A Positive Parenting Program to Enhance Positive Affect in Children of Previously Depressed Mothers

Abstract

Children of mothers with a history of depression are at heightened risk for developing depression and other maladaptive outcomes. Deficits in parenting are one putative mechanism underlying this transmission of risk from mother to child. The current study evaluated whether a brief intervention with mothers with a history of depression produced greater use of positive parenting behaviors and an increase in observed positive affect in their 8- to 10-year-old children. Mothers with a history of depression (n = 65) were randomly assigned to either a positive parenting intervention or an attention control intervention condition. In addition, a comparison group of 66 mothers with no history of depression was evaluated one time. Results revealed significant increases in positive parenting behaviors (e.g., active listening, praise) immediately post-intervention in mothers randomized to the positive parenting intervention as compared to those in the attention control condition. Children of mothers in the positive parenting intervention showed increases in positive affect as compared to children of mothers in the attention control intervention. Increases in mothers’ active listening and smiling/laughing significantly predicted increases in children’s positive affect. The intervention did not increase the rate of children’s moment-by-moment positive affect contingent on mothers’ positive parenting behaviors. This study showed the short-term effectiveness of a brief parenting intervention for enhancing interactions between mothers with a history of depression and their children by directly targeting mothers’ positive parenting and, indirectly, children’s expressions of positive affect.

Depression affects approximately 7.5 million parents in the United States each year, leaving 15 million children at risk (National Research Council and Institute of Medicine, 2009). Depression in mothers, whether current or lifetime, is a well-established risk factor for their children’s maladaptive outcomes, including internalizing symptoms (Goodman et al., 2011; Goodman et al., 2020). Even prior to the typical age of onset of depression, children of depressed mothers show problems in affective, behavioral, cognitive, and biological functioning that may reflect underlying vulnerabilities and early markers of risk for psychopathology (Goodman et al., 2011). Such markers are potential targets for interventions aimed at preventing children’s development of depression and other negative outcomes. The current study targeted positive affect in children as one such potential marker (Forbes & Dahl, 2005).

Positive Affect

Compared to children of mothers with no history of depression, offspring of mothers with a history of depression show less positive affect during interactions with their mothers, regardless of mothers’ depression being current or remitted (Durbin et al., 2005), and across a large age range from infancy through adolescence (McMakin et al., 2011; Olino et al., 2011). Lower positive affect, measured as quantity or intensity, is concurrently and prospectively associated with worse emotional and cognitive functioning and less positive interpersonal relationships (for a review, see Ramsey & Gentzler) and lower social acceptance from peers (Hartup et al., 1967). Thus, low positive affect is a potential marker of vulnerability for depression, particularly among offspring of mothers with a depression history (Davis & Suveg, 2014; Forbes & Dahl, 2005). Moreover, children’s moment-to-moment expressions of positive affect can be quantified in real time through behavioral observation (Davis & Suveg, 2014). Therefore, children’s positive affect may be a promising target of interventions to reduce the intergenerational transmission of depression. In the current study, we focused on mothers with a history of depression given the elevated risk of depression in their children and the public health urgency to intervene early to prevent the onset of depression (Weissman, 2016).

Middle childhood is an important time to examine and intervene to increase positive affect in offspring of mothers with depression histories because it is prior to the rise in the rates of depression during adolescence (Hankin et al., 1998). Middle childhood is characterized by behavioral and neuroregulatory system changes relevant to emotion regulation and by marked interpersonal growth, and parents continue to play an essential role in their children’s development during this time (Collins et al., 2005). Therefore, intervening to change qualities of parenting may be one mechanism by which to influence children’s positive affect.

Depression is associated with low positivity (Clark & Watson, 1991), including in parent-child interactions (Lovejoy et al., 2000). For example, Foster et al. (2008) reported that mothers’ higher levels of current depressive symptoms and more chronic and severe histories of depression were associated with mothers acting less positively (e.g., smiling, laughing, praise, and positive statements) with their young adolescent offspring. Mothers with higher current depressive symptoms also have been found to use fewer positive affect words and be less responsive to their school-aged children’s positive affect related words (Morrow et al., 2021). Mothers’ expressions of positivity have been prospectively associated with better emotion regulation in their school-aged children (Eisenberg et al., 2003; Morris et al., 2007). Moreover, children of parents with recurrent depression who express more positive emotions are nearly twice as likely to remain free of mental health problems through adolescence (Collishaw et al., 2016). Thus, middle childhood may be an optimal developmental window during which to implement interventions that expand positive parenting behaviors of mothers with histories of depression, and thereby increase their children’s positive affect.

Positive Parenting

Positive parenting includes expressions of positive affect, active listening, praise, positive talk, affectionate touch, and less criticism/more positive requests for change (Luebbe & Bell, 2014; Sanders, 2003). Beyond correlational designs, some intervention studies have shown that increased positive parenting lowers children’s internalizing symptoms in offspring of parents with a history of depression (Compas et al., 2010; McKee et al., 2014) and increases children’s positive affect (Webster-Stratton, 1998). Thus, positive parenting is a plausible target for interventions seeking to enhance children’s positive affect. The current study tested the effect of a brief intervention to increase positive parenting and positive affect in children of mothers with a history of depression. Because we were interested in distinct observable components of positive parenting, we observed seven different parenting behaviors rather than relying on a composite across multiple behaviors.

Importantly, parenting behaviors are modifiable (Nowak & Heinrichs, 2008), even among parents with a history of depression (Compas et al., 2010; Goodman & Garber, 2017). Sanders and McFarland (2000) found that teaching positive parenting skills, in the context of behavioral parent training, was as effective as behavioral parent training integrated with cognitive therapy at reducing both mothers’ depression and their children’s problem behaviors. In an uncontrolled trial, Timmer et al. (2011) showed that Parent-Child Interaction Therapy (PCIT) with currently depressed mothers of children 8 to 10 years old was associated with significant pre- to post-treatment increases in mothers’ encouragement of verbalizations and decreases in her discouragement of verbalizations. Thus, positive parenting is modifiable not only in the general population, but also among parents with depression.

The Current Study

The aims of the current randomized controlled experiment were to test the efficacy of a brief intervention designed to enhance positive parenting behaviors in women with a history of depression, and to examine the extent to which enhancing mothers’ positive parenting predicted increases in positive affect in their children. We hypothesized that: (1) mothers with a history of depression would show a lower rate of positive parenting than mothers without a history of depression (a replication); (2) mothers with a history of depression randomized to a positive parenting intervention would show a greater increase in positive parenting behaviors after training than would mothers assigned to an attention control intervention; (3) children of mothers who received the positive parenting intervention would show an increase in positive affect (PA) as compared to children of mothers who received the attention control intervention; (4) increases in mothers’ positive parenting will predict increases in children’s PA, over and above levels of mothers’ positive parenting and children’s PA at pre-intervention, for mothers in the positive parenting condition as compared to those in the attention control intervention. If the predicted interaction with intervention condition is not significant, we then will explore the main effects of positive parenting over time on increases in children’s PA regardless of intervention condition; and finally, (5) for dyads in the positive parenting condition, there will be an increase in contingent associations between positive parenting behaviors and children’s PA, such that intervention type will be associated with an increase in children’s PA immediately following mothers’ positive parenting behavior(s). Note that Hypothesis 5, relative to the previous hypotheses, switched from a focus on averages across the observations to averaging dyadic temporal contingencies between the occurrence of mothers’ positive parenting behaviors followed by children’s positive affect in intervention versus control condition dyads.

For Hypotheses 2 through 4, given our proof-of-concept approach to examining this novel intervention, we expected each of the predicted increases to be present immediately following the intervention, relative to baseline. We further explored whether intervention effects persisted or emerged following a brief intervention refresher one week later. For Hypotheses 2 and 3, we also conducted exploratory analyses to test whether intervention effects persisted or emerged two weeks following the refresher session at T4.

Method

Participants

This study was conducted at Vanderbilt University and Emory University. We recruited participants from the Emory University Child Study Center database, Kaiser Permanente-Georgia, the Vanderbilt University research listserv, referrals, and a birth record database. During the child’s life, eligible mothers either met DSM-IV-TR (American Psychiatric Association, 2000) criteria for a depressive disorder or had no history of depression (i.e., non-depressed group). Exclusion criteria for mothers were current suicidality, psychosis, substance abuse, or lifetime bipolar I disorder or schizophrenia. Child exclusion criteria were a pervasive developmental disorder, intellectual disability, autism, bipolar disorder, psychosis, ever met criteria for a mood disorder, or non-English speaking. We excluded children with a lifetime history of a mood disorder because we aimed to study processes relevant to children before having an onset of depression. Children were 8- to 10-years-old (M = 9.34; SD = 0.85); about half (53%; n = 70) were female. We included only one child per family, randomly selected if there was more than one child in the age range. This yielded 131 mothers, 66 with no depression and 65 with a depression history; 37 of the mothers with depression history were randomized to the positive parenting intervention and 28 were randomized to the control intervention. With this N (plus 2-tailed alpha = .05), we had .99 power to detect at least a medium effect for Hypothesis 1 and between .78 and .99 power to detect at least a medium effect for the other hypotheses. Table 1 presents the demographic characteristics of the samples by depression history and by intervention condition. There were no significant differences by site on any demographic characteristics.

Table 1.

Demographic Characteristics According to Maternal Depression History and Intervention Condition

	Maternal History of Depression				Intervention Condition
	Yes	No	Statistic	ES (d)	Positive Parenting	Nutrition Control	Statistic	ES (d)
Child sex (% female)	56.9	50.0	X2(1, N = 131) = 0.63	0.14	56.8	57.1	X2(1, 65) = 0.001	0.01
Child age (in years), mean (SD)	9.26 (0.82)	9.44 (0.88)	F(1, 129) = 1.53	0.22	9.36 (0.86)	9.18 (0.79)	F(1, N = 64) = 0.21	0.22
Child ethnicity (% Black, African American, Asian, Hispanic, or Other)	53.8	36.4	X2 (1, N = 131) = 4.04*	0.36	48.6	60.7	X2 (1, 65) = 0.93	0.24
Mother age (in years), mean (SD)	38.86 (6.58)	41.49 (7.73)	F(1,125) = 5.78*	0.43	39.16 (5.67)	38.44 (7.75)	F(1, N = 63) = 0.18	0.11
Mother ethnicity (% Black, African American, Asian, Hispanic, or Other)	47.6	30.8	X2(1, N = 128) = 3.82	0.35	44.4	51.9	X2 (1, N = 63) = .34	0.15
Maternal education, highest grade completed (% completed 4-year college)	68.25	80	X2(1, N = 128) = 2.3	0.20	75.0	59.25	X2(1, N = 63) = 1.76	0.17
Maternal marital status (% married)	70.3	89.2	X2(1, N = 129) = 7.17**	0.49	70.3	70.4	X2(1, N = 64) = 0.00	0.00
Mothers’ depressive symptoms (CES-D) at T1, mean (SD)	10.11 (4.95)	6.15 (1.90)	F(1, 127) = 36.15**	6.37	10.21 (4.67)	10.03 (5.23)	F(1, N = 63) = 0.95	−0.04

Note. CES-D = Center for Epidemiologic Studies – Depression Scale; ES = effect size; d = effect size; T1 = Time 1

^*p < .05

^**p< .01

Procedure

The study protocol was approved by the Institutional Review Boards at each site. Mothers provided written informed consent. The data in the current manuscript were part of a larger protocol that included measures of other constructs not a focus of this study. Figure 1 shows the flow of study enrollment, randomization, and participation. Prior to the first visit, all mothers completed an online survey comprised of questionnaires, including the Center for Epidemiologic Studies - Depression Scale (CES-D; Radloff, 1977). The comparison group of mothers without a history of depression visited the laboratory one time and engaged in the mother-child interaction once, for ten minutes (Time 1; T1); none received the intervention.

Figure 1.

Flow of Study Enrollment, Randomization, and Participation

Mothers with a history of depression (N = 65) were randomized to either the active intervention (positive parenting) or the control (nutrition) intervention condition. Randomization was based on a computer-generated randomization list with blocking by child gender and child age, such that mothers of sons and daughters and of older and younger children were equally likely to be assigned to either condition. Mothers and children participated in three visits to the lab. In Visit 1, Time 1(T1; pre-intervention), we video-recorded mother-child interactions for ten minutes; next, the mother met with a coach to receive the assigned intervention for 90 minutes. After the intervention in Visit 1, we video-recorded the Time 2 (T2; immediate post-intervention), 10-minute mother-child interaction. One week later, at Visit 2, mothers in both intervention conditions met with their same coach to review the homework, discussed material she needed help implementing, and asked questions about the intervention content. The mother-child dyads then were again video recorded interacting, which constituted Time 3 (T3). Two weeks after Visit 2, at Visit 3, Time 4 (T4), the mother-child dyads again engaged in the videotaped interaction. Thus, the four time points were pre-intervention (Visit 1, T1), immediate post-intervention (Visit 1, T2), after the second training visit a week later (Visit 2, T3), and follow-up, two weeks after Visit 2 (Visit 3, T4). At the end of each visit, children were given a small toy and mothers were compensated financially. The study was not preregistered. Of the 65 mothers with a depression history who participated with their child at Visit 1, we recorded mother-child interactions of 59 participant dyads at both Visits 2 and 3. The overall retention rate was: Visit 2: 90.77%; Visit 3: 90.77%; all three visits: 87.69%.

Intervention Conditions

The active intervention (positive parenting) and control intervention (nutrition training) were parallel in format, structure, and duration; thus, the control condition served as an attention control while engaging mothers’ interest. Both interventions included psychoeducation, planning, and homework for between-session practice. The interventions were delivered via one- on-one training with a coach – an advanced graduate student in clinical psychology. Coaches were trained in and provided both interventions.

Intervention sessions were video recorded; a randomly selected 20% were reviewed for fidelity to the parenting and control intervention manuals. Fidelity rating scales were based on a review of checklists used to assess similar interventions (e.g., Triple P parenting intervention; Breitenstein et al., (2010). Observers had to attain an Intraclass Correlation exceeding 70% for 3 sessions relative to the team leader before rating study videos. Coaches delivering the parenting intervention covered, on average, 93.93% of all parenting topics; coaches delivering the control training covered 99.4% of all nutrition topics and 0% of parenting topics.

Positive parenting intervention.

The positive parenting intervention manual (available upon request) was based on evidence-based interventions including: Standard Triple P (Positive Parenting Program; Sanders et al., 2001), ACTION Parenting manual (Stark, 2008), Family Focused Treatment Manual (Miklowitz, 2010), Parent Child Interaction Therapy (Eyberg, 1999), and the Coping with Depression Course (Lewinsohn et al., 1991). The specific positive parenting strategies were active listening, verbal and nonverbal praise, showing affection, reflecting on the positive, positive requests for change, and expressing negative thoughts and feelings about specific child behaviors rather than the child. During the first visit, coaches introduced mothers to the concept of positive parenting and had mothers practice each parenting strategy. Coaches worked with mothers to generate a plan for implementing and tracking these strategies with the study child at home over the next week. Finally, the coach elicited from the mother anticipated challenges to performing the skills and then brainstormed with her solutions to these challenges.

Mothers then reunited with their children and participated in the video-recorded interaction described previously. After five minutes of mothers’ interacting with their child, parent training coaches engaged in a 5-minute in vivo coaching session, using a bug-in-the-ear device to provide verbal input to mothers based on the moment-to-moment observed mother-child interaction. This verbal input focused on shaping mothers’ behaviors with labeled verbal praise contingent on her use of any positive parenting skill. Coaches encouraged mothers to use the skills, modeled examples if the coach noticed a missed opportunity to use a skill, and helped mothers positively reframe negative statements if the mother was not doing so on her own. Following this in vivo coaching, the mother and child were observed for another five-minute interaction, which was the T2 post-intervention observation. A week later, the mother and her coach did a second training session, which was like the first session but without the in vivo coaching. They reviewed the homework, problem solved difficulties in implementation, and anticipated concerns about using the positive parenting strategies in the future. Mothers and children engaged in another 10-minute, videotaped interaction, which was the T3 observation.

Nutrition training intervention.

This active control intervention was similar in structure and length to the parenting program. Coaches provided psychoeducation about principles of good nutrition and described strategies for enhancing nutrition in 8- to 10-year-old children, considering the child’s current diet, and for making nutritious family meals. The manual (available upon request) was based on online information from the U.S. government, the Centers for Disease Control, and established nutrition associations. After training (Visit 1), the mother-child dyads completed the video-recorded play interaction again (T2). At Visit 2, one week later, the mother met with same coach to review homework and problem-solve implementation issues. After the review session, mother and child again engaged in a videotaped interaction (T3).

Measures

Maternal Depression

Mothers’ history of mood disorders and the exclusion diagnoses were assessed with a modified SCID-IV, a semi-structured clinical interview (SCID; First et al., 2002). Interviewers were trained and supervised by the senior investigators (Goodman and Garber) who reviewed all information and made the final diagnosis of a past depressive episode or not.

The Center for Epidemiologic Studies - Depression Scale (CES-D; Radloff, 1977) is a self-report measure that assesses level of depressive symptoms over the past week. A total score of 16 or above is considered the clinical cut-off for depression (Radloff, 1977). Internal consistency reliabilities of the CES-D are about 0.85 in a general population and 0.90 in a clinical sample (Radloff, 1977). In the current sample, coefficient alpha was .88.

Observational measures

Mothers and children were video-recorded engaging in a laboratory play interaction using Legos or art supplies provided to them. The child and mother sat at a small table. One camera faced the mother, and another faced the child. A research assistant introduced the materials and encouraged them to do what they would usually do when engaging in an activity together, like they might at home. Similar unstructured and semi- structured play tasks often have been used in the literature, are ecologically valid, and effectively elicit observable qualities of parent-child interactions and child behaviors (Rusby et al., 2015). Trained students transcribed the video-recorded mother-child play interactions and coded mothers’ parenting behaviors using Procoder (Tapp, 2003) and BXC software developed for this study. BXC enabled observers to code interactions based on specific markers from an imported transcript of the video recording.

Positive parenting.

In the positive parenting condition at Visit 1, the post-intervention assessment (T2) mother-child interaction occurred in two five-minute segments, before and after the in vivo coaching. To keep observers unaware of the mothers’ depression history, the condition to which the mother had been assigned (intervention or control), and the time point being coded, we divided all 10-minute sessions into 5-minute segments and randomly assigned them to observers to code without knowing the time point, condition, or maternal history.

To code parenting behaviors, we followed the procedures outlined in Bakeman and Goodman (2019) for training and determining reliability of observers. Each verbalization by the child was counted as an event; mother talk events were unitized in relation to individual child talk events (i.e., mother talk events were punctuated by new child talk events). Training involved each observer independently coding practice segments and then discussing any disagreements with the coding team to reach consensus. Once an observer consistently demonstrated inter-rater reliability of a kappa above .80 for four consecutive segments, they then coded independently. Observers received weekly feedback on their reliability to minimize observer drift.

The codes of parenting behaviors were consistent with the conceptual model based on the Dyadic Parent-Child Interaction Coding System (Eyberg et al., 2004) and the Family Observation Schedule (Sanders et al., 2000). Coded parenting behaviors were combined into these composites: (a) talk about the positive, (b) physical contact, (c) smiling or laughing, (d) active listening, (e) praise, (f) positive requests for change, and (g) criticism. Observers also indicated if a behavior could not be coded (e.g., mother was out of camera view; or someone entered the room). Table 2 show the definitions and coding metrics for each parenting variable. A randomly selected 30% of segments (193 out of 625) were rated by two observers to assess inter-observer reliability, yielding a kappa of .71 across the parenting behavior codes.

Table 2.

Summary of Coding Categories for Observed Parenting Behaviors used in the Analyses.

Behavioral Code Category	Type of Code	Description
Talk about the Positive	Frequency	Number of times the mother said something positive or encouraging to her child about things in their life, current environment, or the activity she and the child were doing together, e.g., “What was the best part of your day? ...That sounds like fun.”
Physical Contact	Frequency	Number of times the mother was affectionate or accidentally touched her child, initiated by the mother [merged affection and accidental touch]
Smiling/Laughing	Duration	Total duration of time the mother spent smiling or laughing
Active listening	Frequency	Number of times the mother looked at, nodded at, questioned, or repeated or responded to her child
Praise	Frequency	Number of times the mother provided unlabeled verbal praise (i.e., a non-specific verbalization that expressed favorable judgment), labeled verbal praise (i.e., a specific verbalization that expressed favorable judgment), or non-verbal praise (i.e., gestured in a way that conveys approval of her child’s specific behavior) to child
Positive requests for change	Dichotomous (yes/no)	Whether or not the mother requested that the child do a specific behavior or expressed negative thoughts or feelings about specific behaviors without using criticism or anger, e.g., “Please tell me what you want using a pleasant tone of voice.”
Criticism	Frequency	Number of times mother verbally expressed disapproval of the child or the child’s attributes, products, or choices, or the mother corrected the child, e.g., “I don’t like it when you make that face.”

Child positive affect.

A different set of coders rated children’s observed affect, on a second-by-second basis from the recorded interactions, using Interact-9 (Mangold, 2020). Observers were unaware of maternal depression history, intervention condition, time point, coded parenting behaviors, and study hypotheses. To train, observers independently coded practice segments and discussed disagreements with the coding team to reach consensus. Observers were required to achieve a kappa above .80 for four consecutive segments, relative to the team leader’s codes. To assess inter-rater reliability of positive affect codes, two observers coded a randomly selected 21% of segments, yielding a kappa of .81. To prevent observer drift, observers received weekly feedback on their reliability.

The Child Affect Coding Rubric (in the online supplementary materials) was based on established systems (e.g., Dougherty et al., 2010); a 7-point scale reflected both valence and intensity. Intensity of children’s affective displays was based on facial expressions (e.g., smiling or laughing), physical gestures (e.g., open or engaged posture), and tone of voice (e.g., warm). Observers also noted when affect was not coded (e.g., child was out of view, or someone entered the room). A positive affect score reflected the proportion of the codable time the child spent in positive affect (regardless of intensity) in each segment, yielding the percent of time child expressed positive affect for each 5-minute segment.

To assess contingency responses, we used Mangold Interact to generate a transitional probability statistic for each mother-child dyad at each time point. This statistic reflected the likelihood that child positive affect (relative to non-positive affect) would follow a positive parenting code action within a 10-second window. The use of a 10-second lag was consistent with other research on parent-child positive affect (e.g., Thomassin & Suveg, 2014).

Data Analyses Plan

We compared demographic characteristics and CES-D scores of mothers with and without a history of depression, and between intervention conditions with Chi-squares for categorical variables and one-way ANOVAs for continuous variables. We inspected the frequency distributions and bivariate correlations among all variables, considered data reduction based on frequency distributions (e.g., drop or merge low frequency codes), and addressed data skew for parenting behaviors. Mauchly’s test of the assumption of sphericity (i.e., the Greenhouse-Geisser estimate was greater than .75) had been violated for two parenting variables – Physical Contact and Positive Requests for Change; we used the Huynh-Feldt correction (Field, 2018). We corrected for multiple tests using family-wise Bonferroni (.05/7 = .007).

We analyzed Hypothesis 1 with t-tests and Chi-squared tests. To test Hypothesis 2, that mothers in the positive parenting intervention would increase in positive parenting, and Hypothesis 3, that their children would increase in PA, we conducted two separate condition by time Repeated Measures Analyses of Variance (ANOVA). We examined increases in mothers’ positive parenting (Hypothesis 2) and increases in children’s PA (Hypothesis 3) from pre- to post-intervention (i.e., T1 to T2), and from pre-intervention (T1) to the short-term follow-up (T3). We also conducted exploratory analyses of changes from pre-intervention (T1) to the follow up three weeks post intervention (T4).

To test Hypothesis 4, we conducted hierarchical regression analyses, with one parenting behavior and child PA from T1 in block 1, the same parenting behavior at the concurrent time point and the intervention condition in block 2. In the final block, we entered the intervention condition by parenting behavior interaction, which if significant, would indicate that the strength of the effect differed for mother-child dyads in the positive parenting versus control intervention. We also compared mother-child dyads in the positive parenting versus the control condition regarding increases in children’s PA, over and above mothers’ positive parenting and children’s PA at pre-intervention, with a focus on changes from T1 to T2 (baseline to immediate post-intervention). To test the hypothesis that intervention effects would persist or emerge after a brief intervention refresher one week later, we explored changes from T1 to T3. We neither anticipated nor tested changes from T1 to T4. When an interaction was not significant, we then tested whether increases in mothers’ positive parenting over time predicted increases in children’s PA regardless of intervention (i.e., main effects). Finally (Hypothesis 5), we generated transitional probabilities in Mangold Interact (Mangold, 2020) to ask: to what extent did changes in children’s PA follow mothers’ use of a positive parenting behavior – the sequential, moment-to-moment relation between mothers’ positive parenting and children’s PA, and whether these probabilities differed by condition.

Data Reduction and Skew of Observed Parenting

Prior to hypothesis testing, we examined the distributional properties of the variables and then conducted data reduction steps based on the frequency distributions of thirteen parenting variables coded from the interactions. Due to low frequencies, we dropped one variable (Expressing Negative Thoughts/Feelings about Specific Behaviors) and merged others into superordinate constructs. See Table 2 for definitions of the seven parenting variables used in the analyses. See Table S1 for correlations among the parenting composite variables. We then examined the distributional properties of the composite parenting variables, separately at each time point. All but one (Active Listening) of the seven variables was not normally distributed across time points. Raters rarely coded Positive Requests for Change more than once per segment; therefore, we recoded this variable as either 0 = no occurrences or 1 = one or more occurrences. The other five non-normally distributed variables were positively skewed. We used a square-root transformation for subsequent analyses because it provided the best fit to the count data and reduced skew. Coding manuals are in the Supplementary Materials.

Results

Demographic Characteristics and Current Maternal Depressive Symptoms

Table 1 presents demographic characteristics and CES-D scores for mothers with and without a history of depression, and for mothers randomized to the positive parenting versus the control intervention. Mothers with a depression history were younger, less likely to be married, more likely to have ethnically diverse children and had significantly higher levels of current depressive symptoms (CES-D) than did mothers with no depression history. Of mothers with a history of depression, 12.31% met or exceeded the established depression cutoff score of 16 on the CES-D, indicating clinically significant levels of current depressive symptoms (Radloff, 1977). Mothers randomized to positive parenting did not differ significantly from those in the control intervention on any demographic variable or any positive parenting behaviors at baseline (see Table 3), thus indicating that randomization was successful.

Table 3.

Descriptive and Test Statistics, and Effect Sizes for Differences in Observed Parenting between Mothers with and without a History of Depression at Time 1, and for Comparisons by Intervention Conditions at Time 1

Maternal History of Depression					Intervention Condition
	Yes (n = 65)	No (n = 66)	Statistic	ES (d)	Positive Parenting (n = 37)	Nutrition (Control) (n = 28)	Statistic	ES (d)
Talk about the Positive, M (SD)	3.55 (3.58)	3.68 (3.73)	t(1, 129) = .39	0.07	3.84 (3.67)	3.18 (3.48)	t(1, 63) = .71	0.18
Physical Contact, M (SD)	1.49 (2.63)	1.74 (2.32)	t(1, 129) = 1.10	0.19	1.43 (2.30)	1.57 (3.06)	t(1, 63) =.293	0.07
Smiling/Laughing, M (SD) (duration in seconds)	49.35 (53.26)	60.52 (68.48)	t(1, 129) = 1.24	0.22	46.31 (43.10)	53.36 (64.94)	t(1, 63) = .021	0.01
Active Listening, M (SD)	52.74 (20.10)	65.80 (26.48)	t(1, 129) = 3.18**	0.55	55.14 (18.44)	49.57 (22.04)	t(1, 63) = 1.11	0.28
Praise, M (SD)	3.69 (3.15)	3.86 (3.64)	t(1, 129) = .25	0.04	3.08 (3.19)	4.50 (2.96)	t(1, 63) =−1.86	−0.47
Positive Requests for Change (% with at least one)	9.2	4.5	X2(1, N = 131) = 1.12	0.19	13.51	0.04	X2 (1, N = 65) = 1.88	0.35
Criticism, M (SD)	1.54 (3.17)	1.55 (2.87)	t(1, 129) = .29	0.05	1.35 (2.98)	1.79 (3.44)	t(1, 63) = −.75	−0.19

Note. All observed parenting behaviors are frequency counts except for Smiling/Laughing and Positive Requests for Change.

ES = effect size

^*p < .05

^**p<.01

Correlations among Demographic Characteristics and Observed Parenting Variables

Pearson’s correlations among continuous variables and the observed parenting variables yielded few significant associations (5.4% of all comparisons tested across all time points); all but one had small effect sizes. More frequent Positive Requests for Change at T3 were associated with younger child age (r = −.28, p = .03); more frequent maternal Criticism at T4 was associated with older child age (r = .36, p = .01). Older maternal age correlated with more frequent Physical Contact with their children at T3 (r = .27, p = .04).

Parenting Behaviors of Mothers with and without a History of Depression (Hypothesis 1)

Table 3 presents the descriptive statistics, test statistics, and effect sizes for differences in observed parenting between mothers with and without a history of depression, and for mothers with depression histories assigned to parenting versus the control intervention. Mothers with a depression history engaged in significantly less Active Listening at T1 (d = 0.55) than did mothers with no depression history. No significant group differences were found on the other parenting behaviors.

Effects of the Parenting Intervention on Parenting Behaviors (Hypothesis 2)

The time by intervention condition interactions were significant for the following parenting variables. Significant differences for T1 versus T2 (i.e., immediate effect of the intervention) were: Active Listening (F = 26.16, p < .001), Smiling/Laughing (F = 4.93, p = .03), and Praise (F = 5.07, p = .03). For T1 versus T3 (effect of the intervention one week later, after a refresher): Active Listening (F = 16.37, p < .001) and Praise (F = 8.74, p = .005).

Active Listening.

The significant time by condition interaction indicated that increases in Active Listening from T1 to T2, F (1, 53) = 21.71, p = .00, and from T1 to T3, F (1, 53) = 8.70, p = .01, were greater for mothers in the parenting intervention as compared to mothers in the control intervention condition (see Figure 2a). Given that Active Listening differed between mothers with and without a depression history at T1, we conducted an exploratory analysis to compare mothers with no depression at T1 with mothers in the parenting intervention at T2 and T3. The t-tests indicated that mothers with depression histories who were in the parenting intervention engaged in significantly more Active Listening at T2 (p = .01) and at T3 (p = .02) as compared to the amount of Active Listening at T1of mothers with no depression history.

Figures 2a – 2c.

Changes in Mothers’ Positive Parenting Behaviors over Time as a Function of Intervention Condition. Results of Tests of the Intervention Condition by Time Interactions

Smiling/Laughing.

The repeated measures ANOVA for Smiling/Laughing at T1 compared to T2 yielded a significant interaction, F (1, 53) = 6.13, p = .02, indicating that mothers in the parenting intervention showed a significant increase in Smiling/Laughing from T1 to T2 as compared to mothers in the control intervention condition (see Figure 2b).

Praise.

The repeated measures ANOVA for Praise was significant from T1 to T2, F (1, 53) = 8.19, p = .01; and T1 to T3, F (1, 53) = 13.06, p = .00 (see Figure 2c). Increases in Praise from T1 to T2 and T1 to T3 were significantly greater for mothers in the parenting than control intervention.

Intervention Effects on Children’s Positive Affect (Hypothesis 3).

The two-way Repeated Measures ANOVAs revealed that the time by intervention condition interaction was not significant; thus, there was not a significant difference between the parenting and control interventions on children’s PA when analyzing across all four time points. Separate regression analyses testing the effect of intervention condition on children’s PA at each time point, controlling for children’s PA at T1, however, yielded a significant effect of the intervention on children’s PA at T2, controlling for child PA at T1, B = .091, t = 2.069, p = .043, 95.0% CI (.003 - .180). The effects of intervention condition on child PA at T3 and T4, controlling for child PA at T1 were not significant. Thus, there was a significant effect of the parenting program on children’ PA immediately after the intervention.

Intervention related increases in mothers’ positive parenting over time will predict increases in children’s PA (Hypothesis 4).

We tested whether the strength of the association between changes in mothers’ parenting behaviors and changes in children’s PA differed by intervention condition. The interactions between intervention condition and changes in parenting predicting additional variance in children’s PA at T2, over and above the main effects of parenting or the intervention, were not significant (see Table S2 in the online supplemental materials). Thus, the strength of the associations between changes in parenting and changes in child PA did not differ significantly as a function of the intervention at the immediate post-intervention observation.

Regression models revealed significant main effects of two positive parenting behaviors on children’s PA at T2, over and above mothers’ positive parenting and children’s PA at T1, regardless of intervention condition. Mothers’ Active Listening at T2 accounted for unique variance in child PA at T2, over and above mothers’ T1 Active Listening and T1 child PA, ΔF(1,58) = 5.29, p = .03, f² = .31, indicating that increases in mothers’ Active Listening significantly predicted increases in children’s PA, β = .29, F = 6.04, p = .00. Similarly, mothers’ Smiling/Laughing at T2 accounted for unique variance in T2 child PA over and above mothers’ T1 Smiling/ Laughing and T1 child PA, ΔF(1,58) = 14.60, p = .00, f2 = .48, indicating that increases in mothers’ Smiling/Laughing significantly predicted increases in child PA, β = .48, F=9.35, p = .00. Tests of the relation between the other parenting behaviors at T2 and T2 child PA, controlling for the parenting variable and child PA at T1 were not significant.

Although our focus was on the immediate effects following the intervention relative to baseline (T1 to T2), we also explored whether intervention effects occurred following the brief intervention refresher one week later (T3), over and above the main effects of parenting or the intervention. Like at T2, we found that mothers’ Active Listening and Smiling/Laughing at T3 accounted for unique variance in T3 child PA over and above mothers’ T1 Smiling/ Laughing and T1 child PA. Results of the analyses of the interactions as well as main effects of parenting on children’s PA at T3, controlling for PA at T1 are presented in the supplementary materials.

Contingency of Positive Parenting with Children’s Positive Affect (Hypothesis 5).

A two-way Repeated Measures ANOVA tested whether children of mothers in the positive parenting intervention, relative to children of mothers in the control intervention, would show more contingent responses to their mothers’ positive parenting across time. The distribution of transitional probability values had one extreme outlier that fell beyond 1.5 times the interquartile range above the third quartile, which we excluded from these analyses. Mauchly’s test indicated that the assumption of sphericity had not been violated. Results of the ANOVA showed no significant time by intervention condition interactions, indicating that the extent to which children in the positive parenting intervention showed contingent responses to their mothers’ positive parenting did not significantly differ as compared to those in the control intervention and over time. Thus, there was no evidence that the intervention increased the rate of children’s moment-by-moment positive affect contingent on mothers’ positive parenting.

Discussion

The current study demonstrated that in a sample of mothers with a history of depression, a brief positive parenting intervention was more effective than an active control intervention in enhancing mothers’ use of positive parenting behaviors and increasing their children’s positive affect. These findings are consistent with other studies that have shown positive effects of parenting interventions on parent and child outcomes in families with a depressed parent, even when for some parents their depression was in the past (e.g., Compas et al., 2010). Building upon the earlier studies, we showed changes in specific positive parenting behaviors for mothers receiving the parenting intervention as compared to mothers in the control intervention. The positive parenting intervention increased mothers’ active listening, smiling/laughing, and praise (verbal and non-verbal) during a laboratory interaction with their child, in contrast to mothers in the control intervention. These gains were observed immediately post-intervention, a week later for active listening and praise, and even three weeks later for praise. Moreover, the post-intervention level of active listening among mothers in the positive parenting intervention was greater than levels seen among mothers with no history of depression and no intervention. Thus, a brief intervention targeting positive parenting can produce observable changes in parenting behaviors of mothers with a history of depression and in their children’s PA.

Consistent with evidence from more extensive positive parenting training programs with parents with a history of depression (e.g., Compas et al., 2010), our findings showed that specific parenting behaviors are modifiable and promising intervention targets for mothers with a history of depression (Goodman & Garber, 2017). Of note was the finding that after the positive parenting intervention, mothers with a history of depression showed even higher levels of active listening than mothers with no history of depression. Changes in mothers’ praise were the most persistent over time, lasting for as long as three weeks post intervention. Thus, mothers with a history of depression might be particularly responsive to an intervention that enhances their expressions of praise to their children, which may have resonated with them and may have been relatively easy for them to implement after receiving instructions.

Low positive affect is a vulnerability for depression in offspring of depressed parents (Olino et al., 2011). The current study demonstrated that relatively simple changes in mothers’ positive parenting produced brief changes in positive affect in offspring of mothers with a history of depression. Compared to children of mothers in the control intervention, children of mothers in the parenting intervention showed significantly more positive affect-immediately post-intervention (T2).

These gains in children’s positive affect, however, were short-lived and were not observed one (T3) or three weeks (T4) after the first session. Moreover, contrary to our hypothesis, examination of transitional probabilities revealed that an increase in positive parenting did not produce an in-the-moment increase in children’s positive affect. Although enhancing mothers’ positive parenting increased children’s positive affect across the session, a longer or more frequent intervention might be needed for mothers’ greater positive parenting to produce sustained effects on children’s positive affect specific to participants in the parenting intervention.

Perhaps positive parenting has a cumulative effect on children’s positive affect, such that more sustained and recurrent exposures to mothers’ positive parenting may be needed for it to elicit moment-to-moment changes in children’s positive affect. Mothers may need to engage in positive parenting consistently and across contexts, involving the larger family, highly emotionally evocative exchanges, and for a longer period, for it to reliably evoke children’s positive affect in the moment. Future studies also might examine the role of children’s attention to, and awareness of, their mothers’ positive parenting, especially in association with intervention related increases in mothers’ parenting behaviors. Interventions might target attention biases that interfere with children processing their mothers’ positive emotions in response to them.

Interestingly, we found positive effects of the intervention on several parenting behaviors despite finding only one statistically significant difference in parenting behaviors between mothers with and without a depression history. A few explanations are possible for our not finding more significant differences in parenting behaviors of mothers with versus without a history of depression. First, a meta-analytic review of associations between depression and parenting yielded a weighted mean r = .15, even with current parental depression (Goodman et al., 2020). We might expect an even lower association for mothers with past rather than current depression. Nevertheless, in addition to the one significant finding, three other maternal behaviors showed clinically meaningful effect sizes ranging from .19 to .22; mothers with histories of depression showed less physical contact and less time smiling/ laughing, and more positive requests for change than mothers with no depression history. Second, our parenting behavior codes focused more on tapping different types of positive parenting behaviors than typical parenting behavior observational coding systems. That is, we relied heavily on codes used to test the effectiveness of parent management training interventions such as Parent-Child Interaction Therapy (Eyberg et al., 1995) and Triple P-Positive Parenting Program (Sanders, 2008).

Study Strengths, Limitations, and Future Directions

The present study had several strengths. First, we demonstrated a “proof of concept” that targeting the hypothesized mechanism of positive parenting can be effective and can increase children’s positive affect. Second, we compared the target intervention (i.e., parenting) to an active comparison program that had an ecologically valid rationale, controlled for amount of children’s positive affect. Second, we compared the target intervention (i.e., parenting) to an active comparison program that had an ecologically valid rationale, controlled for amount of contact with the coach, and had a similar structure with homework assignments. Finally, the current study used observational data and micro-level coding to measure both parenting behaviors and children’s positive affect, whereas many previous studies have relied on parent- and child-report of parenting and child outcomes (e.g., Timmer et al., 2011). Meta-analytic reviews have found that in some studies, observational measures of parenting were more strongly predictive of child outcomes than were self-reports of parenting (Slagt et al., 2016).

Observational assessment of parenting also allowed us to detect changes in specific behaviors, and for a micro-level analyses of sequential relations to child positive affect. The inclusion of micro-coded, observed child positive affect, as opposed to mothers’ or children’s report of child affect, is notable given that such an index of behavioral positive affect is robustly linked with later maladaptive outcomes in children of mothers with histories of depression as compared to other purported features of positive affect, such as hedonic PA (Forbes & Dahl, 2005; Hayden et al., 2013; Olino et al., 2014).

Study limitations provide directions for future research. First, mothers in the sample were mostly well-educated. Thus, replication of these findings is needed with mothers with more variability in their education attainment. Second, the level of current maternal depressive symptoms was low. Although women with a history of depression had significantly higher depressive symptom scores than those with no depression history, most mothers did not exceed the established cutoff for clinically significant depression at the time of the study. Future studies should test these hypotheses in samples of mothers with more severe current depression.

A third limitation concerns dissemination. In the present study, the intervention was delivered under controlled circumstances (i.e.., in university psychology departments, with clinical psychology graduate students as coaches). Future iterations of the intervention should be conducted in settings such as pediatricians’ offices or schools and implemented by paraprofessionals endogenous to the specific environment.

Finally, the mother-child interactions involved a shared play activity; mothers’ behaviors likely differ depending on the context in which the dyads interact – whether neutral or emotionally evocative (e.g., a conflict discussion task) (Eisenberg et al., 1998; Roque et al., 2012). We observed low rates of certain parenting behaviors (e.g., positive requests for change, and criticism); perhaps a more structured or emotionally evocative task would elicit more such parenting behaviors and, thus, yield even greater change in positive parenting and child positive affect as a function of the intervention. Further, we only assessed children’s positive affect in the context of a mother-child interaction. Future studies might also include examination of children’s positive affect outside of this context, consistent with consideration of children’s positive affect as a protective factor.

In summary, this study demonstrated the modifiability of parenting behaviors using a brief, structured intervention with mothers with a history of depression and showed a link between positive parenting and children’s positive affect. Mothers in the positive parenting intervention increased their use of active listening, smiling/laughing, and praise during laboratory interactions with their child. Additionally, their children showed increases in observed positive affect, compared to children of mothers assigned to the control intervention. Although further research and replication will be important for guiding future intervention efforts, these findings suggest that certain positive parenting behaviors can be effectively enhanced with a brief intervention and can lead to a short-term increase in children’s positive affect (Fung et al., 2020).