Latent Profiles of Parental Academic Conditional Positive and Negative Regard

Parental Conditional Regard in the Academic Domain

Academic achievement is a core area for parent–child disagreements (Smetana, Daddis, & Chuang, Citation2016). Parents’ involvement in children’s school work plays an important role in families’ daily lives and has an impact on children’s learning motivation (Barger, Kim, Kuncel, & Pomerantz, Citation2019; Grolnick & Slowiaczek, Citation1994; Pomerantz, Moorman, Litwack, & Litwack, Citation2007). An autonomy-supportive parental style of school involvement, where parents display a sincere interest in the child’s perspective on school work and offer choices on how to deal with school work, benefits children’s motivation and learning (Froiland, Citation2011; Katz, Kaplan, & Buzukashvily, Citation2011; Li, Yao, Liu, & Chen, Citation2019; Viljaranta et al., Citation2018). In contrast, a more controlling parental type of involvement, where parents pressure their children to meet prescribed standards, diminishes motivation and hampers learning (Silinskas & Kikas, Citation2019).

One specific form of controlling parental involvement is parental conditional regard, which means that parents let their love and appreciation depend on the child’s achievement of particular standards for performance (Assor et al., Citation2004). Conditional regard can take two forms (Assor & Roth, Citation2005; Roth et al., Citation2009). The most widely studied form of conditional regard is parental academic conditional negative regard (PACNR). This parenting practice includes withdrawing appreciation, esteem, and interest in the face of failure in school. In contrast, parental academic conditional positive regard (PACPR) is a seemingly more benign form of control. It includes showing more appreciation, esteem, and interest in the face of success in academics.

From a behaviorist perspective, conditional regard may be seen as an effective socialization strategy that increases the likelihood of the respective behavior (Aronfreed, Citation1968; Gewirtz & Peláez-Nogueras, Citation1991). In contrast, from a self-determination theory (SDT) perspective, both PACNR and PACPR are considered autonomy-suppressing and need thwarting strategies. When parents withdraw their love in the face of failure (as in PACNR), the sense of secure attachment to the caregiver and thus the need for relatedness (i.e., the need to feel cared for and to belong) is directly threatened. Furthermore, the child receives direct negative feedback on his or her behavior or performance, which is likely to threaten the child’s need for competence, that is the need to experience a sense of mastery and effectiveness. The need thwarting effect of PACPR is less straightforward because this strategy only includes a surplus in appreciation and caring, which, at first sight, appears to support the need for relatedness. However, conditional positive regard makes children feel compelled to engage in an activity to gain more parental approval. As such, both in the case of conditional negative and positive regard, children feel forced to choose between seeking parental approval and staying true to their personal preferences and interests, thereby creating a tension between the need for relatedness and the need for autonomy, that is the need to feel free in one’s own actions (Assor et al., Citation2004; Cohen, Moed, Shoshani, Roth, & Kanat-Maymon, Citation2020; Ryan & Deci, Citation2017). In turn, this tension between needs is considered harmful to children’s development.

Importance of a Person-Oriented Approach

To the best of our knowledge, all current studies of conditional regard have adopted a variable-oriented approach. When controlling for the respective other strategy, PACNR and PACPR differ somewhat in their psychological mechanisms and their impact on psychological functioning. However, the two strategies are conceptually closely linked and share about 22–28% common variance (.45 < r < .53; Assor & Tal, Citation2012; Roth et al., Citation2009). The overlap between the two dimensions raises issues that cannot be addressed using a variable-oriented approach but require a person-oriented approach to investigate within-person combinations of PACNR and PACPR. Specifically, a person-oriented approach is useful from both a substantive and a statistical point of view.

Substantively, a person-oriented approach allows one to test the assumption that the two dimensions of conditional regard are truly distinct. By identifying profiles of individuals, it is indeed possible to examine whether at least some of these profiles reflect a unique role for either PACNR or PACPR. If a person-oriented approach resulted in only two profiles, with some individuals scoring high and others scoring low on the two dimensions of conditional regard, then both dimensions would always co-occur to a certain extent at the individual level. In this case, PACNR and PACPR would reflect two sides of the same coin, with parents showing their love either unconditionally or conditionally. The distinction between the two dimensions at the variable level would then have little real-life value for individuals. If, instead, at least one profile would emerge with uniquely elevated scores on only one dimension (e.g., PACPR but not PACNR), the distinction between the two dimensions would be more meaningful. In this case, PACNR and PACPR would represent specific and distinct socialization strategies. Theoretically, four different profiles are possible, which reflect each combination of the two dimensions (high on both, low on both, high on PACPR but low on PACNR, and vice versa). A person-oriented approach allows for an examination of the question of whether each of the four profiles exists and whether profiles relate differently to an individual’s motivation and adjustment.

Statistically, one may wonder whether it is appropriate to simply ignore the variance shared between PACNR and PACPR, as is typically done in the variable-centered approach. When both dimensions of conditional regard are entered simultaneously in a variable-oriented approach, the variance shared between the two dimensions is set aside and only the unique parts of the variance are considered in the analysis. However, excluding the shared variance from the analysis may change the meaning of the constructs examined. If PACPR is stripped from the variance it has in common with PACNR, does it then still reflect a conditionally approving parental orientation? Accordingly, unique scores for PACPR (stripped from PACNR) may result in an underestimation of the detrimental consequences associated with this form of conditional regard. A person-oriented approach avoids this problem because it does not pit the two dimensions against each other; instead, it identifies within-person combinations of the dimensions so that isolated or interactional effects at the person level can be explored.

As such, a person-oriented approach may allow for a more accurate test of the unique roles of PACNR and PACPR in child development. This is particularly important for PACPR because this dimension of conditional regard is seemingly beneficial and is considered effective in promoting desired behavior from a behaviorist point of view (Aronfreed, Citation1968; Frost, Citation2005; Gewirtz & Peláez-Nogueras, Citation1991; McGraw, Citation2004). If PACPR is itself harmless or beneficial, individuals who only report elevated scores on PACPR (but not on PACNR) should be comparable to (or perhaps do even better than) individuals low in the two dimensions of conditional regard. If, instead, PACPR is detrimental to development (as assumed in SDT), the psychological adjustment of individuals uniquely high on PACPR should at least be in between the adjustment of individuals high on both dimensions and individuals experiencing no conditional regard at all.

Research Questions

The first goal of this study was to identify within-person combinations of PACNR and PACPR using a person-oriented approach. The person-oriented approach allows the investigation of the nature of PACNR and PACPR profiles and sheds light on their prevalence. Second, we aimed to investigate profile functionality. A comparison of differences between profiles in terms of various outcomes allows us to extend conclusions on the distinct functionality of the two parenting behaviors and their isolated or interactional effects at the person level. SDT suggests that both PACNR and PACPR are uniquely detrimental to development. We expected that individuals who score high on the two dimensions of conditional regard, should display the most unfavorable outcomes. In contrast, individuals who score low on both dimensions should display more favorable outcomes. We expected that if we can identify groups with elevated scores for either PACNR or PACPR, these individuals should score in between the groups scoring low and high on both dimensions. More specifically, individuals with elevated scores only in PACPR should show lower adjustment in developmental outcomes that were found to be associated with PACPR in variable-oriented studies, likewise for PACNR.

To investigate our research questions, we collected data from three of the most widely studied populations in academics: adolescent students (Study 1), university students (Study 2), and parents of school-aged children (Study 3). This approach allowed us to explore profiles in different populations and investigate the profile functionality in a very broad manner. Data collection took place between 2016 and 2018. Ethical guidelines of the American Psychological Association (APA) and the German Association of Psychologists (DGPs) were followed in the study.

Participants and Procedure

The participants were 3,891 adolescent students aged 9 to 19 years with a mean age of 13.13 years (SD = 1.73). Of the adolescents, 49.96% were women. Schools from seven different federal states in Germany were contacted via principals, 22 of whom agreed to participate. We invited all students from fifth to tenth grades to participate in the study. The parents provided active informed consent for their children. Participation was voluntary, and participants did not receive payment. The response rate was 69.5%. The sample included all different tracks within the secondary school system: low track schools (Sekundarschule, 21.3% of the sample), middle track schools (Realschule, 19.7% of the sample), and high track schools (Gymnasium, 58.9% of the sample). In all, 16.2% of the adolescents were in the fifth grade, 16.1% in the sixth, 21.8% in the seventh, 21.3% in the eighth, 12.6% in the ninth, and 12.0% in the tenth grade. Trained student workers administered questionnaires in class. We used pseudonymization to match the data of the adolescent students with data from their parents who were asked to participate in an additional survey (see Study 3).

Measures

To explore a broad range of different outcomes while not overburdening students with an excessively long survey, subsamples provided ratings of different combinations of outcome variables. Hence, sample sizes vary for the different outcome variables (for the exact composition of the total sample and sizes of subsamples, see Supplementary Table S1).

Confirmatory Latent Profile Analysis

We conducted confirmatory latent profile analyses (CLPA) using Mplus 8 (Muthen & Muthén, Citation1998–Citation2018) to examine whether groups of adolescent students with high PACPR/high PACNR, low PACPR/low PACNR, high PACPR/low PACNR, and low PACPR/high PACNR could be identified in the present sample. Intraclass correlations of all study variables indicated that only very small amounts of variance (1.6–6.8%) could be attributed to the classroom level. Lee (Citation2000) proposed that an amount of variance less than 10% can be considered negligible. Because most of the intraclass correlations in Study 1 fell below .05, the nested structure of the data was not considered in the analysis. In most applications of LPA, researchers do not attempt to explicitly test theories on underlying groups in their substantive area, but rather allow the data to suggest the number and nature of such groups (exploratory LPA). In Study 1, however, we draw on substantive theoretical ideas regarding the characteristics of latent groups with different PACNR/PACPR profiles. Akin to confirmatory factor analysis, such conceptual ideas can be explicitly tested by expressing them through a set of parameter restrictions (Finch & Bronk, Citation2011). Using the confirmatory approach permitted a direct examination of the four qualitatively different groups, whereas an exploratory approach may also include profiles that differ only in quantitative terms (e.g., medium PACPR/medium PACNR). Three types of parameter constraints can be used in CLPA modeling: (1) equality restrictions, (2) deterministic restrictions, and (3) inequality restrictions. Here, to explore how many hypothesized subgroups could be identified in the present sample, we used a set of equality and inequality constraints that expressed the presumed ordering of the PACPR and PACNR means in the respective latent classes.

Overall, we specified and compared the fit of four different CLPA solutions, which represent two, three, and four latent classes, respectively. In the two-class model, we used constraints to specify the subgroups with high PACPR/high PACNR and low PACPR/low PACNR. We tested two three-class models and added constraints to build a third latent class with a high PACPR/low PACNR profile and vice versa. In the four-class model, further constraints allowed us to specify subgroups with all four possible combinations of PACPR and PACNR (see Supplementary Material for the Mplus syntax for all CLPA models).

The classification of latent classes was based on the observable response patterns of individuals across the two conditional regard scales (Marsh, Lüdtke, Trautwein, & Morin, Citation2009; Nylund, Asparouhov, & Muthén, Citation2007). The affiliation of each person to the respective latent class and the sizes of the respective classes were calculated as model parameters. In the model specification, the default settings recommended in Mplus were used so that (1) the mean values and variances of the class indicators were freely estimated, (2) the variances of the class indicators between the classes were kept constant, and (3) the covariances between the indicators within the classes were fixed at 0. The different CLPA solutions were compared in terms of relative statistical fit (Nylund et al., Citation2007) and the interpretability of the profile structure (Marsh et al., Citation2009). Based on recommendations by Nylund et al. (Citation2007), we preferred solutions with the lowest Bayesian information criterion (BIC) and sample-size adjusted Bayesian information criterion (SABIC) values, and highest log-likelihood values (LL). Nylund et al. (Citation2007) also recommend the bootstrapped likelihood ratio test (BLRT). However, BLRT requires one latent class without parameter restrictions (Asparouhov & Muthén, Citation2012), which is why we could not use this indicator in Study 1. With respect to our specific research question regarding the prevalence of various PACPR/PACNR combinations, we also focused on the size of the latent classes. Next, classification accuracy was checked. An entropy near 1 indicates a statistically reliable solution (Marsh et al., Citation2009). Finally, we checked the averaged probabilities throughout the participants for the most likely class membership in each latent class. High probabilities that participants belonging to a given class are correctly categorized indicate a high reliability of classification.

In the next step, the identified profiles were related to different correlates. There are different ways on how the effects of latent classes on such “distal outcomes” can be estimated validly. The class-specific mean values of a distal outcome variable can be determined using either a one-step or three-step approach. With the one-step approach, the distal outcome is incorporated into the LPA model as an additional indicator and then estimated as usual. However, the main disadvantage of this approach is that fairly strong assumptions must be made regarding the distribution of the distal outcome within the latent classes. Violating these assumptions can distort the original LPA model completely. Because of these and similar problems, researchers often prefer the three-step approach, where they first build the LPA model without the distal outcomes, determine the class membership probabilities, and then examine relations between class membership and distal outcomes. A known drawback of this approach is that the parameter estimates obtained in the third step may be incorrect because of the measurement error when assigning individuals to the latent classes in step 1. Several solutions have been proposed to reduce this potential bias, including an optimized version of the so-called “class bias.” The BCH method (Bakk & Vermunt, Citation2016; named after Bolck, Croon, & Hagenaars, Citation2004), in which an ANOVA weighted by the inverse classification error probabilities is calculated, has proven to be the most valid estimation method to date (Bakk & Vermunt, Citation2016; Dziak, Bray, Zhang, Zhang, & Lanza, Citation2016). In Study 1, we therefore relied on the BCH method so that the respective correlates were included in the CLPA models as “auxiliary variables” (Marsh et al., Citation2009; see Supplementary Material for the Mplus syntax).

We conducted the latent profile analysis and BCH method using the total sample to ensure that the allocation of individuals to the different profiles was based on all available information. However, this approach does not allow us to estimate on how many participants the estimations for distal outcomes in each profile rely. Therefore, we replicated the latent profile solution for each subsample, giving us the opportunity, first, to conduct an internal replication of our profile solution and, second, to estimate the number of participants in each profile that reported on the respective outcomes.

Number of Latent Profiles

Table 1 displays the fit indices and latent class frequencies for the different CLPA solutions, and Table 2 shows the class-specific means and standard errors (see Supplementary Table S2 for descriptive statistics of all study variables). Adding the high PACPR/low PACNR group as a third class increased the log-likelihood value and decreased BIC and SABIC values. Adding the low PACPR/high PACNR group, however, increased BIC and SABIC. The fit improved slightly for the four-class solution compared to the three-class solution including a high PACPR/low PACNR. However, the subgroup of adolescent students that was assigned to the fourth latent class representing the low PACPR/high PACNR profile was very small (1.4%). On advice of the reviewers, we conducted an additional series of exploratory latent profile analyses (see Supplementary Table S3). The results supported low PACPR/low PACNR, high PACPR/high PACNR, and high PACPR/low PACNR groups. Several profiles emerged that differed in quantitative (but not qualitative ways) along the continuum between low PACPR/low PACNR and high PACPR/high PACNR. However, no low PACPR/high PACNR group emerged, with this finding further informing the decision to reject the four-class solution. Therefore, we chose the three-class solution including a subgroup of students high in PACPR and low in PACNR as the final solution. The classification probabilities that participants were categorized correctly were higher than .85, which indicated good reliability of the classification. Moreover, these results were replicated in all subsamples (see Supplementary Table S4).

The largest group of adolescent students (58.1%) was assigned to the low PACPR/low PACNR profile (lowPACR), followed by the high PACPR/low PACNR profile (onlyPACPR), in which 27.3% of the adolescents were classified. In total, 14.5% of the adolescents showed a high PACPR/high PACNR profile (highPACR, see Table 1). Latent class frequencies for the three classes were comparable throughout the subsamples (see Supplementary Table S4). An inspection of the class-specific means of PACPR and PACNR revealed that the parameter constraints implemented in the CLPA models successfully shaped the theoretically expected profiles (see Table 2). The four-class solution yielded a slightly better fit. Therefore, we provide results for this solution in the supplementary for interested readers (see Supplementary Table S5).

Latent Profile Outcomes

Using the BCH approach, we examined whether adolescent students in the three latent profiles differed in the outcome variables. The results revealed a rather clear and differentiated pattern for the three latent classes (see Table 2). Adolescent students with a lowPACR profile showed the most adaptive configuration of outcomes, that is the relatively highest values in basic need satisfaction, self-esteem level, ability self-concept, and mastery goals as well as the lowest values in self-esteem contingency, performance-approach goals, performance-avoidance goals, test anxiety, and depressive symptoms. Adolescent students in the onlyPACPR profile differed from the lowPACR group in reporting a higher satisfaction of the need for competence, higher self-esteem contingency, and higher scores on performance-approach and performance-avoidance goals. The subgroup of students in the highPACR profile (high PACPR/high PACNR) revealed a remarkably maladaptive outcome pattern compared to the lowPACR and onlyPACPR groups. Specifically, these adolescent students scored relatively lowest on basic need satisfaction, self-esteem level and ability self-concept, highest on self-esteem contingency and depressive symptoms, and lowest on mastery goals but highest on performance-approach and performance-avoidance goals.

Participants and Procedure

Participants were 556 university students (74.3% women; M_age = 22.14 years, SD_age = 4.44), who studied teaching. Participants were recruited via e-mail (n = 180), seminars (n = 135), or lectures (n = 241) on the introduction to psychology for teaching students. Only students from the lecture sample reported the outcome variables.

Measures

For each measure (except for contingent self-esteem), we used the same items as in Study 1 but adjusted to the university context.

Number of Latent Profiles

The statistical plan of analysis was identical to that of Study 1. The fit indices and latent class frequencies for the different CLPA solutions are displayed in Table 1, and the class-specific means and standard errors are displayed in Table 3 (see Supplementary Table S6 for descriptive statistics of all study variables). Adding the high PACPR/low PACNR group increased the log-likelihood value, BIC, and SABIC values decreased. BIC and SABIC increased when the onlyPACNR group was added as the third group. Compared to the three-class model, including high PACPR/low PACNR, decreases in BIC and SABIC values were only small when the fourth group was added. Additionally, the class frequency in the fourth class was very small (0.7%). In accordance with Study 1, these results supported the three-class solution with the onlyPACPR group as the third group. The classification probabilities that participants were categorized correctly were higher than .91, which indicated excellent reliability of the classification.

The largest group of university students (44.4%) was allocated to the low PACPR/low PACNR profile (lowPACR, see Table 1). In total, 31.3% of the university students were classified into the high PACPR/low PACNR profile (onlyPACPR), followed by the high PACPR/high PACNR profile (highPACR), in which 24.3% of the university students were classified. An inspection of the class-specific means of PACPR and PACNR revealed that the parameter constraints implemented in the CLPA models successfully shaped the theoretically expected profiles (see Table 3).

Latent Profile Outcomes

As in Study 1, we used the BCH approach to examine the differences between profiles in several outcome variables. The results show that the lowPACR group displayed a more favorable pattern of outcomes than the highPACR group by reporting lower contingent self-esteem and lower performance-approach and performance-avoidance goal orientation. The onlyPACPR group was in between the highPACR and lowPACR groups concerning contingent self-esteem but did not differ significantly from the other groups. The onlyPACPR group differed from the highPACR group only by showing less performance-avoidance goal orientation. The results revealed no significant differences between the groups for self-esteem level, ability self-concept, and mastery goals.

Participants and Procedure

The adult sample consisted of parents from children who were included in Study 1. Seven (31.8%) schools from three different federal states participated in the parent survey. Schools were chosen for convenience and by considering all the different school tracks. Children from these schools received envelopes with questionnaires for their parents. Parents were invited to complete the survey at home and to return the sealed envelopes to the school via their children. A total of 760 parents participated, and the response rate was 65.7%.

Participants’ ages ranged between 28 and 60 years, with a mean age of 45.0 years (SD = 5.06), and 88.7% were women. This high participation of mothers corresponds to the finding of Study 1, where 92.3% of the adolescents indicated their mother as the main caregiver. For education, 0.8% had no school leaving qualification, 5.2% attended the lowest track of school, 24.0% attended the middle track, and 57.4% attended the highest track. A comparison with data from the Federal Statistical Office of Germany, which indicated that from individuals aged between 25 and 60, 40.2% graduated from the highest track and 21.7% from the lowest (Statistisches Bundesamt [Destatis], Citation2020), revealed that mothers who graduated from the highest track were overrepresented, and mothers who graduated from the lowest track were underrepresented in the present sample.

Measures

The parent dataset consisted of two subsamples that provided reports on self-esteem facets and depressive symptoms, or parenting sense of competence.

Number of Latent Profiles

We conducted the same set of CLPA as in Studies 1 and 2. The fit indices and latent class frequencies for the different CLPA solutions are shown in Table 1, and the class-specific means and standard errors are shown in Table 4 (see Supplementary Table S7 for descriptive statistics of all study variables). In accordance with the results of Studies 1 and 2, our analyses supported the three-class solution with high PACPR/low PACNR as the third group. Including low PACPR/high PACNR as the third group diminished the fit. Again, the four-class solution only showed a slightly better fit compared to the three-class solution including high PACPR/low PACNR and class frequency in the low PACPR/high PACNR class was very small (1.2%).

A very large group of parents (61.4%) was allocated to the low PACPR/low PACNR profile (lowPACR; see Table 1). In contrast to Studies 1 and 2, the second largest group was the high PACPR/high PACNR profile (highPACR), in which 24.5% of the parents were classified. Fewer parents (14.1%) were allocated to the high PACPR/low PACNR profile (onlyPACPR). The classification probabilities were higher than .86, which indicated good classification reliability. An inspection of the class-specific means of PACPR and PACNR revealed that the parameter constraints implemented in the CLPA models successfully shaped the theoretically expected profiles (see Table 4).

Latent Profile Outcomes

The highPACR group scored less favorably than the lowPACR group on all outcomes by showing lower self-esteem level, higher self-esteem contingency, lower parenting sense of competence, and higher depressive symptoms. The onlyPACPR group did not differ from the highPACR group on self-esteem level and contingency, thereby scoring less favorably than the lowPACR group. The onlyPACPR group did not differ from the lowPACR group on parenting sense of competence and depressive symptoms.

Functionality of Parental Conditional Regard Profiles

The results of all three studies are in line with the assumption that the experience of PACNR is accompanied by clear dysfunctional correlates. As proposed in self-determination and attachment theory, love withdrawal is detrimental to development because it constitutes a direct threat to basic psychological needs and attachment (Bowlby, Citation1969; Deci & Ryan, Citation2000). In this respect, all three studies support that individuals reporting high PACNR show unfavorable developmental outcomes and replicate findings from variable-oriented studies for various outcomes (e.g., Assor & Tal, Citation2012; Roth et al., Citation2009; Wouters et al., Citation2018).

The results for individuals who only experience high PACPR are of particular interest because the functionality of this practice is judged differently by behaviorist theories and SDT (Gewirtz & Peláez-Nogueras, Citation1991; Ryan & Deci, Citation2017). In general, individuals who experience only PACPR report a less maladaptive pattern of outcomes than individuals who additionally perceive PACNR. Nevertheless, individuals in the onlyPACPR group fare worse than individuals experiencing generally low levels of conditional regard. Adolescents in the onlyPACPR profile can be described as highly involved and motivated (high mastery orientation, high performance-approach, and high performance-avoidance orientation). At the same time, they do not report general negative consequences for psychological health (no lowered ability self-concept and no heightened test anxiety or depressive symptoms); however, they display a brittle sense of worth. This insecure type of self-esteem was also observed in university students, where the onlyPACPR group scored in between the highPACR and lowPACR groups in contingent self-esteem. The onlyPACPR parent group showed equally high levels of contingent self-esteem as the highPACR group. Even in the absence of conditional negative regard, conditional positive regard shows, if any, associations with detrimental outcomes. This finding is noteworthy considering that more than one in four adolescents and university students and 14.1% of the parents belong to the onlyPACPR group.

Although this fragile and conditional type of self-esteem may have some positive short-term effects on engagement, contingent self-esteem may increase the risk of maladjustment over time, particularly in interaction with negative life events. Indeed, people who experience only PACPR and display contingent self-esteem may adjust well as long as they perform well and receive positive feedback because basic needs can at least be superficially satisfied. However, they may become vulnerable to ill-being and poor adjustment when facing setbacks (e.g., poor grades) or negative feedback. We assume that in these situations, individuals will not be flexible in their striving to obtain approval but are “overcommitted” to their goal of academic achievement and, therefore, have problems deactivating their goal intentions (Hallsten, Josephson, & Torgén, Citation2005; Kudielka, Känel, Gander, & Fischer, Citation2004). Therefore, their striving may become excessive, potentially resulting in overinvestment and difficulties relaxing and calming down, while the expected approval remains missing (Hallsten et al., Citation2005). Individuals may also prefer behavior in the relatedness-relevant domain to behavior in domains of personal interest, thereby hindering the individual’s growth tendencies that are inherent in intrinsic activity (Ryan & Deci, Citation2017). We propose that individuals who only experience PACPR may be prone to workaholism (working compulsively and excessively), and burnout symptoms, such as emotional exhaustion, are more likely to occur because of the effort–reward imbalance (Siegrist et al., Citation2004; Van Wijhe, Peeters, & Schaufeli, Citation2014). Future studies should examine the interplay between PACPR, contingent self-esteem, and negative life events or hassles in predicting adjustment and achievement in the academic context and across the life span.

In addition to the possibility that PACPR increases the risk of maladjustment when facing setbacks, one may wonder about the quality of individual engagement when they experience conditional positive regard and have contingent self-esteem. Parallel to findings on contingent self-esteem and in line with results from Study 1, individuals who only experience PACPR are likely to show higher autonomous and controlled motivation at the same time (Van der Kaap-Deeder et al., Citation2016). Individuals high in contingent self-esteem may enjoy learning until an evaluation is anticipated because evaluation may pose a potential threat. Consequently, they switch to a more controlled form of motivation regulation. In comparison to adolescent students who are mainly mastery-oriented, adolescent students who additionally endorse performance goals report similar vigor, dedication, and absorption, but at the same time report more exhaustion and cynicism during schoolwork (Tuominen-Soini, Salmela-Aro, & Niemivirta, Citation2012). Thus, the motivational constellation found for individuals who only experience PACPR is likely associated with lower learning quality (Spinath & Stiensmeier-Pelster, Citation2003), lower affective functioning in learning (Tuominen-Soini et al., Citation2012; Van der Kaap-Deeder et al., Citation2016), and constant concerns on performance outcomes (Tuominen-Soini et al., Citation2012).

Strengths, Limitations, and Future Directions

Overall, this study has several strengths. It includes three large samples from different populations, which allowed us to replicate findings and investigate functionality in a very broad manner. Using a person-oriented approach enabled us to provide evidence for the real-life value of the distinction between conditional negative and positive regard and their differential associations with individuals’ general and educational adjustment. These findings have potential applied value because the identification of conditional regard profiles in school counseling may help to identify the antecedents of academic striving, lower learning quality, or emotional disaffection.

This study has several limitations as well. First, the cross-sectional and single-informant design restrict conclusions and did not allow us to disentangle possible causal mechanisms of action, which is important because the relation between parenting and children’s adjustment is inherently reciprocal (Otterpohl & Wild, Citation2015; Soenens, Luyckx, Vansteenkiste, Duriez, & Goossens, Citation2008). In addition, the use of self-report measures may have led to an overestimation of associations between conditional regard and outcomes. For example, participants with higher contingent self-esteem may project a conditionally approving attitude toward their parents, even when their parents do not actually engage in conditionally regarding practices. It is also possible that individuals who feel close and loyal toward their parents respond in more socially desirable ways to questions about parenting, particularly in the case of the obviously unfavorable form of conditional negative regard. This social desirability may have led to an overestimation of the onlyPACPR group in size and an underestimation of the prevalence of the highPACR and even the onlyPACNR profiles. In general, longitudinal research is needed to test profile stability and cross-lagged effects. By using multiple informants for parenting measures, future studies could strengthen findings regarding the reliability of profiles and profile functionality.

We deliberately chose to identify profiles only on the basis of PACNR and PACPR, without including other parenting dimensions. This analysis focusing only on the two dimensions of conditional regard is an important first step to determine whether these dimensions are distinct. Having established distinct profiles of conditional regard, future research can now add other parenting dimensions to the analysis of these profiles, such as parental warmth, autonomy support, and structure. Although conditional regard is considered an autonomy-suppressing parental strategy, it is possible that some parents combine the use of conditional regard with episodes of autonomy-support, resulting in a profile combining high conditional regard with autonomy-support. Perhaps this combination occurs more often with PACPR than with PACNR because in PACPR important components of intrusiveness are not evident, which makes it appear less harmful. Parents may combine an autonomy-supportive parenting style with occasional praise that conveys the message of PACPR. Similarly, some parents may combine high levels of conditional regard with parental warmth. Studies taking a variable-oriented approach have shown that this combination can be detrimental to children’s development (e.g., Aunola & Nurmi, Citation2005; Kanat-Maymon & Assor, Citation2010; Wouters, Doumen, Germeijs, Colpin, & Verschueren, Citation2013). Rather than buffering against the maladaptive effects of controlling parenting, high levels of parental warmth have been found to exacerbate effects of parental control, presumably because children get caught in a loyalty conflict with parents who are highly involved and warm yet pressuring. Whereas some parents may combine high levels of conditional regard with high levels of structure (i.e., clear communication of rules) in an attempt to enforce rules through pressuring means, other parents may be high on conditional regard but low on structure. These parents may use conditional regard in a more unpredictable fashion, resulting in a pressuring and at the same time chaotic parenting style with detrimental repercussions for children’s adjustment (Rodríguez-Meirinhos et al., Citation2020).

Our results reveal robust associations between the two dimensions of parental conditional regard and contingent self-esteem. Although they report the same amount of contingent self-esteem, individuals who only perceive high levels of PACPR clearly differ in their affective functioning from individuals who perceive high levels of both strategies. Only experiencing heightened regard in the face of success may specifically foster self-esteem increases following success, whereas PACNR may mainly impact decreases in self-esteem following failure (Li et al., Citation2019).

Conflict of interest disclosures

Each author signed a form for disclosure of potential conflicts of interest. No authors reported any financial or other conflicts of interest in relation to the work described.