Method

Definitions Adopted

A standard definition of child sexual abuse has not yet been reached (Genuis, 1991; Kassim & Kassim. 1995; Violato & Genuis, 1993, 1994; Violato & Travis, 1995). Although some researchers have restricted their definition of CSA exclusively to acts of sexual intercourse and penetration. others have included a broader range of victimization, such as the witnessing of asexual act between others, being fondled, or being spoken to in a sexual manner (Metcalfe. Oppenheimer, Dignon, & Palmer. 1990). For the present meta-analysis, we have defined CSA as any unwanted sexual contact (ranging from genital touching and fondling to penetration) during the period in which the victim is considered a child by legal definition and the perpetrator is in a position of relative power vis a vis the victim (Violato & Genuis, 1993. p. 37).

In the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-JV; American Psychiatric Association. 1994), PTSD is defined as

[Page 22]

the development of characteristic symptoms following exposure to an extreme traumatic stressor involving a direct personal experience. To receive a PTSD diagnosis, an individual must demonstrate the persistence (i.e., lasting more than 1 month) of various symptoms that have caused or are causing impairment or distress in one's social, occupational, or other important areas of functioning. Some of these symptoms may include re-experiencing the trauma, avoidance of stimuli associated with the trauma, numbing of general responsiveness, increased arousal and anxiety, sleep disruption, hyper-vigilance, difficulty concentrating, experiences of dissociative states, and diminished interest or participation in previously enjoyed activities (American Psychiatric Association, 1994).

Depression is the display of a despondent mood or the loss of interest or pleasure in nearly all activities; irritability, sadness; changes in appetite or weight, sleep, psychomotor activity; decreased energy; feelings of worthlessness or guilt; and difficulty thinking, concentrating, or making decisions (American Psychiatric Association, 1994).

Suicidality consists of recurrent thoughts of death or suicidal ideation and plans, and attempts or gestures of self-harm with death as a possible end result.

Sexual promiscuity is defined as early involvement in sexual activity and/or prostitution.

The victim-perpetrator cycle involves acts of sexual victimization directed against others following a single or repeated instances of having been sexually victimized.

Finally, academic performance in the present study is defined by scores on intelligence or achievement tests and teacher ratings or self-reports of student learning.

Based on the adopted definition of CSA and each of the constructs just mentioned, we included studies that assessed any of the foregoing symptoms and outcomes for coding and analysis in the current meta-analysis.

Literature Search

The studies selected for this meta-analysis were located by computer and hand searches of the published research from 1976 to 1996. Using the term child sexual abuse, we found a total of 860 studies from the search of the CD-ROM PsychLIT, Sociofile, and ERIC databases. This initial search included documents of published journal articles and conference papers. Each of the 860 original studies was evaluated against the inclusion criteria. Those lacking the necessary requirements were eliminated, thereby resulting in a total of 37 empirical investigations that were accessible, retrieved, and included in the final analyses.

Inclusion Criteria

To be included in the meta-analysis, studies were required to meet five main criteria.

	First, studies had to focus on at least one of the six identified outcome effects (as defined earlier) of sexual abuse on persons victimized in childhood.
	Second, the authors needed to empirically report their findings on the specific dependent variable measure,

[Page 23]

	Third, studies were required to have included empirical data on a contrast group of persons who were not known to have experienced CSA.
	Fourth, only studies that used psychometric measures were included (standardized measurement instruments, observation techniques. checklists, etc.).
	Fifth, studies were required to have sample sizes of 12 or greater.

Data Coding

The effect sizes from the

	PTSD,
	depression,
	suicide,
	sexual promiscuity,
	victim-perpetrator cycle, and
	academic performance measures

were the dependent variables; the experience of CSA was the independent variable.

All the studies were coded for six mediating variables, which included

	participant gender,
	socioeconomic status (SES),
	type of sexual abuse,
	age of abuse,
	relationship to perpetrator of sexual abuse, and
	number of CSA incidents.

Studies received

	a score of 1 if no statistical differences were tested,
	a score of 2 if no statistical differences were found, and
	a score of 3 if a statistical difference was reported between the groups on the specific mediator variable.

Additional recorded and coded factors for each study included

	the source of study,
	year, country of the first author,
	clinical group sample size,
	contrast group sample size,
	total sample size,
	type of design (i.e., longitudinal, cross-sectional, retrospective, and retrospective-longitudinal),
	sampling method (i.e., random, stratified random, cluster. systematic, convenience, quota, and dimensional),
	assessment method (interview, questionnaire, rating scales, standardized test, observation, and self-report),
	type of instrument (structured interview. standardized questionnaire, and developed questionnaire),
	highest level of statistics computed (i.e., percent- ages, univariate, multivariate, and structural equation modeling), and
	reliability of the assessment measure.

A summary of the coding results of some of the descriptive characteristics of the studies and the effect sizes is presented in Table 1.

The effect size for the dependent variable was calculated from

	means and standard deviations,
	correlations,
	chi squares,
	p values,
	t scores, and
	F ratios,

as is conventional in meta-analysis of d values (Glass, McGaw, & Smith, 1981; Hunter & Schmidt, 1990; Rosenthal, 1991; Wolf, 1986).

Moreover. effect sizes for percentages were computed from a table of probit transformations of differences in proportions to effect sizes (Glass et al.). The baseline prevalence rate for CSA in the general population was conservatively adopted at 20% (Violato & Genuis, 1994). Quality weightings based on study sample size were used as an adjustment mechanism in computing an average weighted effect size (Rosenthal, 1995).