See discussions, stats, and author profiles for this publication at:

https://www.researchgate.net/publication/26275889

Persistence and Desistance

of the Perpetration of
Physical Aggression Across
Relationships: Findings
From a National Study of...

Article in Journal of Interpersonal Violence · July 2009

DOI: 10.1177/0886260509334402 · Source: PubMed

CITATIONS READS

42 77

3 authors, including:

Daniel J Whitaker Phyllis Niolon

Georgia State Univer… Centers for Disease C…
85 PUBLICATIONS 3,785 21 PUBLICATIONS 216
CITATIONS CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Phyllis Niolon on 29 May 2015.

The user has requested enhancement of the downloaded file.

J Interpers Violence OnlineFirst, published on June 8, 2009 as doi:10.1177/0886260509334402

Journal of Interpersonal
Violence
Volume XX Number X

Persistence and Desistance Month XXXX xx-xx

© 2009 The Author(s)
10.1177/0886260509334402
of the Perpetration http://jiv.sagepub.com

of Physical Aggression
Across Relationships
Findings From a National Study of Adolescents
Daniel J. Whitaker
Marcus Institute
Brenda Le
Phyllis Holditch Niolon
Centers for Disease Control and Prevention

This study examined the persistent perpetration of physical intimate partner

violence (IPV) across relationships. Based on the National Longitudinal
Study on Adolescent Health, data were analyzed on 6,446 young adults, who
reported on two recent relationships. Frequency and logistic regression
analyses were used to examine the persistence of physical IPV perpetration
across relationships and the predictors of persistent perpetration. Among
individuals who perpetrated physical violence in their first relationship,
29.7% persisted in their perpetration in the second relationship and 70.3%
desisted. Significant predictors of persistent physical IPV in the final multi-
variate model were as follows: IPV frequency in the first relationship, age,
living together versus apart in the subsequent relationship, respondent being
better educated than the partner, and being an IPV victim in second relation-
ship. The persistence of physical IPV across relationships was relatively low,
with desistance being much more common. Factors specific to the second
relationship were the strongest predictors of persistence.

Keywords: intimate partner violence; perpetration; persistence; desistance;

trajectory

N ational estimates of youth’s engagement in partner violence show an

annual victimization rate of almost 9% (Straus & Gelles, 1995), with
studies of high-risk youth showing much higher rates (Watson, Cascardi,
Avery-Leaf, & O’Leary, 2001). Partner aggression is believed to peak in
adolescence and early adulthood (K. O’Leary, 1999), but relatively little is
1
2 Journal of Interpersonal Violence

known about trajectories of aggression perpetration, such as when it starts

and stops and when it increases and decreases. The tendency to measure
partner aggression as, for example, simply occurring or not occurring over
the past year (or lifetime) or with a specific partner leads to the potentially
false conclusion that perpetration is a static construct. That is, those who
have perpetrated partner aggression are referred to as perpetrators or abus-
ers or batterers, thereby suggesting that they will inevitably perpetrate
partner aggression (Corvo & Johnson, 2003). But the perpetration of physi-
cal partner aggression may fluctuate over time and across relationship part-
ners. This study examined (a) the persistence of the perpetration of physical
partner aggression across relationships and (b) the predictors of persistence
versus desistance.
Studies examining the persistence of partner aggression primarily
focused on persistence within a relationship and on adults. Those studies
show substantial variation in persistence. A study of 272 couples who were
engaged to be married found that 51% of physically aggressive men and
59% of physically aggressive women remained aggressive 18 months into
marriage (O’Leary et al., 1989). A study using a nationally representative
sample of couples found 5-year recurrence rates of physical partner aggres-
sion, as broken down by race/ethnicity: 37% for the White population;
52%, Black; and 58% Hispanic (Caetano, Field, Ramisetty-Mikler, &
McGrath, 2005). Data from another national survey indicated that about a
third of aggressive couples remained aggressive over 5 years (Jasinski,
2001), and a nonrepresentative sample found a persistence rate of more
than 75% among physically aggressive husbands 1 to 2 years into marriage
(Quigley & Leonard, 1996).
Although the persistence of physical aggression may be relatively high
within a relationship, it is unclear how physical partner aggression per-
petration persists across relationship partners. This is a critical question
because intimate partner violence (IPV) perpetration is highest among young
people (K. O’Leary, 1999) and young people change partners frequently.
Three studies have examined the persistence of physical aggression across
relationships, all using nonrepresentative samples of young adults. One found
that persistence of physical aggression was more common among men who

Authors’ Note: The findings and conclusions of this study are those of the authors and do not
necessarily represent the views of the Centers for Disease Control and Prevention. We would
like to thank Linda Dahlberg and Janet Saul for their comments and suggestions on this article.
Address correspondence to Daniel J. Whitaker, PhD, Marcus Institute, 1920 Briarcliff Road,
Atlanta, GA 30329; e-mail: [email protected].
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 3

stayed with a partner (about two thirds persisted) than among those who
changed partners (42% persisted; Capaldi, Shortt, & Crosby, 2003). A sec-
ond study focused on at-risk males and found the opposite result: The
overall persistence rate among men who stayed with a partner (51%) was
nominally but not significantly lower than that of men who changed part-
ners at the follow up (58%; Woffordt, Mihalic, & Menard, 1994). A third
study of young adults (early twenties) found that over 5 years, the persis-
tence rates did not differ between respondents who had the same partners
versus those who had different partners (Robins, Caspi, & Moffitt, 2002).
This study used the National Longitudinal Study of Adolescent Health
(Add Health), a nationally representative sample of young adults, to exam-
ine persistence of physical partner aggression. In Add Health, more than
7,000 young adults reported on their two most important relationships, and
we ordered those relationships temporally to estimate the persistence of
perpetration of physical partner aggression and to examine whether persis-
tence varied between men and women.
In addition, we examined predictors of persistence among those who were
physically violent in their first relationship. We focused on three classes
of predictors: individual-level variables (socioeconomic status, age, drug/
alcohol use), variables associated with the nature of the IPV in the first
relationship (e.g., severity, frequency), and variables associated with the sub-
sequent relationship. Each class of variables has a theoretical and/or empiri-
cal basis for hypothesizing why they might predict physical partner aggression
persistence. Theories that emphasize stable individual-level factors, such as
attitudes, personality traits (psychopathy, borderline traits), and child abuse
victimization (Dutton & Nicholls, 2005), would predict IPV perpetration to
be reasonably consistent across partners. In contrast, theories that emphasize
the dynamic nature of IPV suggest that IPV arises from interactional patterns
between partners (Capaldi & Kim, 2007) and that new relationships with
different interactional patterns might result in less or more IPV perpetration.
Finally, research has found that the one of the strongest predictors within
relationships of physical aggression is the nature of the aggression, with more
severe aggression being more likely to persist (O’Leary et al., 1989).
In general, we selected variables for inclusion that had been shown in
prior research to relate to the perpetration of IPV, to examine whether these
variables were related to the persistence of IPV perpetration. Research
reviews suggest that relationship variables (e.g., concurrent partner aggres-
sion, relationship conflict) are the most strongly variables related to the
perpetration of physical partner aggression (Stith, Smith, Penn, Ward, &
Tritt, 2004), thus suggesting that they strongly relate to the persistence of
4 Journal of Interpersonal Violence

physical aggression as well. In selecting variables for inclusion in this

analysis, we selected variables that were found in prior research to predict
perpetration of physical partner aggression.

Method

Participants
All respondents were part of Add Health and participated in the third
wave of data collection, during 2001. Add Health designers used a multi-
stage stratified cluster design to identify a nationally representative sample
of adolescents (complete details regarding Add Health are found elsewhere;
see Harris et al., n.d.). In 1995, 18,924 adolescents in middle school and
high school (ages 12–21) participated in the first wave of Add Health’s in-
home interview. Six years later, 14,322 participants (77.4% of those who
completed the Wave 1 survey) completed the in-home survey of Wave 3
(ages 18–26). In unpublished analyses, the Add Health study team deter-
mined that participant nonresponse for Wave 3 has minimal impact on the
sample’s representativeness (Chantala, Kalsbeek, & Andraca, 2004).
The current analyses focus primarily on Wave 3 data and so include
participants who reported information about at least two relationships (n
7,180). Interview participants were asked to report an inventory of all their
sexual and romantic relationships during the past 5 years (the mean number
of relationships among those who reported any was 3.08, SD 2.7); then,
they were asked a short series of questions about each relationship, with
specific types of relationships selected for more detailed questions.
Relationships that were selected for questions on physical partner aggres-
sion were “important” relationships. Importance was defined by a preset
algorithm based on such as marital status, recency, and duration of relation-
ship. We used the two most recent important relationships in this analysis
(only 39 participants reported on more than two relationships). There were
6,446 participants for whom we could establish a clear temporal ordering
of relationships, as based on the start date and whether the relationship was
current. Table 1 presents sample characteristics and comparisons of the
initial relationship (R1) and the subsequent relationship (R2).

Measures
All measures were taken from Wave 3 of the Add Health survey, with two
exceptions (described later). In the following sections, we detail the variables
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 5

Table 1
Individual Characteristics by Sex
and Relationship Characteristics by Relationship Number
Total Men Women
(n 6,446) (n 2,896) (n 3,550)

Individual Characteristics n (%) n (%) n (%)

Age, mean years 21.8 21.9 21.7

Race/ethnicity
White 3,879 (73.7) 1,743 (73.5) 2,136 (73.9)
Black 1,205 (13.2) 483 (12.8) 722 (13.7)
Hispanic 892 (9.4) 452 (10.3) 440 (8.6)
Other 442 (3.7) 203 (3.4) 239 (3.8)
Education
Less than high school 499 (9.2) 256 (10.1) 243 (8.4)
High school 2,129 (34.6) 1,012 (37.5) 1,117 (32.1)
Some college 2,944 (44.9) 1,295 (42.8) 1,649 (46.8)
College graduate 752 (11.2) 286 (9.6) 466 (12.8)

Relationship 1 Relationship 2
Relationship Characteristics (n 6,446) (n 6,446) p

Participant age at 17.4 19.9 .001

relationship, mean years
Relationship length 89.4 90.7 .03
> 3 months, %
Cohabitation status, % .001
Not living together 75.3 56.5
Living together 20.5 30.4
Married 4.2 13.1
Relationship current at time 5.1 68.7 .001
of interviewa
Education differences, % .001
Same 42.5 45.0
Respondent more 39.7 32.7
Respondent less 17.8 22.3
Age difference, % .001
Same 74.2 63.3
Respondent older 5.6 9.0
Respondent younger 20.2 27.7
Ever had sex, % 88.5 94.5 .001
Love, mean on 0–4 scale 1.48 2.49 .001
Material commitment, % 12.3 31.5 .001

a. All participants who indicated that Relationship 1 was current also indicated that
Relationship 2 was current but that the former began before the latter.
6 Journal of Interpersonal Violence

used for the perpetration of physical partner aggression during R1 and R2 and
the variables at each level previously described. Predictor variables are
grouped into three categories: individual-level variables, variables associated
with R1 partner physical aggression, and variables associated with the subse-
quent relationship. The number of variables included in each category was in
part a function of what was available and appropriate for use in the data set.

Respondents’ physical partner aggression perpetration in each relation-

ship. To assess the perpetration of physical partner aggression, respondents
answered two questions about physical aggression in the past year of the
relationship, for each relationship (e.g., being threatened, pushed, shoved,
or thrown something; being slapped, hit, or kicked). Respondents used a
frequency scale, and we used responses to create dichotomous measures of
any physical partner aggression perpetration for each relationship.

Individual-level variables. The analyses included nine self-reported

individual-level variables. Sex and race/ethnicity were self-reported via
single items, the latter of which was collapsed into the categories of White,
Black, Hispanic, and other. Four items from Wave 1 assessed the most recent
high school grades in English, math, social studies, and science, and we
averaged those to measure grade point average as the indicator of educa-
tional achievement. Four items from Wave 3 measured history of child
maltreatment, including neglect and abuse (e.g., not having basic needs met,
being hit or kicked, being touched in a sexual way). Responses were com-
bined into a dichotomous variable of either experiencing any child abuse or
not. Early peer violence perpetration was measured with three items from
Wave 1 that asked whether the participant had ever got into a fight, pulled a
knife or gun on anyone, or shot or stabbed anyone during the past 12 months.
Recent peer violence perpetration was similarly measured with four items
from Wave 3 on fighting and weapon use. Combined responses were zero-
inflated and thus dichotomized into none versus any early and recent peer
violence. A continuous index of problem drinking was created by averaging
responses to eight items (alpha .77) assessing whether drinking caused
individual problems during the past year (e.g., with school, work, friends,
family). A dichotomous index of drug use (none versus any) was created
from five questions assessing use of marijuana, cocaine, methamphetamines,
and other illegal drugs since the study began (approximately 6 years).

R1 partner physical aggression factors. The predictive analyses included

three characteristics of respondents’ physical partner aggression during R1.
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 7

A continuous measure of the frequency of physical partner aggression per-

petrated by the respondent was created by averaging the two questions on
physical partner aggression perpetration in R1. Similarly, a continuous
measure of respondent’s victimization in R1 (i.e., how much the respon-
dent’s partner perpetrated against him or her) was created using the two
parallel questions on the respondent’s victimization in R1. Injury to the
partner was a dichotomous measure created from a single item assessing
whether the respondent had caused an “injury, sprain, bruise, or cut” to the
partner with his or her physical aggression.

R2-specific factors. Nine variables were examined specific to R2.

Respondent age at the beginning of R2 was measured with a single
item. Relationship length was assessed with a single question: Has the rela-
tionship lasted more than 3 months? Almost 90% of relationships were
greater than 3 months, and no further information was available in the data
set. Cohabitation status was assessed on the basis of self-reports (married,
living together but not married, and not living together nor married), as was
sex in the relationship (yes/no). Age difference between respondent and
partner was classified as follows: respondent older by 2 or more years,
respondent and partner the same age, respondent younger by 2 or more
years. Educational difference was examined by assessing the respondent’s
and partner’s highest grade levels completed (less than high school, high
school, some college, college graduate or higher) and by classifying the
respondent as higher, equal, or lower. Respondent victimization in R2 was
determined by questions identical to those described for R1 to create a
dichotomous index (any victimization versus none). A dichotomous mea-
sure of material commitment was created by combining two yes/no ques-
tions: whether the couple had a joint bank account and whether the couple
had together purchased anything costing more than $500. A continuous
measure of love was created by averaging two items (r .77) assessing the
respondent’s and partner’s love for each other. Because of the high correla-
tion between the two variables, it was not possible to examine discrepan-
cies in love between respondent and partner.

Analyses
We first examined the persistence and desistance of physical partner
aggression perpetration via frequency analyses; then, we examined whether
there were sex differences in rates of persistence via logistic regression
(i.e., testing whether sex interacted with R1 perpetration to predict R2
8 Journal of Interpersonal Violence

Table 2
Physical Partner Aggression and Its Persistence
and Desistance, by Relationship and Sex
All Men Women
(N 6,446) (n 2,896) (n 3,550)

Percentage perpetrated
Relationship 1 18.1 11.2 24.5
Relationship 2 15.8 9.8 21.3
Percentage perpetrated in Relationship 2
Relationship 1 perpetrators (n 1,212) 29.7 26.4 31.1
Relationship 1 nonperpetrators (n 5,234) 12.7 7.8 18.1

perpetration). To examine the predictors of persistence, we first examined

each predictor variable individually and whether it related to persistence
differently between men and women. We did so by regressing persistence
onto each predictor, sex, and the interaction between the variables, with
the interaction term indicating significant sex differences or not. Only 2 of
the 20 Predictor Sex interactions were significant, and results for those
variables are presented separately for men and women. We also examined
the three categories of variables (individual factors, R1 partner physical
aggression factors, and R2 factors) in separate multivariate models. For
each model, we first constructed a main effects model; then, in a second
step, we added any significant Sex × Predictor interactions from the bivar-
iate models. For the final multivariate model, we entered all variables sig-
nificant at p .10 from each main effect model.

Results
Table 2 shows the frequency and weighted proportion of individuals
reporting perpetrating physical partner aggression in R1 and R2. Perpetration
was more common in the initial relationship (18.1%) than in the second
(15.8%), 2 76.64, p .001. Women were more likely than men to report
perpetration in both R1 (24.5% versus 11.1%), 2 94.67, p .001, and R2
(21.3% versus 9.8%), 2 72.76, p .001.
Across relationships, 29.7% of R1 perpetrators reported perpetrating
physical aggression in R2 and 70.3% reported not perpetrating in R2.
Among R1 nonperpetrators, 12.7% reported perpetrating in R2 and 87.3%
reported not perpetrating in R2. Thus, not surprisingly, R1 perpetration was
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 9

associated with increased R2 perpetration (odds ratio [OR] 2.9, 95%

confidence interval [CI] 2.4, 3.5). Examining these trends by sex revealed
a Sex × R1 Perpetration × R2 Perpetration interaction, 2 76.64, p .001;
that is, R1 perpetration was a stronger predictor of R2 perpetration for men
(OR 4.3, 95% CI 2.9, 6.3) than for women (OR 2.0, 95% CI 1.6,
2.6). However, the interaction was driven by a greater proportion of
women nonperpetrators in R1 beginning perpetration in R2, as compared
to men (18.1% versus 7.8%), 2 60.39, p .001. Among R1 perpetrators,
women and men were equally likely to persist in their perpetration in R2
(31.1% and 26.4%, respectively), 2 1.56, p .21.
Table 3 shows proportions of persisting perpetrators of physical aggres-
sion (or means for persisting and desisting perpetrators) and unadjusted ORs
from logistic regression models. As noted above, we tested sex interactions
for each predictor variable, and where those interactions were significant,
we present findings separately for men and women; specifically, 2 of the 20
sex interactions were significant. Among individual-level variables, those
who persisted in physical aggression across relationships, compared to
desisters, were more likely to have a history of child maltreatment (OR
1.5), have perpetrated recent peer violence (OR 1.6), and have reported
drug use (OR 1.7). Past peer violence was associated with a lower persis-
tence of physical partner aggression among men (OR 0.5) but not among
women (OR 1.2). Among R1 violence-related variables, predictors of per-
sistence include violence frequency (OR 1.3) and injuring the partner (but
only for men: OR 3.3). Among R2-specific variables, six of the nine vari-
ables were associated with persistence: younger respondent age at the start of
R2, a longer relationship, living with or being married to the partner, greater
material commitment, greater love, and respondent victimization. Respondent
victimization in R2 was strongly related to persistence (OR 26.1).
The first three columns of Table 4 present the results from the three mul-
tivariate models focusing on category of predictors. The first model included
the eight individual-level predictors, plus the interaction between past peer
violence and sex, and it accounted for 4.5% of the variance in the persistence
of partner aggression. The only two significant predictors were drug use
(adjusted odds ratio [AOR] 1.7) and the interaction between past peer
violence and sex (for men, AOR 0.3, p .05; for women, AOR 1.0, ns).
The second model included the three R1 partner physical aggression vari-
ables plus the Sex × Injuring the Partner interaction. The interaction was not
significant (p .14) and was thus dropped. The model accounted for 1.6% of
the variance in persistence, and the only significant predictor was R1
10 Journal of Interpersonal Violence

Table 3
Bivariate Results of Physical Partner
Aggression Persistence Across Relationships
Mean for Odds Ratio
Percentage Persisters (95% Confidence
Factors/Variables Subgroup Persistent and Desisters Interval) p

Individual-level
factors
Sex Male 26.4 0.8 (0.6, 1.2) .22
Female 31.1 —
Race/ethnicity White 27.0 —
Black 34.0 1.4 (0.9, 2.1) .13
Hispanic 36.3 1.5 (0.9, 2.6) .10
Other 37.8 1.6 (0.6, 4.1) .28
Grade point Persisters 2.8 1.1 (0.8, 1.4) .58
average Desisters 2.8
History of child Yes 32.8 1.5 (1.1, 2.1) .01
maltreatment No 24.4 —
Past history
of peer violencea
Men Yes 20.3 0.5 (0.3, 0.9) .03
No 34.5
Women Yes 33.3 1.2 (0.8, 1.7) .45
No 30.0
Recent history Yes 37.9 1.6 (1.0, 2.5) .05
of peer violence No 27.9 —
Problem drinking Persisters .50 1.3 (1.0, 1.7) .10
Desisters .42 —
Drug use Yes 33.3 1.7 (1.2, 2.3) .002
No 23.1 —
R1 partner physical
aggression factors
Violence frequency Persisters 1.37 1.3 (1.1, 1.5) .001
in R1 Desisters 1.08
Victimization at R1 Yes 31.51 1.3 (0.9, 1.8) .16
No 26.40 —
Injured partner in R1a
Men Yes 49.49 3.3 (1.4, 8.2) .01
No 22.74
Women Yes 34.68 1.2 (0.8, 1.9) .38
No 30.38
R2-specific variables
Age at R2 Persisters 19.7 0.9 (.81, .94) .001
Desisters 20.3 —
(continued)
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 11

Table 3 (continued)
Mean for Odds Ratio
Percentage Persisters (95% Confidence
Factors/Variables Subgroup Persistent and Desisters Interval) p

Romantic R2 lasted Yes 31.11 2.2 (1.1, 4.2) .03

3 months No 17.24 —
Cohabitation status Not living 17.27 —
together
Living 43.73 3.7 (2.6, 5.4) .001
together
Married 32.00 2.3 (1.3, 3.8) .003
Educational Respondent 34.99 1.48 (0.9, 2.1) .11
differences at R2 higher
Equal 28.01 —
Respondent 26.72 0.9 (0.6, 1.5) .78
lower
Age difference Respondent 35.93 1.5 (0.7, 3.1) .24
at R2 older
Same 27.56 —
age ( 2)
Respondent 31.94 1.2 (0.9, 1.8) .30
younger
Sex in R2 Yes 30.12 2.3 (0.5, 10.0) .28
No 16.07 —
Material Yes 36.95 1.6 (1.2, 2.2) .002
commitment No 26.33 —
Love Persisters 2.60 1.3 (1.1, 1.7) .007
Desisters 2.39
Respondent Yes 80.76 26.4 (15.7, 44.6) .001
victimization No 13.72 —

Note: n 1,212 participants who perpetrated physical partner aggression in R1. R1

Relationship 1; R2 Relationship 2.
a. Interaction with sex significant at p .05; thus, results are displayed separately for men and
women.

violence frequency (AOR 1.3). The third model included the nine R2 fac-
tors. The overall model accounted for 35.6% of the variance in persistence.
Four variables were significant predictors of persistence: younger age at R2
(AOR 0.9), living together versus apart (AOR 2.4), respondent more
educated than partner (AOR 1.9), and respondent victimization in R2
(AOR 25.7). By itself, respondent victimization accounted for 31% of the
12
Table 4
Multivariate Logistic Regression Models Predicting
Physical Partner Aggression Persistence Versus Desistance
Individual Factors R1 Violence Factors R2 Factors Final Multivariate Model
(R2 .045) (R2 .016) (R2 .350) Model (R2 .355)

AOR (95% CI) p AOR (95% CI) p AOR (95% CI) p AOR (95% CI) p

Individual variables
Male (ref female) 0.8 (0.5, 1.2) .23
Race (ref White) .11
Black 1.6 (1.0, 2.5) .07
Hispanic 1.6 (0.9, 2.6) .09
Other 1.6 (0.6, 4.4) .36
Grade point average 1.1 (0.9, 1.4) .29
Child maltreatment 1.3 (0.9, 1.8) .20
(ref none)
Past peer violencea
Men (ref none) 0.3 (0.2, 0.8) .01
Women (ref none) 1.0 (0.6, 1.6) .98
Recent peer violence 1.7 (1.0, 2.9) .06 0.8 (0.5–1.4) .51
(ref none)
Problem drinking 1.1 (0.8, 1.5) .56
(higher more)
Drug use 1.7 (1.2, 2.4) .004 0.9 (0.6–1.5) .76
R1 partner physical
aggression variables
Violence frequency 1.3 (1.1, 1.5) .009 1.4 (1.1–1.6) .002
in R1
Victimized during R1 1.1 (0.8, 1.7) .56
Injured partner in R1 1.2 (0.8, 2.0) .41

(continued)
 Table 4 (continued)
Individual Factors R1 Violence Factors R2 Factors Final Multivariate Model
(R2 .045) (R2 .016) (R2 .350) Model (R2 .355)

AOR (95% CI) p AOR (95% CI) p AOR (95% CI) p AOR (95% CI) p

R2 variables
Age at R2 0.9 (.80–.96) .007 0.9 (.80–.97) .02
R2 lasted 3 months 0.9 (0.4–2.2) .73
(ref no)
Cohabitation status .002 .004
(ref not living
together)
Living together 2.4 (1.5–4.0) .001 2.6 (1.6–4.2) .001
Married 1.4 (0.7–3.0) .35 1.5 (0.8–2.8) .20
Educational .01 .02
differences at R2
(ref same)
Respondent higher 1.9 (1.1–3.2) .02 1.7 (1.1–2.9) .03
Respondent lower 0.8 (0.4–1.5) .51 0.7 (0.4–1.3) .25
Age difference at .59
R2 (ref same)
Respondent older 0.8 (0.2–2.5) .66
Respondent younger 1.3 (0.7–2.1) .41
Sex in R2 (ref no) 1.1 (0.1–13.4) .93
Material commitment 0.8 (0.4–1.4) .43
(ref no)
Level of love in R2 1.2 (0.9–1.7) .25
Partner violence 25.7 (14.8–44.5) .001 26.1 (15.2–44.6) .001
at R2 (ref no)

Note: AOR adjusted odds ratio; CI confidence interval; R1 Relationship 1; R2 Relationship 2.

13
a. Interaction with sex significant at p .05; thus, results are displayed separately for men and women.
14 Journal of Interpersonal Violence

variance, but removal of this variable from the multivariate model had little
overall impact on the other coefficients.
In a final multivariate regression model, we simultaneously entered all
variables from each of the prior three models that were significant at the
p .10 level: recent peer violence, drug use, R1 violence frequency, age at
R2, relationship status, educational differences, and respondent victimiza-
tion. We also included the Sex × Past Peer Violence interaction, but it was
not significant (p .26) and was thus dropped with the main effects of sex
and past peer violence, which were not significant in the initial models. The
last column of Table 4 displays the results from the final model. The overall
model explained 35.5% of the variance, and five variables were significant
predictors of persistence: frequency of physical partner aggression in R1
(AOR 1.4), respondent age at the start of R2 (AOR 0.9), living together
versus not living together (AOR 2.6), the respondent being more edu-
cated than the partner (AOR 1.7), and concurrent R2 victimization (AOR
26.1). Again, removing concurrent R2 victimization had little effect on
the other predictors.

Discussion

In one of the first large-scale studies to examine the persistence of

physical partner aggression across relationships, we found that desistance
was much more common (70.3%) than persistence (29.7%). We also found
that men’s and women’s persistence of physical partner aggression perpe-
tration was nearly identical. This is important because men are considered
more likely to be chronic perpetrators of partner aggression, whereas
women’s perpetration has been attributed to circumstantial factors such as
self-defense (Corvo & Johnson, 2003; Hamberger & Potente, 1994).
In predictive analyses, variables from each class of predictors related to
the persistence of physical partner aggression, but the strongest predictors
were those of the second relationship. This finding is suggestive of the
importance of relationship-specific factors in determining whether persons
with a history of physical partner aggression perpetration continue that
behavior in new relationships. Concurrent victimization was a strong pre-
dictor, with an OR of 26. Perhaps, partners of persistent perpetrators hit
back after being victim to aggression or retaliated if their partners initiated
aggression. Other studies have found that partners’ IPV perpetration rates
tend to change together over time in similar ways (Robins et al., 2002),
which may reflect the reciprocal nature of some partner violence. There
were relatively few sex differences in the predictors of the persistence of
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 15

physical aggression across relationships (although liberal statistical methods

were used to maximize the likelihood of finding significant sex interactions);
as such, the processes leading to persistence are similar for the men and
women in this study.
Our findings suggest that the persistence of physical aggression across
relationships may be lower than the persistence within a relationship. Rates
for the persistence of physical partner aggression within a relationship have
been shown to vary, from 33% to 75% over 1 to 5 years (Caetano et al.,
2005; Jasinski, 2001; O’Leary et al., 1989; Quigley & Leonard, 1996), but
in this analyses, persistence across relationships was much lower. Our find-
ings suggest that some of the predictors of persistence in physical partner
aggression across relationships are similar to factors that predict persis-
tence within a relationship. Persistence of physical aggression within rela-
tionships has been predicted by variables such as young age (Jasinski,
2001), severity of aggression (Robins et al., 2002), verbal aggression
(Aldarondo & Sugarman, 1996), concurrent partner’s aggression (Feld &
Straus, 1989; Robins et al., 2002; Schumacher & Leonard, 2005) cohabi-
tation (Jasinski, 2001), and marital conflict (Aldarondo & Sugarman,
1996). We found that frequency of physical partner aggression, concur-
rent partner aggression, and cohabitation in the second relationship pre-
dicted persistence across two relationships. This makes sense because,
even across relationships, individuals’ partner choices (Capaldi & Crosby,
1997) and relationship behaviors may remain somewhat consistent
(Robins et al., 2002). Limits of the Add Health methods and measures did
not allow us to fully test all relevant variables that have been examined
within relationships.
A final point of interest is that many of the variables that typically pre-
dict the occurrence of partner aggression (e.g., early peer violence, child
maltreatment; Fang & Corso, 2007) were either weakly related or unrelated
to the persistence of physical partner aggression. Thus, variables that pre-
dict the occurrence of physical partner aggression may not predict its per-
sistence (Jasinski, 2001), which suggests that relationship interventions
may be fruitful intervention points. Dyadic interventions have been contro-
versial owing to fears about victim safety, but a growing consensus believes
that they are appropriate under some circumstances (Bograd & Mederos,
1999; O’Leary, 2002). Although dyadic approaches to partner aggression
prevention are still being developed, initial results are promising (Whitaker,
Baker, & Arias, 2007).
From a theoretical perspective, the results suggest that physical partner
aggression may be viewed as a relationship-specific phenomenon. Most
16 Journal of Interpersonal Violence

perpetrators of partner aggression in Relationship 1 became nonperpetra-

tors in Relationship 2, and the best predictors of persistence of partner
aggression across relationships were relationship variables; as such, the
determinants of partner violence are relationship specific. Theories may
overemphasize individual-level characteristics as determinants of IPV per-
petration (Stith et al., 2004; Capaldi & Kim, 2007) at the expense of rela-
tionship-level characteristics. Individual characteristics may also interact
with relationship-level characteristics to result in partner violence. There
has been much discussion in the literature on the way in which relationship
types fit together to result in more or less functional relationships. The present
study did not examine interactions between levels of predictors, however.

Limitations and Strengths

This study has a number of limitations, most of which pertain to the
study methods and measures. All data were collected via self-report from
one partner; thus, all the problems and biases of self-report measures are
relevant, including recall bias, underreporting of negative behaviors, and
inflated relationships of variables owing to increased method variance. In
addition, reports of partner violence were collected with some variables
(material commitment, level of love), and other variables were collected
with varying time frames (e.g., drug and alcohol use, peer violence); as
such, it is difficult, if not impossible, to determine whether certain variables
preceded or proceeded partner violence. Second, the analyses were con-
ducted using a nationally representative sample of young adults aged 18 to
26 and may thus be nongeneralizable to populations outside that age range.
Some argue that IPV among teens and young adults is common, low level,
and not serious, although several studies have observed serious partner
violence among younger samples. Here, the levels of reported partner
aggression perpetration were relatively low. Furthermore, some have
argued that community samples include few cases of serious clinical levels
of partner violence (Johnson, 1995). Others have argued the opposite, how-
ever (Ehrensaft, Moffitt, & Caspi, 2004). In addition, respondents reported
on their two most important relationships, with importance based largely
based on length of relationship; thus, the results may be applicable only to
such types of relationships. This limitation may be consequential to rates of
persistence; that is, respondents may have had other relationships that dis-
solved quickly because of relationship aggression and were thus not included
as important relationships. The measures also have limitations. Only two
questions assessed physical partner aggression perpetration; thus, many
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 17

types of violent behaviors were not measured, including sexual violence,

more serious forms of physical violence (beating up, choking, weapon use),
and the controlling behaviors that can accompany severe physical IPV.
Motives, meaning, and the context of aggression were also not measured.
Finally, only one self-report item assessed partner’s injury, and it did not
assess injury severity. Self-report measures of inflicted injury are problem-
atic because an injury may not be readily apparent to the perpetrator.
The study does have a number of strengths. The Add Health sample is
a nationally representative sample of adolescents that has been adequately
retained over a period of 6 years. The study is comprehensive and so
allows for a range of variables to be included in the analyses, including
relationship-level variables. The size of the sample allowed us to not only
focus analyses on the subset of individuals who perpetrated physical
aggression in the first relationship but to do so with sufficient power to
conduct most subgroup analyses. To our knowledge, this study is the first
to examine the persistence of violence across relationships in a nationally
representative sample.

Conclusion

The perpetration of physical partner aggression does not necessarily

persist across relationships. This is encouraging news for prevention
efforts—especially, those concerned with primary prevention. Although
primary prevention with adolescents seeks to begin before violent behavior
has occurred, persons who have already perpetrated partner aggression
against one partner may not do so to a future partner. Interventions for those
perpetrators may further reduce their tendency to aggress against a partner
(Foshee et al., 2005), and we must seek to further understand the course of
partner aggression across relationships.

References
Aldarondo, E., & Sugarman, D. B. (1996). Risk marker analysis of the cessation and persis-
tence of wife assault. Journal of Consulting and Clinical Psychology, 64(5), 1010-1019.
Bograd, M., & Mederos, F. (1999). Battering and couples therapy: Universal screening and
selection of treatment modality. Journal of Marital and Family Therapy, 25(3), 291-312.
Caetano, R., Field, C. A., Ramisetty-Mikler, S., & McGrath, C. (2005). The 5-year course of
intimate partner violence among White, Black, and Hispanic couples in the United States.
Journal of Interpersonal Violence, 20 (9), 1039-1057.
18 Journal of Interpersonal Violence

Capaldi, D. M., & Crosby, L. (1997). Observed and reported psychological and physical
aggression in young, at-risk couples. Social Development, 6(2), 184-206.
Capaldi, D. M., & Kim, H. K. (2007). Typological approaches to violence in couples: A cri-
tique and alternative conceptual approach. Clinical Psychology Review, 27(3), 253-265.
Capaldi, D. M., Shortt, J. W., & Crosby, L. (2003). Physical and psychological aggression in
at-risk young couples: Stability and change in young adulthood. Merrill-Palmer Quarterly,
49 (1), 1-27.
Chantala, K., Kalsbeek, W. D., & Andraca, E. (2004). Non-response in Wave III of the Add Health
Study. Chapel Hill, NC: University of North Carolina, Carolina Population Center. Retrieved
March 22, 2007, from http://www.cpc.unc.edu./projects/addhealth/files/W3nonres.pdf
Corvo, K., & Johnson, P. J. (2003). Vilification of the “batterer”: How blame shapes domestic
violence policy and interventions. Aggression and Violent Behavior, 8(3), 259-282.
Dutton, D. G., & Nicholls, T. L. (2005). The gender paradigm in domestic violence research
and theory: Part 1—The conflict of theory and data. Aggression and Violent Behavior,
10 (6), 680-714.
Ehrensaft, M. K., Moffitt, T. E., & Caspi, A. (2004). Clinically abusive relationships in an
unselected birth cohort: Men’s and women’s participation and developmental antecedents.
Journal of Abnormal Psychology, 113(2), 258-271.
Fang, X., & Corso, P. S. (2007). Child maltreatment, youth violence, and intimate partner violence
developmental relationships. American Journal of Preventive Medicine, 33(4), 281-290.
Feld, S. L., & Straus, M. A. (1989). Escalation and desistance of wife assault in marriage.
Criminology, 27, 141-161.
Foshee, V. A., Bauman, K. E., Ennett, S. T., Suchindran, C., Benefield, T., & Linder, G.
(2005). Assessing the effects of the dating violence prevention program “safe dates” using
random coefficient regression modeling. Prevention Science, 6(3), 245-258.
Hamberger, L., & Potente, T. (1994). Counseling heterosexual women arrested for domestic
violence: Implications for theory and practice. Violence and Victims, 9 (2), 125-137.
Harris, K. M., Florey, T., Tabor, J., Bearman, P. S., Jones, J., & Udry, J. R. (n.d.) The National
Longitudinal Study of Adolescent Health: Research design. Retrieved September 15, 2005,
from http://www.cpc.unc.edu/projects/addhealth/design
Jasinski, J. L. (2001). Physical violence among Anglo, African American, and Hispanic couples:
Ethnic differences in persistence and cessation. Violence and Victims, 16(5), 479-490.
Johnson, M. P. (1995). Patriarchal terrorism and common couple violence: Two forms of
violence against women. Journal of Marriage & the Family, 57(2), 283-294.
O’Leary, K. (1999). Developmental and affective issues in assessing and treating partner
aggression. Clinical Psychology: Science and Practice, 6(4), 400-414.
O’Leary, K. D. (2002). Conjoint therapy for partners who engage in physically aggressive
behavior: rationale and research. Journal of Aggression, Maltreatment, and Trauma, 5,
145-164.
O’Leary, K. D., Barling, J., Arias, I., Rosenbaum, A., Malone, J., & Tyree, A. (1989).
Prevalence and stability of physical aggression between spouses: A longitudinal analysis.
Journal of Consulting and Clinical Psychology, 57(2), 263-268.
Quigley, B. M., & Leonard, K. E. (1996). Desistance of husband aggression in the early years
of marriage. Violence and Victims, 11(4), 355-370.
Robins, R. W., Caspi, A., & Moffitt, T. E. (2002). It’s not just who you’re with, it’s who you
are: Personality and relationship experiences across multiple relationships. Journal of
Personality, 70 (6), 925-964.
 Whitaker et al. / Perpetration of Physical Aggression Across Relationships 19

Schumacher, J. A., & Leonard, K. E. (2005). Husbands’ and wives’ marital adjustment, verbal
aggression, and physical aggression as longitudinal predictors of physical aggression in
early marriage. Journal of Consulting and Clinical Psychology, 73(1), 28-37.
Stith, S. M., Smith, D. B., Penn, C. E., Ward, D. B., & Tritt, D. (2004). Intimate partner phy-
sical abuse perpetration and victimization risk factors: A meta-analytic review. Aggression
and Violent Behavior, 10 (1), 65-98.
Straus, M. A., & Gelles, R. J. (1995). How violent are American families? Estimates from the
National Family Violence Resurvey and other studies. In M. A. Straus & R. J. Gelles
(Eds.), Physical violence in American families: Risk factors and adaptations to violence in
8,145 families (pp. 95-112). New Brunswick, NJ: Transaction.
Watson, J. M., Cascardi, M., Avery-Leaf, S., & O’Leary, K. D. (2001). High school students’
responses to dating aggression. Violence and Victims, 16(3), 339-348.
Whitaker, D. J., Baker, C. K., & Arias, I. (2007). Interventions to prevent intimate partner
violence. In L. S. Doll, S. E. Bonzo, J. A. Mercy, D. A. Sleet, & E. N. Haas (Eds.),
Handbook of injury and violence prevention (pp. 203-221). New York: Springer.
Woffordt, S., Mihalic, D. E., & Menard, S. (1994). Continuities in marital violence. Journal
of Family Violence, 9 (3), 195-225.

Daniel J. Whitaker received his doctorate in social psychology from the University of
Georgia in 1996. He worked as a behavioral scientist and team leader at the Division of HIV/
AIDS Prevention and the Division of Violence Prevention at the Centers for Disease Control
and Prevention from 1997 to 2007. In 2008, he became director of the National SafeCare®
Training and Research Center and a visiting professor of public health at Georgia State
University. His research interests focus on the areas of child maltreatment and intimate partner
violence prevention.

Brenda Le received her master of science in public health in biostatistics at the Tulane School
of Public Health and Tropical Medicine, and she works as a health scientist at the Centers for
Disease Control and Prevention’s Division of Violence Prevention.

Phyllis Holditch Niolon received her doctorate in community psychology at Georgia State
University. She is a behavioral scientist in the Prevention Development and Evaluation Branch
of the Division of Violence Prevention at the Centers for Disease Control and Prevention. Her
work focuses on the prevention of intimate partner violence and adolescent dating violence.

For reprints and permissions queries, please visit SAGE’s Web site at http://www.sagepub.com/
journalsPermissions.nav

View publication stats