressive Sx	3.9	2.5			-	−.10	.00	.01	.16†	−.22**	.02	.07
4. PR ECR Avoidance	2.6	1.1				-	.43***	.01	−.09	−.09	.00	−.20*
5. PR ECR Anxiety	2.7	1.1					-	.04	−.06	−.10	.03	−.13
6. PR Depressive Sx	5.3	4.5						-	−.10	−.11	.35***	−.18*
7. TR Negative Parenting	2.8	.81							-	−.11	.27**	−.01
8. TR Positive Parenting	4.6	1.3								-	−.10	.20*
9. PR Negative Parenting	2.5	.83									-	−.13
10. PR Positive Parenting	5.6	.81										-

Note. AR = Adolescent-reported. PR = Parent-reported.

^†p = .058.

^*p = .05.

^**p < .01.

^***p < .001

Table 4.

Study 2: Descriptive Statistics and Intercorrelations Among Friend and Adolescent Variables

Variable	M	SD	1	2	3	4	5	6	7	8	9
1. AR ECR Avoidance	3.1	.97	-	.23**	.22**	−.03	−.18*	.04	.24	.00	.13
2. AR ECR Anxiety	3.2	1.2		-	.21**	.30***	−.15†	.26**	−.15	−.44**	.00
3. AR Depressive Sx	3.9	2.5			-	.08	−.20*	.07	−.08	.02	−.24
4. AR Conflict	1.6	.69				-	−.19*	.76***	.08	−.24	.25
5. AR Reliable Alliance	3.9	1.1					-	−.17*	.03	.33*	.11
6. AR Antagonism	1.7	.75						-	.02	−.02	.29*
7. FR Conflict	1.4	.46							-	.05	.51***
8. FR Reliable Alliance	4.0	1.0								-	−.12
9. FR Antagonism	1.6	.52									-

Note. AR = Adolescent-reported. PR = Parent-reported.

^†p = .08.

^*p ≤ .05.

^**p < .01.

^***p < .001

Measures

Coping with Children’s Negative Emotions Scale – Adolescent Version (Fabes & Eisenberg, 1998).

Parents were asked to respond to nine hypothetical situations in which their adolescent may experience distress (e.g., adolescent was hurt by a friend). For each item, parents reported their likelihood of responding in six different ways using a 7-point scale. We calculated two summary scores: Negative Parenting Responses (mean of the minimizing, punitive, and parental distress subscales; α = .73) and Positive Parenting Responses (mean of the expressive encouragement, emotion-focused, and problem-focused subscales; α = .87). Adolescents completed the same items about their parents’ responses to their distress, and we computed the summary scores described above (α’s = .67 and .94, respectively). Previous research with the child version of this scale demonstrated good psychometric properties (Eisenberg, Fabes, & Murphy, 1996).

Experiences in Close Relationships Scale (Brennan et al., 1998).

Adolescents and parents completed the ECR (described above; α’s > .86).

Youth Self Report: Depression subscale (YSR; Achenbach & Rescorla, 2001).

Adolescents reported about their symptoms of depression using the 8-item subscale from the YSR (α = .65). T Scores of 70 and greater are typically considered clinically significant (Achenbach, 1995), which characterized only 2.0% of this sample.

Center for Epidemiological Studies Scale for Depression (CES-D; Radloff, 1977).

Parents provided information about their own depressive symptoms with the CES-D, a widely used 20-item assessment of recent depressive symptoms using a 4-point scale (α = .80). We calculated total scores for parent CES-D depressive symptoms such that higher scores reflect higher depressive symptoms. Parent depressive symptoms scores above 15 are typically considered clinically significant (McDowell, 2006), which characterized 3.3% of the sample.

Network of Relationships Inventory (NRI; Furman & Buhrmester, 1985).

Adolescents and their friends completed the NRI, a 30-item questionnaire measuring perceptions of social support and negative interactions in relationships. Three of the ten conceptually distinct scales measure qualities about the relationship rather than the friend (α’s > .74), and therefore were appropriate for this investigation: conflict (e.g., “how much do you and this person disagree and quarrel?”), antagonism (e.g., “how much do you and this person get on each other’s nerves?”), and reliable alliance (e.g., “how sure are you that your relationship will continue in the years to come?”).

Absolute and directional discrepancy score calculation.

As described in Study 1, we calculated discrepancy scores by taking the difference between two informants’ standardized scores. Absolute discrepancy scores are ≥ 0, with larger values indicating greater discrepancies in informants’ reports. For directional discrepancies, positive scores indicate that adolescents under-reported, and negative scores indicate that adolescents over-reported, relative to other informants.

Results and Discussion

Parent and adolescent reports about parenting responses.

Our examinations of (a) absolute and (b) directional discrepancies in adolescent and parent reports of parents’ responses to adolescent distress were conducted using multiple regressions with informant depressive symptoms and adolescent attachment as predictors (see Table 5).

Table 5.

Study 2: Discrepancies in Adolescent and Parent Reports of Parents’ Negative and Positive Responses to Adolescent Distress

	Negative Parenting Responses						Positive Parenting Responses
	Absolute			Directional			Absolute			Directional
	β	R2	sr 2	β	R2	sr 2	β	R2	sr 2	β	R2	sr 2
Step 1		.03			.04			.03			.10*
Adolescent Gender	.09		.01	−.10		.01	.03		.00	.20*		.04
Adolescent Age	.09		.01	.20*		.04	.06		.00	−.20*		.04
Adolescent Race: Black	.15		.02	.01		.00	.20†		.03	.20†		.03
Adolescent Race: Other	.06		.00	.03		.00	.13		.01	.19†		.03
Step 2		.11			.24***			.10			.13†
AR ECR Avoidance	.29**		.07	.02		.00	.05		.00	.03		.00
AR ECR Anxiety	.02		.00	−.31***		.09	−.15		.02	−.01		.00
AR Depressive Sx	−.06		.00	−.11		.01	.05		.00	.19*		.03
PR ECR Avoidance	−.02		.00	.01		.00	.20*		.03	−.02		.00
PR ECR Anxiety	−.06		.00	.09		.01	.04		.00	.01		.00
PR Depressive Sx	−.04		.00	.28**		.08	.02		.00	.00		.00

Note. AR = Adolescent-reported. PR = Parent-reported.

^†p < .09.

^*p ≤ .05.

^**p < .01.

^***p ≤ .001.

Missing data.

Analyses using Little’s MCAR test indicated that data were “missing completely at random,” χ²(44) = 37.94, p = .73. As such, we used listwise deletion to handle missing data (Hayes et al., 2008).

Inclusion of covariates.

As in Study 1, we included adolescent gender and race as covariates in the following analyses. In this study, however, we also included adolescent age in analyses because there was sufficient variability in this sample (unlike Study 1).

Preliminary analyses.

Examination of covariates in the regression analyses revealed that adolescent males reported less positive parenting than their parents reported (See Table 5). Older adolescents appeared to have more positive views of their parents, relative to younger adolescents: Older teens reported more positive parenting and less negative parenting than their parents reported. Relative to White adolescents, minority adolescents reported less positive parenting than their parents reported.

Principal analyses.

Adolescent depressive symptoms were associated only with directional discrepancies in reports of positive parenting responses, such that adolescents with higher symptoms reported less positive parenting responses, relative to parents. Parent depressive symptoms were associated only with directional discrepancies in their negative parenting responses, such that higher symptoms were associated with reporting increasingly negative responses relative to adolescents. These findings support the depression distortion hypothesis, with both parents and adolescents viewing parental responses less positively or more negatively as their depressive symptoms increased.

Examination of parent and adolescent attachment avoidance and anxiety as predictors of discrepancies about the parent-adolescent relationship revealed several significant findings. Adolescent attachment avoidance was associated with greater absolute discrepancies in reports of negative parenting responses; directional analyses were nonsignificant. Adolescent attachment anxiety was associated with greater directional discrepancies in reports of parents’ negative responses, such that as adolescent attachment anxiety increased, adolescents’ reports of negative parenting increased, relative to parent reports. This pattern of discrepant reporting is in line with previous findings that, relative to secure individuals, attachment anxious individuals view family members as expressing more negative and punitive responses (Bell, 1998; Priel & Besser, 2001). Parents’ attachment avoidance was associated with absolute discrepancies in reports of positive parenting; parent’s attachment anxiety was unrelated to discrepancies. It is interesting to note that a common finding across these analyses is the association between attachment avoidance and absolute discrepancies, a pattern that is consistent with the idea that attachment avoidance is associated with restricted communication in relationships (Mikulincer & Nachshon, 1991).

Adolescent and friend reports about friendship.

Finally, in relation to adolescent and friend reports of the friendship, our examinations of (a) absolute and (b) directional discrepancies were conducted using multiple regression (controlling for gender, age, and race), with adolescent depressive symptoms and attachment as predictors. (Friends’ attachment and depressive symptoms were unavailable in this study.)

Preliminary analyses.

Examination of covariates revealed that older adolescents had more absolute discrepancies in their reports of the reliable alliance of the friendship (see Table 6). In addition, boys’ reports were marginally more discrepant than girls’ reports of the reliable alliance. White and Black adolescents had marginally fewer absolute discrepancies in their reports of the reliable alliance relative to other minority adolescents.

Table 6.

Study 2: Absolute and Directional Discrepancies in Adolescent and Friend Reports of Friendship Qualities

	Conflict in the Friendship						Reliable Alliance
	Absolute			Directional			Absolute			Directional
	β	R2	sr 2	β	R2	sr 2	β	R2	sr 2	β	R2	sr 2
Step 1		.13			.09			.23*			.08
Adolescent Gender	.21		.04	.01		.00	.28b		.07	−.09		.01
Adolescent Age	.01		.00	−.24		.05	.32*		.10	−.17		.03
Adolescent Race: Black	.17		.02	.25		.05	−.06		.00	−.04		.00
Adolescent Race: Other	−.24		.05	.00		.00	.29†		.07	.16		.02
Step 2		.25			.31a			.27			.27
AR ECR Avoidance	.33*		.10	.45**		.18	−.16		.02	.28c		.07
AR ECR Anxiety	−.17		.03	−.19		.03	.11		.01	−.35*		.11
AR Depressive Sx	−.09		.01	−.16		.02	.06		.00	.12		.01

Note. AR = Adolescent-reported. No significant findings emerged for absolute or directional analyses about antagonism in the friendship.

^ap = .058.

^bp = .07.

^cp = .08.

^*p < .05.

^**p < .01.

Principal analyses.

Adolescent depressive symptoms did not predict absolute or directional discrepancies. For attachment, significant associations emerged for two of the three friendship qualities. First, in relation to reports about conflict in the friendship, analyses indicated that as adolescents became more avoidant, their reports of conflict became more discrepant from their friends’ reports, and their reports of conflict became increasingly lower, relative to friend reports. This pattern of under-reporting conflict in the friendship relative to the friend’s report is consistent with the notion that avoidance is associated with suppression of negative emotions in close relationships and idealization of relationship partners (Cassidy & Kobak, 1988; Mikulincer & Shaver, 2007). Second, in relation to reports of the reliable alliance of the friendship, there was a marginally significant (p = .08) association between adolescent attachment avoidance and directional discrepancies, such that greater attachment avoidance was associated with adolescents reporting less confidence in the long-term reliability of the friendship relative to friends’ reports. This finding supports the notion that attachment avoidant individuals prefer not to rely on others, and that they are more likely than secure individuals to engage in short-term (rather than long-term) relationships (Mikulincer & Shaver, 2007). In contrast, there was a significant association between adolescent attachment anxiety and directional discrepancies, such that greater attachment anxiety was associated with reporting greater confidence in the long-term reliability of the friendship relative to friends’ reports. Given research showing that individuals high in attachment anxiety prefer their relationships to be characterized by high commitment (Mikulincer & Erev, 1991), it may be that these preferences contribute to a relative over-reporting on this dimension, compared to friend reports. Interestingly, data from several studies indicate that these individuals tend to perceive that their friends are less committed to the relationship than they themselves are (evident in endorsing ECR items such as “I often wish that close relationships partners’ feelings for me were as strong as my feelings for them”); the finding from the present study that the friends of these individuals do in fact report relatively lower confidence that the relationship will endure indicates that such perceptions may indeed be correct. No significant absolute discrepancies emerged for reports of reliable alliance. In relation to reports of antagonism in the friendship, no significant findings emerged.

General Discussion

The findings from these two studies have implications for researchers who rely on informant reports about adolescent relationships. Our analyses suggest that discrepancies in reports about adolescent relationships are not random and are often predictable as a function of individuals’ depressive symptoms and attachment avoidance and anxiety in ways consistent with theory and previous research. Across three sets of informant comparisons and across two very different samples, our ability to predict informant discrepancies from individuals’ depressive symptoms and attachment avoidance and anxiety suggests that these factors play a significant role in contributing to discrepant reports about their relationship experiences.

It is likely that reporter discrepancies about adolescent relationships reflect, in part, the ways in which reporters process relationship-related information; as such, it is interesting to consider the ways that these findings mesh with previous findings about patterns of attachment-related information processing (see Dykas & Cassidy, 2011, for a review). For instance, the finding that attachment avoidance was associated with more negative perceptions of the self than adolescents’ peers perceived them to be converges with previous findings that avoidance is associated with negatively biased attention to and memory for information about the self (e.g., Pereg & Mikulincer, 2004). Moreover, the extent to which social information is shared versus restricted (i.e., open vs. closed communication) is likely to influence absolute discrepancies; the findings that avoidance was associated with absolute discrepancies with (a) parents about positive and negative parenting and (b) friends about conflict in the friendship (similar to findings of Berger et al., 2005) suggests that avoidant individuals may be restricting communication with others, as has been proposed in previous attachment research (Mikulincer & Nachshon, 1991; see Bretherton, 1990, for a theoretical discussion of links between secure attachment and open communication). Finally, the findings that attachment avoidance is associated with discrepancies in which negative emotions are minimized in relation to another reporter and attachment anxiety is associated with discrepancies in which positive and negative emotions are maximized in relation to another reporter can be viewed as reflections of biases in perceptual processing that are consistent with much previous research (Dykas & Cassidy, 2011; Mikulincer & Shaver, 2007).

Additional research is needed to examine whether particular correlates of attachment, such as the open communication or emotion regulation capacities that typically characterize secure attachment (Cassidy, 1994), can be identified as best predicting discrepant reports. It could be that one of these attachment-related factors is most central to predicting whether informants will report in similar or discrepant ways. For example, if an adolescent is upset with a relationship partner but never expresses these negative feelings, then reports from the adolescent and the relationship partner might be discrepant as a direct result of the lack of open communication. Alternatively, it may be that no single component can account for informant discrepancies. Instead, it could be the constellation of emotions, cognitions, and behaviors that make up an individual’s attachment style that ultimately contributes to informant discrepancies. Examination of these proximal factors will be particularly informative for clinicians, who want to help relationship partners minimize disagreement in perceptions of their interactions. This clinical application remains to be tested empirically but will be important to examine in future studies.

Across studies, we found some support for the depression distortion hypothesis. Yet the fact that our findings sometimes diverged from previous work showing support of the depression distortion hypothesis (e.g., Treutler & Epkins, 2003) may reflect our community samples’ relatively low depressive symptomatology. Alternatively, it may be that informant depressive symptoms are less relevant when the focus of reporting is on interpersonal relationships rather than on symptoms. Replications of this work in clinical samples will be important.

We identified a number of racial differences in adolescents’ reports about relationships, which, although not the focus of the present study, merit a brief discussion. When discrepant reports emerged as a function of race, minority adolescents had more negative views, relative to White adolescents. In Study 1, minority adolescents had negative perceptions about their social acceptance (relative to peer reports), a finding that is consistent with previous research on minority children’s expectations of social exclusion based on race (Killen, Henning, Kelly, Crystal, & Ruck, 2007). In Study 2, minority adolescents’ reports of positive parenting responses were lower than their parents’ reports. This finding may be related to ethnic differences in parenting styles, with Black, Hispanic, and Asian adolescents reporting that their parents engage in more authoritarian and less authoritative parenting than White adolescents reported about their parents (e.g., Dornbusch, Ritter, Liederman, Roberts, & Fraleigh, 1987). It may be that minority adolescents compare their experiences with parents to their perceptions of White friends’ parents (who display less authoritarian parenting), resulting in ratings of positive parenting that are lower than their parents’ reports. Additional research on discrepant reporting and the role of race will be important.

Several study limitations should be addressed in future research. In Study 1, we did not have identical measures to compare adolescent and peer reports of adolescent social acceptance. Although assessments were comparable across informants, future research could use an identical question to measure reports about adolescent social acceptance. We note that across analyses, effect sizes are generally in the small to medium range, so it will be important for future research to consider additional explanatory factors that may contribute to informant discrepancies. Further, in some analyses, one attachment dimension but not the other (e.g., attachment avoidance but not anxiety) significantly predicted discrepancies; future work should look at the conditions under which each dimension might play a central role in contributing to discrepancies. In addition, our study design included only one parent’s reports for each adolescent, and it will be important for future studies to compare the ways in which attachment and depressive symptoms predict discrepancies in mother-adolescent and father-adolescent dyads separately. Similarly, the small sample size for the friendship analyses prevented us from examining whether discrepancies in reports differed depending on whether the dyad mutually reported that it was a “best” friendship, and it may be that the pattern of discrepant reporting is quite different for best friends versus non-best friends. Relatedly, time constraints prohibited us from measuring friends’ depressive symptoms and attachment, and it will be important to examine how both friends’ characteristics contribute to discrepancies in reports.

The analyses presented here cannot reveal whether one informant is more “accurate” than the other. In some research, each individual reports on his or her perception of an unobservable state, and assessment of a “gold standard truth” is not possible. In many cases, however, especially when the topic of reports is behavior, creation of a relatively accurate approximation of “truth” may be possible. In the present study, for instance, parents and adolescents reported about parental responses to adolescent distress – behaviors that observers presumably would be able to observe given sufficient naturalistic observations of the dyad. Parent and adolescent reports could then be examined to assess whose report more closely matched behavioral observations. Such work will be important for the future.

In sum, these findings suggest that informants may report about adolescent relationships with an “attachment-colored lens.” This research on the role of attachment as a contributor to informant discrepancies will be important to extend to additional age groups, such as childhood, early adolescence, and adulthood. Although attachment may influence parent-adolescent informant discrepancies at any age, it is also possible that attachment only becomes a salient predictor of discrepancies during adolescence, when communication between parents and adolescents decreases (Collins, 1990). Similarly, in light of the fact that adolescents have more interactions with peers than they did in childhood (Larson & Richards, 1991) and thus have more experiences to draw on when reporting, it could be that systematic discrepancies with peers are not evident before adolescence. Similarly, examination of the role of attachment in informant discrepancies of reports about additional relationships in adolescence beyond those examined here, such as sibling and romantic relationships, will be useful in order to learn whether these findings generalize across other important adolescent relationships.