• Aucun résultat trouvé

PHENOTYPIC AND GENETIC-ENVIRONMENTAL ASSOCIATIONS BETWEEN EARLY PARENTING AND CALLOUS-UNEMOTIONAL TRAITS IN PRIMARY SCHOOL

Résumé

Cette étude avait deux objectifs: 1) documenter les associations prospectives entre des conduites parentales préscolaires (chaleureuses/gratifiantes et hostiles réactives) et les CI d’âge scolaire, au- delà des conduites externalisées préscolaires de l’enfant; 2) étudier l’étiologie génétique- environnementale partagée entre ces conduites parentales et les CI. Les données provenaient de 662 paires de jumeaux provenant de l’Étude des Jumeaux Nouveau-nés du Québec (ÉJNQ). Les conduites parentales et externalisées étaient rapportées par la mère. Les CI étaient rapportés par des professeurs. Les conduites chaleureuses/gratifiantes et hostiles réactives étaient toutes les deux associées aux CI. Après le contrôle des conduites externalisées préscolaires, seules les conduites chaleureuses/gratifiantes prédisaient les CI. L’association entre l’hostilité réactive et les CI était expliquée par une étiologie génétique commune. Cette association génétique devenait non significative après le contrôle des conduites externalisées préscolaires. Les composantes positives des conduites parentales sont particulièrement susceptibles de contribuer de façon unique au développement des CI.

Abstract

This study aimed: 1) to document prospective associations between parenting in preschool (hostile- reactive and warm/rewarding) and CU traits at school age, over and above child preschool externalizing problems, and; 2) to investigate shared genetic-environmental aetiology between both parenting practices and CU traits. Data were collected in a population sample of twins composed of 662 twin pairs (Quebec Newborn Twin Study – QNTS). Parenting and externalizing problems were reported by the mother. CU traits were reported by teachers. Both hostile-reactive and warm/rewarding parenting were correlated with CU traits. After controlling for externalizing problems, only warm/rewarding parenting predicted CU traits. The association between hostile-reactive parenting and CU traits was mostly explained by a shared genetic aetiology. This genetic association was reduced to non- significance when externalizing problems were included as a control variable. These results suggest that positive aspects of early parenting contribute to CU traits through an environmental pathway.

Introduction

Callous-unemotional (CU) traits include lack of guilt, lack of empathy and shallow affect (Hare, 2003). Adults with elevated levels of CU traits and antisocial behavior are considered psychopaths (Hare & Neumann, 2006). CU traits refer to the affective dimension of adult psychopathy (Cleckley, 1976), and several studies support their importance in the development of antisocial personality (Frick, Ray, Thornton, & Kahn, 2014). Indeed, CU traits characterize severe, stable and aggressive trajectories of antisocial behavior in childhood (Christian et al., 1997; Frick, Cornell, Barry, Bodin & Dane, 2003a, Frick et al., 2003b; Frick & Dickens, 2006; Frick, Stickle, Dandreaux, Farrell & Kimonis, 2005; Vincent, Vitacco, Grisso & Corrado, 2003). There is also some evidence pointing to the unique risk of CU traits for adult psychopathy (Barry et al., 2000; Loney, Taylor, Butler & Iacono, 2007; McMahon, Witkiewitz & Kotler, 2010). Therefore, it is important to clarify the early risk factors and developmental mechanisms underlying CU traits.

Conscience Development: A Theoretical Framework. Current developmental theories regard CU

traits as the result of a failure in the normal development of the moral emotions of guilt and empathy (i.e., conscience; Frick et al., 2014). This proposition is informed by the existing conceptual and empirical links between CU traits and various aspects of conscience: empathy (Dadds, Cauchi, Wimalaweera, Hawes, & Brennan, 2012; Dadds et al., 2009; Frick, 2009), guilt/remorse (Lotze, Ravindran, & Myers, 2010; Pardini & Byrd, 2012), and prosocial behavior, which presumably reflect normative empathy (Roose, Bijttebier, Decoene, Claes, & Frick, 2010; Sakai, Dalwani, Gelhorn, Mikulich-Gilbertson, & Crowley, 2012). It has further been suggested that conscience development in high CU children is impeded because they process affect atypically (Frick et al., 2014), as evidenced by their poorer behavioral modulation following punishment (Frick & Viding, 2009), as well as their difficulties in processing – and reduced amygdala response to – affective stimuli (e.g., fearful faces; Blair & Viding, 2008; Jones, Laurens, Herba, Barker, & Viding, 2009; Marsh & Blair, 2008; Sebastian et al., 2014; Viding et al., 2012). Thus, conscience development may be disrupted in these children because they have difficulty associating their transgressions and subsequent punishments, and they do not find other people’s distress aversive. Consequently, they have difficulties in developing empathy and guilt/remorse (Blair, 2010; Blair & Viding, 2008).

Early Parenting and Callous-Unemotional Traits. While this model does not specify the cause of a

child processing affect atypically, evidence from twin studies suggests that this may be partly due to genetic vulnerabilities (see Viding & McCrory, 2012). This genetic predisposition may notably underpin individual differences in amygdala reactivity to signals of distress (e.g., Sebastian et al., 2014). It is however clear from past research that not all children with atypical affective processing will show later deficits in empathy and guilt. Neither genetic vulnerability nor differences in neurobiological function preclude the role of social factors in the environment, and particularly parenting practices (Viding & McCrory, 2012). Indeed, early parenting patterns have been shown to hinder or promote conscience development, both in the general population (Cornell & Frick, 2007; Kiang, Moreno & Robinson, 2004; Kochanska, 1995, 1997; Kochanska & Murray, 2000) and in children with high CU traits (Dadds et al., 2012).

Two general categories of parenting practices have been most investigated (Waller, Gardner, & Hyde, 2013). First, negative forms of parenting (termed hostile-reactive in the current study to reflect the way this construct was measured) – encompassing a reliance on verbal aggression and/or corporal punishment in managing conflict with the child – were considered (Waller et al., 2013). Hostile-reactive parenting may imply low levels of child exposure to prosocial actions that may help solve stressful situations, which would be positively associated with CU traits (Pardini, Lochman, & Powell, 2007). These parenting practices may also be a proxy for low levels of rewards by the parent, including rewards for prosocial behavior (Pardini et al., 2007). Relatedly, the second form of parenting that was considered is positive parenting (termed warm/rewarding in the present study), which includes high levels of rewards and shared positive affect between parent and child (Waller et al., 2013). Warm/rewarding parenting may be particularly important for the development of a prosocial conceptualization of relationships and emotional responsiveness, both of which are negatively related to the construct of CU traits (Kochanska, 1997). Both hostile-reactive and warm/rewarding parenting may be particularly crucial in preschool, a period during which children show the first signs of understanding right vs. wrong (Eisenberg & Fabes, 1998; Kochanska, Gross, Lin & Nichols, 2002).

Empirical Evidence. Unsurprisingly, extant published reports indicate that both hostile-reactive and

warm/rewarding parenting practices in preschool prospectively predict later levels of CU traits (N.B.: higher and lower, respectively; Barker, Oliver, Viding, Salekin & Maughan, 2011; Hawes, Dadds, Frost & Hasking, 2011; Hyde et al., 2016; Lopez-Romero, Romero & Gomez-Fraguela, 2015; Mills-Koonce

et al., 2016; Wagner, Mills-Koonce, Willoughby, Zvara & Cox, 2015; Waller et al., 2012, 2014, 2016a, 2016b; Waller, Shaw, Hyde, 2016c; for a narrative review, see Waller et al., 2013; for an exception, see Brown, Granero & Ezpeleta, 2015). All published studies have considered child levels of externalizing problems or CU traits at baseline. In a normative sample aged 3–10 years, it was found that positive reinforcement, but not punishment at baseline predicted lower CU traits one year later (Hawes et al., 2011). In another representative sample, both harsh and positive parenting at age four years were associated with CU traits at age 13 years (Barker et al., 2011). Two other studies conducted on normative samples have reported negative associations between warm/rewarding forms of parenting in preschool and CU traits at school age (Lopez-Romero et al., 2015; Wagner et al., 2015). In a high-risk sample, Waller and colleagues found that parental harshness at ages two and three years predicted higher CU traits one year later, and that parental warmth at two years predicted lower CU traits a year later (Waller et al., 2012, 2014). In another high-risk sample, the same group has found an association between parental warmth at two years and lower CU traits at ten and twelve years (Waller et al., 2016c). Another group has found, in a high-risk sample, associations between parental warmth and harshness at 6, 15, 24 and 36 months, and CU traits at six years (Mills-Koonce et al., 2016). Finally, in a recent adoption study, it was found that adoptive mother positive reinforcement at 18 months negatively predicts CU traits at 27 months – even in children at genetic risk for such traits (i.e., history of severe antisocial behavior in the biological mother; Hyde et al., 2016; see also Waller et al., 2016b).

These studies support the notion that early parenting is involved in the development of CU traits, even after accounting for initial child vulnerabilities. However, as implied above, diverging results were found when hostile-reactive and warm/rewarding forms of parenting were entered as competing predictors, one report highlighting the role of hostile-reactive parenting (Waller et al., 2012), and others emphasising that of warm/rewarding parenting (Hawes et al., 2011; Lopez-Romero, Romero & Gomez- Fraguela, 2015; Waller et al., 2016c). This particular question is of interest as past prospective studies on adolescent samples have found positive forms of parenting to be more closely related to conduct problems in youths with high CU traits (Kroneman, Hipwell, Loeber, Koot, & Pardini, 2011; Pasalich, Dadds, Hawes, & Brennan, 2012), while negative forms of parenting were more predictive of conduct problems in youths with normative levels of CU traits (e.g., Hipwell et al., 2007; Oxford, Cavell, & Hughes, 2003; Pasalich et al., 2012; Wootton, Frick, Shelton, & Silverthorn, 1997).

The heterogeneity in results from the above-mentioned reports is thus unforeseen as a sizeable amount of data leads us to expect warm/rewarding parenting/CU associations to be considerably more robust than hostile-reactive parenting/CU associations. This heterogeneity may reflect differences across studies (e.g., normative vs. high-risk samples, observed vs. reported parenting), but also raises potential limitations in the methods of said studies. First and foremost, some studies were limited either by the fact that parents rated CU traits, which is more prone to rater bias than other measures (e.g., teacher reports, observations), or by the fact that parents rated both their own parenting and their children’s CU traits, possibly inflating correlations via shared method variance (Barker et al., 2011; Hawes et al., 2011; Lopez-Romero et al., 2015). Second, few of the available reports have crossed the preschool/school age barrier in examining longitudinal associations between parenting and CU traits (Barker et al., 2011; Lopez-Romero, Romero & Gomez-Fraguela, 2015; Wagner et al., 2015; Waller et al., 2016c). Evidence from reports relying on cross-sectional designs or contiguous parenting/CU assessments may only capture co-occurrence and reflect chaos within high-risk parent-child dyads (or adaptation in low-risk families). In probing the developmental role of early parenting, it appears crucial to determine whether this preschool experience predicts CU traits assessed after school entry – a critical period during which other forms of environmental stress typically emerge (e.g., deviant peers, bullying; Barker & Salekin, 2012; Kimonis, Frick, & Barry, 2004). Finally, the available results are difficult to interpret because most studies did not consider positive and negative forms of parenting simultaneously in prediction models. In brief, no study to date has met all of the following conditions in examining predictive associations between early parenting and CU traits: 1) to consider positive and negative parenting simultaneously; 2) to cross the preschool/school age barrier; 3) to include different raters for parenting vs. CU traits; 4) to include a CU traits rater that is not the parent. Thus, the extent to which early parenting contributes to CU traits remains uncertain.

The Possibility of Child Effects. Another relevant issue is the correlational nature of these reports

that precludes interpretations regarding the direction – child and/or parent-driven – of associations between early parenting and CU traits. In this regard, twin studies, which allow disentangling genetic and environmental sources of variance of a given phenotype, are very useful; they have already highlighted strong genetic contributions to CU traits (Viding & McCrory, 2012). Environmental- developmental associations between putative stressors and phenotypes known to be heritable can sometimes overshadow more complex gene-environment processes – for instance, a gene- environment correlation (rGE; Scarr & McCartney, 1983). Passive rGE occurs when these associations

are mainly a function of shared genetic vulnerabilities between parent and child, while evocative rGE is indicated when children elicit reactions from the caregiving environment because of initial genetic characteristics (Scarr & McCartney, 1983). The twin design helps determine whether the covariance between a given parental practice with a specific child and the child’s behavior is attributable to shared genetic or environmental aetiology. Significant shared genetic variance is generally indicative of rGE, but the type of rGE is not specified.

In the case of early parenting and CU traits, a more accurate examination of evocative rGE would be, first, to test shared genetic variance between these two phenotypes, and second, to determine whether controlling for early child externalizing problems or CU traits reduces this shared genetic variance to non-significance. If that is the case, shared genetic variance between parenting and CU traits is largely attributable to initial child features, making evocation likely. No study to date has employed a twin design to investigate this question pertaining to CU traits. This may prove helpful in determining whether warm/rewarding and hostile-reactive forms of parenting are linked to CU traits through similar or distinct processes.

The Present Study. Initial reports suggest that hostile-reactive and warm/rewarding parenting

prospectively predict CU traits in middle and late childhood. However, previous research in this area relied on contiguous assessments of parenting and CU traits, and was potentially affected by shared method bias. These limitations may contribute to the uncertainty as to which environment is the best predictor of CU traits after accounting for initial externalizing problems in the child. Moreover, the correlational nature of the available reports precludes interpretations regarding the direction of parenting/CU associations. For all these reasons, the present study had two aims: 1) to examine the prospective associations between hostile-reactive and warm/rewarding practices in early childhood, and CU traits across middle and late childhood, accounting for parent-rated child externalizing problems in the preschool years; 2) to test, through multivariate twin modelling, the presence of shared genetic-environmental aetiology (i.e., rGE) between both early parenting practices and later CU traits. Parenting practices were assessed at age five years. This choice mainly relied on the integrity of our rGE testing. Because children’s initial disruptive tendencies tend to consolidate with age (Scarr & McCartney, 1983), evocative rGE (Scarr & McCartney, 1983) is more likely in the late preschool years than in infancy. Moreover, measuring parental practices at this period will likely help capture enduring

patterns in how parents interact with their children. During infancy, parents generally react to punctual challenges as part of rearing young children (e.g., feeding problems, teething); their behaviors towards their child may not accurately reflect their style as parents. To provide a realistic test of rGE and to capture enduring parenting tendencies, we used the latest preschool measure available for hostile- reactive and warm/rewarding practices; these are most likely to be affected by child genetic makeup.

Methods

Participants. The study sample was part of an ongoing prospective study of a normative birth cohort

of twins from the Great Montreal area (Quebec Newborn Twin Study; QNTS) who were recruited between November 1995 and July 1998 (N = 662 twin pairs; Boivin et al., 2013). Zygosity was initially assessed via questionnaire (Goldsmith, 1991) and confirmed with DNA tests on a subsample of same- sex pairs showing a 96% correspondence (Forget-Dubois et al., 2003). Informed consent was obtained for all participating families and teachers. The procedures were approved by relevant ethics review boards of Laval University and St.-Justine Hospital, as well as by the participating schools. The number of twins varies across measures; we therefore resorted to a Full Information Maximum Likelihood (FIML) approach in order to maximize the number of participants in each analysis (see Table 1).

Measures.

Hostile-Reactive Parenting. Hostile-reactive parenting at five years was measured through a seven-

item self-reported questionnaire assessing mothers’ perceptions of their hostile reactions to conflict with the child. Most of these items were derived from the ‘hostile-reactive’ subscale of the Parental Cognitions and Conduct Toward the Infant Scale (PACOTIS; Boivin et al., 2005; “I have been angry with my child when he/she was misbehaving”, “I have spanked my child when he/she was particularly fussy”, “I have lost my temper when punishing my child”, “I have shaken my child when he/she was particularly fussy”, “I had to punish my child multiple times for the same misconduct”, “I have raised my voice or shouted at my child when he/she was particularly fussy” and “I resorted to corporal punishment towards my child”). The PACOTIS subscales have been shown to be valid and internally consistent (Forget-Dubois et al., 2007; Pierce et al., 2010). Items were assessed on a 0–5 Likert scale. In the present study, all internal consistencies were verified by selecting one twin per pair. For this scale,

internal consistency was acceptable (α = .77). Mean scores were computed by averaging the items. High mean scores indicate high levels of hostile-reactive parenting.

Warm/Rewarding Parenting. Warm/rewarding parenting at five years was assessed through a four-

item self-reported questionnaire assessing mothers’ perceptions of how they reward and spend quality time with their child (i.e., “I have talked and/or played games with my child”, “I told my child I was proud of him/her”, “I have spent time with my child playing sports and/or doing activities”, “I have praised my child for a good deed”). The items were assessed on a 0–5 Likert scale. Internal consistency was acceptable (α = .70). Mean scores were computed by averaging the items. High mean scores indicate high levels of warm/rewarding parenting.

Externalizing Problems. Externalizing problems were measured repeatedly at three (32 months), four

(50 months) and five years (63 months) via an eight-item questionnaire assessing mothers’ perceptions of behaviors that reflect externalizing problems in the target children. The items were assessed on a 3- point Likert scale (0 = never; 1 = sometimes; 2 = often or always). Selection of items was based on documented behavioral correlates of CU traits: punishment insensitivity (Dadds and Salmon, 2003; Frick & White, 2008), physical aggression (Fanti, Frick, & Georgiou, 2009; Frick et al., 2003a) and CU traits (including but not limited to core affective traits). Although each time point included the same number of items, division of items across behavioral spectrums slightly differed between the three vs. four and five years’ assessments. That is because availability of certain items differed across ages. For example, some items included in the four- and five-year scales were not developmentally appropriate at three years (e.g., “he/she broke his/her own things”, “he/she broke other children’s things”).

At three years, two items pertained to punishment insensitivity (“he/she was not obedient, refused to comply”, “punishment did not change his/her behavior”), four items pertained to physical aggression (“he/she kicked other children”, “he/she fought with other children”, “he/she pushed other children to get what he/she wanted”, “he/she physically attacked others”), and two items pertained to CU tendencies (“he/she did not seem to feel guilty after misbehaving”, “he/she was cruel to other children”). At four and five years, two items pertained to punishment insensitivity (“he/she was not obedient, refused to comply”, “punishment did not change his/her behavior”), three items pertained to physical aggression (“he/she hit, bit and/or kicked other children”, “he/she fought with other children”, “he/she

physically attacked others”), and three items pertained to CU tendencies (“he/she did not seem to feel guilty after misbehaving”, “he/she broke his/her own things”, “he/she broke other children’s things”). Mean scores were computed by averaging the items. A high mean score indicates high levels of externalizing problems. Internal consistencies were acceptable to good (α = .64–.81). An aggregate externalizing score was created by averaging the mean scores from each time point. The internal consistency of the aggregate score was acceptable (α = .74).

Callous-Unemotional Traits. Teachers provided CU data at seven, nine, ten and 12 years. The five

items were assessed on a 3-point Likert scale (0 = never; 1 = sometimes; 2 = often or always). Three items were selected from the Inventory of Callous-Unemotional Traits, and captured central dimensions of CU traits (Frick, 2003): “he/she did not seem to feel guilty after misbehaving”, “his/her emotions