Typologies of Daily Relationship Quality in Marital and Parent–Child Subsystems: Implications for Child Adjustment

Abstract

The ways in which marital relationships (MR) affect parent–child relationships (PCR) vary from day-to-day and differ from one family to another. The day-to-day fluctuations of MR–PCR associations and the between-family differences, however, have been largely overlooked in the literature. Using daily diary data from 152 mother–father couples (with a 7- to 17-year-old child) across three consecutive years, we identified typologies based on parents’ daily relational dynamics and examined the association between family typologies and child adjustment. State space grids of daily relationship quality (i.e., MR and PCR) were constructed for each parent at each wave of assessment. Grid-sequence analysis was subsequently conducted to identify multiple family typologies, including cohesive, fluctuating cohesive, spillover, compensatory, and moderate relationship typologies. Some typologies changed over the years and differed for fathers and mothers (e.g., the compensatory typology). Child adjustment problems were the highest in children from families with poor MR and PCR quality (i.e., the spillover typology). Overall, our results evince the heterogeneous nature of family relationship dynamics and the developmental implications of these typologies. For practitioners and therapists who work with families, our findings highlight the value of improving marital and parent–child relationship quality in promoting positive child outcomes.

Children flourish in the context of positive family relationships. Positive interconnections between family members, including marital (MR) and parent–child (PCR) relationships, have long-term benefits on children’s socioemotional outcomes (Buehler & Gerard, 2002; Cummings & Davies, 2010; Easterbrooks & Emde, 1988). However, studies devoted to this area of investigation have adopted a relatively static view of family dynamics. Understanding of the day-to-day changes and fluctuations in the association between MR and PCR, as well as the between-family differences in these associations, is limited. Based on longitudinal daily diary assessments, the present study provides a dynamic, developmental picture of these family relationships, as well as its implications for child adjustment.

How MR is associated with PCR differs across families, and these associations have been commonly addressed by two competing hypotheses: the spillover and compensatory hypotheses. According to the spillover hypothesis, MR problems are linked with PCR difficulties (Almeida et al., 1999; Kaczynski et al., 2006; Krishnakumar & Buehler, 2000). In contrast, the compensatory hypothesis holds that parents become more engaged and supportive of the child in the context of marital difficulties (Belsky et al., 1991; Brody et al., 1986). Despite the many empirical efforts for a refined understanding of these associations (Gao & Cummings, 2019; Ponnet et al., 2013; Sherrill et al., 2017), no consensus has been reached. One way to advance this research is to investigate how MR and PCR interact day-to-day and to articulate profiles of family dynamics.

Some parents may have a difficult time with their child on the same day when they encounter marital distress with their spouse, possibly as a way to deflect marital tensions (Minuchin et al., 1978). Through day-to-day transactions between MR and PCR, a pattern of low MR–low PCR may constitute a pattern of family relationship dynamics. This pattern is consistent with the spillover hypothesis whereby emotion negativity is transmitted from one family subsystem to another (Almeida et al., 1999; Kaczynski et al., 2006; Krishnakumar & Buehler, 2000). Conversely, other families may exhibit an opposite pattern in which parents become more supportive of the child on days with marital difficulties, in align with a compensatory hypothesis (Belsky et al., 1991). Taken together, these and other patterns may reflect diverse forms of interrelatedness between MR and PCR that may exist in families’ daily lives. A person-oriented approach is a promising avenue for identifying family typologies that emerge from daily relationship dynamics.

Person-oriented approaches cluster individuals by focusing on specific patterns of interrelatedness between the marital and parent–child subsystems. Family typologies have been identified based on differences in functioning across different family subsystems (e.g., Belsky & Fearon, 2004; Malinen et al., 2010; Sturge-Apple et al., 2010, 2014). For example, Sturge-Apple and colleagues (2014) detected several family types in a sample of families with toddlers: spillover, adequate-functioning, and compartmentalization. Families in the spillover profile were characterized by the coupling of high levels of marital hostility and harsh parenting. The adequate-functioning group exhibited low levels of marital violence and conflict, punitive parenting, and average levels of parental warmth. Families in the compartmentalization group exhibited high levels of marital conflict and anger, while displaying high levels of warmth and low levels of punitive parenting toward the child. Similar typologies were identified in families with kindergarteners in another study (Sturge-Apple et al., 2010), including cohesive, enmeshed, and disengaged families, which are similar to the adequate-functioning, compartmentalization, and spillover profile, respectively, found in Sturge-Apple et al. (2014). Family typologies, however, are by no means constrained to three. For example, based on marital and parenting functioning during infancy, toddler, and/or preschool years of families in a national study, Belsky and Fearon (2004) identified five patterns, including good marriage/good parenting, moderate marriage/moderate parenting, poor marriage/good parenting, good marriage/poor parenting, and poor marriage/poor parenting.

These findings from family-oriented studies have typically been based on observations or questionnaires during a single laboratory visit. The day-to-day emotional exchanges and influences between marital and parent–child subsystems, which provide a more extensive delineation of family processes, could not be captured by these studies. Additionally, the achieved typologies may only apply to the characterization of family functioning in broad strokes. Typologies based on the ecologically valid context of the home may be needed for understanding the intricate and complex dynamics of daily family transactions. Toward this goal, we applied the daily diary approach to the identification of family typologies based on parents’ daily perceptions of MR and PCR quality in the home. Daily diaries provide information about not only the content (i.e., the relative level compared to other families) but also the structure (i.e., how the MR–PCR association varies from day to day) of daily family functioning, which may not otherwise be detected with traditional designs (Bolger et al., 2003; Repetti et al., 2015). How stable (or variable) a family is in (or across) certain emotional states reflects another important way in which families may differ from each other. Therefore, applying the typological approach to daily diary data allows for the examination of inter-family differences in intra-family dynamics (Brinberg et al., 2017, 2018; Lichtwarck-Aschoff et al., 2009).

Maternal and Paternal Family Typologies

Studies based on variable-oriented approaches have identified gender differences in the MR–PCR link (e.g., Davies et al., 2009; Gao & Cummings, 2019; Stroud et al., 2011). Although fathers and mothers from the same family may experience marital and parenting ups-and-downs together, they may also exhibit different patterns of interplay between the marriage and parenting domains. For example, a profile indicating a compensatory process may be more common for mothers than fathers. One hypothesis is that as the emotional gatekeeper of the family, mothers regard a poor mother–child relationship as a personal failure (De Luccie, 1995; Pleck, 1983) such that more effort is devoted to connecting with the child and to compensate for marital problems. In fact, empirical support is found for the compensatory MR–PCR link more consistently for mothers than for fathers (Belsky et al., 1991; Brody et al., 1986; Engfer, 1988). However, only one empirical study (Malinen et al., 2010) to our knowledge has investigated family typologies for mothers and fathers simultaneously. Accordingly, we seek to identify both maternal and paternal family typologies based on daily reported MR–PCR relations.

Implications for Child Adjustment

The present study also seeks to break new ground by examining how different configurations of MR and PCR interrelations relate to child development. Studies adopting a typological approach have identified patterns of MR–PCR linkages to children’s developmental sequelae, including externalizing and internalizing behaviors, social skills, academic achievement, and physiological functioning (e.g., Belsky & Fearon, 2004; Johnson, 2003; Lamela et al., 2018; Lindblom et al., 2017; Sturge-Apple et al., 2014). In general, children from families that are characterized by poor relationship quality in both marital and parent–child subsystems are most likely to exhibit fewer competencies and more problems in various developmental domains.

Family typologies may change over time. Parents may become better at separating their role as a spouse from that as a parent, such that one’s parenting is less likely to be compromised by negative experiences in the marital subsystem (Gao & Cummings, 2019). Accordingly, a family identified as a spillover family at one point in time may change into a compensatory family at a later time point. As another example, an adequate-functioning family characterized by relatively high levels of MR and PCR functioning may become encumbered with unexpected life stressors (e.g., unemployment, health problem) over time and thus change to another family profile. Consistent with a developmental psychopathology perspective on family process (Cummings et al., 2000), the changes in family environment over time merit consideration in accounting for children’s functioning. However, only one study has tested how changes in family typology were associated with changes in children’s functioning: Johnson (2003) found that children’s externalizing behaviors decreased from Grade 1 to Grade 4, when their family type changed from not cohesive type to cohesive type. It remains to be examined how family typologies and changes in typologies inform children’s adjustment. Investigation of this question has practical implications: for example, family types that confer the greatest risk for child developmental problems may be prioritized for receiving intervention.

The Present Study

The present study had two aims. First, we aimed to investigate whether multiple profiles of the MR–PCR links could be identified using daily diary reports of emotional quality in marital and parent–child subsystems. Grid-sequence analysis was applied to data obtained with three bursts of 15-day daily diaries. We were interested in examining whether the typologies described in the literature would be replicated in the current daily diary study, including the spillover and compensatory typologies, and whether additional typologies would be identified. This study was not preregistered.

Second, we aimed to examine the developmental implications of each family typology for child adjustment outcomes. Specifically, we explored whether family topologies identified at each wave would be related to children’s adjustment assessed at the same time. Furthermore, we explored whether family typologies identified at Wave 1 would be related to children’s adjustment outcomes at Wave 3. Lastly, we examined whether changes in typologies would be related to changes in children’s adjustment across the two-year assessment period.

Method

Participants

Couples and their child participated in a three-year prospective study in a small Midwestern city in the U.S. Participants were recruited from the community through flyers, newspaper, TV, and radio advertisements, community events, and letters distributed to local schools and neighborhood residents. Eligibility criteria included: the couples had to be living together for at least two years, a 7- to 17-year-old child had to live in the home for majority of the time. The sample includes 299 couples at Wave 1 (W1), from which 250 remained at Wave 2 (W2) and 248 were retained at Wave 3 (W3).

Of the 299 W1 families, 237 agreed to participate the daily diary portion in addition to the laboratory portion of the study. Because one goal of the present study was to examine changes of family type, it was essential that the analytical sample had participated in all three assessment occasions. Therefore, families in which at least one parent missed one diary assessment period (out of three waves) were removed from final analyses. The final analytical sample consisted of 152 father–mother dyads. The reduction in sample size (from 237 to 152 families) reflected the strict requirements for complete data by both mothers and fathers across the many times points assessed to implement our analytic approach. However, the sample for analyses did not differ from the rest of the full sample on most demographic variables measured at W1, including child gender, family income, parents’ age, education, marital status, marriage length, and marital satisfaction (all ps > .07). The excluded sample had more parents from racial minority groups (i.e., non-White; χ²(1) = 14.84, p < .01 for mothers, and χ²(1) = 10.57,p <.01for fathers) than the final analytic sample We also compared the analytic sample (N = 152) and the sample who completed daily diaries (N = 237). The two samples did not differ in any of the key study variables (i.e., averaged levels of marital relationship, averaged levels of parent-child relationship, and child adjustment) in any of the three assessment waves (all ps > .10).

Of the 152 couples, the majority were married for an average of 13 years (SD = 6.68), and one couple was living together but not married. Fathers’ mean age was 40.50 years (SD = 6.84) and mothers’ mean age was 37.99 years (SD = 6.17). Most mothers were White (94.1%), 3.3% were Black, 1.3% were Hispanic/Latina, and two mothers did not report their ethnicity or race. With regard to fathers, 92.1% were White, 4.6% were Black, 2.6% were Hispanic/Latino, and one father was mixed race. The families were primarily middle class. Nine families had a household income less than $25,000 per year, 8 between $10,001 and $25,000, 32 between $25,001 and $40,000, 75 between $40,001 and $65,000, 19 between $65,001 and $80,000, and 18 above $80,001. Children were 7 to 17 years old (M = 11.10 years, SD = 2.31; 53.3% girls).

Procedure

Parents and their child visited the laboratory at each of the three waves spaced one year apart; W1 occurred during 1999–2000. During their visits, parents completed questionnaires to report on marital satisfaction, marriage length, and child adjustment. In addition, they completed a daily paper diary entry for 15 days, beginning the next day following the laboratory visits¹. All families completed the diary at the end of each day and parents were instructed not to discuss their answers with each other. Fathers and mothers mailed back their diaries separately when 15 entries were completed². Each family received monetary compensation for their participation. The study was approved by University of Notre Dame’s institutional review board, and parent consent and assent from children were obtained.

Measures

Relationship quality.

We used daily diaries to assess participants’ MR and PCR. On a 0 (negative) to 9 (positive) scale, both fathers and mothers responded each day to the question “what is the emotional quality of your relationship with your spouse that day?” On the same scale, they also rated the overall emotional quality of their relationship with their child that day. Across all waves and both relationship quality, over 70% (range 70% to 91.3%) of participants completed all 15 diaries. For mothers, the average scores of MR and PCR were 6.94 (SD = 1.12) and 7.27 (SD = 1.09), 6.95 (SD = 1.29) and 7.25 (SD = 1.26), 6.93 (SD = 1.20) and 7.33 (SD = 1.12) respectively at W1, W2, and W3. Correspondingly, the average scores for fathers’ MR and PCR were 6.73 (SD = 1.26) and 6.82 (SD = 1.27), 6.82 (SD = 1.30) and 6.93 (SD = 1.28), 6.69 (SD = 1.35) and 6.90 (SD = 1.27) respectively at each wave.

Child adjustment.

Mothers and fathers completed the internalizing and externalizing problems subscales of the Child Behavior Checklist (CBCL; Achenbach, 1991) at all three measurement occasions. The internalizing problems subscale (30 items) reflects children’s anxious, withdrawn, and depressive symptoms. The externalizing problems subscale (33 items) reflects children’s aggressive and delinquent behaviors. Parents rated children’s behaviors during the previous six months on a 3-point Likert scale (0 – not true, 1 – somewhat or sometimes true, 2 – very true or often true). Scores were summed and averaged across parents to create a parent-composite score of internalizing (Cronbach’s α_s ranged from .73 to .87, W1 to W3) and externalizing problems (α_s ranged from .75 to .89, W1 to W3).

Analytical plans

Two phases of data analyses were conducted. The first phase involved applying grid-sequence analysis (Brinberg et al., 2017, 2018) to the identification of family typologies. Here we provide a brief summary of the analytic steps involved in this approach. Readers interested in technical details beyond this brief description are referred to Brinberg and colleagues’ work (Brinberg et al., 2017, 2018). First, a state space grid (SSG) was created for each parent. Specifically, father-report MR was assigned to x-axis and father-report PCR was assigned to y-axis to construct an SSG for each father. Mothers’ SSGs were constructed in the same way. Each grid was further separated into “cells”. As shown in Figure 1 (produced using the base, ggplot2, and reshape packages in R; R Core Team, 2015; Wickham, 2007, 2009), the distribution of both MR (on the x-axis) and PCR (on the y-axis) were separated into 3 sections: 0–5, 6–7, and 8–9) with 2 cut points (i.e., rating of MR or PCR = 5.5 and 7.5). The two cut points were chosen after considering the distribution and interpretability tradeoffs (that is, more cut points result in more cells therefore harder to interpret).³ For each parent, the trajectory (i.e., sequence of MR–PCR states) at each wave was drawn as it proceeded in a temporal sequence over the 15 diary-reporting days, capturing all emotional states between marital and parent–child subsystems experienced by this parent. Variability of a family’s MR–PCR states across days was conceptualized as structural characteristics of each trajectory, including transitions and cell range. Specifically, more transitions (i.e., number of movements between cells on an SSG) and wider cell range (i.e., the count of the total number of unique cells visited by the trajectory) indicated high variability of a system.

Figure 1.

Mothers’ Daily Reports of MR and PCR Mapped onto the SSG in the Order of the Daily Occurrence

Note. ggplot2 and reshape packages in R (R Core Team, 2015; Wickham, 2007, 2016) were used to obtain these images. Mother A (in Figure 1a) reported consistent high levels of both MR and PCR, whereas Mother B (in Figure 1b) in general reported good MR and PCR but with higher variability. Extracted sequences of each mother’s MR–PCR states are presented at the bottom of their SSG, in the order of their temporal occurrence across the 15 assessment days (grey cells indicate missing data). Additionally, the two variability measures for each mother are presented next to each SSG. For example, Mother A visited 2 unique cells (i.e., cell B and cell C; cell range = 2) and moved twice between cells (i.e., transitions = 2).

Next, each cell of the SSG was labeled with a letter, as shown in Figure 1. The letters (A through I) were simply categorical names assigned to each cell and were not meaningful. A sequence of each parent’s MR–PCR states across the 15 diary-reporting days can be represented by a string of letters of the visited cells in a temporal order. See Figure 1 for two examples of extracted sequences. The order of colors represented the order of cells visited, and grey cells were inserted whenever missing. Dissimilarity between two sequences was quantified as the minimum cost of transforming one sequence to another via any combination of three actions: insertion, deletion, and substitution. Sequences that were more similar to one another had lower transformation cost. Following typical procedures and previous research, we set insertion and deletion costs equal to 1 (Brinberg et al., 2017; MacIndoe & Abbott, 2004) and determined substitution cost using Manhattan (city-block) distance (travels are only allowed along the sides of cells). For example, the cost of substitute A with A was 0, A with B was 1, A with C was 2, and A with I was 4, and so forth. The weight of substitution cost (ranges from 0 to 4) aligned with the intuition that larger moves in the emotional SSG required much more energy (i.e., more costly) than smaller moves. We first established a 9 (Number of Cells) × 9 (Number of Cells) substitution cost matrix. We then added an additional row and column to accommodate missingness, such that the cost of substitution to or from missingness to any state was half the highest substitution cost (i.e., cost of substitution cost of missingness is 2). Taken together, the dimension of the resulting substitution cost matrix was 10 ×10 (Number of Cells + 1 × Number of Cells +1). After establishing the cost matrix, we found the minimum transformation cost using the three aforementioned actions in an optimal matching algorithm (Needleman & Wunsch, 1970) for every pair of sequences. A 152 ×152 (there were 152 families in the present study) dissimilarity matrix was constructed, with each element indicating the cost of transforming any given sequence into another. TraMineR and TraMineRextras packages in R (Gabadinho et al., 2011; Studer & Ritschard, 2016) were used to construct the dissimilarity matrix.

Finally, cluster analytic techniques were applied to the dissimilarity matrix to identify typologies for fathers and mothers, using the agglomerative hierarchical cluster analysis in R’s cluster package (Maechler et al., 2016). The clustering procedure involved representing the dissimilarity a dendrogram, which was further explained in Figure S1 in the supplemental material. Following Brinberg and colleagues’ (2017, 2018) suggestion, the number of clusters selected were based on two criteria: 1) dissimilarity between subgroups was large, such that each subgroup represented distinct patterns of relationship dynamics, and 2) the size of each subgroup was sufficient, and the number of subgroups was interpretable. We named each subgroup by evaluating their average levels of MR and PCR and variability measures around the SSG.

In the second phase of data analysis, we examined whether the identified typologies of MR–PCR dynamics were concurrently correlated with and longitudinally predictive of child adjustment, based on a series of multivariate analyses of covariance (MANCOVA). Child age and child biological sex were included as covariates. Additionally, we examined how families transitioned from one family type to another and explored how these changes in family typology were related to changes in child adjustment. Theoretically, there would be m × (m − 1) ways of changing typology from W1 to W3 (m is the number of typologies identified in each wave). However, large values of m would result in a great number of transition patterns between the m family typologies, which may not be feasible for the current study due to its sample size. Therefore, we regroup families as either adequate-functioning or not adequate-functioning at each time point (m = 2) based on the family typology identified via grid-sequence analysis. A mixed 2 (between-family type at W1) × 2 (between-family type at W3) × 2 within-adjustment scores at W1 and W3) ANOVA was used to test whether changes in children’s adjustment over 2 years were systematically related to changes in family typology during the same time frame.

Results

Identifying Typologies of Daily Relationship Quality

A typology of four-cluster intra-family relationship dynamics was indicated for both fathers and mothers at Wave 1. Figure 2 presents the identified groups of sequences for mothers. Groups for fathers can be found in Figure S2 in supplemental material. A family from each group was selected and its SSG was presented next to the grouped sequence to depict the prototypical relationship dynamics of that group. In the top panel of Table 1, we summarized characteristics (i.e., levels of emotional quality in MR and PCR) of each group.

Figure 2.

Profiles of Intra-family Relationship Dynamics for Mothers at Each Assessment Occasion (i.e., W1 to W3), with an Accompanying Exemplar Mother’s SSG.

Note. In each profile, all trajectories are grouped together with each row representing one family (e.g., there are 33 rows for Type 1 Cohesive mothers at W1, and each row represents one mother’s sequence of MR–PCR states).

Table 1.

Characterization of the Five Cluster Groups with Different MR–PCR Dynamics for Mothers and Fathers at Each Wave and the Mean Differences among Cluster Groups

Typology Name					Cohesive	Spillover	Fluctuating Cohesive	Moderate Relationship	Compensatory
				MR	High MR	Low MR	High MR	Moderate MR	Low MR
Characteristics		Relationship Quality		PCR	High PCR	Low PCR	High PCR	Moderate PCR	High PCR
Regroup into two groups based on overall relationship quality					Adequate functioning	Less than adequate functioning	Adequate functioning	Less than adequate functioning	Less than adequate functioning
					Type 1	Type 2	Type 3	Type 4	Type 5
		df	F	p	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Wave 1	Group Size (n)				33	12	53	45
Mother	Relationship Quality
	MR	3,139	75.22	< .001	8.28 (.47) 2,3,4	5.15(.70) 1,3,4	6.96 (.62) 1,2,4	6.50 (.87) 1,2,3	–
	PCR	3,139	86.51	< .001	8.41 (.49) 2,3,4	5.28(.81) 1,3,4	7.51(.65) 1,2,4	6.70(.70) 1,2,3	–
	Variability Measures
	Transitions	3,139	40.00	< .001	4.64 (2.38) 2,3,4	8.08 (3.50) 1	9.62 (2.00) 1	9.98 (2.33) 1	–
	Cell range	3,139	26.84	< .001	2.97 (1.13) 2,3,4	4.75 (1.71) 1	5.25 (1.40) 1	5.22 (1.00) 1	–
Wave 1	Group Size (n)				35	44	32	27
Father	Relationship Quality
	MR	3,134	88.27	< .001	8.14(.53) 2,3,4	5.59(.98) 1,3,4	7.27(.64) 1,2,4	6.61(.42) 1,2,3	–
	PCR	3,134	114.1	< .001	8.34(.53) 2,3,4	5.78(.88) 1,3,4	7.38(.41) 1,2,4	6.37(.55) 1,2,3	–
	Variability Measures
	Transitions	3,134	17.72	< .001	5.37 (3.03) 2,3,4	9.55 (2.85) 1	8.81 (2.19) 1	9.48 (3.01) 1	–
	Cell range	3,134	14.53	< .001	3.06 (1.39) 2,3,4	4.84 (1.26) 1	4.63 (1.21) 1	4.85 (1.54) 1	–
Wave 2	Group Size (n)				72	24	40		11
Mother	Relationship Quality
	MR	3,143	69.52	< .001	7.65(.86) 2,4	5.56(.84) 1,3,4	7.25(.64) 2,4	–	4.67(1.14) 1,2,3
	PCR	3,143	112.7	< .001	8.19(.51) 2,3,4	5.29(1.09) 1,3,4	6.88(.58) 1,2	–	7.06(1.00) 1,2
	Variability Measures
	Transitions	3,143	6.64	< .001	6.85 (3.27) 3,4	8.25 (3.08)	9.23 (2.58) 1	–	9.55 (2.73) 1
	Cell range	3,143	9.94	< .001	3.74 (1.42) 3,4	4.38 (1.25)	5.15 (1.55) 1	–	5.09 (.94) 1
Wave 2	Group Size (n)				44	25	46	30
Father	Relationship Quality
	MR	3,141	85.26	< .001	8.01(.52) 2,3,4	4.97(1.04) 1,3,4	7.01(.75) 1,2,4	6.53(.85) 1,2,3
	PCR	3,141	203.8	< .001	8.27(.49) 2,3,4	5.06(.66) 1,3,4	7.27(.52) 1,2,4	6.18(.59) 1,2,3
	Variability Measures
	Transitions	3,141	8.74	< .001	5.32 (2.63) 2,3,4	7.48 (4.14) 1,3	9.54 (1.96) 1,2	9.20 (2.85) 1	–
	Cell range	3,141	8.95	< .001	3.11 (1.06) 3,4	3.76 (1.67) 3,4	4.98 (1.22) 1,2	4.77 (1.43) 1,2	–
Wave 3	Group Size (n)				74	14	48		11
Mother	Relationship Quality
	MR	3,143	89.09	< .001	7.75(.69) 2,3,4	5.35(.62) 1,3	6.61(.57) 1,2,4	–	4.79(1.43) 1,3
	PCR	3,143	103.4	< .001	8.09(.61) 2,3	5.24(.75) 1,3,4	6.72(.59) 1,2,4	–	7.68(.76) 2,3
	Variability Measures
	Transitions	3,143	15.69	< .001	6.61 (2.90) 3,4	8.50 (3.01)	9.98 (2.63) 1	–	9.27 (1.19) 1
	Cell range	3,143	9.91	< .001	3.76 (1.30)	4.50 (1.40)	5.13 (1.53)	–	4.36 (.81)
	Group Size (n)				49	28	40	28
Wave 3	Relationship Quality
Father	MR	3,141	115.0	<.001	7.97(.64) 2,3,4	4.79(.74) 1,3,4	6.66(.88) 1,2	6.52(.59) 1,2
	PCR	3,141	89.87	<.001	8.10(.62) 2,3,4	5.41(.78) 1,3,4	7.13(.77) 1,2,4	6.20(.87) 1,2,3
	Variability Measures
	Transitions	3,141	20.21	<.001	5.39 (2.71) 2,3,4	7.96 (3.33) 1,3	10.07 (2.72) 1,2	8.75 (3.13) 1	–
	Cell range	3,141	16.09	<.001	3.27 (1.20) 3,4	3.71 (1.38) 3	5.20 (1.51) 1,2	4.39 (1.37) 1	–

Note. Top panel: names and characteristics of the five identified cluster groups. Bottom panel: Means and standard deviations of the relationship quality for each identified cluster, as well as the mean differences among all clusters. F and p values are from the overall MANOVA tests with degrees of freedom.

Significant pairwise comparisons tested with Tukey’s honest significance test correction were indicated by numeric subscripts: 1,2,3,4,5. Numbers 1–5 indicate family type 1–5. As an example, a subscript of 2 represents that the current family type is significantly different from families of Type 2 (i.e., Spillover). If a family type was found to exhibit significant difference from other types (as indicated by subscripts), the values and subscripts were also bolded. On the contrary, if the value of one family type was not significantly different than any of the other types, no subscripts were added, and these values were not bolded.

Type 1, cohesive families, included parents who reported high levels of emotional quality both with their child and with their partner. By comparison, spillover families (Type 2) reported low levels of both MR and PCR every day. Fluctuating cohesive (Type 3) families were similar to cohesive families in that they reported high levels of emotional quality, but differed in variability over time; that is, parents in the fluctuating cohesive subgroup were more variable in their reported daily relationship quality than parents in the cohesive group⁴. The differences among these subgroups were also indicated post hoc by significant differences in MR and PCR, as shown in the bottom panel of Table 1. For example, Type 1 parents showed significantly higher relationship quality (both MR and PCR, p_s <.001) than other family types at W1. Type 2 parents had the lowest levels of MR and PCR compared to other family types (p_s <.001).

Interestingly, although the same four-cluster typology was found for fathers at Wave 2 and 3, a new cluster was found for mothers (see Figure 2). Specifically, the moderate-relationship group (Type 4) from W1 was not identified, while a new cluster was present and was given the name compensatory, based on the characteristics of their relationship dynamics (i.e., Type 5). These mothers reported high levels of relationship quality with their child and low quality with their spouse on the same day.

The identified clusters at each wave did not differ as a function of child age or child gender (p_S > .05) for either fathers or mothers, with only one exception: at W2, the four types of paternal typology differed significantly in child age, F (3,141) = 3.83, p < .05. Post hoc analysis with Tukey’s honest significance test correction indicated that fathers in the fluctuating cohesive (Type 3) group tended to have older children than fathers in the cohesive (Type 1) group.

Fair agreement was found for father’s and mother’s affiliated clusters, Cohen’s κ = .236, p <.001 at W1, κ = .127, p <.01 at W2, and κ = .179, p <.001 at W3. Fathers and mothers from the same family typically had similar patterns of relationship dynamics. Although κ_s were different from 0, the strengths of agreement were not strong (κ > .6 is generally considered to indicate a substantial agreement, Landis & Koch, 1977).

Changes in family typology.

Moderate levels of consistency in family type across time were indicated for mothers and fathers. Most mothers (n = 25, 75.8%) and fathers (n = 22, 62.9%) who were clustered into the cohesive profile at W1 continued to be clustered into cohesive type at W3. Other mothers (n = 4) and fathers (n = 10) who were in the cohesive group at W1 were later classified as in the fluctuating cohesive group at W3. Additionally, most mothers (n = 28, 84.9%) and many fathers (n = 16, 25.7%) in the fluctuating cohesive group at W1 changed their group membership to cohesive at W3, although some (18 mothers and 9 fathers) continued to be identified as fluctuating cohesive at W3. Mothers clustered into spillover subgroup, however, exhibited little stability over the years: only two mothers (out of 11) remained in the spillover profile at W3. Moreover, only 19 fathers (out of 44) remained in the spillover profile at W3. More detailed breakdown of how parents’ membership in a cluster change from W1 to W3 can be found in supplemental material (Table S2).

Associations between Family Typologies and Child Adjustment

Associations between family typologies and child adjustment at all waves can be found in Table S1 in the supplemental material. Here we summarize key findings. At W1, parents’ report of child externalizing problems differed significantly across the four types of maternal typology, F (3,137) = 3.55, p < .05. Children from cohesive families (M_T1 = 5.50) had lower externalizing problems than children from spillover (M_T2 = 10.21) or moderate-relationship (M_T4 = 8.73) families. Similar results emerged for relations between fathers’ typology and child externalizing problems at W1, F (3,132) = 2.67, p < .05. Fathers from the cohesive group had children with fewer externalizing problems (M_T1 = 5.93) than fathers from the spillover (M_T2 = 8.63) and moderate-relationship (M_T3 = 8.54) groups. At W2, typologies were not related to children’s internalizing or externalizing problems. At W3, mothers’ typologies were related to children’s externalizing problems, F (3,141) = 2.90, p < .05. Specifically, children of mothers characterized by the spillover (M_T2 = 9.68) profile had higher levels of externalizing problems than children of mothers in cohesive (M_T1 = 5.55) or fluctuating cohesive (M_T3 = 6.69) profiles.

As for longitudinal predictions of W1 typology for W3 outcomes, MANCOVA results showed that W1 typology did not predict any of the adjustment outcomes assessed at W3, controlling for child age, gender, and adjustment at W1. We also tested whether W2 typology would predict W3 child adjustment, controlling for W2 adjustment, age, and gender. No significant effects were found.

Changes in family typology and child adjustment.

Following the analytical plan, we regrouped families into two groups at each assessment wave: families who exhibited adequate-functioning relationship dynamics (i.e., consisted of cohesive and fluctuating cohesive families) and families who exhibited less than adequate functioning (i.e., consisted of spillover or moderate-relationship/compensatory families) at W1 and W3. A significant 2 (between-family factor: family type at W1) × 2 (between-family factor: family type at W3) × 2 (within-family factor: adjustment scores at W1 and W3) 3-way interaction was found for parent-reported child internalizing problems, F (1,136) = 5.91, p < .05. Although children from the other three groups had fewer internalizing problems at W3 than W1, an increase in internalizing problems was observed in children who came from families that were identified as adequate-functioning at W1 yet became non adequate-functioning at W3. In other words, a family that was unstable in its type of daily relationship dynamics, particularly when it ceased to be identified as an adequate-functioning family, tended to confer risk for the child’s development of internalizing problems.

Discussion

Applying a novel grid-sequence analysis to daily diary data, we identified family typologies derived with daily diary data and found associations between family typologies and child adjustment. The intertwined marital and parent–child relationships (MR and PCR) constitute a family’s emotional environment, which affords an immediate developmental context for the child in the household.

Four specific family types were identified in our study at W1. The most common one was a relationship profile (cohesive) that featured parents who showed high levels of emotional quality in both marital and parent–child subsystems, which was also found in previous empirical and clinical classification (Johnson, 2003; Lindblom et al., 2017; Minuchin, 1974; Sturge-Apple et al., 2010). A second family profile identified was a fluctuating cohesive type, characterized by highly variable MR–PCR states but also high levels of relationship quality. The simultaneous identification of these two typologies highlights ways in which the current understanding of family relationship dynamics may be refined. For example, prior studies with a typological perspective have consistently found a well-functioning group characterized by high levels of warmth and closeness in both marital and parent–child subsystems (e.g., Cohesive subgroup in Struge-Apple et al 2010; Good-Marriage/Good-Parenting subgroup in Belsky & Fearon, 2004; Adequate-Functioning subgroup in Sturge-Apple et al., 2014). Taking advantage of the repeated measures over time provided by daily diaries and the grid-sequence methodology, our study captured heterogeneity in these well-functioning families and identified two ways in which “cohesiveness” may manifest itself: families may consistently exhibit positive relationship quality every day (cohesive), or their relationship quality may fluctuate from day to day while maintaining a high level of relationship quality in general (fluctuating cohesive). This finding advocates for a more nuanced understanding of family relationship dynamics (Brinberg et al., 2017; Laurenceau & Bolger, 2005; Ram & Gerstorf, 2009). Not only the content (i.e., the relative level in comparison to other families) but also the structure (i.e., intra-family variability) of family relationship dynamics matter when it comes to clustering families.

Following the calls for incorporating longitudinal design in understanding family typology (Lamela et al., 2018; Sturge-Apple et al., 2010), we employed measurement burst design to empower the examination of stability and changes in family typologies. A new typology emerged for mothers at later assessment points: a compensatory profile was identified for mothers at W2 and W3. Mothers from the compensatory group reported high emotional quality with their child in spite of the low emotional quality with the spouse on the same day. These mothers might seek to form strong emotional ties with the child to compensate for the negativity that they experienced with their spouse, supporting the compensatory hypothesis (Engfer, 1988). Although this “compensatory” association between marital and parent–child subsystems has been long forwarded by theoretical propositions and clinical classifications, empirical findings supporting this family profile have been mixed (Belsky et al., 1991; Gao & Cummings, 2019; Ponnet et al., 2013; Sherrill et al., 2017). It may be that previous findings were mostly obtained by variable-centered approaches; only the most dominant pattern of family dynamics was picked up, leaving the less commonly seen profiles, such as the compensatory type, undetected (Bergman & Magnusson, 1997). With a family-centered, pattern-based approach, the present study provides some of the first evidence in support of a family typology that is characterized by high PCR regardless of marital distress, especially for mothers.

There existed a moderate level of consistency in family typologies across the three assessment occasions (i.e., over two years). Although most families with adequate functioning in both marital and parent–child subsystems (i.e., cohesive and fluctuating cohesive families) stayed in their original or similar profile, families with poor MR and PCR (spillover) were more likely to change profiles at later time points. This result corroborates other findings (e.g., Johnson, 2003) to indicate that family typologies are subject to change over time. Families identified as spillover at one time point may function well at a later time point and be identified as the cohesive type. This finding delivers an encouraging message for social workers and family therapists that there is plasticity to work with for promoting positive family functioning.

Only mothers evidenced the profile characterized by consistent high levels of PCR despite marital negativity on the same day (i.e., compensatory).This result coalesced with existing findings (Belsky et al., 1991; Gao & Cummings, 2019; Stroud et al., 2011; Sturge-Apple et al., 2014) to show that marital distress may have a greater impact on father–child than mother–child relationships. Mothers may be the “gatekeepers” of caregiving practices (Parke, 2002) and tend to put forth more effort to counteract the unfolding cascade of marital negativity into child-rearing contexts (Belsky et al., 1991; Denham et al., 2010). Therefore, compared to fathers, mothers may be especially invested in good relations with the child in face of marital distress, reflected in the typology of compensatory in the present study. The different typologies observed for mothers and fathers highlight the importance of differentiating the experienced relationship dynamics for fathers and mothers.

Implications of Family Typology for Child Adjustment

Overall, children from the spillover families had the worst adjustment outcome, as indicated by the highest levels of externalizing problems reported by both parents. Possible interpretations merit consideration. The general emotional climate in a spillover family may not provide a supportive environment for children’s emotional and behavioral development. Emotionally exhausted from constantly feeling distressed toward their spouse and child, parents who were identified in this profile may no longer possess enough emotional resources to patiently interact with the child, address their emotional needs (Thompson & Meyer, 2007), or discipline them when necessary. Additionally, children from these families are likely to be drawn to family difficulties, revealed by heightened distress, dysregulated behaviors, and negative reappraisals (Cummings & Davies, 1996). In turn, such preoccupation and concern about family difficulties may compromise children’s adjustment and detract them from demands at school (Cummings et al., 2013). Moreover, although we described the spillover hypothesis (Kaczynski et al., 2006; Krishnakumar & Buehler, 2000) as a theoretical explanation of the positive associations between MR and PCR, other processes, such as enmeshed relational boundaries, may also account for this link. This significant MR–PCR link may indicate blurred boundaries between the marital and parent–child subsystems, which may intensify patterns of caregiving that overly involve children into the dysfunctional family unit and pose developmental risks.

Whereas spillover families seemed to confer the highest risk for children’s adjustment, cohesive families may provide a desirable family context. This is supported by the least adjustment problems of children from cohesive families, which replicated findings from previous research utilizing person-centered approaches (e.g., Johnson, 2003; Sturge-Apple et al., 2010). Hence, our study adds further evidence to the beneficial developmental utility of the family profile characterized by high relationship quality in both marital and parent–child subsystems. In addition to the general relationship quality of each family, the variability of family dynamics may be another factor related to child adjustment problems. Although not explored in the current study, we plan to investigate the implication of across-day variability of MR–PCR states for adolescent development in future work.

The undesirable developmental implications of the spillover families have been repeatedly reported in the larger extant literature. That is, compromised relationship quality among the father, mother, and child within a family is often associated with less-than-optimal child socioemotional outcomes, such as increased risk for psychopathology and loneliness (Heshmati et al., 2021; Sturge-Apple et al., 2014). This association seems to be rather robust because research employing a diverse array of methodologies, including latent profile analysis, state space grids, and social network analysis, have consistently reached the same conclusion. It may be time to move beyond merely linking family contexts to child adjustment concurrently and start exploring over-time changes in family dynamics and how these changes may impact later trajectories of child development (Teti & Fosco, 2021).

Accordingly, we found in the present study that changes in family typology across two years were linked with child internalizing problems. Consistent with a transactional model of development (Cummings et al., 2000), this finding shows that children’s adjustment outcomes are jointly influenced by their current and previous family experiences. When families were seen as adequate functioning at W1 but were classified as non-adequate functioning at W3, children’s internalizing problems increased. Changes of a family emotional environment toward an undesirable direction may negatively affect children’s functioning. As for the developmental sequel of families who remained in the inadequate group, our finding was surprising: children from these families exhibited fewer internalizing problems. This result may indicate an adaptive mechanism developed by those children who were continuously exposed to family contexts with unceasing negativity.

Surprisingly, we did not find longitudinal associations between earlier family typology and child adjustment problems at W3. This null finding may be attributed to our stringent tests in which children’s adjustment problems at earlier waves were controlled for. Our sample of low-risk families, consisted with mostly married, middle-class couples, may also account for the lack of longitudinal findings. In the context of a relatively stable and resourceful socioeconomic background, children’s long-term developmental outcomes may not be as sensitive to different family typologies. Moreover, other youth outcomes, such as prosocial behaviors, may be more sensitive to different family typologies. Future studies are encouraged to include more measures of youth outcomes, focus on high-risk families, and investigate the role of socioeconomic disparity in the longitudinal association of family type and child adjustment.

An interesting result worth noticing is that there was a significant, but weak correspondence between mothers’ and fathers’ cluster membership within family. That is, mothers and fathers from the same family do not always exhibit the same pattern of relationship dynamics. Limited by our current sample size, we did not explore how different configurations of relationship dynamics between the two parents from the same family would impact the child’s adjustment outcome. It is likely, however, the negative impact of having one parent in the spillover profile on the child’s socioemotional development could be buffered by having the other parent in the cohesive profile; or it could be exacerbated if the other parent was also in a non-adequate profile, such as the spillover profile. These possibilities should be tested empirically in future studies.

Strengths, Limitations, and Future Directions

Methodology used in the present study highlights new avenues for understanding intra-family relationship dynamics in the context of inter-family differences and their developmental implications. First, daily diary studies produce intensive longitudinal for delineating intra-family variability (Laurenceau & Bolger, 2005), which complements traditional longitudinal panel methods to capture family dynamics at the daily level. Second, SSG extract and plot MR and PCR simultaneously, re-representing bivariate time series data into univariate time series, which allows for capturing interrelated, transactional family dynamics. Third, we identified family typologies with distinct patterns of intra-family dynamics and explored whether certain family typologies indicate risks or opportunities for intervention efforts.

Despite the strengths, our study is not without limitations. First, the participating families had children within a wide range (i.e., 7–17 years old). Typologies may be different for families with middle-childhood children and late-adolescence teenagers, as adolescence has been theorized as a transitional period when fast-pacing changes in the family emotional states may occur (Granic et al., 2003; Lichtwarck-Aschoff et al., 2009). Second, findings may not be generalizable to other populations, such as divorced families and populations with diverse racial compositions. Third, family relational dynamics may be susceptible to the influence of other daily-level characteristics, such as the number of interactions/conflicts a parent had with the spouse/child on a certain day, which was not assessed in this study. Fourth, our assessment of PCR did not incorporate the child’s perspective. Fortunately, it would not be difficult to add child’s input in future studies, because state space grids can be easily extended to state space cubes in order to accommodate an additional dimension. Relatedly, our reliance on parents to report both MR, PCR, and child adjustment may introduce same-reporter bias. Ups and downs across subsystems reported by the same person could indicate individual characteristics of emotionality, rather than family characteristics. Although father- and mother-report of child adjustment problems were averaged to obtain a single indicator, there could be method variance inflation that may account for the significant associations between family typology and child adjustment. Future studies should include multiple informants on family relationships and child adjustment outcomes and incorporate assessment of individual differences, such as temperament or personality traits.

Several constraints of our methodology warrant further discussion. To begin with, although participants were instructed to complete one diary every day before bed, and record the date when diaries were completed each day, we cannot rule out the possibility that in some instances more than one diary was completed on the last day of the burst. This is a limitation of many early diary studies, such as ours. We acknowledge that best practices have changed due to technological development. Using stamps on paper-pencil diaries to track date/time or employing electronic diaries to automatically capture time of completion are strongly encouraged in future diary research. Additionally, the loss of data in the final analyses due to attrition or incomplete data from either partner (i.e., mother or father) may pose problems for interpretation. The final sample and the rest of the full sample may differ in certain contextual factors that were not available in our dataset. For example, families excluded in final analysis may experience more life events, such as moving, job change, or bereavement, limiting parents’ ability to provide diary reports. These families may manifest different day-to-day relationship dynamics. It is highly recommended that future studies with similar designs should include additional contextual measures. Third, our study used a 3 × 3 (9 cell) grid after considering the distribution of daily diary scores as well as the interpretability of achieved typologies. We acknowledge that a finer grid configuration (e.g., the 10 × 10 grid) may offer additional information; after all, degrading data from a 10-point Likert scale to 3-point scale results in loss of information. Too many cells, however, can invoke interpretation challenges. Therefore, we encourage future researchers to carefully design their measurement tools so that their gridding choice can be mapped to their theoretical framework.

In summary, the present study applied a family-centered approach to daily diary data and identified family types with distinct relationship patterns. Children exhibited the highest level of adjustment problems if they were affiliated with a family type that were characterized by low relationship quality. Therapists and social workers who work with families are encouraged to address the need in lessening emotional negativity in the marital and parent–child relationships, which may have a long-term effect on children’s developmental outcomes.