Are trauma memories state-dependent? Intrusive memories following alcohol-involved sexual assault.

Background: Sexual assault (SA) frequently occurs under the influence of alcohol, and is often followed by both drinking and posttraumatic stress symptoms, including intrusive memories. Although many theories attempt to explain the co-occurrence of alcohol use and posttraumatic stress, one possibility not yet considered is that SA memories may be more likely to occur when there is an encoding-retrieval match in alcohol intoxication state. Objective: The aim of this study was to examine the potential for intrusive memories of SA to be state-dependent, such that intrusive memories for alcohol-involved SA may be more likely to occur in the context of subsequent alcohol intoxication. Method: Participants were 100 college women (age range = 18 to 24 years; 73% White/Caucasian, 89% heterosexual) with a history of alcohol-involved SA (67%) or other, non-alcohol-involved SA (33%). Participants completed daily questionnaires for 30 days assessing past-day drinking and intrusion symptoms.
Results: A random-intercept, negative binomial multilevel model revealed that, after controlling for overall frequency of drinking and perceived threat during SA, women with a history of alcohol-involved SA reported more severe intrusion symptoms on drinking days than on non-drinking days. No such difference in intrusions was observed for women who were not intoxicated at the time of the assault. Conclusions: Findings are consistent with the possibility of state-dependent intrusive memories. Additional research is needed to determine whether alcohol intoxication might serve as a discriminative cue preceding intrusive memories of alcohol-involved SA.

1. Introduction

Sexual assault (SA) is a pervasive problem for women, not just because of its high prevalence (19% of women in the United States; Smith et al., 2017), but also because of the mental health consequences that can result, such as posttraumatic stress disorder (PTSD) and substance abuse or dependence (Dworkin, 2018; Dworkin, Menon, Bystrynski, & Allen, 2017; Rhew, Stappenbeck, Bedard-Gilligan, Hughes, & Kaysen, 2017). Alcohol is commonly involved in SA (Testa & Livingston, 2009), with over half of SA incidents involving drinking by the victim and/or perpetrator (Abbey, Zawacki, Buck, Clinton & McAuslan, 2001). Perpetrators may attempt to get an individual drunk in order to facilitate an assault (Zinzow et al., 2010) or target women whom they perceive as vulnerable due to visible intoxication (Davis, Danube, Stappenbeck, Norris, & George, 2015). Acute alcohol intoxication can in turn impair women’s capacity to use effective resistance techniques (Norris et al., 2006). Given the centrality of alcohol in women’s SA experiences, further research is needed to better understand how alcohol intoxication at the time of the assault may impact longer-term mental health outcomes.

A small, but growing literature has begun to examine survivor experiences after alcohol-involved SA. These studies suggest that alcohol-involved SA is related to increased risk for revictimization (relative to no SA; Messman-Moore, Ward, & Zerubavel, 2013) and is associated with greater post-assault drinking than non-alcohol-involved SA (Bedard-Gilligan, Kaysen, Desai, & Lee, 2011). Less is known, however, about distress-related outcomes like PTSD. Of the studies conducted, results have been mixed. Compared to non-alcohol-involved SA, some have found that alcohol-involved SA is associated with less severe PTSD symptoms (Peter-Hagene & Ullman, 2018), including fewer initial intrusive symptoms (Kaysen et al., 2010), whereas others indicate no differences (Aakvaag, Strøm, & Thoresen, 2018; Brown, Testa, & Messman-Moore, 2009; Littleton, Grills-Taquechel, & Axsom, 2009; Zinzow et al., 2010). Over time, however, alcohol-involved assault survivors report more persistent PTSD intrusive symptoms than non-alcohol assault survivors (Kaysen et al., 2010), especially when there is greater intoxication at the time of the SA (Jaffe et al., 2017).

Post-assault distress is often accompanied by heavy drinking, as alcohol use disorders frequently co-occur with PTSD (Kessler, Chiu, Demler, & Walters, 2005). In addition to being consistently positively associated with avoidance and hyperarousal symptoms, alcohol misuse has been linked to greater PTSD-related intrusion symptoms in some, but not all studies (see Debell et al., 2014). Theories that explain this overlap focus on both pre-existing factors and post-trauma reactions as maintaining processes. For example, shared vulnerability models (Breslau, Davis, Andreski, Peterson, & Schultz, 1997) posit that there are common underlying biological, affective, or cognitive processes predisposing individuals to develop both alcohol misuse and PTSD. The high-risk hypothesis (Chilcoat & Breslau, 1998) proposes that individuals who misuse alcohol are more likely to be exposed to trauma, and thus develop PTSD. The self-medication hypothesis (Stewart, Pihl, Conrod, & Dongier, 1998), which is the most widely studied and supported explanation of the overlap, suggests individuals use alcohol after trauma to alleviate distressing symptoms associated with PTSD (Simpson, Stappenbeck, Varra, Moore, & Kaysen, 2012), including intrusive symptoms (Kaysen et al., 2014).

The overlap between alcohol misuse and trauma-related symptoms, including intrusive memories (Johnson, Cottler, O’Leary, & Abdallah, 2010; Taft et al., 2007), can also be viewed in light of alcohol’s well-documented effects on memory. When present at the time of encoding, alcohol impairs memory formation (Mintzer, 2007). However, after intoxicated encoding, recollection is enhanced by intoxication at retrieval (see Eich, 1980; Poling & Cross, 1993). For example, female social drinkers who learned a word list while intoxicated subsequently recalled more words when intoxicated (vs. sober) at retrieval (Weingartner, Adefris, Eich, & Murphy, 1976). In this way, alcohol intoxication has been supported as an internal cue that can facilitate explicit memory retrieval (Duka, Weissenborn, & Dienes, 2001; Weissenborn & Duka, 2000). Termed state-dependent memory, this phenomenon is consistent with broader research demonstrating improved recollection when context (e.g. drug/alcohol state, mood, environment) is matched at encoding and retrieval (Bower & Forgas, 2000; Eich, 1980; Herz, 2004), consistent with Tulving and Thomson's (1973) encoding specificity principle of memory. Although efforts to observe state-dependent effects have revealed mixed findings, substance state-dependent effects have been most consistently demonstrated in studies that ask participants to recall a long-term memory (Eich, 1980; Weingartner, Putnam, George, & Ragan, 1995).

Long-term autobiographical memories, such as trauma memories may be susceptible to alcohol state-dependent effects. That is, the subjective sense of being intoxicated during a trauma may be encoded as a sensory-based trauma memory, and in turn, subsequent intoxication could serve as an internal cue to trigger involuntary recollections of the trauma. Indeed, other forms of sensory-based cues like being touched or a particular posture, or internal sensations like physiological arousal or a feeling of being trapped, can contribute to intrusive reexperiencing of the trauma memory (Ehlers & Clark, 2000). In addition, alcohol might serve as a discriminative stimulus for the conditioned fear response to other environmental cues present during the trauma. In these ways, it is plausible that a trauma encoded in the presence of alcohol could be involuntarily recalled in more detail when subsequently intoxicated, potentially evoking more intrusion-related distress.

In sum, women who experience SA are often under the influence of alcohol during and after the trauma (Abbey et al., 2001; Kaysen, Neighbors, Martell, Fossos, & Larimer, 2006), but it remains unclear how this match in state might affect the occurrence of PTSD-related intrusive memories. Should alcohol use after alcohol-involved-SA increase intrusions, this would be valuable information for clinicians and patients, and could contribute to understanding the high comorbidity between alcohol misuse and PTSD. Therefore, we examined the potential for intrusions to be alcohol state-dependent by examining a sample of college women with a history of SA who completed daily reports of drinking and intrusion symptoms for one month. We hypothesised that women with alcohol-involved SA would report more intrusion symptoms on drinking days than on non-drinking days. We did not expect such differences for participants who experienced non-alcohol-involved SA. Because the distinctiveness of a cue (such as drinking) is important to determine its predictive value for evoking a memory (Goh & Lu, 2012; Nairne, 2002), we also considered frequency of drinking during the study as a covariate. To further differentiate alcohol involvement during SA from other characteristics of the SA, we also considered perceived threat during SA as a covariate, which has been associated with both assault characteristics (Kaysen, Morris, Rizvi, & Resick, 2005) and subsequent PTSD symptoms in past work (Kaysen, Rosen, Bowman, & Resick, 2010).

Method

Participants and procedures

Participants were drawn from a larger study on trauma exposure (Kaysen et al., 2014) in undergraduate women who reported (1) a history of SA with at least one PTSD intrusive symptom and one PTSD hyperarousal symptom and (2) heavy episodic drinking (4+ drinks in one occasion) at least twice in the past month. Participants received $20 for an online screening survey, $25 for an online baseline survey, and up to $160 for daily diary procedures. The daily diary portion of the study consisted of a one-on-one training session followed by 30 days of morning and evening 5-minute surveys on a study personal digital assistant. Participants selected a 2-hour block in the morning and evening to complete their daily surveys, with an alarm set on the personal digital assistant as a survey reminder. Daily PTSD symptoms and past-day alcohol use were assessed only in the morning surveys. All procedures were approved by the University’s Institutional Review Board.

A total of 108 women who identified an index SA since the age of 14 participated. Those who did not respond (n = 3) or were not sure (n = 4) whether they were under the influence of alcohol at the time of this index SA were excluded from analyses. An additional 1 participant was excluded for missing data on perceived threat during the SA. The final sample therefore included 100 women, 67 of whom reported an alcohol-involved SA and 33 who reported a non-alcohol-involved SA. Participants were 18 to 24 years old (M = 20.10, SD = 1.22). The sample consisted of 73% White/Caucasian, 8% Asian/Pacific Islander, 2% Black/African American, 1.0% Native American/American Indian, 12% multi-racial, and 3% other-race students (1% did not indicate a racial identity). Regarding ethnicity, 6% identified as Hispanic/Latina. In addition, 89% identified as heterosexual.

Screening measures

Sexual assault history

SA in adolescence or adulthood (since age 14) was assessed via the Sexual Experiences Survey (SES; Koss & Gidycz, 1985; Koss & Oros, 1982). Unwanted sexual experiences were defined as ‘unwanted oral-genital contact, vaginal/anal intercourse, and/or vaginal/anal penetration by objects’ (response options: 0 = no, 1 = yes). Participants were considered to have a history of SA if they responded ‘yes’ to any item on the SES.

Baseline measures

Alcohol-involved sexual assault

Focusing on the ‘most upsetting’ or index SA experience, participants were asked whether they ‘were under the influence of alcohol or drugs at the time of the assault.’ Response options were 0 = no, 1 = yes, and 2 = not sure (as detailed above, the four participants who responded ‘not sure’ were excluded from analyses). Participants who responded ‘yes’ were asked to specify which of nine substance(s) they used (e.g., alcohol, marijuana, amphetamines, etc.), and if alcohol was endorsed, how many drinks they consumed prior to the assault. Whether individuals were under the influence of alcohol at the time of the SA was coded as 0 = non-alcohol-involved SA or 1 = alcohol-involved SA.

Perceived threat

Regarding the index SA, participants were also asked, ‘During the worst unwanted sexual experience, did you think about being killed or seriously injured?’ with response options ranging from 0 = no, not at all to 4 = I thought about it all the time. In addition, participants were asked, ‘During the incident, how certain were you that you were going to be killed?’ with response options ranging from 0 = I was completely certain that I was not going to be killed to 4 = I was completely certain that I was going to be killed. Consistent with past research on perceived threat (Kaysen et al., 2005, 2010), these two items were summed to create a total score with a possible range from 0 to 8.

Daily diary measures

Alcohol use

After viewing a standard drink chart, participants were asked to report the number of drinks they consumed in the last 24 hours (open ended response). For this study, responses were dichotomized to represent whether any drinks were consumed on a given day. Percent of drinking days was computed out of the total number of diaries completed.

Intrusion symptoms

The PTSD Checklist – Specific version (PCL-S; Weathers, Litz, Herman, Huska, & Keane, 1993) was modified for daily use to assess symptoms related to the index sexual assault. Participants were asked how much each symptom bothered them in the past 24 hours on a scale from 0 = not at all to 4 = extremely. Five items corresponding to the intrusion/reexperiencing symptom cluster of DSM-IV PTSD were summed. Acceptable internal reliability was observed in this sample (α = 0.83).

Data analysis

To determine the amount of between- and within-person variance in intrusion symptoms, an intraclass correlation was computed via an empty-means, random intercept model. Intrusion symptoms were then modeled in SAS PROC GLIMMIX using maximum likelihood estimation to retain participants who did not complete all daily diaries. Although intrusion symptoms were measured using a sum score, this variable was not normally distributed. Therefore, to represent this positively skewed distribution of non-negative integers, we utilised a count distribution and a log link to keep predicted values above 0. A Poisson distribution (in which the variance is assumed to equal the mean) was compared to a negative binomial distribution (which includes a dispersion or scale parameter allowing the variance to differ from the mean) using a deviance difference test. Intrusion symptoms were expected to show within-person fluctuation, but no systematic change over the course of the 30-day monitoring period. Therefore, models with a random intercept but no random or fixed effects of time were considered. Fixed effects were specified as follows:

Specifically, between-person fixed effects were examined for percent of drinking days during the study, perceived threat during SA, and alcohol involvement during SA. Any drinking in the past day was also included as a within-person main effect. The cross-level interaction between alcohol involvement during SA and past-day drinking was included and pairwise comparisons were examined, regardless of the significance of the omnibus test (Howell, 2013).

Results

Descriptive statistics

Among the 100 participants, 67% reported they were under the influence of alcohol at the time of their index SA. These participants reported consuming between 1 and 20 drinks (M = 6.84, SD = 3.76) at the time of assault. In addition, 13% of all participants reported being under the influence of drug(s) in addition to alcohol at the time of the SA, such as marijuana (n = 7), marijuana in combination with hallucinogens, barbiturates, cocaine and/or other drugs (n = 4), or cocaine and/or amphetamines (n = 2). No participants reported other drug use in the absence of alcohol use.

The average participant completed 20.06 daily diaries (SD = 7.40; range = 1 to 29). Diary completion did not differ significantly between those with alcohol-involved and non-alcohol-involved SA, t(98) = 1.56, p = .121, and was not associated with average intrusion severity, r = .11, p = .257, or percent drinking days, r = – .05, p = .609. Participants reported drinking alcohol on 0% to 90.9% of the days that diaries were completed (M = 29.9%, SD = 21.1%). Percentage of drinking days did not significantly differ between participants with an alcohol-involved SA (M = 32.4%, SD = 22.4%) and those with non-alcohol-involved SA (M = 24.8%, SD = 17.5%), t(98) = – 1.71, p = .090.

With regard to perceived threat, participants with an alcohol-involved SA (M = 0.67, SD = 1.38) did not significantly differ from participants with a non-alcohol-involved SA (M = 0.60, SD = 1.30), t(98) = 0.25, p = .806. However, regardless of alcohol involvement, greater perceived threat during the SA was associated with more severe average intrusion symptoms, r = .31, p = .002, and greater percentage of drinking days during the study, r = .20, p = .047.

Prediction of intrusion symptoms

The intraclass correlation revealed that 50.4% of the variance in intrusion symptoms was due to between-person mean differences, and the remaining 49.6% was due to within-person variation. A negative binomial count distribution fit significantly better than a Poisson distribution, – 2ΔLL(1) = 741.45, p < .001, and was therefore retained. Model results revealed a significant interaction between alcohol-involved SA and drinking day, B = 0.30, SE = 0.15, p = .046, after controlling for percentage of drinking days and perceived threat during SA (see Table 1). As can be seen in Table 2, pairwise comparisons revealed that women whose index SA involved alcohol reported significantly more severe intrusion symptoms on drinking days than non-drinking days, t(91) = 3.09, p = .003, after controlling for between-person differences in percentage of drinking days and perceived threat during SA. There was no such effect of drinking on intrusion symptoms for those whose SA did not involve alcohol, t(91) = – 0.15, p = .885. Moreover, there were no significant differences in intrusion symptoms between alcohol-involved and non-alcohol-involved SA during drinking days, t (91) = 0.13, p = .898, or non-drinking days, t(91) = – 0.67, p = .503. Although pairwise comparisons involved contrasting model estimates in log units, exponentiated estimates are represented in Figure 1 to more clearly show model-predicted values on the original scale of the intrusion severity sum score.

Table 1.

Results of Negative Binomial Multilevel Model Predicting Intrusion Severity.

		B	SE	p
Model for the means
γ_00	Intercept	–0.30	0.36	.413
γ_10	Drinking Day	–0.02	0.12	.885
γ_01	Percent Days Drinking	–0.48	0.85	.574
γ_02	Perceived Threat during SA	0.39	0.13	.003**
γ_03	Alcohol-Involved SA	–0.25	0.37	.503
γ_11	Alcohol-Involved SA x Drinking Day	0.30	0.15	.046*
Model for the variance
τ^2_U_0	Random intercept variance	2.63	0.47
	Scale	0.74	0.06

SE = standard error; SA = sexual assault. *p < .05. **p < .01.

Table 2.

Simple Effects Comparisons for Intrusion Severity.

	Non-Drinking Day			Drinking Day
	B	SE	exp(B)	B	SE	exp(B)	t(df = 91)	p
Alcohol-Involved SA	–0.30	0.22	0.74	–0.02	0.23	0.98	3.09	.003**
Non-Alcohol-Involved SA	–0.05	0.30	0.95	–0.07	0.31	0.94	–0.15	.885

SE = standard error; SA = sexual assault. For all estimates, percent of drinking days was held constant at 0.30 and perceived threat was held constant at 1. Estimates (B and SE) are in log units. Exp(B) represented predicted intrusion severity scores plotted in Figure 1. **p < .01.

Figure 1.

Simple effects comparisons for intrusion severity holding percentage drinking days constant at 0.30 and perceived threat constant at 1. **p < .01.

Discussion

This study is the first to consider potential state-dependent effects of alcohol on intrusive SA trauma memories. Consistent with expectations, college women with a history of alcohol-involved SA reported more severe intrusion symptoms on drinking days than on non-drinking days, whereas no such difference was found for non-alcohol-involved SA after controlling for overall drinking frequency and perceived threat during SA. Among alcohol-involved SA survivors, the experience of feeling intoxicated at the time of the assault may have been encoded as a sensory component of the trauma memory. Subsequent intoxication may have served as a sensory-based cue, evoking involuntary memories of the trauma. Because cues for trauma-related memories often go unidentified, the memory may be experienced as spontaneous, distressing, and intrusive (Ehlers & Clark, 2000). In addition, when there is a match (vs. no match) in alcohol state between encoding and retrieval, more details of the trauma memory might be available for recall, and therefore, involuntary intrusive memories may be experienced as particularly detailed, current (i.e. with a greater sense of ‘nowness’), and distressing. It is also possible that drinking episodes may involve exposure to environmental cues (e.g. a bar or party, the smell of a certain drink, witnessing peers’ intoxicated behavior) that are similar to the original alcohol-involved SA, which may allow for context-dependent memory access or serve as cues for intrusive memories. Additional research is needed to determine whether intrusions might be differentially affected by internal subjective cues of intoxication and external environmental cues of drinking context.

These findings also shed light on the overall associations between alcohol-involved SA and subsequent intrusive memories. Prior studies examining alcohol involvement during SA and subsequent PTSD-related intrusions have revealed mixed findings (Aakvaag et al., 2018; Brown et al., 2009; Jaffe et al., 2017; Kaysen et al., 2010; Littleton et al., 2009; Peter-Hagene & Ullman, 2018; Zinzow et al., 2010), and we found no difference by SA type in overall intrusion symptoms. However, intrusions after alcohol-involved SA depended on daily consumption of alcohol in our study, which suggests that post-SA alcohol use may be an important factor not considered in previous work. Frequency of post-SA alcohol use, and therefore exposure to alcohol-related cues, is likely to vary between previously studied samples and might shed light on discrepant findings. We encourage future studies to examine post-SA drinking as a potential moderator of the association between peritraumatic alcohol use and subsequent intrusion symptoms.

In this study, we found no difference in intrusions between drinking and non-drinking days for participants with non-alcohol-involved SA. Although true state-dependent effects would suggest that a sober state at encoding of the trauma would also be followed by better recollection while sober (vs. intoxicated) at retrieval, completeness of voluntary recall for the trauma memory was not assessed in this study. Instead, we assessed the occurrence of intrusive or involuntary memories, which likely reflects both the availability of any complete or fragmentary memory and the presence of cue-driven retrieval. A sober state was not expected to be a distinctive cue specific to the trauma, and therefore would not be expected to functionally improve retrieval (Nairne, 2002).

Although we focused on the potential for state-dependent intrusions in this study, findings could also be interpreted within a self-medication framework. Due to the daily nature of the assessments in our study, we were unable to determine whether drinking or intrusions occurred first on a given day. We focused on intrusion symptoms as our outcome, but it is plausible that individuals who experienced an alcohol-involved SA were more likely to drink in response to having more intrusions, consistent with the self-medication model (Stewart et al., 1998). Indeed, experimental studies have shown that trauma-related cues, including exposure to the trauma memory itself, elicit alcohol cravings in individuals with PTSD and alcohol use disorder (e.g. Coffey et al., 2010). Those who report alcohol-involved SA are also particularly likely to report high rates of drinking both before and after the SA (Kaysen et al., 2006), and therefore, may be especially likely to turn to alcohol as a means of coping with trauma-related distress. Future studies with more time-sensitive assessments of drinking and intrusions are needed to determine the temporal ordering of drinking and intrusions in survivors of alcohol-involved SA. Still, self-medication models and the potential for state-dependent intrusions should not be seen as competing hypotheses, but instead may work in tandem. For example, intrusions for alcohol-involved SA may occur more frequently in the context of alcohol use, which may then lead to increased distress and drinking to cope. We recommend the examination of these reciprocal associations within drinking episodes.

Limitations and future directions

This study utilised archival data collected as part of a larger study, and as such, there are limitations to the conclusions that can be drawn. First, we only considered the presence or absence of alcohol during the assault because blood alcohol concentration could not be reliably estimated with the available data (duration of the pre-SA drinking episode and weight at the time of the SA were not assessed). However, prior research has indicated that alcohol state-dependent effects have been most consistently demonstrated for moderate doses; low doses may not be sufficient to observe an effect and high doses may severely impair memory encoding (Eich, 1980). We recommend considering dose in future examinations of post-SA drinking on intrusions. Second, we only considered alcohol effects in this study, although a small subset of participants (n = 13) also endorsed marijuana or other drugs in addition to alcohol at the time of SA. It is possible that the presence of other drugs at the time of encoding, and potentially at the time of retrieval, could complicate state-dependent effects. Given the low base rates of other drugs during SA in this study, a larger sample is needed to consider this possibility. Relatedly, although we only had data on participants’ alcohol use prior to the assault, alcohol-involved SA commonly involve alcohol use by both the victim and perpetrator (e.g. Aakvaag et al., 2018). Future work is therefore needed to disentangle whether perpetrator alcohol use might also contribute to post-assault intrusions when in drinking contexts, perhaps because being around other intoxicated individuals could serve as a reminder of the perpetrator. In addition, sexual revictimization is common (Walker, Freud, Ellis, Fraine, & Wilson, 2019) and although we did not assess the number of adolescent/adulthood SA experiences, or alcohol-involvement for non-index assaults, such event-level assessments for each SA would allow for more specificity in future studies of SA-related intrusions. Similarly, we included perceived threat as a proxy for SA severity, but incorporating more event-level questions would allow for more direct consideration of SA characteristics when testing the unique effect of alcohol intoxication on intrusions. Moreover, future studies should examine whether patterns are specific to intrusions or related to general increases in PTSD symptoms in settings associated with the SA. Finally, more detailed assessment (e.g. intrusion diaries; Holmes, Brewin, & Hennessy, 2004) would help to clarify whether intoxication affects the occurrence of intrusive memories or resulting distress.

Conclusions

In sum, this study was the first to examine whether alcohol use can act as a trauma-related cue for intrusive memories of SA on a given day using a daily diary methodology. Although prior event-level studies have found daily relationships between increases in same-day PTSD and drinking (Gaher et al., 2014; Kaysen et al., 2014), none have tested whether drinking itself can provide a contextual cue around the traumatic event. Our results suggest that for women whose worst SA involved alcohol, drinking on a given day was associated with same day increases in intrusive symptoms of PTSD. This pattern was not seen for women whose index SA did not involve alcohol. Although replication and additional research is needed, these results have important implications for interventions aimed at addressing alcohol use and PTSD for women who have experienced alcohol-involved SA. Results point to the importance of working with survivors to anticipate how alcohol use may exacerbate, instead of mitigate PTSD symptoms in cases where alcohol was part of the original trauma memory. Therapists may also wish to discuss how using alcohol to manage these symptoms may lead to downstream escalations in alcohol use. Intervening to build up situational coping skills may help prevent this escalation. Just-in-time interventions that reach survivors in distressing situations may be a useful direction for future research.