258 JOURNAL OF STUDIES ON ALCOHOL AND DRUGS / MARCH 2018

Pre-Assault Personality Predicts the Nature of Adverse

Outcomes Among Sexual Assault Victims
JESSICA L. COMBS, PH.D.,a ELIZABETH N. RILEY, M.S.,b,* SARAH J. PETERSON, M.S.,b CAROL E. JORDAN, M.S.,c &
GREGORY T. SMITH, PH.D.b,*
aMichael E. DeBakey Veterans Affairs Medical Center, Houston, Texas
bDepartment of Psychology, University of Kentucky, Lexington, Kentucky
cOffice for Policy Studies on Violence Against Women, University of Kentucky, Lexington, Kentucky

ABSTRACT. Objective: Exposure to sexual assault results in ongoing year. Results: A premorbid personality disposition to act impulsively
harms for women. After an assault, some women engage in higher levels when distressed (negative urgency) interacted positively with sexual
of externalizing behaviors, such as problem drinking, and others experi- assault experience to predict subsequent increases in drinking behavior;
ence higher levels of internalizing dysfunction, such as symptoms of a premorbid personality disposition toward internalizing dysfunction
anxiety and depression. We sought to understand the role of premorbid positively interacted with sexual assault experience to predict increased
factors on the different post-assault experiences of women. Method: symptoms of anxiety and depression. Conclusions: Women with dif-
We studied 1,929 women prospectively during a period of high risk ferent personalities tend to experience different forms of post-assault
for sexual assault (the first year of college): women were assessed in consequences. (J. Stud. Alcohol Drugs, 79, 258–268, 2018)
July before arriving at college and in April near the end of the school

S EXUAL ASSAULT, sexual acts that are perpetrated

against a person’s will or when a person is incapable
of giving consent (U.S. Department of Education, 2011), is
The highest estimates of sexual assault are found in
women ages 16–24 years, who are four times as likely to
be sexually assaulted as any other age group (Danielson
known to have a significant impact on the mental health of & Holmes, 2004). Major maladaptive outcomes of sexual
assaulted women. One important and, as of yet, unexplained assault are drinking problems, clinical anxiety, and clinical
phenomenon is that some women respond to the trauma of depression. These outcomes can be understood to reflect ex-
sexual assault with symptoms of externalizing disorders, ternalizing dysfunction (drinking problems) and internalizing
most notably heavy alcohol consumption, whereas others re- dysfunction (clinical anxiety, clinical depression; Krueger &
spond to trauma with increases in internalizing dysfunction, Markon, 2006).
such as major depression and generalized anxiety disorder
(Kilpatrick et al., 2003; Ullman et al., 2013). It is crucial Outcomes following sexual assault
to be able to identify which women are likely to respond to
sexual trauma in which way, given that (a) heavy alcohol Externalizing dysfunction. Sexual assault is consistently
consumption is associated with many problems and (b) there associated with maladaptive drinking behavior. Reports of
are different empirically supported treatments for different alcohol dependence in women post-assault range from 13%
kinds of dysfunction (Hayes et al., 1999; Linehan, 1993). to 49% (Petrak et al., 1997). College student and other adult
The intent of this study was to test whether pre-assault sexual abuse survivors report more overall drinking after as-
personality characteristics interact with sexual trauma to sault as well as more drinking problems than other women
predict differential outcomes. Findings in support of this (Bedard-Gilligan et al., 2011; Blayney et al., 2016; Griffin
model suggest the value of applying different interven- et al., 2013; Ullman & Najdowski, 2009). Of course the
tions to different assault victims, based on their personality responsibility of engaging in sexual assault rests with the
characteristics. perpetrator, but increased use of substances is of particular
note because such use increases risk for re-victimization
(Littleton & Ullman, 2013; Ullman & Najdowski, 2009). The
Received: October 24, 2016. Revision: August 18, 2017.
This research is supported by National Institute on Alcohol Abuse and
identification of factors that lead to such high-risk behaviors
Alcoholism (NIAAA) Grant RO1AA016166 (to Gregory T. Smith), NIAAA after an assault experience may facilitate more effective care
Grant 1F31AA020767-01A1 (to Jessica L. Combs), and National Institute for assault survivors.
on Drug Abuse Grant DA035200 (to Craig Rush). Internalizing dysfunction. Sexual assault is also highly
*Correspondence may be sent to Elizabeth N. Riley or Gregory T. Smith associated with subsequent anxiety and depressive disorders
at the Department of Psychology, 105 Kastle Hall, University of Kentucky,
Lexington, KY 40506, or via email at: enriley1231@gmail.com or gsmith@
(Breslau, 2002, Frazier et al., 2009). Reportedly, 13%–51%
email.uky.edu. At the time the research was conducted, Jessica L. Combs was of women develop depression after an assault, whereas
with the Department of Psychology, University of Kentucky, Lexington, KY. 73%–82% develop fear and/or anxiety symptoms (Acierno

258
 COMBS ET AL. 259

et al., 2002; Clum et al., 2000). If a post-assault expression victimized. Second, we can test for the modifying effects of
of anxiety is to go out less, stay home, and perhaps have less prior sexual assault. Third, we can control for prior drinking
social interaction, one result is reduced access to positive or internalizing symptoms and thus test whether personality
reinforcement and social support (e.g., Hopko et al., 2003). interacts with sexual assault to predict changes in symptom
Such reduced access can serve to suppress positive affect. In- levels.
dividuals who are experiencing distress and negative affect, We tested whether each of these effects varied as a func-
but reduced positive affect, are likely to experience higher tion of whether women were experiencing their first assault
levels of depressive symptomatology. or re-victimization. There is evidence for negative sequelae
Personality correlates of substance use and anxiety and to both first assaults and repeated assault among adolescent
depression. The fact that only a portion of women who have and adult women (Walsh et al., 2014). In general, multiple
experienced a sexual trauma go on to develop substance traumas exacerbate negative outcomes (Ullman & Naj-
abuse problems, and only a portion experience heightened dowski, 2009). However, those in our sample with a sexual
anxiety and depression symptoms, suggests that there are im- assault history may not be fully representative of women
portant individual differences contributing to risk for those with such a history because we studied only college students.
negative sequelae to sexual trauma. One source of such in- For example, it may be the case that a higher percentage of
dividual differences may be personality. The trait of negative women who enter college were fortunate enough to receive
urgency, the tendency to act rashly when distressed (Cyders support or intervention following the trauma. Therefore, we
& Smith, 2008; Whiteside & Lynam, 2001), predicts sub- had no a priori hypotheses as to whether effects would vary
stance abuse and other externalizing behavioral dysfunction based on the number of assaults experienced.
(Fischer et al., 2013; Settles et al., 2010; Smith & Cyders,
2016). In contrast, personality dispositions to respond to Method
events in anxious or depressive ways (trait anxiety and trait
depression) predict the development of symptoms of clinical Participants
anxiety and clinical depression (Hettema et al., 2006; Settles
et al., 2012). Perhaps elevations in one or the other of these Participants were recruited the summer before their fresh-
traits increase the likelihood of externalizing or internalizing man year of college to participate in a longitudinal study. We
behavioral responses to trauma. limited the participant sample to women entering college
within 3 years of graduating high school. The sample par-
Current study ticipated at two different time points: the month before the
school year began (July) and near the end of the freshman
To assess personality before sexual assault exposure, we year (April). The mean age of participants at the initiation of
followed women across their first year of college, a period in the study was 18.04 years. Most participants were European
which approximately 17% of women experience some form American (88.1%) and identified as heterosexual/straight
of sexual assault (Wilcox et al., 2007). We assessed 1,929 (96.7%).
women before they began their first year of college and
again near the end of that first year. This sample included Procedure
both women who were first assaulted during the first year of
college and women who were assaulted before college and Time 1. The Time 1 assessment took place in July before
experienced re-victimization over the course of the year. the participants’ first day of move-in. We sent an e-mail to all
We tested two hypotheses with regard to interactions incoming female freshmen with instructions on accessing the
between personality and the experience of sexual assault: online survey system. Eligibility was determined by ques-
(a) Negative urgency would interact with sexual assault to tions regarding gender, nature of enrollment, and English-
predict increased alcohol consumption across the first year, speaking ability. On completion, participants were entered
above and beyond the effect of initial alcohol use. In con- in a raffle to win one of eight $250 gift cards to Target.
trast, (b) trait depression and anxiety would interact with Time 2. The Time 2 assessment took place in late April
assault victimization to predict increases in symptoms of of the participants’ freshman year; participants were paid
clinical depression and anxiety across the first year of col- $10 for their participation. Participants completed the same
lege. The interaction between negative urgency and sexual group of measures with modifications to assess exposure
assault would not predict increased internalizing dysfunction. to sexual assault, problem drinking, and symptoms of
A prior cross-sectional study (Combs et al., 2014) found anxiety/depression for the period of the academic year. All
the hypothesized associations to be present at one point in participants received information about various ways to
time. There are advantages to testing these associations using receive help from community or university clinics; those
a longitudinal design. First, we can measure personality be- who disclosed a history of sexual assault received additional
fore the experience of sexual assault for those not previously reminders about community resources.
260 JOURNAL OF STUDIES ON ALCOHOL AND DRUGS / MARCH 2018

Measures measure that consists of 21 items used to assess depressive

symptoms. In this sample, the reliability of the scale was .93
Demographic information. The participants filled out a at Time 1 and .98 at Time 2.
demographic questionnaire providing information on esti- Beck Anxiety Inventory. The Beck Anxiety Inventory
mated household income, age, ethnicity, parents’ education, (BAI; Beck & Steer, 1990) is a 21-item measure of different
and sexual orientation. symptoms of anxiety. In this sample, the reliability of the
Sexual Experiences Survey. The Sexual Experiences scale was .94 at Time 1 and .97 at Time 2.
Survey (SES; Koss & Oros, 1982) is a 14-item measure of UPPS-P Impulsivity Scale: Negative Urgency Subscale.
different dimensions of sexual assault and rape. The experi- The UPPS-P Impulsivity Scale: Negative Urgency Subscale
ences asked about a range of sexual assault experiences, (Lynam et al., 2006) (12 Likert-type items) was administered
from unwanted touching to rape with a foreign object, and at Time 1 (α = .89). Negative urgency can be understood as
the questions asked about the participant’s age at which the the tendency to perform rash acts while in a negative mood.
experience occurred and the number of times the experience Revised NEO Personality Inventory, Trait Depression and
occurred. For Time 1, the SES asked about any incidents Trait Anxiety Scales. Trait depression and trait anxiety were
that occurred between age 14 years and the current age of measured at Time 1 with eight items each from the Revised
the participant. Time 2 asked if any assaults had occurred NEO Personality Inventory, Trait Depression and Trait Anxi-
between Time 1 (July) and the present (April); it assessed in ety Scales (NEO-PI-R; Costa & McCrae, 1992) and are part
which specific month the assault occurred. The SES is con- of the Neuroticism domain of personality (α = .83 and .75,
sistent with verbal reports of victimization at r = .73 (Koss & respectively, and when combined, α = .86.). Because the two
Gidycz, 1985). A revised version of the SES was published scales were so highly correlated (r = .62), and because when
in 2007 (Koss et al., 2007), but we used the original version treated individually the two scales produced the same results
because, to date, it rests on a more extensive body of validity in all of the analyses described below, we combined them.
evidence. We did modify the SES by adjusting the wording
of some items to reduce the level of responsibility placed on Data analysis
the potential victims; for example, the phrase “sex play” was
removed in favor of strict behavioral descriptions (“fondling, Model tests. We tested the central hypotheses using re-
touching, petting”), and questions referring to intoxication gression models in IBM SPSS Statistics for Windows, Ver-
were edited to reflect the possibility that the respondent was sion 22 (IBM Corp., Armonk, NY). The predictor variables
intoxicated before initially encountering an assailant (rather were as follows: Step 1: personality (negative urgency and
than mandating that the assailant provided alcohol/other trait anxiety/depression measured at Time 1; both variables
drugs to the respondent in order to qualify as assault). These centered); sexual assault occurring between August and
changes are consistent with revisions made to the SES. March of freshman year (SA); sexual assault history before
There is good evidence for the convergent validity of college (SAHx); and report of each outcome variable in the
different scoring methods of the SES (Davis et al., 2014). month before college began (QF in July and symptoms of
Because the current sample was a nonselected sample with clinical anxiety/depression in July). Step 2: interaction terms
a low base rate of assault experiences, we scored the total for SA × negative urgency, SA × trait anxiety/depression,
number of sexual assaults experienced by the participant SA × SAHx, SAHx × negative urgency, and SAHx × trait
across the first year of college. As we describe below, these anxiety/depression. Step 3: three-way interaction terms for
data were highly positively skewed. We describe analytic negative urgency × SA × SAHx and trait anxiety/depression
decisions made based on the skewness of these results later × SA × SAHx. We probed interactions using computational
in this section. tools provided by Preacher et al. (2006). We present interac-
Drinking Styles Questionnaire. We constructed a standard tions by plotting one standard deviation above and below
quantity/frequency (QF) index by summing the responses the mean for centered interval scale and using 0 or 1 for
to two items in the Drinking Styles Questionnaire (Smith dichotomous variables.
et al., 1995) that were responded to in relation to drinking Treatment of missing data. There was no significant ef-
frequency and quantity during the current month. There is fect for missingness: retained and nonretained participants
considerable evidence for the internal consistency, stability did not differ on any study variables. Because listwise and
over time, and validity of this measure (cf. Settles et al., pairwise deletion of missing data produces biased popula-
2010). Coefficient alpha in this sample was .87 at Time 1 tion parameter estimates (Allison, 2003; Enders & Peugh,
and .90 at Time 2. We measured QF for the current month 2004), we used expectation maximization (EM), a maximum
in July, before the start of freshman year, and in April, at the likelihood process, to impute values for missing cases us-
end of freshman year. ing SPSS. This method is a single imputation procedure.
Beck Depression Inventory-II. The Beck Depression Although single imputation has historically produced under-
Inventory-II (BDI-II; Beck et al., 1996) is a self-report estimates of standard errors, the SPSS procedure includes a
 COMBS ET AL. 261

TABLE 1. Descriptive statistics of reported sexual assault data across fresh- college. Of those, 31.8% reported experiencing an unwanted
man year (August through April)
sexual event during the school year, which is a higher fre-
Reported no history Reported history of quency than for those with no assault history, χ2(1) = 80.79,
Sexual assault across of assault before assault before
freshman year college (n = 1,376) college (n = 553)
p < .001. Descriptive statistics of outcome variables and
bivariate correlations between key study variables can be
Reported no sexual assault 86.0% (n = 1,183) 68.2% (n = 377)
Unwanted touching 12.2% (n = 168) 29.1% (n = 161) found in Table 2 and Table 3, respectively.
Attempted sex 3.5% (n = 48) 7.8% (n = 43) Given the high levels of skew and kurtosis in the
Pressured/coerced sex 2.5% (n = 34) 9.8% (n = 54) measurement of sexual assault, we present our main hy-
Forced rape 2.5% (n = 35) 6.7% (n = 38)
pothesis test results using sexual assault experiences as a
Notes: These numbers and percentages represent the number of people dichotomous variable: a 1 was assigned to any instance of
who experienced this type of sexual assault at all during their first year, as
a dichotomous variable. Individuals who experienced more than one type of unwanted touching, coerced/forced attempted intercourse,
sexual assault are represented in multiple categories; thus, the percentages and coerced/forced intercourse and a 0 was assigned when
add up to more than 100%. no such instances were reported. We analyzed the sexual
assault data in two additional ways. First, we coded scores
correction to the standard error to avoid/reduce this problem. into three categories reflecting the number of sexual assault
Use of EM procedures has been shown to produce parameter experiences: 0 = no reported assaults (n = 1,560, 80.9%), 1
estimates accurate to within two decimal points and of equal = one reported sexual assault experience (n = 133, 6.9%),
accuracy to full-information maximum likelihood methods and 2 = more than one reported sexual assault in the first
(Enders & Peugh, 2004). year of college (n = 236, 12.2% of the sample). We found
no statistically significant differences between having ex-
Results perienced one assault or multiple assaults with regard to
our main hypothesis tests. Second, we ran the same models
Retention, descriptive statistics, and measurement of sexual treating the number of sexual assaults as an interval scale.
assault Whether we then transformed the data to reduce skew (we
ran square root and fourth root transformations) or analyzed
The retention rate was 63% from Time 1 to Time 2, which the skewed data without transformation, we found virtually
is comparable to most longitudinal studies of sexual assault the same effects as we found when treating assault experi-
(average rate = 70%; Campbell et al., 2011). Descriptive sta- ence dichotomously.
tistics for sexual assault broken down by those who had been
previously assaulted and those who had not are displayed Hypothesis 1
in Table 1. Of those women who reported that they had not
experienced an assault before college, 14.0% reported ex- The first hypothesis was that the effect of negative ur-
periencing an unwanted sexual event during the school year gency on drinking QF will be stronger among those having
(August through March). A total of 553 women (28.67% of experienced sexual assault. This hypothesis was supported:
the sample) reported having been sexually assaulted before The significant, positive main effects for negative urgency

TABLE 2. Descriptive statistics of outcome variables (drinking quantity/frequency and clinical anxiety/depression symptom count)
T1 drink T2 drink T1 T2
Variable QF QF BDI/BAI BDI/BAI T1 NU T1 NEO
Reported no history of assault
before college (n = 1,376)
Mean 2.71 3.91 9.22 6.44 2.01 2.74
Variance 3.86 4.75 137.23 122.27 0.24 0.33
Range 0–8 0–9 0–92 0–94 1–3.84 1–4.81
Reported history of assault
before college (n = 553)
Mean 3.89 4.31 19.19 10.39 2.33 2.98
Variance 4.82 4.06 387.29 285.85 0.37 0.40
Range 0–8 0–9 0–101 0–116 1–4 1–4.75
Notes: N = 1,929. T1 drink QF = drinking quantity frequency, Time 1, July; T2 drink QF = drinking quantity frequency, T2, April; T1
BDI/BAI = composite score of results on Beck Anxiety Inventory (BAI) and Beck Depression Inventory (BDI), T1, July; T2 BDI/BAI
= composite score of results on BAI and BDI, T2, April; NU = negative urgency levels; NEO = scores on the trait anxiety/depression
measure. The range for drinking frequency and quantity represents a sum total of participant responses to questions about their drinking
habits. A 9 represents the highest level of drinking seen in this population; this person would report drinking a lot of alcohol almost
daily. The range for clinical anxiety/depression represents the sum total of responses on measure of anxiety and depression. A person
who received a 116 likely rated all anxiety and depression symptoms as being severe. Negative urgency and trait anxiety/depression are
both moderators, and are thus reported at Time 1 only.
262 JOURNAL OF STUDIES ON ALCOHOL AND DRUGS / MARCH 2018

TABLE 3. Correlations between key study variables

T1 T2 T1 T2 SA
T1 NU T1 NEO drink QF drink QF BDI/BAI BDI/BAI history
T1 NU
T1 NEO .52**
T1 drink QF .23** .00
T2 drink QF .13** -.04 .46**
T1 BDI/BAI .43** .58** .06** -.04
T2 BDI/BAI .20** .33** .01 -.04 .42**
SA history .27** .18** .25** .09** .30** .17**
SA college .09** .09** .14** .14** .12** .17** .21**
Notes: T1 drink QF = drinking quantity frequency, Time 1, July; T2 drink QF = drinking quantity frequency, T2, April; T1
BDI/BAI = composite score of results on Beck Anxiety Inventory (BAI) and Beck Depression Inventory (BDI), T1, July;
T2 BDI/BAI = composite score of results on BAI and BDI, T2, April; NU = negative urgency levels; NEO = scores on the
trait anxiety/depression measure; SA history = reported sexual assault before college; SA college = reported sexual assault
in college. Negative urgency and trait anxiety/depression are both moderators, and are thus reported at Time 1 only.
**p < .01.

and assault experience in predicting increased drinking QF women in each quadrant of the interaction is provided in the
were modified by an interaction between the two (Table 4). Figure 2 caption.
As shown in Figure 1, for those women who were assaulted None of the effects were modified by whether women had
during the first year of college, the association between nega- experienced an assault before college. We therefore have no
tive urgency and drinking QF was significantly greater than basis for concluding that the joint effects of sexual assault
zero, slope = .86, t(358) = 3.37, p < .001. The slope relating and the traits studied operate differently for first assault suf-
negative urgency to drinking QF for women not assaulted ferers and those experiencing re-victimization.
was nonsignificant, slope = .23, t(1549) = 1.81, p = .07. The
women who reported the highest levels of drinking were Hypothesis 2
those who had been assaulted and had high levels of nega-
tive urgency. The number of women in each quadrant of the The second hypothesis was that trait depression and
interaction is provided in the Figure 1 caption. anxiety would interact with assault victimization to predict
Unexpectedly, when we controlled for the direct effects increases in symptoms of clinical depression and anxiety
and the Negative Urgency × Sexual Assault interaction, there across the first year of college. As shown in Table 5 and
was a crossover interaction between trait anxiety/depression Figure 3, this hypothesis was supported. For women who
and assault experience in predicting drinking QF (Table 4). were assaulted sexually during freshman year, trait anxiety/
As shown in Figure 2, among those women who were as- depression significantly predicted increases in symptoms of
saulted freshman year, elevations in the trait predicted lower clinical anxiety/depression, slope = 5.74, t(358) = 3.25, p <
levels of drinking behavior than for women low on the trait, .001, and more strongly than for women who were not as-
simple effects slope = -4.71, t(358) = -21.74, p < .001. For saulted, for whom the relationship was still positive, slope =
women who were not assaulted, there was no evidence of 1.85, t(1549) = 2.60, p < .01. The women who reported the
a relationship between trait levels and drinking QF, simple most anxiety/depression symptoms were those who had been
effects slope = -.09, t(1549) = 0.90, p = .37. The number of assaulted and had high levels of trait anxiety/depression.

TABLE 4. Linear regression results for prediction of drinking QF scores at Time 2

Variable B SE B t Significance [95% CI] R2
Final step .23
T1 NU 0.23 0.13 1.79 .074 [-0.02, 0.48]
T1 NEO -0.09 0.11 -0.87 .387 [-0.31, 0.12]
SA history -0.13 0.12 -1.10 .271 [-0.37, 0.10]
SA college 0.56 0.15 3.79 .000 [0.27, 0.85]
T1 drink QF 0.45 0.02 20.57 .000 [0.41, 0.49]
SA college × T1 NU 0.63 0.25 2.54 .011 [0.14, 1.12]
SA college × T1 NEO -0.46 0.22 -2.10 .035 [-0.89, -0.03]
SA college × SA history -0.17 0.23 -0.72 .474 [-0.63, 0.29]
SA history × T1 NU -0.20 0.20 -0.97 .333 [-0.60, 0.20]
SA history × T1 NEO -0.30 0.18 -1.64 .101 [-0.66, 0.06]
Notes: Bold indicates predictors significant at the .05 level. CI = confidence interval; T1 = Time 1 (July); NU = negative
urgency levels; NEO = scores on the trait anxiety/depression measure; SA history = reported sexual assault before college;
SA college = reported sexual assault in college; drink QF = composite score of drinking quantity and drinking frequency.
Step 2, the final step of the regression model, is presented.
 COMBS ET AL. 263

FIGURE 1. Prediction of changes in drinking quantity and frequency (QF) by the interaction of negative urgency
and freshman year sexual assault experience. Negative urgency scores are plotted 1 SD above and below the
mean, and sexual assault experience is coded dichotomously. For women who were sexually assaulted, the mean
drinking QF increase from low to high levels of negative urgency was 1 point on the 9-point scale. Examples of
a 1-point increase would be increasing drinking frequency from monthly to once or twice a week, or increasing
drinking quantity from 1 beer or 1 drink or less to between 2 and 3 beers or drinks. For women who were not
sexually assaulted, the change was very small. Participant frequency by quadrant is as follows: above the mean
on negative urgency/positive for sexual assault n = 179; above the mean on negative urgency/negative for sexual
assault n = 599; below the mean on negative urgency/positive for sexual assault n = 190; below the mean on
negative urgency/negative for sexual assault n = 961.

FIGURE 2. Prediction of changes in drinking quantity and frequency (QF) by the interaction of trait anxiety/
depression and freshman year sexual assault experience. Trait anxiety depression scores are plotted 1 SD above
and below the mean, and sexual assault experience is coded dichotomously. For women who were sexually as-
saulted, the mean drinking QF drop was 5.7 points. An example of this magnitude of change would be shifting
from drinking between 4 and 8 beers or drinks to not drinking at all. For women who were not sexually assaulted,
there was no change. Participant frequency by quadrant is as follows: above the mean on trait anxiety/depression/
positive for sexual assault n = 211; above the mean on trait anxiety/depression/negative for sexual assault n = 706;
below the mean on trait anxiety/depression/positive for sexual assault n = 158; below the mean on trait anxiety/
depression/negative for sexual assault n = 854.

See the Figure 3 caption for the number of women in each anxiety/depression symptoms. We did, however, observe
quadrant of the interaction. a three-way interaction among negative urgency, sexual
We had anticipated no association between negative assault experience during freshman year, and pre-college
urgency and assault experience in predicting increases in assault history (Table 5). As shown in Figure 4, for women
264 JOURNAL OF STUDIES ON ALCOHOL AND DRUGS / MARCH 2018

TABLE 5. Linear regression results for BDI/BAI scores (anxiety/depression) at Time 2

Variable B SE B t Significance [95% CI] R2
Final step .22
T1 NU 0.71 0.81 0.88 .382 [-0.88, 2.30]
T1 NEO 1.85 0.72 2.57 .010 [0.44, 3.27]
SA history 0.45 0.74 0.61 .545 [-1.00, 1.90]
SA college 3.36 0.91 3.70 .000 [1.58, 5.14]
T1 BDI/ BAI 0.28 0.02 12.16 .000 [0.23, 0.33]
SA college × T1 NU -2.96 2.21 -1.34 .180 [-7.30, 1.37]
SA college × T1 NEO 3.89 1.91 2.04 .042 [0.15, 7.64]
SA college × SA history -1.85 1.47 -1.25 .210 [-4.74, 1.04]
SA history × T1 NEO -4.84 1.42 -3.42 .001 [-7.61, -2.06]
SA history × T1 NEO 0.99 1.29 0.77 .445 [-1.54, 3.52]
SA college × T1 NU × SA history 10.67 3.07 3.48 .001 [4.66, 16.68]
SA college × T1 NEO × SA history 1.21 2.71 0.45 .656 [-4.11, 6.53]
Notes: Bold indicates predictors significant at the .05 level. CI = confidence interval; T1 = Time 1 (July); NU = negative urgency
levels; NEO = scores on the trait anxiety/depression measure; SA history = reported sexual assault before college, SA college =
reported sexual assault in college; BDI/BAI = composite score of results on Beck Anxiety Inventory (BAI) and Beck Depression
Inventory (BDI). Step 3, the final step of the regression model, is presented.

with a pre-college history of sexual assault, after we con- three-way interaction among pre-college assault, college
trolled for main effects and the interaction of trait anxi- assault, and trait anxiety/depression in analyses for either
ety/depression, elevations in negative urgency predicted hypothesis.
increased anxiety/depression symptoms for women who
were assaulted again freshman year, slope = 3.58, t(162) Discussion
= 2.03, p < .05, and no significant change for those who
were not, slope = -2.25, t(162) = 1.11, p = .27. For women The findings of this study provide clear support for the
with no pre-college assault history, there was no interac- hypothesis that variations in premorbid personality interact
tion between negative urgency and college assault in pre- with the experience of sexual assault to predict variations in
dicting anxiety/depression symptoms. This unanticipated the nature of post-assault experience. Specifically, the trait
finding should be replicated before theoretical and clinical of negative urgency interacted with sexual assault to predict
inferences are drawn. Last, we found no evidence of a increases in one marker of externalizing behavior, which was

FIGURE 3. Prediction of changes in anxiety/depression symptoms by the interaction of trait anxiety/depression

and freshman year sexual assault experience. Trait anxiety depression scores are plotted 1 SD above and below
the mean, and sexual assault experience is coded dichotomously. For women who were sexually assaulted, the
mean increase in symptoms endorsed is 6.4. An example of a change of this magnitude would be increasing the
frequency of between 6 and 7 symptoms from 1–2 days in the preceding week to 3–4 days in the preceding week.
For women who were not assaulted, there was a mean increase in symptoms of 2.4. Participant frequency by
quadrant is as follows: above the mean on trait anxiety/depression/positive for sexual assault n = 211; above the
mean on trait anxiety/depression/negative for sexual assault n = 706; below the mean on trait anxiety/depression/
positive for sexual assault n = 158; below the mean on trait anxiety/depression/negative for sexual assault n = 854.
 COMBS ET AL. 265

FIGURE 4. Depiction of three-way interaction between negative urgency, freshman year sexual assault experience,
and pre-college sexual assault history. The top panel depicts results for women positive for pre-college sexual as-
sault history who were or were not assaulted freshman year. The bottom panel depicts results for women negative
for pre-college sexual assault history who were or were not assaulted freshman year.

drinking QF. Sexual assault has been shown to predict subse- clinical anxiety and depression, which is a marker of inter-
quent increases in substance use (Ullman et al., 2013). This nalizing behavior. Sexual assault has been shown to predict
effect was significantly stronger for women with elevated subsequent increases in anxiety or depressive symptoms
levels of negative urgency. Among women who are sexually (Acierno et al., 2002; Clum et al., 2000). This effect was sig-
assaulted, those high in negative urgency are most likely to nificantly stronger for women with elevations in trait anxiety/
increase their drinking behavior. This effect operated whether depression. As was true for the Negative Urgency × Sexual
women were experiencing their first sexual assault or were Assault interaction, this effect operated whether women were
re-victimized. experiencing their first sexual assault or were re-victimized.
In contrast, trait anxiety/depression (the disposition to Together, these findings may help explain why some
respond to events in anxious or depressive ways) interacted women experience increases in externalizing behaviors and
with sexual assault to predict increases in symptoms of others in internalizing behaviors following sexual assault.
266 JOURNAL OF STUDIES ON ALCOHOL AND DRUGS / MARCH 2018

Pre-assault personality characteristics that are predictors self-report questionnaire using a web-based format; there
of externalizing (negative urgency) or internalizing (trait was thus no opportunity to clarify questions or responses.
anxiety/depression) dysfunction may help shape the form However, confidential self-report is likely the most effective
of women’s specific reactions to the trauma. We qualify this method for obtaining data of this sensitive nature because
conclusion in two ways. First, given the many intrapersonal, women may be more likely to under-report sexual assaults
interpersonal, and contextual factors that influence drinking in person (Ongena & Dijkstra, 2007), and questionnaire
behavior and anxiety/depressive symptoms, the impact of data are often more reliable than interview data (Testa et al.,
personality in shaping responses to sexual assault is mod- 2005). Fifth, we were unable to examine the role of child-
est. Second, an important future direction is to investigate hood sexual abuse history, which is surely important to this
whether and how internalizing and externalizing experiences process (Classen et al., 2005). Finally, the sample is made
transact to influence the risk process. up of predominantly White college women, which limits
An unanticipated finding of potential interest was that generalizability.
high levels of trait anxiety/depression in women who report- The finding that different pre-assault personality charac-
ed previous assault interacted with sexual assault experience teristics predict different patterns of post-assault distress is
to predict lower levels of drinking. We highlight two related novel and important. It can contribute to etiological models
possible explanations for this finding. First, the predictive of post-assault distress and lead to more effective, person-
influence of trait anxiety/depression was above and beyond
specific interventions for assaulted women.
that of negative urgency; perhaps elevations in that part of
trait anxiety/depression that is unrelated to negative urgency
References
capture only that part of distress that is unassociated with
risk taking or acting out. Related to this, perhaps women Acierno, R., Brady, K., Gray, M., Kilpatrick, D. G., Resnick, H., & Best, C.
with such elevations were more likely to withdraw from L. (2002). Psychopathology following interpersonal violence: A com-
social occasions and were thus less likely to be in situations parison of risk factors in older and younger adults. Journal of Clinical
where drinking is occurring. Geropsychology, 8, 13–23. doi:10.1023/A:1013041907018
We also observed an unanticipated three-way interac- Allison, P. D. (2003). Missing data techniques for structural equa-
tion modeling. Journal of Abnormal Psychology, 112, 545–557.
tion, which suggested that the interaction between negative
doi:10.1037/0021-843X.112.4.545
urgency and freshman year sexual assault actually predicts Angelo, F. N., Miller, H. E., Zoellner, L. A., & Feeny, N. C. (2008). “I need
increases in symptoms of anxiety/depression for those to talk about it”: A qualitative analysis of trauma-exposed women’s rea-
women who were re-victimized. We consider it best to with- sons for treatment choice. Behavior Therapy, 39, 13–21. doi:10.1016/j.
hold speculation as to the meaning of this finding, pending beth.2007.02.002
its replication. Beck, A. T., & Steer, R. A. (1990). Beck Anxiety Inventory manual. San An-
tonio, TX: The Psychological Corporation, Harcourt Brace & Company.
There are empirically supported interventions that re-
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). BDI-II manual (2nd ed.).
duce an individual’s risk for engaging in distress-driven San Antonio, TX: Harcourt Brace & Company.
rash action (Linehan, 1993) and their risk for engaging in Bedard-Gilligan, M., Kaysen, D., Desai, S., & Lee, C. M. (2011).
substance abuse (Marlatt et al., 1998). There are other in- Alcohol-involved assault: Associations with posttrauma alcohol use,
terventions most effective for internalizing expressions of consequences, and expectancies. Addictive Behaviors, 36, 1076–1082.
distress (Chambless & Ollendick, 2001, Hayes et al., 1999). doi:10.1016/j.addbeh.2011.07.001
Blayney, J. A., Read, J. P., & Colder, C. (2016). Role of alcohol in college
A clearer understanding of the differences between the two
sexual victimization and postassault adaptation. Psychological Trauma:
types of responses to trauma can facilitate the choice of ef- Theory, Research, Practice, and Policy, 8, 421–430. doi:10.1037/
fective treatment, and developing an empirically supported tra0000100
combination of personality-driven treatments with validated Breslau, N. (2002). Epidemiologic studies of trauma, posttraumatic stress
treatments aimed to reduce the risk of sexual assault re-vic- disorder, and other psychiatric disorders. Canadian Journal of Psychia-
timization and alcohol use could improve treatment success try, 47, 923–929. doi:10.1177/070674370204701003
(Angelo et al., 2008; Gilmore et al., 2015; Orchowski et al., Campbell, R., Sprague, H. B., Cottrill, S., & Sullivan, C. M. (2011).
Longitudinal research with sexual assault survivors: A method-
2008). ological review. Journal of Interpersonal Violence, 26, 433–461.
It is important to note the limitations of this study. One is doi:10.1177/0886260510363424
the rate of retention, although missing participants did not Chambless, D. L., & Ollendick, T. H. (2001). Empirically supported psy-
vary from retained participants on study variables, and reten- chological interventions: Controversies and evidence. Annual Review of
tion rates for longitudinal sexual assault studies tend to be Psychology, 52, 685–716. doi:10.1146/annurev.psych.52.1.685
lower than others (Campbell et al., 2011). Second, the lon- Classen, C. C., Palesh, O. G., & Aggarwal, R. (2005). Sexual revictimiza-
tion: A review of the empirical literature. Trauma, Violence & Abuse, 6,
gitudinal period of 1 academic year may not have been long
103–129. doi:10.1177/1524838005275087
enough to see more maladaptive outcomes develop. Third, Clum, G. A., Calhoun, K. S., & Kimerling, R. (2000). Associations
we were unable to examine the effects of the characteristic among symptoms of depression and posttraumatic stress disor-
of the assault experience. Fourth, all data were collected by der and self-reported health in sexually assaulted women. Jour-
 COMBS ET AL. 267

nal of Nervous and Mental Disease, 188, 671–678. doi:10.1097/ tion. Journal of Consulting and Clinical Psychology, 50, 455–457.
00005053-200010000-00005 doi:10.1037/0022-006X.50.3.455
Combs, J. L., Jordan, C. E., & Smith, G. T. (2014). Individual differences in Krueger, R. F., & Markon, K. E. (2006). Reinterpreting comorbidity: A
personality predict externalizing versus internalizing outcomes follow- model-based approach to understanding and classifying psychopathol-
ing sexual assault. Psychological Trauma: Theory, Research, Practice, ogy. Annual Review of Clinical Psychology, 2, 111–133. doi:10.1146/
and Policy, 6, 375–383. doi:10.1037/a0032978 annurev.clinpsy.2.022305.095213
Costa, P. T., & McCrae, R. R. (1992). NEO PI-R. Professional manual. Linehan, M. M. (1993). Cognitive behavioral treatment of borderline per-
Odessa, FL: Psychological Assessment Resources. sonality disorder. New York, NY: Guilford Press.
Cyders, M. A., & Smith, G. T. (2008). Emotion-based dispositions to rash Littleton, H., & Ullman, S. E. (2013). PTSD symptomatology and hazard-
action: Positive and negative urgency. Psychological Bulletin, 134, ous drinking as risk factors for sexual assault revictimization: Examina-
807–828. doi:10.1037/a0013341 tion in European American and African American women. Journal of
Danielson, C. K., & Holmes, M. M. (2004). Adolescent sexual assault: An Traumatic Stress, 26, 345–353. doi:10.1002/jts.21807
update of the literature. Current Opinion in Obstetrics & Gynecology, Lynam, D. R., Smith, G. T., Whiteside, S. P., & Cyders, M. A. (2006). The
16, 383–388. doi:10.1097/00001703-200410000-00005 UPPS-P: Assessing five personality pathways to impulsive behavior
Davis, K. C., Gilmore, A. K., Stappenbeck, C. A., Balsan, M. J., George, W. (Technical Report). West Lafayette, IN: Purdue University.
H., & Norris, J. (2014). How to score the Sexual Experiences Survey? Marlatt, G. A., Baer, J. S., Kivlahan, D. R., Dimeff, L. A., Larimer, M. E.,
A comparison of nine methods. Psychology of Violence, 4, 445–461. Quigley, L. A., . . . Williams, E. (1998). Screening and brief intervention
doi:10.1037/a0037494 for high-risk college student drinkers: Results from a 2-year follow-
Enders, C. K., & Peugh, J. L. (2004). Using an EM covariance matrix to up assessment. Journal of Consulting and Clinical Psychology, 66,
estimate structural equation models with missing data: Choosing an 604–615. doi:10.1037/0022-006X.66.4.604
adjusted sample size to improve the accuracy of inferences. Structural Ongena, Y. P., & Dijkstra, W. (2007). A model of cognitive processes and
Equation Modeling, 11, 1–19. doi:10.1207/S15328007SEM1101_1 conversational principles in survey interview interaction. Applied Cogni-
Fischer, S., Peterson, C. M., & McCarthy, D. (2013). A prospective test of tive Psychology, 21, 145–163. doi:10.1002/acp.1334
the influence of negative urgency and expectancies on binge eating and Orchowski, L. M., Gidycz, C. A., & Raffle, H. (2008). Evaluation of a
purging. Psychology of Addictive Behaviors, 27, 294–300. doi:10.1037/ sexual assault risk reduction and self-defense program: A prospective
a0029323 analysis of a revised protocol. Psychology of Women Quarterly, 32,
Frazier, P., Anders, S., Perera, S., Tomich, P., Tennen, H., Park, C., & 204–218. doi:10.1111/j.1471-6402.2008.00425.x
Tashiro, T. (2009). Traumatic events among undergraduate students: Petrak, J., Doyle, A., Williams, L., Buchan, L., & Forster, G. (1997). The
Prevalence and associated symptoms. Journal of Counseling Psychol- psychological impact of sexual assault: A study of female attenders of
ogy, 56, 450–460. doi:10.1037/a0016412 a sexual health psychology service. Sexual and Marital Therapy, 12,
Gilmore, A. K., Lewis, M. A., & George, W. H. (2015). A randomized con- 339–345. doi:10.1080/02674659708408177
trolled trial targeting alcohol use and sexual assault risk among college Preacher, K. J., Curran, P. J., & Bauer, D. J. (2006). Computational tools for
women at high risk for victimization. Behaviour Research and Therapy, probing interaction effects in multiple linear regression, multilevel mod-
74, 38–49. doi:10.1016/j.brat.2015.08.007 eling, and latent curve analysis. Journal of Educational and Behavioral
Griffin, M. J., Wardell, J. D., & Read, J. P. (2013). Recent sexual victimiza- Statistics, 31, 437–448. doi:10.3102/10769986031004437
tion and drinking behavior in newly matriculated college students: A Settles, R. F., Cyders, M., & Smith, G. T. (2010). Longitudinal validation
latent growth analysis. Psychology of Addictive Behaviors, 27, 966–973. of the acquired preparedness model of drinking risk. Psychology of Ad-
doi:10.1037/a0031831 dictive Behaviors, 24, 198–208. doi:10.1037/a0017631
Hayes, S. C., Strosahl, K. D., & Wilson, K. G. (1999). Acceptance and com- Settles, R. E., Fischer, S., Cyders, M. A., Combs, J. L., Gunn, R. L., &
mitment therapy: An experiential approach to behavior change. New Smith, G. T. (2012). Negative urgency: A personality predictor of exter-
York, NY: Guilford Press. nalizing behavior characterized by neuroticism, low conscientiousness,
Hettema, J. M., Neale, M. C., Myers, J. M., Prescott, C. A., & Kendler, K. and disagreeableness. Journal of Abnormal Psychology, 121, 160–172.
S. (2006). A population-based twin study of the relationship between doi:10.1037/a0024948
neuroticism and internalizing disorders. American Journal of Psychiatry, Smith, G. T., & Cyders, M. A. (2016). Integrating affect and impulsivity:
163, 857–864. doi:10.1176/ajp.2006.163.5.857 The role of positive and negative urgency in substance use risk. Drug
Hopko, D. R., Lejuez, C. W., Ruggiero, K. J., & Eifert, G. H. (2003). Con- and Alcohol Dependence, 163, Supplement 1, S3–S12. doi:10.1016/j.
temporary behavioral activation treatments for depression: Procedures, drugalcdep.2015.08.038
principles, and progress. Clinical Psychology Review, 23, 699–717. Smith, G. T., McCarthy, D. M., & Goldman, M. S. (1995). Self-reported
doi:10.1016/S0272-7358(03)00070-9 drinking and alcohol-related problems among early adolescents: Dimen-
Kilpatrick, D. G., Ruggiero, K. J., Acierno, R., Saunders, B. E., Resnick, sionality and validity over 24 months. Journal of Studies on Alcohol, 56,
H. S., & Best, C. L. (2003). Violence and risk of PTSD, major depres- 383–394. doi:10.15288/jsa.1995.56.383
sion, substance abuse/dependence, and comorbidity: Results from the Testa, M., Livingston, J. A., & VanZile-Tamsen, C. (2005). The impact of
National Survey of Adolescents. Journal of Consulting and Clinical questionnaire administration mode on response rate and reporting of
Psychology, 71, 692–700. doi:10.1037/0022-006X.71.4.692 consensual and nonconsensual sexual behavior. Psychology of Women
Koss, M. P., Abbey, A., Campbell, R., Cook, S., Norris, J., Testa, M., . . . Quarterly, 29, 345–352. doi:10.1111/j.1471-6402.2005.00234.x
White, J. (2007). Revising the SES: A collaborative process to improve Ullman, S. E., & Najdowski, C. J. (2009). Revictimization as a moderator
assessment of sexual aggression and victimization. Psychology of of psychosocial risk factors for problem drinking in female sexual as-
Women Quarterly, 31, 357–370. doi:10.1111/j.1471-6402.2007.00385.x sault survivors. Journal of Studies on Alcohol and Drugs, 70, 41–49.
Koss, M. P., & Gidycz, C. A. (1985). Sexual Experiences Survey: Reli- doi:10.15288/jsad.2009.70.41
ability and validity. Journal of Consulting and Clinical Psychology, 53, U.S. Department of Education (2011). Dear colleague letter: Sexual vio-
422–423. doi:10.1037/0022-006X.53.3.422 lence. Washington, DC: Office of Civil Rights. https://www2.ed.gov/
Koss, M. P., & Oros, C. J. (1982). Sexual Experiences Survey: A re- about/offices/list/ocr/letters/colleague-201104.pdf
search instrument investigating sexual aggression and victimiza- Walsh, K., Resnick, H. S., Danielson, C. K., McCauley, J. L., Saunders,
268 JOURNAL OF STUDIES ON ALCOHOL AND DRUGS / MARCH 2018

B. E., & Kilpatrick, D. G. (2014). Patterns of drug and alcohol use as- ity. Personality and Individual Differences, 30, 669–689. doi:10.1016/
sociated with lifetime sexual revictimization and current posttraumatic S0191-8869(00)00064-7
stress disorder among three national samples of adolescent, college, Wilcox, P., Jordan, C. E., & Pritchard, A. J. (2007). A multidimensional
and household-residing women. Addictive Behaviors, 39, 684–689. examination of campus safety: Victimization, perceptions of dan-
doi:10.1016/j.addbeh.2013.12.006 ger, worry about crime, and precautionary behavior among college
Whiteside, S. P., & Lynam, D. R. (2001). The Five Factor Model and impul- women in the post-Clery era. Crime and Delinquency, 53, 219–254.
sivity: Using a structural model of personality to understand impulsiv- doi:10.1177/0097700405283664