A meta-analytic review of findings from national samples on psychological correlates of child sexual abuse.

Child sexual abuse (CSA) has received considerable attention over the last two decades by the media, the lay public, mental health care professionals, the legislature, law enforcement personnel, and the judiciary. Much of this attention has concerned the possible psychological consequences of this experience, as is shown by the surge in recent scientific and popular publications (Pope & Hudson, 1995). Our purpose in the current research was to examine these possible consequences by reviewing an important body of literature that has not been systematically examined in previous literature reviews on the psychological correlates of CSA.

First, it is important to discuss definitions. Kilpatrick (1987) argued that, in scientific discussions, abuse (i.e., harm) is something to be established as a conclusion, not to be accepted as a premise. Although we are in agreement with this position, because the term child sexual abuse is used pervasively in the literature under consideration, we have retained this term as a matter of convenience. Based on the manner of use of this term in the studies to be reviewed, CSA is generally defined in the current article as a sexual interaction involving either physical contact or no contact (e.g., exhibitionism) between either a nonadult (i.e., child or adolescent) and someone significantly older (e.g., an older adolescent or an adult), or between two nonadults in which coercion is employed.

Empirical investigations into the psychological correlates of CSA began in earnest during the late 1970s and grew rapidly during the 1980s. Given the emerging literature on correlates, a number of researchers began conducting literature reviews (e.g., Beitchman, Zucker, Hood, DaCosta, & Akman, 1991; Beitchman et al., 1992; Black & DeBlassie, 1993; Briere & Elliot, 1994; Briere & Runtz, 1993; Browne & Finkelhor, 1986; Constantine, 1981; Glod, 1993; Jumper, 1995; Kendall-Tackett, Williams, & Finkelhor, 1993; Kilpatrick, 1987; Mendel, 1995; Neumann, Housekamp, Pollock, & Briere, 1996; Urquiza & Capra, 1990; Watkins & Bentovim, 1992). Among the literature reviews just cited, the authors typically have concluded that CSA is associated with a wide range of psychological problems, including anger, depression, anxiety, eating disorders, alcohol and drug abuse, low self-esteem, relationship difficulties, inappropriate sexual behavior, aggression, self-mutilation, suicide, dissociation, and posttraumatic stress disorder, among others. Additionally, the typical literature review has asserted or implied that (a) CSA causes these problems, (b) these problems occur pervasively throughout the population of persons who have experienced CSA, (c) these problems are generally intense (i.e., severe), and (d) the CSA experiences of boys and girls are equivalent in terms of pervasiveness and intensity of harm. Next we discuss these four fundamental claims in more detail.

Four Basic Conclusions or Implications about CSA

Causality. Conclusions or implications that CSA causes psychological problems have been indicated in the literature reviews by the consistent use of terms such as effects and impact and by a failure to qualify discussions of symptoms associated with CSA by considering alternative explanations for this association (e.g., Black & DeBlassie, 1993; Briere & Elliot, 1994; Browne & Finkelhor, 1986; Kendall-Tackett et al., 1993; Mendel, 1995; Watkins & Bentovim, 1992). Conclusions of causality have also appeared in several reviews in which the authors briefly discussed the methodological principle that correlation is not causation but then argued or implied that the evidence supports causality (e.g., Briere & Runtz, 1993; Glod, 1993; Urquiza & Capra, 1990). In only a minority of the reviews have the authors argued or acknowledged that causality cannot be inferred from the studies they reviewed because of problems such as the presence of confounding variables (e.g., Beitchman et al., 1991; Beitchman et aL, 1992; Constantine, 1981; Kilpatrick, 1987; Neumann et al., 1996).

Pervasiveness. In addition to concluding or implying that CSA causes later maladjustment, authors of the typical literature review have implied that these negative "consequences" are pervasive throughout the population of persons with these early sexual experiences. For example, several a ors started out with the claim that they were conducting a "comprehensive" review of the literature (e.g., Briere & Runtz, 1993; Glod, 1993; Urquiza & Capra, 1990) or were summarizing "what is currently known" (e.g., Briere & Elliot, 1994; Watkins & Bentovim, 1992). Despite claims of comprehensiveness, these authors examined only a small portion of available studies, ones that were mainly based on clinical or legal samples. After listing symptom after symptom to demonstrate the harm "caused" by CSA, these authors then failed to discuss limitations on the generalizability of these findings that were based on highly unrepresentative samples, thereby implying that their findings of harm characterize the population of persons who have experienced CSA. Other authors, although not explicitly claiming to have conducted a comprehensive review, nevertheless also implied that their findings of harm, based mostly on clinical and legal samples, apply to the entire population by similarly failing to qualify the generalizability of these findings (e.g., Beitchman et al., 1992; Black & DeBlassie, 1993; Kendall-Tackett et a., 1993; Mendel, 1995). A number of authors reported that sizable minorities of persons who experienced CSA are asymptomatic (e.g., Briere & Elliot, 1994; Kendall-Tackett et al., 1993). Rather than conclude that reactions and outcomes are variable (i.e., positive, neutral, or negative), Kendall-Tackett et al. (1993) argued that an absence of symptoms may be attributable to insensitive measures of trauma, delayed symptoms that have yet to manifest themselves, or less damaging abuse with which highly resilient children in highly supportive environments have been able to cope. Briere and Elliot (1994, p. 55) similarly rejected the possibility that lack of symptoms indicates no harm by speculating that studies using clinical or forensic samples of abused children may not be generalizable to all sexually abused children because "[t]hese studies may underestimate the impact of abuse in children who are motivated to deny their abuse or children whose reaction to abuse is significantly delayed" (italics added). These lines of argument imply that harm is pervasive. In only a minority of the reviews as it been made clear that findings of harm from the studies examined cannot safely be extended to the entire population of persons who have experienced CSA, implying that harm may not be pervasive in this population (e.g., Beitchman et al., 1991; Constantine, 1981; Kilpatrick, 1987; Neumann et al., 1996).

Intensity of harm. As a group, these literature reviewers have further implied that the harm resulting from CSA is generally intense. By listing long series of symptoms, including such reactions and outcomes as posttraumatic stress, suicide, self-mutilation, and dissociation, many reviewers have implied that effects are severe (e.g., Black & DeBlassie, 1993; Briere & Elliot, 1994; Briere & Runtz, 1993; Browne & Finkelhor, 1986; Glod, 1993; Kendall-Tackett et al., 1993; Mendel, 1995; Urquiza & Capra, 1990; Watkins & Bentovim, 1992). Briere and Runtz; (1993, pp. 320-21) discussed the "extreme psychic pain" and the "extreme dysphoria" that result from CSA Browne and Finkelhor (1986, p. 163) concluded that "sexual abuse is a serious mental health problem, consistently associated with very disturbing significant portion of its victims." Glod (1993, p. 171) concluded that the "consequences of early childhood abuse are far reaching and have been linked to the development of significant impairment in daily functioning and severe psychiatric and medical disorders." Mendel (1995, p. 101) concluded that "child sexual abuse has pronounced deleterious effects on its victims." Only a minority of the reviewers presented findings to show that reactions and outcomes are highly variable, rather than being typically highly negative (e.g., Constantine, 1981; Jumper, 1995; Kilpatrick, 1987; Neumann et al., 1996).

Equivalence of male and female CSA experiences. A fourth implication of the literature reviews is that reactions and outcomes for boys and girls are the same (i.e., highly negative). Black and DeBlassie (1993) reviewed studies concerned with outcomes for boys and cited research that indicates that "sexual abuse in childhood has, at the very least, an equivalent impact on males and females" (p. 128). Watkins and Bentovim (1992), who also limited their review to outcomes for boys, began by listing a number of "prevalent myths about the sexual abuse of male children or adolescents." One of these "myths" was that "[w]hen boys are abused they are less psychologically affected than girls, both initially and in the long term" (p. 27). Mendel (1995) similarly focused on "consequences" for boys. He was critical of the opinions of other researchers (e.g., Rush, 1980) who maintained that boys are less affected by sexual experiences with adults than girls are. He argued that "[i]t appears to be an exercise in futility to attempt to determine whether males or females are affected more adversely by sexual abuse. As most researchers conclude, however, and as the studies reviewed here indicate, child sexual abuse has pronounced deleterious effects on its victims, regardless of their gender" (p. 101).

All other reviews cited previously consisted mostly or exclusively of studies using female participants. Nevertheless, reviewers generally stated or implied that reactions and outcomes are similarly negative for males and females. Briere and Runtz (1993, p. 312) stated that "[a]though most research on long-term effects of sexual abuse has been limited to female samples, it is likely that sexually abused males experience similar problems in psychological functioning." Briere and Elliot (1994) equated the sexes in their reactions and outcomes by using expressions such as "he or she" when referring to persons who experienced CSA. These authors, as well as others (e.g., Glod, 1993), have further given the impression of equivalence of outcome between the sexes by failing specifically to discuss sex differences. Typically, after evidence for harm was provided based on female samples, the effects of CSA on "children" were discussed, implying that boys are similarly affected because boys are children, too. Kendall-Tackett et al. (1993) noted that their studies indicated an absence of consistent gender differences, although they did speculate that only the most symptomatic boys may have ended up in their samples. Only one review was explicit in not equating the experiences of boys and girls, concluding that sex differences are unclear because they have not been adequately addressed in studies (Beitchman et al., 1991).

Problems with Conclusions and Implications of the Literature Reviews

Sampling biases. These literature reviews have predominantly consisted of studies conducted on clinical or legal samples. In a sizable minority of these reviews, the studies that were included were restricted entirely to those using clinical or legal samples (e.g., Beitchman et al., 1991; Black & DeBlassie, 1993; Glod, 1993; Kendall-Tackett et al., 1993; Watkins & Bentovim, 1992). Studies sampling from other populations were much less frequently included, despite the fact that a large literature exists on psychological correlates of CSA based on college samples. At least 60 such college studies are currently available, locatable through databases such as PsycLIT or Dissertation Abstracts International, most of which were available to researchers who conducted literature reviews in the 1990s.

Clinical and legal samples are highly problematic because they cannot be assumed to be representative of the general population of persons who have experienced CSA (Okami, 1991; Rind, 1995), and because data coming from these samples are vulnerable to a number of biases that threaten their validity (Pope & Hudson, 1995). Okami (1991) examined a group of adults who had experienced CSA as negative, neutral, or positive. Negative responders included both clinical and nonclinical participants. He found that clinical participants evidenced a variety of adjustment problems that were substantially more pronounced compared to other participants, including nonclinical negative responders. He concluded that clinical patients with adverse CSA experiences appear to constitute the negative polar extreme of reactions and outcomes. Rind (1995) reviewed the results of a large number of studies based on college samples from which researchers examined the psychological correlates of CSA. He reported that most researchers either found no adjustment differences between sexually abused (SA) students and control students or found only a few differences among many measures of adjustment--differences that were small in terms of effect size. Rind also reported that, as a group, the college samples showed that females and males react differently to CSA, with about two thirds of females reacting negatively compared to only one third of males. These results are important in terms of trying to understand correlates of CSA in the population of persons with this experience, because about half of U.S. adults have some college exposure (Fritz, Stoll, & Wagner, 1981; U.S. Bureau of the Census, 1995). These findings suggest that clinical samples are not representative of the general population.

Pope and Hudson (1995) argued that clinical samples are not only unrepresentative of the general population but are also unrepresentative of nonclinical persons suffering from the same presenting symptoms because of selection bias. These researchers were critical of the validity of results from CSA research using clinical samples because of a number of methodological considerations. They pointed to information bias as a major threat to the internal validity of the findings from these samples. Aside from possible investigator biasing effects (cf. Rosenthal, 1977), they argued that participant effects posed serious problems for making causal inferences concerning the relation between CSA and later maladjustment. Patients searching for causes of their problems (termed effort after meaning) are more likely to recall CSA experiences than are controls, thereby inflating the correlation between CSA and maladjustment, they argued. Further, they were critical of the general failure of CSA researchers to take into account confounding variables that could be responsible for the CSA-maladjustment link. They concluded their methodological critique by commenting that many of these studies "are so severely vulnerable to selection bias, information bias, and lack of consideration of confounding variables that they are rendered almost valueless" (p. 378).

Pope and Hudson (1995) methodological argument that maladjustment in general cannot safely be attributed to past CSA experiences because of confounding variables has received support from recent empirical research, which has demonstrated that CSA tends to be confounded with other factors that are likely to affect later psychological adjustment (Alexander & Lupfer, 1987; Bergdahl, 1982; Briere & Runtz, 1988a,b, 1990; Cole, 1987; Collings, 1994; Edwards & Alexander, 1992; Fishman, 1990; Fromuth, 1986; Greenwald, 1993; Harter, Alexander, & Neimeyer, 1988; Higgins & McCabe, 1994; Jackson, Calhoun, Amick, Maddever, & Habif, 1990; Kinzl, Traweger, Guenther, & Biebl, 1994; Nash, Hulsey, Sexton, Harralson, & Lambert, 1993a; Ney, Fung, & Wickett, 1994; Pallotta, 1991; Pizzolo, 1989; Yama, Tovey, & Fogas, 1993). Researchers who have statistically controlled for confounding variables such as emotional neglect, physical abuse, psychological abuse, and general family disruption have tended to find that the relationship between CSA and maladjustment disappears (e.g., Cole, 1987; Fromuth, 1986; Harter et al., 1988; Higgins & McCabe, 1994; Nash et al., 1993a; Pallotta, 1991).

Subjectivity and imprecision. Another shortcoming of previous literature reviews is that most have been narrative and qualitative. Jumper (1995) argued that conclusions drawn from this method of review are vulnerable to being subjective and imprecise. In an attempt to be more objective and precise, Jumper (1995) and Neumann et al. (1996) recently conducted meta-analyses of CSA studies. A meta-analysis is a quantitative analysis of the statistical findings from a collection of studies. From each study, a statistic is obtained (e.g., a t or an F statistic assessing the difference in adjustment between a CSA and control group). All statistics are then converted to a common statistic (e.g., to standard normal deviate zs or Pearson rs). The new statistics are then combined into one single overall statistic that can then be used to (a) infer whether, in the population, one variable (e.g., CSA) is significantly associated with another (e.g., adjustment) and (b) estimate the strength of this association. This latter use is a frequent focus of meta-analyses, and the strength of association is referred to as the effect size--effect size can also be conceptualized as the standardized size of the difference in mean responses (e.g., adjustment) between two groups (e.g., a CSA versus control group). Two common measures of effect size are Pearson's r and Cohen's d; the former measure is simply the correlation between group membership and response, whereas the latter represents the number of standard deviations between the means of the responses of two groups. Cohen (1988) suggested that rs of .10, .30, and .50 and ds of .20, .50, and .80 can be interpreted as small, medium, and large effect sizes, respectively.

Jumper (1995) meta-analyzed 26 published studies containing 30 samples (5 male and 25 female), of which approximately one fourth were community, clinical, student, and mixed. She found that CSA respondents consistently performed more poorly than controls on various measures of psychological adjustment. Jumper used Pearson's r to assess the strength of association between CSA and adjustment problems. Averaging her findings across measures revealed a medium effect size in the case of clinical samples (r = .31), but only a small effect size in the case of college samples (r = .07). Neumann et al. (1996) meta-analyzed 38 studies consisting of 38 female samples, about half of which were clinical and half of which were nonclinical. They also found that CSA respondents were more poorly adjusted than controls. Using Cohen's d, they found that the size of this difference was medium in the case of clinical samples (d = .50), but small in the case of nonclinical samples (d = .32). The results of Jumper's (1995) and Neumann et al.'s (1996) meta-analyses, taken together, imply that clinical samples cannot be assumed to be representative of other populations. For example, in the case of Jumper's study, whereas CSA accounted for nearly 10% of the adjustment variance in the clinical samples, it accounted for only one half of 1% of the adjustment variance in the student samples (percent variance accounted for is obtained by squaring the effect size r and then multiplying this result by 100; although appropriate for describing percent of variance explained, [r.sup.2] is not currently considered to be a good measure of effect size by leading methodologists because it underestimates this metric, Rosenthal & Rosnow, 1991).

Current Study

The goal for the current study was to examine the basic conclusions and implications that are contained in many previous literature reviews, which hold that CSA causes pervasive and intense harm, regardless of gender, in the population of persons with a history of CSA. The findings of Okami (1991), Pope and Hudson (1995), and Rind (1995), as well as the results of the meta-analyses by Jumper (1995) and Neumann et al. (1996), suggest that these conclusions and implications, based mainly on findings from clinical and legal samples, may not be valid for the population of persons with a history of CSA, because clinical and legal samples are not representative of this population. Most appropriate for addressing causality, pervasiveness, intensity, and gender equivalence in the population of persons with a history of would be studies based on samples that are representative of the general population.

Although it may not be possible to obtain a perfectly representative sample of the general population, good approximations have been made by a number of researchers who have employed national probability samples. National probability samples are selected in such a way so as to be representative of the (usually adult) population in the nation where the sampling is done. Because response rates are never 100%, and because certain segments of the adult population are generally excluded from the sampling frame (e.g., prisoners or college students living in dormitories), these samples cannot be considered to represent precisely the national population. Nevertheless, these samples serve as good approximations of the national population and are more representative of this population than any other types of samples that have been selected. Therefore, to examine whether CSA causes pervasive, intense harm, regardless of gender, in the population of persons with a history of CSA, we attempted to collect and analyze all currently available studies conducted on national probability samples that included data relevant to psychological correlates or effects of CSA. Relevant data could come from measures of psychological or sexual adjustment or from self-reported effects of the CSA experience. We also employed meta-analytic techniques in an attempt to achieve greater objectivity and precision compared to previous qualitative and narrative reviews.

Method

Sample of Studies

To obtain studies using national probability samples in which researchers examined self-reported effects or psychological correlates of CSA, several strategies were employed. Studies using national probability samples already known to the authors were included. Computer searches of Dissertation Abstracts International, Sociofile, PsycLIT, and ERIC were made to locate additional studies. The keywords and phrases entered were (adjustment or effect or effects) and (national or nationally) and (sex abuse) or (sexual abuse) or (child and adult and sexual). Finally, obtained articles were examined for mentions of additional studies.

These strategies yielded 17 studies, of which 7 were included for analysis in the current article. The remaining 10 studies were eliminated either because they were unobtainable (3 studies), or because they did not satisfy one or more criteria for inclusion in the current article (7 studies). These criteria were (a) the study had to include both male and female respondents, (b) the study had to present either summaries of self-reported effects or measures of psychological or sexual adjustment, (c) these summaries had to be reported separately for male and female respondents, (d) researchers who reported results of measures of adjustment had to have included a control group, and (e) the self-reported effects or the adjustment data had to be presented quantitatively. The inclusion of both male and female respondents and the separate reporting of their results were required so as to be able to address the question of whether the CSA experiences of males and females are equivalent. Most obtained studies were not retained because they contained no outcome data. Self-reported effects refers to asking respondents whether they thought their experience of CSA had an effect (i.e., impact on psychological or sexual adjustment) upon their lives that could be classified into the two mutually exclusive and exhaustive categories of negative and nonnegative. Nonnegative effects could include perceptions of neutral or positive impact. Self-reported effects did not include reports of emotional reactions at the time (e.g., fear or pleasure), because such reactions are not necessarily mutually exclusive and do not necessarily indicate that effects occurred. Finally, quantitative presentation of the self-reported effects or adjustment data was required so as to be able to compare and combine the results across studies via meta-analysis.

Four of the seven included studies came from the United States, and one each came from Great Britain, Canada, and Spain. Table 1 lists the seven studies along with the populations of inference, methods of gathering data, definitions of CSA, sample sizes reported for males and females, prevalence rates of CSA, and response rates. In the case of studies that included measures of adjustment (Bigler, 1992; Boney-McCoy & Finkelhor, 1995; Laumann, Gagnon, Michael, & Michaels, 1994; Lopez, Carpintero, Hernandez, Martin, & Fuertes, 1995; Los Angeles Times Poll, as reported by Finkelhor, Hotaling, Lewis, & Smith, 1989), sample size refers to the number of respondents who were included in data analysis. In the case of studies that included only self-reported effects (Badgley et al., 1984; Baker & Duncan, 1985), sample size refers to the number of respondents who agreed to participate.

Table 1

Attributes of Seven Studies Using National Probability Samples to Examine Psychological Correlates or Effects of Child Sexual Abuse

                     Population          Data
Study                of Inference        Gathering(a)

Badgley et al.       Canada:             SAQ
  (1984)             ages 18+
Baker & Duncan       Great Britain:      FTF
  (1985)             ages 15+
Bigler               U.S.:               Mail
  (1992)             ages 30 to 55
Boney-McCoy &        U.S.:               Tele
  Finkelhor (1995)   ages 10 to 16
Finkelhor            U.S.:               Tele
  et al. (1989)      ages 18+
Laumann              U.S.:               FTF
  et al. (1994)      ages 18 to 59
Lopez et al.         Spain:              FTF
  (1995)             ages 18 to 60       SAQ

                     Definition              Sample Size(c)
Study                of CSA(b)             Males      Females

Badgley et al.       any unwanted sex;     1002       1006
  (1984)             C, NC
Baker & Duncan       < 16 ("sexually        834        923
  (1985)             mature"); C, NC
Bigler               < 18 (5+, family,      140        174
  (1992)             or coerced); C, NC
Boney-McCoy &        any unwanted sex;      987        911
  Finkelhor (1995)   C, NC
Finkelhor            < 19; any sex now     1142       1476
  et al. (1989)      seen as SA; C, NC
Laumann              < puberty (past       1311       1608
  et al. (1994)      puberty); C only
Lopez et al.         < 17 (5+, or           462        433
  (1995)             coercion); C, NC

                      CSA Prevalence(d)       Response
Study                Males      Females       Rate

Badgley et al.       31%        53%           94%
  (1984)
Baker & Duncan        9%        14%           87%
  (1985)
Bigler               36%        51%           33%
  (1992)
Boney-McCoy &         6%        15%           72%
  Finkelhor (1995)
Finkelhor            15%        28%           76%
  et al. (1989)
Laumann              12%        17%           79%
  et al. (1994)
Lopez et al.         15%        22%           82%
  (1995)

(a) FTF = face-to-face interviews; SAQ = self-administered questionnaires; Mail = mail survey; Tele = telephone survey

(b) Ages qualifying as "child" given first; in parentheses, ages for other person and any other conditions; C = contact sex, NC = noncontact sexual experience

(c) Includes number of respondents used in data analyses in studies assessing adjustment; otherwise, indicates number of actual participants

(d) Based on actual number of respondents who participated

The sampling frames (i.e., method for selecting a representative sample) in these studies generally consisted of area probability sampling from nearly all households throughout the countries in which the studies were conducted. Populations of inference generally consisted of adults aged 18 and over in the countries studied, although a number of researchers used restricted age ranges (e.g., 18 to 59 and 30 to 55); one study was based on respondents from 10 to 16 years of age. Data gathering consisted of face-to-face interviews in three studies, a mail survey in one study, telephone interviews in two others, and a self-administered questionnaire in another study. Two studies with face-to-face interviews also included a self-administered questionnaire. Definitions of CSA varied from study to study. In all studies but one, child included both children and adolescents--i.e., young persons aged younger than 19, 18, 17, or 16 years, depending on the study. In one study (Laumann et al., 1994), child referred to prepubescent persons. In all studies but one, events classified as CSA included both contact and noncontact sexual experiences. Laumann et al. (1994) included only sexual experiences involving physical contact. CSA was restricted to age-discrepant sexual experiences in only two studies. The other five included abusive peer interactions of a sexual nature as well, although the focus was generally on age-discrepant sexual experiences.

Total sample sizes of respondents who actually participated in these studies ranged from 314 to 3,432. Excluding the low return rate of 33% in the mail survey study, response rates ranged from 72% to 94% (M = 82%, SD = 8%). Prevalence rates of CSA varied considerably from study to study. For male respondents, these rates ranged from 6% to 36% (M = 17.7%, SD = 11.3%), whereas for female respondents, prevalence rates ranged from 14% to 53% (M = 28.6%, SD = 16.7%). This variability in prevalence rates is attributable to differing definitions of CSA. Two studies employed widely inclusive definitions of CSA. In addition to defining older-younger sexual experiences to be CSA, Bigler (1992) defined CSA to include all sexual experiences persons under 18 years had with family members, regardless of their age. Badgley et al. (1984) included persons over 18 years of age if this was the earliest age at which they had their first unwanted sexual experience. As will be discussed later, most unwanted sexual experiences first occurred when respondents in this study were under 18, thus qualifying as CSA. Excluding these two studies, prevalence rates ranged from 6% to 15% for males (M = 11.4%, SD = 3.9%) and from 14% to 28% for females (M = 19.2%, SD = 5.8%).

Measures

Self-reported effects. In three studies researchers collected and reported data on self-reported effects (Badgley et al., 1984; Baker & Duncan, 1985; Laumann et al., 1994). Badgley et al. (1984) asked their Canadian respondents whether they had experienced sex with someone when they "didn't want this." If they answered affirmatively, they were then asked to indicate at what age or ages this occurred and whether they had been physically injured or emotionally or psychologically harmed by the first such incident of this type at the time it occurred. Baker and Duncan (1985) asked their British respondents with a history of CSA to indicate the effect on them of this experience by choosing one of four options: (a) unpleasant and harmful at the time but had no lasting effects, (b) permanently damaging with long-term effects, (c) had no effect at all, or (d) had improved the quality of their life. Laumann et. (1994) asked their U.S. respondents with a history of CSA whether this experience had affected their lives since it happened.

Psychological or sexual adjustment. In five studies researchers used various measures to assess psychological or sexual adjustment among their control and CSA respondents (Bigler, 1992; Boney-McCoy & Finkelhor, 1995; Finkelhor et al., 1989; Laumann et al., 1994; Lopez et al., 1995). Bigler (1992) measured sexual functioning in his sample of U.S. respondents with two instruments: (a) the Sexual Esteem subscale of the Sexuality Scale (Snell & Papini, 1989) and (b) the Golombok-Rust Inventory of Sexual Satisfaction (Rust & Golombok, 1986), which measured sexual dysfunction. Bigler also used the Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979) to measure the level of trauma associated with CSA. Boney-McCoy and Finkelhor (1995) created an instrument to measure trauma related to posttraumatic stress disorder (PTSD), which contained items asking their sample of U.S. children and adolescents how often they experienced in the past week each of 10 symptoms. These symptoms were all associated with PTSD and were modified from the SCL-90-R. Boney-McCoy and Finkelhor also assessed lifetime depression, occurrence of sadness in the past month, and occurrence of trouble with a teacher in the last year. In Lopez et al.'s (1995) study conducted on a national Spanish sample, the Self Reporting Questionnaire was used to assess current psychological adjustment.

Finkelhor et al. (1989) reviewed the results of the Los Angeles Times Poll, which was conducted by telephone on a national sample of U.S. residents, asking them to respond to a series of items about CSA. Finkelhor et al. argued that three items were relevant to respondents' current level of functioning and adjustment: extent of marital disruption, satisfaction with intimate relationships with the opposite sex, and being a religious nonpractitioner. Finkelhor et al. argued that this last item was a valid indicator of long-term harm because "Russell ... found that victimized women were more disillusioned with religion than were nonvictimized peers" (p. 393). Despite this argument, we did not retain this last item in our analysis because, unlike marital disruption and sexual satisfaction, religious nonpractice is not a face-valid measure of adjustment. Finkelhor et al. also analyzed respondents' attitudinal responses to questions concerning CSA. They argued that attitudinal differences between sexually abused (SA) and control respondents were also indicative of the long-term negative impact of CSA.

We analyzed the 89 attitude items presented by Finkelhor et al. (1989) to determine which items were face-valid indicators of long-term negative impact. Four judges (the two authors and two other sex researchers, all of whom are familiar with the CSA literature and are currently involved in CSA research) judged the validity of each item by answering two questions. The first read, "Assuming that sexual abuse has a long-term negative impact, would you expect to find a difference in the proportion of SA vs. control respondents' responses to this item (and in what direction)?" If a judge did expect a difference, then the judge answered the second question, which read, "If there is a difference in the expected direction, was it most likely caused by a long-term negative impact from the SA experience, or could some other factor reasonably account for this difference?" If the judge decided that the difference was most likely caused by the SA experience, then the judge considered the attitude item to be a valid measure of long-term harm. Each judge rated the items independently. The mean pair-wise interjudge agreement across all items was 74%. Next, the judges convened to discuss discrepancies on items where their judgments were not unanimous. For 43 items, unanimous agreements of validity were reached; these items were retained as the final set of valid indicators of harm.

In Laumann et al.'s (1994) study using a U. S. sample, respondents were asked a series of questions relevant to their current psychological and sexual adjustment, as well as to their level of sexual activity. Both authors independently examined these questions to determine which ones were valid measures of adjustment and which ones measured only sexual activity without implication of adjustment. We reached 100% agreement on our first evaluation of the items. We judged 11 items for males and 10 items for females to be valid measures of adjustment--most of these items assessed sexual difficulties. We judged the remaining seven items to be measures of sexual activity without implication of adjustment problems.

Procedure

For the self-reported effects data, the percentages of males and females with a history of CSA who reported negative effects resulting from their CSA experiences were tabulated. For each study reporting self-reported effects data, males and females were compared by (a) contrasting the proportion of each gender that reported negative effects and (b) computing the effect size of this contrast. The effect size used for these comparisons was Pearson's r. Formulas for calculating r were taken from Rosenthal (1984, 1995). Positive rs indicated that males reported fewer negative effects, or more neutral or positive effects, than females. The effect sizes comparing the genders were then meta-analyzed using formulas from Rosenthal (1984) and Shadish and Haddock (1994). The meta-analysis involved several steps. First, the effect sizes were combined by (a) transforming each r to a Fisher z, (b) multiplying each Fisher z by the degrees of freedom associated with its sample (N-3), (c) summing these weighted Fisher zs, (d) summing the degrees of freedom across all samples, (e) dividing the sum of the weighted Fisher zs by the sum of the degrees of freedom, and (f) converting the resulting mean Fisher z to a Pearson r. The resulting r represents the mean weighted effect size and is referred to as the unbiased effect size estimate ([r.sub.u]). The unbiased effect size estimate is used to estimate the effect size in the population and is considered to be unbiased because it weights more heavily larger samples whose effect sizes are generally considered to be more precise population estimates.

In addition to combining the effect sizes, they were compared as well. Comparing effect sizes determines whether they are all of the same general magnitude and in the same direction (i.e., positive or negative). If the set of effect sizes has these commonalities, then the effect sizes are said to be homogeneous, and their combined value can be taken as a valid population estimate. Comparing a set of effect sizes is achieved by summing the products of each sample's degrees of freedom (N-3) times the square of the difference between a sample's Fisher's z and the mean Fisher z across all samples. The resulting statistic is distributed as Chi-square with k-1 degrees of freedom, where k represents the number of samples. A non-significant result indicates that the effect sizes are homogeneous. Next, a 95% confidence interval around the unbiased effect size estimate was computed using the formula presented by Shadish and Haddock (1994). This interval provides the range of effect sizes that has a probability of .95 of containing the population effect size. This interval is also useful for evaluating the statistical significance of the effect size estimate; if the interval does not contain zero, the effect size estimate is significant.

For the adjustment data, the statistics assessing the difference between the CSA and control groups in terms of psychological or sexual adjustment were converted to effect sizes (Pearson rs). Positive rs indicated that CSA was associated with poorer adjustment, whereas negative rs indicated the reverse relation. In most studies reporting adjustment data, more than one measure comparing the CSA and control respondents in terms of their adjustment was used. For studies with multiple measures of adjustment, an effect size r was computed for each measure, an en se rs were averaged using Fisher z transformations to obtain a single mean effect size, termed the study-level effect size. This practice has been used in other meta-analyses (e.g., Erel & Burman, 1995; Neumann et al., 1996) and has been recommended by Rosenthal (1984). Individual and study-level effect sizes were computed separately for males and females. The study-level effect sizes were then meta-analyzed for males and females separately. Meta-analysis consisted of combining and comparing effect sizes, as well as computing 95% confidence intervals.

Results

Self-Reported Effects

Badgley et al.'s (1984) Canadian sample. Of the 307 men who reported an unwanted sexual experience, 3.9% said they had been physically injured, and 6.8% said they had been emotionally or psychologically harmed at the time by the first such incident. For the 538 women who reported an unwanted sexual experience, these figures for physical injury and psychological or emotional injury were 19.9% and 24%, respectively. About two thirds of both male and female respondents who reported unwanted sexual experiences were younger than age 18 when the first such experience occurred. Thus, if we assume that self-perceived harm was most likely to occur with younger persons, an upper boundary of self-reported psychological or emotional harm at the time resulting from CSA for persons less than 18 would be 10% for males and 36% for females. An analysis of sex differences revealed that men reported significantly fewer physical injuries than did women, z = 6.43, p [is less than] .001, two-tailed, and significantly less emotional or psychological harm than did women, z = 6.29, p [is less than] .001, two-tailed. The effect sizes for sex differences for these two indices of harm were small to medium, r = .22 in both cases.

Baker and Duncan's (1985) British sample. For the 79 males with a history of CSA who responded to the question concerning self-perceived effects of their CSA experience, the following distribution resulted: 4% said their experience caused permanent damage; 33% said it was harmful at the time, but with no lasting effects; 57% said it had no effect; and 6% said it improved the quality of their life. The distribution for the 119 females with a history of CSA who responded to this question was the following: 13% reported permanent damage; 51% said it was harmful at the time, but with no lasting effects; 34% said it had no effect; and 2% said it improved the quality of their life. These distributions show that reports of permanent harm were far from pervasive (4% of males and 13% of females), whereas reports of no negative effects at all were common (63% of males and 36% of females). We contrasted the proportion of male versus female respondents who reported either temporary or permanent harm and found that females perceived significantly greater harm from these experiences than males, z = 3.73, p [is less than] .001, two-tailed. The effect size for this sex difference was medium, r = .26. Baker and Duncan found that girls on average experienced CSA at a younger age (10.74 years) than boys (12.03 years) and that they were three to four times more likely to experience incest than boys. Bringing together these situational differences in CSA with the differences in reported harm between the genders, Baker and Duncan concluded that "reports of harmful outcome were associated more with close family relationships, a younger age at the time of first abuse and being female" (p. 465).

Laumann et al.'s (1994) U.S. sample. In Laumann et al.'s (1994) survey o Americans, respondents who acknowledged being sexually touched by an older person when they were pre-pubescent were asked whether this experience had affected their lives since it happened. Laumann et al. found that "Where was a striking difference between the reports of women and men" (p. 343). Of the 273 sexually touched women who answered this question, 70% said the experience had affected them. Of the 134 sexually touched men who responded to this item, 45% said the experience had affected them. Reports of effects were nearly all negative. A significantly greater proportion of women than men reported effects, z = 4.91, p [is less than].001, two-tailed, effect size r = .24.

Synthesis of the three samples. Results of the three studies providing data on self-reported effects are summarized in Table 2. The percentage of males and females reporting harm was considerably lower in the Canadian study than in the U.S. or British studies. These lower percentages may be attributable to several factors. As discussed previously, the Canadian percentages apply to persons older, as well as younger, than 18. Even with adjusted percentages (10% for males and 36% for females), the Canadian percentages still fall far below those of the other two studies. This remaining difference may be attributable to the fact that respondents in the Canadian study were asked only about harm at the time and for the first such experience. Broadening the time frame and the number of experiences would likely have increased the percentages, bringing them closer to those obtained in the other two studies.

Table 2

Percentage of Male and Female Self-Reports of Negative Psychological
Effects of Child Sexual Abuse

                  Time      Males       Females
Study             Frame     %   N       %    N       Z        r

Badgley et al.    then      7   307     24   538   6.29(*)   .22
  (1984)(a)       only
Baker & Duncan    then &   37   79      64   119   3.73(*)   .26
  (1985)(b)       since
Laumann et al.    then &   45   134     70   273   4.91(*)   .24
  (1994)(c)       since

(a) Data based on first unwanted sex, about two thirds of which occurred prior to age 18

(b) Data based on CSA under age 16

(c) Data based on sexual touching before puberty with older persons

(*) p < .001, two-tailed

Averaging the percentages of self-reported harm across the three studies is not straightforward because of differences in ages included for CSA and because of differences in time frames for perceived effects. Using only the British and U.S. studies suggests that a minority of males (42%) and a majority of females (68%) perceive harm from CSA at some point after it occurs. The Baker and Duncan (1985) data suggest that, of those who report harm, only a small portion perceive themselves to be seriously harmed to the extent that they perceive lasting effects. These results imply that, in the population of individuals who have experienced CSA, perceive arm is not pervasive, particularly for males. Moreover, when harm is perceived, these data suggest that, although it may be perceived as serious (i.e., causing lasting damage) for some individuals, it is not typically perceived this way by individuals who perceive harm.

We conducted a meta-analysis of the effect sizes contrasting the N = 1,450 males and females in the three studies who reported CSA experiences to determine an overall assessment of the gender difference. This analysis revealed that the effect sizes were homogeneous, [chi square](2) = .34, p [is greater than] .80. Combining the effect sizes produced a small to medium unbiased effect size estimate, [r.sub.2] = .23, which was significantly different from zero, z = 7.13, p [is less than].001, two-tailed. The 95% confidence interval was from r = .18 to .28. These results imply that, in the population of persons who experience sexual encounters classified as CSA, males on average perceive it to be less negative, or conversely more neutral or positive, than females do.

Additional support for the conclusion that males and females experience CSA differently was provided by Lopez et al. (1995) in their national study conducted in Spain. These researchers provided emotional reaction data regarding how respondents felt at the time of their CSA experiences. These data are not useful for inferring effects for reasons discussed previously (e.g., pleasure and fear are not mutually exclusive; reactions do not necessarily lead to effects). Nevertheless, they are useful for shedding more light on sex differences in how CSA is experienced. Moreover, they are suggestive of effects, because reactions have been shown to be related to effects (Long & Jackson, 1993). Although males and females responded similarly in reactions of shame and guilt, a significantly smaller proportion of males than females experienced fear (45% vs. 69%, z = 4.39, p [is less than] .001, two-tailed, effect size r = .24), disgust (3% vs. 70%, z = 12.15, p [is less than] .001, two-tailed, effect size r = .66), distrust (0% vs. 77%, z = 13.87, p [is less than] .001, two-tailed, effect size r =.76), feelings of isolation or marginalization (4% vs. 19%, z = 4.00, p [is less than] .001, two-tailed, effect size r = .22), hostile or aggressive feelings toward the older person (41% vs. 67%, z = 4.71, p [is less than] .001, two-tailed, effect size r = .26), and anxiety (29% vs. 48%, z = 3.48, p [is less than] .001, two-tailed, effect size r = .19). Moreover, a significantly greater proportion of males experienced pleasure and satisfaction than females (27% vs. 10%, z = 4.09, p [is less than] .001, two-tailed, effect size r = .22). The mean effect size for all nine reactions, including shame (r = .02) and guilt (r = -.05), was r = .31, which is consistent with sex differences on self-reported effects. The self-reported effects data and the reaction data from four national samples thus consistently indicate that males and females do not respond to experiences classified as CSA equivalently.

Measures of Psychological and Sexual Adjustment

Bigler's (1992) U.S. sample. In Bigler's (1992) mail survey, 314 respondents (140 men and 174 women) provided sufficient data to be retained for analysis. The SA and control males did not differ in sexual self-esteem (effect size r = -.01), but SA and control females did, with a small effect size (r = .10). Bigler speculated that this gender difference may have been because females and males are socialized differently and because females (56%) were more likely than males (30%) to have been forced or coerced into their CSA experience in this sample. Both SA males and females showed more sexual dysfunction than did control respondents, with effect sizes that were small to small-medium (r = .15 for males and r = .24 for females). The study-level effect sizes, based on averaging the two effect sizes for males and females separately using Fisher z transformations, were r = .07 for males and r = .17 for females.

Bigler provided additional information on the distributions of SA males and females for sexual dysfunction. Distributions were slightly positively skewed, with means and medians within the nonproblematic range. The sexual dysfunction scale was scored from 1 to 9, where scores of 5 and higher indicated sexual difficulty. Mean scores were 3.63 for SA males and 3.25 for SA females. Bigler used the Impact of Event Scale to assess the level of trauma associated with CSA; the trauma scale runs from 15 (no trauma) to 60 (extreme trauma). Bigler found that the distributions for both males and females with CSA experiences were strongly positively skewed, with means close to the nontraumatic end of the scale (20 for males and 21 for females) and medians at the nontraumatic end (15 for males and 16 for females). These results indicate a high concentration of SA individuals with nontraumatic reactions and a low concentration of SA individuals with traumatic reactions.

Boney-McCoy and Finkelhor's (1995) U.S. sample. Based on telephone interviews with children and adolescents 10 to 16 years of age, Boney-McCoy and Finkelhor (1995) found that respondents who reported unwanted sexual experiences with peers or older persons showed poorer adjustment compared to controls in all four areas measured: PTSD-related symptomatology, sadness, lifetime depression, and trouble with teachers. Although Boney-McCoy and Finkelhor reported adjusted means from an ANCOVA, in which a number of background factors were statistically held constant, we used the unadjusted means to compute effect sizes. We did this because effect sizes from all other studies examined in the current review were based on unadjusted means. For boys, the effect sizes for the four measures just listed were, respectively, r = .12, .09, .16, and .11, For girls, the effect sizes were, respectively, r = .26, .13, .20, and .20, with a study-level effect size of r = .20. All these results were statistically significant and remained so after statistical control. The background factors held constant were child's age, whether child lived with both parents, highest level of education by child's head of household, size of town child lived in, child's race, and a parent-child index, tapping non-abuse-related family dynamics. Boney-McCoy and Finkelhor argued that the results of the statistical control provide more confidence in the claim that CSA causes harm.

Laumann et al.'s (1994) U.S. sample. We computed effect sizes for the items judged to be relevant to current adjustment, as well as those relevant only to sexual activity with no implication for adjustment. Table 3 presents the items along with their effect sizes computed separately for males and females. Effect sizes with asterisks indicate a significant difference between the control and sexually touched groups. The .0025 level of significance was used because of the large number of variables (the Bonferroni correction was p = .05/20 = .0025); one-tailed tests were used to test the hypothesis that proportionately more sexually touched than control respondents would evidence sexual problems or activity. The table shows that males who were sexually touched when prepubescent by older individuals had more problems than control males in 4 of 11 problem areas; sexually touched females had more problems than control females in 4 of 10 areas. The size of these differences was small in all cases. For males, effect sizes ranged from r =.01 to .12; for females, the range was from r = -.12 to .14. Study-level effect sizes were r = .07 for males and r = .05 for females. Laumann et al. (1994) cautioned that the finding regarding more negative adjustment for sexually touched persons needs to be interpreted carefully because "[a]t no point is there any evidence that a majority of those with childhood sexual encounters [with older persons] have problems in their adult sexual lives. The differences, although consistent, are modest" (p. 346).

The bottom section of Table 3 provides the items and their effect sizes comparing sexually touched versus control respondents on different types and levels of sexual activity (e.g., more than 10 sex partners ever; ever had group sex; masturbated last year). For most items for both males (5 of 7 items) and females (6 of 7 items), significant differences emerged, with sexually touched individuals consistently experiencing more varied sexual activities at greater levels. For males, effect sizes ranged from r = .04 to .13; for females, these effect sizes ranged from r = .06 to. 17. The study level effect sizes for sexual activity were r = .10 for males and r = .12 for females. Because sexual activity and CSA were confounded, Laumann et al. cautioned against attributing the somewhat poorer adjustment in sexually touched respondents to their CSA.

Table 3

Effect Sizes for Items Measuring Adjustment and Sexual Activity from
Laumann et al. (1994)
                                           Men             Women
                                           (N = 1311)   (N = 1608)
Item                                       [r.sup.a]    [r.sup.a]

General Adjustment
  Not healthy last year                        .01         .02
  Not happy last year                          .12(*)      .07
Sexuality-Related Problems
  Lacked interest in sex, last year            .05         .06
  Unable to experience orgasm, last year       .06         .10(*)
  Experienced orgasm too soon, last year       .05        -.01
  Experienced pain during sex, last year       .05        -.12
  Sex was not pleasurable, last year           .03         .12(*)
  Was anxious about sexual
   performance, last year                      .10(*)      .06
  Experienced difficulty
   lubricating, last year                       --         .11(*)
  Emotional problems interfered
   with sex, last year                         .11(*)      .14(*)
  Erection difficulties, last year             .10(*)       --
  Ever forced a woman sexually                 .07          --
Sexual Activity
  > 10 sex partners ever                       .12(*)      .17(*)
  Ever performed oral sex                      .06         .12(*)
  Ever received oral sex                       .04         .12(*)
  Ever had anal sex with
    member of opposite gender                  .11(*)      .13(*)
  Ever had group sex                           .12(*)      .12(*)
  Often think about sex, last year             .13(*)      .06
  Masturbated in last year                     .11(*)      .12(*)

(a) Positive effect size (r) indicates proportionately more sexually touched respondents reporting adjustment problems or greater sexual activity than control respondents.

(*) p < .0025, one-tailed (in direction of higher proportion for sexually touched respondents)

Lopez et al.'s (1995) Spanish sample. In Lopez et al.'s (1995) study conducted on a national Spanish sample, Lopez et al. found that SA respondents, as a group, were more poorly adjusted than controls but noted that additional analyses revealed that this poorer adjustment applied only to female SA versus control respondents. No difference in adjustment was found between male SA and control respondents. Using their data, we calculated effect sizes for males and females, yielding a very small effect size for males (r = .04) and a small effect size for females (r = .09).

Los Angeles Times telephone survey of a U.S. sample. Finkelhor et al. (1989) divided respondents into three groups: victims with intercourse, victims with no intercourse, and nonvictims. They found significant differences in adjustment for both males and females, in which victims reported more marital disruption and a lower level of satisfaction with their current heterosexual relationships. We contrasted the two victim groups with the nonvictim group and obtained small effect sizes for both measures for both sexes. For males, effect sizes were r = .07 for marital disruption and r = .06 for sexual satisfaction; for females, effect sizes were r = .13 for marital disruption and r = .08 for sexual satisfaction. Finkelhor et al. acknowledged these small differences by noting that the discussion of long-term impact needs to be tempered wit caution because "[w]hat has often been neglected ... is that many victims do not suffer ... impairments" (p. 395). In this national sample, most victims did not suffer impairments. Finkelhor et al. also conducted analyses on these data to control for a variety of background factors (e.g., age, race, father's education, single parent or step-parent family, happiness of family life) and found that satisfaction with sexual relationships for men became nonsignificant. They argued that, because the other differences remained significant, the assertion that CSA causes harm was strengthened.

For the attitudinal items, we calculated effect sizes separately for males and females. If the actual difference in proportions between SA and control respondents was in the direction expected by the judges, the effect size was assigned a positive sign. If the actual difference was in the opposite direction from the judges' expectations, then the effect size was assigned a negative sign. Table 4 contains the set of 43 valid items, along with their effect sizes for male and female respondents. Effect sizes with asterisks indicate a statistically significant negative impact of CSA. The .0012 level of significance (one-tailed) was used because of the large number of items (the Bonferroni correction was p = .05/43 = .0012).

Table 4

Effect Sizes for Los Angeles Times Poll Attitude Items Judged
to be Valid Measures of Negative Impact of CSA (Taken from
Finkelhor et al., 1989)

                                                    Expected
Item                                               Difference (a)
Difficulty discussing SA                                +
Laws adequate                                           -
Fingerprint cheeks needed                               +
Most McMartin victims really abused                     +
More worried that innocent are tried                    -
More worried that guilty escape trial                   +
Attempted SA = SA                                       +
14-yr-old girl seduced by 18-yr-old boy = abuse         +
14-yr-old girl seduced by 22-yr-old man = abuse         +
14-yr-old boy seduced by 18-yr-old girl = abuse         +
14-yr-old boy seduced by 22-yr-old woman = abuse        +
14-yr-old boy forcing 14-yr-old girl = abuse            +
14-yr-old girl seducing 24-yr-old man = abuse           +
14-yr-old girl consenting to 24-yr-old man = abuse      +
14-yr-old prostitute with 22-yr-old man = abuse         +
Indecent suggestions to child = abuse                   +
Naked photos of child = abuse                           +
Exposing to child = abuse                               +
10-yr-old victim: A lot/very great harm                 +
SA has great permanent effect                           +
Worst damage mental                                     +
Community has crime problem                             +
Community has sex crime problem                         +
Main reason for abusers = mentally ill                  +
Main reason children submit
   = fear                                               +
   = force                                              +
   = no one to help                                     +
   = powerful abuser, must obey                         +
Best solution
   = more control of media                              +
   = less permissiveness                                +
   = more police                                        +
   = more public agencies                               +
   = heavier punishment of abuser                       +
In incest, priority
   = remove child                                       +
   = remove father                                      +
   = keep family together
Proper punishment for abuse
   = death                                              +
   = 20+ years                                          +
   = 6 months to 5 years                                -
   = probation                                          -
   = punishment not appropriate                         -
Rehabilitation of abusers likely                        -
Victims not responsible                                 +

Item
Difficulty discussing SA                                -.02
Laws adequate                                           -.02
Fingerprint cheeks needed                               -.01
Most McMartin victims really abused                      .00
More worried that innocent are tried                     .00
More worried that guilty escape trial                    .00
Attempted SA = SA                                       -.01
14-yr-old girl seduced by 18-yr-old boy = abuse          .03
14-yr-old girl seduced by 22-yr-old man = abuse          .00
14-yr-old boy seduced by 18-yr-old girl = abuse          .05
14-yr-old boy seduced by 22-yr-old woman = abuse         .02
14-yr-old boy forcing 14-yr-old girl = abuse             .02
14-yr-old girl seducing 24-yr-old man = abuse            .05
14-yr-old girl consenting to 24-yr-old man = abuse       .03
14-yr-old prostitute with 22-yr-old man = abuse          .07
Indecent suggestions to child = abuse                    .00
Naked photos of child = abuse                            .02
Exposing to child = abuse                                .05
10-yr-old victim: A lot/very great harm                  .02
SA has great permanent effect                            .02
Worst damage mental                                     -.05
Community has crime problem                              .11(*)
Community has sex crime problem                          .07
Main reason for abusers = mentally ill                  -.07
Main reason children submit
   = fear                                               -.06
   = force                                              -.01
   = no one to help                                     -.03
   = powerful abuser, must obey                          .05
Best solution
   = more control of media                               .02
   = less permissiveness                                 .00
   = more police                                        -.03
   = more public agencies                                .04
   = heavier punishment of abuser                       -.04
In incest, priority
   = remove child                                        .08
   = remove father                                      -.05
   = keep family together                               -.02
Proper punishment for abuse
   = death                                               .01
   = 20+ years                                          -.01
   = 6 months to 5 years                                 .04
   = probation                                           .00
   = punishment not appropriate                         -.08
Rehabilitation of abusers likely                        -.01
Victims not responsible                                 -.02

Item
Difficulty discussing SA                                -.02
Laws adequate                                            .06
Fingerprint cheeks needed                                .00
Most McMartin victims really abused                      .10(*)
More worried that innocent are tried                     .01
More worried that guilty escape trial                    .01
Attempted SA = SA                                        .02
14-yr-old girl seduced by 18-yr-old boy = abuse         -.05
14-yr-old girl seduced by 22-yr-old man = abuse         -.03
14-yr-old boy seduced by 18-yr-old girl = abuse         -.03
14-yr-old boy seduced by 22-yr-old woman = abuse         .03
14-yr-old boy forcing 14-yr-old girl = abuse             .03
14-yr-old girl seducing 24-yr-old man = abuse            .03
14-yr-old girl consenting to 24-yr-old man = abuse       .03
14-yr-old prostitute with 22-yr-old man = abuse          .06
Indecent suggestions to child = abuse                    .04
Naked photos of child = abuse                            .05
Exposing to child = abuse                                .01
10-yr-old victim: A lot/very great harm                  .01
SA has great permanent effect                           -.05
Worst damage mental                                      .05
Community has crime problem                              .03
Community has sex crime problem                          .10(*)
Main reason for abusers = mentally ill                  -.02
Main reason children submit
   = fear                                               -.05
   = force                                              -.03
   = no one to help                                      .02
   = powerful abuser, must obey                          .08(*)
Best solution
   = more control of media                              -.03
   = less permissiveness                                 .02
   = more police                                        -.03
   = more public agencies                                .08(*)
   = heavier punishment of abuser                        .00
In incest, priority
   = remove child                                        .03
   = remove father                                      -.04
   = keep family together                               -.02
Proper punishment for abuse
   = death                                               .04
   = 20+ years                                          -.06
   = 6 months to 5 years                                -.01
   = probation                                           .03
   = punishment not appropriate                         -.03
Rehabilitation of abusers likely                         .01
Victims not responsible                                  .06

(a) Plus sign (+) indicates greater proportion and negative sign (-) indicates smaller proportion of SA than control respondents were expected to endorse item, assuming that SA respondents were harmed by their CSA experience.

(b) A positive effect size (r) indicates difference in proportions between SA and control respondents was in the expected direction; a negative sign indicates proportions were in the opposite direction.

(*) p < .0012, one-tailed

The effect sizes were consistently small for both males and females. For males, they ranged from r = -.08 to .11 (mean r = .006). For females, they ranged from r = -.06 to. 10 (mean r = .013). Only 1 out of 43 items was significant in the case of males, and only 4 out of 43 items were significant in the case of females. Because attitudes are feelings people have about themselves, other people, objects, or issues (Petty & Cacioppo, 1986), it would be expected that attitudes should be affected by traumatizing events that produce long-term harm. Although this relation undoubtedly occurs for some SA persons, the results of our analysis indicate that, as a group, this relation was negligible in the LA Times national sample. We calculated the study-level effect sizes for males and females separately, computing the mean of the effect sizes for marital dissatisfaction, sexual satisfaction, and the mean attitudes. The study-level effect size was r = .05 for males and r = .07 for females.

Synthesis of adjustment measures in the five samples. As a final step in examining the various measures of adjustment across the five national samples, we conducted two meta-analyses of these results, one for male respondents and the other for female respondents. Table 5 summarizes the five studies, showing the study-level effect sizes and sample sizes used to determine the unbiased effect size estimates for males and females separately.

Table 5

Study-Level Effect Sizes of Measures of Adjustment from Studies
Using National Probability Samples to Examine Correlates of
Child Sexual Abuse

                                     Males       Females
Study                              N      r       N      r
Bigler (1992)                      140   .07     174    .17
Boney-McCoy & Finkelhor (1995)     987   .12     911    .20
Finkelhor et al. (1989)           1142   .05    1476    .07
Laumann et al. (1994)             1311   .07    1608    .05
Lopez et al. (1995)                462   .04     433    .09

Note: Study-level effect sizes for each study were computed as means of all adjustment measures in that study.

Table 6 shows the results of the meta-analyses. For males, the effect sizes were homogeneous, [chi square](4) = 3.63, p [is greater than].40. Combining the effect sizes yielded an unbiased effect size estimate of [r.sub.u] = .07, which was statistically significantly different from zero, z = 4.56,p [is less than].001, two-tailed. The 95% confidence interval ran from r = .04 to .10. For females, the effect sizes were not homogeneous, [chi square](4) = 15.01, p [is less than].005. The unbiased effect size estimate was [r.sub.u] = .10, which was statistically significantly different from zero, z = 6.51, p [is less than] .001, two-tailed. The 95% confidence interval ran from r = .07 to .12. We compared the male and female effect sizes using a weighted contrast analysis, as described by Rosenthal (1984). SA females compared to their controls showed somewhat poorer adjustment than SA males did compared to their controls, although this difference was not statistically significant, z = 1.12, p [is greater than] .20, two-tailed.

Table 6

Meta-Analyses of Psychological Correlates of CSA by Gender
from National Probability Samples

                                        95% Confidence
                                        Interval for
Gender          k      N     [r.sub.u]    [r.sub.u]       H

Males           5      4042     .07     .04 to .10      3.63
Females         5      4602     .10     .07 to .12     15.01(*)

Note: k represents the number of effect sizes (samples); N is the total number of participants in the k samples; [r.sub.u] is the unbiased effect size estimate (positive values indicate better adjustment for control subjects); H is the within-group homogeneity statistic (Chi-square).

(*) p < .05 in the Chi-square test

Because the female effect sizes were not homogeneous, we examined study attributes in an attempt to account for the heterogeneity. The effect size from the Boney-McCoy and Finkelhor (1995) study was substantially larger than most other effect sizes for females--a similar discrepancy obtained in the case of the male effect sizes. These researchers asked their respondents only about unwanted CSA, in contrast to the other four studies, in which respondents were questioned about CSA events that could have been unwanted or willing. Rind (1996) showed in a meta-analysis of college samples that willingness of CSA participation moderated a variety of psychological correlates. Therefore, we examined whether willingness of participation would significantly account for the variance in effect sizes for females in the national samples. To examine the moderating effect of willingness of participation, we used categorical model testing (Hedges & Olkin, 1985). In categorical model testing, a categorical moderator variable adequately accounts for the heterogeneity in effect sizes when there is (a) heterogeneity between categories (i.e., the mean effect sizes of the categories defined by the moderator variable differ significantly), but (b) homogeneity within categories (i.e., effect sizes within each category are homogeneous). Results of the categorical model testing were that (a) the mean effect sizes of the unwanted and all types of participation categories were significantly different, [chi square](1) = 14.38, p [is less than] .001, and (b) the effect sizes within each category were homogeneous, [chi square](3) = 2.62, p [is greater than] .45, for the all types category. These findings indicate a moderating influence of respondents' level of participation. The unbiased effect size estimate for female respondents for the all types of participation category was [r.sub.u] = .07, z = 4.27, p [is less than] .001, with a 95% confidence interval of r = .04 to .10. For comparison's sake, we compared male effect sizes between the two categories of level of participation and found that the categories were marginally heterogeneous, [chi square](1) = 3.08, p [is less than] .08. Excluding the Boney-McCoy and Finkelhor result yielded an effect size estimate of [r.sub.u] = .06, z = 3. 10, p [is less than] .001, with a 95% confidence interval of r = .02 to .09.

The results of the meta-analyses show that, in the national samples that included measures of psychological or sexual adjustment, CSA was related to adjustment problems for both genders, but the magnitudes of these relations were small. Therefore, if these CSA-adjustment relations reflected effects, rather than just correlates, of CSA, then the effects were on average of low intensity.

Discussion

Previous literature reviews examining the psychological correlates of CSA have generally (a) relied mostly or exclusively on data coming from clinical or legal samples to draw conclusions about the "effects" of CSA on later adjustment and (b) been narrative and qualitative. Clinical and legal samples are problematic because they are not representative of the general population, and because findings from these samples are vulnerable to information and selection biases that threaten their validity (Pope & Hudson, 1995). Conclusions based on narrative and qualitative reviews are vulnerable to subjectivity and imprecision (Jumper, 1995). Therefore, the conclusions and implications presented in most of these literature reviews cannot be accepted as scientifically valid for the entire population of persons with experiences that are labeled as CSA without further scrutiny. These conclusions and implications are that, in the general population, (a) CSA causes psychological harm, (b) this harm occurs in most persons with this experience, (c) the harm is typically intense, and (d) male and female experiences of CSA are equivalent. To evaluate these basic conclusions and implications about population characteristics of CSA, we examined only studies that used national probability samples, because these samples are most appropriate for addressing questions about population characteristics. Further, we evaluated the results from these studies meta-analytically, in the attempt to reduce subjectivity and strengthen precision.

We located seven studies using national probability samples that contained data concerning correlates or effects of CSA. Four of these studies were based on U.S. samples, and one each was based on a Canadian, British, and Spanish sample. Not including two studies in which researchers used overly broad definitions of CSA, we found that prevalence rates for CSA were approximately 11% for males and 19% for females. In the following sections, we evaluate the basic conclusions and implications presented in previous literature reviews by reviewing the findings from the national samples.

Gender Equivalence

A number of national studies presented data on respondents' self-reported effects from or reactions to their CSA experiences. A consistent finding across these studies was the greater proportion of negative responses by females. Whereas a majority of females (about two thirds) reported negative effects, only a minority of males (about two in five) did so. Across the three studies that provided self-reported effects data, the size of this sex difference was small to medium. Data on reactions at the time of the CSA experience, presented in a fourth study Upez et al., 1995), further indicated that the CSA experiences of males and females are not equivalent, with a greater proportion of females reacting negatively. This sex difference was medium size. The contrast analysis of CSA-adjustment relations showed a nonsignificant trend in the direction of poorer adjustment for SA females compared to SA males. The more reliable difference between SA females and males in self-reported effects or reactions compared with current adjustment is not contradictory, because effects or reactions may be temporary and therefore may not be reflected in measures of current adjustment, and measures of current adjustment assess correlates of CSA, which may or may not reflect effects.

Results from several studies reviewed suggest possible reasons for the consistent differences in self-reported effects and reactions. Bigler (1992) found that fewer than a third of his male respondents reported being coerced in their CSA experience, whereas more than half of his female respondents reported being coerced. Baker and Duncan (1985) found that female CSA experiences were three to four times more likely to be incestuous than male CSA experiences. They also found that females were younger than males on average when they experienced CSA. These findings imply that, in the population of persons with a history of experiences classifiable as CSA, males and females have different responses to these experiences because they typically do not have the same type of experience. The finding of nonequivalence of male and female CSA experiences in the national samples matches findings in college samples in which only one third of SA males reports negative reactions compared to two thirds of SA females (Rind, 1995). This finding of nonequivalence, however, contradicts clinical findings, upon which most researchers have based their conclusions.

Pervasiveness and Intensity

The self-reported effects data contradict the conclusions or implications presented in previous literature reviews that harmful effects stemming from CSA are pervasive and intense in the population of persons with this experience. Baker and Duncan (1985) found that, although some respondents reported permanent harm stemming from their CSA experiences (4% of males and 13% of females), the overwhelming majority did not (96% of males and 87% of females). Severe or intense harm would be expected to linger into adulthood, but this did not occur for most respondents in this national sample, according to their self-reports, contradicting the conclusion or implication of intense harm stemming from CSA in the typical case.

Meta-analyses of CSA-adjustment relations from the five national studies that reported results of adjustment measures revealed a consistent pattern: SA respondents were less well adjusted than control respondents. Importantly, however, the size of this difference (i.e., effect size) was consistently small in the case of both males and females. The unbiased effect size estimate for males and females combined was [r.sub.u] = .08, which indicates that CSA, assuming that it was responsible for the adjustment difference between SA and control respondents, did not produce intense problems on average. This effect size estimate is comparable to the mean effect size estimate for college students ([r.sub.u] = .07) in Jumper's (1995) meta-analysis, but it is considerably smaller than the effect size estimates for clinical samples in Jumper's (1995) and Neumann et al.'s (1996) meta-analyses, which were [r.sub.u] = .31 and d = .50 (roughly equivalent to [r.sub.u] = .30), respectively. Thus, whereas CSA accounted for only about one half of 1% of the adjustment variance for respondents in the national and college samples, CSA accounted for nearly 10% of the adjustment variance in the clinical samples. This substantial difference implies that clinical samples cannot be used to make valid estimates of psychological correlates of CSA in the general population. The results of the current meta-analyses and those of previous meta-analyses (Jumper, 1995; Neumann et al., 1996) taken together imply that clinical samples overestimate the adjustment variance accounted for by CSA in the general population by a factor of 15 and thereby substantially exaggerate the intensity of CSA correlates in the general population. Researchers' reliance on clinical samples in most previous literature reviews to estimate CSA-adjustment relations is therefore problematic.

The effect size estimates obtained from the national studies ([r.sub.u] = .07 for males and [r.sub.u] = . 10 for females) cannot be interpreted as estimates of the effects of CSA on adjustment for several reasons. First, these values estimate correlates, of which only a portion is likely to be effects. Therefore, the actual effects of CSA on adjustment in the general population are likely to be lower than the obtained effect size estimates, implying that the intensity of CSA effects in the population of persons who have experienced CSA is of very small magnitude. Second, the effect sizes from Boney-McCoy and Finkelhor's (1995) study should not be combined with the effect sizes from the other studies because these researchers restricted the SA population to persons with unwanted experiences. If the goal is to estimate CSA psychological correlates in the population, then the sample of SA respondents obtained must accurate reflect the population of SA persons, which Boney-McCoy and Finkelhor's sample did not. Thus, the lower unbiased effect size estimates obtained without Boney-McCoy and Finkelhor's effect sizes ([r.sub.u] = .06 for males and [r.sub.u] = .07 for females) are likely to be more precise population estimates.

Bigler (1992) provided information on the distributions of his participants' responses, which are relevant to examining whether CSA produces intense and pervasive effects. The distributions were positively skewed for both sexual dysfunction and level of trauma, with most scores falling in the nonproblematic range. This information suggests that, assuming a causal role for CSA affects a small portion of individuals in an intensely negative way but as a smaller negative effect, if any, on most individuals. This finding is consistent with Baker and Duncan's (1985) data on self-reported effects, in which only a small minority of SA individuals reported permanent harm. Clinical and legal samples are the sample types that are mostly likely to consist of these extreme negative cases (Okami, 1991). This bias highlights the problem of focusing on clinical and legal samples to describe the correlates CSA--a focus that has characterized nearly all previous literature reviews.

Causality

The results of several studies reviewed in this article are relevant to the issue of the role of CSA in causing harm. Laumann et al. (1994) found that, although SA respondents consistently reported more sexuality-related problems than control respondents, they were also consistently more sexually active and had consistently experienced more diverse types of sex than control respondents. This confounding variable of extent and type of sexual experience renders the difference in adjustment between the SA and control respondents uninterpretable in terms of cause and effect. Laumann et al. argued that greater sexual activity may have subjected SA respondents to more untoward outcomes, which may then have led to poorer adjustment.

Two studies used statistical control in assessing the relationship between CSA and current adjustment in an attempt to overcome problems of confounding variables (Boney-McCoy & Finkelhor, 1995; Finkelhor et al., 1989). Boney-McCoy and Finkelhor controlled for a variety of demographic variables as well as an index of parent-child relations and found that CSA respondents remained statistically significantly more poorly adjusted than controls. They argued that this result strengthens the assertion that CSA, and not some other variable, caused the poorer adjustment.

However, several considerations suggest that their control variables may not have been adequate to support this assertion. First, demographic variables have been found not to be important correlates of CSA in other national probability samples (Ageton, 1988; Laumann et al., 1994). Second, the parent-child index excluded nonsexual abusive experiences, which may have weakened its effect in statistical control. This possibility is suggested by recent research that has shown that CSA often occurs along with physical abuse or emotional neglect (Ney et al., 1994), and that when these other experiences are held constant, CSA-symptom relations frequently disappear (e.g., Cole, 1987; Eckenrode, Laird, & Doris, 1993; Higgins & McCabe, 1994; Ney et al., 1994; Wisniewski, 1989). For example, Wisniewski (1989), who used path analysis to examine CSA-adjustment relations among female students from 32 U.S. colleges and universities, chosen to be representative of all institutions of higher learning in the U.S., concluded that the "data do not support child sexual abuse as a specific explanation of current emotional distress. The data are best interpreted as supportive of other factors such as family violence ... as having the greatest impact on current emotional adjustment" (p. 258). Finkelhor et al.'s (1989) control variables were similar to those of Boney-McCoy and Finkelhor (1995), and hence their conclusion that a causal link between CSA and adjustment was supported because most CSA-adjustment relations remained statistically significant after statistical control is subject to the same criticisms.

The findings from another study using a national sample, not included in the current review because it did not report psychological correlates of CSA and because of its exclusive focus on female respondents, are particularly relevant for interpreting causation in national samples (Ageton, 1985, 1988). Ageton reported the results of longitudinal data collected on U.S. respondents aged 11 to 17 at the beginning of the study and found that girls experiencing unwanted sexual contacts differed from those who did not in terms of family, school, and social factors. The family environments of SA girls compared with control girls were characterized by more disruption and instability. SA girls had substantially more involvement with delinquent friends and received more support from these peers for delinquent and antisocial acts. SA girls also displayed far more tolerant attitudes toward delinquent behavior. Compared with controls, SA girls were less well integrated into their families and schools. We computed effect sizes for these measures for the last year for which data were taken to obtain estimates of the magnitude of confounding between CSA and social environment. Effect sizes ranged from small-medium to large. Some examples were these: disruptive events in home (r = .21), school-related delinquency (r = .30), exposure to delinquent peers (r = .48), support from delinquent peers for deviance (r = .30), delinquent behavior (r = .30), peer pressure for drinking and drugs (r = .47), attitudes toward deviance (r = .28), and nonsexual victimization (r = .34). These results suggest that the small differences in adjustment between CSA and control respondents observed in other national samples may be attributable to larger differences in social environment, rather than to the CSA.

To address the issue of whether CSA was the result or cause of negative social factors, Ageton (1988) analyzed the longitudinal data prospectively. She investigated whether any social factors measured in earlier years of the study were predictive of CSA that occurred in later years of the study and found that a number of factors were predictive. We computed effect sizes for the significant predictors, averaged them over the two years reported for each predictor, and obtained the following: family normlessness (r = .18), school normlessness (r = .15), peer support for delinquent behavior (r = .22), exposure to delinquent peers (r = .25), and attitudes toward deviance (r = .22). These results, with effect sizes of small to medium magnitude, imply that negative social factors predispose children and adolescents to CSA, rather than the reverse. These results are consistent with the possibility that negative social factors lead to poorer adjustment and to CSA and that the relation between CSA and adjustment in the general population is either spurious or of lower magnitude than we estimated.

Qualifications

Several qualifications of the findings from the national probability samples need to be discussed. The first concern involves the measures of effects used in these studies. It may be argued that self-reports of effects of CSA are not valid in terms of describing actual effects. For example, it may be argued that persons with CSA experiences are sometimes not aware of the negative impact that these experiences have had on their current adjustment, leading them to underreport this impact. This argument is consistent with the recovered memory view (see Loftus & Ketchum, 1994).

Although these false negatives may occur in some cases, their occurrence may be counterbalanced by the occurrence of false positives in other cases. Persons with current adjustment problems who attribute these problems to their experiences may make invalid causal inferences because of their "effort after meaning," in which they have been motivated to explain their current difficulties and have chosen their CSA experience as the causative factor because of the salience of CSA in our culture at the present time as an explanation for maladjustment (Pope & Hudson, 1995). Thus, false negatives and false positives may both occur, with the result that overall bias in measurement may be small and may not necessarily be in the direction of underestimation of negative effects.

Self-reports of harm stemming from CSA experiences are generally accepted by child abuse researchers as valid indicators of harm. This being the case, it seems inappropriate to dismiss self-reports of no harm. Baurmann (1983), who examined psychological correlates of CSA using the largest sample ever employed in CSA research of children and youths who experienced CSA, argued for the inclusion of self-reports of effects. One basis for his argument was that self-reported effects corresponded well to standardized measures of adjustment in his study.

Another possible shortcoming of the studies reviewed in the current article is that a number of these researchers did not use standardized measures of adjustment but instead used just a few common sense items. Although this criticism applies to two of the five studies with measures of adjustment (Finkelhor et al., 1989; Laumann et al., 1994), it does not apply to the other three (Bigler, 1992; Boney-McCoy & Finkelhor, 1995; Lopez et al., 1995). Bigler and Lopez et al. used standardized measures; Boney-McCoy and Finkelhor constructed a measure of PTSD, but its items were taken and modified from the well-established SCL-90-R. The results across studies were fairly consistent, regardless of measurement instrument, suggesting the validity of the nonstandardized measures.

Another criticism that might be made is that the measures employed were not sensitive to the type of harm that is assumed to result from CSA, because CSA is assumed to have particular effects (Briere & Elliot, 1993). Nash, Hulsey, Sexton, Harralson, and Lambert (1993b, p. 290) argued, however, that because "almost every conceivable form of pathology" has been attributed to CSA, any given measure of adjustment should be sufficiently sensitive. Furthermore, the majority of studies using measures of adjustment in the current review specifically assessed sexual adjustment, which is one type of adjustment that has been frequently hypothesized by child abuse researchers to be vulnerable to CSA experiences.

Aside from measurement concerns, another possible shortcoming is response rate. Bigler (1992) showed the poorest response rate (33%), which casts doubt on the generalizability of his findings. In Lopez et al. (1995) study, although the initial response rate was 82%, only 49% of these respondents returned their questionnaire, which contained the adjustment measure. The validity of the results demonstrated, by the fact that SA and control respondents returned their questionnaires at approximately the same rate. The validity of Lopez et al.'s (1995) results, as well as those of Bigler (1992), is implied by the consistency of these findings with those of the other studies in which response rates were quite high.

Finally, it is important to consider Boney-McCoy and Finkelhor's (1995) results more closely, because their effect sizes were larger than those in the other national studies. It is possible that their effect sizes were larger not because they asked only about unwanted CSA, in contrast to the other national studies in the meta-analysis, but because their participants were younger. The effects of CSA might have been more salient and thus detectable for their SA respondents because of the recency of the CSA; these effects may tend to diminish over time, resulting in the smaller effect sizes in the other national samples based on older respondents.

This possibility would suggest that CSA has effects that are more intense, at least initially, than is indicated in the current review. Several points, however, argue against this possibility. First, when age of participants is held constant (i.e., only college samples are used), unwanted CSA is associated with larger effect sizes (Rind, 1996). Second, respondents in Boney-McCoy and Finkelhor's sample who reported CSA were likely to have been unrepresentative of all respondents in their sample who experienced events classified by our society as CSA. Only 1% of boys and 5.6% of girls aged 16 or less reported contact CSA involving peers or older persons. In sharp contrast to this in their national sample that 12% of men and 17% of women reported contact CSA with older persons (peer sex was excluded) before they reached puberty. They found further that prevalence rates did not vary as a function of age cohort. These findings suggest that many respondents in Boney-McCoy and Finkelhor's sample who experienced contact CSA may not have reported it. Respondents who did report CSA may have tended to have had more negative experiences, resulting in parents or guardians being more likely to be aware of the CSA. This greater awareness, in turn, may have led these respondents to be more willing to report experiences of CSA. Respondents with CSA whose families were unaware of this experience may have been more reluctant to disclose it, given at their parents or guardians knew about study and may have been present or nearby while they answered questions over the telephone--respondents participated only after parents or guardians were informed about the study and gave permission. Therefore, CSA correlates may have been inflated by a bias in willingness to report the CSA. These considerations suggest that CSA does not on average have effects that are more intense, at least initially, than is indicated in the current review.

Summary and Conclusions

Our goal in the current study was to examine whether, in the population of persons with a history of CSA, this experience causes pervasive, intense psychological harm for both genders. Most previous literature reviews have favored this viewpoint. However, their conclusions have generally been based on clinical and legal samples, which are not representative of the general population. To address this viewpoint, we examined studies that used national probability samples, because these samples provide the best available estimate of population characteristics. Our review does not support the prevailing viewpoint. The self-reported effects data imply that only a small proportion of persons with CSA experiences are permanently harmed and that a substantially greater proportion of females than males perceive harm from these experiences. Results from psychological adjustment measures imply that, although CSA is related to poorer adjustment in the general population, the magnitude of this relation is small. Further, data on confounding variables imply that this small relation cannot safely be assumed to reflect causal effects of the CSA.

Browne and Finkelhor (1986, p. 178) cautioned "advocates not [to] exaggerate or overstate the intensity or inevitability of [CSA] consequences," because such exaggeration has iatrogenic potential. Despite this caution, child abuse researchers have tended depict CSA as a "special destroyer of adult mental health" (Seligman, 1994, p. 232). McMillen, Zuravin, and Rideout (1995, p. 1037) recently commented that the "experience of child sexual abuse is a traumatic event for which there may be few peers." Results of analyses of the national samples show that such characterizations are exaggerated at the population level. This exaggeration may stem from our culture's tendency to equate wrongfulness with harmfulness in sexual matters (Money, 1979). CSA is violative of norms and laws in our culture; these facts, however, do not imply its harmfulness in a scientific or psychological sense (Kilpatrick, 1987). It is important to add to this discussion of exaggeration that understatement is also problematic. CSA is potentially harmful for young persons because of their vulnerability to being misused. The current findings should thus not be interpreted by lay persons as condoning abusive behavior.

Finally, analysis at the population level may obscure characteristics of particular segments of the population. In the current review, the effect size estimate of the relation between CSA and adjustment, which was of low magnitude, cannot be interpreted as applicable to every case. When CSA is accompanied by particular dispositional or situational factors, including variables such as temperamental vulnerability, the use of force, or the presence of close familial ties between participants, then CSA might produce intense harm; on the other hand, if temperamental factors are favorable, if the child or adolescent perceives his or her participation to have been willing, or if the sexual experience is essentially trivial or transient, then harm may be absent (Constantine, 1981). Combining the former and latter types of experiences into one category labeled CSA is problematic, because both negative and neutral effects can become obscured. By moving beyond sociolegal definitions of CSA and employing strictly scientific definitions (cf. Ames & Houston, 1990; Rind & Bauserman, 1993), researchers can better describe psychological correlates of the heterogeneous collection of experiences currently labeled as CSA.

References

Ageton, S. S. (1985). A research report for adults who work with teenagers: Facts about sexual assault. National Institute of Mental Health (DHHS), Rockville, MD. Report no. DHHS-ADM-85-1398.

Ageton, S. S. (1988). Vulnerability to sexual assault. In A. W. Burgess (Ed.), Rape and sexual assault II (pp. 221-243). New York: Garland.

Alexander, P. C., & Lupfer, S. L. (1987). Family characteristics and long-term consequences associated with sexual abuse. Archives of Sexual Behavior, 16, 235-245.

Ames, M. A., & Houston, J. D. (1990). Legal, social, and biological definitions of pedophilia. Archives of Sexual Behavior, 19, 333-342.

Badgley, R., Allard, H., McCormick, N., Proudfood, P., Fortin, D., Rae-Grant, Q., Gelinas, P., Pepin, L., & Sutherland, S. (1984). Sexual offenses against children. Ottawa, Canada: Canadian Government Publishing Centre.

Baker, A. W., & Duncan, S. P., (1985). Child sexual abuse: A study of prevalence in Great Britain. Child Abuse & Neglect, 9, 457-467.

Baurmann, M. C. (1983). Sexualitat, Gewalt und Folgen. Wiesbaden: Bundeskriminalamt.

Beitchman, J. H., Zucker, K. J., Hood, J. E., DaCosta, G. A., & Akman, D. (1991). A review of the short-term effects of child sexual abuse. Child Abuse & Neglect, 15, 537-556.

Beitchman, J. H., Zucker, K. J., Hood, J. E., DaCosta, G. A., Akman, D., & Cassavia, E. (1992). A review of the long-term effects of child sexual abuse. Child Abuse & Neglect, 16, 101-118.

Bergdahl, J. B. (1982). Differences in demographic features and present anxiety level among college women who report type, extent and rated severity of child and adolescent sexual contacts with adults. Doctoral dissertation, Ohio University.

Bigler, M. O. (1992). Childhood sexual trauma as a predictor of adult sexual self-esteem and sexual dysfunction. Doctoral dissertation, New York University.

Black, C. A., & DeBlassie, R. R. (1993). Sexual abuse in male children and adolescents: Indicators, effects, and treatment. Adolescence, 28, 123-133.

Boney-McCoy, S., & Finkelhor, D. (1995). Psychosocial sequelae of violent victimization in a national youth sample. Journal of Consulting and Clinical Psychology, 63, 726-736.

Briere, J. N., & Elliot, D. M. (1993). Sexual abuse, family environment, and psychological symptoms: On the validity of statistical control. Journal of Consulting and Clinical Psychology, 61, 284-288.

Briere, J. N., & Elliot, D. M. (1994). Immediate and long-term impacts of child sexual abuse. The Future of Children, 4, 54-69.

Briere, J. N., & Runtz, M. (1988a). Symptomatology associated with childhood sexual victimization in a nonclinical adult sample. Child Abuse & Neglect, 12, 51-59.

Briere, J. N., & Runtz, M. (1988b). Multivariate correlates of childhood psychological and physical maltreatment among university women. Child Abuse & Neglect, 12, 331-341.

Briere, J. N., & Runtz, M. (1990). Differential adult symptomatology associated with three types of child abuse histories. Child Abuse & Neglect, 14, 357-364.

Briere, J. N., & Runtz, M. (1993). Childhood sexual abuse: Long-term sequelae and implications for psychological assessment. Journal of Interpersonal Violence, 8, 312-330.

Browne, A., & Finkelhor, D. (1986). Initial and long-term effects: A review of the research. In D. Finkelhor (Ed.), A sourcebook on child sexual abuse (pp. 143-179). London: Sage.

Cohen, J. (1988). Statistical power analyses for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.

Cole, C. B. (1987). The specificity of long-term effects of sexual abuse and factors mediating outcome: A comparison of sexually and physically abused young adults. Doctoral dissertation, University of Washington.

Collings, S. J. (1994). The long-term effects of contact and noncontact forms of child sexual abuse in a sample of university men. Child Abuse & Neglect, 19, 1-6.

Constantine, L. L. (1981). The effects of early sexual experience: A review and synthesis of research. In L. L. Constantine & F. M. Martinson (Eds.), Children and sex (pp. 217-244). Boston: Little, Brown.

Eckenrode, J., Laird, M., & Doris, J. (1993). School performance and disciplinary problems among abused and neglected children. Developmental Psychology, 29, 53-62.

Edwards, J. J., & Alexander, P. C. (1992). The contribution of family background to the long-term adjustment of women sexually abused as children. Journal of Interpersonal Violence, 7, 306-320.

Erel, O., & Burman, B. (1995). Interrelatedness of marital relations and parent-child relations: A meta-analytic review. Psychological Bulletin, 118, 108-132.

Finkelhor, D., Hotaling, G. T., Lewis, I. A., & Smith, C. (1989). Sexual abuse and its relationship to later sexual satisfaction, marital status, religion, and attitudes. Journal of Interpersonal Violence, 4, 379-399.

Fishman, J. D. (1990). Prevalence, impact, and meaning attribution of childhood sexual experiences of undergraduate males. Doctoral dissertation, University of Massachusetts.

Fritz, G., Stoll, K., & Wagner, N. N. (1981). A comparison of males and females who were sexually molested as children. Journal of Sex and Marital Therapy, 1, 54-60.

Fromuth, M. E. (1986). The relationship of childhood sexual abuse with later psychological and sexual adjustment in a sample of college women. Child Abuse and Neglect, 10, 5-15.

Glod, C. A. (1993). Long-term consequences of childhood physical and sexual abuse. Archives of Psychiatric Nursing, 7, 163-173.

Greenwald, J. P. (1993). Romantic and sexual self-esteem, a new construct: Its relationship to current and early childhood sexual experiences. Doctoral dissertation, University of Nebraska.

Harter, S., Alexander, P. C., & Neimeyer, R. A. (1988). Long-term effects of incestuous child abuse in college women: Social adjustment, social cognition, and family characteristics. Journal of Consulting and Clinical Psychology, 56, 5-8.

Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. San Diego, CA: Academic Press.

Higgins, D. J., & McCabe, M. P. (1994). The relationship of child sexual abuse and family violence to adult adjustment: Toward an integrated risk-sequelae model. The Journal of Sex Research, 31, 255-266.

Horowitz, M. J., Wilner, N., & Alvarez, W. (1979). Impact of event scale: A measure of subjective stress. Psychometric Medicine, 41, 209-218.

Jackson, J. L., Calhoun, K. S., Amick, A. E., Maddever, H. M., & Habif, V. L. (1990). Young adults women who report childhood intrafamilial sexual abuse: Subsequent adjustment. Archives of Sexual Behavior, 19, 211-221.

Jumper, S. (1995). meta-analysis of the relationship of child sexual abuse to adult psychological adjustment. Child Abuse & Neglect, 19, 715-728.

Kendall-Tackett, K. A., Williams, L. M., & Finkelhor, D. (1993). Impact of sexual abuse on children: A review and synthesis of recent empirical studies. Psychological Bulletin, 113, 164-180.

Kilpatrick, A. (1987). Childhood sexual experiences: Problems and issues in studying long-range effects. The Journal of Sex Research, 23, 173-196.

Kinzl, J., Traweger, C., Guenther, V., & Biebl, W. (1994). Family background and sexual abuse associated with eating disorders. American Journal of Psychiatry, 151, 1127-1131.

Laumann, E. O., Gagnon, J. H., Michael, R. T., & Michaels, S. (1994). The social organization of sexuality: Sexual practices in the United States. Chicago: The University of Chicago Press.

Loftus, E., & Ketchum, K. (1994). The myth of repressed memories. New York: St. Martin's Press.

Long, P., & Jackson, J. (1993). Initial emotional response to childhood sexual abuse: Emotion profiles of victims and relationship to later adjustment. Journal of Family Violence, 8, 167-181.

Lopez, F., Carpintero, E., Hernandez, A., Martin, M., & Fuertes, A. (1995). Prevalencia y consecuencias del abuso sexual al menor en Espana [Prevalence and consequences of sexual abuse of minors in Spain]. Child Abuse & Neglect, 19, 1039-1050.

McMillen, C., Zuravin, S., & Rideout, G. (1995). Perceived benefit from child sexual abuse. Journal of Consulting and Clinical Psychology, 63, 1037-1043.

Mendel, M. P. (1995). The male survivor. Thousand Oaks, CA: Sage.

Money, J. (1979). Sexual dictatorship, dissidence, and democracy. International Journal of Medicine and Law, 1, 11-20.

Nash, M. R., Hulsey, T. L., Sexton, M. C., Harralson, T. L., & Lambert, W. (1993a). Long-term sequelae of childhood sexual abuse: Perceived family environment, psychopathology, and dissociation. Journal of Consulting and Clinical Psychology, 61, 276-283.

Nash, M. R., Hulsey, T. L., Sexton, M. C., Harralson, T. L., & Lambert, W (1993b). Reply to comment by Briere and Elliot. Journal of Consulting and Clinical Psychology, 61, 289-290.

Neumann, D. A., Houskamp, B. M., Pollock, V E., & Briere, J. (1996). The long-term sequelae of childhood sexual abuse in women: A meta-analytic review. Maltreatment, 1, 6-16.

Ney, P. G., Fung, T., & Wickett, A. R. (1994). The worst combinations of child abuse and neglect. Child Abuse & Neglect, 18, 705-714.

Okami, P. (1990). Sociopolitical biases in the contemporary scientific literature on adult human sexual behavior with children and adolescents. In J. R. Feierman (Ed.), Pedophilia: Biosocial dimensions (pp. 91-121). New York: Springer-Verlag.

Okami, P. (1991). Self-reports of "positive" childhood and adolescent sexual contacts with older persons: An exploratory study. Archives of Sexual Behavior, 20, 437-457.

Pallotta, G. (1991). Intrafamilial and extrafamilial sexual abuse vulnerability factors and long-term psychological adjustment in a college population. Doctoral dissertation, West Virginia University.

Petty, R., & Cacioppo, J. T. (1986). Communication and persuasion. New York: Springer-Verlag.

Pizzolo, C. P. (1989). The relationship between childhood sexual abuse and symptoms of eating disorders among college women. Doctoral dissertation, State University of New York at Stony Brook.

Pope, H. G., & Hudson, J. I. (1995). Does childhood sexual abuse cause adult psychiatric disorders? Essentials of methodology. Journal of Psychiatry and Law, Fall, 363-381.

Rind, B. (1995). An analysis of human sexuality textbook coverage of the psychological correlates of adult-nonadult sex. The Journal of Sex Research, 32, 219-233.

Rind, B. (1996). Correlates of child sexual abuse: A meta-analytic review of results from studies based on nationally representative and college samples. Paper presented at the Western Region Annual Conference of the Society for the Scientific Study of Sexuality (SSSS) in San Diego, April.

Rind, B., & Bauserman, R. (1993). Biased terminology effects and biased information processing in research on adult-nonadult sexual interactions: An empirical investigation. The Journal of Sex Research, 30, 260-269.

Rosenthal, R. (1977). Biasing effects of experimenters. ETC: A Review of General Semantics, 34, 253-264.

Rosenthal, R. (1984). Meta-analytic procedures for social research. London: Sage.

Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bulletin, 118, 183-192.

Rosenthal, R., & Rosnow, R. (1991). Essentials of behavioral research: Methods and data analysis (2nd ed.). New York: McGraw-Hill.

Rush, F. (1980). The best kept secret. New York: Prentice Hall.

Rust, J., & Golombok, S. (1986). The GRISS: A psychometric instrument for the assessment of sexual dysfunction. Archives of Sexual Behavior, 15, 157-165.

Seligman, M. E. P. (1994). What you can change and what you can't. New York: Knopf.

Shadish, W. R., & Haddock, C. K. (1994). Combining estimates of effect size. In H. Cooper & L. V. Hedges (Eds.), The handbook of research synthesis (pp. 261-281). New York: Russell Sage Foundation.

Snell, W. E., & Papini, D. R. (1989). The sexuality scale: An instrument to measure sexual-esteem, sexual-depression, and sexual-preoccupation. The Journal of Sex Research, 26, 256-263.

Urquiza, A. J., & Capra, M. (1990). The impact of sexual abuse: Initial and long-term effects. In M. Hunter (Ed.), The sexually abused male: Prevalence, impact, and treatment (Vol. 1) (pp. 105-135). Lexington, MA: Lexington Books.

U.S. Bureau of the Census. (1995). Statistical abstract of the United States: 1995 (115th ed.). Washington, DC: Department of Commerce.

Watkins, B., & Bentovim, A. (1992). Male children and adolescents as victims: A review of the current knowledge. In G. C. Mezey & M. B. King (Eds.), Male victims of sexual assault (pp. 27-66). New York: Oxford University Press.

Wisniewski, N. (1989). A path analytic model of the aftereffects of childhood sexual victimization. Doctoral dissertation, Kent State University.

Yama, M. F., Tovey, S. L., & Fogas, B. S. (1993). Childhood family environment and sexual abuse as predictors of anxiety and depression in adult women. American Journal of Orthopsychiatry, 63, 136-141.

Manuscript accepted October 14, 1996

Correspondence should be sent to Bruce Rind, Ph.D., Department of Psychology, Temple University, Philadelphia, PA 19122. E-mail: [e-mail].