Mothers’ Implicit Appraisals of their Adolescent as Unlovable: Explanatory Factor Linking Family Conflict and Harsh Parenting

Abstract

This study examined the consequences of negative change in mothers’ implicit appraisals of their adolescents after engaging in a family disagreement. Participants included 194 mothers and their early adolescents (Mage = 12.4 at Wave 1; 50% female) followed over one year. Mothers’ implicit appraisals of her child as “unlovable” were assessed using the GNAT-Child (Sturge-Apple et al., 2015), an associative word-sorting task, before and after engaging in a family conflict task. Mothers’ implicit appraisals, on average, did not become more negative following conflicts with their teen. However, substantial variability was evident, suggesting that important individual differences exist in mothers’ cognitive reactivity to conflict. Greater susceptibility to implicit change predicted more harsh and insensitive parenting in response to their adolescents’ bids for support one year later. This effect held over and above mothers’ emotional reactivity to the conflict, their explicit negative attitudes about their adolescent, and maternal harshness at Time 1. Harsh and insensitive parenting, in turn, mediated the link between maternal implicit reactivity and subsequent increases in adolescent internalizing and externalizing problems. The results suggest that individual differences in maternal susceptibility to changes in implicit appraisals following conflictual interactions serve as a unique determinant of parenting in adolescence.

Sensitive and responsive parenting as children negotiate the developmental challenges of early adolescence fosters more effective emotion regulation skills, promotes healthy psychological adjustment, and buffers teens from extrafamilial stressors, such as peer relationship difficulties (Hazel, Oppenheimer, Technow, Young, & Hankin, 2014; Raudino, Fergusson, & Horwood, 2013). By contrast, overly harsh and controlling parenting during this period contributes to adolescents’ poorer psychological and school adjustment (Branje, van Doorn, van der Valk, Meeus, 2009; Weymouth, Buehler, Zhou, & Henson, 2016). Unfortunately, the transition to adolescence commonly co-occurs with a general decline in parental warmth and the increasing use of intrusive control tactics (Barber, Maughan, & Olsen, 2005; McNally, Eisenberg, & Harris, 1991). These changes highlight the need to better understand potential determinants of harsh parenting across the adolescent transition.

Behavioral systems conceptualizations emphasize the importance of parents’ implicit relationship schemas in guiding how they process past and current experiences as a caregiver (George & Solomon, 2008; Mayseless, 2006). Developing out of parents’ subjective experience and memories of parent-child interactions, these integrated “caregiving representations” act as a set of metaphorical blueprints guiding parents’ interpretation and response in subsequent parent-child interactions. Based on the proposition that changing parents’ caregiving representations can shift parenting, several therapeutic programs consider parental caregiving representations to be target mechanisms underlying intervention effects (e.g., Hoffman, Marvin, Cooper, & Powell, 2006; Kobak & Kerig, 2015). For example, parents’ internalized beliefs and attitudes about their child form an integral component of their broader caregiving representations (Scharf, Mayseless, & Kivenson-Baron, 2015; Snarr, Slep, Smith, & Grande, 2009; Vreeswijk, Maas, & van Bakel, 2012). Accordingly, an intervention found to be effective in reducing problem behavior in adolescents specifically targeted parents’ implicit perceptions, appraisals, and expectations of their teen (Moretti & Obsuth, 2009; Moretti, Obsuth, Mayseless, & Scharf, 2012). Despite the promise of these programs, empirical tests of the underlying assumption that changes in parents’ caregiving representations predict changes in parenting behavior remain relatively rare. In addition, there is a paucity of research examining parental caregiving representations beyond early childhood (Scharf et al., 2015; Vreeswijk et al., 2012).

In response to this gap, the current study tests the hypothesis that short-term changes in parents’ implicit appraisals of their adolescent can set the stage for longer-term detriments in caregiving. A novel measure, the GNAT-Child (Sturge-Apple, Rogge, Skibo, Peltz, & Suor, 2015a), was employed to capture changes in mothers’ implicit appraisals. First, it was posited that mothers’ implicit appraisals would become increasingly negative following a family conflict. Given the emotional nature of parent-adolescent disagreements, conflicts may present a context in which maintaining a positive attitude towards one’s teen is a significant challenge. Second, we tested whether short-term changes in implicit appraisals would impact mothers’ parenting. More reactive mothers, those who evidenced negative shifts in implicit appraisals pre-to-post conflict, were expected to demonstrate lower sensitivity and greater intrusiveness in the face of adolescent distress one year later as short-term cognitive reactivity canalizes into more stable and long-term shifts in parental caregiving representations. Lastly, and building on extant research (e.g., Branje et al., 2009; Weymouth et al., 2016), harsh and insensitive parenting was proposed to contribute to poorer adolescent adjustment.

Caregiving representations as determinants of parenting

The behavioral systems approach posits that parenting is heavily influenced by the caregiving system, an underlying latent goal system evolved to motivate protection and care for dependents (Bowlby, 1969/1982; George & Solomon, 2008). Like all behavioral systems, the caregiving system guides behavior in large part through implicit working models (i.e., representations), that draw on past caregiving experiences to organize interpretation and response to on-going experience. The term “implicit” is used in recognition that the beliefs, biases, and expectations that make up parental caregiving representations operate automatically and largely outside of conscious awareness (Bretherton & Munholland, 2008). Parents’ implicit appraisals of their child as lovable and deserving of care make up an important aspect of their broader caregiving representations, which also include beliefs about their role as parents and about the quality of the parent-child relationship (Scharf et al., 2015).

Caregiving representations are proposed to directly shape parents’ behavior in interactions with their child. Research with infants and young children has shown that parents who attribute their child’s misbehavior to internal, stable characteristics are more likely to use harsh and overreactive parenting practices (Snarr et al., 2009). Similarly, mothers with more negative caregiving schemas were found to be less sensitive, less engaged, and less encouraging in parent-child interactions, (Huber, McMahon, & Sweller, 2015; Slade, Belsky, Aber, & Phelps, 1999). However, despite the well-established findings with parents of young children, when it comes to adolescence, only one study has tested for a direct link between parental caregiving representations and parenting behavior. Scharf and colleagues (2015) found that mothers with more positive representations of their adolescent daughter behaved in a more autonomy-supportive way in a subsequent discussion. The mounting evidence in research with young children suggesting that parents’ caregiving representations play a role in shaping their risk for employing harsh and insensitive parenting tactics, and the paucity of research on parental representations in adolescence, together highlight the need to better understand how parental caregiving representations impact parenting during this developmental period.

Capturing change in parents’ caregiving representations

An in-depth understanding of change may be hampered, in part, by conceptualizations of parental caregiving representations as relatively stable, trait-like, and highly influenced by the early history of parents’ relationships with their children (George & Solomon, 2008). Given the tendency of internal representations to hierarchically organize and solidify over time, parents’ caregiving representations may be fairly stable by the time their child enters adolescence (Allen et al., 2003). However, theory suggests that the process whereby individuals assimilate new experiences into existing representations is complemented by the potential for representations to change as a function of on-going experience (Bretherton & Munholland, 2008). For example, several studies provide support for an evocative effect whereby mothers’ caregiving representations become increasingly negative following exposure to children’s oppositional and challenging behavior (Sturge-Apple et al., 2015a; Theran, Levendosky, Bogat, & Huth-Bocks, 2005). Nonetheless, little is known about how this change occurs, or what the implications of representational changes are for caregiving.

The transition to adolescence may serve as a salient source of discontinuity in the parent-child relationship (Scharf & Shulman, 2006; Steinberg & Silk, 2002). For one, adolescence represents a move from dependent child to independent adult, pushing parents to shift their perceptions of their child’s need for care. At the same time, adolescents undergo significant identity exploration and development, some of which may include the active rejection of parental values. This is likely to present a significant challenge for parents (Koepke & Denissen, 2012). Together, the changes characterizing early adolescence may mark this as a period of heightened plasticity for parents’ internal representations of caregiving.

Changes in parents’ broader caregiving representations may be preceded by smaller, short-term shifts in parents’ responses to on-going interactions with their adolescent. For example, implicit cognition research suggests that affective priming can produce robust, if brief changes in implicit attitudes (Evans, 2010). Affective priming involves the rapid and repeated pairing of an unconditioned stimulus (e.g., child-relevant stimuli) and an emotionally-valenced affective prime (e.g., happy faces, positive words). Following exposure, participants’ implicit appraisals of the unconditioned stimulus become more strongly associated with the emotional valence of the conditioned stimulus (e.g., more pleasant). In a similar way, challenging interactions with one’s adolescent may strengthen associations between the child and negative feelings in parents’ minds, potentially resulting in short-term shifts in parents’ implicit appraisals of their teen. Over time, repeated small shifts in parents’ implicit appraisals may canalize into more substantial changes in parents’ broader caregiving representations, ultimately leading to sustained changes in parenting. Thus, we propose that the changes characterizing early adolescence will heighten parents’ reactivity to family conflicts in ways that increase the negativity of their implicit appraisals and, over time, disrupt their ability to provide sensitive, autonomy-supportive parenting.

Parent-adolescent conflict as a context for change

Parent-child conflict is a hallmark of the adolescent period and typically peaks in early adolescence (Laursen, Coy, & Collins, 1998). Conflicts are not inherently negative for adolescents and disagreements with parents can provide a context for establishing greater autonomy and renegotiating their role in the family (Adams & Laursen, 2007; Smetana, Campione-Barr, & Metzger, 2006). Compared with teens, we know relatively little about the experience of parent-adolescent conflict from the parents’ perspective. Prior research suggests that conflict tends to be more distressing for parents than for teens and difficulties in the parent-adolescent relationship are a primary source of parenting stress during this period (Branje et al., 2009; Steinberg, 2001). Parents may view their adolescents’ attempts to gain greater autonomy as a personal rejection, or as a worrisome sign that their child is getting out of control rather than an opportunity to negotiate changing family roles (Bugental, Brown, & Reiss, 1996). For some parents, conflict may signal an important shift in family dynamics. With this in mind, the current study examines individual differences in the degree to which participating in family conflicts increase the negativity of parents’ implicit appraisals of their adolescent.

Representational change and parental caregiving behavior

Given the role of caregiving representations in shaping parenting, the second aim of this study was to test whether individual differences in parents’ implicit reactivity to conflict shape parenting in a caregiving context. Similar to other levels of reactivity (e.g., behavioral, physiological), repeated, short-term increases in the negativity of parents’ implicit appraisals may canalize, over time, into more entrenched negative beliefs and expectations for adolescents’ behavior (Repetti, Robles, & Reynolds, 2011). In this way, even small changes in implicit appraisals may come to influence parents’ broader caregiving representations and, ultimately, the functioning of the parental caregiving system. Negative representations, in turn, may blunt parents’ sensitivity to their adolescents’ distress and result in caregiving being subsumed under other socialization goals (e.g., control/obedience), goals that may lead to harsh and intrusive tactics. Therefore, we propose that parents who evidence stronger reactivity to conflict, as evidenced by an increase in the negativity of their implicit appraisals of their teen, will demonstrate more harsh and insensitive parenting in future caregiving interactions with their adolescent. In turn, parenting characterized by insensitive, hostile, and overly controlling behavior has been shown to undermine adolescent adjustment (Branje et al., 2009; Wang & Kenny, 2014; Weymouth et al., 2016). Harsh parenting may be particularly detrimental in limiting adolescents’ growing needs for autonomy and capacity for self-regulation (Collins & Steinberg, 2006). Therefore, we propose that disrupted caregiving, as evidenced by harsh and insensitive parenting in response to adolescents’ bids for support, will serve as a mediator linking parents’ reactivity to family conflicts (i.e., negative changes in their implicit appraisals) and adolescents’ psychological adjustment.

The Current Study

A computerized task, the GNAT-Child (Sturge-Apple et al., 2015a), was developed to capture parents’ implicit appraisals of their adolescent. Using the GNAT-Child, we assessed maternal appraisals across two time points, before and after mothers participated in a family disagreement with their adolescent. Three primary research questions guided our approach. First, we tested whether mothers’ implicit appraisals of their adolescent become increasingly negative following parent-adolescent conflict. Second, we tested whether mothers’ susceptibility to changes in their implicit appraisals would ultimately undermine their caregiving behavior in future parent-child interactions, as evidenced by mothers’ use of harsh and insensitive parenting strategies in response to adolescents’ distress. Lastly, given prior research with children and adolescents suggesting that harsh and insensitive parenting is detrimental for adolescent adjustment, we examined whether negative changes in maternal appraisals following conflict ultimately influenced adolescents’ psychological adjustment over time, mediated through mothers’ harsh parenting in a caregiving context. To provide a rigorous test of the generalizability of our proposed model, we included adolescent gender and family socioeconomic status as covariates. In addition, we included maternal distress as a covariate, as a rough indicator of the intensity and negativity of the family conflict from the mother’s perspective. Lastly, mothers’ self-reported attitudes about her adolescent were included to test the potential value of using the GNAT-Child procedure to capture mothers’ implicit appraisals of her adolescent over and above explicit, self-report measures.

Methods

Participants

Participants for this study included 194 families recruited from a city in a Northeastern area of the U.S. Interested families were included in the project if they met five criteria: (1) they had an adolescent between the ages of 12 and 14; (2) the target adolescent and two parental figures had been living together for at least the previous three years; (3) at least one parental figure was the biological parent of the target teen; (4) all participants were fluent in English; and (5) the target adolescent had no significant cognitive impairments. Families participated in two measurement occasions spaced one year apart. The majority of parents were married or engaged (87%) and another 12% reported being in a committed, long-term relationship. Adolescents lived with their biological mother in the vast majority of cases (94%). Girls comprised 50% (n = 97) of the sample and adolescents averaged 12.4 years of age at Time 1. The median household income for this sample ranged from $55,000 to $74,999 with 14% of the sample reporting a household income under $23,000. Median parental education was an Associate’s degree (i.e., completed two years of college), with most parents (85%) attending at least some college. A smaller subset of the adults in this sample (12 %) earned a high school diploma or GED as their highest degree. The sample largely identified themselves as White (74%), followed by Black (13.5%) and mixed race (10%), and a number identified as being of Hispanic or Latino ethnicity (12%). The retention rate from Time 1 to Time 2 was 93% (180 families).

Procedures

At each wave of data collection, families (i.e., two parental figures and their adolescent) visited the laboratory for a single, three-hour visit. The laboratory included one room designed to resemble a living room and equipped with audiovisual equipment to record family interactions, as well as other comfortable rooms for participants to complete confidential interviews, computerized assessments, and survey measures. The study was approved by the Institutional Review Board. Families received monetary payments for their participation.

During the first measurement occasion, families participated in a structured discussion task (e.g., Cummings, Koss, & Davies, 2015; Saxbe, Margolin, Spies Shapiro, & Baucom, 2012). Prior to the task, each family member chose a topic they felt was a common source of disagreement. They were then brought together and given seven minutes to (a) choose one topic to discuss and (b) attempt to reach a solution. The most common topics chosen were chores (36%), followed by use of electronics (e.g., TV, computer, videogames, phone) (13%), fighting with siblings (10%), and schoolwork (10%). Families were instructed to discuss the topic as they would at home, working toward a resolution. Following the seven-minute discussion period, each family member was asked to report, separately, their emotional experience during the interaction. Mothers completed the computerized GNAT-Child (Sturge-Apple et al., 2015) task, both before and directly following this family disagreement task. Mothers and their adolescents completed all other survey measures independently on a computer.

During the second wave of data collection, mothers and their adolescents participated in a dyadic interaction task designed to elicit maternal support (e.g., Furman & Shomaker, 2008; Gavin & Furman, 1996). Prior to the interaction, adolescents were asked to think of three issues that they felt upset, stressed, or worried about. They were then asked to choose one of these topics to discuss with their mom. The most common topics chosen involved school/grades (30%), family (17%), friends (10%), and sports/extracurricular activities (7%). Adolescents were asked to share the topic they chose with their mother and to discuss with her how to resolve it. Both participants were asked to speak about the topic as they might at home. All interactions were video recorded for later coding. Again, mothers and adolescents completed all other survey measures independently.

Measures

Mothers’ implicit appraisals (Wave 1)

Mothers completed a computerized, fast-paced word sorting task called the Go/No-Go Association Task-Child (GNAT-Child; Sturge-Apple et al., 2015a). The GNAT-Child is designed to measure mothers’ implicit representations of her child by determining the strength of her associations between words representing her child (e.g., the child’s name, shortened name, nickname) and either of two categories with positive and negative valence. For the GNAT-Child, these two poles are captured by eighteen “lovable” (e.g., good, loving, easygoing, wonderful) and eighteen “unlovable” (e.g., ungrateful, awful, annoying, hateful) descriptors (Sturge-Apple et al., 2015a). In this, the GNAT resembles the Implicit Association Task (Greenwald et al., 1998), but has the methodological advantage of comparing associations between a single target and two poles of an attribute dimension rather than contrasting implicit attitudes between two separate target categories (e.g., the target adolescent vs. all other teens).

For the current study, mothers completed two practice blocks of sixteen trials each, followed by two experimental blocks of 49 trials each. In accord with established GNAT procedure (i.e., Nosek & Banaji, 2001), mothers were told to respond by pressing the spacebar (a “go” response) when viewing a word in one of the designated “target” categories (e.g., child and lovable) and to refrain from responding (a “no-go” response) when viewing a word not in the “target” categories (e.g., unlovable). The designated target categories change with each trial and which category was presented first was counterbalanced. To limit mothers’ use of controlled processing, the interval during which each word was presented and the participant could respond was kept very short (650 milliseconds), with a 400-millisecond rest in between. Before each trial, mothers were given instructions on which categories represented targets vs. distractors. These category labels then remained in the upper corners of the screen. After each response, feedback was briefly (i.e., 100 milliseconds) flashed on the screen indicating a correct (a green “O”) or incorrect (a red “X”) response. Instructions before each block of trials identified the target categories for that block, and category labels for those target categories (e.g., child and lovable) were presented in the upper left-hand corner of the screen during the trials. The average sorting accuracy rate was 80%.

The GNAT requires participants to classify stimuli within a short time interval so that task performance represents a trade-off between speed and accuracy. However, this allows for respondents to artificially inflate their hit rates on “go” responses by indiscriminately pressing the spacebar (thereby also increasing their false-positive error rates). To adjust for this possibility, the GNAT requires the calculation of d′ scores (Green & Swets, 1966), which use signal detection theory to account for both accuracy (proportion of correct “go” responses) and error rate (number of incorrect “go” responses) (Nosek & Banaji, 2001). Using a similar method, the GNAT has been shown to demonstrate good internal consistency, test–retest reliability, and discriminant validity in recent studies (e.g., Bar-Anan & Nosek, 2014; Sturge-Apple at al., 2015a). As per recommendation, d′ values that were negative (i.e., reflecting accuracy below chance levels) were treated as missing data. Higher d′ values for the block of trials pairing “child” and “unlovable” words reflected more negative implicit appraisals. Given our interest in change in mothers’ implicit appraisals of her adolescent, mothers completed the GNAT-CHILD both before and directly following the family disagreement task

Maternal harsh parenting (Wave 1)

Our measurement battery did not contain a mother-adolescent support-seeking interaction in Wave 1. Therefore, we used mother and adolescent reports as proxy measures of initial maternal harshness. Mothers completed the Overreactivity subscale of the Parenting Scale (Arnold, O’Leary, Wolff, & Aker, 1993; Rhoades & O’Leary, 2007) which included seven items reflecting mothers’ use of harsh, punitive behaviors in interactions with her teen (e.g., “When my child does something I don’t like, I insult my child, say mean things, or call my child names”). Response options ranged from 1 to 7, with one reflecting the least harsh choice and seven the most harsh. Adolescents reported on their mother’s use of intrusive control tactics on the 10-item Psychological Control Scale (Barber, 1996). Teens respond on a scale of 1 (not like my mother) to 3 (a lot like my mother) to statements such as “My mother is a person who is always trying to change how I think or feel about things.” Internal consistencies for both scales were acceptable: α = .74 and .84, respectively and the scales were correlated, r = .28, p < .001. Each scale was standardized and averaged into a single variable reflecting maternal harsh parenting at Wave 1.

Maternal harsh and insensitive parenting (Wave 2)

At Wave 2, observer ratings were collected using the Caregiving Assessment Scale (CAS; AUTHOR CITATION), a continuous rating system designed to capture parental behavior during a supportive interaction task. The CAS required trained observers to watch the entire parent-adolescent support interaction and provide a continuous rating from 1 (not at all characteristic) to 9 (highly characteristic) based on the frequency, pattern, and intensity of each dimension of parent behavior. Hostility referred to mothers’ use of harsh, angry, critical, and/or rejecting behavior toward the adolescent. Controlling reflected attempts to direct the adolescent to conform to the behaviors, opinions, or points of view desired by the parent. At high levels, the controlling dimension captured intrusive, domineering forms of parental control. Lastly, the Sensitivity dimension assessed the extent to which the parent behaved in a sensitive, engaged, and responsive manner. High levels of sensitivity were reflected in: (a) recognition of and empathy for the adolescent’s perspective and feelings, (b) displays of genuine interest in and attention to the adolescent, and (c) engaging in a way that promotes adolescents’ autonomy. This rating was reverse-scored so that higher values reflected low sensitivity. Two independent coders overlapped on 20% of the interactions, resulting in intraclass correlation coefficients of .79, .88, and .77 across the three dimensions, respectively. The maternal behaviors observed resemble those found in previous studies using a similar observational methodology to assess parental harshness (e.g., Buehler & Welsh, 2009; Schofield, Conger, Gonzales, & Merrick, 2016; Sturge-Apple, Davies, Cicchetti, & Manning, 2012).

Adolescent adjustment (Waves 1 and 2)

In building on our multi-method, multi-informant design, adolescents completed self-report assessments of their psychological adjustment at each wave using the subscales of the Youth Self-Report (YSR11–18: Achenbach & Rescorla, 2001). The YSR is widely used and well-validated (see Achenbach & Rescorla, 2001). Teens first reported on the anxious/depressed (15 items; “I cry a lot,” “I fear going to school”) and withdrawn (8 items; “I prefer to be alone rather than with others”) subscales. These were then summed to form a single internalizing symptoms scale (α = .80 and .90 at Waves 1 and 2, respectively). Next, teens completed the delinquent behavior (12 items; “I don’t feel guilty after doing something I shouldn’t”), and aggressive behavior (17 items; “I get in many fights”) subscales, which were summed to create an assessment of externalizing symptoms (α = .87 and .89 at Waves 1 and 2, respectively).

Covariate: Socioeconomic status

Mothers and fathers completed a short demographic survey in Wave 1. First, both parents reported on their family’s average yearly income. Mothers’ and fathers’ reports were averaged and ranged from less than $6000 to over $125,000 (Median = $55,000 – $74,999). Each parent also reported their highest degree of education. Mothers ranged from 10^th or 11^th grade to doctoral degree (Median = Associate degree) and fathers from 8^th or 9^th grade to doctoral degree (Median = some college). Each parents’ education and the average family income were standardized and averaged to yield a single score, where higher values reflect higher socioeconomic status.

Covariate: Mothers’ negative emotions during disagreement task

Mothers completed a short questionnaire immediately following the conflict reporting on the degree to which they felt a number of negative emotions (i.e., anger, sadness, worry, upset, helpless) during the interaction. The intensity ratings for each emotion ranged from 0 (not at all) to 5 (a whole lot). We took the highest rating reported across these five emotions, resulting in a single variable reflecting mothers’ most intense negative emotion during the disagreement from 0 to 5.

Covariate: Mother’s explicit negative attitudes

During the first wave of data collection, mothers completed nine items of the Parent Cognition Scale (PCS; Snarr et al., 2009) to assess self-reported, child-centered attributions about misbehavior. Statements range from beliefs about youths’ negative traits (e.g., “My child is very demanding”) to youths’ willful intent to negatively affect the parent (e.g., “My child purposefully tries to make me angry;” “Likes to see how far s/he can push me”). Mothers were asked to report the degree to which they believe each statement accurately reflects the reason for their child’s misbehavior from 1 (never true) to 6 (always true). The nine items evidenced acceptable internal consistency (α =. 91).

Results

Preliminary Analyses

Missing data in this study came from two sources. First, fourteen families (7% of the sample) did not participate in the second wave of data collection. A comparison of these families with those who returned for the second wave evidenced no differences on any of the variables collected at Wave 1. Second, out of 194 mothers who participated in Wave 1, six (3%) did not complete the post-conflict GNAT due to a computer malfunction. Not counting these six, 33 moms’ (17% of the sample) scores resulted in a negative d′ value for at least one of the four GNAT trials (i.e., Child-Lovable Before, Child-Lovable After, Child-Unlovable Before, Child-Unlovable After). Only three moms (1%) had a negative d′ value for more than one trial. These values were treated as missing data. We performed Little’s MCAR test for all variables included in the analyses. This test suggested that the data was missing completely at random: χ² (148) = 170.43, p = .10. Therefore, to maximize our sample size, missing data was estimated using full information maximum likelihood (FIML; Enders, 2001) estimation in Amos 21.0 (Arbuckle, 2012). Table 1 includes descriptive statistics and correlations among variables in the study.

Table 1.

Descriptives and Bivariate Correlations Among Primary Variables

Variable	Mean	SD	1	2	3	4	5	6	7	8	9	10	11
1. SEX			----
2. SES	0.00	0.83	.08	----
Wave 1
3. EXPL	2.73	0.95	−.10	−.01	----
4. EMOT	4.34	4.70	−.03	−.03	.22**	----
5. HARSH	0.00	0.78	−.20**	−.03	.39**	.26**	----
6. CINT	6.91	5.07	−.13	.00	.08	.19**	.27**	----
7. CEXT	7.04	5.78	−.10	−.15*	.21**	.09	.36**	.62**	----
Wave 2
8. CNTRL	4.50	2.66	.03	−.17*	.09	.10	.22**	.17*	.22**	----
9. HOST	2.70	2.22	.03	−.31**	.19*	.05	.20*	.00	.13	.70**	----
10. INSENS	3.75	2.13	.08	−.24**	.14	.08	.14	.13	.09	56**	.68**	----
11. CINT	8.61	7.13	−.10	−.09	−.03	.14	.13	.47**	.35**	.20*	.14	.13	----
12. CEXT	8.18	6.89	−.01	−.17*	.19*	.04	.21**	.33**	.61**	.22**	.26**	.24**	.52**

Note. SEX = Adolescent gender; SES = Family socioeconomic status; EXPL = Mothers’ explicit negative attitudes; EMOT = Mothers’ negative emotion during the conflict task; HARSH = Mom and teen reports of maternal harsh parenting; CINT = Teen self-reports of internalizing symptoms; CEXT = Teen self-reports of externalizing problems; CNTRL = Observer reports of mothers’ control; HOST = Observer reports of mothers’ hostility; INSENS = Observer reports of maternal insensitivity.

^*p < .05;

^**p < .01

Change in Mothers’ Implicit Appraisals

First, we were interested in whether mothers, as a group, would exhibit increasingly negative implicit appraisals of their adolescent following a disagreement. Consistent with the broader literature, mothers demonstrated better performance pairing child stimuli (e.g., adolescents’ name) with Lovable words (e.g., loving, good) compared with Unlovable words (e.g., awful, annoying), both before, t(163) = 12.78, p < .001, and after the family conflict, t(180) = 11.74, p < .001. This would suggest that, on average, mothers held more positive than negative implicit appraisals of their adolescent.

An assessment of the d′ scores over time suggests that mothers’ were more accurate in sorting both Lovable, t(184) = 2.69, p = .01, and Unlovable words, t(153) = 3.82, p < .001, the second time they completed the GNAT. However, the d′ scores for mothers’ performance on the Lovable and Unlovable trials were positively correlated (r = .48 and r = .42 before and after conflict, respectively). This is a common occurrence in the use of implicit measures, including the GNAT (Greenwald et al., 1998; Lee, Rogge, & Reis, 2010; Nosek & Banaji, 2001). The positive correlation is proposed to reflect shared method variance, such as mothers’ general verbal ability, effort expended, amount of sustained attention, comfort with computers, or grasp of the task. This interpretation is supported by a positive correlation between SES, which includes maternal education, and mothers’ performance on all four trials: rs ranged from .19 to .36. The recommended procedure for removing this common method variance is to regress performance for one valence onto performance for the opposite valence (e.g., Boldero, Rawlings, & Haslam, 2007; Lee et al., 2010; Sturge-Apple et al., 2015a).

Since our primary interest in this study was whether participation in family conflict increased mothers’ negative implicit appraisals of their adolescent, we regressed mothers’ d′ scores for Lovable trials onto her performance for Unlovable trials within a latent difference score (LDS) model of change in implicit appraisals pre- and post-conflict. This model is illustrated in Figure 1. The end result is that the “Unlovable before” and “Unlovable after” negative implicit attitudes variables reflect mothers’ accuracy in the Child-Unlovable trials (i.e., the strength of her negative implicit appraisals of her child), controlling for the overlapping variance accounted for by the d′ scores for Child-Lovable trials (Keith, Bennetto, & Rogge, 2015; Sturge-Apple et al., 2015a). The LDS was specified so that change in mothers’ negative implicit appraisals could be estimated and statistically predicted from her earlier implicit appraisals. LDS models include a latent intercept and latent change score, modeled as a function of both the linear slope (i.e., constant change) and the previous score (i.e., proportional change), providing an assessment of interindividual differences in intraindividual change (McArdle, 2009). Thus, the LDS model can determine whether: (a) mothers’ implicit appraisals, as a whole, increase in negativity pre-to-post conflict; and (b) mothers exhibit significant individual differences in the degree to which their implicit appraisals change following conflict.

Figure 1.

Latent difference score modelling change in maternal unlovable appraisals of her adolescent pre-to-post conflict, using residualized scores that control for maternal lovable appraisals assessed using the GNAT-Child (Sturge-Apple et al., 2015). The path coefficients are standardized. **p < .001.

To test our first hypothesis, an unconditional LDS was run in which the association between initial status and latent change was modeled as a correlational path (see Figure 1). An unconditional LDS facilitates the examination of the mean change score. Results indicated that the mean initial level (.33; z = 1.06) and change (.10; z = 0.24) in mothers’ negative implicit appraisals were not significantly different from zero at p < .05. Thus, participating in family conflicts did not, on average, increase mothers’ negative appraisals of their adolescent. However, variance was evident in both initial levels (1.00; z = 9.00, p < .001) and change (1.69; z = 8.90, p < .001), suggesting that while mean level change was not apparent, there may be important individual differences in the susceptibility of mothers’ implicit appraisals to change post-conflict.

Implicit Appraisals Predicting Parenting

Therefore, we sought to test our second hypothesis that mothers whose negative implicit appraisals of their adolescent increased post-conflict would be more likely to adopt harsh and insensitive parenting strategies in future caregiving contexts. A structural equation model was specified in which the LDS modeling change in mothers’ negative implicit appraisals after family conflict predicted a latent construct of maternal harsh and insensitive parenting one year later. In this analysis, we adopted a conditional LDS approach by specifying a regression path in lieu of a correlation between Time 1 to Time 2 Unlovable scores on the GNAT. This approach provides a more conservative test of whether change in mothers’ implicit appraisals predict change in parenting by controlling for the initial negativity of their implicit appraisals (McArdle, 2009). In addition, the latent construct for harsh and insensitive parenting was derived from three manifest indicators: hostility, controlling behavior, and insensitivity. Each loaded adequately onto the latent variable; β = .94, .75, and .74, respectively, all p < .01. Mother and adolescent reports of maternal harsh parenting at Wave 1, as well as the study covariates (i.e., adolescent gender, family SES, maternal explicit negative attitudes, maternal distress during the conflict) were specified as predictors in the model, which evidenced good fit: Χ²(21) = 21.56, n.s., CFI = 1.0, TLI = .99, RMSEA = .01, CMIN/df = 1.03. As predicted, increasingly negative appraisals pre- to post-conflict were uniquely associated with greater maternal harshness towards the adolescent one year later, β = .24, p = .002. This held over and above mothers’ emotional reactivity, explicit attitudes, and the autoregressive pathway between early harsh parenting and maternal harshness and insensitivity at Wave 2, β = .19, p = .03. Lastly, higher family SES predicted less maternal harshness and insensitivity, β = −.31, p < .001. No other pathways were significantly different from zero.

Indirect Effect of Implicit Change on Adolescent Adjustment Through Parenting

Lastly, we proceeded to test the hypothesized mediational model, in which we propose that increasingly negative maternal appraisals will ultimately undermine adolescents’ psychological adjustment by increasing mothers’ use of harsh and insensitive parenting. Two separate structural equation models were run, one examining adolescent internalizing symptoms and one for externalizing. In each model, adolescent psychological adjustment problems were added for Waves 1 and 2, and an autoregressive pathway was specified. Additional pathways were specified between each predictor (i.e., latent change in maternal appraisals, maternal harsh parenting) at Wave 1, as well as between the latent construct for maternal harsh and insensitive parenting and Wave 2 adjustment. Adolescent gender, family SES, mothers’ explicit attitudes, and maternal distress were retained as covariates.

Internalizing Symptoms

The model including adolescent self-reported internalizing symptoms evidenced good fit: Χ²(28) = 35.97, p = .14, CFI = .98, TLI = .94, RMSEA = .04, CMIN/df = 1.29. Adolescent adjustment remained relatively stable from Wave 1 to 2, β = .43, p < .001. Maternal harshness at Wave 1 continued to uniquely predict harsh and insensitive parenting at Wave 2, β = .17, p = .05. Similarly, mothers’ increasingly negative implicit appraisals continued to predict more harsh and insensitive parenting one year later, β = .26, p < .001. Supporting the second link in the proposed mediational chain, maternal harsh and insensitive parenting at Wave 2 predicted increasing adolescent internalizing symptoms at Wave 2, controlling for initial symptoms, β = .19, p = .03 (see Figure 2). Change in maternal appraisals did not demonstrate a direct effect on adolescent internalizing symptoms. As further support for the proposed model, we conducted a test of the indirect effect using Tofighi and MacKinnon’s (2011) RMediation program. Results indicated that the proposed indirect pathway (μ = 0.27) was significantly different from zero, 95% CI [.02, .61]. In addition, family socioeconomic status, β = −.30, p < .001, and adolescent gender, β = −.15, p = .05, uniquely predicted mothers’ harsh and insensitive parenting at Wave 2. Mothers from families low in SES and with adolescent boys were more likely to be harsh and insensitive during the support-seeking task. No other pathways were significantly different from zero.

Figure 2.

A structural equation model testing the proposed mediational pathway from change (increases) in maternal unlovable appraisals of her adolescent post-conflict to harsh parenting to teen adjustment problems. Light gray paths were included in the model, but did not reach significance at p < .05. Path coefficients are standardized. *p < .05 **p < .001.

Externalizing Symptoms

A second structural equation model was run testing adolescent self-reported externalizing symptoms. This model also evidenced good fit: Χ²(28) = 32.02, p = .27, CFI = .99, TLI = .97, RMSEA = .03, CMIN/df = 1.14. The results are illustrated in Figure 3. As with the previous model, adolescent adjustment remained relatively stable, β = .59, p < .001. However, the association between maternal harshness at Wave 1 and observed harshness and insensitivity at Wave 2 was no longer significant at p < .05: β = .16, p = .07. The hypothesized indirect effect was similar in this model. Change in mothers’ implicit appraisals predicted harsh and insensitive parenting at Wave 2, β = .25, p = .002, which, in turn, was associated with adolescent externalizing problems controlling for earlier levels, β = .16, p = .04. We again conducted a test of the indirect effect, indicating that the indirect pathway (μ = 0.21) was significantly different from zero, 95% CI [.01, .49]. Family socioeconomic status continued to evidence a direct effect on Wave 2 harshness and insensitivity, β = −.30, p < .001. No other pathways were significantly different from zero.

Figure 3.

A structural equation model testing the proposed mediational pathway from change (increases) in maternal unlovable appraisals of her adolescent post-conflict to harsh parenting to teen adjustment problems. Light gray paths were included in the model, but did not reach significance at p < .05. Path coefficients are standardized. ^†p = .07 *p < .05 **p < .001.

Discussion

This study was the first to apply the GNAT-Child to capture mothers’ implicit appraisals of her child in an adolescent sample and the first to examine change in these appraisals over a short period of time (i.e., approximately 10–12 minutes). Given our hypothesis that the developmental transitions characterizing early adolescence would mark a time of heightened plasticity in parents’ implicit appraisals of their teen, we proposed that mothers’ participation in family conflicts would alter their implicit appraisals for the worse. Furthermore, we predicted that more negative implicit appraisals would, over time, result in significant disruptions to the parental caregiving system, ultimately contributing to adolescent psychological adjustment problems by undermining mothers’ ability to provide sensitive and supportive care.

Both stability and change in implicit appraisals

In contrast to our first hypothesis that their child’s adolescence would represent a period of developmental change in caregiving representations for most mothers, participation in a family conflict did not result in a mean increase in the negativity of mothers’ implicit appraisals of their adolescent. Combined with the fact that most mothers viewed their adolescent as more lovable than unlovable, these findings are consistent with prior research suggesting that many, if not most, families move through a child’s adolescence with no significant disruptions in the parent-child relationship (Adams & Laursen, 2007). The majority of parents may not experience conflicts with their teen that are intense or disruptive enough to meaningfully disturb caregiving motivations. For most parents, caregiving representations may be relatively stable by the time their child reaches adolescence (Allen et al., 2003). Strong, positive global caregiving representations may buffer many parents from relatively transient shifts in implicit appraisals. Alternatively, the structured nature of the family interaction task in the lab may have limited the intensity of parent-adolescent conflicts. Parents and teens may have felt greater pressure to reach a resolution or to curtail more extreme responses (e.g., shouting) compared to typical interactions at home, potentially decreasing the likelihood of eliciting change in parental implicit appraisals.

However, substantial variability in the degree to which mothers’ appraisals changed following conflict was also evident, indicating that implicit appraisals vary in their sensitivity to conflict and openness to change. Implicit reactivity does not appear to be solely a matter of emotional reactivity, as mothers’ self-reported negative affect during the conflict did not predict change. Neither maternal harshness at Wave 1 nor mothers’ explicit negative attitudes about their adolescent predicted susceptibility to implicit change either. This suggests that implicit reactivity, parents’ susceptibility to rapid shifts in implicit appraisals, serves as a distinct mechanism accounting for individual differences in parenting. Future work is needed to understand what drives individual differences in parents’ susceptibility to change in their implicit appraisals. For example, research on the factors associated with parents’ explicit attitudes and beliefs about their child point to parental executive functioning and physiological regulation as potential moderating influences (Sturge-Apple, Skibo, & Suor, 2014; Wang, Deater-Deckard, & Bell, 2016).

At a conceptual level, the variability in mothers’ implicit reactivity to conflict contributes to our understanding of the role caregiving representations play in parenting process models. Internal representations are thought to be relatively stable, integrated relationship schemas (Mayseless, 2006). However, these findings suggest that parents’ implicit appraisals of their adolescent, considered a component of broader representations, can change relatively quickly. It is important to point out that we are not suggesting that a single family conflict will elicit prolonged representational change. Instead, we propose that parents’ immediate response to even relatively small stressors in the family reflect variations in their underlying sensitivity to negative interpersonal cues in the environment. Given the frequency of conflicts during adolescence, parents who are more sensitive to conflict may be experiencing repeated, small shifts in the negativity of their implicit appraisals. If repeated over time, these small changes may have the potential to build towards larger developmental shifts that, ultimately, result in prolonged disruptions to their caregiving motivation (Repetti et al., 2011). Thus, implicit appraisals may reflect a source of “cognitive reactivity to conflict” that operates uniquely, but similarly, to emotional or physiological reactivity in parenting process models. The recurring nature of parent-adolescent conflicts was supported by mothers’ self-reports of conflict frequency in the current study. Following the family disagreement task, mothers were asked: “In the past year, how frequently have disagreements about [the chosen conflict topic] occurred?” On a scale of 1 (never) to 5 (very often), mothers reported a mean of 3.57 (SD = 1.00), indicating that these parent-child disagreements occur often. An important next step for research will be to directly test the proposed unfolding cascade by examining changes in the relationship between parents’ implicit appraisals of their adolescent, conflict, and caregiving over longer periods of time.

At a methodological level, this was the first use of the GNAT-Child with mothers of adolescents. The GNAT-Child is uniquely designed to provide a quick and relatively simple method of tapping into parents’ automatic, implicit appraisals of their child. It also has the advantage of asking parents about their child directly, as opposed to recalling memories of past caregiving interactions. By capturing automatic associations, the GNAT-Child is not limited to what mothers can consciously recall and avoids concerns for impression management (Stone et al., 2000; Sturge-Apple et al., 2015a). This may be especially important when asking parents to report on their child’s “unlovability.” The utility of the GNAT is supported by the finding that change in implicit appraisals predicted parenting over and above mothers’ self-reported negative attitudes. This adds to a growing literature demonstrating the value of implicit assessments in predicting actual behavior (e.g., Cameron, Brown-Iannuzzi, & Payne, 2012; Sturge-Apple, Rogge, Pelts, Suor, & Skibo, 2015b). With these strengths in mind, the GNAT-Child represents a practically useful complement to relatively time-intensive interview assessments.

Implicit appraisals, harsh parenting, and adolescent adjustment

Mothers who viewed their adolescent as increasingly unlovable post-conflict were more likely to exhibit harsh and insensitive parenting one year later. This connection may represent a developmental process whereby parents’ initial implicit reactivity in response to conflictual interactions, over time coalesces into more stable negative beliefs and expectations about the adolescent that bias parents towards behaving in more defensive or coercive ways. This interpretation is supported by the results linking mothers’ implicit reactivity to conflict and their parenting behavior during a support-seeking interaction. The support task was designed to place the adolescent in a position of vulnerability, with the goal of minimizing parent-adolescent antagonism and maximizing the saliency of contextual cues (e.g., adolescent distress) motivating parents to provide care. In spite of this, mothers who viewed their adolescent in an increasingly negative light following the conflict at Time 1 responded to their adolescents’ distress a year later with greater hostility, intrusive control, and insensitivity.

Prior research demonstrates a clear link between harsh and insensitive parenting and adolescent psychological and school adjustment problems (Branje et al., 2009; Weymouth et al., 2016). The current findings corroborate and extend this work to show that maternal harsh and insensitive parenting in a caregiving context is similarly associated with adolescent adjustment. Moreover, previous research indicates that there may be differential effects of harsh parenting on adolescents’ internalizing and externalizing symptoms (e.g., Reitz, Dekovic, & Meijer, 2006). By contrast, we found the association to be similar across both types of adjustment problems. In addition, although this represents only a first step, the indirect effect suggests that short-term changes in mothers’ implicit appraisals may be a determining factor in this parenting process. Lastly, the results may have implications for attachment-based therapeutic (ABT) interventions. On the one hand, these findings support the assumption that parental appraisals of their adolescent are a relevant therapeutic target. On the other hand, existing ABTs commonly rely on parent-adolescent disagreements as the context for training and shifting parental cognitions (e.g., Ewing et al., 2015; Moretti & Obsuth, 2009). Although the current findings await replication, they do point to the potential value of extending the context of training to include non-conflictual (i.e., caregiving) family interactions.

Limitations and Future Directions

Interpretation of these findings must be considered in light of the study’s limitations. First, mothers and their adolescents only participated in the support-seeking discussion task during the second wave of data collection. Although adolescent and mother reports of maternal harsh parenting in Wave 1 were controlled for, we cannot definitively conclude that change in mothers’ representations in Wave 1 predicted subsequent change in parenting in a caregiving context. A related limitation is the concurrent assessment of maternal harsh and insensitive parenting and adolescent adjustment problems at Wave 2. Although the results support an indirect effect, future research is needed that can fully disaggregate maternal harshness and adolescent adjustment in time. Second, as a first step in examining change in parent implicit appraisals, this study included only mothers. Exploring the potential for change in fathers’ implicit appraisals is an important future direction, and caution should be exercised in translating these findings to the development of fathers’ representations of their adolescent. Third, it will be important for future research to contextualize changes in parents’ implicit appraisals within a broader network of co-occurring response processes. This may include intra-personal processes, such as parents’ physiological reactivity and interpersonal processes such as parent-child interaction dynamics. Lastly, it will be important to expand the assessment of parents’ implicit appraisals of their adolescent to incorporating other aspects of parents’ broader caregiving models: the self as caregiver and the parent-child relationship. This will include comparing implicit appraisals as assessed using the GNAT with interview assessments of parental caregiving representations (e.g., Scharf et al., 2015; Vreeswijk et al., 2012).

Conclusion

Despite these limitations, the current study serves to advance the literature in several ways. It provides initial evidence of the value of adopting implicit methods for assessing parents’ appraisals of their adolescent, and potentially their caregiving representations. The results signify the importance of considering change in parental appraisals as a significant contributor to parenting behavior. It also begins to build on the scarce knowledge regarding the nature of parent-adolescent conflict from the parents’ perspective. In addition, this study contributes to the theory and research on the caregiving system in organizing parenting. Although parent-adolescent conflict may serve the desirable function of helping adolescents to negotiate greater autonomy, conflict may also have negative consequences for adolescents’ adjustment if it serves to undermine parents’ ability to provide sensitive and responsive support.