Title: A meta-analytic examination of assumed properties of child sexual abuse using college samples
Authors and Affiliations:
Bruce Rind, Department of Psychology, Temple University, Philadelphia
Philip Tromovitch, Graduate School of Education, University of Pennsylvania
Robert Bauserman, Department of Psychology, University of Michigan
Citation: Rind, B., Tromovitch, P., & Bauserman, R., “A meta-analytic examination of assumed properties of child sexual abuse using college samples,” Psychological Bulletin, vol. 124, no. 1, 1998, pp. 22-53.
Many lay persons and professionals believe that child sexual abuse (CSA) causes intense harm, regardless of gender, pervasively in the general population. The authors examined this belief by reviewing 59 studies based on college samples. Meta-analyses revealed that students with CSA were, on average, slightly less well adjusted than controls. However, this poorer adjustment could not be attributed to CSA because family environment (FE) was consistently confounded with CSA, FE explained considerably more adjustment variance than CSA, and CSA-adjustment relations generally became nonsignificant when studies controlled for FE. Self-reported reactions to and effects from CSA indicated that negative effects were neither pervasive nor typically intense, and that men reacted much less negatively than women. The college data were completely consistent with data from national samples. Basic beliefs about CSA in the general population were not supported.
Publication of this article resulted in considerable controversy. Details about that controversy can be found here.
The media and scientific literature commonly express the view that (a) CSA causes harm, (b) this harm is pervasive in the population of persons with a history of CSA, (c) this harm is likely to be intense, and (d) CSA is an equivalent experience for boys and girls in terms of its widespread and intensely negative effects. The purpose of this study was to address this question: “In the population of persons with a history of CSA, does this experience cause intense psychological harm on a widespread basis for both genders?”
Much of the literature suffers from a confusion of terminology, and the term “child sexual abuse” is used to capture such a wide range of experiences that a valid understanding of its harmfulness is threatened.
In addition, studies designed to determine whether or not harm is a consequence of CSA often use language that presumes harmfulness, possibly biasing findings. Much of the literature is pervaded by value-loaded terms such as “abuse,” “victim,” and “perpetrator,” in which violation of social norms or “wrongfulness” is often conflated with harmfulness.
Previous reviews of the literature—both qualitative and quantitative—have addressed questions of causality, pervasiveness, intensity, and gender differences.
Qualitative reviews have come to one of two conclusions regarding causality. Some have taken a cautious stance, noting that confounding family environment problems preclude the definite conclusion that CSA is the cause of harm. Others have either directly argued that causality is likely, or used language that implied causation without actually proving it.
With regard to pervasiveness, some reviewers have found that the effects of CSA vary, with some incidents causing harm and others not. Others have implied that negative results are prevalent.
Results regarding intensity have been similar; some reviews have concluded that the intensity of effects varies, while others have implied that intense effects were always present through the descriptive language they used or by listing severe disorders associated with CSA.
As for gender differences, some reviewers have concluded that the data is insufficient to determine whether CSA has different effects on girls and boys. Others have concluded that girls react more negatively to CSA than boys, and many have concluded or implied that the effects are equivalent for boys and girls.
The lack of agreement in the literature on any of the above four issues points to a need for further study. Problems of sample bias, subjectivity, and imprecision are reasons for caution when interpreting the results of these studies. The common use of legal and clinical samples makes generalizing results problematic since such samples do not represent the general population and may be biased in several ways.
In addition, qualitative studies are narrative, so that biases of the author can more easily enter into the studies than is possible in quantitative studies. For these reasons, it is important to examine quantitative reviews.
Only three recent quantitative reviews (Jumper, 1995; Neumann et a., 1996; Rind & Tromovitch, 1997) included a significant number of studies with non-clinical and non-legal samples and avoided subjective biases by using the statistical technique of meta-analysis.
Regarding causality, all three reviews called for caution when making causal inferences from data. They noted that the studies they reviewed did not control for the possibility of other causes of problems for CSA victims, such as other traumatic events or family variables. They also noted that a lack of comparison groups precluded a finding of causality.
Only one of the reviews commented on the question of pervasiveness, suggesting that lasting effects of CSA were rare. However, that finding was based on the results of only one study.
As for intensity, two of the reviews concluded that “the relation between CSA and adjustment in the general population is small.” The third study (Jumper 1995) found a medium sized effect. However, an attempt to re-calculate the effect size in that review yielded discrepancies that could not be resolved, suggesting that the study’s finding might be erroneous.
Only two of the reviews (Jumper 1995, Rind and Tromovitch 1997) examined gender differences, and they used a small number of samples to do so, pointing to a need for further research. Both suggested that sex differences existed.
The small number of studies that had dealt with these issues and their mixed results demonstrated the need for further work. Thus, the purpose of the current review was to address the shortcomings of previous research and to achieve a more accurate and precise understanding of CSA in the general population.
Studies of college samples were chosen for review for the following reasons: (1) College samples provide the largest group of studies on nonclinical populations, which are essential for understanding CSA in the general population. Since half of the general population attends college, such samples are more reflective of the general population than clinical samples. (2) Studies of college samples have the most extensive data on adjustment relations and take confounding variables like family environment into account. (3) They more effectively study male subjects, allowing for gender differences to be better measured.
No systematic review of college studies had been conducted before. Although there are limitations regarding the generalizability of findings based on college samples, these concerns can and will be addressed later in the article.
The authors located studies that exclusively examined college students, or that provided results for college students separate from any other populations. These studies had to meet the following criteria: (a) Include a control group that contained no students with CSA experiences; (b) use a distinct CSA group, rather than a general “abused” group that could include participants without a history of CSA; (c) report on at least one of 18 symptom categories commonly studied in CSA research; and (d) provide sufficient data to compute one or more effect sizes.”
The authors found 59 studies which met these criteria. They contained 70 independent samples (comprised of 13,704 men and 21,999 women) from which prevalence could be measured, and 54 samples (comprised of 3,254 men and 12,570 women) from which effect sizes could be computed.
Altogether, these studies used well over 50 different standard instruments to measure 18 symptoms or categories of psychological correlates of CSA. Some studies also asked students about their reactions at the time of the incident, their current feelings about it, or any effects they believed to have resulted from the incident. Pearson correlation coefficients were used to determine effect sizes for the symptoms.
Based on 26 samples of males from studies providing usable data, 14% of 13,704 participants reported CSA. Based on 45 samples of females, 27% of 21,999 participants reported CSA. Frequency of CSA experiences in these college samples were similar to those found in an earlier review of national population studies.
To determine whether the severity of college students’ CSA experiences mirrored those in the general population, four basic types of CSA were identified—exhibitionism, fondling, oral sex, and intercourse—and their prevalences were compared with those in the earlier review of national population studies. The rates of reported cases of intercourse were comparable for women, but for men they were much higher in the college samples than in the national samples.
In addition, closeness of family relationship between the sexually abused child and the abuser was examined since that factor is often viewed as an aspect of the severity of CSA. The college students reported higher rates of CSA with close family members than did those in the population samples. Thus, the college students in these studies did not experience less severe CSA than those in the general population. However, it was not possible to compare the amount of force or threats used since the national samples did not provide such data.
Turning to the magnitude of the relationship between CSA and psychological adjustment, calculations showed that CSA accounted for less than 1% of the adjustment variance. The authors write, “The obtained small unbiased effect size estimate implies that, in the college population, the magnitude of the relationship between CSA and adjustment is small, which contradicts the assumption that CSA is associated with intense harm in the typical case.”
After calculating the correlations between the eighteen different suspected symptoms and the occurrence of CSA, they conclude that “for all symptoms but one, CSA participants as a group were slightly less well adjusted than control participants. The small magnitude of all effect size estimates implies that CSA effects or correlates in the college population are not intense for any of the 18 meta-analyzed symptoms.”
The impact of gender and degree of consent on effects was examined by conducting semipartial correlational and contrast analyses. Results showed that for males, psychological adjustment was associated with level of consent; that is, unwanted sex was associated with poorer adjustment, while willing sexual activity was not. However, for females, adjustment was not associated with level of consent; CSA was associated with poorer adjustment among females regardless of whether it was wanted.
Thus, the association between CSA and adjustment problems was not the same for men and women. If the definition of CSA was restricted to unwanted sex only, however, then the results implied a gender equivalence between men and women in the association between CSA and adjustment problems.
There was also a possibility that variations among the studies might be related to effect sizes. All studies were similar in that they were based on questionnaires, and almost all were conducted at state universities. They did vary in that some used convenience samples of students, while others used wider samples. However, calculations showed that this variation had no relation to effect size. Nor did variation in mean age of students. On the other hand, published studies did result in slightly larger effect sizes than unpublished ones.
Existence of force, occurrence of penetration, duration of abuse, frequency of incidents, and occurrence of incest were each variables in at least two studies. Some researchers reported correlations between these moderators and symptoms or effects of CSA.
Meta-analysis determined that 3 of the 10 moderator-outcome relations reached statistical significance. Existence of force had a medium-sized relation to negative reactions and self-reported effects, and incest had a small relation to them. Force was unrelated to symptoms, and penetration, frequency, and duration were not related to symptoms or reactions.
Three kinds of self-report data were examined: immediate reactions, current reflections, and self-reported effects of CSA. Fifteen studies, consisting of 10 female and 11 male samples, asked students to recall their immediate reactions to CSA. Seventy-two percent of females and 33% of males reported that they reacted negatively at the time. Additionally, 11% of females and 37% of males reported that they reacted positively.
Seven female and three male samples contained data regarding current reflections. Females currently reported 59% of experiences as negative and males reported 26% as negative. Sixteen percent of females and 42% of males reported current positive reflections.
Eight studies involving 11 samples presented usable data regarding self-reported effects of CSA. Five studies found from 0.4% to 16% of men reported negative effects on their current sex lives. Two studies found that 2.2% and 24% of women made the same report. In one study, 17% of males and 26% of females reported a temporary negative effect on their sexual attitudes.
The authors summarize the self-reports:
The overall picture that emerges from these self-reports is that (a) the vast majority of both men and women reported no negative sexual effects from their CSA experiences; (b) lasting general negative effects were uncommon for men and somewhat more common for women, although still comprising only a minority; and (c) temporary effects were more common, reported by a minority to a majority of women. These data imply that, in the college population: (a) CSA affects males and females differently; (b) lasting negative effects are not prevalent; and (c) when negative effects occur, they are often temporary, implying that they are frequently not intense. These findings are inconsistent with the assumption that CSA has the properties of gender equivalence, prevalence, and intensity in terms of harmful effects.
Meta-analyses were performed to examine the size of sex differences in the three kinds of self-report data. Results showed that women reported significantly more negative immediate reactions, significantly more negative current reflections, and significantly more negative lasting effects than men did.
After finding a correlation between CSA and poorer adjustment as adults, the question of causality arises. Past studies had found that family environment and CSA were confounded, so that family environment may have contributed to effects that had been attributed to CSA.
Studies which included data regarding family environment were identified. These data were classified into the following categories: nonsexual abuse and neglect, adaptability, conflict and pathology, family structure and bonding, and traditionalism. Effect sizes measuring the relationships between these variables and CSA were calculated. Meta-analysis of the effect sizes showed that all were statistically significant. That is, CSA and family environment were confounded in these samples.
A meta-analysis was then performed to determine the degree to which family environment was related to the 18 measures of psychological adjustment. Results showed that with the exception of two measures based on single samples, the effect sizes were generally medium in size. On the other hand, the CSA-symptom and CSA-family environment effect sizes were small. Thus, family environment was a more important predictor of symptoms than was CSA.
To determine the extent to which symptoms were related to CSA apart from family environment, CSA-symptom relations were recalculated while statistically controlling for family environment. Results showed that out of 83 CSA-symptom relations, 34 (41%) were significant before statistical control, but only 14 (17%) remained significant after statistical control.
The authors concluded that “the majority of significant CSA-symptom relations examined in the college samples may have been spurious. These results imply that significant CSA-symptom relations in studies based on college samples cannot be assumed to represent effects of CSA.”
However, they identify three reasons why these findings may have underestimated the association of CSA and symptoms. Firstly, the un-standardized measures of CSA used in the studies may have had less reliability than measures of family environment. Secondly, CSA was usually measured as a dichotomous variable (i.e., present or absent), whose distribution was skewed toward the absent category. Resulting low base rates for CSA may have produced attenuated correlations. Thirdly, the artificial dichotomization of CSA may have directly attenuated correlations.
The authors addressed these threats to their findings by detailing several studies which supported the reliability of the studies in this review, showed that base rates for CSA were not positively correlated to effect size estimates, and found that when CSA was measured on a continuum rather than taken dichotomously, results were consistent with those in this review.
The findings of this review do not support the commonly assumed view that CSA possesses the four properties stated at the beginning of the article. CSA explained less than 1% of the variance in adult adjustment, and family environment and statistical control issues made it impossible to even attribute that variance to CSA.
The fact that these findings are consistent with those of prior studies (Constantine 1981) and those using national samples (Rind and Tromovitch 1997) suggest that concerns about using college samples are not well founded; college samples do adequately reflect the general population to give generalizable results. There is also no evidence to support the hypothesis that college may be too soon to determine whether long-term effects have appeared; one study found that persons under 30 did not differ from those over 30 in CSA-adjustment relations.
The relation between CSA and adjustment was stronger for women than men. Two thirds of male CSA experiences, but less than one third of female CSA experiences, were reported not to have been negative at the time. Three of every eight male experiences, but only one of every 10 female experiences, were reported to have been experienced positively at the time. These results were virtually identical to those found in studies of national samples, further supporting their generalizability.
These gender differences were consistent with the findings by other researchers that males and females perceive sexuality and sexual experiences (e.g., loss of virginity) quite differently. In addition, men and women tended to experience different types of CSA. It was notable that when only unwanted sex was considered, men and women did not differ significantly in adjustment.
Family environment was found to be confounded with CSA and explained nine times more adjustment variance than CSA did, and most CSA-adjustment relations became nonsignificant when family environment was controlled for. Thus, for the most part, adjustment may be an effect of family environment rather than CSA. Five other studies had shown similar results.
Previous attempts to control statistically for family environment suffered from issues of validity, but the authors demonstrated that the current study did not fall prey to those issues. However, the findings of this study do not preclude the possibility that CSA is legitimately a cause of adjustment difficulties in some cases. Evidence suggests that in a minority of instances, CSA can cause significant harm.
Self-reported effects of CSA revealed that lasting psychological harm was uncommon among college students. Self-reported temporary harm, although more common, was not at all pervasive. The self-reported effects data do not support the assumption of wide-scale psychological harm from CSA.
The results of this review failed to support the common belief that contact sex is more harmful than noncontact sex. However, this finding should be viewed with caution because non-contact incidents were not specifically identified in the studies reviewed. Some of the studies examined only contact incidents, while others mixed contact and non-contact incidents together.
Conclusions regarding consent should also be viewed cautiously, since willing incidents were not specifically identified by any of the studies. Some studies examined only unwanted incidents, while others mixed willing and unwanted incidents together.
However, judging by self-reported reactions, most of the mixed incidents for women may have been unwanted, while most for men may have been willing. This may explain why for women, CSA was related to poorer adjustment for both unwanted and mixed incidents, while for men, it was not related to adjustment for mixed incidents.
Only force and incest were found to moderate outcomes. Force was related to self-reported reactions but not to symptoms, while incest moderated both. Penetration, duration, and frequency were not found to moderate outcomes or reactions. However, this conclusion can only be tentative due to the small number of samples upon which it is based.
The studies included in this review typically defined CSA based on moral and legal considerations rather than on empirical and phenomenological criteria. As other researchers have noted, calling behavior “abuse” on these grounds can be problematic for scientific studies.
Historically, clinicians and researchers have conflated morality and science in their study and treatment of masturbation, homosexuality, fellatio, cunnilingus, and sexual promiscuity, hindering a scientifically valid understanding of the phenomena and creating iatrogenic victims. The authors write that the history of confusing morality and law with science in the area of human sexuality suggests a need for caution in scientific studies of sexual behaviors that remain taboo, with child sexual abuse being a prime example.
Especially in the case of males, the positive reactions reported by many subjects seems inconsistent with the use of the term “abuse.” Combining positive and negative experiences into a single category of abuse may incorrectly suggest harm for the former and simultaneously cause an underestimation of harm for the latter. Citing the work of another researcher (Fishman 1991) who reached similar conclusions, the authors argue that the term “child sexual abuse” should be used more cautiously and with greater precision.
It is widely believed that CSA is likely to cause harm, that most children or adolescents who experience it will be affected, that this harm is typically severe, and that it has an equivalently negative impact on both boys and girls. This review failed to support any of these assumptions, in line with prior studies, including those based on national population samples. Current definitions of CSA might well be a factor in this finding, since it is overly inclusive by including willing experiences.
The authors suggest a possible alternative to current definitions of CSA—one suggested by Constantine (1981). Such a definition would take into account a young person’s perception of his or her willingness to participate in sexual activity, as well as his or her perceptions and reactions. They suggest the use of the term “adult-child sex” in cases where the child is a willing participant and perceives it positively.
They also suggest a further precision of language by restricting the use of the term “child” to preadolescents. “Adult-adolescent” sex may be different in character from “adult-child” sex. According to several authors, the former has been commonplace historically, often socially approved, and might fall within the range of “normal” sexual behavior.
The authors close by noting that their conclusions need not have implications for moral or legal decisions. To the extent that such decisions are based on perceptions that CSA is harmful, they might be in need of revision, but other considerations may also play a part.