The Children, Intimate Relationships, and Conflictual Life Events (CIRCLE) Interview for Simultaneous Measurement of Intimate Partner and Parent to Child Aggression

Abstract

Despite substantial rates of parent to child aggression (PCA) and intimate partner aggression (IPA) co-occurrence within families, the co-occurrence of PCA and IPA within incidents of aggression has not previously been examined. To do so, we developed the Children, Intimate Relationships, and Conflictual Life Events (CIRCLE) interview to simultaneously measure incidents of psychological and physical PCA and IPA. The CIRCLE interview was administered quarterly for approximately one year to 109 women and 94 men from 111 couples with a first born child approximately 32 months of age at study initiation. Demonstrating the CIRCLE interview’s ability to yield new knowledge about the nature of family aggression, we describe the frequency of aggressive incidents, the average number of aggressive behaviors within incidents, the daily occurrence of multiple aggressive incidents, and rates of within-incident PCA and IPA co-occurrence. With the exception of men’s physical IPA, aggression scores derived from the CIRCLE interview exhibited a relatively high degree of inter-partner reporting concordance, as well as structural validity and convergent validity with common aggression measures. Aggression reports via repeated testing were not influenced by social desirability or attempts to avoid aggression. Participants who perceived enhanced memory for aggression as a function of study participation reported increasing PCA and IPA frequencies over time. In the prediction of child conduct and emotional problems, the CIRCLE interview demonstrated predictive validity and incremental validity over traditional aggression measures. For the first time, within-incident co-occurrence of PCA and IPA was documented and shown to uniquely impact child outcomes.

Aggression within families is a common phenomenon, particularly among relatively young couples and parents of young children. For example, among a community sample of parents of children age 3 to 7, 87% of parents engaged in physical parent to child aggression (PCA) and 49% engaged in physical intimate partner aggression (IPA; Slep & O’Leary, 2005). Similar, or even higher, rates of PCA and IPA have been observed in other community samples of young families (Langhinrichsen-Rohling, Selwyn, & Rohling, 2012; Straus & Stewart, 1999). Moreover, PCA and IPA frequently co-occur within families, such that approximately 95% of parents who engage in IPA also engage in PCA (Slep & O’Leary, 2005).

The physical and mental health consequences of experiencing IPA are well-known, and often occur following exposure to psychological IPA, even in the absence of physical IPA (Coker, Davis, Arias, & Desai, 2002; Lawrence, Orengo-Aguayo, Langer, & Brock, 2012). Similarly, for children, the negative consequences of PCA are substantial (El-Sheikh, Cummings, Kouros, Elmore-Staton, & Buckhalt, 2008; Gershoff, 2002; Keiley, Howe, Dodge, Bates, & Pettit, 2001) and occur even with relatively commonplace acts such as spanking (Gershoff, 2002; Repetti, Taylor, & Seeman, 2002). At the same time, exposure to IPA is a risk factor for a wide variety of negative child outcomes (El-Sheikh et al., 2008; English et al., 2009; Saltzman, Holden, & Holahan, 2005; Ybarra, Wilkens, & Lieberman, 2007). These relations occur for both father- and mother-perpetrated IPA (El-Sheikh et al., 2008) and are more prevalent and serious with exposure to physical IPA (Kitzmann, Gaylord, Holt, & Kenny, 2003). Exposure to psychological IPA, however, also predicts children’s adverse outcomes above and beyond the effects of physical IPA (Jouriles, Norwood, McDonald, Vincent, & Mahoney, 1996; Wang, Wang, & Liu, 2016). Importantly, children exposed to both PCA and IPA exhibit more severe behavioral and emotional problems than children exposed to only one form of family aggression (McDonald, Jouriles, Tart, & Minze, 2009).

Despite high co-occurrence rates and similar theoretical perspectives, PCA and IPA are typically studied separately, with little consideration of how they function together (O’Leary, Slep, & O’Leary, 2000; Slep & O’Leary, 2001). Integration of the two fields can serve to strengthen theoretical convergence and parsimony, limit empirical redundancy, and promote study of the causes and consequences of co-occurring PCA and IPA. It is commonly thought that the co-occurrence is due to shared etiological factors (e.g., individual characteristics that cause the use of aggression in both dyads) or aggression in one dyad indirectly leading to aggression in the other dyad (e.g., the experience of IPA decreases parents’ ability to deal effectively with child misbehavior, resulting in PCA; O’Leary et al., 2000; Slep & O’Leary, 2001). However, according to family systems theory (e.g., Erel & Burman, 1995; Minuchin, 1985), aggression in one dyad may also directly (i.e., within the same incident) lead to aggression in the other dyad. To document the existence of such within-incident “spillover” of aggression, examine its potential impact on children, and permit research on the antecedents and consequences of within-incident aggression spillover, development of an instrument that simultaneously measures incidents of PCA and IPA is necessary and, thus, the focus of the current report.

The most frequently used measures of family aggression are the self-report Parent-Child Conflict Tactics Scales (CTSPC; Straus, Hamby, Finkelhor, Moore, & Runyan, 1998) and Revised Conflict Tactics Scales (CTS2; Straus, Hamby, Boney-McCoy, & Sugarman, 1996). The Conflict Tactics (CT) scales’ popularity may be owed to their relative brevity, ease of administration, focus on behaviorally-based acts, assessment of behaviors varying in severity (i.e., potential for physical injury), and, in the case of IPA, symmetry of measurement (i.e., assessment of self- and partner-perpetrated behaviors). However, these scales also have a number of limitations. For example, the CT scales’ ability to produce false positives (partly due to misidentification of some nonaggressive behaviors, such as playful wrestling, as aggressive) is recognized in recent literature (Hamby, 2016; Lehrner & Allen, 2014). Additionally, unaided self-report of behavior frequencies over a one-year referent period can be systematically biased such that reporting errors increase linearly with time since events and, although memory for events decreases with time, more personally salient events may be erroneously reported within a requested time frame (Rubin & Baddeley, 1989). These biases may be amplified by the need to remember behaviors on an act basis, rather than an event basis, as accessibility biases appear to be eliminated when participants tie the occurrence of behaviors to other events in their lives (Rubin & Baddeley, 1989). Finally, the CT scales do not provide a means to ask nuanced questions about the nature of aggression because they do not include a means to measure the immediate context (e.g., antecendents, consequences) of aggression, the sequence of behaviors within incidents, or the within-incident co-occurrence of PCA and IPA.

To address these issues and continue moving the assessment of family aggression forward, we developed an interview-based instrument that uses a repeatedly administered 13-week event history calendar methodology to simultaneously measure incidents of psychological and physical PCA and IPA. The event history calendar methodology allows interviewers to clarify whether behaviors meet definitions of aggression and allows participants to remember aggressive behaviors on an incident, rather than act, basis with incidents tied to other personally relevant events. The event history calendar methodology has previously been used to measure physical IPA (e.g., Testa, Quigley, & Leonard, 2003), but not psychological IPA or physical or psychological PCA, and only within a past-year time frame. Additionally, measurement has been confined to days of physical IPA, rather than incidents of aggression, which may occur repeatedly during a single day. Consequently, we currently do not know how frequently incidents of family aggression occur or how many aggressive behaviors typically occur during separate incidents of aggression. Importantly, we also do not know how frequently (if at all) PCA and IPA co-occur during the same aggressive incident.

In recognition of the interdependency and circuitous nature with which aggression likely occurs within families, we refer to our new instrument as the Children, Intimate Relationships, and Conflictual Life Events (CIRCLE) interview. The CIRCLE interview includes two parts: Part I measures aggressive behaviors that occurred during each incident of family aggression, and Part II measures the context of select aggressive incidents. In this first report, we focus on providing evidence for the validity of Part I of the interview. We first describe the primary types of data that the CIRCLE interview can capture, including the number of incidents of psychological and physical PCA and IPA, the range and average number of aggressive behaviors within and across incidents, and the rate of within-incident co-occurrence of PCA and IPA. To determine the utility of focusing on incidents, rather than days, of aggression, we describe how often separate incidents of family aggression occur on the same day. Next, we provide data on the reliability and validity of the CIRCLE interview aggression scores. To support inter-rater reliability, we examine the inter-partner concordance of reports of PCA and IPA from the CIRCLE interview. To support structural validity, we examine correlations across measures of psychological and physical PCA and IPA, including total number of behaviors, total number of incidents, and average number of behaviors per incident. To support convergent validity, we examine the concordance of reports of aggression obtained from the CIRCLE interview with reports on the CT scales over approximately the same time frame as well as prior to the time frame sampled by the CIRCLE interview. To support discriminant validity and to test for systematically biased reporting, we examine associations between responses to the CIRCLE interview and social desirability; we also compare such associations with those obtained for the CT scales. To examine the impact of repeated testing, we report whether social desirability, attempts to avoid aggression, or enhanced recognition or memory of aggression due to study completion affect reports of aggression over time. Finally, to support incremental and predictive validity, we examine whether aggression reports obtained from the CIRCLE interview predict child outcomes (i.e., conduct and emotional problems) above and beyond that predicted by the CT scales. We also examine whether the identification of children who have experienced within-incident co-occurrence of PCA and IPA predicts child outcomes beyond estimates of total PCA and IPA measured by the CIRCLE interview and CT scales.

Methods

Participants

Participants include 203 individuals (109 women, 94 men) from 111 heterosexual couples whose first-born child was approximately 32 months old at the beginning of the study (in 19 families only one parent participated). Most participants were married (87% of women, 84% of men) and self-identified as non-Hispanic, White (90.8% of women, 92.6% of men). On average, women were 31.02 years of age (SD = 3.81, range = 23 – 45), with 15.78 years of education (SD = 1.38, range = 12 – 17), an annual family income of $85,500 (SD = $34,429, range = $7,500 – $162,500), and worked 28.07 hours per week (SD = 19.17, range = 0 – 80). On average, men were 32.56 years of age (SD = 5.12, range = 25 – 50), with 15.43 years of education (SD = 1.75, range = 11 – 17), an annual family income of $85,698 (SD = $34,564, range = $22,500 – $162,500), and worked 43.44 hours per week (SD = 14.17, range = 0 – 80).

Procedures

Participants were originally recruited to take part in a randomized controlled trial of Family Foundations, a psychosocial prevention program for first-time expectant parents (Feinberg et al., 2016). To participate, couples had to be expecting a first child and cohabitating or married, with each parent over 18 years of age. Parents were primarily recruited from childbirth education programs and Ob/Gyn clinics in two mid-Atlantic states. Couples in the intervention group participated in nine classes from mid-pregnancy through several months post-birth focused on fostering positive coparenting (Feinberg & Kan, 2008). Couples in the control group were mailed literature on selecting quality childcare and stages of child development. In addition to other procedures, as part of the Family Foundations study, participants completed self-report measures in their homes at pretest (during pregnancy) and a follow-up assessment at child age 24 months.

After completion of the Family Foundations 24-month follow-up assessment, participants were recruited into the Children, Intimate Relationships, and Conflictual Life Events (CIRCLE) study. To be eligible, participants needed to be in a cohabitating intimate relationship, but not necessarily with their original Family Foundations study partner. Only Family Foundations participants who recently completed the 24-month follow-up assessment were recruited into the CIRCLE study. To capture a family developmental period when rates of IPA remain high and rates of PCA begin to peak (Langhinrichsen-Rohling et al., 2012; Quigley & Leonard, 1996; Straus & Stewart, 1999), we aimed to have participants begin the CIRCLE study at child age 30 months, thus measuring family aggression when children were approximately 2.5 to 3.5 years of age. With the exception of one participant who did not complete the first interview until his child was 47 months of age, the average child age at wave 1 was 31.86 months (SD = 2.70, range = 25–38).

Participants who completed at least one CIRCLE interview (n = 203) did not differ from those who declined the invitation to participate or never completed the first interview (n = 113; 49 women and 64 men from 66 couples) in terms of age, race, education, number of hours worked per week, family income, intervention group status, or social desirability. Based on data gathered at the child age 24-month follow-up assessment, participants who accepted versus declined participation also did not differ in terms of perpetration of psychological IPA, physical IPA, or physical PCA. Men who accepted versus declined participation did not differ in terms of psychological PCA, but women who participated in the CIRCLE study reported more psychological PCA than those who declined participation, t(128.80) = −3.85, p < .001, d = 0.68 (full results available upon request).

Participants were asked to complete four telephone interviews during which the CIRCLE interview was administered. Because aggression that occurred during the prior 13 weeks was assessed, interviews could not be scheduled less than 13 weeks (91 days) apart. Time lags naturally occurred between each interview such that the four interviews covered 52 nonconsecutive weeks. The median time between interviews ranged from 112 to 115 days.

After completion of the final (wave 4) CIRCLE interview, before ending the telephone call, participants provided feedback regarding their experiences with the interviews and completed via telephone the Strengths and Difficulties Questionnaire (Goodman, 1997), Revised Conflict Tactics Scales (Straus et al., 1996), and Parent-Child Conflict Tactics Scales (Straus et al., 1998). Participants were paid $40 for each interview, plus a $40 bonus for completion of all four interviews.

Non-identifiable data collection

To facilitate valid self-reports, particularly of physical PCA, we utilized a system for non-identifiable data collection. In brief, at the beginning of each interview, interviewers were given a piece of paper that included the participant’s phone number and gender. As soon as the interview began (using a telephone that did not store numbers of placed calls), the paper was shredded. The interviewer manager immediately recorded completion of the interview (using a two-week interval rather than a date), deleted a 20-digit ID number from the calendar of scheduled interviews, and mailed the participant’s payment without observation of the participant’s name or address. During the interviews, participants answered three personal, but non-identifying, questions used to merge their data across the four interviews. We obtained no redundancies across participants, thus allowing us to link each interview within each participant. Participants answered the same three questions about their partners, which allowed us to link all partners within couples. Participants were informed of the conditions under which they might be identified (e.g., the first participants to complete the study). When abusive behaviors were reported, interviewers urged participants to seek services and offered to assist in the process. Participants were reminded of available services after each interview, both during the telephone calls and in the form of local resource lists included with payments.

Attrition

From the 203 participants who completed wave 1 interviews, 181 completed wave 2 interviews (11% attrition from wave 1), 160 completed wave 3 interviews (10% attrition from wave 2), and 133 completed wave 4 interviews (13% attrition from wave 3), resulting in a total attrition rate of 34%. Only three participants (two from the same couple) actively withdrew from the study; the remaining cases of attrition were a function of repeatedly not being able to contact participants to schedule or complete interviews. The higher rate of attrition at wave 4 was partly due to a strict end date for data collection. In total, 677 interviews were completed.

Participants who completed all four CIRCLE interviews did not differ from those who did not complete all interviews in terms of age, race, education, income, number of hours worked per week, intervention group status, social desirability, or perpetration of psychological or physical PCA or IPA at child age 24 months (full results available upon request).

Measures

Demographic characteristics

During the first Family Foundations study assessment (i.e., during pregnancy), participants reported on personal characteristics such as race, ethnicity, and years of education. During the child age 24 months assessment of the Family Foundations study, participants again reported on their age, average hours spent working outside the home, and family income. Family income was measured on an ordinal scale recoded to the midpoint of the range (e.g., $50,000–$54,999 was recoded to $52,500) to create a continuous scale with a potential range of $0 – $162,500.

Social desirability

During the first Family Foundations study assessment, participants completed Reynold’s (1982) 13-item (true/false) short form C of the Marlowe-Crowne Social Desirability Scale (MCSD; Crowne & Marlowe, 1960). The MCSD and Reynold’s short forms are widely used, and their reliability, validity, and factor structure have been supported using a variety of approaches (Loo & Thorpe, 2000). In the current study, ∝ = .68 for men and ∝ =.66 for women.

Parent-Child Conflict Tactics Scales (CTSPC)

As a measure of PCA, the CTSPC (Straus et al., 1998) includes a nonviolent discipline subscale as well as a 13-item physical assault subscale that ranges from spanked him/her on the bottom with a bare hand to threatened him/her with a knife or gun and a 5-item psychological aggression subscale that ranges from threatened to spank or hit him/her but did not actually do it to said you would send him/her away or kick him/her out of the house. For each item, participants indicate how many times in the past year they engaged in the behavior on a 7-point response scale: Never, once, twice, 3–5 times, 6–10 times, 11–20 times, and more than 20 times. The midpoint of each range is used and more than 20 times is recoded as 25 times. The subscales have a reasonable degree of internal consistency and factorial validity (Straus et al., 1998), as well as convergent validity across a wide range of measures (e.g., Gershoff, 2002). During the Family Foundations study, the CTSPC was self-administered on paper, excluding two items assessing nonviolent discipline and the six most severe physical assault items (due to expected low frequency of endorsement). As such, it yielded ∝ = .66 and ∝ = .60, respectively, for men’s and women’s psychological PCA and ∝ = .80 and ∝ = .75, respectively, for men’s and women’s physical PCA. The full physical assault and psychological aggression subscales were administered via telephone after the final CIRCLE interview and yielded ∝ = .41 and ∝ = .39, respectively, for men’s and women’s psychological PCA and ∝ = .61 and ∝ = .64, respectively, for men’s and women’s physical PCA.

Revised Conflict Tactics Scales (CTS2)

As a measure of IPA, the CTS2 (Straus et al., 1996) includes physical assault, psychological aggression, sexual coercion, injury, and negotiation subscales. For each of 39 items, participants indicate how many times in the past year they or their partners engaged in the behavior on a 7-point response scale: Never, once, twice, 3–5 times, 6–10 times, 11–20 times, and more than 20 times. The midpoint of each range is used and more than 20 times is recoded as 25 times. The 12-item physical assault subscale ranges from threw something at my partner that could hurt to used a knife or gun on my partner, and the 8-item psychological aggression subscale ranges from insulted or swore at my partner to threatened to hit or throw something at my partner. These subscales have good test-retest reliability (O’Leary & Williams, 2006; Vega & O’Leary, 2007), factorial validity (Connelly, 2005), and convergent validity across a wide range of measures (e.g., Schumacher, Feldbau-Kohn, Smith Slep, & Heyman, 2001a; Schumacher, Slep, & Heyman, 2001b; Straus, 2004). During the Family Foundations study, the entire measure was self-administered on paper and yielded ∝ = .66 and ∝ = .99, respectively, for men’s and women’s psychological IPA and ∝ = .99 and ∝ = .94, respectively, for men’s and women’s physical IPA. After the final CIRCLE interview, the physical assault and psychological aggression subscales were administered via telephone and yielded ∝ = .70 and ∝ = .62, respectively, for men’s and women’s psychological IPA and ∝ = .70 and ∝ = .50, respectively, for men’s and women’s physical IPA.

Children, Intimate Relationships, and Conflictual Life Events (CIRCLE) interview

The CIRCLE interview uses an event history calendar methodology to measure incidents of psychological and physical PCA and IPA. During Part I of each telephone-administered interview, participants note on a provided calendar the dates being assessed (the prior 13 weeks), days of personal significance (e.g., holidays, illnesses, birthdays), and days of no face-to-face contact with their partner and/or child. Interviewers record the same days and events to help administer the interviews. Using these events to aid memory, participants work backwards in time to review all incidents of self- and partner-perpetrated psychological and/or physical aggression that occurred during the prior 13 weeks. Participants are provided a list of aggressive behaviors from the CTSPC (Straus et al., 1998) and CTS2 (Straus et al., 1996) and are urged to consider additional aggressive acts not listed. For each aggressive incident, they report the topic of the conflict (if any) as well as the order, victim, and perpetrator of each aggressive behavior. For acts of partner-perpetrated PCA, participants indicate whether they were present to observe the behavior. Incidents are defined as terminated if emotional arousal and conflict behaviors returned near to baseline for one hour. Participants also identify “regular patterns” of aggression and, if present, report the usual conflict topic, order and perpetrator/victim of each aggressive behavior, and frequency of the pattern (with start and end dates if it did not occur continuously).

For Part II of the interview, a computer algorithm is used to select three incidents for review of the immediate context, antecedents, and consequences of aggression. For the current study, this algorithm prioritized the most severely aggressive incidents and incidents in which both PCA and IPA occurred while balancing incidents of PCA with incidents of IPA. Participants who identify at least three aggressive incidents also identify their most stressful or upsetting nonaggressive conflict. Participants who do not identify three incidents of aggression identify their most stressful or upsetting nonaggressive conflicts such that all participants review four incidents, at least one of which was nonaggressive. Data from Part II of the CIRCLE interview are not included in the current report.

Administration of both parts of the interview and debriefing takes approximately 45 minutes. Interviewers (clinical psychology graduate students) were trained to administer the interview and respond to participant distress and disclosures using a behaviorally-based manual and interview protocol, role plays, group discussions, and live interviews with pilot participants.

Impact of interview completion

At the end of the final CIRCLE interview, participants were asked the following questions: 1) Over the course of completing this study’s four telephone interviews, did you find yourself recognizing or remembering more incidents of aggression in your family? [If yes:] Do you think this was due to answering the interview questions every three months?, and 2) Over the course of completing this study’s four telephone interviews, did you find yourself putting in more effort to prevent or avoid incidents of aggression in your family? [If yes:] Do you think this was due to answering the interview questions every three months?

Child outcomes

The “early years” Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997) was administered via telephone to parents after the final CIRCLE interview. The SDQ includes five 5-item scales with wording appropriate for 2–4 year old children intended to screen for prosocial behavior, hyperactivity, and peer, conduct, and emotional problems. It has demonstrated convergent validity, moderate sensitivity, and good specificity (Goodman, Ford, Simmons, Gatward, Meltzer, 2000; Goodman & Scott, 1999). In this report, we focus on the conduct and emotional problems subscales, which yielded ∝ = .67 and ∝ = .42, respectively.

Statistical Analyses

Total aggression scores were adjusted for missing interviews using mean replacement and log transformed to reduce statistical skew. In most analyses, men and women’s data were analyzed separately; in multi-level models, main effects of gender were included. With the exception of tests of inter-partner concordance, we focus on participants’ reports of their own perpetration because 1) not all participants’ partners completed the study and, among those who did, interview dates (thus the days assessed) oftentimes did not converge across partners, and 2) reports of partners’ PCA perpetration are not commonly examined in existing literature, partly due to concerns that such reports would provide a limited, and potentially biased, representation of the overall aggression that has occurred if the participant was not present during the incidents.

To describe the data that emerged from the CIRCLE interview and to examine convergent, discriminant, and structural validity, we calculated descriptive statistics, bivariate correlations, partial correlations (i.e., to adjust for time between CIRCLE interviews and, when relevant, time between completion of the pre-CIRCLE CT scales and the CIRCLE study), comparisons of correlations, and group comparisons (i.e., t-tests, chi-square analyses).

To examine inter-partner concordance of reports, we calculated percent agreement (among all couples and, to avoid inflation of agreement due to aggression nonoccurrence, among only those couples in which at least one partner reported the occurrence of the indicated form of aggression), kappa, and Yule’s Y (Spitznagel & Helzer, 1985). Kappa and Yule’s Y are less influenced by chance than percent agreement; however, incorrect estimates may arise when base rates are skewed. The likelihood of an incorrect estimate is smaller for Yule’s Y than kappa (Hoffmann & Ninonuevo, 1994). Both of these measures are interpreted on a scale from −1 to +1, with 0 indicating agreement at the level of chance. They can generally be interpreted as follows: 0.0 = poor, 0.0 − 0.2 = slight, 0.2 – 0.4 = fair, 0.4 – 0.6 = moderate, 0.6 – 0.8 = substantial, and 0.8 – 1.0 = perfect (Hoffman & Ninonuevo, 1994; Landis and Koch, 1977). We additionally used bivariate correlations to compare partner reports of aggressive behavior frequencies.

To examine effects of repeated testing, in a series of multilevel models we tested the main effects of social desirability, avoidance of aggression, and enhanced memory of aggression due to study completion, as well as their interactions with time, to predict participants’ total number of aggressive behaviors and average number of aggressive behaviors per incident (separately for PCA and IPA). Due to a great degree of redundancy between participants’ number of aggressive behaviors and number of aggressive incidents, we did not predict participants’ total number of aggressive incidents. Participants’ number of psychologically and physically aggressive behaviors were combined in the multilevel models in order to reduce the number of analyses conducted and because such behaviors are commonly conceptualized as on a single continuum of aggression severity (e.g., El-Sheikh et al., 2008). Models were structured whereby repeated measures were nested within family dyads. We additionally included participants’ gender, Family Foundations treatment condition, and CIRCLE interview wave (i.e., “time”) in each model.

To examine the predictive validity of the CIRCLE interview PCA and IPA scores and identification of children exposed to within-incident spillover of aggression, as well as the incremental validity of these measures beyond that provided by the CT scales, we tested two multilevel models in which these variables were used to predict child conduct problems and emotional problems. Gender of the reporting parent, Family Foundations treatment condition, social desirability, and child age were included in each model. In addition to examining main effects, we tested the moderating effect of aggression spillover exposure on the associations between CIRCLE interview measured aggression and child outcomes.

Results

Descriptive Information from the CIRCLE Interview

Rates, frequencies, and incidents of aggression

As displayed in Table 1, across interviews, parent to child aggression (PCA) commonly occurred in this sample, such that 83% of mothers and 80% of fathers reported engaging in psychological PCA and 59% of mothers and 50% of fathers reported engaging in physical PCA. Of those who engaged in physical PCA, 13% of mothers and 11% of fathers engaged in severe physical PCA. Rates of psychological intimate partner aggression (IPA) were similar, such that 80% of women and 79% of men reported engaging in such aggression. However, only 6% of women and 4% of men reported engaging in physical IPA. Because only 11 participants reported physical IPA, subsequent results regarding physical IPA should be considered examples not necessarily representative of the population.

Table 1.

Descriptive information derived from the CIRCLE interview.

Type of aggression	N (% of sample)	Behaviors/Year M (SD, range)	Incidents/Year M (SD, range)	Behaviors/Incident M (SD, range)
Mothers’ psychological PCA	91 (83.49%)	85.43 (162.43, 1 – 1217)	45.46 (66.99, 1 – 337)	1.73 (0.94, 1 – 6)
Mothers’ physical PCA	64 (58.72%)	36.39 (123.13, 1 – 929)	27.13 (77.69, 1 – 581)	1.15 (0.37, 1 – 2.73)
Fathers’ psychological PCA	75 (79.79%)	79.11 (141.03, 1 – 724)	52.34 (84.95, 1 – 507.50)	1.38 (0.62, 1 – 4.02)
Fathers’ physical PCA	47 (50.00%)	29.16 (67.16, 1 – 393.75)	23.18 (45.60, 1 – 212.50)	1.14 (0.34, 1 – 2.50)
Women’s psychological IPA	87 (79.82%)	41.30 (46.45, 1 – 210)	18.67 (20.87, 1 – 115)	2.24 (1.77, 1 – 10.75)
Women’s physical IPA	7 (6.42%)	1.64 (1.18, 1 – 4)	2.86 (2.67, 1 – 8)	1.00 (0.00, 1 – 1)
Men’s psychological IPA	74 (78.72%)	27.91 (61.78, 1 – 468)	10.76 (18.64, 1 – 139)	2.24 (2.34, 1 – 17.50)
Men’s physical IPA	4 (4.26%)	5.50 (4.65, 1 – 12)	3.50 (1.73, 1 – 5)	1.50 (1.00, 1 – 3)

Note. Scores adjusted for missing interviews, thus sometimes resulting in fractions of behaviors. Similarly, behaviors/incident represents the average number of aggressive behaviors per aggressive incident that each participant reported perpetrating, thus sometimes resulting in fractions of behaviors. M, SD, and range calculated among those who reported the indicated type of aggression.

PCA = Parent to child aggression; IPA = Intimate partner aggression.

The CIRCLE interview captured very high behavior counts, with the number of behaviors ranging up to 1,217 acts of mothers’ psychological PCA. Additionally, some participants reported that incidents of PCA occurred very frequently, ranging up to 581 incidents of mothers’ physical PCA. Among those who reported any such aggression, on average, mothers and fathers, respectively, reported engaging in physical PCA during 27 and 23 separate incidents. Although often co-occurring with physical PCA, on average, mothers and fathers, respectively, also reported engaging in 45 and 52 incidents of psychological PCA. IPA, particularly physical IPA, was reported to occur less frequently, but the CIRCLE interview still captured high frequencies among some participants, including up to 139 incidents of men’s psychological IPA. Finally, although aggression occurred frequently in this sample, on average, few aggressive behaviors occurred during each incident of aggression. Still, some participants reported typically engaging in many aggressive behaviors during each incident, including one man who averaged more than 17 acts of psychological IPA per aggressive incident.

Structural validity

Due to the high frequency of aggressive incidents that included only one aggressive behavior, measures of the number of aggressive behaviors and number of aggressive incidents were highly inter-correlated within genders for each type of aggression (ranging from r = .94 to r = 1.00). Men’s and women’s average numbers of aggressive behaviors perpetrated per incident were relatively more moderately correlated with their total number of aggressive behaviors and number of aggressive incidents, and such correlations were more modest for psychological aggression (ranging from r = .60 to r = .85) than physical aggression (ranging from r = .72 to r = .97). Measures of psychological and physical PCA were consistently moderately correlated (ranging from r = .30 to r = .49), while measures of psychological and physical IPA were not consistently significantly correlated (ranging from r = .10 to r = .23).

Days of multiple aggressive incidents

At least one day during which multiple separate aggressive incidents occurred was reported by 38% (n = 77) of participants. Most (58%, n = 45) of these participants reported that this occurred only once, though the frequency ranged up to 8 times during the reporting period. Twelve participants reported that more than two incidents of aggression occurred on a single day, ranging up to five separate incidents in a single day.

Rates of within-person and within-incident co-occurrence of PCA and IPA

Most participants (72% of women, 70% of men; n = 145) reported perpetrating both PCA and IPA, though not necessarily within the same aggressive incident. Of participants who engaged in IPA, 90% engaged in PCA. Of participants who engaged in PCA, 83% engaged in IPA. Among all participants, 17% (n = 34) reported at least one instance of within-incident co-occurrence of PCA and IPA. At least one partner within 24% (n = 27) of families reported the occurrence of within-incident co-occurrence of PCA and IPA.

Inter-Partner Concordance of Aggression Reports

Given attrition and differential scheduling of interviews, on average, partners reported on 236 (SD = 98.64, range = 33 – 364) of the same days. As displayed in Table 2, inter-partner concordance regarding the occurrence/non-occurrence of psychological PCA and IPA was fair to moderate. Inter-partner behavior frequency correlations ranged from r = .25 to r = .37 for psychological PCA and from r = .49 to r = .55 for psychological IPA. For physical PCA, agreement was similarly fair to moderate, with inter-partner behavior frequencies correlating r = .60 and r = .50, respectively, for mothers’ and fathers’ physical PCA. For women’s physical IPA, agreement was moderate to substantial, with an inter-partner behavior frequency correlation of r = .47. Couples typically agreed regarding the absence of men’s physical IPA, but no partners agreed regarding the occurrence of such aggression. With the exception of physical IPA, partner disagreement typically included individuals reporting their own aggression when their partners did not make such reports; this was especially the case for PCA.

Table 2.

Inter-partner concordance of reports on the CIRCLE interview.

Type of aggression	Aggression Occurrence/Nonoccurrence						# Behaviors
	% agreement (n)	% occurrence agreement (n)a	% ♀, not ♂ (n)b	% ♂, not ♀ (n)c	Kappa	Yule’s Y	r
Mothers’ psychological PCA	60.9% (56)	55.6% (45)	39.5% (32)	4.9% (4)	.18	.33	.25*
Mothers’ physical PCA	57.6% (53)	29.1% (16)	67.3% (37)	3.6% (2)	.22	.48	.60***
Fathers’ psychological PCA	68.5% (63)	63.7% (51)	8.8% (7)	27.5% (22)	.26	.33	.37***
Fathers’ physical PCA	67.4% (62)	44.4% (24)	16.7% (9)	38.9% (21)	.34	.37	.50***
Women’s psychological IPA	78.3% (72)	75.6% (62)	15.9% (13)	8.5% (7)	.37	.45	.55***
Women’s physical IPA	93.5% (86)	33.3% (3)	33.3% (3)	33.3% (3)	.47	.68	.47***
Men’s psychological IPA	73.9% (68)	71.1% (59)	13.3% (11)	15.7% (13)	.26	.32	.49***
Men’s physical IPA	92.4% (85)	0.0% (0)	57.1% (4)	42.9% (3)	−.04	−1.00	−.03

Note. PCA = Parent to child aggression; IPA = Intimate partner aggression; ♀ = female; ♂ = male.

^aPercent of couples in which both partners reported each form of aggression among only those couples in which at least one partner reported the indicated form of aggression.

^bPercent and number of aggressive couples in which the female, but not male, partner reported each form of aggression.

^cPercent and number of aggressive couples in which the male, but not female, partner reported each form of aggression.

^*p < .05

^***p < .001

Convergent Validity with the CT Scales

The number of psychological PCA, physical PCA, and psychological IPA behaviors reported during the CIRCLE interview correlated strongly with that reported on the CT scales at the end of the final CIRCLE interview (women: r = .72 – .82; men: r = .70 – .72). Demonstrating expected variability in aggressive behavior over time, frequencies of such behaviors reported during the CIRCLE interviews correlated significantly less strongly with the pre-CIRCLE study CT scales (women: r = .41 – .57; men: r = .45 – .50; z values ranged from 1.83 to 4.67). Reports of physical IPA during the CIRCLE interview did not correlate consistently with CTS2 reports at the end of the final CIRCLE interview (women: r = .22, p = .066; men: r = .94, p < .001) or prior to the CIRCLE study (women: r = .16, p = .200; men: r = −.02, p = .881). This is likely because physical IPA was reported by only 6 women and 4 men on the pre-CIRCLE CTS2, 7 women and 4 men during the CIRCLE interview, and 9 women and 4 men on the post-CIRCLE CTS2.

Table 3 presents the degree of concordance in reports of the occurrence or non-occurrence, as well as the number of aggressive behaviors, across the CIRCLE interview and post-CIRCLE study CT scales. For each form of aggression, when used to classify participants as aggressive or not, the CIRCLE interview and CT scales very frequently agreed, with the exception of physical IPA perpetration. In cases of disagreement, the CIRCLE interview identified more cases of mothers’ physical PCA than did the CTSPC, but the CT scales identified more cases of psychological PCA and psychological and physical IPA. Participants identified as aggressive by the CT scales, but not the CIRCLE interview, typically reported less psychological IPA perpetration on the CTS2 than did other participants, t(19.69) = 5.45, p < .001, d = 2.46. In contrast, participants identified as aggressive by the CT scales, but not the CIRCLE interview, typically reported more physical IPA perpetration on the CTS2 than did other participants, t(8.31) = −4.96, p = .001, d = 3.44. Finally, as displayed in Table 3, compared to the CT scales, the CIRCLE interview yielded higher behavior counts of mothers’ psychological PCA, but differences on other scales were statistically nonsignificant.

Table 3.

Concordance of reports between the CIRCLE interview and Conflict Tactics (CT) scales.

Type of aggression	Aggression Occurrence/Nonoccurrence				Behavior Counts
	All % agreement (n)	Aggressive % agreement (n)	% CIRCLE, not CT scale (n)a	% CT scale, not CIRCLE (n)b	t (df)	d
Mothers’ psychological PCA	97.3% (71)	97.2% (69)	0.0% (0)	2.8% (2)	2.30 (72)*	0.54
Mothers’ physical PCA	86.3% (63)	80.8% (42)	13.5% (7)	5.8% (3)	1.08 (72)	0.24
Fathers’ psychological PCA	90.7% (49)	90.2% (46)	2.0% (1)	7.8% (4)	−0.68 (53)	−0.19
Fathers’ physical PCA	74.1% (40)	65.0% (26)	17.5% (7)	17.5% (7)	−0.63 (53)	−0.17
Women’s psychological IPA	87.7% (64)	87.1% (61)	2.9% (2)	10.0% (7)	1.76 (72)	0.41
Women’s physical IPA	86.3% (63)	16.7% (2)	25.0% (3)	58.3% (7)	−1.95 (72)	−0.46
Men’s psychological IPA	90.7% (49)	90.0% (45)	2.0% (1)	8.0% (4)	0.44 (53)	0.12
Men’s physical IPA	96.3% (52)	50.0% (2)	0.0% (0)	50.0% (2)	−1.75 (53)	−0.48

Note. CT scales include the Parent-Child Conflict Tactics Scales and Revised Conflict Tactics Scales psychological aggression and physical assault scales. Positive t-values indicate that more aggression was reported in the CIRCLE interviews than on the CT scale.

^aPercent and number of aggressive participants who reported aggression in the CIRCLE interviews, but not on the indicated CT scale.

^bPercent and number of aggressive participants who reported aggression on the indicated CT scale, but not in the CIRCLE interviews.

PCA = Parent to child aggression; IPA = Intimate partner aggression.

^*p < .05

Impact of Social Desirability and Repeated Testing on Aggression Reports

Social desirability was consistently nonsignificantly correlated with men’s and women’s reports of psychological and physical PCA and IPA reported during the CIRCLE interviews and the post-CIRCLE study CTS2 (range: r = −.17 to r = .07). The correlations did not differ significantly between reports provided via the CIRCLE interviews and the CTS2 with one exception: Social desirability differentially correlated with men’s reports of psychological IPA on the CTS2 (r = −.16) compared to the CIRCLE interviews (r = .03), z = 1.86, p = .032.

Most participants (i.e., 79.5% of wives and 74.5% of husbands) reported recognizing or remembering more aggression over the course of the study period. Of these, 94.8% of wives and 100% of husbands attributed their enhanced memory to completing the CIRCLE interviews. Additionally, 54.8% of wives and 52.7% of husbands reported that they tried to avoid aggression over the course of the study period. Of these, 90.2% of wives and 73.3% of husbands attributed their avoidance of aggression to completing the CIRCLE interviews.

In multilevel regression models (Table 4), the main effects of time, social desirability, avoidance of aggression, and enhanced remembering of aggression were not significantly associated with reports of the number of PCA or IPA behaviors or the average number of PCA or IPA behaviors per incident, with two exceptions: 1) Social desirability was negatively associated with reports of the number of PCA behaviors reported, and 2) the number of IPA behaviors decreased over time. Attempts to avoid aggression did not interact with time to significantly predict the number of PCA or IPA behaviors or the number of PCA or IPA behaviors per incident. In contrast, enhanced memory for aggression interacted with time to significantly predict the number of PCA and IPA behaviors, indicating that repeated testing was associated with increased reports of PCA and IPA over time among participants who reported enhanced recognition of, or memory for, aggression as a function of study completion.

Table 4.

Multilevel regression models testing the impact of social desirability, avoidance of aggression, and enhanced memory for aggression due to study completion on aggression reports.

	PCA Number Behaviors			PCA Number Behaviors Per Incident			IPA Number Behaviors			IPA Number Behaviors Per Incident
	B	SE	t	B	SE	t	B	SE	t	B	SE	t
Time	.03	.09	0.31	−.02	.03	−0.63	−.15	.07	−2.00*	−.05	.04	−1.10
Social Desirability (SD)	−.08	.04	−2.29*	−.01	.01	−1.12	−.04	.02	−1.54	−.02	.01	−1.65
SD × Time	.01	.01	0.47	.01	.01	0.52	.01	.01	0.47	.00	.01	0.31
Avoid Aggression (AA)	.07	.22	0.32	.09	.06	1.54	−.05	.15	−0.32	−.02	.08	−0.22
AA × Time	−.10	.08	−1.20	.01	.03	0.44	−.05	.07	−0.75	−.01	.04	−0.22
Remember Aggression (RA)	.27	.25	1.08	.05	.07	0.71	.21	.17	1.25	.15	.09	1.64
RA × Time	.20	.10	2.02*	.04	.03	1.19	.21	.08	2.48*	.07	.05	1.40

Note. All models include controls for the effects of gender and Family Foundations treatment condition. PCA and IPA scores include psychologically and physically aggressive behaviors measured during the CIRCLE interview.

PCA = Parent to child aggression; IPA = Intimate partner aggression.

^*p < .05

The dearth of statistically significant effects does not appear to be due to over-fitting of models. When excluding social desirability and the interaction between social desirability and time from the models, the strength of the effects for avoidance of aggression, enhanced remembering of aggression, and their interactions with time remained essentially unchanged. When excluding avoidance of aggression, enhanced remembering of aggression, and their interactions with time, the effects of social desirability and the interactions between social desirability and time were consistently nonsignificant (results available upon request).

Incremental and Predictive Validity of CIRCLE Interview Aggression Estimates and Identification of Within-Incident Aggression Spillover

As displayed in Table 5, when allowing the CT scales and CIRCLE measures of PCA and IPA to compete in a single model, the CT scales did not significantly predict children’s conduct problems, while the CIRCLE measure of PCA significantly predicted more severe conduct problems. Although the main effect of the CIRCLE measure of IPA was not statistically significant, it interacted with children’s exposure to within-incident spillover of aggression to significantly predict children’s conduct problems. That is, among children with a history of experiencing within-incident co-occurrence of PCA and IPA, the severity of their parents’ IPA as measured by the CIRCLE interview predicted more severe conduct problems. These effects did not generalize to the prediction of children’s emotional problems. However, unlike the model predicting conduct problems, children’s exposure to within-incident spillover of aggression significantly predicted more severe emotional problems, above and beyond the main effects of PCA and IPA as measured by both instruments. When spillover status and interactions with spillover status were removed from the models, the comparative effects of the CT scales and CIRCLE interview remained essentially the same, though the incremental validity of the CIRCLE interview aggression scores appeared slightly stronger as statistical trends emerged for CIRCLE measured IPA to positively predict conduct problems (t = 1.81, p = .072) and CIRCLE measured PCA to positively predict emotional problems (t = 1.66, p = .099).

Table 5.

The impact of within-incident spillover of aggression and total PCA and IPA measured by the CT scales and CIRCLE interviews on child outcomes.

	Conduct Problems			Emotional Problems
	B	SE	t	B	SE	t
CTSPC PCA	−.65	.39	−1.68	.48	.38	1.26
CTS2 IPA	.25	.44	0.56	.00	.32	0.01
Spillover	.24	.43	0.56	.61	.30	2.02*
CIRCLE PCA	.64	.29	2.22*	.03	.18	0.16
CIRCLE PCA × Spillover	−.33	.36	−0.92	.29	.29	1.00
CIRCLE IPA	.40	.24	1.67	−.15	.23	−0.63
CIRCLE IPA × Spillover	.99	.49	2.04*	−.02	.54	−0.03

Note. Models include controls for the effects of parent gender, Family Foundations treatment condition, child age, and social desirability. Spillover represents whether either parent ever reported that the child experienced within-incident spillover of PCA and IPA. PCA and IPA scores include psychologically and physically aggressive behaviors.

CTSPC = Parent-Child Conflict Tactics Scales; CTS2 = Revised Conflict Tactics Scales; PCA = Parent to child aggression; IPA = Intimate partner aggression.

^*p < .05

Discussion

To address the substantial rates of PCA and IPA co-occurrence within families (Slep & O’Leary, 2005) and to support the theoretical, empirical, and treatment implications of studying them together (O’Leary et al., 2000; Slep & O’Leary, 2001), we aimed to develop an instrument to simultaneously measure incidents of psychological and physical PCA and IPA. In particular, our goal was to develop an instrument that would estimate rates and frequencies of PCA and IPA at least as well as the most commonly used measures of family aggression (the CT scales), while additionally measuring the within-incident co-occurrence (“spillover”) of PCA and IPA and providing a means to examine the context of aggressive incidents. Reflecting the circuitous nature with which aggression occurs in families, we refer to this instrument as the Children, Intimate Relationships, and Conflictual Life Events (CIRCLE) interview.

Initial construct validity of the CIRCLE interview aggression estimates was supported in several ways. First, reports of psychological and physical PCA and IPA (including total number of behaviors, total number of incidents, and average number of behaviors per incident) generally intercorrelated as expected, supporting structural validity. Second, consistent with prior findings of only small negative correlations between social desirability and reports of IPA on the CTS2 (Bell & Naugle, 2007; Straus, 2004), IPA and PCA reports during the CIRCLE interview did not exhibit meaningful associations with social desirability, supporting discriminant validity and suggesting that reports were not systematically biased in this way. Third, the CIRCLE interview exhibited convergent validity with the CTSPC and CTS2 such that, across measures, aggression frequencies mostly correlated highly when assessing aggression over a similar time frame and less strongly when assessing aggression during separate time frames.

The CIRCLE interview differs from other measures of family aggression in that four quarterly assessments are used to estimate rates and frequencies of aggression over approximately one year. As such, we considered both potentially useful and detrimental effects of repeated testing. We found that social desirability and attempts to avoid aggression due to study completion were not associated with mean levels or changes in reports of aggression over time. Thus, it appears that the potential influence of social desirability did not change as participants conceivably became more comfortable with completing the interviews. Similarly, despite reported attempts to avoid aggression due to study completion, the structure of repeatedly completing CIRCLE interviews did not appear to change engagement in future aggression. In contrast, participants who reported that completion of multiple CIRCLE interviews enhanced their recognition of, and memory for, aggression reported more PCA and IPA over time. Thus, repeated testing positively impacted some participants’ ability to accurately report aggression. Although repeated assessments over a relatively short time frame add complexity to study protocols, increase participant burden, and potentially result in higher attrition rates, in conjunction with results suggesting incremental validity of the CIRCLE interview aggression estimates over aggregate measures of past-year aggression, these results suggest that repeated assessments may be a worthwhile means of obtaining the most precise data possible.

The CIRCLE interview also generally demonstrated a sufficient degree of inter-partner concordance of aggression reports, at least compared to that previously reported for the CT scales. Only one known examination of parents’ concordance of reports of PCA exists (Lee, Lanford, Pettit, Bates, & Dodge, 2012). Compared to that report, the CIRCLE interview yielded a higher degree of concordance for reports of mothers’ and fathers’ psychological PCA and mothers’ physical PCA. Although the CIRCLE interview yielded a lower degree of concordance for frequency of fathers’ physical PCA (i.e., r = .50 as compared to r = .60), the degree of concordance was adequate. Inter-partner concordance of IPA reports on the CTS2 has been examined much more extensively. With the exception of men’s physical IPA, the CIRCLE interview performed in the high range of what has previously been observed (e.g., LaMotte, Taft, Weatherill, Panuzio, & Eckhardt, 2014; Marshall, Panuzio, Makin-Byrd, Taft, & Holtzworth-Munroe, 2011; O’Leary & Williams, 2006). Although inter-partner concordance of aggression reports was not substantial (a common observation that in many ways has puzzled family researchers), results indicating that the CIRCLE interview yielded a degree of inter-partner concordance in line with, or higher than, the CT scales (with the exception of men’s physical IPA) is notable given that, on average, partners reported on only 65% of the same days during the CIRCLE interviews. In addition, when partners disagreed about the occurrence of aggression, it was generally in the direction of individuals reporting their own aggression when their partners did not make such reports. Thus, lack of concordance does not appear to simply be due to the common concern of perpetrators underreporting their own aggression.

Given that the CT scales are widely used and easy to administer, it was important to directly examine the incremental validity of the CIRCLE interview’s aggression scores beyond those of the CT scales. Compared to the CT scales, the CIRCLE interview yielded significantly higher frequencies of women’s psychological PCA perpetration. This is likely partly a function of the CIRCLE interview yielding precise estimates of the number of behaviors that occurred, whereas on the CT scales, each type of behavior (e.g., threatened to spank) receives a maximum past-year frequency estimate of 25 times. Therefore, participants’ scores on the CT scales are capped based on the number of types of behaviors they engaged in. At the same time, when used to identify aggressive versus nonaggressive individuals, the CIRCLE interview and the CT scales very frequently agreed; when they disagreed, with the exception of mothers’ and fathers’ physical PCA, the CT scales identified more participants as aggressive. Because the CT scales were administered in the same manner as the CIRCLE interview (i.e., via telephone) and social desirability was not significantly associated with aggression reports during the CIRCLE interview, it does not appear that participants purposefully avoided reporting aggression to a greater extent during the CIRCLE interviews than on the CT scales. Instead, three alternative explanations are plausible. First, participants may not interpret relatively low frequency behaviors as “aggressive,” thus not reporting them during the CIRCLE interview, which is openly presented to participants as a measure of aggression. This may be particularly the case for psychological IPA as participants who reported psychological IPA on the CT scales but not during the CIRCLE interview reported fewer such behaviors than other participants. Second, as Hamby (2016) suggests, the CT scales may be especially prone to identifying false positives. Indeed, nearly 20% (5 out of 26) of women have been misclassified as violent by the CTS2 due to their reports of behaviors such as playful wrestling or mock fighting (Lehrner & Allen, 2014) and 100% of physical IPA victims have reported experiencing CTS2 physical IPA behaviors in a joking or playful context (though some such reports may be due to participants’ reinterpretation of truly violent acts; Arriaga, 2002). Such behaviors would not be recorded by CIRCLE interviewers. Third, participants may have reported behaviors on the CT scales that occurred outside of the time frame assessed by the CIRCLE interview. They also may have reported behaviors that occurred outside of the time frame assessed by the CT scales as the magnitude of reporting errors increases with time since events, often allowing salient events to be erroneously reported during a requested time frame (Rubin & Baddeley, 1989).

In further tests of the incremental validity of the CIRCLE interview scores that were also designed to test their predictive validity, we demonstrated that CIRCLE measurement of PCA predicted children’s conduct problems beyond the CT scales’ measurement of PCA and IPA. Additionally, within-incident aggression spillover exposure predicted children’s emotional problems above and beyond measures of PCA and IPA, and it interacted with CIRCLE measured IPA to predict greater conduct problems. Results for the CIRCLE interview were not consistent across child outcomes, potentially due to the low internal consistency of the emotional problems scale, which may reflect parents’ greater difficulty reporting on the internal moods and states of children compared to their behaviors. However, overall, these results demonstrate stronger predictive power of the CIRCLE interview’s aggression estimates than that of the CT scales and support the incremental value of what the CIRCLE interview can offer beyond estimates of aggression frequencies. Now, with the existence of a measure of within-incident spillover, we can begin to better understand what causes such spillover, the ways in which it is uniquely detrimental for children, and why this might be the case. For example, in our ongoing work using Part II of the CIRCLE interview, we are addressing questions of how children’s immediate emotional reactions to IPA may differ based on their prior exposure to within-incident spillover.

In this first report, we additionally demonstrated how the CIRCLE interview provides valuable new knowledge about the nature of family aggression that other measures cannot. First, by measuring distinct incidents of aggression, we found that, in this sample, PCA occurred very frequently with few aggressive behaviors per incident. In contrast, although incidents of psychological IPA generally occurred less frequently than PCA, they often included more acts of aggression. Second, we found that 38% of participants experienced multiple incidents of aggression on a single day. Thus, measurement of aggressive incidents more accurately represents how family aggression occurs than measurement of days of aggression (as has been done with other uses of the event history calendar methodology; e.g, Testa et al., 2003). Finally, supporting the utility of incident-level measurement, as well as the importance of measuring PCA and IPA together, we found that a meaningful portion of participants reported the within-incident co-occurrence of PCA and IPA. This is the first time the existence of within-incident spillover of family aggression has been documented, thus supporting family systems conceptualizations of family conflict (e.g., Erel & Burman, 1995; Minuchin, 1985).

Our ability to validate the CIRCLE interview as a measure of physical IPA was limited by the dearth of participants (7 women and 4 men) who reported physical IPA. The low rates of physical IPA in this sample compared to other samples of parents of young children may be due to a number of selection factors. Sixty-one percent of participants completed a psychosocial prevention program focused on fostering positive coparenting, all participants were followed as research participants for nearly three years, and, on average, participants exhibited demographic factors that can be protective against physical IPA (e.g., relatively high income and education levels, mostly married; Schumacher et al., 2001a). Thus, this may have been a particularly low-risk sample compared to other parents of young children. Additionally, women reported more psychological PCA perpetration prior to the CIRCLE study than women who declined participation. Therefore, within the context of the Family Foundations study that focused on parenting and child development, women may have been more motivated to discuss problems in their relationships with their children than in their intimate relationships.

Additional study limitations warrant consideration. First, comparisons between the CIRCLE interview and CT scales were limited in two primary ways: 1) many of the CT scales exhibited low internal consistency, and 2) although adjusted for in statistical analyses, time lags between CIRCLE interviews were sometimes quite long, thus making comparisons between the four waves of CIRCLE interviews and past-year aggression estimates from the CT scales imperfect. Second, due to participant attrition, nearly 17% of interviews were missing. Although such attrition does not appear to have occurred systematically, at a minimum, this means that reported percentages of participants exhibiting particular patterns of aggression (e.g., multiple incidents of aggression on the same day, within-incident co-occurrence of PCA and IPA) are likely underestimates. Third, given difficulties accurately reporting on internal processes (Nisbett & Wilson, 1977), participants’ reports of their enhanced memory for aggression or avoidance of aggression as a function of completing the CIRCLE interviews may have been subject to substantial error. Finally, as described above, the current sample is unique in many ways. It will be important to test the validity and feasibility of the CIRCLE interview with demographically representative and high-risk samples, including those that exhibit higher rates of physical IPA.

Family aggression is notoriously difficult to assess. Although further tests of the CIRCLE interview as a measure of physical IPA are needed, it appears to be a reasonably valid measure of family aggression with incremental validity over the most widely used measures of PCA and IPA. Additionally, these initial findings indicate that measurement of within-incident co-occurrence of PCA and IPA may be important for better understanding the impact of aggression exposure on children. It is hoped that by providing a means to simultaneously measure incidents of PCA and IPA, the CIRCLE interview will facilitate movement of the family aggression field toward more comprehensive yet specific and process-oriented knowledge of family aggression.

Public Significance Statement.

To help integrate the fields of intimate partner aggression and parent to child aggression research, we developed an interview to assess the occurrence and severity of incidents of each form of aggression. In this study we demonstrate that the interview is reliable and valid in a variety of ways and, for the first time, we document that aggression toward partners and children can occur within the same aggressive incident, with unique negative implications for children’s functioning.