Adolescent Dating Violence Stability and Mutuality: A 4-Year Longitudinal Study

Abstract

This 4-year longitudinal study explored the stability of dating violence (DV) during adolescence and the reciprocal associations between perpetration and victimization over time. Participants were 991 high school students (52.4% females; mean age at baseline = 14.80 years) from Bizkaia (Spain), who completed a measure of DV perpetration and victimization at four measurement points spaced 1 year apart. Findings evidenced stability of teen perpetration and victimization of DV, which appears to increase in late adolescence. Moreover, longitudinal reciprocal influences were demonstrated, but in general, the cross-lagged paths from one’s partner’s aggression to one’s own perpetration and vice versa were lower than the autoregressive paths obtained from stability. The model showed an adequate fit for both females and males, although some paths were significantly higher for the females than for the males. Preventive interventions should consider these findings about stability and longitudinal reciprocal associations of DV during adolescence.

Keywords

dating violence adolescence stability mutuality dyadic influences

Dating violence (DV) includes any act of physical, emotional, or sexual violence that can take place in person or electronically (Centers for Disease Control and Prevention [CDC], 2016). Numerous studies in various countries have evidenced the presence of aggressive behaviors in adolescent dating relationships (CDC, 2010; Connolly et al., 2010; Hird, 2000; O’Leary, Slep, Avery-Leaf, & Cascardi, 2008), with mutual violence being the most prevalent pattern of aggression (Chiodo et al., 2012; O’Leary et al., 2008). These findings are similar to those of several studies conducted in Spain with samples of high school students (Fernandez-Fuertes & Fuertes, 2010; Fernández-González, O’Leary, & Muñoz-Rivas, 2014; Muñoz-Rivas, Graña, O’Leary, & González, 2007). Although psychological aggression and mild physical aggression are the most prevalent types of adolescent DV, more severe types of violence (such as sexual aggression and severe physical aggression) are also present (Foshee, 1996; Foshee, Benefield, Ennett, Bauman, & Suchindran, 2004), and research has revealed the negative impact of DV on adolescents’ physical and mental health (Fernández-González et al., 2014; Foshee, Reyes, Gottfredson, Chang, & Ennett, 2013) and its relationship with other health-compromising risk behaviors (Reingle, Staras, Jennings, Branchini, & Maldonado-Molina, 2012; Wolfe et al., 2009). Moreover, available research suggests that some stability already exists in partner aggression during adolescence (e.g., Fritz & Slep, 2009), although the scarcity of longitudinal studies makes it very difficult to draw well-founded conclusions.

Knowledge about the stability and longitudinal mutual influence of DV in adolescent dating relationships is important to better understand the nature of intimate partner aggression during adolescence and to improve preventive interventions. That is, to what extent does partner violence persist during adolescence? Which type of cross-lagged associations among perpetration and victimization take place over time? To answer these questions, the main objective of the present study was to examine DV stability and reciprocal associations between perpetration and victimization over the course of 4 years in a sample of Spanish high school students.

Stability of DV Through Adolescence

Romantic relationships during adolescence appear to be characterized by a lower stability of intimate partner aggression than during adulthood (O’Leary & Slep, 2012). This may be related to several factors. First, research on the trajectory of DV throughout adolescence has found that physical partner aggression peaks around the ages of 16 and 17 years (Foshee et al., 2009; Nocentini, Menesini, & Pastorelli, 2010). This peak age of DV during middle-to-late adolescence was also found in Spanish adolescents (Fernández-González et al., 2014). Second, teen dating relationships are characterized by a lesser level of commitment and a shorter length than romantic relationships in later stages (Wekerle & Wolfe, 1999). However, although at a lower level than in adult life, partner violence may be a moderately steady behavior during adolescence according to the findings of the few longitudinal studies that have examined partner aggression stability in teens (Choi & Temple, 2016; Fritz & Slep, 2009; O’Leary & Slep, 2003). A recent study conducted by Choi and Temple (2016) in Texas (USA) found that the majority of adolescents (between the ages of around 16 to 18 years) tend to stay in the same status of DV victimization (nonvictims, emotional/verbal victims, and physical/psychological victims) over 3 years, although some of them did transition between statuses.

Two additional studies explored both the perpetration and victimization of DV. O’Leary and Slep (2003) demonstrated the substantial stability of physical DV perpetration and victimization over 14 weeks in a sample of high school students (ages around 16-17 years) from Long Island (New York). Correlations between physical aggression perpetration at Time 1 and Time 2 were .57 for females and .55 for males, while they were .62 and .69 (females and males, respectively) for physical aggression victimization. However, this study was carried out with adolescents who stayed in the same relationship and only two assessment points in a short time range of three and a half months. In a further study with a longer time span (three assessment points at weeks 1, 14, and 52), Fritz and Slep (2009) explored the stability of physical and psychological DV across both time and partners (whether or not they remained with the same partner across the three waves). Results showed moderate levels of stability of physical and psychological DV perpetration and victimization, and increases in psychological aggression were related to consistency in dating partners across time. Mean correlations across the three waves were .28 for perpetration and .23 for victimization, which were lower than those found in O’Leary and Slep’s (2003) previous study conducted during 14 weeks. The findings of these latter two studies with high school students from the United States suggest that partner aggression is already a relatively steady behavior in middle adolescence during at least 1 year, although results may vary if we consider a longer period of time, as we did in our 4-year longitudinal study.

Reciprocal Influences on Adolescent DV Stability

Although some of the predictors of adolescent DV vary across gender (Foshee, Linder, MacDougall, & Bangdiwala, 2001; Reingle, Jennings, Maume, & Komro, 2013), partner aggression during adolescence is strongly characterized by a reciprocal pattern of aggression with both members of the couple acting simultaneously as perpetrators and victims (Archer, 2000; Chiodo et al., 2012; Fernández-González et al., 2014; O’Leary et al., 2008). This was found even when the context and reasons for aggression were considered (Foshee, Bauman, Linder, Rice, & Wilcher, 2007). Therefore, it seems reasonable to assume that perpetration and victimization can influence each other over time. This perspective is supported by the results of previous longitudinal studies that have highlighted the influence between perpetration and victimization trajectories of adolescent DV (Nocentini et al., 2010; Orpinas, Nahapetyan, Song, McNicholas, & Reeves, 2012). Orpinas et al.’s (2012) 7-year study conducted in Georgia (USA) showed that 91% of the adolescents who experienced and perpetrated psychological DV followed concordant trajectories (e.g., increasing victimization, increasing perpetration). Moreover, Nocentini et al.’s (2010) 3-year longitudinal study conducted in Italy showed that being a victim of the partner’s aggression retards the normative decrease of physical DV over adolescence. These findings emphasize the mutual relationship between being an aggressor and being a victim of DV, although studies exploring bidirectional longitudinal associations between victimization and perpetration of DV in adolescents are scarce. O’Leary and Slep’s (2003) study tested a dyadic longitudinal model of DV, and they found that when the cross-lagged paths between victimization and perpetration were included in the model, the strength of the autoregressive paths for both perpetration and victimization dropped to nonsignificance. Thus, they concluded that dyadic influences were more predictive of one’s own physical aggression than stability. Nevertheless, as noted before, O’Leary and Slep’s study was conducted with a sample of adolescents who remained with the same partner and it only included two waves in a short range of three and a half months.

The Present Study

Although our knowledge about the developmental pattern of DV through adolescence has grown considerably in the last decade (Foshee et al., 2009; Nocentini et al., 2010), we still know very little about adolescent DV stability and longitudinal reciprocal associations between victimization and perpetration. Due to the high overlapping between victimization and perpetration, longitudinal research is needed to examine whether victimization predicts future perpetration or vice versa, although both reciprocal influences could coexist. Furthermore, our study expands upon the adolescent DV stability literature in examining whether the results found in shorter time periods also apply to what occurs in a longer time period over 4 years in adolescence. Data were collected at four measurement periods (Time 1 [T1], Time 2 [T2], Time 3 [T3], and Time 4 [T4]), which were spaced 1 year apart. Since adolescent dating relationships are short in length (Wekerle & Wolfe, 1999) and the study lasted 4 years, our objective was to examine aggression stability regardless of whether or not participants were in the same relationship. Moreover, given that the different types of aggression tend to co-occur (Cornelius & Resseguie, 2007; Sears, Byers, & Price, 2007), we analyzed a global measure of DV that included the five types of aggression (physical abuse, threatening behavior, verbal/emotional abuse, relational abuse, and sexual abuse) assessed by the Conflict in Adolescent Dating Relationships Inventory (CADRI; Wolfe et al., 2001). Nevertheless, the model was also estimated separately for psychological and physical DV.

Based on the above, this study aimed to accomplish the following: (a) analyze the stability of perpetration and victimization of DV during 4 years in a sample of Spanish high school students (ages from around 15 to 18 years), (b) examine the longitudinal reciprocal associations between victimization and perpetration over time, and (c) explore sex differences in both stability and longitudinal reciprocal associations between victimization and perpetration. With regard to the first aim, we hypothesized that DV perpetration at each time point will be predicted by DV perpetration in the previous time point, while DV victimization at each time point will be predicted by DV victimization in the previous time point. However, coefficients are expected to be lower than in previous studies (Fritz & Slep, 2009; O’Leary & Slep, 2003) because of the longer period of time assessed. Regarding the second aim, based on previous research about the dyadic nature of DV, we assumed a strong relationship between DV perpetration and victimization, and our hypothesis was that an individual’s own aggression will be influenced by their previous partner’s aggression and vice versa (i.e., DV perpetration at each time point will be influenced by DV victimization in the previous time point, and DV victimization at each time point will be influenced by DV perpetration in the previous time point). Last, regarding the effect of sex, we had no specific hypothesis as previous studies have found similar results for females and males as well as some specific inconclusive differences (Choi & Temple, 2016; Fritz & Slep, 2009; O’Leary & Slep, 2003).

Method

Participants

The whole sample consisted of 2,262 adolescents (1,642 in T1; 1,579 in T2; 1,271 in T3; and 1,058 in T4). Of these, 1,796 participants reported having had a dating relationship in the past year in at least one of the four waves of the study. However, only adolescents who had had a dating relationship in at least two of the four waves of the study were selected (N = 991; 55.18%). This criterion was proposed to have at least two data points for each participant but, at the same time, to avoid the further loss of participants due to a stricter criterion of three or four waves. Therefore, after excluding the participants who did not meet the mentioned criterion, the final sample comprised a total of 991 adolescents. The mean age of the participants at baseline was 14.80 years (SD = 1.11), and there were 519 females (52.4%) and 472 males (47.6%). Regarding ethnicity, the majority of the participants were Spanish (91.4%). With respect to the remaining participants, 7.2% were from South America and 1.4% was from Eastern Europe, Africa, and other countries. The socioeconomic class of the participants was as follows: 13.5% low, 17.9% medium-low, 31.9% medium, 29.7% medium-high, and 7% high (Spanish Society of Epidemiology and Family and Community Medicine, 2000). This distribution is similar to the overall distribution in the community (National Statistics Institute of Spain, 2013).

Procedure

All the participants came from 9 public and 13 private high schools in Bizkaia, a province in northern Spain with a total population in 2015 of 1,141,442 (Basque Institute of Statistics [Eustat], 2015). Most of the schools were from Bilbao, which is a metropolitan area and the capital of Bizkaia. The sample was first stratified by school type (i.e., private vs. public). The schools were then selected randomly by means of a cluster sampling procedure. First, we contacted the schools to explain the objectives of our study. About 40% of the schools agreed to take part in the study. Thereafter, we sent information letters and consent forms to parents. Although parents had the option of refusing to allow their child’s participation in the study, all parents agreed to let their children participate (i.e., parental consent rate = 100%). All adolescents agreed to participate in this study (i.e., assent rate = 100%), and they completed the questionnaires in their classrooms at the four measurement periods. Participants were asked to provide some data (first initial of mother’s name, first initial of father’s name, and her/his birthday date) to match their questionnaires over time. The Ethics Committee of the University of Deusto (Spain) approved this study and confidentiality was guaranteed. All missing values were imputed using the expectation–maximization (EM) imputation algorithm with IBM SPSS 23.0.

Measure

The CADRI (Wolfe et al., 2001) is a self-report questionnaire made up of 25 bidirectional questions (perpetrator/victim), which assesses five types of abusive behaviors among adolescent dating partners: physical abuse (e.g., “I slapped my partner or pulled my partner’s hair”/”My partner slapped me or pulled my hair”), threatening behavior (e.g., “I threatened to hurt my partner”/”My partner threatened to hurt me”), sexual abuse (e.g., “I forced my partner to have sex when he/she didn’t want to”/”My partner forced me to have sex when I didn’t want to”), relational abuse (e.g., “I tried to turn my partner’s friends against them”/”My partner tried to turn my friends against me”), and verbal or emotional abuse (e.g., “I insulted my partner with put-downs”/”My partner insulted me with put-downs”). Response choices for each item were defined as never (this has never happened), seldom (this has happened only 1 to 2 times), sometimes (this has happened about 3-5 times), and often (this has happened 6 times or more). Participants were asked to answer the questionnaire only if they had had a dating partner during the last year. They must indicate whether the stated aggressive behaviors had happened in the last year during the course of a disagreement with their current dating partner, or with their previous partner in case they were not in a dating relationship at the time of the study.

The Spanish version of the CADRI has shown adequate psychometric properties and confirmation of its structure in a sample of Spanish high school students (Fernández-Fuertes, Fuertes, & Pulido, 2006). At T1, we used a shorter version of the CADRI, which consisted of six items of perpetration and six of victimization from the subscales of physical abuse, threatening behavior, sexual abuse, and verbal/emotional abuse. The reason for doing so was that several items were considered not appropriate for adolescents at that age. Cronbach’s alphas in this first wave were .67 for the perpetration subscale and .69 for the victimization subscale. The full version of the CADRI was used in the subsequent waves. Cronbach’s alphas were .86 for the perpetration subscale and .87 for the victimization subscale in T2, .86 for the perpetration subscale and .83 for the victimization subscale in T3, and .83 for the perpetration subscale and .87 for the victimization subscale in T4. Mean global scores were calculated for perpetration of DV and victimization of DV.

Data Analyses

We used path analysis with LISREL 8.8 (Jöreskog & Sörbom, 2006) to examine stability and longitudinal associations between DV perpetration and victimization. Because our data did not follow a multivariate normal distribution, we used the robust maximum likelihood (RML) method, which requires an estimate of the asymptotic covariance matrix of the sample variances and covariances and includes the Satorra-Bentler scaled χ² index (S-B χ²; Chou, Bentler, & Satorra, 1991). Goodness of model fit was examined using the comparative fit index (CFI), the nonnormed fit index (NNFI), the standardized root mean square residual (SRMR), and the root mean square of error of approximation (RMSEA). Generally, CFI and NNFI values of .90 or higher and SRMR and RMSEA values lower than .08 indicate adequate fit for longitudinal data (Little, 2013).

The model included autoregressive paths between T1, T2, T3, and T4 DV perpetration and between T1, T2, T3, and T4 DV victimization. The autoregressive paths assess whether one’s own aggression toward the partner at one time point is related to one’s own aggression toward the partner at the next time point, and whether aggression by the partner at one time point is related to aggression by the partner at the next time point. Therefore, these paths represent the stability of these variables over time. The model also included cross-lagged paths between perpetration and victimization over time. These cross-lagged paths assess whether aggression by the partner at one time point is related to one’s own aggression toward the partner at the next time point, and whether one’s own aggression toward the partner at one time point is related to aggression by the partner at the next time point. Although data about perpetration and victimization come from one member of the partner, consistent with previous research in the field, we interpret the cross-lagged paths as longitudinal dyadic influences. In addition to the autoregressive and cross-lagged paths, direct paths from T1 participants’ age to T2, T3, and T4 DV perpetration and to T2, T3, and T4 DV victimization were estimated to control for the effect of age. The exogenous variables at T1 were all free to covary, and residuals of the endogenous perpetration and victimization variables at each time were also allowed to covary.

We also explored whether the longitudinal model of DV perpetration and victimization was invariant across female and male subsamples through a multiple-group analysis. For this purpose, the following steps were conducted. First, the final model was tested both for females and males separately. Second, we tested the configural invariance of the model to demonstrate that the pattern of fixed and free parameters was equivalent across subsamples. Third, we tested the invariance of the variances and covariances of the exogenous factors. Finally, we tested the invariance of the autoregressive and cross-lagged paths linking DV perpetration and victimization over time.

Results

Descriptive Statistics

Means and standard deviations of DV perpetration and victimization at each time point are displayed in Table 1. In addition, correlation coefficients between mean scores of DV perpetration and victimization within and across time are depicted in Table 2. All the correlations were statistically significant. The highest correlations were those within the same time point between perpetration and victimization, which ranged between .80 and .89. Correlations across time between perpetration and victimization, which are related to longitudinal dyadic influences, ranged from .31 to .65. Regarding the participants’ own perpetration and victimization stability, correlations across time for perpetration ranged between .32 and .75, while correlations across time for victimization ranged between .36 and .62. For all the correlations across time, the largest ones were found between T3 and T4, which were higher than correlations between T2 and T3, and these were in turn higher than correlations between T1 and T2. Overall, correlations were higher for females than for males.

Table 1.

Means and Standard Deviations of Dating Violence Perpetration and Victimization (N = 991).

	Total Sample	By Sex		t(989)
	Total Sample	Females	Males	t(989)
T1 perpetration	0.18 (0.27)	0.18 (0.29)	0.17 (0.26)	0.99
T1 victimization	0.19 (0.29)	0.19 (0.30)	0.19 (0.29)	−0.10
T2 perpetration	0.18 (0.22)	0.20 (0.25)	0.15 (0.18)	3.89***
T2 victimization	0.18 (0.23)	0.20 (0.25)	0.15 (0.20)	3.12**
T3 perpetration	0.16 (0.18)	0.17 (0.17)	0.14 (0.19)	2.70**
T3 victimization	0.15 (0.17)	0.16 (0.17)	0.14 (0.18)	2.40*
T4 perpetration	0.17 (0.17)	0.17 (0.17)	0.17 (0.17)	0.63
T4 victimization	0.18 (0.18)	0.18 (0.17)	0.18 (0.19)	−0.38

Note. T1 = Time 1; T2 = Time 2; T3 = Time 3; T4 = Time 4. Female subsample: n = 519; male subsample: n = 472.

p < .05. **p < .01. ***p < .001.

Table 2.

Pearson Correlations Among the Study Variables (N = 991).

Variable	1	2	3	4	5	6	7	8
1. T1 perpetration		.80	.47	.49	.39	.46	.32	.37
2. T1 victimization	.83/.77		.45	.49	.33	.44	.31	.36
3. T2 perpetration	.49/.45	.50/.39		.89	.53	.52	.59	.46
4. T2 victimization	.52/.46	.53/.46	.93/.81		.55	.56	.57	.50
5. T3 perpetration	.43/.34	.40/.26	.64/.40	.64/.44		.88	.75	.63
6. T3 victimization	.47/.44	.49/.40	.60/.42	.61/.49	.87/.90		.65	.62
7. T4 perpetration	.37/.27	.37/.23	.69/.48	.67/.45	.83/.68	.72/.58		.85
8. T4 victimization	.45/.29	.44/.27	.56/.35	.61/.40	.69/.57	.68/.56	.80/.90

Note. The correlations for the whole sample (n = 991) are presented above the diagonal; below the diagonal are correlations across sex: females (n = 519)/males (n = 472). T1 = Time 1; T2 = Time 2; T3 = Time 3; T4 = Time 4. All the correlations were statistically significant (p < .001).

Longitudinal Model for DV Perpetration and Victimization

The path analysis results showed that DV perpetration and DV victimization were significantly associated at T1. Neither of them were significantly associated with participants’ age, although DV perpetration showed a marginally significant association (p = .067). All the autoregressive paths were statistically significant. With regard to the cross-lagged paths, all of them were statistically significant, with the exception of the cross-lagged path from DV perpetration at T2 to DV victimization at T3 and the cross-lagged path from DV victimization at T3 to DV perpetration at T4. In addition, age at T1 predicted an increase of DV perpetration at T2 and T3, an increase of DV victimization at T2, and a decrease of DV victimization at T4. Modification indices indicated that the model would improve by estimating the direct path from DV perpetration at T1 to DV perpetration at T3 and from DV victimization at T1 to DV victimization at T3. Therefore, the model was re-estimated by including these paths and removing the nonsignificant paths. This model decreased χ² significantly, Δχ²(3, N = 991) = 34.92, p < .001. Figure 1 displays the significant model paths. All the fit indices reflected good fit, S-B χ²(15, N = 991) = 126.83, p < .001, CFI = .99, NNFI = .97, SRMR = .06, with the exception of RMSEA = .087 (90% confidence interval [CI] = [.073, .10]), which was slightly above .08. The model accounted for 25%, 32%, and 56% of the variance in DV perpetration at T2, T3, and T4, respectively; and 29%, 34%, and 41% of the variance in DV victimization at T2, T3, and T4, respectively. This model was also estimated separately for psychological DV and physical DV. The results were similar, except for the following paths that dropped to nonsignificance: the path from perpetration at T1 to perpetration at T3, for the psychological DV model; and also the paths from victimization at T1 to victimization at T3, victimization at T1 to perpetration at T2, victimization at T1 to victimization at T2, and perpetration at T3 to victimization at T4, for the physical DV model. Fit indices reflected an adequate fit for both models. For psychological aggression, S-B χ²(15, N = 991) = 136.86, p < .001, RMSEA = .091 (90% CI = [.077, .10]), CFI = .99, NNFI = .97, SRMR = .07. For physical aggression, S-B χ²(15, N = 991) = 83.82, p < .001, RMSEA = .068 (90% CI = [.054, .083]), CFI = .98, NNFI = .96, SRMR = .06.

Figure 1.

Longitudinal model of dating violence perpetration and victimization.

Sex Differences in the Longitudinal Model

Similarly to the results of the whole sample, all the fit indices were adequate for both sexes, with the exception of RMSEA for females. For females, S-B χ²(15, N = 519) = 99.35, p < .001, CFI = .98, NNFI = .96, SRMR = .056, RMSEA = .10 (90% CI = [.086, .12]). For males, S-B χ²(15, N = 472) = 48.64, p < .001, CFI = .99, NNFI = .98, SRMR = .062, RMSEA = .069 (90% CI = [.048, .091]). Comparison between both samples indicated that some paths were not statistically significant for both sexes (see Figure 2). The paths from DV perpetration at T1 to T2 and from DV perpetration at T1 to DV victimization at T2 were statistically significant only for males, while the paths from DV victimization at T1 to DV perpetration at T2 and from DV perpetration at T2 to T3 were significant only for females. In addition, the association between DV perpetration at T1 and participants’ age was only significant for females.

Figure 2.

Longitudinal model of dating violence perpetration and victimization considering participants’ sex.

Second, to examine whether the model was invariant across females and males, we also tested the configural invariance of the pattern of fixed and free model parameters, χ²(30, N = 991) = 293.21, p < .001, CFI = .99, NNFI = .97, RMSEA = .084 (90% CI = [.070, .099]). Third, we tested the invariance of the variances and covariances of the exogenous factors. This constriction increased chi-square significantly, Δχ²(5, N = 991) = 24.39, p < .001. Therefore, the hypothesis of invariant exogenous factor variances and covariances must be rejected. We examined each parameter separately to identify the differences, and results showed that the association between DV perpetration and victimization at T1 and the association between DV perpetration and participants’ age at T1 were higher for females than males, Δχ²(1, N = 991) = 4.23, p < .05, and Δχ²(1, N = 991) = 15.5, p < .001, respectively. Finally, we tested the invariance of the autoregressive and cross-lagged paths. This constriction also increased χ²significantly, Δχ²(16, N = 991) = 51.81, p < .001, indicating that the overall pattern of paths was different between females and males. The examination of each path separately evidenced that three paths were significantly higher for females than males: the cross-lagged path from DV victimization at T1 to DV perpetration at T2, Δχ²(1, N = 991) = 7.50, p < .01; the autoregressive path from DV perpetration at T3 to T4, Δχ²(1, N = 991) = 21.62, p < .001; and the cross-lagged path from DV perpetration at T3 to DV victimization at T4, Δχ²(1, N = 991) = 5.90, p < .05. All the other paths were similar for the female and male subsamples.

Discussion

The present study explored the stability and longitudinal reciprocal associations between DV perpetration and victimization during 4 years in a sample of high school students from around 15 to 18 years old. Because adolescent dating relationships are characterized by being less stable and shorter in time, stability was assessed regardless of whether or not the participant was with the same partner. Correlation coefficients across time were all statistically significant, and the longitudinal model of DV stability and mutuality showed a good fit to the data (all the model-fit indices reflect a good fit with the exception of RMSEA for females), with all paths being statistically significant except for two cross-lagged paths. Furthermore, the model accounted for a considerable amount of variance, particularly in T4. Thus, our findings evidenced stability of DV, for both females and males, across a broad range of time throughout adolescence. Moreover, longitudinal dyadic influences were demonstrated, but in general the coefficients from one’s partner’s aggression to one’s own perpetration and vice versa were lower than those obtained from stability, two of which were nonsignificant. Next, we will further discuss these findings and their implications.

All the autoregressive paths over time from perpetration and victimization were statistically significant. These results are consistent with our hypothesis about DV perpetration at each time point being predicted by DV perpetration at the previous time point (and the same for DV victimization), as well as the findings of previous studies that found moderate stability of DV (Fritz & Slep, 2009; O’Leary & Slep, 2003). However, contrary to our expectations, both autoregressive paths and correlation coefficients were higher than expected given the much longer time period covered by our study. The coefficients obtained across 4 years were considerably high, even larger than those found by O’Leary and Slep (2003) for physical DV in a sample of teens in stable relationships over 14 weeks, and those found in Fritz and Slep’s (2009) study over approximately 1 year. In fact, they were close to those obtained for physical intimate partner aggression in young adults (O’Leary et al., 1989). Our findings suggest a greater stability in this sample of Spanish teens, although a possible explanation is related to the types of partner aggression assessed. O’Leary et al.’s (1989) and O’Leary and Slep’s (2003) studies explored stability of physical partner aggression, while our study examined different types of aggression (physical, psychological, and sexual) all together, with 17 out of 25 items assessing different forms of psychological aggression (verbal/emotional, relational, and threats). In fact, when our model was estimated separately for physical DV, five of the paths dropped to nonsignificance. In this sense, Fritz and Slep demonstrated a moderate level of stability for both physical and psychological aggression, but they found higher correlations across time for psychological aggression than for physical aggression. Thus, physical aggression might be less stable than psychological aggression during adolescence. This potential greater stability for psychological aggression may be related to the higher prevalence of this type of aggression (Muñoz-Rivas et al., 2007; O’Leary et al., 2008) and to the adolescents’ possible lower perception of its antisocial nature and negative consequences (Fernández-González et al., 2014; Foshee et al., 2009). Future research should clarify whether DV stability is influenced by the type of partner aggression.

Consistent with previous studies (e.g., Chiodo et al., 2012; Fernández-González et al., 2014), the predominance of a mutual pattern of DV was supported by our results. Both females and males reported acts of perpetration and victimization of DV, and of all the correlations, the highest ones were those between perpetration and victimization within time (around .80-.90). Regarding longitudinal reciprocal associations between perpetration and victimization, we hypothesized that an individual’s own aggression will be influenced by the partner’s previous aggression and vice versa. The findings partially supported our hypothesis as all but two of the cross-lagged paths in the tested model were significant. Furthermore, when considering both the autoregressive and cross-lagged paths in the model tested, overall we found the cross-lagged paths to be lower, reflecting a greater influence of the autoregressive paths. These results are contrary to those reported by O’Leary and Slep (2003), who found that the cross-dyad influences were more predictive of aggression than stability. These mixed findings may be related to the differences in the studied time frame and the dating relationship characteristics. The study by O’Leary and Slep was conducted in a sample of adolescents in steady dating relationships during 14 weeks, whereas our study covered 4 years of adolescence regardless of whether or not the participants were with the same partner. Therefore, our findings suggest that in longer time periods throughout adolescence and across partners, the longitudinal dyadic influences are not as strong as in shorter and steady dating relationships, but they are still relevant and reflect the influence of the partner’s aggressive behavior.

Overall, the findings of our study noted both the relevance of considering dyadic influences and the presence of some DV stability in adolescence despite the presumable partner changes. On one hand, one key risk factor for DV perpetration is being a victim of DV and for DV victimization is being a perpetrator of DV. This highlights the relevance of the reciprocal influences of DV, although it may also be a result of the presence of other common risk factors that increase the probability of both being a perpetrator and a victim of partner violence (Reingle, Jennings, Connell, Businelle, & Chartier, 2014). On the other hand, the stability of DV might be explained by characteristics of the adolescent himself or herself, who will accumulate a number of risk factors (related to both individual traits and his or her family and social environment) that predispose him or her to behave aggressively in partner relationships. In addition, assortative partnering by antisocial behavior (Capaldi & Gorman-Smith, 2003) would explain why, even though adolescents change partners frequently, aggressive behaviors remain as a result of continuously selecting partners who also have a higher likelihood of behaving aggressively.

Another noteworthy result is related to the fact that, although all the autoregressive paths were statistically significant, the highest one was the path from DV perpetration at T3 to T4. Similarly, correlation coefficients increased over time and the highest ones were found between T3 and T4 for both perpetration and victimization. These results may reflect a tendency toward increasing DV stability in late adolescence, which supports the idea that partner violence stability appears to increase from initial dating to early marriage (O’Leary & Slep, 2012) and emphasizes the relevance of early preventive interventions.

With regard to the role of participants’ sex, we found a better model fit for males than females, mainly for the RMSEA index. Nevertheless, the longitudinal model of DV perpetration and victimization showed an adequate fit for both sexes, which allows us to conclude that stability and mutuality of DV during adolescence occurred for both females and males. Some paths—mostly between T3 and T4—were significantly higher for females than males, which is consistent with the overall higher correlations found for the females. These results suggest a greater stability of DV perpetration and victimization in late adolescence for females. A possible explanation may be related to the fact that adolescent girls tend to date older boys and that they reported longer and more serious dating relationships during adolescence (e.g., Muñoz-Rivas et al., 2007; O’Leary et al., 2008). However, future research is needed to clarify this finding.

Participants’ age was positively correlated to DV perpetration at T1 (although this association was nonsignificant for males). In addition, a higher age of the participants at T1 predicted an increase of DV perpetration at T2 and T3, higher levels of DV victimization at T2, and lower levels of DV victimization at T4. These results suggest that until T3 (mean age around 16.8 years), the older participants are at an increased risk for DV, while at T4 (mean age around 17.8 years), this trend is reversed such that the older participants are at a lower risk for DV victimization. This could have something to do with the overall trajectories of DV and the peak of physical aggression found around the ages of 16 and 17 years (Foshee et al., 2009; Nocentini et al., 2010).

Limitations

Interpretation of the findings should be conducted in light of some limitations. First, the measure of DV perpetration and victimization is based on a self-report questionnaire that collects information about the participant’s own and their partner’s aggression from only one member of the relationship. This could have introduced biases into the data, mainly with regard to the dyadic results (see studies on couple agreement about dating aggression with aged-college samples; Perry & Fromuth, 2005; Schnurr, Lohman, & Kaura, 2010). However, despite the potential constraints, this assessment methodology is frequent in the research area of intimate partner aggression due to the difficulty of obtaining sizeable samples of couples, which further applies to studies with adolescent samples. Another limitation is related to the use of a shorter version of the CADRI in T1. Some of the items were considered inappropriate for adolescents at that age, but as a consequence of the fewer number of items, this shorter version had lower Cronbach’s alphas. DV stability between T1 and T2 might have been underestimated, although it should also be noted that mean scores did not seem to be influenced by the fewer number of items at T1 (see Table 1). Last, these findings may not be completely generalizable to severe types of aggression, as most of the CADRI items refer to acts of psychological aggression (verbal/emotional abuse, relational abuse, and threatening behavior). Moreover, a number of the paths dropped to nonsignificance when the model was estimated only for physical DV.

Implications for Diversity

Regarding diversity, our results suggest, on one hand, that DV perpetration and victimization are more stable among girls than among boys. Prevention and intervention programs should take this into account as discussed below. On the other hand, the results suggest an increased DV stability in late adolescence. Therefore, prevention and intervention programs should start before this age and be intensified during this period. Our sample had very little diversity as regards race/ethnicity, and therefore, we were unable to examine differences by race or ethnic affiliation. Moreover, our study did not include measures for diversity in terms of sexual preference or religion. The inclusion of these variables could have been interesting for further analyses. For example, in a recent study, sexual minority identity was associated with higher initial levels of DV at baseline, but also with greater decreases in DV across time (Martin-Storey & Fromme, 2016). In another study, religious service attendance was associated with a lower prevalence of DV (Howard, Qiu, & Boekeloo, 2003). Finally, this study was conducted in Spain. Some studies have found higher prevalence rates of DV among Spanish adolescents compared with adolescents from the United States (Muñoz-Rivas et al., 2007) and trajectories could also be different in diverse countries.

Conclusions and General Implications

This study extends the current research on adolescent DV stability and mutuality and is thus one of the few studies in this area covering a broad range of time, that is, 4 years, throughout adolescence. The findings allow us to conclude that partner aggression between adolescents already shows some stability, which appears to increase in late adolescence and thus indicates the relevance of early preventive interventions. In the absence of such interventions, adolescents will accumulate increasingly more risk factors that predispose them to engage in aggressive and unhealthy relationships. Another main conclusion of the study is related to the key role of the partner’s behavior via the reciprocal influence of DV over time even when re-partnering. Therefore, a reasonable way to break this stability is for prevention programs to consider this dyadic influence of DV by increasing the awareness of younger people about the impact of their behavior on their partner and the processes that lead them to select their partners. This does not imply denying that the consequences of partner aggression (especially for physical abuse) are also worse for women than men during adolescence (Fernández-González et al., 2014), but rather that a female partner’s aggression increases the likelihood that she will be a victim of the male partner’s aggression. Finally, some suggestions for future research can be derived from the current study, such as the need to explore the stability of adolescent DV depending on the different types of aggression and their interrelationship. Another important issue is to understand which risk variables are related to a greater stability of DV over time, which is essential in order to develop more effective preventive and intervention programs.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a grant from the Ministerio de Economía y Competitividad (Spanish Government, Ref. PSI2015-68426-R).

Author Biographies

Liria Fernández-González is a postdoctoral researcher at the Faculty of Psychology and Education of University of Deusto, Spain. Her research interest is mainly focused on the etiology, development, and prevention of dating violence.

Esther Calvete is a senior lecturer of psychology at the University of Deusto, Spain, and principal researcher of Deusto Stress Research. Her research focuses on violence against women and child-to-parent violence.

Izaskun Orue is an assistant professor of psychology at the University of Deusto, Spain. Her research focuses on personality, cognitive, and contextual factors associated with aggressive behavior among adolescents.

References

Archer

(2000). Sex differences in aggression between heterosexual partners: A meta-analytic review. Psychological Bulletin, 126, 651-680. doi:10.1037/0033-2909.126.5.651

Basque Institute of Statistics. (2015, November 13). Municipal population statistics 2015. Available from http://www.eustat.eus

Capaldi

D. M.

Gorman-Smith

(2003). The development of aggression in young male/female couples. In Florsheim

(Ed.), Adolescent romantic relations and sexual behavior: Theory, research, and practical implications (pp. 243-278). Mahwah, NJ: Lawrence Erlbaum.

Centers for Disease Control and Prevention. (2010). Youth risk behavior surveillance United States, 2009. Morbidity and Mortality Weekly Report, 59(No. SS-5), 1-148.

Centers for Disease Control and Prevention. (2016). Understanding teen dating violence. Fact sheet 2016. Retrieved from http://www.cdc.gov/violenceprevention/pdf/teen-dating-violence-factsheet-a.pdf

Chiodo

Crooks

C. V.

Wolfe

D. A.

McIsaac

Hughes

Jaffe

P. G.

(2012). Longitudinal prediction and concurrent functioning of adolescent girls demonstrating various profiles of dating violence and victimization. Prevention Science, 13, 350-359. doi:10.1007/s11121-011-0236-3

Choi

H. J.

Temple

J. R.

(2016). Do gender and exposure to interparental violence moderate the stability of teen dating violence? Latent transition analysis. Prevention Science, 17, 367-376. doi:10.1007/s11121-015-0621-4

Chou

C. P.

Bentler

P. M.

Satorra

(1991). Scaled test statistics and robust standard errors for nonnormal data in covariance structure-analysis: A Monte-Carlo study. British Journal of Mathematical & Statistical Psychology, 44, 347-357.

Connolly

Nocentini

Menesini

Pepler

Craig

Williams

T. S.

(2010). Adolescent dating aggression in Canada and Italy: A cross-national comparison. International Journal of Behavioral Development, 34, 98-105. doi:10.1177/0165025409360291

10.

Cornelius

T. L.

Resseguie

(2007). Primary and secondary prevention programs for dating violence: A review of the literature. Aggression and Violent Behavior, 12, 364-375. doi:10.1016/j.avb.2006.09.006

11.

Fernandez-Fuertes

A. A.

Fuertes

(2010). Physical and psychological aggression in dating relationships of Spanish adolescents: Motives and consequences. Child Abuse & Neglect, 34, 183-191. doi:10.1016/j.chiabu.2010.01.002

12.

Fernández-Fuertes

A. A.

Fuertes

Pulido

R. F.

(2006). Assessment of violence in adolescent couples. Validation of the Conflict in Adolescent Dating Relationships Inventory (CADRI)—Spanish version. International Journal of Clinical and Health Psychology, 6, 339-358.

13.

Fernández-González

O’Leary

K. D.

Muñoz-Rivas

M. J.

(2014). Age-Related changes in dating aggression in Spanish high school students. Journal of Interpersonal Violence, 29, 1132-1152. doi:10.1177/0886260513506057

14.

Foshee

V. A.

(1996). Gender differences in adolescent dating abuse prevalence, types and injuries. Health Education Research, 11, 275-286. doi:10.1093/her/11.3.275-a

15.

Foshee

V. A.

Bauman

K. E.

Linder

Rice

Wilcher

(2007). Typologies of adolescent dating violence: Identifying typologies of adolescent dating violence perpetration. Journal of Interpersonal Violence, 22, 498-519. doi:10.1177/0886260506298829

16.

Foshee

V. A.

Benefield

T. S.

Ennett

S. T.

Bauman

K. E.

Suchindran

(2004). Longitudinal predictors of serious physical and sexual dating violence victimization during adolescence. Preventive Medicine, 39, 1007-1016. doi:10.1016/j.ypmed.2004.04.014

17.

Foshee

V. A.

Benefield

Suchindran

Ennett

S. T.

Bauman

K. E.

Karriker-Jaffe

K. J.

. . . Mathias

(2009). The development of four types of adolescent dating abuse and selected demographic correlates. Journal of Research on Adolescence, 19, 380-400. doi:10.1111/j.1532-7795.2009.00593.x

18.

Foshee

V. A.

Linder

MacDougall

J. E.

Bangdiwala

(2001). Gender differences in the longitudinal predictors of adolescent dating violence. Preventive Medicine, 32, 128-141. doi:10.1006/pmed.2000.0793

19.

Foshee

V. A.

Reyes

H. L. M.

Gottfredson

N. C.

Chang

L.-Y.

Ennett

S. T.

(2013). A longitudinal examination of psychological, behavioral, academic, and relationship consequences of dating abuse victimization among a primarily rural sample of adolescents. Journal of Adolescent Health, 53, 723-729. doi:10.1016/j.jadohealth.2013.06.016

20.

Fritz

P. A. T.

Slep

A. M. S.

(2009). Stability of physical and psychological adolescent dating aggression across time and partners. Journal of Clinical Child & Adolescent Psychology, 38, 303-314. doi:10.1080/15374410902851671

21.

Hird

M. J.

(2000). An empirical study of adolescent dating aggression in the U.K. Journal of Adolescence, 23, 69-78. doi:10.1006/jado.1999.0292

22.

Howard

Qiu

Boekeloo

(2003). Personal and social contextual correlates of adolescent dating violence. Journal of Adolescent Health, 33, 9-17. doi:10.1016/S1054-139X(03)00061-2

23.

Jöreskog

K. G.

Sörbom

(2006). LISREL 8.8 [Computer software]. Lincolnwood, IL: Scientific Software International.

24.

Little

T. D.

(2013). Longitudinal structural equation modeling. New York, NY: Guilford Press.

25.

Martin-Storey

Fromme

(2016). Trajectories of dating violence: Differences by sexual minority status and gender. Journal of Adolescence, 49, 28-37. doi:10.1016/j.adolescence.2016.02.008

26.

Muñoz-Rivas

M. J.

Graña

J. L.

O’Leary

K. D.

González

M. P.

(2007). Aggression in adolescent dating relationships: Prevalence, justification, and health consequences. Journal of Adolescent Health, 40, 298-304. doi:10.1016/j.jadohealth.2006.11.137

27.

National Statistics Institute of Spain. (2013, September 23). Socioeconomic levels 2012. Available from http://www.ine.es

28.

Nocentini

Menesini

Pastorelli

(2010). Physical dating aggression growth during adolescence. Journal of Abnormal Child Psychology, 38, 353-365. doi:10.1007/s10802-009-9371-8

29.

O’Leary

K. D.

Barling

Arias

Rosenbaum

Malone

Tyree

(1989). Prevalence and stability of physical aggression between spouses: A longitudinal analysis. Journal of Consulting and Clinical Psychology, 57, 263-268.

30.

O’Leary

K. D.

Slep

A. M. S.

(2003). A dyadic longitudinal model of adolescent dating aggression. Journal of Clinical Child & Adolescent Psychology, 32, 314-327.

31.

O’Leary

K. D.

Slep

A. M. S.

(2012). Prevention of partner violence by focusing on behaviors of both young males and females. Prevention Science, 13, 329-339. doi:10.1007/s11121-011-0237-2

32.

O’Leary

K. D.

Slep

A. M. S.

Avery-Leaf

Cascardi

(2008). Gender differences in dating aggression among multiethnic high school students. Journal of Adolescent Health, 42, 473-479. doi:10.1016/j.jadohealth.2007.09.012

33.

Orpinas

Nahapetyan

Song

McNicholas

Reeves

P. M.

(2012). Psychological dating violence perpetration and victimization: Trajectories from middle to high school. Aggressive Behavior, 38, 510-520. doi:10.1002/ab.21441

34.

Perry

A. R.

Fromuth

M. E.

(2005). Courtship violence using couple data: Characteristics and perceptions. Journal of Interpersonal Violence, 20, 1078-1095. doi:10.1177/0886260505278106

35.

Reingle

J. M.

Jennings

W. G.

Connell

N. M.

Businelle

M. S.

Chartier

(2014). On the pervasiveness of event-specific alcohol use, general substance use, and mental health problems as risk factors for intimate partner violence. Journal of Interpersonal Violence, 29, 2951-2970. doi:10.1177/0886260514527172

36.

Reingle

J. M.

Jennings

W. G.

Maume

M. O.

Komro

K. A.

(2013). The substance-related etiology of teen dating violence victimization: Does gender matter? Women & Criminal Justice, 23, 185-208. doi:10.1080/08974454.2013.802269

37.

Reingle

J. M.

Staras

S. A. S.

Jennings

W. G.

Branchini

Maldonado-Molina

M. M.

(2012). The relationship between marijuana use and intimate partner violence in a nationally representative, longitudinal sample. Journal of Interpersonal Violence, 27, 1562-1578. doi:10.1177/0886260511425787

38.

Schnurr

M. P.

Lohman

B. J.

Kaura

S. A.

(2010). Variation in late adolescents’ reports of dating violence perpetration: A dyadic analysis. Violence and Victims, 25, 84-100. doi:10.1891/0886-6708.25.1.84

39.

Sears

H. A.

Byers

E. S.

Price

E. L.

(2007). The co-occurrence of adolescent boys’ and girls’ use of psychologically, physically, and sexually abusive behaviours in their dating relationships. Journal of Adolescence, 30, 487-504. doi:10.1016/j.adolescence.2006.05.002

40.

Spanish Society of Epidemiology and Family and Community Medicine. (2000). Una propuesta de medida de la clase social [A proposed measure of social class]. Atención Primaria, 25, 350-363.

41.

Wekerle

Wolfe

D. A.

(1999). Dating violence in mid-adolescence: Theory, significance, and emerging prevention initiatives. Clinical Psychology Review, 19, 435-456.

42.

Wolfe

D. A.

Crooks

Jaffe

Chiodo

Hughes

Ellis

. . . Donner

(2009). A school-based program to prevent adolescent dating violence: A cluster randomized trial. Archives of Pediatrics & Adolescent Medicine, 163, 692-699. doi:10.1001/archpediatrics.2009.69

43.

Wolfe

D. A.

Scott

Reitzel-Jaffe

Wekerle

Grasley

Straatman

A.-L.

(2001). Development and validation of the Conflict in Adolescent Dating Relationships Inventory. Psychological Assessment, 13, 277-293. doi:10.1037//1040-3590.13.2.277