[Page 708 continued]
Much of the concern regarding Rind et al. (1998) has centered on the small CSA-outcome effect sizes they report, leading to criticism of their methodology and findings (e.g., Dallam, Gleaves, Cepeda-Benito, Silbe,. Kraemer, & Spiegel, 2001). Although we do not endorse Rind et al.'s data (and in fact have some specific methodological criticisms), it is our belief that the findings themselves are not cause for anything other than normal scientific skepticism. Our deeper concerns - like those of many - lie less with the data than with their presentation. We will thus consider the data and their presentation separately, as well as the definitional issues that the authors raised.
As stated above, a number of methodological criticisms have been leveled against the Rind et al. (1998) meta-analysis. In one of the few published commentaries, Spiegel (2000) criticized Rind et al.'s methodology in a number of ways. For example, he emphasized the tendency for victims of CSA to show only a subset of all possible symptoms; thus, anyone symptom may not be significantly elevated in the CSA population as a whole, even though the majority of individuals demonstrate some symptoms. He also noted that the abuse-specific outcome of posttraumatic stress disorder (PTSD) was absent in the list of dependent measures used by Rind et al. and that failure to prove a relationship does not prove that a relationship does not exist.
Close reading of the original sources (of which we were able to gather approximately 75%) suggests that there may be further reason for closer analysis of Rind et al. ' s ( 1998) findings regarding self-reported reactions to and effects of CSA. For example, 11 of 15 studies listed in Table 7 (p. 36) and 7 of 11 studies listed in Table 8 (p. 37) used definitions of CSA that include both contact and non-contact sex. In another study, fully 83% of males' "CSA" experiences involved a proposition by an adult, but not actual sexual contact (Landis. 1956). Appropriate interpretation of Tables 7 and 8, in our opinion, would have been greatly facilitated by a caveat emphasizing the types of experiences included in the studies listed.
In spite of the above, we are most troubled by the firm conclusions drawn from the partialization of family environment when virtually all data were derived from self-report, retrospective, quasi-experimental studies. Although the use of retrospective quasi-experimental designs is common in the absence of better (i.e., prospective) data, retrospective designs are particularly problematic for assessing the relative contributions of risk factors such as CSA and family environment. Family environment is of special concern in this regard, in part because it may be at once a risk factor, a correlate, and an outcome of CSA. These and other concerns are described in detail by Briere and Elliot (1993).
In this vein, Garbarino and Eckenrode (1997) have likened events of abuse or neglect to a fever, often of concern in its own right, but also an intrinsic part of a larger pathology. Thus, to remove variance associated with a negative family environment is to artificially separate family context from the events that are part of that context.
There are multiple possible pathways by which CSA, family environment, and child mental health can be interrelated. Having only quasi-experimental retrospective data and the very blunt instrument of covariance analysis, Rind et al. ( 1998) emphasized only one of many possible interpretations - that family environment is a confound, the removal of which allows a "clean" analysis of the relationship between CSA and outcomes.
Rind et al. did not describe other possibilities in handling the CSA - family-environment relationship, such as that CSA can lead to more negative assessments of family environment. They took care to address concerns regarding the validity of their partialization procedure, but concluded that these concerns "do not appear to be problematic in the current review" (p. 43).
We believe that this is an overstatement of the extent to which supporting data can mitigate the inherent weaknesses of partialization procedures when used with self-report, retrospective, quasi-experimental data, especially when environment and CSA are so thoroughly intertwined.
Presentation and Interpretation of Findings
In spite of these suggestions, we wish to reemphasize that our concerns regarding Rind et al. (1998) are not predicated solely or even primarily on their methodology and findings, which should be accorded the same blend of trust and skepticism as any other study. In our view, the primary flaw in the Rind et al. article is not the science that it used but its use of science. Through its emphasis on certain key points and its omission of others, the article could be interpreted as using science to inappropriately imply that key moral assumptions about CSA should be reconsidered. We take issue with the basis as well as with the logic and nature of these implications.
the foundation of their discussion is limited by how narrowly harm was defined in this study. For example, it is common and acceptable to study anyone of many possible aspects of harm. Regarding CSA, this might include general psychological correlates such as
It could also include
Both long- and short-term outcomes are perfectly appropriate for study. Rind et al. (1998) chose to study long-term (i.e., young adulthood) general psychological effects, a common and reasonable focus of study.
However, mental health symptoms alone, especially when measured years later, are only one aspect of harm and by no means a necessary or sufficient definition of harm. If proving the existence of harm at all requires the demonstration of effects (with or without intervention) lasting into young adulthood, it would seem that other clearly negative childhood experiences - for example, being beaten by an adult or having leukemia - might not qualify as harmful either. Moreover, harm does not require that the victim perceive the experience negatively. For example, the possibility that a child might learn from an abuser that such experiences are normal and positive is one of the most concerning possible outcomes of CSA.
the effect sizes derived in the Rind et al. (1998) study must be considered in context. The effect sizes reported may seem small and are accurately described as small under Cohen's (1988) suggested definitions.
For example, prior to co-varying family environment, Rind et al reported many effect sizes below .10, and values between .11 and .13 for relations between CSA and primary mental health outcomes such as anxiety, depression, paranoia, psychotic symptoms, and general adjustment. They noted that these relations are small, and that "CSA effects or correlates in the college population are not intense for any of the 18 meta-analyzed symptoms" (p. 32).
However, small effect sizes can reflect very important effects for many people and impact large numbers of people if a phenomenon is relatively common, as CSA appears to be. From a public health perspective, even miniscule effects can have huge personal and societal costs when one extrapolates to a societal level. For instance, the effects of aspirin in preventing heart attacks in one major study was only r = .03. Nonetheless, this meager effect size translated into nearly half as many heart attacks in the experimental
group as in the placebo group (Rosnow & Rosenthal, 1989), an outcome of tremendous personal and financial significance.
A similar example comes from a recent retrospective national survey of 4,000 adult women regarding rape in childhood and its mental health correlates (Saunders, Kilpatrick, Hanson, Resnick, & Walker, 1999). Using data from this study, we calculated the effect size of a reported history of child rape on current diagnosis of PTSD to be r = .12, similar to the mean value reported by Rind et al. (1998) before co-varying family environment. However, women in the Saunders et al. study who reported experiencing child rape were over 4 times more likely to meet Diagnostic and Statistical Manual of Mental Disorders (4th ed; American Psychiatric Association, 1994) criteria for current PTSD than those not reporting child rape (8.5% vs. 1.9%). Effect sizes do not convey clinical , significance, and in this regard can be misleading.
A third interpretive concern
is the authors' broad conclusions on the basis of findings from college samples. Regardless of the "similarity", (p. 27; this is not clarified further by Rind et al.. 1998) in effect sizes between a meta-analysis of college samples and three meta-analyses of broader samples, findings from a college sample may not clearly generalize to the broader population. In addition to the obvious tendency of college samples to include higher functioning individuals, studies of this group can easily obscure effects on more vulnerable populations such as those at risk for psychopathology. If CSA is more harmful to certain subgroups, the more global analysis conducted by Rind et al. would obscure those effects.
Although some tempering comments are made, Rind et al. ( 1998) failed to highlight these and other caveats in discussing their data. For example, they did not point out
They did not suggest that the data may have looked different if they had not collapsed on age at which the CSA occurred or that positive reactions to CSA may be related to age. They did not address
Neither did they highlight
They also did not note that most of the studies used in Tables 7 and 8 utilized particularly broad definitions of CSA that include sexual solicitation; this is important given evidence that boys and girls report equally negative reactions to and effects from CSA when CSA is more narrowly defined to include physical contact of a sexual nature (Haugaard & Emery, 1989).
Instead of appropriately qualifying their findings, Rind et al. ( 1998) emphasized aspects of their data that suggest CSA is not harmful, made allusions implying that CSA can be morally benign, and suggested that researchers should not characterize CSA as a negative phenomenon unless it is unwanted and produces long-term harm. Such a presentation appears to represent, at its core, an attempt to erode current societal views regarding CSA.
For instance, their emphasis on adults' recollections and perceptions of whether CSA was wanted implies that perhaps children and adolescents can make informed decisions about having sex with an adult.
A second example of this advocacy for relaxed moral standards occurs where the authors draw parallels between society's current attitudes toward CSA (including use of the term abuse) and 19th century attitudes toward masturbation. The authors asserted that there is "a strong need for caution in scientific inquiries of sexual behaviors that remain taboo, with child sexual abuse being a prime example" (p. 45). The authors went on to note that adult- adolescent sex "has been commonplace cross-culturally and historically, often in socially sanctioned forms, and may fall in the 'normal' range of human sexual behaviors" (p. 46). It is difficult to avoid interpreting this and other language in the article as meaning that first masturbation and soon CSA may be revealed as simply another "sexual behavior" that must shake itself free of outdated moral baggage. Making such a comparison without highlighting the extreme and obvious differences between masturbation and CSA is troublesome, especially when other caveats are also omitted.
Definition of CSA
A great deal of controversy has surrounded attempts to define CSA, and Rind et al.'s (1998) suggestions in this regard have proved particularly controversial; consequently, some exploration of definitional issues in CSA appears warranted.
Rind et al. followed the above-critiqued presentation by questioning the basis on which CSA is defined in science. They wrote, "Classifying a behavior as abuse simply because it is generally viewed as immoral ... is problematic, because such a classification may obscure the true nature of the behavior and its actual causes and effects" (p. 45).
However, at issue for the majority of researchers is not whether sex with children should be considered abuse but rather the gray areas of how child should be defined and whether acts not involving contact (such as exhibitionism and exposure to pornography) should be included in the definition.
A number of different definitions of CSA have been used in the published literature, with variations leading to great differences in prevalence estimates (Fromuth & Burkhart, 1987; Wyatt & Peters, 1986). Holmes and Slapp ( 1998) listed some of the various requirements that investigators have included in their operational definitions of CSA:
Such variations reflect attempts to approximate the basis on which acts are considered abuse. Rind et al.'s (1998) suggestion regarding the basis on which science should consider certain acts to be abusive, and our proposed alternative. are described below.
The empirical harm standard.
Rind et al. (1998) argued that scientific clarity demands a definition based on empirical (evidence of harm) rather than legal or moral criteria. They stated:
However, basing definitions of abuse on empirical evidence of resultant harm is problematic for several reasons. A primary reason
is that such a definition equates abuse and harm in a tautological manner that overlooks the essential nature of abuse. If abuse is only what is harmful, then abuse as a concept ceases to be necessary; one need only speak of harmful behavior or child harm.
The term abuse is used to specify acts that are "corrupt" or "improper" in nature (Merriam-Webster, Inc., 2000). A wide range of intentional and harmful, but not abusive, acts (e.g., surgery) would be wrongly classified if harmfulness was the only standard applied (Finkelhor, 1979). It is unclear why this key aspect of abuse was eschewed by Rind et al.; operational definitions should approximate the constructs they reflect as accurately as possible, rather than seek to alter the essential nature of those constructs.
Further, under the harm standard proposed by Rind et al. ( 1998 ), labeling behaviors as abuse requires scientific evidence that a significant number of previously abused adults are still affected many years later. The implications of such a requirement quickly become absurd. How big of an effect size, using what methodology, is necessary before sex with children becomes abuse under a long-term harm standard? Is it abuse if negative effects are present 5 years after the event, but not if they can be documented only1I year after the event? Should rape be relabeled as "unilaterally consenting adult-adult sex" by scientists if many victims do not show mental health problems years later (co-varying for other events in their lives)? A parent who administers crack cocaine to a 6-year-old may very well not cause long-term or even short-term harm. Under a harm standard, even a reprehensible act such as this would not qualify as abuse.
Abuse definitions that require empirical evidence of harm are also problematic in that the causation of harm cannot be proven experimentally. In this way, without randomly assigning children to abuse conditions (obviously impossible), the acceptance of a harm standard makes abuse nonexistent. Further problems arise given the inevitable limitations of available measures. Abusive acts could be classified as "non-abusive" simply because of the inability of available measures used to capture important sequelae.
The moral standard.
Ultimately, we believe that child maltreatment in all forms - not just CSA - may best be determined sociologically through the consensus of a given society (Barnett, Manly, & Cicchetti, 1991). The afore-mentioned problems arise because CSA is not and was never meant to be primarily a scientific construct. It is amoral and legal term, and as such its definition should have a sociological rather than an empirical foundation (Barnett, Manly, & Cicchetti, 1993).
Finkelhor (1979) has argued cogently that the inability of children to provide full and informed consent is the proper basis on which sexual acts with children are appropriately described as abusive. Finkelhor clarified two preconditions to true consent:
We would argue that children are fundamentally incapable of meeting these preconditions. They cannot fully understand the ramifications of sexual acts, and they are never fully free to accept or decline when interacting with an adult. Clear agreement on this point is evident in the principles of the scientific and legal communities: Children are seen as incapable of free and informed consent to
These positions are not based on evidence of inevitable long-term harm or children's unwilling- ness to make these decisions for themselves but rather on societal beliefs that children lack the maturity to make major life decisions and need to be protected from those, who would exploit their immaturity.
One possible criticism of this standard is the lack of clarity in a societal definition of abuse; it is true that gray areas will inevitably arise from such a definition. Setting an age of consent for behaviors as disparate as driving, voting, or having sexual relationships is by its nature arbitrary because same-age children will vary in their capacities. However, the abusiveness of the vast majority of sexual acts involving children is quite unambiguous from a societal perspective. For example, a survey of a representative community sample about parental behaviors found sexually abusive acts to be of more concern than all other forms of abuse (Giovannoni & Becerra, 1979). Further, operational definitions that approximate the larger construct of CSA without altering it (e.g., by including only "unwilling" encounters) are easily implemented (e.g., Wyatt, 1985).
A second possible criticism is that many nonsexual acts involving children (e.g., riding a roller coaster, being thrown in a pool) could be considered mildly abusive if unwanted and damaging, and benign if wanted and harmless. Thus, critics may argue, why is it that sexual activity with an adult cannot be held to the same standard? Ultimately, no amount of explaining why the vast majority of persons see sexual acts between adults and children as qualitatively different from other activities that children and adults engage in may ever be convincing to those holding the minority view.
A number of important and widely held values appear to converge to make CSA uniquely and consistently abusive - for example,
On the basis of their findings regarding harm and self-reported reactions, Rind et al. (1?98) concluded that, "it is appropriate to reexamine the scientific validity of the construct of CSA as it has been generally conceptualized" (p. 45). They suggested renaming the construct with a value-neutral term such as adult-child sex, suggesting that the term abuse may obscure the behavior's "true nature" (46).
Although the need for more consistent operational definitions is clear (and we agree with Rind et al. on this point), scientists studying a range of social behaviors - from rape to robbery to gangs - have not previously found a need to alter these value-laden terms. Although some may choose to use neutral terminology in research (e.g.: fatal intentional injury rather than murder), such terms in and of themselves do nothing to advance the field; there is a difference between a euphemism and a precise definition. A stranger who provides a willing child with heroin may not cause short- or even long-term harm; further, that child could report the experience as positive and might grow to see heroin use as a normal and natural part of life. In our opinion, that adult's act would still be child endangerment, would still be corrupt, and could not be either profitably or appropriately labeled adult-child drug sharing.
Further, the use of value-neutral terminology may only be feasible when there is little risk that certain individuals might infer that widely condemned behaviors are acceptable. This is particularly true when a small but vigorous minority is actively seeking to justify illegal acts that contradict consensual public morality.
For example, the effect on society at large, and especially its fringe elements, could be immediate if science and respected scientific societies were to define only unwanted sexual acts as abuse.
This, in fact, appears to be exactly what has happened; NAMBLA and other pedophilia advocates continue to trumpet the Rind et al. (1998) meta-analysis as supportive of their views and as a rationalization for engaging in sex with minors.
Rind et al.'s suggestion thus overlooks the possibility that classifying an exploitive act in neutral terms also obscures much of that behavior's true nature because of the values such terms omit (e.g., that children cannot consent to sex or that it is wrong for adults to use children for sexual gratification). The term adult-child sex lends itself to a set of values that are far more troublesome and disturbing than those Rind et al. sought to avoid.
Science and Morality
This leads to what may be the crux of the matter in understanding where Rind and colleagues (1998) went astray and ironically, is a point that the authors themselves highlighted briefly in their discussion: Science cannot provide answers to moral and legal questions. Science is a method for studying relations between observables, and is no more able to offer ultimate answers to questions regarding morality than it is able to address the purpose of life. Scientific research can inform moral issues (e.g., contributing to a new moral value that parents should place infants in car seats) but can never be the sole arbiter of them.
Similarly, philosophical beliefs and consensual values should never be used to make assumptions about relations between observables (e.g., CSA and long-term psychological harm). Society's moral stance on CSA, as with a wide range of other actions involving children (e.g., child labor, child management of large sums of money), is appropriately based only in part on the potential for harm (which is quite clear, in Rind et al., 1998, and elsewhere). The negative response to Rind et al. among the public and many scientists is thus something very different from an attempt to censor unpopular data. It is instead a rejection of the way that those data are used to make implications in a sphere in which they have no authority.
In urging the abandonment of terminology implying moral judgment, in comparing taboos against CSA now to those against masturbation previously, in their failure to fully qualify their findings, in their emphasis on certain aspects of their data, and in their reminder that other societies have endorsed adult-adolescent sex, Rind et al. ( 1998) appeared to make a crucial extra-scientific assertion - that data suggesting a certain relation between CSA and functioning in young adulthood allow one to question moral judgments regarding CSA.
We suspect that the authors themselves would contest that they made this assertion. In fact, they acknowledged that "lack of harmfulness does not imply lack of wrongfulness," (p. 41) and went on to say the findings "do not imply that moral or legal definitions of or views on behaviors currently classified as CSA should be abandoned or even altered" (p. 47). However, these caveats appear insufficient in balancing the overall presentation, a suggestion that is supported by the strong public reaction to the article.