Measurement of women's child sexual abuse experiences: an empirical demonstration of the impact of choice of measure on estimates of incidence rates and of relationships with pathology.

Studying child sexual abuse (CSA) is a difficult task due to factors such as the sensitivity of the topic; the related legal, ethical, and practical restrictions; and the difficulty of determining the reliability and validity of data provided years after a sometimes traumatic childhood event. In general, there is little that researchers can do about these challenges. However, researchers have contributed to the problems of conducting research in this area and interpreting research results through the proliferation of methods used to define and measure CSA. For instance, reports of the incidence of CSA have ranged from 5% (Stein. Golding, Siegel, Burnam, & Sorenson, 1988) to 66% (Boyer & Fine, 1992) due primarily to variations in the measures used.

In addition to determining incidence rates, many investigators are interested in whether CSA is related to adult adjustment. Several studies have reported a relationship between CSA and adult adjustment among women, including depression (e.g., Ernst, Angst, & Foldenyi, 1993: Fromuth, 1986; Gidycz & Koss, 1989; Mullen, Martin. Anderson, Romans, & Herbison, 1993; Roesler & McKenzie, 1994; Saunders, Villeponteaux, Lipovsky, Kilpatrick, & Veronen, 1992; Stein et al., 1988). Despite this common finding, the strength and nature of the relationship between CSA and adjustment remains unclear, and studies have contradicted each other regarding whether CSA is related to adult mental health problems (Beitchman et al., 1992), in part due to the wide variety of measures used in researching this question. In particular, the broad variations in definitions of CSA used in research have made it difficult to interpret results or to compare one study to another (Beitchman, Zucker, Hood, DaCosta, & Akman, 1991; Beitchman et al., 1992; Briere, 1992; Haugaard & Emery, 1989).

There is accumulating evidence that the relationship between CSA and adult adjustment may not be a causal relationship but may be due to common background factors that place some women at risk for both sexual abuse and adjustment problems (e.g., Finkelhor, Hotaling, Lewis, & Smith, 1990; Higgins & McCabe, 1994; Rind & Tromovitch, 1997). The argument is that a variety of background factors, such as a hostile, nonsupportive family environment, greatly increase the likelihood that a child will be sexually abused as well as the likelihood that she will have adjustment problems as an adult. Thus, the relationship of CSA to adjustment may be spurious due to common risk factors. However, because of the variations in definitions of CSA, the nature of this relationship remains unclear.

Much of the argument that CSA may not be related to adult adjustment is based on research that defines CSA solely by the ages of the "partners" in a sexual event without regard for characteristics of the event (e.g., whether the female wanted the event to occur; whether force was threatened or used; whether the event involved penetration). Although such status definitions reflect legal definitions of sexual abuse with minors, there is little theoretical or empirical reason to expect desired or voluntary, sexual experiences during childhood or adolescence to be related to adult adjustment. Beitchman et al. (1992) argued that CSA may be related to adult adjustment only when force, threat of force, or penetration are involved, we empirically explored the implications of several measurement issues for estimates of the incidence of CSA and for understanding the relation of CSA to depression among young women. We considered only definitions of CSA that involve unwanted sexual events.

Sexual abuse has been operationalized in a wide variety of ways. Most commonly, researchers have dichotomized women into those who have been sexually abused, however defined, and those who have not. Other researchers have measured sexual abuse in ways that take into account the severity of the experience (e.g., differentiate between molestation and rape). In other studies, forced or unwanted sexual events with similar-aged peers have not been considered sexual abuse experiences. Finally, some studies have included noncontact experiences (e.g., being seen nude against one's wishes, exhibitionism) in their definitions, whereas others have not. Readers of the CSA literature are faced with the considerable challenge of trying to interpret and compare the results of studies using these and other very different measures of this variable.

In one of the simplest approaches (see Table 1, column 1), researchers ask women, often with a single question, whether they have had unwanted sexual experiences, such as touching or fondling, that made them feel uncomfortable (e.g., Alexander & Lupfer, 1987) or sexual touching or intercourse that involved force (e.g., Laumann, Gagnon, Michael, & Michaels, 1994; Stein et al., 1988). These and similar approaches then dichotomize respondents into those who were abused and those who were not abused. In these studies, women who had been touched against their will are categorized along with women who had been raped as victims of sexual abuse with no consideration For differences in the severity of their experiences.

Table 1. Examples of Variations in Definitions of Childhood Sexual Abuse Across Studies

                  Dichotomous

Any Sexual Abuse               Age of Perpetrator a
vs. None                       Factor

Alexander & Lupfer,            Ernst et al., 1993
1987                           Finkelhor. 1984
Bendixen, Muus, &              Fromuth, 1986
Schei, 1994                    Gold, 1986
Conte & Schuerman, 1987        Greenwald, Leitenberg.
Stem et al., 1988              Cado, & Tartan, 1990
                               Higgins & McCabe, 1994
                               Nash, Hulsey, Sexton,
                               Harralson & Lambert, 1993
                               Parker & Parker, 1991
                               Peters, 1988
                               Wyatt, 1985
                               Yama, Fogas, Teegarden,
                               & Hastings, 1993

                  Multilevel

Contact Molestation            Noncontact Molestation
to Rape                        to Rape

Baler, Rosenzweig,             Anderson, Martin
& Whipple, 1991                Mullen, Romans, &
Gidycz & Koss. 1989            Herbison, 1993
                               Boyer & Fine. 1992
                               Finkelhor et al., 1990
                               Mullen et al., 1993
                               Rew, Esparza, & Sands, 1991
                               Saunders et al., 1992

Another dichotomous approach to measuring CSA takes into account the victim's age, the perpetrator's age, or both (see Table 1, column 2). Although there are a number of variations within this status definition approach, CSA is typically defined as sexual events occurring before the female reaches a particular age and in which the male involved is an adult (over age 16 or 18) and/or a certain number of years older than the female (e.g., Briere & Runtz, 1993; Finkelhor, 1984; Parker & Parker, 1991). For instance, any sexual event between a female who is less than 12 or 13 (depending on the study) and a male who is 3 or 5 years older than the female (depending on the study) is considered CSA. Similarly, any sexual event involving a female older than 12 or 13 and under age 15, 16, or 18 (depending on the study) and a male who is 10 years older is considered CSA. These studies argue that young females lack the developmental ability or the legal standing to consent to sexual acts, or lack enough power in such age-discrepant relationships to refuse to consent. Neither the female's willingness to participate nor the male's use of force or coercion are considered in defining CSA in most of these studies.

Another purpose of taking the relative ages of males and females into consideration in measures of CSA is to prevent sexual experiences with peers from being considered sexual abuse. In most of these studies, sexual experiences with peers are excluded from definitions of sexual abuse even if they involve force. Thus, if there is a significant amount of sexual abuse within high school dating relationships (e.g., date rape) or other sexual abuse from similar-aged peers, studies that use only status definitions of CSA will underreport the incidence of CSA. Women who were sexually abused by a boyfriend or other similar-aged peer would not be considered sexually abused and would be included in the comparison group in these studies. Assuming that unwanted sexual experiences with peers are not free from trauma (Finkelhor, 1979), placing these women in comparison groups could reduce the likelihood of finding differences between abuse victims and nonvictims by inflating the trauma or mental health symptom scores for nonvictims.

Despite the popularity of this approach, we know of no empirical justification for the argument that sexual abuse by peers should not be considered when reporting incidence rates, nor that sexual abuse by peers would be any more or less related to women's adult adjustment than would sexual abuse by older persons. Finkelhor (1979) reported that the trauma of CSA was greatest when the perpetrator was 5-10 years older than the victim. However, Finkelhor also reported that trauma from sexual abuse was greatest for women who were victimized between ages 16 and 18, a time when sexual abuse during dating is possible for most females. Finkelhor did not present data on the relation of the relative ages of the victim and perpetrator to women's mental health. Kilpatrick (1992) reported that the relative ages of females (willing or otherwise) and their partners were not related to long-term outcomes. Thus, it remains an empirical question whether CSA by same-aged peers is related to adult adjustment.

While many studies have limited the experiences that might be considered CSA, other studies have expanded the definition to include noncontact forms of abuse (e.g., exposure to exhibitionism, exposure to pornography, being seen by a "peeping Tom," having nude photographs taken against one's will; e.g., Boyer & Fine, 1992; Finkelhor, 1984). By using perpetrator-victim relative age restrictions in dichotomous measures along with broader definitions of what experiences constitute sexual abuse, some studies have created very curious labeling situations. In these studies, women who were raped by older men--that is, men who met the absolute or relative age criteria for that study--are categorized along with women who had a nude picture taken against their wishes as victims of sexual abuse. In the same studies, women who were raped by similar-aged dates in early to mid-adolescence are classified as nonvictims because the perpetrator did not meet the relative age restriction. Although there is no estimate of how many women fall into these categories, the relation of CSA to adjustment would be weaker for studies using this approach because of these anomalies.

Koss and Oros (1982) were the first to develop a measure of sexual abuse that considers the severity of the experience (see Table 1, column 3). In studies that use their measure, participants can be categorized as having experienced no sexual abuse, contact molestation, coercion, attempted rape, or rape. From a statistical point of view, expanding the range of a measure from two levels (dichotomous) to five levels should increase the sensitivity of the measure and, therefore, increase the strength of its correlations with related variables (Cohen & Cohen, 1983). In addition, distinguishing between the varying levels of sexual abuse allows researchers to make more fine-grained analyses of the relationships between sexual abuse and various antecedents or consequences. For instance, researchers can address questions about whether the consequences for contact molestation are similar to those for rape.

More recently, some studies have used multiple-level measures of CSA that include noncontact molestation (see Table 1, column 4; e.g., Boyer & Fine, 1992). Once again, expansion of the range of a measure should increase the strength of its correlations with related variables. However, if the experience of noncontact sexual abuse is not related to elevated levels of adjustment problems compared to not being sexually abused (Haugaard & Emery, 1989), adding this category to the definition would not increase the overall relation of CSA to adjustment. Adding noncontact experiences to definitions of CSA would increase the incidence rate for CSA.

Finally, CSA researchers have disagreed on the upper age limit of childhood. Age cutoffs used in definitions of CSA should be strongly related to the incidence rates reported in studies: All else being equal, studies using lower age limits should report lower incident rates than studies using higher limits. Although age 18 seems to be the most popular upper limit for definitions of CSA, particularly among studies shown in Table 1, age 16 is almost as common, and a few studies use ages 12, 13, or 14 or puberty (e.g., Thornhill & Thornhill, 1990). Research has shown no clear relation between age at which CSA occurs and adult adjustment (Beitchman et al., 1992). However, Finkelhor (1979) reported that sexual abuse between ages 16 and 18 may be particularly traumatic: Thornhill and Thornhill (1990) found that sexual abuse after puberty was more problematic for women than sexual abuse before puberty. These results provide support for using an older age as the upper limit in definitions of CSA, at least in studies of relations between CSA and later adjustment. We defined CSA as unwanted sexual events that occurred before age 18, the age at which people are legally considered adults for most purposes.

Previous studies have examined some of the measurement problems in CSA research. Haugaard and Emery (1989) reported that the experience of exhibitionism was significantly different from forms of CSA involving contact in terms of pressure experienced and feelings of guilt. As a result, they argued that researchers might want to exclude exhibitionism from definitions of CSA. Similarly, Kilpatrick (1992) found that the relative ages of a girl and her partner were not related to her later adjustment. However, neither of these studies explored whether the sexual event was wanted despite evidence that the characteristics of CSA events, not .just its occurrence, may be related to later adjustment (Beitchman et al., 1992). Haugaard and Emery defined CSA based solely on the relative ages of the partners; Kilpatrick studied all child and adolescent sex events regardless of the age of the partner, the initiator of the event, or the female's perceived voluntariness. In addition, both studies used college student samples.

Given the diversity in definitions used in research on CSA, researchers are faced with tremendous difficulties when comparing or interpreting multiple studies. The purpose of our study was to use a single large, ethnically and socioeconomically diverse sample from the general population to demonstrate the importance of these measurement issues and to make recommendations for future studies on the relationship of CSA on women's adjustment.

METHODOLOGY

Sample

Participants were 2,003 women, 18-22 years old (m = 19.9 years), living in Arizona. Selective sampling was used to obtain sufficient representation of the major ethnic groups ill Arizona for subgroup analyses. Within the sample, 39.2% (n = 785) were Anglos, 14.1% (n = 283) African Americans, 27.4% (n = 549) Hispanics, and 16.0% (n = 320) Native Americans: 3.2% (n = 66) categorized themselves as "other" or did not answer the ethnicity question. In addition, 47.2% (n = 946) had attended college and 34.4% (n = 689) were married or cohabiting.

Procedures

Women were recruited at 44 sites in urban and rural areas throughout Arizona, including reservations, to participate in a study of women's sexual histories (Roosa, Tein, Reinholtz. & Angelini, 1997). These sites were health clinics, community colleges, vocational schools, private health-care providers, universities, shelters, parenting programs, and a mental-health agency. Because less than 3% of the sample came from a mental-health agency, this was not a clinical sample. Women in the targeted age group were approached at each site by a female, either the project manager or a member of the agency who explained the purposes of the study, the anonymity of responses, and participants' rights.

The questionnaire was available in English or Spanish to match the language preference of respondents; 3% (n = 61) completed the survey in Spanish. To assure equivalence in meaning and interpretation of the two versions, we used two translators. The first translated the complete survey (except the depression scale, for which a validated Spanish version existed) into Spanish. The second translated the first draft of the Spanish instrument back into English. Then the original document and the back-translated document were compared. All discrepancies went through repeated translations and back-translations until equivalent documents were obtained.

Participants completed the questionnaire alone or in groups with assistance from the project manager or an agency representative as needed. Responses were recorded on computer-scored answer sheets to facilitate data entry with minimal errors. Upon completion of the questionnaire, participants placed their answer sheets in an envelope, sealed the envelope, and gave it to either the project manager or agency representative. In turn, the project manager or agency representative gave the participants $15.00 for participating. Each participant was given a referral list of local behavioral health agencies that provided counseling services on a sliding fee basis. The sealed envelopes were collected and were not opened until it was necessary to send them to be scored. This was done quarterly over the 18-month data-collection period.

QUESTIONNAIRE

Sexual Abuse

A modified version of Koss and Oros's (1982) Sexual Experiences Survey (SES) was used to determine whether participants had experienced CSA. The SES was developed in studies of college students and has been used almost exclusively with these samples. We modified the SES by simplifying the language to make it understandable to women with limited reading abilities. The original scale included 12 questions about contact molestation (e.g., "Has a male ever used force to kiss or pet you?"), coercion (e.g., "Has a male ever had sex with you when you didn't want to because you felt pressured by the reasons he gave?"), attempted rape (e.g., "Has a male ever used force to try to get you to have sex with him when you didn't want to, but sex didn't happen?"), and rape (e.g., "Has a male ever had sex with you when you didn't want to because he used force?"). A major advantage of using the SES is that each item provides a clear behavioral description of the behavior in question. It does not assume that all women understand the meaning of technical terms such as coercion and rape, or use a vague, all-encompassing single statement (e.g., unwanted sexual events) and allow each respondent to interpret the statement.

However, two SES items ("Has a male ewer gone farther sexually than you wanted to go?" and "Has a male ever become so turned or a that you felt that it was useless to stop him even though you did not want to have sex?") were dropped from analyses because we decided that they were too vague or that they did not clearly distinguish between sexual abuse and poor judgement; all other items describe something clearly being done to the woman against her will. A third item ("Has at male ever had sex with you when you didn't want to by saying things that he didn't really mean?") was dropped from analyses because of an error in wording in the printed questionnaires that changed its meaning. Cronbach's alpha for the 9-item scale was .84. A woman who reported that she had experienced an event in the SES was asked for her age the first time the event occurred. Another item asked participants to identify their relationship to the perpetrator the first time any of these events happened. These latter two items are not part of the SES but were used to create the various measures of CSA.

Molestation Experiences

Boyer and Fine (1992) added items to the original SES to assess noncontact and contact molestation, expanding the definition of sexual abuse, particularly for contact molestation. Five questions were asked about specific experiences (e.g., "Did a male ever take sexual pictures of you when you did not want him to?"). A woman who answered "yes" to an item was asked for her age at the first occurrence. One question, "Did a male ever make you kiss him when you did not want to?", was dropped from analyses because of its almost universal endorsement. Cronbach's alpha for the remaining 4-item scale was .71.

Severity of Sexual Abuse

Two scales of the severity of CSA were created by combining information from the measures of molestation and sex abuse. Using the SES items, we have a five-level measure of sexual abuse: no abuse, contact molestation, coercion, attempted rape, and rape. Adding items from Boyer and Fine's measure, we have six severity levels, including noncontact molestation and a broader definition of contact molestation. Each woman was assigned to the category representing the most severe level of molestation or sexual abuse she had experienced.

Depression

Current levels of depression were assessed by the 20-item Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977). The items of the CES-D are symptoms of depression that had been used in previously validated scales. The major components of depressive symptomatology assessed in the CES-D were identified from clinical literature and factor-analytic studies: depressed mood, feelings of guilt and worthlessness, feelings of helplessness and hopelessness, psychomotor retardation, loss of appetite, and sleep disturbance. A few items to represent each of these components are included on the scale. Because the purpose of the scale is to assess current state of depression, the directions read "How often this past week did you ...?". Validated English and Spanish versions of the CES-D were available (Moscicki, Locke, Rae, & Boyd, 1989). Cronbach's alpha for the current study was .89.

Plan of Analysis

To demonstrate the implications of the various approaches to measuring CSA, we created eight measures of sexual abuse similar to those described in Table 1. All measures defined CSA as unwanted sexual experiences occurring before age 18, but the measures differed in how they categorized samples. One measure categorized women into five severity levels of sexual abuse (no abuse, contact molestation, coercion, attempted rape, and rape) using the SES (Koss & Oros, 1982). A second measure categorized women into six severity levels by adding noncontact molestation to the previous measure and by using a broader definition of contact molestation (Boyer 8, Fine, 1992). A third measure dichotomized women into those who reported any CSA and those who reported none using the five-level SES. A fourth measure dichotomized women into those sexually abused and those not sexually abused using the six-level Boyer and Fine measure. For each of these four measures, we created another measure that mimicked one effect of using relative age restrictions in CSA definitions--that is, treating women who were victimized by boyfriends and friends as though they were not sexually abused (i.e., as members of the comparison group). Women who were sexually abused as adults (at age 18 or older) but not as children were excluded from all analyses.

First, we examined the incidence of CSA generated by each of these measures. Next, we compared mean levels of depression for nonvictims and victims for each measure using ANOVA. Finally, we examined the correlations of the various measures to depression. Although it is unclear whether CSA causes depression or other adjustment problems for its victims, there is strong evidence that CSA and depression are correlated, at least at the zero-order level.

RESULTS

In Table 2, we see dramatic differences in the incidence of CSA across definitions. When the six-level measure of the severity of sexual abuse (column 1) was used, almost 59% of women were identified as victims of CSA compared to about 39% for the five-level measure (column 3). Only 2% (n = 36) reported only noncontact molestation (columns 1 and 2). The broader definition of contact molestation in the six-level measure added about 18% (n = 327) to those identified as sexually abused compared to the five-level measure. When victims of boyfriend and friend perpetrators were defined as not sexually abused, the six-level measure (column 2) identified 34% of women as victims compared to 18% for the five-level measure (column 4). When boyfriends and friends were eliminated as perpetrators, the number of rape victims declined by over 50%, attempted rape victims declined by over 50%, victims of coercion declined by 50%, and victims of contact molestation declined by about one-third. Thus, restricting CSA definitions by eliminating events perpetrated by same-aged peers greatly reduced the number of women defined as victims.

Table 2. Percentage of Respondents Experiencing Sexual Abuse Using Different Types of Measures

                                         Measure

                               Six Levels      Six Levels; No
                                                (Boy) Friend
                                                Perpetrators
Multiple Level Measures
No Sexual Abuse                    41%               66%
Noncontact molestation              2%                2%
Contact molestation                21%               16%
Coercion                            6%                3%
Attempted rape                      7%                2%
Rape                               23%               11%
Two Level Measures
No sexual abuse                    41%               66%
Sexual abuse                       59%               34%

                                         Measure

                               Five Levels     Five Levels; No
                                                (Boy) Friend
                                                Perpetrators

Multiple Level Measures
No Sexual Abuse                    61%               82%
Noncontact molestation
Contact molestation                 3%                2%
Coercion                            6%                3%
Attempted rape                      7%                2%
Rape                               23%               11%
Two Level Measures
No sexual abuse                    61%               82%
Sexual abuse                       39%               18%

Note. N = 1,806

Table 3 shows that all definitions of sexual abuse clearly differentiated victims from nonvictims on depression scores. When we compared the five- and six-level measures of CSA when boyfriends and friends were included as perpetrators, each severity level of sexual abuse, with the exception of noncontact molestation, was significantly different from the no sexual abuse group on depression scores. The small group of noncontact abuse victims had a mean depression score that was elevated compared with that of the comparison group, but not enough to be significantly different; the small size of this group (i.e., the lack of statistical power) might account for the lack of statistical significance. The mean depression score was lower for contact molestation victims when the six-level measure was used than when the five-level measure was used; items used in the broader definition of contact molestation were not as strongly related to depression as the original items in the SES.

Table 3. Analysis of Variance of Depression Scores by Sexual Abuse Experience for Different Types of Measures

                                      Measure
                           Six Levels           Six Levels; No
                                                 (Boy Friend
                                                 Perpetrators
Multiple Level Measures
No Sexual Abuse              .79(1,2,3,4,)        .90(6,7)
Noncontact molestation       .90                 0.92
Contact molestation          .95(1,5)             .93(8,9)
Coercion                    1.08(2)              1.15
Attempted rape              1.08(3)              1.28(6,8)
Rape                        1.21(4,5)            1.25(7,9)
F value                    31.43(****)          17.06(****)

Two Level Measures
No sexual abuse             0.79                 0.90
Sexual abuse                1.08                 1.08
F value                   109.19(****)          38.66(****)

                                      Measure
                           Five Levels          Five Levels: No
                                                 (Boy) Friend
                                                 Perpetrators

Multiple Level Measures      .84(10,11,12,13)    0.90(14,15,16)
No Sexual Abuse             --                   --
Noncontact molestation      1.08(10)             1.14
Contact molestation         1.08(11)             1.15(14)
Coercion                    1.08(12)             1.28(15)
Attempted rape              1.21(13)             1.25(16)
Rape                       36.48(****)          22.61(****)
F value

Two Level Measures          0.84                 0.90
No sexual abuse             1.15                 1.22
Sexual abuse              135.55(****)          88.07(****)
F value

Note. Means with the same superscripts are significantly different from one another (p < .05). N = 1,772

(****) p < .0001.

When women sexually abused by boyfriends and friends were placed in the nonvictim category (columns 2 and 4), we see an increase in the level of depression for women in the nonvictim category and an increase in depression for almost every sexual abuse severity level. The most dramatic change occurred for victims of attempted rape. Attempted rape by boyfriends and friends apparently is not as strongly related to depression as is attempted rape by others. However, because of the increased depression score for the "no sexual abuse" group, differences in depression scores between victims and nonvictims were not as large when CSA by peers was excluded compared to when this restriction was not used. Thus, sexual abuse by peers, although less strongly related to adjustment than was sexual abuse by others, had a meaningful negative relationship to adjustment. Depression scores for contact molestation victims were no longer significantly different from scores for nonvictims when boyfriend and friend perpetrators were excluded from either the five- or six-level measures of CSA; for the six-level measure, the same was true for coercion victims.

An examination of the differences in depression scores for victims and nonvictims using the dichotomous measures (bottom panel of Table 3) showed that these measures generally understated the relationship of CSA and depression for at least some victims, those who were raped. When the less severe forms of sexual abuse were combined with rape in a single category of CSA, the mean depression score was lower than were scores for rape victims only.

Table 4 shows the correlations between the various measures of CSA and depression. Correlations for the multiple-level measures with depression were not significantly higher than correlations for the dichotomous measures. However, when women sexually abused by a boyfriend or friend were treated as nonvictims, correlations decreased by 18% to 37%.

DISCUSSION

The development of science depends a great deal on the quality of measures available to, and used by, researchers (Torgerson, 1958). We used a single data set to illustrate the challenges that face those who conduct research on CSA because of the proliferation of measures that define CSA so differently from one another. The simple choice of measure could lead to reports of incidence rates that differ by over 300%, as we found when comparing the five-level measure when victims of boyfriend and friend perpetrators were not included among those considered sexually abused with the six-level measure when victims of boyfriend and friend perpetrators were included. This dramatic difference is more conservative than it would have been if we had not eliminated the "unwanted kiss" item from the six-level measure; this item would have resulted in almost every woman being labeled a victim of CSA.

Differences in CSA incidence rates between studies are partly due to differences in methods of gathering data, levels of confidentiality provided to respondents, degrees of rapport established between respondents and interviewers, or differences in samples (Koss, 1993). However, the dramatic 300% difference was obtained using data from a single study so that only differences in definitions of CSA accounted for this variation. When studies differ on the measures used as well as on other factors, differences in incidence rates of CSA as large as 1300% are possible (e.g., comparing the rates reported by Boyer and Fine, 1992, and Stein et al., 1988).

All the measures of CSA clearly distinguished women who were not sexually abused from those who were on depression scores. However, the measures that considered severity of sexual abuse provided much more information than did the dichotomous measures. First, noncontact molestation involved a small percentage of women, and this experience was not significantly related to depression scores. Second, although it has been widely accepted that rape victims are likely to experience adjustment problems, the elevated depression scores of women who experienced contact molestation, coercion, and attempted rape (in the latter case, especially when the perpetrator was not a friend or boyfriend) support arguments that many of these women also might be at risk for adjustment problems. Correlational evidence clearly showed that definitions that do not consider CSA experiences involving boyfriend or friend perpetrators are overlooking a potentially important correlate of women's adjustment.

IMPLICATIONS

Studies that use multiple-level measures of sexual abuse provide more information to researchers and practitioners about women's sexual abuse experiences. Using multiple-level measures, researchers can examine the processes that contribute to women's adjustment for each specific type of CSA. Measures that place all CSA victims into a single category are limited in their contributions to our understanding of the processes involved in women's adjustment after CSA. The experiences of women who were touched against their wishes and of those who were raped are vastly different from each other. By treating them as a homogeneous group, we lose sight of critical differences in experiences. Furthermore, because rape victims constituted the largest category of CSA victims, reporting incidence rates or adjustment scores for a dichotomous measure provides a distorted view of the numbers of people who might be seriously affected by CSA and of the relationship of CSA to adjustment. Because of the valuable information gained by using multiple-level measures of CSA, there is little justification for the continued use of dichotomous measures.

We examined two versions of multiple-level measures of sexual abuse: the five-level SES (Koss and Oros, 1982) and the six-level measure, developed by Boyer and Fine (1992). Koss and Oros chose the categories for their five-level measure based on legal definitions that apply to women of all ages and developed behavioral items to assess sexual abuse consistent with these laws. Use of Boyer and Fine's expanded version of this measure greatly increased the number of women labeled as sexually abused. However, rationale for Boyer and Fine's noncontact molestation category is not clear; it apparently represented the experience of very few women and this experience was not strongly related to depression. In other words, neither the validity nor the utility of the noncontact molestation category has been established, at least for understanding the relationship of CSA to women's adjustment. Similarly, Boyer and Fine's use of a broader definition of contact molestation added over 300 women (18% of the sample) to the sexually abused group, but the experiences covered by the new contact-molestation items were less strongly related to depression scores than were the original items (i.e., the mean depression score using the Boyer and Fine items in addition to the Koss and Oros items was lower than the mean score using just the Koss and Oros items). Once again, the rationale for broadening this definition is not clear, and the lower correlations raise questions about the validity of such measures in research concerned with women's adjustment. Given the lack of a clear rationale for either the noncontact-molestation category or the broader contact-molestation definition and the weak relationship of either to depression, researchers interested in women's adjustment may want to take a conservative approach by using the Koss and Oros measure until the utility or validity of the expanded definition of CSA is established.

Studies of CSA that eliminate most or all abuse by boyfriends or friends by using relative age restrictions in their definitions would appear to underreport the incidence of CSA greatly and to eliminate from study or treatment a group that may be at risk for adjustment problems. Using instruments that provide behavioral definitions of sexual abuse, such as Koss and Oros' SES, should eliminate the likelihood that casual or consensual sexual behavior may be counted as sexual abuse, which appears to be at least part of the rationale for using status definitions. Most important, researchers cannot justify discounting or overlooking sexual experiences involving force and/or violence regardless of the victims' or perpetrators' ages.

We used a single large sample from the general population to demonstrate the influence of a researcher's choice of definition of CSA on the results of a study. Although we have addressed a few CSA measurement issues, several others remain to be explored. In particular is the relationship between CSA, when defined by unwanted sexual experiences, and adult adjustment explained by background and family characteristics as appears to be the case when CSA is defined by status measures (Rind & Tromovitch, 1997). Furthermore, how do single-item measures of CSA compare to multiple-level severity measures in both incidence rates and relationships to adjustment?

Research, theory development, and interventions will be enhanced to the degree that consensus is reached among researchers about which types of definitions or measures of CSA are most appropriate and helpful, and the current proliferation of competing but woefully inconsistent measures is reduced. Knowledge to guide theory and intervention development will accumulate much faster when researchers and clinicians can readily interpret results and compare results across studies. Because CSA is a topic of interest to people from many fields, some level of consensus in measurement and definition is likely only if experts from the various areas can openly discuss the merits and costs of different measures. It is unlikely, and perhaps unhealthy, to expect a single definition or measure to meet the needs of all interested parties (Haugaard & Emery, 1989). For instance, those interested in accurate reports of incidence rates of CSA as a legal matter or to monitor efforts to reduce these rates may need a broader measure, including status definitions and definitions of unwanted sexual experiences, than will those who are interested in the relationship of CSA to women's adjustment. Some behaviors that legally constitute CSA may be found to have little or no relationship to women's adjustment and therefore would not be of interest to clinicians or researchers interested in exploring this relationship. Researchers interested in the relationship of CSA to adult adjustment, on the other hand, may be better served by focusing on unwanted sexual experiences that involve contact, threat of force, or force. There appears to be little justification for using any dichotomous measure of CSA. The current diversity in measurement serves as a barrier, a virtual Tower of Babel, to accumulating and communicating useful information about this phenomenon. Substantial progress in research, theory, and intervention is likely to follow the use of fewer and more consistent measures of CSA.

REFERENCES

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

Anderson, J., Martin, J., Mullen, P., Romans, S., & Herbison, E (1993). Prevalence of childhood sexual abuse experiences in a community sample of women. Journal of the American Academy of Child and Adolescent Psychiatry, 32, 911-919.

Baier, J. L., Rosenzweig, M. G., & Whipple, E. G. (1991). Patterns of sexual behavior, coercion, and victimization of university students. Journal of College Student Development, 32. 310-322.

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 and 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 and Neglect, 16, 101-118.

Bendixen, M., Muus, K. M., & Schei, B. (1994). The impact of child sexual abuse--A study of a random sample of Norwegian students. Child Abuse and Neglect, 18, 837-847.

Boyer, D., & Fine, D. (1992). Sexual abuse as a factor in adolescent pregnancy and child maltreatment, Family Planning Perspectives, 24, 4-11, 19.

Briere, J. (1992). Methodological issues in the study of sexual abuse effects. Journal of Consulting and Clinical Psychology, 60, 196-203.

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

Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioral sciences. (2nd ed.) Hillsdale, NJ: Lawrence Erlbaum Associates.

Conte, J. R., & Schuerman, J. R. (1987). Factors associated with an increased impact of child sexual abuse. Child Abuse and Neglect, 11, 201-211.

Ernst, C., Angst, J., & Foldenyi, M. (1993). The Zurich study: XVII. Sexual abuse in childhood: Frequency and relevance for adult morbidity data of a longitudinal epidemiological study. European Archives of Psychiatry, and Clinical Neuroscience, 242, 293-300.

Finkelhor, D. (1979). Sexually victimized children. New York: Free Press.

Finkelhor, D. (1984). Child sexual abuse: New research and theory. New York: Free Press.

Finkelhor, D., Hotaling, G., Lewis, I. A., & Smith, C. (1990). Sexual abuse in a national survey of adult men and women: Prevalence, characteristics, and risk factors. Child Abuse and Neglect, 14, 19-28.

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.

Gidycz, C. A., & Koss, M. P. (1989). The impact of adolescent sexual victimization: Standardized measures of anxiety, depression, and behavioral deviancy. Violence and Victims, 4, 139-149.

Gold. E. R. (1986). Long-term effects of sexual victimization in childhood: An attributional approach. Journal of Consulting and Clinical Psychology, 54, 471-475.

Greenwald, E., Leitenberg, H., Cado, S., & Tartan, M. J. (1990). Childhood sexual abuse: Long-term effects on a psychological and sexual functioning in a nonclinical and nonstudent sample of adult women. Child Abuse and Neglect, 14, 503-513.

Haugaard, J. J., & Emery, R. E. (1989). Methodological issues in child sexual abuse research. Child Abuse and Neglect, 13, 89-100.

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

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

Kilpatrick, A. C. (1992). Long-range effects of child and adolescent sexual experiences: Myths, mores, and menaces. Hillsdale. NJ: Lawrence Erlbaum Associates.

Koss, M. P. (1993). Detecting the scope of rape: A review of prevalence research methods. Journal of Interpersonal Violence, 8, 198-222.

Koss, M. P., & Oros, C. J. (1982). Sexual Experiences Survey: A research instrument in investigating sexual aggression and victimization. Journal of Consulting and Clinical Psychology, 50, 455-457.

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

Moscicki, E. K., Locke, B. Z., Rae, D. S., & Boyd, J. H. (1989). Depressive symptoms among Mexican Americans: The Hispanic Health and Nutrition Examination Survey. American Journal of Epidemiology, 130, 348-360.

Mullen, P. E., Martin, J. L., Anderson, J. C., Romans, S. E., & Herbison, G. P. (1993). Childhood sexual abuse and mental health in adult life. British Journal of Psychiatry; 163, 721-732.

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

Parker, S., & Parker, H. (1991). Female victims of child sexual abuse: Adult adjustment. Journal of Family Violence. 6, 183-197.

Peters, S. D. (1988). Child sexual abuse and later psychological problems. In G. E. Wyatt & G. J. Powers (Eds.), Lasting effects of child sexual abuse (pp. 101-117). Newbury Park, CA: Sage.

Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401.

Rew, L., Esparza, D., & Sands, D. (1991). A comparative study among college students of sexual abuse in childhood. Archive of Psychiatric Nursing, 5, 331-340.

Rind, B., & Tromovitch, P. (1997). A meta-analytic review of findings from national samples on psychological correlates of child sexual abuse. The Journal of Sex Research, 34, 237-255.

Roesler, T. A., & McKenzie, N. (1994). Effects of childhood trauma on psychological functioning in adults sexually abused as children. Journal of Nervous and Mental Disease, 182, 145-150.

Roosa, M. W., Tein, J., Reinholtz, C., & Angelini, P. J. (1997). Relationship of childhood sexual abuse to teenage pregnancy. Journal of Marriage and the Family; 59, 119-130.

Saunders, B. E., Villeponteaux, L. A., Lipovsky, J. A., Kilpatrick, D. G., & Veronen, L. J. (1992). Child sexual assault as a risk factor for mental disorders among women. Journal of Interpersonal Violence. 7, 189-204.

Stein, J. A., Golding, J. M., Siegel, I. M., Burnam, M. A., & Sorenson, S. B. (1988). Long-term sequelae of child sexual abuse: The Los Angeles Epidemiological Catchment Area Study. In G. E. Wyatt & G. J. Powers (Eds.), Lasting effects of child sexual abuse (pp. 135-154). Newbury Park, CA: Sage.

Thornhill, N. W., & Thornhill, R. (1990). An evolutionary analysis of psychological pain following rape: I. The effects of victim's age and marital status. Ethology and Sociobiology, 11, 155-176.

Torgerson, W. S. (1958). Theory and methods of scaling. New York: Wiley & Sons.

Wyatt, G. E. (1985). The sexual abuse of Afro-American and white-American women in childhood. Child Abuse and Neglect, 9, 507-519.

Yama, M. F., Fogas, B. S., Teegarden, L. A., & Hastings, B. (1993). Childhood sexual abuse and parental alcoholism: Interactive effects in adult women. American Journal of Orthopsychiatry, 63, 300-305.

Manuscript accepted August 14, 1997

This study was funded in part by the Arizona Disease Control Research Commission and Arizona State University's Investigator Incentive Award program. We extend thanks to Joanne Gersten, Shirley Johnson, Karen Dubner, and Jennifer Wood for contributions to the study, to cooperating agencies and their staffs, and to all the young women who made this study possible.

An early version of this paper was presented at tile National Council on Family Relations Annual Conference in Kansas City, MO, November 5-10, 1996.

All correspondence should be addressed to Mark W. Roosa, Family Resources and Human Development. Arizona State University, Tempe, AZ 85287-2502; e-mail: [e-mail].