Introduction

Parent–child discrepancies in reports of family dynamics are pervasive. Parents and adolescents routinely disagree in their reports of many family constructs, including cohesion and conflict (Laursen et al. 1998; Ohannessian et al. 1995), parental monitoring and teen disclosure (De Los Reyes et al. 2008; Keijsers et al. 2010), parental authority legitimacy (Rote and Smetana 2016; Smetana 2011), and parents’ influence over child behaviors and family decisions (Holmbeck and O’Donnell 1991; Smetana et al. 2004). These differences are theorized to occur due largely to parents’ and adolescents’ different generational stakes (Bengtson and Kuypers 1971). More specifically, their larger investment in the family leads parents to view family relationships more positively than do teens, whereas because youth desire more autonomy and individuality, they view the family as less cohesive and more conflictive than do their parents (Laursen et al. 1998). At moderate levels, such discrepancies can be adaptive for adolescent autonomy development and the realignment of parent–adolescent relationships (Holmbeck and O’Donnell 1991; Smetana 2011; Steinberg 2001).

Parent–child discrepancies about family constructs also vary somewhat by issue. Parents and adolescents typically agree that parents have the least authority over children’s choice of friends and free-time activities (described from the social domain theory framework as personal issues; see Smetana et al. 2014), and the most authority over behaviors that violate social norms, moral codes, or are dangerous or risky for the child (referred to as social-conventional, moral, and prudential issues, respectively). However, they disagree most about parents’ authority over behaviors that involve both personal choice and elements of risk or conventionality, such as dating, observing curfews, or room cleanliness (all referred to as multifaceted issues; Smetana 2011). Discrepancies in adolescents’ and parents’ reports about family decision-making, or whether parents or teens make decisions alone, jointly, or with the other’s input, likewise vary; they are greatest surrounding personal issues in early adolescence but around risky behaviors in late adolescence (Smetana et al. 2004). Variations in parent–adolescent discrepancies about parental behavioral control have not yet been examined by topic, but levels of parental control vary across these same issues and differ accordingly in links with adjustment (Arim et al. 2010).

A certain amount of disagreement between parents and adolescents, as assessed regarding different family constructs, therefore appears to be normative and potentially adaptive. Greater discrepancies in levels of parent–adolescent disagreement about family functioning are nevertheless associated with more adolescent internalizing and externalizing problems (De Los Reyes et al. 2010; Guion et al. 2009; Ohannessian 2012). Therefore, discrepancies in perceptions of family functioning beyond those explained only by generational differences must be considered, as they may be particularly relevant to adolescent and family functioning. Furthermore, the extent to which informant discrepancies stem from normative generational differences versus more problematic sources may depend on the family construct considered, and such variation must be taken into account.

Consistency and Sources of Discrepancies Across Family Constructs

Family-related constructs, such as beliefs about parental authority, parental knowledge, and parent–child relationship quality are intimately related (Darling et al. 2008; Keijsers and Laird 2014; Kuhn and Laird 2011). Nevertheless, research on parent–child discrepancies has focused on these constructs in isolation. Little research has examined whether there is consistency across constructs within a family in the size and direction of discrepancies or in their associations with parenting and adjustment. In support of moderate consistency, Guion et al. (2009) successfully modeled parent–child discrepancies about harsh discipline, inconsistent discipline, and nurturance as a single latent factor. Likewise, De Los Reyes et al. (2010) found that mother–adolescent discrepancies on three monitoring-related constructs (disclosure, solicitation, and parental knowledge) formed an internally consistent scale and that teens consistently over- or under-reported on all three constructs relative to mothers. These analyses focused on highly related family practices, however. We do not know the extent of consistencies in informant discrepancies when more disparate aspects of family functioning are considered.

Theorizing about the sources of informant discrepancies sheds some light on these questions. A frequent explanation for informant discrepancies in child adjustment is that different informants observe children’s behavior in different contexts (De Los Reyes 2013). Family dynamics function somewhat differently from child adjustment, as they often reflect shared interactions between parents and children. However, there is considerable variation in the extent to which family constructs involve shared experiences and knowledge. Specifically, certain family constructs of concern to many developmental scientists, such as parents’ monitoring knowledge, have both a shared context (i.e., parents know about teens’ activities when adolescents disclose the relevant information), but also unshared contexts that are available only to teens or parents (e.g., adolescents’ actual behavior in situations or information parents gather from third parties; Waizenhofer et al. 2004). Other family constructs have no shared context in that they are not directly observable but must be inferred (e.g., authority legitimacy) or consist entirely of reports about one shared context (e.g., positive parent–child interactions).

Indeed, there is evidence that the extent to which a context is shared, or the objectivity of the construct, impacts the extent to which parents and adolescents provide discrepant reports about family behaviors. Jessop (1982) ranked family topics according to their perceived level of subjectivity and found that parents’ and adolescents’ reports about individual family members’ behaviors (which they ranked as most objective) showed greater agreement than reports about family interactions or relationships, which were ranked as moderately subjective. These, in turn, showed greater agreement than adolescents’ and parents’ self-reported attitudes, which were seen as completely subjective. Likewise, a recent intervention study empirically demonstrated that drawing mothers’ and adolescents’ attention to the (differential) contexts they perceive as prime sources of parental knowledge exacerbates parent–adolescent informant discrepancies on monitoring-related constructs (De Los Reyes et al. 2013). Finally, when considering mothers’ and adolescents’ perceptions of family communication behaviors, discrepancies appear to increase more over time for family constructs involving potentially covert behaviors (e.g., extent of open communication) than overt behaviors (e.g., communication problems; De Los Reyes et al. 2016). We therefore expected that parent–adolescent discrepancies about family constructs that rely less on shared contexts, specifically parental knowledge and parental right to know, would be particularly influenced by factors linked with adolescent disclosure, such as adolescent problem behavior and maternal psychological and behavioral control (Darling et al. 2008; Soenens et al. 2006).

Another explanation for informant discrepancies is perceptual bias—that is, that persons notice, interpret, and recall the same behaviors differently (De Los Reyes 2013). For instance, maternal dysphoria accounts for a small, but significant portion of the variation in mothers’ reports of their children’s positive and negative behaviors and emotions (Youngstrom et al. 1999). Although informant biases do not seem to account for a great deal of variation in informant discrepancies in reports of child adjustment (De Los Reyes 2013), such biases may have more of an impact on ratings of family interactions, in which depression-related distortions in experienced and expressed emotion (Nicholson et al. 2011), as well as in emotion perception (Richters 1992), may come into play. Indeed, a recent study shows that both maternal and adolescent depressive symptoms independently relate to greater mother–adolescent discrepancies about dyadic family monitoring behaviors such as parental knowledge and disclosure (De Los Reyes et al. 2008). Thus, mothers’ and teens’ poor psychological health would be expected to relate more to discrepancies in family constructs involving shared experiences (i.e., parental knowledge and positive interactions) than parental authority beliefs.

Use of Person-Based Analyses

Person-based analyses are well-suited for examining the consistency and sources of informant discrepancies across multiple constructs. Unlike variable-based regression analyses, in which the direction and magnitude of difference scores often must be examined separately (e.g., Ohannessian 2012), person-based analyses allow both the magnitude and direction of informant discrepancies to be modeled within a single analysis. Furthermore, in person-based analyses, dyads are assigned a probability of belonging to each discrepancy profile. When these probabilities are considered across multiple constructs, they can be used to estimate the consistency of discrepancy profile membership across constructs within families. Therefore, the present study employed person-based analyses to model and predict patterns of informant discrepancies across multiple family constructs.

Current Study

The current study had three aims. The first was to use latent profile analysis (LPA) to describe patterns of discrepancies in mother–adolescent dyads over each of three different types of family issues that vary in their level of subjectivity and shared context: beliefs about parents’ RTK, maternal knowledge, and positive mother–adolescent relationships. Based on previous person-based analyses of family relationship discrepancies (e.g., De Los Reyes et al. 2010), we expected to find three profiles for each of these constructs: one reflecting adolescents’ lower ratings than mothers, one reflecting adolescents’ higher ratings than mothers, and one reflecting relative agreement in adolescents’ and mothers’ ratings. Our profiles were based on standardized difference scores (SDS), which center informants’ scores relative to the mean of their groups before comparison (De Los Reyes and Kazdin 2004) and reflect variations in discrepancy scores that exist beyond those which would be normatively expected based on generational stakes. Therefore, we expected that relative agreement would be the most frequent profile, as it reflects a normative generational discrepancy but no additional mother–adolescent disagreement, and that discrepancies would be relatively consistent across topics within a dyad. Based on the extent to which contexts are shared, we additionally hypothesized that discrepancies (beyond those that would be normatively expected) would be greatest for maternal right to know, less for maternal knowledge, and least for positive interactions.

The second aim of the current study was to examine whether similar adjustment and relational factors were associated with dyad membership for each construct. We examined whether membership in the different profiles was predicted by adolescent adjustment (depressed mood and problem behavior), maternal adjustment (well-being), and parenting (adolescents’ and mothers’ perceptions of maternal psychological and behavioral control). Adolescents typically view parents as less knowledgeable about their activities (Keijsers et al. 2010), less supportive (Guion et al. 2009), and having less of a right to know (Rote and Smetana 2016) than do mothers, but adolescent problem behavior appears to further reduce teens’ evaluations of such constructs (Branje et al. 2008; Chan et al. 2015; Laird et al. 2003a, b). Therefore, regardless of family construct, we hypothesized that adolescents would evidence poorer adjustment (particularly more problem behavior) and mothers would be seen as more psychologically controlling in dyads belonging to profiles where adolescents rated these dimensions lower than mothers. As previously noted, we also expected that lower levels of adolescent depressed mood and greater maternal well-being would be more strongly associated with discrepancies about constructs involving shared contexts (positive interactions, and to a lesser extent, knowledge) and that teen problem behavior and parental control (psychological and behavioral) would be more strongly associated with constructs involving unshared contexts and maternal supervision (right to know and maternal knowledge).

Finally, the third aim of the current study was to examine dyads’ stability or variability in discrepancies across these different types of areas. There is little comparable person-centered research on discrepancies to draw on in formulating hypotheses. Based on evidence that discrepancies across family constructs tend to cohere and are associated with similar negative outcomes (De Los Reyes et al. 2010; Guion et al. 2009; Ohannessian 2012), we expected to find considerable stability in profile membership across the three family-constructs.

The overarching goal of the current study was therefore to compare and predict profiles of mother–adolescent discrepancies across family constructs that vary in the extent to which they involve shared family contexts. These analyses were considered exploratory, given the relative paucity of person-centered analyses in this area, as well as the relatively homogenous sample and cross-sectional nature of the data.

Methods

Sample

Participants were 167 mothers or stepmothers (M age = 46.37 years, SD = 5.77) and their 10th or 11th grade adolescents (M age = 15.68 years, SD = .64, 53 % female), drawn from two suburban high schools in a Northeastern city and studied three times over 1 year. Adolescents were 75 % European American, 8 % Asian or Pacific Islander, 8 % African American, 1 % Native American, and 4 % other ethnicities; 5 % identified as Latino. Families were mostly married (81 %); nearly all youth (98 %) lived with their mothers; the remaining youth lived with fathers and stepmothers or other family members. Most mothers had at least some college education; 15 % of mothers had only a high school education or less. The majority of mothers (57 %) worked full-time, 17 % reported not being currently employed, and the remainder worked part-time; median household income was between $70,000 and $89,000 a year and 6 % of the sample had household incomes less than $30,000 a year.

Procedures

Data for this study were taken from the first wave of a 1-year longitudinal study of families with adolescents. Families were recruited through letters sent home to parents of 10th and 11th graders from two high schools and presentations at school during the spring of 2010. Interested parents replied on a secure online site or phoned the project office and were mailed permissions. Families were enrolled in the study once these were returned; at least one parent had to agree to participate along with their adolescent.

Although both mothers and fathers were encouraged to participate, only 64 % of study families (n = 112) included fathers. Because the analyses were computationally intense, only adolescents and their mothers were examined here. Importantly, however, a focus on mother–adolescent discrepancies is consistent with previous work on informant discrepancies regarding family variables (e.g., De Los Reyes et al. 2016; Laird and De Los Reyes 2013; Laird and LaFleur 2016).

Due to limitations in the funding available for honoraria, participation in the study was capped, resulting in a participation rate of approximately 15 % of all eligible families in the two districts. Of the families expressing interest and subsequently sent consent and assent forms, the participation rate was 59 %. The demographic background of participating students matched the profiles of the two high schools, although the average GPA was somewhat higher. Families received a $35 honorarium for their participation.

The surveys were administered online using SurveyMonkey. Although families could complete paper versions, nearly all families (97 %) chose to respond online. Families were sent separate e-mail links for each participating family member and weekly reminders until the surveys were completed or they chose to discontinue participation.

Measures

Parents’ Right to Know (RTK) About Teens’ Activities

Adolescents and mothers rated how much parents have a right to know about 18 domain-specific items (see “Appendix”) on a 5-point Likert scale ranging from 1 (definitely not) to 5 (definitely yes). Alphas are in Table 1.

Table 1 Means, standard deviations, reliability, and cross-informant correlations and comparisons for variables contributing to difference scores
Full size table

Mothers’ Knowledge of Teens’ Activities

Adolescents and mothers rated how much mothers really know about the same 18 items on a 5-point Likert scale ranging from 1 (don’t know at all) to 5 (know all or mostly all). Alphas are in Table 1.

Positive Interactions

Adolescents and mothers rated their positive interactions with each other on the Companionship, Instrumental Aid, and Affection subscales (3 items each) of the Network of Relationships Inventory (NRI; Furman and Buhrmester 1985). Items include “how much do you play around and have fun with your mother?” “how much does your mother protect and look out for you?” and “how much does your mother like or love you?” Responses were rated on a 5-point scale ranging from 1 (little or none) to 5 (extremely much). Alphas are in Table 1.

Maternal Psychological Control

Adolescents and mothers rated the mother’s behavior on the 8-item Psychological Control Scale-Youth Self Report (PCS-YSR; Barber 1996). Items include “my mother is always trying to change how I feel or think about things” and “my mother blames me for other family members’ problems.” All items were rated on a 4-point Likert scale ranging from 1 (not at all like my mother/me) to 4 (very much like my mother/me). Cronbach’s alphas were. 82 and .80 for adolescents’ and mothers’ ratings; Ms (SD) = 1.64 (.55) and 1.48 (.44), respectively.

Behavioral Control

Adolescents and mothers rated the mother’s use of behavioral control on 19 domain-specific items modified from Arim et al. (2010) and shown in “Appendix”. Items were rated on a 4-point scale ranging from 1 (not at all like my mother/me) to 4 (very much like my mother/me). Cronbach’s alpha across all items was .77 for adolescents’ ratings and .78 for mothers’ ratings; Ms (SD) = 2.54 (.39) and 2.71 (.36), respectively.

Problem Behavior

Adolescents reported how often they engaged in 10 acts of minor deviance (e.g., marijuana or alcohol use, smoking, vandalism, minor theft, fighting, truancy) drawn from Mason, Cauce, Gonzales, and Hiraga (1996) and rated on a 5-point scale ranging from 1 (never happens) to 5 (happens very often). Cronbach’s alpha was .72, M (SD) = 1.35 (.36).

Depressed Mood

Adolescents rated their depressive symptoms on the 20-item Center for Disease Control—Depression Scale (CES-D; Radloff 1977). Teens rated how often they had felt or behaved regarding each item over the past week on a 4-point scale ranging from 1 (rarely or none of the time) to 4 (most or all of the time); Cronbach’s alpha was .90, M (SD) = 1.79 (.52). Example items include “I felt that everything I did was an effort” and “I had crying spells.” When scores were recoded (0-3), summed, and compared to a cutoff score of 16 (Radloff 1977), 65 adolescents (39 %) met the threshold for clinically significant “mild” depressive symptoms.

Well-Being

Mothers rated their well-being on the 18-item Psychological Well-Being Scale (Ryff and Keyes 1995). Mothers reported the extent to which psychological well-being items (e.g., “I am quite good at mastering the many responsibilities of my daily life” “I like most aspects of my personality”) characterized them on a 6-point scale ranging from 1 (strongly disagree) to 6 (strongly agree); Cronbach’s alpha was .82; M (SD) = 4.63 (.60).

Analytic Plan

Based on conceptual formulations (Smetana et al. 2014) and previous work on domain-specific ratings (Darling et al. 2005; Rote and Smetana 2016; Smetana and Rote 2015), parental knowledge and parental right to know items were grouped into personal, romantic, multifaceted, and prudential issues. Positive interactions items were grouped according to their Network of Relationships Inventory subscale. Reliability was acceptable for all issues and subscales (see Table 1). Items in each category or subscale were averaged and standardized for both mothers’ and teens’ reports. In line with recommendations for the computation of difference scores (De Los Reyes and Kazdin 2004) and current practices in their person-centered analysis (De Los Reyes et al. 2010), SDS for right to know beliefs and parental knowledge about each issue type and for all positive interaction subscales were computed by subtracting the standardized mother’s score from the standardized teen’s score. Positive SDS therefore represent dyads where teens’ standardized scores are higher than mothers’; negative SDS represent the reverse.

Standardized difference scores were utilized for analyzing informant discrepancies because they are equally correlated with and are statistically discernible from the reports of both informants (De Los Reyes and Kazdin 2004), properties which are critical when examining associations with informant discrepancies. Although multi-informant statistical interactions are recommended in place of SDS in variable-centered analyses (Laird and De Los Reyes 2013; Laird and LaFleur 2016; Laird and Weems 2011), interaction terms cannot practically take the place of difference scores when profiles of informant discrepancies across items or constructs within a family are to be modeled (i.e., in person-centered analyses). Further, because latent profile analyses use profile membership (a pattern of standardized difference scores) as the predictor or outcome of analyses, rather than the SDS themselves, these methods circumvent some of the statistical misspecification inherent in using difference scores directly in covariance-based analyses (Laird and De Los Reyes 2013; Laird and Weems 2011).

Next, exploratory latent profile analyses were conducted separately on mother–adolescent SDS for each of the three family constructs. LPA is a person-centered approach that identifies profiles of individuals (or in this case, dyads) that exhibit similar patterns of scores across continuous indicators and estimates the probability of each participant belonging to each profile. Indicators in the analyses were mother–adolescent SDS on issue types or subscales.

Profiles of informant discrepancies might alternatively be obtained by simultaneously modeling mother and adolescent reports as separate indicators, as has been suggested for regression analyses; Laird and Weems 2011. However, SDS were chosen as the indicators in these latent profile analyses for two reasons. First, as previously discussed, these scores inherently control for normative generational differences between parent and adolescent reports as part of the standardization process. Therefore latent profiles determined using standardized difference score indicators naturally separate potentially maladaptive informant differences from those that would be normatively expected; profiles based on raw scores do not have this ability. Second, prior latent profile analyses of informant discrepancies in the literature have utilized SDS (De Los Reyes et al. 2010, 2011); therefore, a similar procedure was followed here in order to more directly compare results across studies.

All person-centered analyses were conducted using MPlus 7.4 with FIML estimation of missing data and the MLR estimator, which is robust to nonnormality (Muthén and Muthén 1998–2012). The number of profiles were empirically determined based on fit indexes such as the Bayesian Information Criterion (BIC), for which lower scores represent better fit, and significant values of the Vuong Lo-Mendal Rubin LRT test or Bootstrapped LRT test, which evaluate whether a model with k profiles provides a significant improvement in fit over a model with k-1 profiles). Starting with a one-profile solution, increasing numbers of profiles were modeled until there was no substantive improvement in model fit (BIC scores reached their lowest value or were not substantively declining), LRT tests were non-significant, or additional profiles were small or conceptually unclear. Models with high entropy values and reasonable sample sizes in each profile were also given preference.

Once latent profiles for each family construct were identified, predictors of profile membership were examined for each construct by first regressing profile membership on adolescent age, gender, ethnicity (coded as white versus non-white) and maternal education (as a proxy for socioeconomic status). Significant demographic predictors were then included in logistic regressions predicting class membership from either mother or adolescent reports of adjustment (depression and problem behavior for adolescents, wellbeing for mothers) and parenting (behavioral and psychological control). Logistic regressions were conducted using a three-step procedure that determines and holds profile membership constant before introducing the predictor variables (Asparouhov and Muthén 2014). Maternal- and adolescent-reported predictors were examined in separate regression analyses to avoid methodological artifacts.

Finally, a latent transition analysis was conducted to examine the patterning of class membership within dyads across constructs. Latent transition analysis is an extension of LPA that calculates individuals’ patterns of transitions between latent profiles, providing a probability parameter of being in a specific latent profile in one analysis, given membership in a specific latent profile in a different LPA. Although each transition probability references only two latent profile analyses, these probabilities are combined to identify an individual’s most likely pattern of class membership across all latent profile analyses in the model. Latent transition analysis is traditionally used to model change in individuals’ profile membership over time within the same construct, but can also be used to model patterns of profile membership across constructs within a single time-point (Abar 2012). Chi square tests compared the percentage of dyads showing specific patterns of profile membership to the percentage expected to show that pattern based on profile frequencies.

Results

Descriptive Statistics and Preliminary Analyses

Means and standard deviations for mothers’ and adolescents’ ratings of maternal knowledge, parental right to know, and positive mother-teen interactions are presented in Table 1, as are cross-informant correlations and paired t-tests comparing mother and adolescent means. Consistent with past research (Smetana 2011), mothers reported higher mean levels of right to know and knowledge than did teens, but mothers did not report more positive interactions. Cross-informant correlations also were consistent with our expectations: mother–adolescent correlations were close to 0 for right to know, high for some issues but not others for maternal knowledge, and highest and consistently significant for reports of positive interactions.

Latent Profile Solutions

Latent profile analyses on mother–adolescent SDS for maternal right to know, maternal knowledge, and positive interactions each indicated that a three-profile solution fit the data well (see Table 2).

Table 2 Fit indexes for latent profile analyse
Full size table

Consistent with findings for informant discrepancies of parental monitoring (De Los Reyes et al. 2010), the three-profile solution for all three constructs consisted of the following (see Fig. 1). In the first profile, named Teen Over Mother, adolescents reported consistently higher standardized values than mothers (significantly positive standardized difference scores) across issues/subscales (20, 13, and 30 % of dyads for RTK, knowledge, and positive interactions, respectively). In the second profile, named Teen Under Mother, adolescents reported consistently lower standardized values than mothers (significantly negative standardized difference scores) across issues/subscales (28, 41, and 7 % of dyads for right to know, knowledge, and positive interactions, respectively). In the third profile, named No Disagreement, adolescents and mothers reported relatively equal standardized values (SDS not significantly different from zero) across issues/subscales (52, 46, and 63 % of dyads for right to know, knowledge, and positive interactions, respectively).

Fig. 1
figure 1

Profiles of mother–adolescent standardized difference scores (SDS) for maternal right to know, maternal knowledge, and mother–adolescent positive interactions. Number and proportion of dyads in each profile are based on the estimated model in Mplus; dyads are rounded to the nearest whole number

Full size image

Importantly, because these discrepancy scores reflect standardized values of mother and adolescent ratings, the resulting profiles represent differences between adolescents’ and mothers’ scores relative to the mean of their groups; they control for normative differences. That is, an adolescent from the Teen Over Mother profile for right to know may still believe his parents have less of a right to know about the teen’s activities than does his mother, but that difference is smaller than would be found among his peers (because his mother is lower than average on right to know beliefs and/or he is higher than average on right to know beliefs). Therefore, given mother and adolescent means on each construct (see Table 1), the following is true: The Teen Over Mother profile has the smallest raw difference between mothers and adolescents for the right to know and Knowledge constructs; the No Disagreement profile has the smallest raw difference between mothers and adolescents for reports of positive interactions.

Associations with Latent Profiles

Demographic Associations

Except for adolescent gender, no demographic differences emerged between profiles for any family construct. Dyads with female adolescents were more likely to belong to the No Disagreement profile than the Teen Over Mother profile for reports of parental right to know; they were less likely to belong to the No Disagreement profile than the Teen Over Mother profile for reports of positive interactions. Teen gender was therefore controlled in the subsequent regression analyses predicting discrepancy profiles for parental right to know and positive mother–adolescent interactions from family member adjustment and parenting. Tables 3 and 4 present the results for all final three-step logistic regressions.

Table 3 Results of three-step logistic regressions predicting latent profiles from adolescent reports of teen adjustment and parenting
Full size table
Table 4 Results of three-step logistic regressions predicting latent profiles from maternal reports of adjustment and parenting
Full size table

Parents’ Right to Know

Youth higher in problem behavior were most likely to belong to the Teen Under Mother profile, less likely to belong to the No Disagreement profile, and least likely to belong to the Teen Over Mother profile. Mothers reporting more behavioral control were more likely to belong to the Teen Under Mother profile than either of the other profiles.

Maternal Knowledge

Teens reporting more problem behavior were less likely to belong to the Teen Over Mother profile than either of the other two profiles. Adolescents reporting higher levels of maternal psychological control were more likely to belong to the Teen Under Mother profile than the Teen Over Mother profile. Mothers higher in wellbeing were more likely to belong to the Teen Under Mother profile than the No Disagreement profile.

Positive Mother–Adolescent Interactions

Adolescents reporting more maternal psychological control and less maternal behavioral control were more likely to be in the Teen Under Mother profile than the Teen Over Mother profile.

Latent Transition Analyses

The majority of dyads maintained consistent membership in a discrepancy profile across two or three of the family constructs (see Table 5). Only 10 % of dyads were classified into different discrepancy profiles for all three types of family variables. Of the remaining 90 % of dyads, 20 % had similar levels of standardized disagreement (were in the same profile) across all family variables. When only two discrepancy profiles were consistent (70 % of the sample), the greatest number shared a discrepancy profile between right to know and maternal knowledge (31 % of the sample), fewer shared a discrepancy profile between right to know and positive interactions (22 % of the sample), and the least shared a discrepancy profile between maternal knowledge and positive interactions (16 % of the sample). An omnibus Chi square analysis demonstrated that these distributions were significantly different than what would be expected based on the frequencies of profile membership for each construct individually, χ2(4) = 18.50, p < .01. Specifically, fewer dyads were classified into different profiles for all three constructs than would be expected, χ2(1) = 9.43, p < .01, and more dyads shared a discrepancy profile among right to know and knowledge than would be expected, χ2(1) = 10.80, p < .01.

Table 5 Number, observed proportion, and expected proportion of dyads demonstrating each profile pattern based on the estimated model
Full size table

Alternate Model Analyses

To check the robustness of our findings, we conducted alternate model analyses. Latent profile analyses were run using separate standardized mother and adolescent reports as the indicators (as opposed to standardized difference scores). The makeup of profiles appeared approximately the same as those obtained with the standardized difference score indicators (i.e., a Teen Over Mother, a Teen Under Mother, and a No Disagreement profile) for each family construct considered. However, the proportion of families assigned to each profile differed somewhat from analyses using difference scores; dyads were generally more equally distributed among profiles. Furthermore, fit indexes more often indicated support for a two-profile solution than a three-profile solution. Specifically, the Lo-Mendel Rubin Test showed a non-significant increase in fit between the two- and three-profile solutions for parental knowledge and parental support, despite lower BIC values for the three-profile solutions. These alternate analyses therefore indicated that the use of SDS over individual indicators did not result in fundamentally different profiles of informant discrepancies as long as the same number of profiles were selected, but they might have led to a different conclusion about the number of profiles present in the data more broadly.

Discussion

Different theoretical explanations have been offered for informant discrepancies in child adjustment. For instance, informant discrepancies may stem largely from contextual factors (De Los Reyes 2013), such as when different informants observe children’s behavior in different contexts. However, when studying family dynamics, the meaning of shared contexts is complex. Researchers have not adequately considered variations in the degree to which different contexts involve shared versus unshared experiences, knowledge, and beliefs. In distinguishing among different characteristics of such situations and comparing parent–adolescent discrepancies and their correlates across different family constructs, the present study made several important theoretical and methodological contributions to our understanding.

Theoretically, we extended prior research to elaborate on how differences in shared and unshared contexts are associated with discrepancies. Based on Jessop’s (1982) analysis of how family topics vary in their subjectivity, we distinguished among situations where information may not be directly observable but must be inferred (measured here in terms of right to know beliefs), contexts where parents may have differential access to and recollection of information (here, mothers’ knowledge of teens’ activities), and most objectively, entirely shared contexts (positive parent–child interactions). Bivariate correlations between mothers’ and teens’ ratings for the three constructs studied here confirmed our hypotheses regarding the relative objectivity of these reports: they were near zero for right to know beliefs and highest for positive interactions.

Another novel contribution of the present study was that, rather than focusing on a single area of parent–child discrepancies, we examined adolescents’ and mothers’ reports of three family constructs that varied in their theoretical level of objectivity and further, compared their links with parenting and adjustment. Methodologically, we contributed to the literature by using a person-centered approach to examine consistency in patterns of mother–adolescent discrepancies across constructs. This allowed us to avoid some of the statistical pitfalls commonly associated with the use of difference scores in regression analyses (Laird and De Los Reyes 2013; Laird and Weems 2011) while comparing discrepancies within families. Finally, we capitalized on a heretofore undiscussed property of SDS—their ability to differentiate individual differences in parent–adolescent discrepancies from those normatively expected due to generational stakes—to further explicate when parent–child informant discrepancies may and may not be adaptive for adolescent functioning (Ohannessian et al. 1995).

Profiles of Discrepancies

Separate latent profile analyses of discrepancies in adolescents’ and mothers’ ratings of right to know beliefs, maternal monitoring, and positive mother-teen interactions each yielded three profiles. Consistent with hypotheses, mother-teen dyads were most frequently identified as members of the No Disagreement profile for all three constructs, but slightly more often for positive interactions (63 %) than for mothers’ right to know (52 %) or maternal knowledge (46 %). These findings suggest that there is only moderate variation in mothers’ and teens’ ratings beyond normatively expected differences. They are also consistent with our description of positive interactions as the most objective of the three family constructs.

Membership in the Teen Under Mother profile, which reflected dyads where adolescents’ standardized scores were consistently lower than mothers’, varied according to the family construct assessed. A considerable proportion of adolescents viewed mothers as knowing less and having less of a right to know than did mothers (40.5 and 27.9 %, respectively), even controlling for normatively expected differences, whereas only a very small proportion of dyads rated their interactions with mothers less positively than did mothers (6.5 %). These findings are also consistent with the different explanations outlined previously for discrepancies in parents’ and adolescents’ perceptions of family relationships. Adolescents know what their mothers do not know, because to a large extent, they control access to the information. Mothers do have other sources of information about adolescents’ activities beyond teen disclosure or observation of their behavior, but most use these resources relatively infrequently (Waizenhofer et al. 2004). In contrast, only a small proportions of dyads were classified in the Teen Over Mother profiles. This is consistent with past research indicating that teens rarely view mothers as having more of a right to know, being more knowledgeable about their activities, and having more positive interactions than mothers rate themselves or their interactions (Steinberg 2001).

Correlates of Discrepancies

Consistent with hypotheses and De Los Reyes et al. (2008), mothers higher in psychological well-being were more likely to be in the Teen Under Mother profile for maternal knowledge and, marginally, for positive interactions. As psychological well-being has strong negative correlations with depressive affect (Ryff and Keyes 1995), these findings are consistent with evidence that maternal depression reduces the normative positivity bias mothers usually show in ratings of positive child interactions and family characteristics (Youngstrom et al. 1999). However, our results extend this finding to show that mothers who are psychologically healthier may display a protective bias mainly in perceptions of family behaviors involving a joint context (e.g., parent–child communication). Additionally, given the cross-sectional nature of the data, it is unclear whether greater psychological health leads mothers to over-perceive positive aspects of their interactions with adolescents (a true perceptual bias) or, given mothers’ greater stake in maintaining family harmony (Bengtson and Kuypers 1971), an overly positive perception of family dynamics contributes to feeling psychologically healthier for mothers. Indeed, the second option may be more likely, as positive relations with others is a component of the psychological health measure utilized in this study.

Although a significant proportion of our sample (39 %) was above the clinical threshold for mild depressive symptoms, adolescent depressive symptoms were not associated with profile membership for any of the three family constructs. This result was unexpected but supports past research showing that adjustment-based perceptual biases are not a consistent predictor of parent–adolescent discrepancies (De Los Reyes 2013). Inconsistencies between our and De Los Reyes’ et al. (2008) results may be due to the use of person-centered analyses, the older age of adolescent participants, or because De Los Reyes examined unique associations of adolescent depression (controlling for mothers’ symptoms), whereas we did not. In support of the current results, adolescents tend to be less bothered than mothers by parent–adolescent conflict (Steinberg 2001), resulting in potentially more significant associations between parent–adolescent discrepancies and mothers’ than teens’ psychological health.

As hypothesized, adolescent problem behavior was associated with membership in profiles where adolescents reported lower maternal right to know and maternal knowledge relative to mothers, but not with profile membership for positive interactions. This is consistent with research showing that youth engaging in problem behaviors are more likely to disregard parental authority and withhold information from parents (Laird et al. 2010), resulting in greater mother–adolescent discrepancies about family constructs involving unshared contexts. Additionally, youth problem behavior was least frequent among youth in the Teen Over Mother profile for maternal right to know and knowledge. As this profile demonstrates the smallest raw (unstandardized) discrepancy in mother and adolescent ratings for these variables, these findings imply that greater parent–adolescent agreement remains most adaptive even when discrepancy is a normative aspect of development. Such findings are in line with evidence that other normative behaviors in adolescence, such as an increase in delinquent behavior, are nonetheless not always the most adaptive developmental pathway (Moffitt et al. 2002).

Dyads with adolescents reporting greater maternal psychological control were more likely to belong to the Teen Under Mother profile for maternal knowledge and positive interactions, but maternal reports of psychological control were not associated with profile membership. This is consistent with findings that adolescent reports of parenting are often better predictors of adolescent outcomes than are parent reports (Cottrell et al. 2003).

Although not hypothesized, the finding that adolescents reporting more maternal behavioral control were less likely to belong to the Teen Under Mother profile for reports of positive interactions with mothers is consistent with evidence that moderate levels of behavioral control are beneficial for family relationships and adolescent adjustment (Steinberg 2001). In contrast, dyads with mothers reporting higher levels of behavioral control were more likely to belong to the Teen Under Mother profile for right to know beliefs. Behavioral control (particularly the aspects mothers focus on) may therefore be especially associated with mothers’ (but not teens’) stronger beliefs about their own legitimate authority (i.e., their right to know). The lack of association between behavioral control and discrepancies about maternal knowledge was counter to hypotheses, but it was consistent with past research demonstrating that parental rules and solicitation have a complex relation with adolescent disclosure (Keijsers and Laird 2014).

Consistency of Associations Across Constructs

Overall, predictors of profile membership were only moderately consistent across family constructs. As expected, adolescent problem behavior was primarily associated with profile membership for constructs involving unshared contexts (right to know and maternal knowledge), whereas maternal adjustment was more strongly associated with profile membership for constructs involving shared contexts (maternal knowledge and positive interactions). Surprisingly, however, adolescent depression did not follow this pattern. Rather, teen reports of both psychological and behavioral control were associated with discrepant perceptions in shared but not unshared contexts. Further, mother-reported behavioral control was associated with discrepancies only for beliefs about parental right to know. These findings highlight the importance of taking into account the subject matter of the discrepancy and source of information, even when considering associations with informant discrepancies themselves.

We expected and found considerable consistency in mother–adolescent dyads’ membership in discrepancy profiles across the family constructs studied here. Most dyads were in the same discrepancy profile across at least two of the family constructs studied here, and fewer dyads than expected were classified into different profiles for all three constructs. The greatest match in profile membership was in dyads’ right to know beliefs and maternal knowledge. Although this may be because both right to know beliefs and maternal knowledge include unshared contexts, it also may be that disclosure, and in turn, parental knowledge, are both partly based on teens’ beliefs about parental authority legitimacy (Chan et al. 2015; Rote and Smetana 2015).

Limitations and Future Directions

Although the sample was relatively small for the analyses conducted (Nylund et al. 2007), the latent profile analyses did converge into stable and replicable solutions that mirror those obtained for similar constructs in other research (De Los Reyes et al. 2010). The sample was also relatively homogeneous and only included mothers; future research should examine whether similar discrepancy profiles obtain across constructs for minority adolescents, children of varying ages, and father–adolescent dyads. Furthermore, the present analyses were cross-sectional, so the stability of the profiles over time or causal relationships between the correlates and the profiles could not be determined. Further research would be needed to examine these questions and might focus on a wider variety of constructs (potentially including discrepancies in reports on family members’ adjustment as well) and emphasize individual factors hypothesized to lead to differences in profile membership among family constructs. Finally, other contextual effects—specifically the extent to which parents and adolescents recall different settings even when asked to report on shared interactions (De Los Reyes et al. 2013)—were not considered here. Such recall biases likely contribute further to discrepancy profile variations within dyads and may vary across family constructs. Therefore they should be incorporated into future research.

Conclusion

The present study employed theoretically grounded analyses using novel methods to model discrepancies in adolescents’ and mothers’ perceptions of the family. The results of this research contribute to our understanding of the role of informant discrepancies in family functioning and further our knowledge of the specific constructs studied here. The analyses demonstrated that there is moderate consistency within mother–adolescent dyads in discrepant perceptions of different family constructs (more than would be expected by chance), but also that individual and familial factors are differentially associated with discrepancies for different family variables. As has been the case in the majority of research on informant discrepancies (De Los Reyes 2013), as well as in research on parent–adolescent conflict (Steinberg 2001), families evidencing greater disagreement (especially beyond that which is normative) showed poorer individual and dyadic functioning regardless of the construct considered. However, discrepancies in family constructs involving less of a shared context (parental right to know, maternal knowledge) appeared more related to adolescent behavior problems, whereas discrepancies in family constructs involving more shared interaction (maternal knowledge, positive interactions) were mainly associated with parenting practices. The present study extends our theorizing and research regarding different sources of parent–adolescent discrepancies in family functioning and provides preliminary evidence that greater mother–adolescent agreement may be adaptive even when it is not normative. Given the importance of parent–adolescent disagreement and (moderate) conflict for autonomy development however (Holmbeck and O’Donnell 1991; Smetana 2011), further research is needed to better understand the implications of the findings (and other sources of discrepancies) for adjustment and adolescent development.