Refining Social-Information Processing Theory: Predicting Maternal and Paternal Parent-Child Aggression Risk Longitudinally

Abstract

Background:

Clarifying the pathways leading parents to engage in parent-child aggression (PCA) would benefit child abuse prevention efforts during the perinatal period.

Objective:

The present investigation empirically tested whether a social information processing (SIP) model could predict PCA risk from factors assessed in new mothers and fathers.

Participants and Setting:

This study recruited a diverse sample of 201 primiparous mothers in the last trimester of their pregnancy along with 151 fathers.

Methods:

Using a prospective longitudinal study, the hypothesized SIP model was refined statistically using SIP factors measured prenatally to predict PCA risk when their children were 18 months. This refined model was then validated with SIP factors assessed when infants were 6 months to predict PCA risk when toddlers were 18 months.

Results:

In general, findings indicated poor empathy related to greater overreactivity and more negative child behavior attributions. Moreover, approval of PCA use, negative child attributions, less knowledge of non-physical discipline alternatives, and higher child compliance expectations predicted subsequent PCA risk. The proposed SIP model for mothers demonstrated considerable stability. Although SIP processes predicted paternal risk, several SIP relations changed over time for fathers.

Conclusions:

Findings suggest comprehensive theoretical models like SIP theory can guide the specific processes to target for prevention and clarify how processes may be interconnected. SIP processes appear relevant and relatively stable targets for prevention and early intervention, particularly for mothers. SIP processes were applicable for fathers although the model was less consistent, suggesting work in examining paternal PCA risk remains an important research direction.

Physical child abuse is substantiated by protective services when physical force results in injury, representing over 18% of the nearly 700,000 cases of maltreatment in the U.S. annually (U.S. Department of Health and Human Services, 2019). Nonetheless, official statistics on abuse typically underestimate the scope of physical abuse (Fallon et al., 2010; Sedlak et al., 2010). Research concentrated on parents identified through protective services captures a fraction of abusive parents because of the underreporting observed in official statistics, and such work also misses the opportunity to clarify avenues for averting child abuse in the first place. Indeed, child abuse prevention programs typically target pregnant and perinatal mothers (e.g., Chartier et al., 2017; Eckenrode et al., 2017; Pajer et al., 2014) consistent with priorities to promote early identification and primary prevention of child abuse (Eckenrode, 2011). Thus, the current study adheres to a prevention approach by evaluating child abuse risk in new parents.

Parents typically administer physical discipline hoping to deter child misbehavior by inflicting pain but not injury (Straus, 2000), thereby distinguishing physical abuse from physical discipline largely because of the outcome of using physical force—injury. As parents’ physical punishment becomes more intense and frequent, child abuse becomes more likely (Benjet & Kazdin, 2003; Durrant, Trocmé, Fallon, Milne, & Black, 2009; Gershoff & Grogan-Kaylor, 2016; Zolotor, Theodore, Chang, Berkoff, & Runyan, 2008). Given such links, physical discipline and abuse have been construed as opposing endpoints on a parent-child aggression (PCA) continuum (Gershoff, 2010; Greenwald, Bank, Reid, & Knutson, 1997; Rodriguez, 2010a; Rodriguez & Richardson, 2007; Straus, 2000, 2001; Whipple & Richey, 1997). As physical discipline intensifies along this continuum, physical abuse becomes more likely. Parents who use harsh physical discipline often adopt an authoritarian parenting style (Kochanska, Kuczynski, & Radke-Yarrow, 1989; Rodriguez, 2010a), a parenting approach characterized by high parental demands with limited warmth (Smetana, 2017). Further, a parent’s likelihood of becoming physically abusive has been termed child abuse potential (Milner, 1994), which has been linked to a harsh and authoritarian parenting style (Conners et al., 2006; Margolin, Gordis, Medina, & Oliver, 2003; Rodriguez, Smith, & Silvia, 2016a, 2016b) and physically abusive discipline tactics (Rodriguez, 2010a). Therefore, a parent’s risk to engage in PCA was operationalized in this study inclusively, incorporating indicators across the PCA continuum, including child abuse potential, physical discipline use, and harsh and authoritarian parenting. The current investigation aimed to empirically test a theoretical model to predict PCA risk from factors identified in new parents utilizing a prospective longitudinal design.

Social Information Processing theory (SIP) is one of the leading models that has been applied to understand parents’ risk to engage in child abuse (Milner, 2000; Rodriguez, Silvia, & Gaskin, 2019). According to the SIP model, before parents enter into discipline situations, they hold preconceived beliefs and emotions about parenting and their children, which are considered preexisting schema. Then, when confronted with a potential discipline decision, parents may misperceive the situation (Stage 1) and form biased, negative appraisals and expectations regarding child behaviors (Stage 2). Parents may then fail to integrate all relevant information before engaging in PCA, including considering their nonphysical discipline options (Stage 3). Once they begin implementing physical discipline, they fail to adequately monitor their intensity, escalating toward physical abuse (Stage 4).

SIP theory applied to PCA has empirical support in terms of its individual elements (Milner, 2000) and more collectively (e.g., Rodriguez et al., 2019), in which cognitive and affective preexisting schemas may link to subsequent SIP stages. For example, SIP preexisting cognitive schemas approving of PCA use are associated with greater PCA risk (McCarthy, Crouch, Basham, Milner, & Skowronski, 2016; Rodriguez, Bower-Russa, & Harmon, 2011). A preexisting affective schema may be manifest as empathy, which is associated with parents’ positive emotions toward their child (Light et al., 2009), whereas lower empathy has been linked with greater child abuse risk (de Paúl, Pérez-Albéniz, Guibert, Asla, & Ormaechea, 2008; Francis & Wolfe, 2008; Rodriguez, 2013). SIP Stage 1 involves processes that compromise parents’ accurate perceptions, which can arise when parents are easily frustrated and emotionally dysregulated. Indeed, poor frustration tolerance (McElroy & Rodriguez, 2008; Rodriguez, Russa, & Kircher, 2015; Rodriguez, Baker, Pu, & Tucker, 2017) and emotion regulation difficulties (Hien, Cohen, Caldeira, Flom, & Wasserman, 2010; Hiraoka et al., 2016) are associated with heightened child abuse risk. For Stage 2, negative attributions about children’s behavior have been observed in abusive mothers (Haskett, Scott, Willoughby, Ahern, & Nears, 2006) and in those with higher PCA risk (Azar, Okado, Stevenson, & Robinson, 2013; Berlin, Dodge, & Reznick, 2013; Rodriguez, Cook, & Jedrziewski, 2012). Also at Stage 2, at-risk parents appear to expect more compliance from children following discipline (Rodriguez et al., 2016b). In Stage 3, parents’ awareness of nonphysical discipline alternatives represents a frequent goal in child abuse prevention programs (Lundahl, Nimer, & Parsons, 2006; Prinz, Sanders, Shapiro, Whitaker, & Lutzker, 2009) and has been linked to lower maternal and paternal PCA risk (Rodriguez et al., 2016b).

Combined, these SIP factors have been proposed to follow two pathways—one focused on discipline, arising from physical discipline approval preexisting cognitive schema, and a second arising pre-existing affective schema from empathy (Rodriguez et al., 2019). The current study further refines these SIP pathways empirically using data from different time points in a longitudinal design. Rather than having prevention and intervention efforts target random elements in the hopes of reducing child abuse, theoretically grounded models lend insight into how PCA risk factors are interconnected, elucidating how modifying one element may impact other processes. Considering these processes longitudinally also permits more causal inferences than cross-sectional designs by evaluating how early SIP processes predict subsequent PCA risk. Such longitudinal work is needed not just for mothers but fathers as well, given the repeated calls to investigate paternal PCA risk (Lee, Bellamy, & Guterman, 2009; Rodriguez, Silvia, & Pu, 2018; Smith Slep & O’Leary, 2007; Stith et al., 2009).

The SIP processes necessarily would transpire while parents are confronted by other challenges in their life. Such challenges may “tax” them, interfering with their ability to parent optimally. In contrast, parents’ personal strengths may serve as resources that can reduce PCA risk. Personal issues such as psychopathology (Ammerman & Patz, 1996; Pajer et al, 2014; Stith et al., 2009), intimate partner violence (Casanueva & Martin, 2007; Margolin et al., 2003), and substance use (Ammerman, Kolko, Kirisci, Blackson, & Dawes, 1999; Hien et al., 2010; Pajer et al., 2014) are all linked to elevated PCA risk. Yet parents may access resources to manage such taxes, such as social support (Counts, Buffington, Chang-Rios, Rasmussen, & Preacher, 2010; Rodriguez & Tucker, 2015). Alternatively, parents may draw support from their partner, which can also reduce PCA risk (Bryson, 2004; Rodriguez et al., 2016b) particularly because low relationship satisfaction is linked to greater child abuse risk (Florsheim et al., 2003). Further, problem focused coping skills are related to lower PCA risk (Rodriguez et al., 2018) whereas emotion-focused or avoidant coping is associated with greater PCA risk (Cantos, Neale, O’Leary, & Gaines, 1997; Rodriguez, 2010b). To more clearly test the SIP pathways, the current study predicted PCA risk with and without controlling for such personal-level taxes and resources.

Current Study

The current investigation evaluated SIP theory longitudinally to predict maternal and paternal PCA risk independent of parents’ current personal taxes (psychopathology, substance use, intimate partner violence) or resources (social support, partner satisfaction, coping). PCA risk incorporated indicators across the PCA continuum, including greater child abuse potential, harsher authoritarian parenting style and punitive reactions to noncompliant child behavior, and greater PCA use. Extending Rodriguez and colleagues (2019) prior analyses of the SIP model that examined prenatal factors to predict PCA risk during infancy, the current study utilized additional data to empirically evaluate the robustness of that model; data on SIP factors collected prenatally from mothers and fathers were used to predict their PCA risk nearly two years later, when their child was a toddler, with the specific aiming to refine the SIP model and its pathways and to examine its stability. Once the model was refined, this model was tested with data on the same SIP factors collected from parents when their child was an infant to determine if that model still predicted their PCA risk when their child was a toddler. This approach can identify whether the proposed SIP elements and paths in the theoretical model are empirically robust, for mothers and fathers separately, evidencing stability that would provide critical insights for child abuse prevention efforts that target pregnant versus postnatal mothers. We utilized a multiple indicator methodology including self-report as well as analog tasks which more indirectly assess participants in order to reduce socially desirable responding (Camilo, Garrido, & Calheiros, 2016; DeGarmo et al., 2006; Fazio & Olson, 2003). Building on the two previously proposed SIP pathways (Rodriguez et al., 2019), the current study investigated both empirically in a proposed theoretical model (see Fig. 1). In pathway one, preexisting affective schemas from empathy initiates one path to PCA risk, in which lower empathy was predicted to lead to greater over-reactivity (emotion dysregulation, low frustration tolerance) which would detract from parents’ accurately attending to discipline episodes (Stage 1) which would in turn directly predict PCA risk. Lower empathy was expected to promote more negative child behavior attributions (Stage 2) that would directly predict elevated PCA risk. In pathway two, focused on discipline processes, preexisting schema approving of PCA were expected to relate to greater negative child behavior attributions and higher expectations of child compliance after discipline (Stage 2) and less knowledge of nonphysical discipline alternatives (Stage 3), which were each expected to relate to greater PCA risk. The following research goals (RGs) drove this study: (RG1) To evaluate SIP paths, the hypothesized model was refined statistically from SIP factors assessed prenatally to predict PCA risk nearly two years later for mothers and fathers independently. (RG2) To evaluate model stability, this refined model was then applied to data on the SIP factors gathered from mothers and fathers when their infants were 6 months old to predict PCA risk one year later.

Figure 1.

Hypothesized Path Model

Method

Participants

Parents in this study were enrolled in a prospective longitudinal study tracking PCA risk, the Following First Families (“Triple-F”) study, conducted in a large Southeastern US urban city. English-speaking primiparous mothers over age 15 were recruited in the final trimester of their pregnancy, with 203 women and 151 of their male partners (86% of fathers available to participate) enrolled at Time 1 (T1). Mothers and fathers were reassessed when their infant was 6 months old (± 2 weeks) for Time 2 (T2) and when their toddler was 18 months old (± 3 weeks) for Time 3 (T3). By Time 2, two of the families were no longer eligible to continue because the baby died shortly after childbirth. Of the remaining 201 families, 186 mothers and 146 fathers (>92% available fathers) participated in Time 2 and 180 mothers and 146 fathers participated in Time 3. Triple-F allows father-figures to change across the study as this more realistically reflects family composition (n = 6 families). For example, some fathers expected to be involved in raising the child at Time 1 but the mother had a different partner by Time 2; longitudinal analyses of fathers thus only involves data from same-fathers. Data missing at a time point are estimated (see Analytic Plan).

At Time 1, mothers’ mean age was 26.04 years old (SD = 5.87) and fathers’ mean age was 28.87 years old (SD = 6.10). With regard to race/ethnicity, mothers identified as: 50.7% Caucasian, 46.8% African-American, 1.5% Native American, and 1% Asian; of these, 3% also identified as Hispanic/Latina and 5.5% identified as biracial. With respect to race/ethnicity, fathers identified as: 54% Caucasian, 45.3% African-American, and 0.7% Asian; additionally, of these, 4% identified as Hispanic/Latino and 4.7% identified as biracial. With regard to educational attainment, mothers reported as follows: 30.3% high school or less; 20.9% some college or vocational training; 21.4% college degree; and 27.4% beyond college degree. For fathers’ educational attainment: 25.3% high school or less; 24.7% some college or vocational training; 27.3% college degree; 22.7% beyond college degree.

The Triple-F study oversampled mothers evidencing sociodemographic risk, with 53.2% of mothers meeting at least one of the following criteria: (a) 150% below the poverty line; (b) receipt of public assistance; (c) high school education or less; (d) age 18 or younger. Over half of mothers reported a combined annual household income under $40,000, with 49.3% reporting a household income within 150% of the federal poverty line and more than 42% receiving public assistance.

Measures

Using a multiple indicator approach, measures contributing to each construct in the statistical model are described briefly in Table 1, with this sample’s internal consistency coefficients in Table 2.

Table 1.

Measures by Construct

Construct/Measure	Description
Empathy
Interpersonal Reactivity Index (IRI; Davis, 1983)	Empathic Concern and Perspective Taking subscales, each with 7 items on a 5-point scale (1 = does not describe me well, 5 = describes me very well); summed across items, total hi scores = higher empathy
PCA Attitudes
Adult Adolescent Parenting Inventory-2 (AAPI-2; Bavolek & Keene, 2001), Form A	Belief in Corporal Punishment scale; 11 item 5-point scale (1 = strongly disagree, 5 = strongly agree); summed, hi scores = greater PCA approval
Physical Abuse Vignettes (PAV; Shanalingigwa, 2009)	8 vignettes depicting wide range of PCA scenes, 3 questions/vignette: (1) is scene abusive? Yes/No; (2) rate seriousness of PCA, 4-point scale (1 = least serious, 4 = most serious); (3) if they would officially report parent, Yes/No; 3 questions across vignettes summed, standardized for composite score across questions; hi Combined scores = less approval
Parent-Child Aggression Acceptability Movie Task (Parent-CAAM Task; Rodriguez et al., 2011)	Analog task of 90-sec movie clips of varying PCA levels; timed for how long before they stop the clip if/when they consider scene abusive; average response time across 8 clips, hi scores = greater PCA approval
Reactivity
Negative Mood Regulation Scale (NMRS; Catanzaro & Mearns, 1990)	30 items on ability to regulate negative mood on 5-point scale (1 = strongly agree, 5 =strongly disagree); summed across items, total hi scores = lower emotion regulation
Frustration Discomfort Scale (FDS; Harrington, 2005)	7 items on perceived tolerance of frustration on 5-point scale (1 = strongly disagree, 5 = strongly agree); summed across items, total hi scores = poorer frustration tolerance
Child Attributions
Plotkin Child Vignettes (PCV; Plotkin, 1983)	18 vignettes of child misbehavior; 9-point scale how much they view child behavior as intentional (1 = did not mean to annoy me at all, 9 = only reason the child did this was to annoy me); summed across vignettes, total hi scores = more negative attributions
Video Ratings (VR; Leerkes & Siepak, 2006)	Two 1-min videos of babies crying, each followed by questions about why baby is crying; Negative Attributions subscale with 6 items/video rated on 4-point scale (1 = strongly disagree, 4 = strong agree), averaged across items; total low scores = more negative attributions
Infant Crying Questionnaire (ICQ; Haltigan et al., 2012)	General beliefs about infant crying; 9-item Minimization subscale for viewing crying as manipulation on 5-point scale (1 = never, 5 = always); subscale averaged across items, hi scores = more negative attributions
Noncompliance Implicit Association Test (N-IAT; Rabbitt & Rodriguez, 2019)	Analog sorting task timing how quickly participants sort descriptors of child behavior as good/bad or obeying/disobeying; sorting slows for items inconsistent with beliefs; low scores = more negative attributions
Compliance Expectations
Compliance Expectations Scale (Rodriguez et al., 2016b)	6 vignettes of parents disciplining child behavior, vary on 2 dimensions (child culpability, discipline severity); asked whether child will repeat behavior, rated on 5-point scale (1 = will do it again, 5 = learned their lesson); summed across vignettes and reversed, total hi scores = high compliance expectations
Knowledge of Discipline Options
Production of Discipline Alternatives (Rodriguez, Wittig, & Christl, 2019)	For last PCV vignette, open-ended question, type all possible discipline responses; 2 coders categorize each response as physical, nonphysical, psychological; number of nonphysical options generated averaged between coders (ICC = .94); proportion scores control for more total options (total nonphysical options 4 total options), hi proportion scores = more nonphysical options generated
Time 3 Controls--Taxes
Brief Symptom Inventory-18 (BSI; Derogatis & Melisaratos, 1983)	18 items on frequency of depression/anxiety symptoms in past week, 5-point scale (0 = Not at all, 4 = extremely); summed across items, total hi scores = more symptoms
Substance Abuse & Mental Illness Symptoms Screener (SAMISS; Whetten et al., 2005)	Includes 7 items on past year frequency and extent of problematic substance use on 5-point scales (e.g., 0 = never, 4 = daily or almost daily); total hi scores = greater substance use problems
Revised Conflict Tactics Scale-Short (CTS-2S; Straus & Douglas, 2004)	20 items, frequency of past year intimate partner violence perpetration and victimization; 8-item Victimization Scale assesses frequency of physical or psychological assault using variably weighted count scores; summed across 8 items, hi scores = greater experience of victimization
Time 3 Controls--Resources
Coping Self-Efficacy Scale (CSES; Chesney et al., 2006)	Competence in using problem-focused coping, 12 items rated on 11-point scale (0 = cannot do at all, 10=certain I can do); summed across items, total hi scores = greater coping
Couple Satisfaction Index (CSI; Funk & Rogge, 2007)	Satisfaction with partner relationship, 10 items rated on different 6-point scales (e.g., 0 = not at all, 6 =completely); summed across items, total hi scores = greater relationship satisfaction
Social Support Resources Index (SSRI; Vaux & Harrison, 1985)	5 items rated for 2 closest supporters, 5-point scale (1 = not satisfied, 5 = very satisfied); summed, hi scores = greater social satisfaction
Time 3: PCA Risk
Child Abuse Potential Inventory (CAPI; Milner, 1986)	Screens for physical child abuse risk, 160 Agree/Disagree items, 77 items on variably weighted Abuse Scale; hi scores = greater PCA risk
Adult Adolescent Parenting Inventory-2 (AAPI-2; Bavolek & Keene, 2001) Form B	40 items on parenting beliefs/behaviors considered to characterize abusive parenting, each rated on 5-point scale (1 = strongly disagree, 5 = strongly agree); summed across items, total hi scores = greater PCA risk
Response Analog to Child Compliance Task (ReACCT; Rodriguez, 2016)	Analog task with 12 successive scenes portraying child compliance or noncompliance; parent selects from 16 options how they would respond to child in scene; options variably weighted for adaptive v. maladaptive responses; 12-item Noncompliance Scale for harsh reactions to noncompliance, summed, hi scores = greater PCA Risk
Expected Parental Authority Questionnaire (Expected PAQ; Boppana & Rodriguez, 2017)	Modified parenting style questionnaire (Buri, 1991) rephrased in future tense with 30 items, 10 items each for authoritative, authoritarian, permissive styles, rated on 5-point scale (1 = strongly agree, 5 = strongly disagree); Authoritarian subscale used in this study, summed across 10 items, total hi scores = more authoritarian parenting style
Parent-Child Conflict Tactics Scale (CTSPC; Straus et al., 1998)	22 items on past-year use of discipline tactics; 13-item Physical Assault subscale with wide range of PCA with variably weighted count scores; summed frequency counts, hi scores = more frequent use of PCA

Note. See Rodriguez et al. (2019) for more details on these measures, with the exception of the CTSPC.

Table 2.

Means, Standard Deviations, and Internal Consistencies for Mothers and Fathers by Time Point

	Mothers				Fathers
		Time 1	Time 2	Time 3		Time 1	Time 2	Time 3
	α T1/T2a	M (SD)	M (SD)	M (SD)	α T1/T2	M (SD)	M (SD)	M (SD)
Interpersonal Reactivity Index	.82/.84	55.74 (7.98)	56.08 (8.74)		.82/.81	52.95 (8.14)	52.45 (8.57)
AAPI-2 Corporal Punishment	.82/.84	31.87 (8.30)	30.33 (9.11)		.83/.80	32.14 (8.93)	30.44 (8.29)
Physical Abuse Vignette Combined	.79/.81	0.00 (2.54)	0.00 (2.61)		.84/.84	0.00 (2.62)	0.00 (2.71)
Parent-CAAM Task	.81/.79	19.13 (11.81)	17.32 (11.45)		.84/.82	18.50 (12.45)	18.85 (12.18)
Negative Mood Regulation Scale	.90/.91	64.67 (16.67)	65.24 (18.24)		.91/.90	63.36 (17.07)	66.14 (15.79)
Frustration Discomfort Scale	.82/.87	17.94 (5.27)	17.84 (5.96)		.86/.86	17.71 (5.72)	18.01 (5.77)
Plotkin Child Vignette-Attribution	.85/.82	39.96 (16.30)	36.03 (13.67)		.88/.86	38.53 (17.37)	34.28 (14.05)
Video Rating-Negative Attribution	.83/.89	3.57 (.44)	3.57 (.49)		.80/.84	3.55 (.42)	3.55 (.46)
Infant Cry Quest-Minimization	.75/.74	2.33 (.62)	2.25 (.64)		.82/.81	2.44 (.74)	2.39 (.69)
Noncompliance Implicit Association Test		1.05 (.41)	.96 (.47)			1.02 (.49)	.92 (.51)
Compliance Expectations Scale	.70/.77	17.40 (3.69)	17.72 (4.06)		.74/.77	17.97 (3.47)	17.84 (3.85)
Production of Discipline Alternatives		.83 (.23)	.86 (.22)			.79 (.26)	.83 (.25)
Brief Symptom Index	.88/.91	9.81 (8.93)	6.40 (8.48)		.89/.89	5.05 (6.74)	3.64 (5.56)
SAMISS Substance Abuse	.67/.67	1.68 (2.33)	2.61 (2.53)		.72/.70	4.71 (3.59)	4.26 (3.24)
CTS-2S Victimization		6.67 (11.79)	5.67 (9.76)			4.71 (7.61)	5.20 (7.81)
Coping Self Efficacy Scale	.92/.94	97.64 (21.95)	98.60 (22.78)		.90/.94	104.39 (20.28)	105.10 (22.98)
Couple Satisfaction Index	.98/.97	50.40 (12.66)	48.32 (11.95)		.95/.96	53.01 (8.60)	51.91 (9.63)
SSRI-Social Satisfaction	.91/.90	41.66 (7.30)	41.74 (7.03)		.92/.93	41.01 (7.56)	40.63 (8.20)
Child Abuse Potential Inventory				90.53 (76.85)				77.15 (65.06)
AAPI-2 Total	.91			98.76 (22.20)	.89			102.14 (19.72)
ReACCT Noncompliance	.81			1.10 (12.84)	.79			1.21 (13.60)
Expected PAQ Authoritarian	.82			34.11 (6.74)	.86			32.62 (7.15)
CTSPC Physical Assault				6.85 (11.24)				6.35 (10.84)

Note. AAPI-2 = Adult-Adolescent Parenting Inventory-2; Parent-CAAM = Parent-Child Aggression Acceptability Movie Task; SAMISS = Substance Abuse and Mental Illness Scale; CTS-2S = Revised Conflict Tactics Scale, Short Form; SSRI = Social Support Resources Index; ReACCT = Response Analog to Child Compliance Task; PAQ = Parental Authority Questionnaire; CTSPC = Parent-Child Conflict Tactics Scale.

^aAlpha for PCA risk measures at T3. Alpha not computed for: CTS-2S, CTSPC involve low frequency count data; Noncompliance IAT and Production of Discipline Alternatives are single values, Child Abuse Potential Inventory items are variably weighted.

Procedure

Participants were recruited with flyers distributed at local hospital obstetric/gynecological clinics and childbirth classes. Mothers interested in enrolling in Triple-F contacted the lab, scheduling a 2–2½ hour session for themselves and their partner, when available, for Time 1. At Time 2 and Time 3, families were re-contacted for a 3-hour session. Mothers and fathers independently provided consent and completed the study in separate rooms. Measures were administered electronically on laptop computers with headphones. All study procedures were approved by the university’s Institutional Review Board.

Analytic Plan

Preliminary analyses were performed with SPSS 25.0. Subsequent analyses (confirmatory factor analyses and path models) utilized Mplus 8.1 using maximum likelihood estimation with robust standard errors. Missing values for mother and fathers were accommodated using full-information maximum likelihood methods (Enders, 2010). Although a non-significant chi-square suggests good fit, stronger model fit indices were utilized to assess fit: root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), and comparative fit index (CFI) (Kline, 2011). For RMSEA, a lower bound of the confidence interval less than .05 suggests good fit and an upper bound of its confidence interval more than .10 suggests poor fit (Kline, 2011); for SRMR, a threshold of values below .08 is preferred; values above .95 for CFI suggest good fit (Hu & Bentler, 1999; Kline, 2011).

To refine the SIP model, a series of path models were fit to mothers’ and fathers’ data independently. Models were evaluated based on fit indices, modification indices, and Wald tests to identify the most parsimonious model (see Kline, 2011 for discussion on model trimming) for RG1 (Time 1 SIP to Time 3 PCA risk), and this final model was then used as the validation model for RG2 (Time 2 SIP to Time 3 PCA risk). The starting point for RG1 was largely based on the SIP path model identified in Rodriguez and colleagues (2019) as the base model (see also Fig. 1). Pathways were added to this base model when analyses identified theoretically meaningful modification indices; to achieve parsimony, removal of a non-significant pathway was evaluated using the Wald test. A non-significant Wald test supports removing the non-significant pathway for the model without a significant reduction in overall model fit. Pathways with marginal Wald statistics were retained in the model to avoid over-fitting.

Results

Preliminary Analyses

Means, standard deviations, and internal consistencies for each measure, for mothers and fathers separately, are presented in Table 2. Correlations between Time 3 PCA risk with Time 1 and Time 2 SIP elements and Time 3 Taxes and Resources, by parent, appear in Table 3.

Table 3.

Bivariate Association between Time 1 and Time 2 Predictors and Time 3 Taxes and Resources with Time 3 PCA Risk Measures

		Mothers					Fathers
	Time3	31.	32.	33.	34.	35.	31.	32.	33.	34.	35.
Time 1
1.		−.17*	−.26***	−.21**	.03	−.07	−.32***	−.22*	−.17	−.14	.04
2.		.11	.46***	.33***	.37***	.32***	.04	.52***	.37***	.58***	.32***
3.		.01	−.15*	−.03	−.09	−.20**	.04	−.21*	−.20*	−.18*	−.23*
4.		.11	.18*	.14	.12	.22**	.07	.27**	.09	.30***	.07
5.		.30***	.17*	.02	−.03	.06	.34***	.27**	.13	−.10	−.08
6.		.29***	.20**	.00	.13	.03	.16	.09	−.06	−.01	−.04
7.		.24***	.51***	.32***	.28***	.21**	.15	.22*	.13	.09	.02
8.		−.32***	−.40***	−.26***	−.23**	−.14	−.29**	−.27**	−.29**	−.02	−.21*
9.		.20**	.28***	.22**	.19**	.00	.24**	.32***	.15	.16	−.09
10.		−.25***	−.23**	−.16*	−.12	.07	−.27**	−.22*	−.12	.07	.05
11.		−.04	−.32***	−.11	−.33***	.03	−.03	−.28**	−.24***	−.26**	−.00
12.		−.21**	−.29***	−.29***	−.22**	−.27***	.08	−.21*	−.22*	−.36***	−.19*
Time 2
13		−.23**	−.35***	−.31***	−.10	−.01	−.39***	−.20*	−.13	.00	.04
14.		.26***	.61***	.35***	.44***	.27***	.08	.49***	.40***	.44***	.33***
15.		−.01	.22**	−.01	−.05	−.22**	−.12	−.22*	−.28**	−.27**	−.37***
16.		.12	.16*	.21**	.07	.21**	.16	.21*	.12	.09	.13
17.		.33***	.22**	.10	.09	−.08	.48***	.26**	.23*	−.05	−.04
18.		.14	.16*	−.02	.02	.02	.18	.09	.04	.00	−.06
19.		.24***	.47***	.36***	.28***	.34***	.27**	.25**	.15	.28***	.07
20.		−.35***	−.37***	−.25***	−.14	−.11	−.33***	−.31***	−.27**	−.20*	−.08
21.		.32***	.41***	.27***	.16*	−.02	.54***	.39***	.25**	.25***	.08
22.		−.21*	−.17*	−.11	−.09	.04	−.15	−.00	−.11	.01	.15
23.		−.20**	−.34***	−.26***	−.28***	−.10	−.08	−.26**	−.25**	.27**	−.03
24.		−.13	−.33***	−.33***	−.29***	−.33***	.10	−.28**	−.40***	−.27**	−.42***
Time 3
25.		.58***	.02	.17*	−.10	.26***	.35***	.10	.11	−.06	.01
26.		.14	−.12	.04	−.05	.16*	.10	−.12	.01	−.13	−.14
27.		.32***	.13	.21***	.00	.26***	.24**	.03	.06	−.02	.16
28.		−.45***	−.16*	−.25***	−.01	−.03	−.45***	−.24**	−.19*	.03	.03
29.		−.63***	−.25***	−.24***	−.16*	−.14	−.45***	−.10	−.20*	.11	−.04
30.		−.47***	−.11	−.26***	.09	−.06	−.40***	−.06	−.10	−.11	−.06

Note. 1 & 13=Interpersonal Reactivity Index; 2 & 14=Adult-Adolescent Parenting Inventory-2 Corporal Punishment Attitudes; 3 & 15=Physical Abuse Vignettes Combined; 4 & 16=Parent-Child Aggression Acceptability Movie Task; 5 & 17=Negative Mood Regulation Scale; 6 & 18=Frustration Discomfort Scale; 7 & 19=Plotkin Child Vignettes, Attribution; 8 & 20=Video Ratings, Negative Attribution; 9 & 21=Infant Cry Questionnaire, Minimization; 10 & 22=Noncompliance-IAT; 11 & 23=Compliance Expectations Scale; 12 & 24=Production of Discipline Alternative, Nonphysical Proportion; 25=Brief Symptom Inventory; 26=Substance Abuse & Mental Illness Symptoms Screener; 27=Revised Conflict Tactics Scale-Short, Victimization; 28=Coping Self Efficacy Scale; 29 = Couple Satisfaction Index; 30=Social Support Resources Index; 31=Child Abuse Potential Inventory Abuse Scale; 32= Adult Adolescent Parenting Inventory-2; 33= Response Analog to Child Compliance Task Intentional, Noncompliance; 34=Expected Parental Authority Questionnaire, Authoritarian Parenting; 35=Parent-Child Conflict Tactics Scale, Physical Assault.

^*p ≤ .05,

^**p ≤ .01,

^***p < .001

Data reduction of the multiple measures per construct was achieved by creating composite scores. Confirmatory factor analyses (CFA) evaluated whether all proposed measures loaded significantly on their a priori construct, for mothers and fathers separately. Using either Time 1 or Time 2 SIP variables and Time 3 PCA risk, the CFAs confirmed each measure significantly loaded onto its proposed construct for both mothers and fathers at each time point. For the CFA of mothers’ data: T1 RMSEA = .076 [90% CI: .060, .093], SRMR = .075; T2 RMSEA = .083 [90% CI: .066, .099], SRMR = .083. For fathers: T1 RMSEA = .075 [90% CI: .056, .093], SRMR = .089; T2 RMSEA = .083 [90% CI: .066, .099], SRMR = .083. Thus, each variable’s total score was standardized, separately for mothers and fathers, and these standardized scores were combined for each construct to create composite scores for the path analyses. Composites consisted of the following scores (except the constructs of Knowledge of Nonphysical Discipline, Empathy, and Compliance Expectations, which were based on single measures): PCA Attitudes (AAPI-2 Corporal Punishment, PAV Combined, Parent-CAAM); Reactivity (FDS and NMRS); Attributions (PCV Attribution, VR Negative Attribution, ICQ Minimization, N-IAT); and PCA Risk (CTSPC Physical Assault subscale, CAPI Abuse Scale, AAPI-2, ReACCT Noncompliance, Expected PAQ Authoritarian). In addition, to consider whether SIP predicted PCA risk beyond taxes and resources, these variables were also standardized and combined to reflect taxes that would potentially collectively detract from parenting (BSI, CTS-2 Victimization, SAMISS) or serve as resources for parents (CSES, SSRI Social Satisfaction, CSI). Because income and educational level were collinear, these values were standardized for mothers and fathers separately and combined to provide an estimate of socioeconomic status (SES) as a potential demographic covariate.

Research Goal 1: Refinement of the SIP Model (T1 to T3)

Table 4 presents standardized path coefficients and fit indices for each of the estimated models, with indirect standardized path coefficients for final models. The base model for mothers demonstrated reasonable fit but pointed to a theoretically meaningful modification index—covariance between attributions and compliance expectations, both of which are Stage 2 SIP processes. This pathway was thus added in Step 1, resulting in a model with strong fit (see fit indices in Step 1 column) and a significant Wald test, Wald (df =1) = 9.44, p = .002, but with two non-significant pathways. Removing the path from PCA Attitudes to Compliance Expectations demonstrated a non-significant Wald (df =1) = 2.24, p = .135. Removing this path thus continued to demonstrate strong model fit (see fit indices under Step 2). The remaining non-significant path involved T2 Reactivity to Time 3 PCA risk. This path also demonstrated a non-significant Wald (df =1) = .03, p = .856, with no decrease in overall model fit (see fit indices Step 3 column). The final model for mothers (see top Fig. 2) indicated that lower empathy scores significantly predicted more negative child attributions and greater reactivity, and greater reactivity predicted negative attributions. PCA approval attitudes significantly predicted less knowledge of discipline alternatives and more negative attributions. T3 PCA risk was significantly predicted by more negative attributions, less knowledge of nonphysical discipline alternatives, higher compliance expectations, and stronger PCA approval attitudes. The final model R² for PCA risk was 41.2%. Note that all indirect effects proposed in this Step 3 SIP model were significant for mothers’ PCA risk (bottom of Table 3). Adding T3 SES, Taxes, and Resources to this Step 3 model increased R² to 54.0% and each significantly predicted Time 3 PCA risk (β = −27, .20, and −.19, all p ≤ .001 for SES, Taxes, and Resources, respectively). But these additions decreased model fit to unacceptable levels: χ²(22) = 91.11, p < .001, CFI = .768, RMSEA = .133 [.105, .162], SRMR = .108. This model further did not alter the pattern of findings for any significant SIP paths already identified in Step 3. Thus, Table 4 presents path models without the controls.

Table 4.

Standardized Coefficients for PCA Risk

	T1 to T3												T2 to T3
	Mothers								Fathers				Mothers		Fathers
	Base Model		Step 1		Step 2		Step 3		Base Model		Step 1		Validation		Validation
Direct Effects Paths	β	p	β	p	β	p	β	p	β	p	β	p	β	p	β	p
PCA Attitudes → PCA Risk	.19	.001	.19	.001	.19	.001	.19	.001	.32	.001	.32	.001	.26	.001	.11	.176
Attributions → PCA Risk	.42	.001	.42	.001	.43	.001	.43	.001	.27	.001	.28	.001	.40	.001	.43	.001
Compliance → PCA Risk	.13	.046	.13	.044	.13	.055	.13	.039	.20	.019	.19	.019	.14	.044	.14	.081
Knowledge → PCA Risk	−.28	.001	−.28	.001	−.28	.001	−.28	.001	−.22	.001	−.22	.001	−.34	.001	−.35	.001
Reactivity → PCA Risk	.02	.808	.02	.804	−.01	.855			.03	.705
Empathy → Reactivity	−.44	.001	−.44	.001	−.44	.001	−.44	.001	−.26	.001	−.26	.001	−.55	.001	−.45	.001
Empathy → Attributions	−.32	.001	−.35	.001	−.35	.001	−.35	.001	−.34	.001	−.34	.001	−.32	.001	−.28	.001
Reactivity → Attributions	.21	.001	.21	.001	.21	.001	.21	.001	.34	.001	.34	.001	.17	.011	.35	.001
PCA Attitudes → Attributions	.22	.001	.19	.001	.19	.001	.19	.001	.12	.094	.12	.094	.05	.375	−.00	.964
PCA Attitudes → Knowledge	−.25	.001	−.25	.001	−.25	.001	−.25	.001	−.23	.001	−.23	.001	−.24	.001	−.28	.009
PCA Attitudes → Compliance	.01	.808	.01	.804					.17	.070	.17	.070			.03	.770
Attributions ↔ Compliance			.26	.001	.25	.001	.26	.001					.28	.001
Fit Indices
X2	20.78	.023	8.18	.516	8.27	.507	8.13	.616	13.22	.153	13.43	.201	14.90	.136	17.85	.058
RMSEA	.074		.000		.000		.000		.057		.048		.052		.074
90% CI	[.027, .118]		[.000, .075]		[.000, .075]		[.000, .066]		[.000, .117]		[.000, .108]		[.000, .103]		[.000, .129]
SRMR	.055		.040		.041		.041		.051		.050		.058		.065
CFI	.949		1.000		1.000		1.000		.963		.970		.979		.939
Indirect Effects
Empathy → Attribution → PCA Risk							−.15	.001			−.10	.002	−.13	.001	−.12	.006
Empathy → Reactivity → Attribution → PCA Risk							−.04	.003			−.03	.032	−.04	.015	−.07	.002
Reactivity → Attribution → PCA Risk							.09	.001			.10	.002	.07	.015	.15	.001
PCA Attitudes → Knowledge Discipline → PCA Risk							.07	.008			.05	.010	.08	.002	.10	.005
PCA Attitudes → Attribution → PCA Risk							.08	.004			.03	.092	.02	.365	−.00	.965
PCA Attitudes → Compliance → PCA Risk											.03	.124			.01	.772

Figure 2.

Final SIP Path Models for Mothers (Top) and Fathers (Bottom) with significant standardized coefficients

For fathers, the base model demonstrated good fit (see fit indices in Table 4) but with one non-significant pathway, between T1 Reactivity to T3 PCA risk. With no suggested modification indices and one nonsignificant pathway, Step 1 confirmed the removal of that pathway given the non-significant Wald (df =1) = .14, p = .705. Two marginally significant pathways (from PCA Attitudes to Attributions and from PCA Attitudes to Compliance Expectations) resulted in marginally significant Wald statistics and thus no further modifications were made for model refinement. This final model for fathers (see bottom Fig. 2) indicated that lower empathy scores significantly predicted greater reactivity and more negative attributions, and reactivity in turn predicted negative attributions. PCA approval attitudes significantly predicted less knowledge of discipline alternatives, while only marginally predicting more negative attributions and higher child compliance expectations. Higher T3 PCA risk, in turn, was significantly predicted by more negative attributions, less knowledge of discipline alternatives, higher compliance expectations, and higher PCA approval attitudes. The final model R² for PCA risk was 35.8%. In terms of indirect effects, the hypothesized indirect pathways were supported with the notable exceptions from PCA attitudes that mirror the marginal direct effects from PCA Attitudes through Attributions to PCA risk (see bottom Step 1 for fathers). Adding T3 SES, Taxes, and Resources to the model increased R² to 47.6% of T3 PCA risk. However, only SES significantly predicted PCA risk (β = −.35, p ≤ .001), not Taxes (β = .04, p = .480) or Resources (β = −11, p = .25). The marginal path between PCA Attitudes and Attributions was further reduced (β = .10, p = .189) with these controls but other paths were unchanged. The table displays results of path models without these controls.

Research Goal 2: Validation of the SIP Model (T2 to T3)

Using Time 2 SIP predictors with the model identified in Step 3 for mothers, model fit remained good (see fit indices in Table 4). The only pathway that differed involved a reduction to non-significance between PCA Attitudes to Attributions (β = .05, p = .375), also reflected in the indirect effect from PCA Attitudes through Attributions to PCA risk no longer being significant (β = .02, p = .365), but no indication of additional paths identified through modification indices. The model continued to indicate that lower empathy scores significantly predicted greater reactivity and more negative child attributions, and reactivity predicted negative attributions. Attitudes approving of PCA significantly predicted less knowledge of discipline alternatives. Higher Time 3 PCA risk was significantly predicted by more negative attributions, less knowledge of discipline alternatives, higher compliance expectations, and stronger PCA approval attitudes. The model R² for PCA risk was 46.8%. The model including Time 3 SES and Taxes and Resources increased R² to 58.1%. SES and Taxes significantly related to greater PCA risk (β =−.23 and β = .18, p ≤ .01, respectively), with Resources only marginally associated (β = −.09, p = .098) with lower PCA risk. Validation model paths were unaltered with the inclusion of these controls.

Using the model identified in Step 1 for fathers, model fit decreased (see fit indices Table 4), with one suggested modification index between negative attributions and compliance expectations, the two SIP Stage 2 processes which were paths that were identified earlier in mothers’ models. Lower empathy scores still significantly predicted greater reactivity and more negative attributions, and reactivity still predicted negative attributions. PCA approval attitudes significantly predicted less knowledge of discipline alternatives; the two previously marginal effects from PCA attitudes to either attributions or compliance Expectations were clearly not significant paths in this model (and the indirect effects from PCA attitudes were similarly non-significant). Furthermore, PCA risk was directly predicted by negative attributions and knowledge of nonphysical discipline alternatives but PCA attitudes no longer predicted PCA risk directly and the path from compliance expectations became marginal. This model R² for PCA risk was 36.8%. The significant paths in this model were unaltered when including Time 3 SES, Taxes, or Resources; although R² increased to 46.4%, model fit further decreased. Again, only SES predicted fathers’ PCA risk (β = −.33, p ≤ .001) but not taxes (β = .05, p = .511) or Resources (β = −.10, p = .189). In sum, the model for fathers by Time 2 began to adopt similar patterns to mothers’ model (covariance between negative attributions and compliance expectations; PCA attitudes not predicting compliance expectations or attributions) but evidencing several changes from fathers’ original model.

Discussion

Developing theoretically grounded models to illustrate how parents engage in PCA represents an important direction in guiding child abuse prevention efforts. The current study examined whether SIP theory could be validated in a sample of new mothers and fathers to predict PCA risk longitudinally. Overall, findings indicated that the SIP model for mothers demonstrated considerable stability, but several associations shifted in the SIP model for fathers. Regardless, PCA risk for both mothers and fathers was predicted by pathways involving preexisting schemas as well as a sequence of SIP processes.

The first research goal (RG1) was to refine an SIP model predicting PCA risk for mothers and fathers independently. Aligned with child abuse prevention priorities (Eckenrode, 2011), the current study examined whether SIP factors assessed prenatally predict PCA risk nearly two years later. Beginning with mothers, analyses for RG1 supported many elements of the proposed theoretical model (Fig. 1), in which empathy related to overreactivity and negative child behavior attributions and PCA approval attitudes predicted more negative attributions and less knowledge of nonphysical discipline alternatives. PCA risk was also directly predicted by PCA attitudes, negative attributions, less knowledge of discipline alternatives, as well as higher expectations of child compliance following discipline. The findings echo the observations in prior work on maternal PCA risk that has identified the importance of approval of PCA (McCarthy et al., 2016; Rodriguez et al., 2011), low empathy (de Paúl et al., 2008), emotional reactivity in terms of poor emotion regulation and frustration tolerance (Hiraoka et al., 2016; McElroy & Rodriguez, 2008), negative child behavior attributions (Berlin et al, 2013; Haskett et al., 2006), high compliance expectations, and limited knowledge of nonphysical discipline alternatives (Rodriguez et al., 2016b). Compared to the theorized model, similar to prior work (Rodriguez et al., 2019), this final model in RG1 required the inclusion of one theoretically consistent covariance (between higher compliance expectations and negative child attributions—both Stage 2 processes), and the removal of one path that had previously been identified as significant (overreactivity to PCA risk directly). Otherwise, the only other path from the theoretical removed from the final model—between PCA attitudes and compliance expectations—was previously observed to be non-significant for mothers (Rodriguez et al., 2019).

To validate this final model for RG2, using data from new mothers of infants to predict their PCA risk a year later, the identified model continued to demonstrate strong fit, with only one path no longer demonstrating significance—between PCA approval attitudes and negative child attributions. Otherwise, the SIP model and its pathways demonstrated considerable stability across this transition to parenthood. For mothers, this stability may reflect stability in their schemas related to parenting during early parenthood (cf. Bernstein, Laurent, Measelle, Hailey, & Ablow, 2013).

Turning to the findings for fathers, RG1 was accomplished relatively quickly for fathers, which emerged as quite similar to that observed in earlier work (Rodriguez et al., 2019). Only a single pathway—from overreactivity to PCA risk—was removed from the starting model (as was done for mothers). Such findings are consistent with the limited work that has examined such SIP processes in men (e.g., Azar et al., 2013; Francis & Wolfe, 2008; McCarthy et al., 2016; Rodriguez et al., 2016b; Smith Slep & O’Leary, 2007). Notably, two paths were marginal in this model—from PCA approval attitudes to both negative child behavior attributions and compliance expectations. Interestingly, these two are both paths that were removed for mothers; further, when validating the model for RG2, both paths were clearly not significant for fathers using data from when their children were infants. This pattern suggests that the father model became more similar to the model for mothers after the birth of their child. Fathers may approach parenting differently than mothers (Paquette, 2004). Men may be delayed in embracing the fatherhood identity because women more prominently identify with motherhood than men do with fatherhood as new parents (Katz-Wise, Priess, & Hyde, 2010). Therefore, given that fathers’ models were relatively less stable, parenting schemas may be more fluid for fathers across the transition to parenthood.

The analyses indicated that personal challenges such as psychopathology, intimate partner violence, and substance use were associated with greater PCA risk, as previously observed (Ammerman et al., 1999; Casanueva & Martin, 2007; Stith et al., 2009), but only for mothers. Nonetheless, their combined influence simply increased the variance accounted for in explaining PCA risk and did not substantively alter the SIP models. Similarly, the findings supported that resources like social support, partner satisfaction, and coping skills were collectively related to reduced maternal PCA risk, consistent with prior research (Counts et al., 2010; Rodriguez et al., 2016b). Like taxes, however, these resources did not affect the SIP model for either mothers or fathers. These findings imply that such taxes and resources are simply independent of SIP processes, not that they are irrelevant.

Limitations and Future Directions

Despite the longitudinal design and inclusion of both mothers and fathers, a number of limitations are worth noting. Although the sample included a diverse group of parents, Hispanic/Latino parents were under-represented in this sample. Future research should further replicate how robust the findings appear for subgroups of parents based on racial and sociodemographic characteristics, which would likely require a larger longitudinal study. Given the sample size and complexity of the model, the present analyses relied on composite scores rather than latent variables, which could also be addressed with a larger sample size. Further, the study concentrated on the transition to parenthood to match the target group of many abuse prevention programs, but it is unclear if these findings apply to parenting older children. In addition, although half of the sample reflect a typical risk group for prevention services, explicit testing of whether the model holds for low-risk, at-risk, versus ultimately substantiated abusive parents would be an interesting research direction. Finally, because we focused on refining models for mothers and fathers independently, future work will need to continue to pinpoint specific mother-father differences utilizing dyadic analyses. Such dyadic analyses could simultaneously consider cross-over effects, wherein one partner affects the PCA risk of their partner (cf. Tucker, Rodriguez, & Baker, 2017).

Research Implications

Whether the stability apparent in mothers’ SIP model remains evident as their children enter preschool and beyond awaits further research inquiry given the possibility that parenting schemas may indeed shift as parents become more experienced (including experience with additional children given that our sample involved first-time mothers). In fact, parenting may activate more automatic processing over time rather than conscious processing during discipline encounters (see discussion of automatic versus the conscious processing that characterizes the SIP model; Milner, 2000). Future research should examine the extent to which PCA risk is influenced by both conscious and automatic processing because the SIP model emphasizes conscious processing. Future research should also weave emotion into the SIP model—such as the role of anger—as has been proposed for SIP models applied to other forms of aggression (cf. Lemerise & Arsenio, 2000). The SIP elements examined in the current investigation may capture only part of the picture in characterizing PCA until additional research incorporates automatic processes and emotion. In addition, the observation that the SIP model for fathers appears less stable complicates our ability to target which processes are relevant—and when—for fathers. Fathers’ taxes were also not related to their PCA risk and the selected resources were only in the expected inverse direction with PCA risk. The nascent literature on fathers, coupled with findings from the current study, underscore the need to intensify our research efforts to better appreciate what uniquely contributes to fathering and paternal PCA risk. Although the model may be viewed as sequential, the processes likely essentially contemporaneously and influence each other reciprocally—a phenomenon that represents a unique challenge for researchers to test empirically without more complex assessment strategies (e.g., momentary assessment approaches). Finally, studies that modify these SIP processes in parents explicitly would provide additional evidence into whether such efforts accomplish a reduction in PCA risk.

Prevention, Clinical, and Policy Implications

The current findings indicate that prevention and intervention programs would benefit from approaching delivery of their services inclusively with theoretically founded models like the SIP model. Comprehensive approaches minimize the likelihood that relevant parenting processes are overlooked when assessing and intervening with a parent. For both mothers and fathers, their approval of PCA contributed to their PCA risk either directly or indirectly—suggesting that prevention and intervention programs should continue attempts to adjust such attitudes (e.g., Durrant et al., 2014). The observed pathways also suggested that decreasing a parent’s approval of PCA may in turn reduce the likelihood of developing negative attributions and encourage them to consider nonphysical discipline alternatives. Alternatively, training that promotes empathy (e.g., Wilson, Havighurst, Kehoe, & Harley, 2016) could minimize negative child attributions. For other parents, negative child attributions might emerge as a central focus targeted for intervention (e.g., Bugental et al., 2010). By recognizing the connections across different processes, efforts to modify one element of the SIP model may have the added benefit of translating into adjustments at subsequent stages. The implications of stability in mothers’ SIP model is particularly promising for child abuse prevention efforts that explicitly target pregnant and perinatal mothers (Chartier et al., 2017; Eckenrode et al., 2017). For mothers, these findings suggest that assessing SIP factors either before or after birth can provide an estimate of their later PCA risk. Similar abuse prevention efforts for fathers are warranted although the processes for them remain less clear, and the reduced stability in the model for fathers suggests that relevant processes may shift across early fatherhood. For mothers, enhancing their resources and reducing the challenges that tax their ability to parent optimally remains an important adjunct when addressing SIP-relevant processes. However, the current findings suggest that focusing on taxes and resources alone would not be sufficient, and as such, efforts need to also consider the SIP processes investigated in this model. Beyond services to individual families, public awareness campaigns promoting knowledge of positive parenting techniques and nonphysical discipline options can be implemented (see Poole, Seal, & Taylor, 2014 for review). Thus, this study highlights the potential value of considering SIP elements inclusively in prevention and intervention efforts because their modification could reduce PCA risk and thereby prevent physical child abuse, a major public health concern.