Discussion of the Rind meta-analysis
with a British journalist

By Tom O'Carroll

A JOURNALIST invited by Tom O'Carroll to read the famous Rind et al. meta-analysis published in the Psychological Bulletin wrote back with a number of criticisms of the paper and questions about it. The following are extracts from Tom's response.

Introduction by Tom:

"The journalist, Christian Wolmar, had initially written to me in connection with a planned book on alleged 'abuse' in British children's homes and by the time Wolmar read the Rind paper an e-mail correspondence between us had been going on for over six months. Despite attempts by Wolmar to disguise the fact, it became clear to me very early on that the guy was hostile. The fact that he had stated his intention of writing about PIE and my role in it persuaded me that it would be better to engage him in debate rather than ignore him.

I did eventually manage to persuade him that current police methods in Britain sometimes result in serious miscarriages of justice. This is the main theme in Richard Webster's excellent short book The Great Children's Home Panic, to which reference is made.

In a later email Wolmar congratulated me on my 'rigorous' approach to the Rind findings – but then he completely ignored Rind in his simplistic, shock-horror book! I found him a hugely irritating person to deal with: ignorant, pig-headed and arrogant. The fact that I was less than polite towards him in what follows (to put it mildly!) simply reflects my increasing frustration after banging my head against the wall of his stupidity for months on end."

Tom’s email to Wolmar:

You say the paper [the Rind et al meta-analysis] doesn't stand up because "It is only about college students, and it only follows them into adolescence, not adulthood."

Alleged bias as compared with general populations does not hold up statistically. CSA effect sizes are in line with national, community and non-clinical samples derived from other studies. I note you say nothing about these studies, of which details are given in the paper. It looks very much as though you have not taken the trouble to consider the implications of the effect size comparisons in question.

People too distressed and traumatised to make it to college as a result of sexual abuse would be expected to show up in effect sizes from the community etc samples if these people existed, but the evidence suggests they do not. Traumatic cases turn up in clinical samples, but there is nothing to suggest these are made up disproportionately of non-college goers.

Please read the following carefully, from the study: "Comparing the college and national distributions indicates similar prevalence rates for intercourse for women; SA college men, however, show a higher rate (33%) than SA men in the general population (13%). Because intercourse is frequently viewed as the most severe or serious type of CSA, these results imply that SA college students, especially men, do not experience less severe CSA than SA persons in the general population."

Similarly, it is shown that college students do not experience any less incest (which is generally considered at the harmful end of CSA ) than the general population, when compared with general population studies in the US, UK and Spain (the three countries used in various college v national comparisons). In terms of close family sex, college students experience more (16%) than the general population (11%). The college v national figures are also quite similar for those experiencing more than one episode of CSA (52% for college students, 46% in national studies); the unweighted figures in this case were identical at 49% for college and national.

In the authors' discussion of their results, the following is particularly relevant: "Compared with SA persons in national samples, SA college students experienced intercourse, close family CSA, and multiple incidents of CSA just as often, and the overall prevalence of CSA was not lower in the college samples. The magnitudes of CSA-adjustment relations in the college samples and in the national samples meta-analyzed by Rind and Tromovitch (1997) were identical: r _u= ..07 for men, r _u= .10 for women. Thus, college students do not appear to present fewer symptoms, experience less severe CSA, or show better coping."

Younger and older adults did not differ in CSA-adjustment relations in Neumann et al.'s (1996) meta-analysis. sOne of the national studies referred to was a British one, and therefore perhaps particularly relevant to your concerns. This was Baker & Duncan (1985). They showed with their national sample (i.e. sample of adults, not students) that lasting ill-effects of CSA are rare: 13% for women; 4% for men.

Your objection ignores, or fails to comprehend, the rationale for using college student data. The rationale was that these studies comprise the largest body of nonclinical research on CSA, that they are more representative of CSA experiences in the general population than are clinical cases (because some 50% of the population in the US has college experience), and they are rich in family environment data, which are relevant to examining whether CSA causes symptoms.

These points deal adequately with your objection, I believe, as you would have realised if you had really thought about it and have at least a modest grasp of the statistical considerations. Of course, if you still don't know your r s from your elbow there may be no getting through to you.

 [Tom adds for Ipce: This is a very British joke: r s sounds like "arse" = person’s bottom. To say someone does not know his arse from his elbow is a coarse but frequently used way of alleging incompetence.]

 To be honest, I am myself finding it difficult to get a handle on your mental capacity. My assumption until now has been that anyone capable of gaining significant standing with the Independent, a grant from Rowntree etc, must be pretty bright, and the railway articles I have read lend support to this view. But the silly errors you made in your reading of Webster, and your failure in misguidedly looking for "evidence" in a book where strength of argument was the real point, led me to suppose that the problem was one of purblindness arising from lack of intellectual integrity – in plain language, refusing to see what doesn't suit you. Now I find myself once again wondering whether you are really quite all there in the head: it's as though, intellectually, you do have strong points but taken as a whole you somehow seem one loco short of a train set.

If that's true, I suppose it could be considered unkind of me to spell it out. Maybe I should patiently just lead you through everything as many times as it takes until you have finally "got it". Without wishing to aggrandise myself in any way (my own capacity for original ideas is sadly slender) I find myself wondering whether it ever crossed Galileo's mind that the pope must be a bit thick. If so, I guess he would quickly have dismissed the thought, concluding instead that the pope was so sharp he could think of a thousand reasons for avoiding a sensible, logical conclusion. Oh, well, if I keep chipping away I might be able to reduce your own thousand reasons down to about 997 or 996…

I put it this way because Rind et al have performed an amazingly impressive intellectual feat and have been denounced by Congress for their pains. Why were they singled out in this way? Think about it. Every day crap papers are being produced. Does anyone bother to denounce them from on high? Of course not. They simply fade into obscurity (but only after the authors have carved another notch on their CV and, if they are saying something popular, pulled in a bit of loot from chat shows etc). The only people worthy of denunciation are those with something very powerful to reveal, something threatening to orthodox belief and its entrenched interests.

Rind et al are firmly in this category and if you cannot understand the quality of their work – the rigour of it is far, far above the often mediocre standard in the social sciences – I imagine it is because you have not really given it the attention you claim to have done. With Webster you tossed the thing aside because there was no "evidence". With Rind you have evidence in spades. Indeed you have ALL the evidence (from the 59 constituent studies, none of which had been denounced in Congress; on the contrary, their findings were the bedrock of orthodoxy) it was possible to consider in a meta-analytic fashion, but still you are resolutely unimpressed.

Attempts to impress you remind me of attempts to amuse Queen Victoria. "We are not amused!" was invariably the sniffy royal response. I don't know if anyone ever had the balls to ask "Well, Your Majesty, what would amuse you?" but I can ask something comparable: What would impress you? A petition, perhaps, signed by a million mature, certified happily married adults saying they had been unharmed by early sexual experience with a grown-up? A march on Whitehall by legions of children carrying banners saying "We're up for it" and "Lay off our adult lovers" ? There are good reasons why it doesn’t work that way, and why "victims" (some genuine, some not) tend to have the loudest voices.

In reality data have to be worked hard for, and it is entirely typical when they deal with human matters that after being hauled with much sweat out of deep and barely accessible data mines they emerge a bit messy and creased. Hence – very carefully – you have to use recognised techniques to clean off as much muck as possible and iron out the creases. That is what Rind et al have done and your failure to understand the exceptional skill and care they have brought to this process does you no credit. Worse, it is something that really ought to have Rowntree worried about your suitability for the task they have given you. [Tom adds: Wolmar’s planned book on children’s homes had been commissioned by the prestigious Joseph Rowntree Foundation, who gave him a research grant of £50,000.]

I have spelled out the essentials in my numbered points above but I am anxious you will still not grasp what I am ranting on about unless you really come to grips with such things as what "community" and "national" samples mean in this context and for that you need to read the study carefully. For one thing, it means your complaint that the study "only follows them into adolescence, not adulthood" is weak. It is precisely to anticipate objections of this sort that comparisons were made with adult groups (the community and national samples). In any case, are you seriously suggesting that these mere "adolescents" (all of them over 18, presumably, since they were at state universities; we are not talking about high school kids) are not old enough to know their own minds on matters of sexual choice? Maybe you would like to see the age of consent raised to 28? Every last one of the students, I suggest, would by that age be fully aware that legally they had been "assaulted" and would be perfectly capable of indignation and anger about it if the nature of the experience made such feelings appropriate. What's more you know this. For heaven's sake, we live at a time when the gay age of consent is being brought down to 16. For you to talk about "adolescent" students, many of whom in America are in their twenties, smacks almost comically of desperation.

You also say: "More importantly, it seems to lump together different types of sexual experience, some of which would be impossible to define as abusive."

What leapt out at me here was your use of the phrase "seems to". Well, does it or doesn't it? You don't really know, do you? And why don't you know? Because you haven't read the paper carefully you great pillock, that's why. The salient point here is that the Rind paper itself is not responsible for lumping together varieties of CSA that would have been better considered separately in the figures. Meta-analysis has tremendous strengths, which is why it is increasingly being used in many contexts, but every technique also has limitations. One of them in the case of meta-analysis is that inappropriate "lumping" or other methodological flaws in any of the original studies being meta-analysed has to be accepted. You are obliged to take the data as you find them, not as you would wish them. You note the flaws and limitations you find and you use standard techniques where possible to tease out genuine information which will not be confounded by the flaw or limitation in question.

What all this means is that you are right in this respect: if some of the data relate to behaviour so mild one could not (in your view) call it abusive, then it would hardly be any surprise that students considered it not harmful to them. Right? One might in passing pause to wonder why such so-called "abuse" is ever included in studies. There is one reason that is obvious to anyone who has been around the subject for a while: it is because the original researcher(s) in question had an agenda, the aim being to include every conceivable form of "abuse" in their definition so as to bump up the overall abuse figures. They could then expect outcomes such as shock-horror coverage of their report in the media, research grants for further work following these "disturbing" findings, and advancement within the child-abuse industry hierarchy. But where have the eagle-eyed expose reporters been while all this fraudulence has been committed unhindered for the last couple of decades? Climbing on the bandwagon, that's where. Either that or simply being the bunch of boozy, myopic tossers so many of you are. (Sorry, I should say so many of us: you may find it quite understandable that I often feel a certain detachment from my own involvement in the scribbling trade).

The important point though, is not that trivial "abuse" may be included in the definitions of some researchers but that, if this is the case, allowance is made for it in the meta-analysis. This can be done where the appropriate data are available by analysing different kinds of abuse, with different perceived levels of severity, for their effect size. And indeed, as we shall see, this was very much undertaken by Rind et al.

Before coming to that, however, I would be interested to know what exactly you have in mind when you say "different types of sexual experience, some of which would be impossible to define as abusive". The 59 studies examined by Rind have a lot of faults, and some of the authors may be "agenda driven", but they are all fully academic studies and it would be unwise to write-off their definitions in a cavalier way. The details must be considered. This is most probably the passage from Rind that inspired your comment:

"Twenty one (35.6%) of the 59 studies contained a breakdown of the types of CSA that occurred along with their frequencies. Types listed varied from study to study, including acts such as an invitation to do something sexual, exhibitionism, fondling, masturbation, oral sex, attempted intercourse, and completed intercourse. Many authors referred to this increasing level of sexual intimacy as "severity" or "seriousness." Using the reported prevalence rates of the various types of CSA from these studies, we estimated the distribution of four basic types of CSA in the college population: exhibitionism, fondling, oral sex, and intercourse. For exhibitionism, we included reports of being shown or showing sex organs in a sexual context. Researchers assessed exhibitionism by asking participants if someone had shown, exhibited, or exposed to them his or her sex organs, or if they had shown, exhibited, or exposed their sex organs to the other person at the other person's request. For fondling, we included reports of sexual touching and masturbation. Researchers assessed fondling usually by asking participants if they had experienced fondling or genital touching; occasionally they included nongenital touching as examples of fondling."

Would I be right in thinking that the items to which I have given italics are what you had in mind as non-abusive? If not, then can I ask what you are in fact talking about? Assuming I am more or less right, let's consider these items:

An invitation to do something sexual. Ask your pal Francis Wheen about this one. He has devoted more than one column to publicly humiliating a (named) former teacher of his for apparently doing nothing more than what in an adult context would be called "making a pass". If Wheen did not think he had been abused, why was he banging on about it so much? If you think his complaint was trivial you might have an interesting discussion with him about this.

Exhibitionism. I would guess that whether it is traumatic depends very much on the circumstances. Who can doubt, though, that for a little girl suddenly to be confronted with a naked man leaping out from the bushes would be extremely frightening. Trivial? Non-abusive? Hardly.

Non-genital touching. One would have to see the original study to be certain what this means. However, I think we can rule out a "fatherly" hand on the shoulder over clothing, or anything of that sort. What is meant, probably, is fondling a girl's breasts, or touching children of either sex on their bare backside or under their clothing in the same region, or the upper thighs. If you regard these activities as non-abusive, I suggest you imagine trying to argue the point in front of a roomful of heavy abuse pundits like Beatrix Campbell and Andrea Dworkin. I doubt if you'd get out of the place in one piece!

So, if all these activities are indeed abusive where does that leave the strength of your point? Arguably, perhaps, you could say they are abusive but not at the more traumatic end of the scale. I say "arguably" but thanks to the careful work of Rind et al we do not need to argue about it: they have carefully analysed the data on such "moderators" as "level of contact", a fact that has evidently passed you by entirely. Correlations were made for three factors possibly influencing effects. These were "level of contact" (both noncontact and contact sex as opposed to contact sex only ), "level of consent" (both willing and unwanted sex as opposed to unwanted sex only ), and gender (male as opposed to female).

The original studies were inconsistent in providing statistics on aspects of the CSA experience (e.g., force, penetration) that might affect adjustment . The authors examined all studies to search for such moderators and found five types that were reported in at least two studies: force, penetration, duration, frequency, and incest. The authors meta-analysed separately the effect of these factors (a) as self-reported by the participants and (b) in producing symptoms of poor adjustment as measured by standard indicators.

One of the most interesting findings is that there is no relationship between the level of contact and the degree of harm resulting. A mere touch that offends can perfectly well result in more resentment and perceived harm than full intercourse that is willingly engaged in. As the authors say: "The near-zero correlation between penetration and outcome is consistent with the multiple regression analysis finding that contact sex did not moderate adjustment. This result provides empirical support for Finkelhor's (1979, p. 103) observation that our society's view of intercourse as the most damaging form of CSA is "a well-ingrained prejudice" unsupported by research."

David Finkelhor, be it noted, is no hippy libertarian freak. On the contrary, as you should know by now, he is one of the leading academic lights of the CSA industry. Incidentally, you might also care to note that Finkelhor's work and mine are compared together by C K Li in Children's sexual encounters with adults (C. Li, D. West, & T. Woodhouse; New York; Prometheus.).

The above points ought to be convincing enough but what follows is a logically incontrovertible clincher. We can slice the Rind findings another way to demonstrate that the conclusions would hold good even if the original studies that include "lesser" abuse such as non-contact and non-genital fondling were excluded. The 59 original studies almost all showed only a small association between CSA and poorer adjustment. Thus if the studies including non-contact and other "lesser" abuse had been excluded, an analysis of the remaining studies would have produced similar results (eg, if you average ten studies with effect sizes that range from about .04 to about .13, you get a result of about .09; if you average 59 studies with a similar range, you still get an overall effect size of about .09). QED.

The issue of consent

Right, this completes my response to your negative points about Rind but I think it would help before closing to make a couple of positive ones of my own. If I can find the time to deal with your non-Rind remarks it will be in a separate message sent later. My points are these:

(1) If you can see the strength of what I have said so far I hope you will see that the Rind conclusions are indeed worth looking at carefully. I would remind you that "level of contact" was not the only moderator examined and I would draw particular attention to one of the other analyses, namely "level of consent".

(2) Rind et al have attracted condemnation because they have not only come up with strong data analysis but have also stated radical conclusions. These conclusions follow logically from the analysis but their bold and uncompromising statement has been like a red rag to a bull in conservative quarters. What may pass unnoticed unless I bring it to your attention is that this does not mean these authors can be disregarded as mavericks disowned by their peers. Far from it. If you pay careful attention you will see that Rind et al cite a number of other studies that support their findings in key respects.

On the consent moderator, Rind et al came up with findings that should really surprise nobody but which nevertheless are very important and politically dynamite. They did not prove anything conclusively (because the original studies were not constructed in a way that allowed for unambiguous conclusions), and they stated their findings in properly cautious terms, but their data strongly suggested that children are unlikely to be harmed by sexual activities in which they were willing participants.

This leaves those people who say children can never validly consent to anything sexual with some explaining to do. If children's willingness at the time of a sexual event cannot be taken as an important indicator, how come this willingness emerges from the figures as a strong guide both to their future perception of whether they have been harmed and to their objectively measured level of adjustment?

My main reason for raising the consent issue, however, is not to make this point but to let you know there has been an interesting development. In response to criticism of their handling of the consent issue and other matters, the authors gave a talk presented at the 1999 Joint Annual meeting of the Society for the Scientific Study of Sexuality (SSSS) and the American Association of Sex Educators, Counselors, and Therapists (AASECT) on 6 November 1999 at St Louis, Missouri. In this talk, the authors discussed their consent findings and also pointed out a recent British study which they say "cleanly" analysed outcome data as a function of consent. This is the relevant part of what was said at St Louis:

Many studies in our review distinguished between consenting and forced acts. We merely compiled the relevant data and examined the value of consent as a predictor of outcomes. It had utility, it did discriminate, and it was therefore scientifically valid to use as a construct. The studies we reviewed generally defined CSA either as a child or adolescent’s sexual experience that was unwanted regardless of partner’s age, or as wanted or unwanted experiences with someone older—typically, at least 5 years older. We merely contrasted study effects from these two groups to examine the value of "consent" as a predictor of outcomes. This analysis clearly demonstrated the utility of distinguishing unwanted from wanted (i.e., consenting) experiences in terms of predicting outcome.

The procedure that we used, however, was not completely satisfactory, because the second category (which included both wanted and unwanted experiences) overlapped with the first (which only included unwanted experiences). Unfortunately, researchers almost never analyzed outcome data as a function of consent. The first and only study that we are aware of that has cleanly done this was published earlier this year in BMJ (formerly The British Medical Journal).

Coxell and his colleagues, all abuse researchers, examined a nonclinical sample of nearly 2,500 men in Great Britain, recruited from general medical practices. They were interested in psychological correlates of non-consenting sexual experiences, but also inquired about sexual things the men had done prior to age 16 with someone at least 5 years older that they had wanted to do, so as not to miss these "abusive" experiences. Throughout their paper they distinguished repeatedly between consensual sex and non-consensual sex — their terms. They found that 5.3% of the men had had non-consenting sex prior to age 16 (with peers or persons significantly older), but that 7.7% had had consensual sex prior to age 16 with persons significantly older.

We examined the findings reported for their key dependent measure, which was whether the men had reported a psychological problem of at least two weeks duration sometime in their life. We compared their results for three groups of men on this measure: those with no CSA prior to age 16, those with consensual CSA, and those with non-consenting CSA. The results were that the consenting group had no more problems than the control group, with a very small effect size (r = .02). However, the non-consenting group had significantly more problems than either of these groups, with an effect size of r = .10 when compared to the control group and a somewhat larger effect size when compared to the consenting group (r = .15). These results, obtained by abuse researchers using a huge nonclinical sample where consent served as an explicit key moderating variable, provide very strong support for the utility of the simple consent construct.

So, what are you going to do now? Rubbish this "unimpressive" finding by a bunch of weirdos in some trashy foreign rag called The British Medical Journal ? [Tom adds for IPCE: In an earlier email my correspondent had expressed suspicion of research published outside the UK. My comment here is of course sarcastic. I do not share his prejudice against "foreign" journals.] This of course leads into the second of my two points above, namely that Rind et al are not alone in their findings. Studies by Eckenrode et al, for instance, and by Ney et al, support Rind (with non-college data) on family environment being more important than CSA:

Family environment

The finding that family environment is more important than CSA in accounting for current adjustment in the college population is consistent with the results of several recent studies using participants from noncollege populations (e.g., Eckenrode et al., 1993 ; Ney et al., 1994 ). Eckenrode et al. categorized children and adolescents obtained from a large representative community sample in a small-sized city in New York state into six groups: not abused, CSA, physical abuse, neglect, CSA and neglect, and physical abuse and neglect. They found that SA children and adolescents performed as well in school as nonabused controls in all areas measured, including standardized test scores, school performance, and behavior. Neglect and physical abuse, on the other hand, were associated with poorer performance and more behavior problems. Ney et al. (1994) separated their mostly clinical sample of children and adolescents into categories of CSA, physical abuse, physical neglect, verbal abuse, emotional neglect, and combinations of these. They found that the combination of abuse that correlated most strongly with adjustment problems was physical abuse, physical neglect, and verbal abuse. In the top 10 worst combinations, verbal abuse appeared seven times, physical neglect six times, physical abuse and emotional neglect five times each, whereas CSA appeared only once.

These studies support Rind. What's more, the findings have intuitive appeal, do they not? They do not in the least put under strain the commonsense view that what really hurts and damages youngsters is indications that they are hated, or regarded merely as a nuisance or of zero importance. While sexuality can be expressed in a hateful way, as with rape, it is crazy to lose sight of the fact that love and affection are much more salient factors in the way people wish to be intimate. Other studies point in the same direction:

The greater importance of nonsexual negative childhood experiences in explaining later adjustment was clearly demonstrated in a study of a large, representative sample of female college students throughout the United States. Wisniewski (1990) used path analyses to assess the relative contributions of CSA and family environment to current adjustment. She concluded that the data did not support CSA "as a specific explanation of current emotional distress [but instead are] best interpreted as supportive of other factors such as family violence . . . as having the greatest impact" (p. 258). Other researchers who used college samples and used statistical control reached similar conclusions regarding the role of family violence, rather than CSA, in explaining current adjustment problems (e.g., Higgins & McCabe, 1994 ; Pallotta, 1992 ).

We have seen that it is when "deconfounding" procedures are used that underlying truths begin to emerge. This point is put succinctly in relation to Brubaker's work:

Brubaker (1991) imposed control by separating her participants into mutually exclusive categories (i.e., no abuse, CSA only, psychological abuse only, physical abuse only, followed by combinations of these abuse types). This deconfounding procedure has been used recently by other researchers examining noncollege samples, who have shown that when CSA is isolated, its negative correlates tend to shrink considerably or disappear (e.g., Eckenrode, Laird, & Doris, 1993 ; Ney et al., 1994 ).

Neutral terms

Finally, there are other authors whose work has led them to, like Rind et al, to question their assumptions about CSA, such as whether encounters should really always be described as "victimisation" simply because there is an age discrepancy of more than so many years:

Some researchers have questioned their original definitions of sexual abuse after assessing their results. For example, Fishman (1991) borrowed from Finkelhor's (1979) definition to classify sexual abuse of boys mostly on the basis of age discrepancies (i.e., sex between a boy of 12 or less and someone at least 5 years older, or between a boy aged 13 to 16 with someone at least 10 years older), stating that age differences implied sufficient discrepancy in developmental maturity and knowledge to indicate victimization. He found that SA men in his study did not differ from controls on measures of adjustment and reported a wide range of reactions to and effects from their CSA experiences (mostly positive or neutral). In-depth interviews confirmed and elaborated the quantitative findings, leading Fishman to question his original assumptions.

So, you will begin to understand from this that the Rind findings are part of a developing body of mutually consistent research. There has been no outcry against their work in the academic world; far from it, they have enjoyed peer support. The attack on Rind and his colleagues in the US was initiated by an anti-gay group and the battle cry was taken up by right-wing Christian fundamentalists with strong political ties – the kind of people who also mount similar noisy, populist lobbies against Darwin being taught in schools.

And if you can begin to accept that this is indeed genuine research, then you need to start considering its implications for your book. Briefly, I would suggest it means not that CSA in the children's homes is unimportant or should be swept under the carpet. Far from it. I don't like the idea of children being imposed upon sexually any more than you do or the general public does.

But the evidence from Rind (and many other places, but we'll stick to Rind here) is that children will be helped most if the focus of the attack on abuse is sharpened, and directed towards behaviour that is clearly unwanted and damaging. What comes immediately to my mind is the brutal, intimidating atmosphere that is well documented in the case of the Christian Brothers, especially in Ireland. It makes the clearest sense to investigate and prosecute cases in which sexual approaches and contact have been accompanied by intimidation, threats and sheer force. Prevention is better than force, though, and this thought brings in the question of how a bad institutional culture develops in the first place. A bit more from me on that subject soon, I hope. For now, I will finish here.