The Influence of Parenting on Callous-Unemotional Traits and the Implications for the Causes and Treatment of Conduct Disorder

Pezzoli et al. chose to study parenting as the environmental influence of interest in their study of the genetic and environmental contributions to CU traits based on a number of studies showing an association between parenting and CU traits. Parenting has long been considered a critical part of most theories for how children develop serious conduct problems. Furthermore, bringing about improvements in parenting has long been a primary goal of the most effective treatments for children with serious conduct problems (5). For these reasons, the findings from Pezzoli et al. are highly relevant to research on conduct disorder. Their results showed that the variance and stability in negative parenting were influenced primarily by genetics in early childhood but were influenced primarily by shared environmental effects later in childhood. These findings are critical for interpreting previous research on the role of parenting and its influence on conduct disorder, which has typically ignored potential genetic influences on parenting and has not considered potential changes in the factors that influence parenting over the course of development.

Although these are critical findings, I focus my comments primarily on the results related to potential child effects. Specifically, not only was there weak evidence for the presence of child effects (i.e., the influence of children’s CU traits on later parenting), but also the phenotypic associations between the parenting measures and CU traits were modest overall. The authors note a number of possible reasons for these somewhat surprising results, including the age of the children in the sample (e.g., parenting influences may be stronger earlier in development) and the need to study parenting influences using more interactional measures (e.g., observing parent-child interactions). However, one possibility they raise that deserves particular attention is the potential that focusing on other aspects of parenting may have led to different findings.

Specifically, research over the past several decades has suggested that the influence of parenting on children’s conduct problems may differ depending on the child’s level of CU traits. This possibility is best illustrated by results that I published with students over two decades ago, showing that a measure of dysfunctional parenting was strongly associated with conduct problems only in children with low levels of CU traits (6). These findings were interpreted by us and others (7) as showing that parenting has more of an influence on conduct problems in children low on CU traits. Importantly, the measure of parenting that we used in our study included a wide range of parenting behaviors that had been related to conduct problems, including parental involvement with their child, the use of positive reinforcement, the inconsistent use of discipline, poor monitoring and supervision of the child, and the use of corporal punishment. About a decade later, Pasalich and colleagues (8) attempted to replicate our findings, but they separated harsh and inconsistent parenting from warm and responsive parenting. Their findings closely replicated ours for the measure of harsh and inconsistent parenting (i.e., harsh and inconsistent parenting was related to conduct problems only in children with low levels of CU traits). However, their measure of warm parenting showed a very different association with conduct problems; it was more strongly and inversely related to conduct problems in the children high on CU traits.

The importance of warm and responsive parenting for children high on CU traits has been supported by a host of other studies, both in predicting later conduct problems and in their association with CU traits themselves (see reference 9 for a review). Furthermore, these findings are consistent with developmental theories on how empathy, guilt, and other prosocial emotions develop, which emphasize the importance of warmth and responsiveness in the parent-child relationship (10). Although the vast majority of studies linking parental warmth to conduct problems have not used a genetically informed design, Hyde et al. (11) conducted an adoption study showing that responsive parenting moderated the genetic risk for CU traits in young children. Furthermore, there has been evidence for child effects of CU traits in predicting decreases in parental warmth over time. In a study we conducted (12) testing the cross-lagged effects of parental warmth and CU traits in adolescent males who had been arrested and assessed multiple times over a 5-year period, CU traits consistently predicted reductions in parental warmth over time, but parental warmth did not predict reductions in CU traits. Although this study did not use a genetically informed design, it used a methodology that separated between-individual or trait-level associations across individuals (e.g., people higher on CU traits are more likely to have relationships lower in warmth across development) from changes within the individual (e.g., changes in an individual’s level of CU traits at an earlier time point predict changes in warmth in that individual’s relationships at a later time point).

To be clear, this is not a criticism of the Pezzoli et al. study, but it is an important point in interpreting these results. The authors employed measures of negative parental discipline (e.g., smacking and shouting at their children) and negative parenting feelings toward their children (e.g., feeling frustrated and angry at them). Thus, their findings of modest associations between parenting and CU traits and minimal evidence for child effects may be specific to their definition of parenting. Again, the fact that different findings may have been found for measures of warm and responsive parenting was explicitly noted by Pezzoli et al. However, it is important to emphasize that research has accumulated suggesting that how parenting is defined can have a large impact on the associations of parenting with CU traits and conduct problems.

The fact that CU traits may alter the effects of a major risk factor for conduct disorder, such as parenting, highlights a broader issue related to the use of these traits as a specifier. That is, CU traits were included as a specifier in diagnostic classification because they seem to classify important subgroups within the broader diagnostic category of conduct disorder—subgroups that differ on etiologically important processes (e.g., parenting influences and emotional reactivity to other’s emotions) and clinically important characteristics (e.g., severity and types of aggression) (3). Some have argued for other personality dimensions to be included as specifiers for conduct disorder, based largely on the fact that they are highly associated with conduct problems (13). However, constructs can be highly related to the symptoms of conduct disorder for a number of reasons that do not make them good specifiers; for example, the constructs may be a general risk factor for conduct disorder (e.g., problems with impulse control for early-onset conduct disorder) or may be good indicators of the disorder itself (e.g., lying and conning others) (for a more extended discussion of this issue, see reference 2). In fact, if indicators of a construct are highly related to the symptoms of the disorder, it can make them less useful as a specifier (i.e., for distinguishing subgroups within the group of individuals with the disorder) because they may be present for a large number of people with the disorder.

Instead, evidence to support a specifier should focus on those characteristics that best differentiate individuals with the disorder who manifest the specifier from others with the disorder who do not manifest the specifier. This can be done by comparing subgroups within a group of individuals with the disorder defined by a specifier on important criteria that might suggest different causal factors for the group, might suggest different responses to treatment, or might indicate risk for more significant impairment. However, it is important to establish that these differences are not simply a result of variations in the number of symptoms of the disorder itself. For example, an early study from the TEDS sample showed very different patterns of genetic and environmental effects on serious conduct problems at age 7 depending on whether the children also had elevated CU traits, even after controlling for number of conduct problems (14). Alternatively, the usefulness of a specifier could also be tested by determining whether indicators of the specifier are related to important criteria independent of symptoms of the disorder (e.g., CU traits predicting bullying behavior after controlling for number of conduct problems [15]) or whether these indicators interact with symptoms of the disorder in their relationships to important criteria, as has been the case with parenting.

So how does this discussion of specifiers relate to the findings of Pezzoli et al. on the associations between negative parenting and CU traits? As noted above, the measure of hostile parenting behaviors and feelings may not have assessed the types of parenting that are most important for children with elevated CU traits, but instead, these measures assessed parenting behaviors that are either specific to children with conduct problems without CU traits or parenting behaviors that are related to conduct problems more generally. To illustrate this possibility, we showed (16) that in our sample of arrested adolescents followed over 5 years after arrest, parental warmth was uniquely associated with CU traits after controlling for levels of hostile parenting and antisocial behavior, whereas parental hostility was uniquely associated with antisocial behavior, after controlling for CU traits and parental warmth.

In short, not only must studies carefully consider what dimensions of parenting to study, but it is also important for studies of CU traits and conduct problems to consider their shared and unique associations with etiologically and clinically important variables, including parenting. Again, this possibility was recognized by Pezzoli et al. in their statement that their findings may have differed depending on the child’s level of antisocial behavior. However, it is important to consider this possibility as a guide for future research, which could test the separable associations between parental warmth with CU traits and parental hostility with conduct problems, as our group did (15), but by using a genetically informed design like the one employed by Pezzoli et al.

The points I have chosen to highlight here in no way change any of the interpretations made by Pezzoli et al. but were specifically selected because of their potential implications for treating children with conduct disorder, especially those who meet criteria for the “with limited prosocial emotions” specifier that was added in DSM-5. Perlstein et al. (4) conducted an excellent meta-analysis summarizing the results of research on the effectiveness of existing treatments for children with elevated CU traits. Their review and analyses suggested that many evidence-based treatments for conduct problems, especially those that focus on improving parenting, lead to reductions in conduct problems in children with elevated CU traits. Thus, although CU traits are modestly stable and this stability is heavily influenced by genetics, as shown by the results of Pezzoli et al., children with CU traits do respond to treatment. However, the meta-analysis by Perlstein et al. also showed that children with elevated CU traits tended to start treatment with more severe conduct problems than other children in treatment for conduct problems, and despite improving with treatment, those with elevated CU traits often leave treatment with more severe conduct problems than other treated children. Furthermore, although Perlstein et al. noted significant reductions in the level of conduct problems after treatment, there was less clear evidence for significant reductions in CU traits.

It is possible that these findings on existing treatments may be a result of the fact that most evidence-based treatments for conduct problems, especially those with the goal of improving parenting in families of young children with conduct problems, focus only modestly on increasing parental warmth and responsiveness. Instead, they typically include a greater focus on reducing harsh parenting and increasing parental consistency in discipline, the types of harsh parenting studied by Pezzoli et al. As noted by their results, harsh and inconsistent parenting may be related only modestly to CU traits, and based on other research, such parenting may be less related to conduct problems for children with elevated CU traits. In support of this possibility, Kimonis and colleagues (17, 18) showed that when a standard parenting intervention was augmented to teach parents how to increase their warmth and emotional responsiveness in their interactions with their young children, children with elevated CU traits were more likely to leave treatment with their level of conduct problems in a normative range (17), to have their conduct problems remain low over time (18), and to have increases in their level of empathy toward others (17, 18). Thus, to further inform such interventions, research using strong genetically informed and longitudinal designs (such as that used by Pezzoli et al.) is needed. However, to advance the work by Pezzoli et al., researchers should consider carefully how parenting is defined and test whether the role of parenting differs in its relationship with conduct problems depending on the level of CU traits. Such findings could be critical for advancing both causal theories of conduct disorder and for designing more effective treatments for children with this diagnosis.