Understanding Control in Adolescent and Young Adult Relationships

Objectives  To describe the receipt of controlling behaviors in young women's dating relationships and the association with physical and sexual relationship violence (RV) and to ascertain whether young women experiencing controlling from partners support RV screening and respond honestly.

Design  Anonymous audio computer-assisted self-interview.

Setting  Reproductive health center.

Participants  A total of 603 women aged 15 to 24 years seeking reproductive care.

Main Outcome Measures  Self-reported victimization (controlling behaviors and physical and sexual aggression) by a partner in the past year.

Results  Sixty-eight percent of participants reported receiving 1 or more episodes of controlling behavior by a partner: 38.1% reported experiencing only controlling behaviors; 11.4% and 10.0% reported receiving controlling behaviors plus physical or sexual victimization, respectively; and 8.6% reported all forms of RV. Adjusted Poisson regression found that age 15 to 18 years (relative risk, 1.40), Hispanic ethnicity (1.29), childhood exposure to domestic violence (1.11), ever pregnant (1.21), older partner (1.28), recent physical (1.89) or sexual (1.93) victimization, and uncomfortable asking for condom use (1.39) were significantly associated with increased episodes of controlling behavior by a partner. Younger women and those who reported being victimized by controlling behaviors were more than twice as likely to object to screening by a health care provider, and those who reported receiving these behaviors were 2.5 times more likely to report that they might not honestly disclose RV.

Conclusions  Controlling behaviors are strongly associated with physical and sexual RV. Young women experiencing controlling behaviors are more reticent about screening for RV and may not feel that they can answer honestly.

High rates of relationship violence (RV) have been reported among adolescents and young adults. For example, in a nationally representative survey, 32% of 12- to 21-year-olds in heterosexual relationships reported 1 or more episodes of RV, defined as physical or sexual coercion.¹ However, estimates of RV vary widely (8%-59%) depending on the types of violence included in the definition and the setting used to assess violence.^2-7 Rates of physical aggression among adolescent dating partners have been reported to be 25% to 43%,^2,5 whereas rates of sexual victimization have been reported to be 8% to 15%,^2,4 with a high degree of overlap.

Studies have examined the correlates of RV in adolescents and young adults and the relationship between RV and various health outcomes. For example, correlates of being a victim of RV include female sex,⁷ reported minority ethnicity,⁵ past pregnancy,⁵ lower educational level,⁵ concurrent involvement in other violent behaviors, illicit substance use, and sexual risk taking.⁶ Predictors of the onset of RV victimization in young women include being depressed and having a friend who was a victim of RV.^8,9 Health risks that have been associated with RV in young women include poor health-related quality of life,¹⁰ higher risk of substance use,^10,11 unhealthy weight control behaviors,¹¹ pregnancy,¹¹ and higher rates of suicidal ideation and attempts.^10-12 Whether these health risks precede or are consequences of RV is unknown. There is some evidence, however, that early sexual initiation predicts RV¹³ and that in an RV adolescent relationship, sexual intercourse precedes RV.¹⁴

Controlling behavior, defined as a dating partner exerting nonphysical coercion, demands, or constraints on a partner's behaviors, has been studied in adult women and is linked to other forms of intimate partner violence¹⁵ but has been infrequently studied in adolescent and young adult populations.¹⁶ Among adult women, a wide range of controlling behaviors, including monitoring, stalking, social isolation, economic control, and use of children to manipulate a partner,¹⁷ have been related to the most severe cases of partner violence,^18,19 including attempted homicide.²⁰ Moreover, Pulerwitz et al²¹ examined the measurement of relationship power and its association with physical violence.

There is scant literature on controlling behaviors in adolescent and young adult relationships, and behaviors vary considerably in type and definitions. For example, in a college population, Murphy and Hoover²² described 4 types of psychological aggression, including items assessing partner control. In another similar sample, approximately half of the males and females reported being both a perpetrator and a victim of controlling behaviors, with significant overlap reported in physical and sexual victimization.²³ Among adolescents, these behaviors remained stable over a 3-month period and were associated with current and future episodes of physical aggression for dating males and females.²⁴ Finally, Teitelman et al²⁵ examined the relationship among RV, condom use, and sexual relationship power in minority urban girls. As expected, girls who experienced more RV were less likely to use condoms.²⁵

This study aimed to better describe controlling behaviors experienced by adolescent and young adult females in the absence of experiencing RV and when it overlaps with RV. We hypothesized that receiving increased episodes of controlling behavior by a romantic partner represents another element in the constellation of factors that compose RV. Thus, we sought to determine factors associated with increased episodes of controlling behavior in this population. Because previous studies reported that adolescents are unlikely to seek help for RV,^26,27 we investigated whether young women experiencing controlling behaviors from dating partners would object to screening by a health care provider and whether they felt that they could respond honestly to screening questions.

This study uses a smaller sample from a larger study that has been previously described.²⁸ The present sample included a diverse group of adolescent and young adult females aged 15 to 24 years who reported having a heterosexual relationship during the previous year (N = 603). Briefly, the institutional review board–approved parent study occurred at a busy reproductive health center in New York City in 2004. Young women were approached after clinic registration and were asked whether they would be interested in participating in a study about relationships. Trained research assistants spoke with interested participants to obtain oral consent. Each participant completed an anonymous audio computer-assisted self-interview by reading and listening to the questions on headphones and using a laptop with a privacy screen to respond. This approach has been shown to increase disclosure of sensitive information or risk behaviors in adolescent populations.²⁹ Each participant received a $10 gift certificate to a popular store.

The audio computer-assisted self-interview covered a range of areas, including demographic and reproductive characteristics, such as age, race/ethnicity, education, relationship status, and pregnancy history. Participants were asked to choose their current or most recent sexual or romantic relationship within the previous 12 months and were asked about the partner's sex, relationship duration, importance of this relationship to them, current and past use of contraceptives, and their comfort in discussing birth control with their partner. If they had more than 1 partner during the previous year, they were not given guidance as to which to choose.

We assessed RV by using the Conflict in Adolescent Dating Relationships Inventory (CADRI), which measures episodes of victimization and perpetration of physical and sexual violence by a sexual or romantic partner using an ordinal scale—never, seldom (1-2 times in relationship), sometimes (3-5 times), or often (≥6 times).³⁰ Instructions stated, “the following questions ask about things that may have happened while you were having an argument,” and each respondent was asked the number of times that these behaviors occurred in the past year (with the specified current or ex-partner). Physical violence was assessed using 4 items: being kicked, hit, or punched; being pushed, shoved, or shook; being slapped or having hair pulled; or having something thrown at them. Sexual violence was measured with 3 items: forced to have sex when not wanted, threatened in an attempt to have sex, and experiencing unwanted sexual touching. We excluded unwanted kissing, a fourth item used by Wolfe et al,³⁰ who developed the CADRI, because it is a milder form of sexual coercion.

To measure family violence, including child sexual abuse, we used the Dating Violence Inventory and Family Abuse Scale.³¹ The family violence subscale asks about childhood exposure to domestic violence (ie, have you seen your father hit your mother), physical violence from a parent toward self (ie, has your parent hit you as a child), and early sexual abuse (ie, sexually abused by a family member or other nonpartner adult before age 18 years).

Receipt of controlling behaviors by the partner was assessed using 7 items from the World Health Organization Multi-Country Study: tries to keep me from seeing my friends, tries to restrict contact with my family, insists on knowing where I am at all times, ignores me and treats me indifferently, gets angry if I speak with another man, is often suspicious that I am unfaithful, and expects me to ask his permission before seeking health care for myself.³² Each of these statements was assessed using a similar response set as described previously herein for the CADRI.

The final section asked young women about their attitude toward screening practices, both past and future. For example, they were asked whether they would mind if a health care professional asked about violence in their life and whether they could answer the screening questions honestly.

For this study, participant age was categorized into 1 of 3 groups: 15 to 18, 19 to 21, and 22 to 24 years. If the participant reported 1 or more episodes of either perpetrating or receiving physical or sexual violence on the CADRI, she was considered a perpetrator and a victim of physical or sexual violence in the previous year. Physical violence perpetration or victimization was defined as being present if a young woman responded “yes” to any 1 of 4 questions about her or her partner's use of physical aggression. We defined childhood exposure to domestic violence and childhood sexual abuse if she reported 1 or more episodes of these behaviors. The 7 items assessing controlling behaviors were summed to create a total score (range, 0-21) that represented an approximation of the frequency of these behaviors across the previous 12 months. Finally, because some relationships either began or ended during the 12-month window covered in the measures, we used 2 questions from the survey (ie, length of relationship and did your relationship end before the audio computer-assisted self-interview) to calculate a new variable to approximate months of exposure, that is, “months at risk.”

Specific variables associated with receiving controlling behaviors were then identified by comparing demographic, reproductive, and other RV behaviors of those who reported 1 or more episodes of being controlled by a romantic partner with those who did not report this behavior using the χ² or t tests. Variables significant at P < .10 were considered for entry into a forward stepwise Poisson regression designed to identify predictors of the increasing receipt of controlling behaviors. The colinearity between perpetration and victimization for physical and sexual violence was assessed. Given the high correlation (r = 0.71) between victimization and perpetration of each violence type, only sexual and physical victimization were entered into the regression models. Therefore, this article does not comment on any relationship between experience of controlling by a male partner and the perpetration of physical or sexual violence by the woman experiencing controlling behavior. Potential interactions, such as age by race and other demographic variables with RV, did not change the main outcomes and, thus, were excluded from the model. Relative risks, with associated 95% confidence intervals, were computed.

In the present population, 72.0% of women (n = 434) reported experiencing 1 or more episodes of physical or sexual victimization or controlling behavior. The Figure describes the reported victimization in the past year by violence type and the overlap among these behaviors. Of the women experiencing physical and sexual victimization in the past year, most also reported receiving controlling behaviors. Of the women reporting receiving controlling behaviors, approximately 1 in 10 reported receiving all forms of victimization—sexual and physical aggression and controlling behaviors by a partner.

The proportion of young women who reported episodes of controlling behaviors varied across items used to assess these behaviors. For example, only 3.7% of the sample (n = 22) reported that their partner expected them to ask his permission before seeking health care, and 6.3% (n = 38) reported that the partner tried to restrict their contact with family. In contrast, 24.7% of respondents (n = 149) reported that their partner ignored them and treated them indifferently, and 26.5% (n = 160) reported that their partner tried to keep them from seeing friends. The most common forms of control exerted by the male partner reported by this sample included that the partner was suspicious that she was unfaithful (40.5%, n = 244), got angry if she spoke to another man (40.8%, n = 246), and insisted on knowing her location at all times (45.9%, n = 277).

Table 1 describes the demographic characteristics associated with reporting 1 or more episodes of controlling behavior in the past year. The receipt of controlling behaviors by a partner was higher among younger adolescents, Latinas, those living alone, and those reporting 1 or more pregnancies. No other demographic or reproductive characteristics were significantly associated with receiving controlling behaviors from a partner.

Table 2 reports the association between experiencing controlling behaviors, early exposure to family violence, and child maltreatment and partner violence victimization and perpetration. All forms of violence are associated with statistically significantly more frequent episodes of controlling behavior. Table 3 provides the association between dating relationship characteristics and reports of controlling behaviors. A relationship of less than 3 months was associated with fewer reported episodes of controlling behavior. Finally, we examined the mean difference between those who reported 1 or more episodes of controlling behavior and those who reported no occurrences on the created variable months at risk to examine temporal exposure to this dating partner. As predicted, we found that the mean number of months at risk was higher for those who reported receiving controlling behaviors (8.1 vs 7.4 months, P < .04).

To examine factors associated with increasingly frequent episodes of controlling victimization, a multivariable model predicting the reported increased number of episodes of controlling behavior using the Poisson regression model was computed (Table 4), while adjusting for temporal exposure to the dating relationship length within the 12 months asked about in the questionnaire, that is, months at risk. We found that younger age, being Hispanic, having a partner within 3 to 5 years of the young woman's age (compared with a partner >5 years older), being in a longer relationship, discomfort asking a partner to use a condom, having witnessed parental domestic violence, and recent victimization (sexual and physical) were associated with increasing episodes of controlling behaviors by a partner.

Finally, we examined whether women experiencing controlling behaviors object to being screened by a health care provider. Adjusting for known and expected confounders, including race/ethnicity, non–US-born, living arrangements, physical and sexual victimization, childhood sexual abuse, and exposure to parental domestic violence, of participants (n = 75) who reported often (≥6 episodes) being victimized by controlling behaviors from their partners compared with no controlling victimization, 58% felt that they would mind “a bit” or “a lot” being screened (adjusted odds ratio, 2.8; 95% confidence interval, 1.32-6.12).

To identify factors associated with the patient's report of being able to answer questions about RV honestly, we used logistic regression to compare those who reported that they could answer honestly to screening questions about RV with those responding that they might not be able to answer honestly. Adjusting for age, race/ethnicity, gravidity, length of relationship, physical and sexual victimization, childhood sexual abuse, non–US-born, and age discrepancy between male partner and female patient, those who reported 6 or more episodes of controlling behavior from their partner compared with those who did not experience any were 2.66 times more likely (95% confidence interval, 1.33-3.21) to report that they were unsure whether they could answer honestly.

Just more than two-thirds of the young women in this sample reported receiving 1 or more episodes of controlling behaviors. Consistent with the hypothesis, controlling behaviors are clearly present and related to other forms of RV in adolescent and young adult relationships. The most prominent controlling behaviors from partners included knowing a partner's location at all times, having a problem when a partner talks to another male, and suspected infidelity. Because the study is cross-sectional and reports on the same 12-month period for violent and controlling behaviors, we cannot determine whether controlling behaviors are a precursor or a consequence of physical or sexual violence. In 38.1% of young women, however, controlling behaviors occurred without physical or sexual violence from their partners, which is consistent with studies of adolescent, young adult, and adult females.^18,19,23,24

The profile of young women who are at increased risk for experiencing controlling behaviors includes being younger, Latina, exposed to childhood domestic violence, a current victim of physical and sexual RV, and in a relationship with a male 3 to 5 years older than oneself. Adverse health outcomes among female youth, including pregnancy, have been linked to older adult men (≥5 years),³³ but increased sexual risk has also been reported with male partners just 2 to 3 years older.^34,35

Previous research¹⁷ on controlling behaviors has focused on married or adult partnerships where financial control was exerted, but the present population offers the opportunity to investigate controlling behavior impact in a younger group. Although we know that these young women also sustain physical and sexual violence, the temporal and possibly causal relationship between these types of RV and controlling behaviors (ie, concurrent vs preceding) cannot be determined given the cross-sectional nature of the study. For example, does physical or sexual violence arise because of refusal to be controlled? Alternatively, are the controlling behaviors early signs of coercion in relationships that then become physically or sexually violent? The question remains that if these controlling behaviors are, in fact, precursors to RV, what can stop the controlling behavior from progressing to RV? These data suggest that young women in longer relationships report more episodes of controlling behaviors, suggesting that these types of coercive behaviors increase over time. A longitudinal study is necessary to tease out the temporal sequence of behaviors.

Perhaps most disturbing is that young women who receive controlling behaviors are a group that is more likely to express reticence toward screening by a health care provider. They, particularly the younger women, were also more likely to be unsure whether they could answer honestly if they were asked directly by a health care provider. Most professional organizations advocate universal screening for all women, including adolescents.³⁶ However, the present data suggest that physicians serving teens are likely to find themselves in a conundrum. If those who are younger or who have experienced controlling behaviors from a partner are less likely to be able to answer questions about RV honestly, how should a provider evaluate a negative response to screening queries? In particular, how should a physician communicate with the patient who says “no” when there may be a clear suspicion that she is experiencing a violent relationship?

Screening is needed to identify those who are ready and willing to disclose these relationships, and disclosure will likely facilitate adherence with needed services. Some have argued that there may be therapeutic value in screening³⁷ because the health care provider, independent of the response, demonstrates regard for the patient's life and prioritizes violence as an issue that adversely affects health.³⁸ One would not suggest “pushing” a young woman to disclose information but giving her the understanding that the health care provider is ready to hear when she is ready to tell. It will also be important in helping them recognize that the health care setting is a safe and appropriate place to make such a disclosure. In addition, a young woman must understand that no one will judge her for disclosing violence even if she has previously denied it. These findings underscore the need to frame screening so that young women are not further distressed by answering evasively or dishonestly. This can be achieved by health care providers acknowledging how difficult it can be to talk about violence and coercion in a romantic relationship and that it can take young women time to feel comfortable enough to reveal RV to anyone, including health care providers. In addition, the inclusion of RV screening questions might be incorporated as part of a larger self-reported medical history. The use of a comprehensive screening approach in tandem with an audio computer-assisted self-interview method may facilitate greater honesty in disclosing this sensitive information.^39,40

Several limitations deserve comment. First, the temporal relationship between controlling behaviors and physical or sexual violence cannot be determined because these are cross-sectional data. Second, given that this is an urban sample presenting for reproductive health care, these results may not be generalizable to other populations. Third, the nature of the data, the sample size, and the analyses undertaken mean that overall, the study has investigated a mild to moderate level of violence in these relationships. This is not at the level that has been termed “patriarchal terrorism” when examining adult or married relationships.^15,17 Fourth, given that controlling behaviors occur so frequently, they may, in fact, be normative in adolescent relationships. Adolescents may perceive this control to reflect a more committed relationship, a partner's love, or being extremely important to their partner. O’Leary and Smith Slep²⁴ suggest that some nonviolent tactics are normative and have little to do with power inequities or gender in adolescents. Fifth, participants were asked only to report on RV in 1 relationship in the past year so that levels of RV may be underreported if young women had more than 1 relationship in the past year that involved RV. Sixth, there have been substantive changes in technology that make monitoring partners' behaviors easy. The use of mobile telephones, text messaging, e-mail, and social networking sites makes monitoring in relationships less distinguishable from overt attempts to exert control over one's partner. Recently, a survey among 13- to 18-year-old adolescents and their parents reported high rates of control and threats using technology, that is, 30% reported that they received 10 to 30 texts per hour from a partner to find out where they were or who they were with; 25% reported being called names or being put down by a partner through cell phone call or text, and 10% were physically threatened by a partner via e-mail, instant messaging, or text.⁴¹

These data demonstrate the high frequency of controlling behaviors in the relationships of adolescents and young adults and support a nuanced approach to universal screening of controlling behavior. The high numbers of young women in this sample who reported receiving these controlling behaviors suggest that these may be normative. However, the powerful association of early adverse experiences with family violence and with contemporaneous RV is a cause for concern. Perhaps more important, it may alert a provider to the risk or presence of physical or sexual violence in the relationship and prompt additional time in evaluating the risk of violence. In addition, this awareness of the high rates of controlling behavior and the overlap with RV, particularly for young people, may affect how they view health care provider–based screening and how honestly they might answer screening questions. An awareness that young women may not yet be comfortable disclosing information honestly should prompt carefully crafted, repeated, and novel screening to improve identification, referral, and treatment.

Correspondence: Marina Catallozzi, MD, Division of General Pediatrics, Department of Pediatrics, Columbia University, 622 W 168th St, VC-4 Room 449-F, New York, NY 10032 (mc2840@columbia.edu).

Accepted for Publication: September 26, 2011.

Author Contributions: All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Davidson, Breitbart, and Rickert. Acquisition of data: Davidson, Breitbart, and Rickert. Analysis and interpretation of data: Catallozzi, Simon, Davidson, Breitbart, and Rickert. Drafting of the manuscript: Catallozzi and Simon. Critical revision of the manuscript for important intellectual content: Catallozzi, Davidson, Breitbart, and Rickert. Statistical analysis: Catallozzi, Simon, and Rickert. Obtained funding: Davidson. Administrative, technical, and material support: Breitbart. Study supervision: Davidson.

Financial Disclosure: Dr Rickert has received a research grant from and is a National Adolescent and Young Adult Male Advisory board member for Merck and Company, Inc and is an Adolescent Immunization Leadership Council member sponsored in part by Sanofi Pasteur, Inc.

Funding/Support: This study was supported by Cooperative Agreement 5 U49 CE00731 from the Centers for Disease Control and Prevention.

Disclaimer: The contents of this article are the sole responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention.

Previous Presentation: This article was presented as an abstract as 1 of 5 nominees for a New Investigator Award at the Society for Adolescent Medicine; March 2009; Los Angeles, California.