Professor Richard Ramsay
Richard Ramsay: To Home Page
Email to Richard RamsayEmail to Pierre Tremblay
University of Calgary Home Page

To Home of: Increasing Awareness of Gay, Lesbian, Bisexual, Transgender, Two Spirit, Queer... Suicide Issues
T0 Index Page for: Changing Construction of Western Male Homosexuality: Suicide Problems
Faculty of Social Work Home Page

To Search For Anything At This Site!

An updated interim paper prepared in support of a Poster Presentation with the same name at the 2004 CASP (Canadian Association For Suicide Prevention) Conference held in Edmonton, Alberta. Original Paper. Authors: Pierre Tremblay & Richard Ramsay

Appendix A

Four Recent Studies & The Dark Side of Resilience

Four published studies were encountered that lead to a closer examination of the data and results. A better understanding of the following studies was sought, with some implications. After the introduction and a short discussion of resilience, information related to the four studies referenced below is presented in separate sections.

1. Savin-Williams RC (2001). Suicide attempts among sexual minority youths: population and measurement Issues. Journal of consulting and Clinical Psychology, 69(6): 983-91.

2. Russell ST, Joyner K (2001). Adolescent sexual orientation and suicide risk: evidence from a national study. American Journal of Public Health, 91(8): 1276-1281.

3. Borowski IW, Ireland M, Resnick MD (2001). Adolescent suicide attempts: risk and protective factors. Pediatrics, 107(3): 485-93. Used the same data set as Russell and Joyner (2000). Supplementary information from: Russell ST (2002, In Press). Substance use and abuse and mental health among sexual minority youth: Evidence from Add Health. In: Allen Omoto, Howard Kurtzman, Eds. Recent Research on Sexual Orientation, Mental Health, and Substance Use. Dr. Russell supplied a digital copy of the paper.

4. Cochran, Susan, and Mays, Vickie (2000). Lifetime prevalence of suicide symptoms and affective disorders among men reporting same-sex sexual partners: results from NHANES III. American Journal of Public Health, 90(4), 573-78.


The first published of these studies (Cochran and Mays, 2000) was reviewed more closely after reading about the feelings of one of its authors as reported in an related article by Peter Freiberg (2000):

"One of the concerns for those of us who are doing this work," Cochran said, "is that we don’t want to go back to where life was before Evelyn Hooker. So I want to make the point that even though there’s increased risk, most (homosexually active) people don’t report suicide attempts or (emotional) disorders. That’s very, very important to remember."
Earlier in the article, the reason for this concern had been given:
"Fifty or sixty years ago, Cochran noted, the presumption was that gay people were all psychologically disturbed. Then, Dr. Evelyn Hooker demonstrated—and other researchers confirmed—that there is no difference between the rate of mental illness among gays and the rest of the population."
The above indicates that some researchers fear that, if a difference is found with respect to "mental health" measures for homosexual compared to heterosexual people, this outcome will apparently cause a return to the past when all homosexual people were decreed to be mentally disordered. This fear is unfounded, however, because when one group is more at risk for something than another, it would not means that they are "all" in the category. The query in this respect would be: Given such an unfounded fear, if it is believed, could researchers be using data in a such a way that potential differences between homosexual and heterosexual people will be minimized as much as possible, or even eliminated if it was possible?

The issue surfaced again at the end of 2001 in a USA Today article by Elias (2001) referring to the study by Savin-Williams that is scheduled for publication in the Journal of Consulting and Clinical Psychology (December, 2001). To explain why, "gay youths." or the "homosexual students in university" would apparently be reporting more false suicide attempts (but this is not what the study data 'says', as it will be shown later), the following was written:

"'They're trying to communicate that they do have difficult lives,' Savin-Williams says. 'But most gay kids are healthy and resilient.' Poorly designed studies that exaggerate their suicide risk 'pathologize gay youth, and that's not fair to them,' he says."
This response is indicative of the belief that "attempting suicide" is equal to "pathology." That is, if gay youth have elevated rates for attempting suicide, the "pathology" label would therefore apply. A postulated falsely elevated GLB risk for "attempting suicide" is also attributed to apparently "poorly designed studies." Yet, it has been repeatedly reported that about one-third (25% to 35%) of gay-identified or self-labelling male youth have been attempting suicide, compared to only about 5 percent of heterosexual male youth. Therefore, we must inquire about Savin-Williams' definition of homosexual or gay youth in his study that produced the results reported in USA Today:
"The other study of 266 college men and women found that gay youths were not significantly more likely than straight classmates to have tried to take their own lives" (Elias, 2001).
Immediately after the above was written, Marilyn Elias reported research information to apparently support the proposition that gay youth are not more at risk for suicidality, or much less at risk than most studies had previously indicated:
"Another recent national survey of 12,000 teens found about 15% of gay and lesbian youngsters had tried suicide, compared with 7% of straight teens. The study was school-based, so it reached far beyond support groups. But researchers didn't ask youngsters what suicide method they had used. So some false "attempts" could have slipped through, says study leader Stephen Russell, director of the 4-H Center for Youth Development at the University of California-Davis."
This study by Russell and Joyner (2001) is the second one listed above and it was published in August, 2001. This study, of many available, produced "the least difference" between "gay and lesbian youngsters" and "straight teens," and especially the least difference between apparently homosexually oriented adolescent males and their heterosexual counterpart (OR of 2.4, compared to ORs of 5 to 14 in other studies). The Elias article begins with the initial statement that GLB youth apparently report more false attempt that heterosexual youth, followed by citing the results of the study with the least difference in suicide attempt incidences. The lead researcher for the study, Dr. Russell, is then brought into the picture to say that "some false attempts could have slipped through," thus creating a reason for believing that even the minimum difference reported (about 2 times more at risk) in his study would be too high. That is, if Savin-Williams is correct with the assertion that GLB youth report more false attempts than heterosexual youth, then all such studies are reporting inflated differences. This was, in fact, the major conclusion of the Savin-Williams study as emphasised in the first paragraph of the Elias' article:
Gay and lesbian teenagers are only slightly more likely than heterosexual kids to attempt suicide, contrary to past studies that suggest gay youths have about triple the rate of trying suicide, says a Cornell University psychologist in a controversial report due next month.

Studies finding that about 30% of gay adolescents have attempted suicide exaggerated the rates because they surveyed the most disturbed youngsters and didn't separate thoughts from action, says Ritch Savin-Williams. Nearly all research on the topic has drawn teens from support groups or shelters, where the most troubled gather, and has taken at face value the claim of a suicide attempt, he says. [Here, only non-random community studies are mentioned, and the similar results for males in large-scale sampling in YRBS are not mentioned.]

Savin-Williams' own two studies, to appear in The Journal of Consulting and Clinical Psychology, focus on 349 students ages 17 to 25. When they said they had tried to kill themselves, he asked what method they used. He also separated out the small minority that attended support groups.

Among key findings: Over half of reported suicide attempts turned out to be "thinking about it" rather than trying anything.

There are many problems with Savin-Williams' study, including the fact that the "gay youth" or "homosexual" youths he is talking about would not be what everyone hearing these words are thinking these words to mean. Basically, if Savin-William's definition of gay or homosexual had been used by Kinsey et al. (1948), they would have reported that 50 percent of American males studied are gay or homosexual because they were not 100% heterosexual, given that 37 percent had related sexually with other males, and another 13% had related sexual desires but had not acted on them. Although Savin-Williams apparently used the Kinsey scale to classify the respondents, he did not ask them questions related to sexual behavior or desires as was done in the Kinsey study. Nor did he ask them if they self-identified as gay or lesbian. Savin-William's definition for the gay or homosexual youths in his study included all students studied who did not report themselves to be 100% heterosexual. That is, if they reported only being "slightly homosexual" on a 7-point scale, they were classified as "sexual minority," and almost fifty percent of students were in this category.

The Russell and Joyner (2001) study that was spoken about with the words "gay and lesbian youngsters" also did not ask the study subjects if they self-identified as gay or lesbian or if they were relating sexually with same-sex individuals, nor even if they had same-sex desires. The subjects were only asked if they had "romantic attractions" for or "romantic relationship" with same sex individuals. This assumes that adolescents who have such "romantic feelings" would all be self-identifying or self-labelling as gay or lesbian. Their results, based in anomalous, new and questionable measures for "sexual orientation," are used to incorrectly cast doubt on the suicidality results of studies that have "specifically" asked adolescents if they identified and/or self-labelled as gay, lesbian, or bisexual, and/or if they were relating sexually with same-sex individuals. Others have used these results, apparently to minimize the magnitude of suicidality problems reported for gay and lesbian youth. For example, even before two articles based on the same data set by Russell and Joyner (2001) and Russell et al. (2001) were published, the results were receiving preferential status and cited in the following manner in the Human Rights Watch (2001) study:

"Asking about same-sex attraction is a way of measuring the incidence of sexual orientation in a group that includes individuals who are not sexually active. The National Longitudinal Study of Adolescent Health, perhaps the most comprehensive of the studies that include questions about same-sex attraction, includes over 12,000 youth in grades seven through twelve drawn from high schools and middle schools across the United States. Six percent of participants between the ages of thirteen and eighteen reported same-sex romantic attraction, with 1 percent identifying same-sex attraction only and 5 percent reporting attraction to both sexes. 34"

"34. Of the remaining youth, 82.6 percent reported opposite-sex attractions only and 11.4 percent reported attractions to neither sex. See Stephen T. Russell and Brian D. Franz, "Violence in the Lives of Sexual Minority Youth: Understanding Victimization and Violence Perpetration" (paper presented at the annual meeting of the American Sociological Association, Chicago, Illinois, August 1999), p. 6."

Later, in the section titled "Coping with harassment and violence," the subsection titled "suicide" that would be apparently concerned about those who become suicidal does not begin by noting this fact. Instead, it begins with:
As noted, most lesbian, gay, bisexual, or transgender youth never consider suicide. A recent analysis of nationally representative data of U.S. students in the seventh through the twelfth grades found, for example, that 84.5 percent of boys and 71.7 percent of girls who reported same-sex attractions or relationships had never attempted suicide and had never had suicidal thoughts.

But several youth reported to us that the harassment and violence they endured led them to consider suicide. For instance…"

After noting two results based on American school studies (Remafedi et al., 1998, and Garofalo et al, 1999), this short section on suicide problems ends with:
"A 1998 study based on nationally representative data found that youth who report attractions to or relationships with persons of the same sex were more than twice as likely as their heterosexual counterparts to attempt suicide. 210"

"210. Stephen T. Russell and Kara Joyner, "Adolescent Sexual Orientation and Suicide Risk: Evidence from a National Study" (paper presented at the annual meeting of the American Sociological Association, San Francisco, California, August 1998), p. 8."

Therefore, the "suicide" section begins, not by emphasising those at risk, but by emphasising the ones who are apparently not at risk, and it ends with the results of a yet-to-be-published study that will report on the lowest relative risk available for homosexually oriented adolescent, compared to their heterosexual counterpart. Furthermore, the "romantic" qualification was made to disappear and the illusion is created that a valid measure of "sexual orientation" was used in the study. This type of reporting should be unacceptable because it sets up a situation where one may easily think: "Given that most GLB youth do not consider suicide or attempt, what on earth could be wrong with the ones who do? Are they pathological? Are they lacking in resilience? Savin-Williams (above) did answer this question when he stated:
"'They're trying to communicate that they do have difficult lives,' Savin-Williams says. 'But most gay kids are healthy and resilient.' Poorly designed studies that exaggerate their suicide risk "pathologize gay youth, and that's not fair to them,' he says" (Alias, 2001).
The "blaming of the victim" here is close, and this perception of resilience was warned against by James Garborino (2001: 176-7):
"The third concept we need to bring to our toolbox is humility about resilience. I've used the term before, the ability to deal with adversity, the idea that we can cope but we also have limits. In American society, however, the term resilience has a kind of dark side to it. In this judgmental society, it's a short step between celebrating coping and judging those who are not coping. The worse example I encountered was when I was testifying… the prosecutor asked, 'What wrong with this boys that he isn't resilient?' And I realized that instead of being in celebration of coping, we are now making a deficiency judgment.

Two studies really give you an intellectual foundation for the humility we need in thinking about resilience… when the researchers looked at the 2% who did not break down, they found they were not robust healthy, well adjusted men - they were all psychopaths. That gives us a perspective on resilience - the only people who didn't go crazy, already were. Psychologists Pat Tolan from the University of Illinois, has done the domestic equivalent to that study when he asked what percentage of kids would break down under the burden of additional risk factors… Not surprisingly, the answer was 100%. That doesn't mean the kids' lives are over. Many of them find a way back, but anyone with that burden takes a major hit."

Unfortunately, Garborino's assertion that, "in American society, …the term resilience has a kind of dark side to it" that appears to have manifested itself in nothing less than a major American document dealing with suicide: The National Strategy for Prevention of Suicide (USDHHS, 2001). Suicide has typically been blamed on the individual, or on the individual's assumed pathology (metal disorders), but not on social factors that would maybe required major changes if youth suicidality levels are to be reduced. This is the other aspect of resilience much less talked about and related information is readily available, as the following example from the Internet illustrates. After Rouse, Longo, and Trickett (1999) listed the definitions of resilience by the experts in the field, the following was emphasised in the section titled "Definition of Resilience."
"These definitions have practical applications. What they mean to the practitioner is that some children who are exposed to chronic or severe stress will turn out competent. These children will successfully adapt over time. These children will need tremendous biological, psychological, and environmental resources in order to do this. These children cannot do it themselves. They need love, care, and support not only from their parents, but from educational personnel and other community adults as well...

The definition is also not simple because resilience is contextual. The individual characteristics and environmental factors that lead to resilience in one context may not lead to resilience in another. For instance, academic resilience may be related to a certain set of individual characteristics and environmental factors. However, these same factors and characteristics may not equal emotional resilience. Different kinds of resilience are related to different kinds of support.

The definition is not simple because resilience is complex. It takes personal characteristics such as social skills and environmental factors such as mentoring to create the resilience phenomenon. Resilience does not just come from the person. Additionally, it draws on biological (temperament) and psychological (internal locus of control) characteristics of the person. The environment's role cannot be forgotten. Environmental factors also come into play. People, opportunities, and atmospheres all add to the resilience equation. A resilient personality is not sufficient. It takes the person and his or her environment."

In the National Strategy for Prevention of Suicide, the highly restrictive 'we are setting the stage for blaming the victim' definition of resilience is given:
"Resilience – capacities within a person that promote positive outcomes, such as mental health and well-being, and provide protection from factors that might otherwise place that person at risk for adverse health outcomes" (USDHHD, 2001: 201).
Maybe, the architects of the National Strategy for Prevention of Suicide, or even Savin-Williams, should be given the test for GLB adolescents driven to attempting suicide for a number of reasons, including anti-gay harassment in schools. This 'test' would be: "Do you believe that the kids attempting suicide are lacking in personal resilience? If so, what would you be thinking if it was reported that heterosexual-identified kids who are subjected to the same abuses, but only because they are believed to be gay, are also at similar risk for attempting suicide? Would you say they are also lacking in personal resilience? Or would you say that such problems have nothing to do with resilience, in the personal sense, but everything to do with the abuses inflicted on them. Would you not agree that the concept of resilience, no matter how it is defined, is maybe a red herring? That, maybe, 'resilience' is nothing but a word that many will be using or misusing to somehow do nothing about what is the true causes - the socially constructed causes - of these adolescent suicide problems?"

To better understand the above, it is a good learning experience to remember the 1960s when homosexuality was defined to be a mental disorder and also defined to be a crime (before 1969 in Canada), and evaluate the following situation. A male is imprisoned for relating sexually with another male, he is also subjected to aversion shock therapy for his homosexuality, and he attempts suicide. Would we then say that his suicide attempt was related to a lack of resilience on his part? The same applies today given the great abuses still being imposed on adolescents recognising their homosexual desires, or adolescents suspected of being gay. In the pre-1969 example given above, anyone using the "lack of personal resilience" to explain the suicide attempt would essentially be the best friend of the abusers. His abusers were also the entire social system, from the legislative and legal, to the ones still being called mental health professionals who were mostly being trained in universities, much like the situation existing today. Maybe similar professors are still teaching at universities. Apparently, there are professors who would believe, given the facts of the case, that GLB youth who attempt suicide are not only pathological, but that they are also lacking in personal resilience.

There are significant problems related to the results of the three studies listed above, and each one is discussed below. The most recently published study by Savin-Williams (2001) that was deemed "controversial" is discussed first given that the Elias (2001) article was reproduced in many newspapers as noted in the bibliography reference. Information in the article was also summarized for publication in other newspapers and web sites.

1. Savin-Williams R (2001). Suicide attempts among sexual-minority youths: population and measurement issues. Journal of Consulting and Clinical Psychology, 60(6): 983-91

The evaluation of this study is based on a hardcopy final draft of paper and a copy of the questionnaire that was made available by Ritch Savin-Williams. A digital copy of an earlier draft of the paper was made available via email.

Summary and Evaluation of Study Results:

Study 1 Female Sample:

This was a non-random sample of 83 females (15-25 years Mean = 21.9 years, 11 from a community support groups, 72 from "community classes and organization." All report having "same-sex attractions or… are questioning their sexual attraction," 87 percent of the females are White with 64 percent having at least one parent in a professional occupation, and 74 percent identify as lesbian (41%) or bisexual (33%). The remaining females either refused to self-label or they report being unsure about their sexual orientation. Data was solicited via interviews. (p. 986)

Study 1 Results: 23 percent of the females (19/83) report attempting suicide but only17 percent (14/83) made "a real attempt." Five females were defined as having reported "a false attempt." That is, 26.3 percent (5/19) of the suicide attempt reports were deemed to be false attempts. (p. 987)

Study 2 Female Sample - Females replying to an anonymous questionnaire distributed in "several introductory human development and introductory sexuality college courses."

The "sexual-minority" definition in this study is based on the results obtained by including the 7-point Kinsey Scale on a written questionnaire and requesting that females students chose where they would fit on this specified "sexual orientation" scale. The "sexual-minority" classification included all the females that did not rate themselves as 100% heterosexual, meaning that acknowledgement of the least bit of homosexuality on the 7-point scale (being "slightly homosexual") was used to classify that individual as a "sexual-minority female." This application of the Kinsey Scale is not how it was originally used. Its original and correct use involves a researcher soliciting a lifetime of information about one's sexual desires (one part of the scale) and one's sexual behaviours (another part of the scale), and the combined scores produce a Kinsey Scale rating. And additional "sexual orientation" category called "other" was added to the above 7 categories, but it does not seem that any subjects chose this category. Study subjects were not asked about their sexual behaviours or desires. Furthermore, given that many of these students were taking a sexuality course, they may well have learned about the Kinsey scale in such a course and they may therefore have attempted to be more honest and accurate with themselves with respect to supplying a Kinsey self-rating. If this applies, it would greatly bias the results, likely making them incomparable to results obtained from a similar Kinsey rating request made to other students, or to individuals in an average population who would generally not be familiar with the Kinsey scale and how related ratings are done.

Typically, in studies of sexuality, or demography based on sexual orientation, and especially in studies of adolescent/youth, the measures used to identify those with a homosexual sexual orientation have been "gay," "lesbian," or "bisexual" self-identification or self-labelling and/or questions that asked the subjects about same-sex sexual activity. Generally, these studies have produced demographic results indicating that no more than 5% of females were deemed to be 'sexual minority' either on the basis of lesbian/bisexual self-identification and/or because they reported having engaged in same-sex sexual behaviour. At best, some degree of acknowledged homosexuality may apply for about 10 to 20 percent of the population if a pencil-and-paper questionnaire is used to solicit relevant information.

The "sexual-minority" definition in study 2 produced a non-random sample of 73 "sexual-minority females," representing 48 percent of the females answering the questionnaire. This new definition of "sexual minority" would therefore mean that all associated attributes (such as suicidality results) cannot be used to make any statements about what has been traditionally called "sexual-minority youth." This "sexual-minority" definition produces a population subsample that barely qualifies as a minority that is to be compared to just over 50% in the "sexual majority" group. Sexual minority definitions in other studies usually generate a 5% (or less) population subsample that is compared to about 95% in the majority group. In the Bagley and Tremblay (1997) suicidality study that used improved computer methodology, a definition that included self-identification and reported behaviours produced a slightly higher ratio of about 12% to 88%. The Savin-Williams’ definition in Study 2 generated a much larger than usual subsample with reported results that will require serious consideration when researchers begin to carry out youth suicidality research using a similarly expanded "sexual-minority" definition.

Results of Study 2: 152 Female University Students

100% Heterosexual Females (n = 79, 53% of the female students): 11 percent (9/79) report an attempt, but only 8 percent (6/79) were"true attempts," meaning that 33.3 percent (3/9) of the reported suicide attempts by these females were "false attempts." (p. 986)

"Sexual Minority" Females (n = 73, 48% of the female students): 22 percent (16/73) report an attempt, but only 10 percent (7/73) were"true attempts," meaning that 56.2 percent of the reported suicide attempts (9/16) by these females were "false attempts." (p. 986)

Conclusion 1: When the definition of "sexual minority" is applied to all female students who report the slightest amount of homosexuality (about 48% of females in Study 2) the results suggest that a serious problem may occur in reports of "true attempts" by the sexual minority group. Their reports of attempting suicide will apparently consist of more false reports of "attempting suicide" than is the case for the 100% heterosexual females in such samples.

Conclusion 2: The "sexual minority" female student subsample in Study 2 is not equivalent (nor is it close to being equivalent) to community samples of lesbian and bisexual identified females (youth or adult samples) in other studies that report "suicide attempt" incidences. The same non-equivalence applies to more random population samples (e.g. Youth Risk Behavior Surveys) in which the "sexual minority" definition was based on self-identification and/or reporting of same-sex sexual activity. The female "sexual minority" sample in Study 1, however, is similar to previously studied populations given that most of the females (74%) in this study self-identified as lesbian or bisexual. For them, as noted above, only 26.2 percent of the reported "suicide attempts" were "false attempts." This result is comparable to the 33.3% of "false attempts" reported for the 100% heterosexual females in Study 2. In this case, the difference in "false attempts" reporting is not statistically significant even though the incidence of false reporting is a little higher for the 100% heterosexual females.

Conclusion 3: Evidence was not presented by Savin-Williams to indicate or suggest that previous studies of mostly self-identified lesbian/bisexual females who have reported "suicide attempt" incidences would have a greater percentage of "false attempts" in their reports compared to the 100% heterosexual female group, as indicated in Study 2. In fact, if anything one could conclude there a little more honesty in "true suicide attempt" reporting for the lesbian and bisexual identified females based on Study 1 results. Given these results, an important question must be asked: "Which of the sexual minority females in Study 2 would account for the higher level of "false attempts"? The evidence indicates that they are likely those who would not otherwise self-identify as lesbian or bisexual. If this is correct, as Study 1 results suggest, it could be assumed that the number of "false attempts" in studies that used a more conservative definition of sexual minority would have been greater on the "heterosexual" side of the compared groups, not on the predominantly lesbian and bisexual side. The Study 2 results therefore lead to the informed suspicion that the higher risk ratios for attempting suicide by lesbian and bisexual identified female youth found in Youth Risk Behavior Surveys, for example, may be underestimates as opposed to the overestimates that are believed and proposed by Savin-Williams.

The Report on 114 Male University Students:

Savin-Williams did not seek a sample of predominantly gay/homosexual or bisexual identified males that would be somewhat equivalent to the predominantly lesbian or bisexual identified females identified (the group having the lowest percentage of "false attempts") in Study 1. The only males studied are part of Study 2, meaning that the results are related to a highly expanded "sexual minority" category that essentially has no equivalent (or near-equivalence) in the research that has reported "suicide attempt" incidences for non-random community samples of gay/homosexual and/or bisexual identified males. This non-equivalence also applies for the studies reporting relative risks for attempting suicide by male "sexual-minority" youth or adults compared to their heterosexual counterparts. "Sexual minority" in these studies was based on gay/homosexual or bisexual self-labelling/identification and/or reports of engaging in same-sex sexual activity. The "other males" were generally classified as "heterosexual" based on heterosexual self-labelling/identification and/or because they did not report having engaged in same-sex sexual activity and/or because they reporting having only engaged in opposite-sex sexual activity.

Results of Study 2:

100% Heterosexual Males (n = 61, 53.5% of the male students): approximately 2 percent (1/61) report an attempt, and "true attempts" apply to 2% (1/61), meaning that 0.0 percent of the one reported suicide attempt (0/1) was a "false attempt." However, a count of "one" means that nothing can be said, with any statistical significance, about what be the percentage of "false attempts" might be in this population of males. For such a determination, one would need a greater number of males who are reporting having attempted suicide. If, for example, their incidence of "false attempts" reporting was 50 percent (meaning about the same as it is for "sexual minority" males), the probability that the first "suicide attempt" count would be a "true attempt" or a "false attempt" is 50/50. Therefore, the fact that the first "suicide attempt" count happens to be a "true attempt" essentially means nothing. On the basis of the "one" count, nothing can be said about the degree of "false attempt" reporting by 100% heterosexual males. Their percentage of "false attempt" reporting could also be higher than it is for "sexual minority" males reported on below.

"Sexual Minority" Males (n = 53, 46.5% of the male students): 23 percent (12/53) report an attempt, but "true attempts" only apply to 9 percent (5/53) of these males, meaning that 58.3 percent of their reported suicide attempts (7/12) were "false attempts."

Conclusion 1: When the concept of "sexual-minority" is defined as applying to all males reporting the slightest amount of homosexuality (46.5% of the male university student sample), there likely will be a serious problem in this group of males with respect to reports of suicide attempts. About 60% of the reported suicide attempts may be "false attempts." The same, however, may also apply for the 100% heterosexual males. Data was not produced to permit any speculation on what their percentage of "false attempts" may be.

Conclusion 2: The "sexual minority" male subsample in Study 2 is neither equivalent (nor is it close to being equivalent) to the subsamples used in studies that have reported "suicide attempt" incidences for community samples of gay/homosexual and bisexual identified males (youth and/or adult samples). The same applies for comparisons to studies of more random or representative population samples of sexual minority based on self-identification and/or having engaged in same-sex sexual activities. These studies include all the Youth Risk Behaviour Surveys; Bagley and Tremblay, 1997; Fergusson et al., 1999; Cochran and Mays, 1999 and all non-random community-based study samples of gay and/or bisexual identified males.

Conclusion 3: Evidence was not presented by Savin-Williams to indicate or suggest that studies of gay/homosexual/bisexual identified males that have reported incidences of suicide attempts would have a greater percentage of "false attempts" reporting by these homosexually oriented males, compared to the percentage for the 100% heterosexual males. In fact, for males in Study 2, evidence is not produced to even indicate that the magnitude of "false attempts" reporting by "sexual minority" males (as defined) would be different than whatever would have been the proportion of "false attempts" reporting for the 100% heterosexual males. A count of "one" is insufficient to offer even a hint as to what the result would be for 100% heterosexual males.

Conclusion 4: The suicidality results for males in the Savin-Williams study do have some redeeming value. When the definition of "sexual minority" is made to apply to a much greater percentage of males than has been the case in previous suicidality studies, the greatly expanded "sexual minority" category is likely include an even greater proportion of "true suicide attempters" than reported in the worse-case scenarios. For example, Bagley and Tremblay (1997) reported that 62.5 percent of the "true suicide attempters" in the stratified random sample of young adult males living outside of Calgary's gay community area were homosexually oriented. Their definition of homosexual orientation was on the basis of homosexual/bisexual self-labelling and/or being currently homosexually active (in the last 6 months). The "sexual minority" males in Study 2 account for an even greater proportion of "true suicide attempters" in the non-random sample of male university students: 83 percent (5/6) in a sample of 114 university male students. In this study, the expanded definition of "sexual minority" males also includes all three males deemed to have performed a "life-threatening" suicide attempt. These attempts were also reported to have resulted in seeking medical care


Savin-Williams' major point appears to be that if the definition of "sexual minority" is increased to something close to 50% of the population, the higher levels of suicidality currently being reported for gay/homosexual/bisexual identified individuals and/or for the ones reporting same-sex sexual activities will likely be reduced. (The 50% figure is close to the Kinsey et al. 1948 estimate for the incidence of manifested homosexuality in males based on honestly reported homosexual desires and/or sexual activities.) This conclusion is true, and all differences could certainly be made to disappear if the definition of 'sexual minority' was expanded even more, to include maybe 100% of the population. This, however, is not the issue. As reported in many studies, males who self-label and/or self-identify with the "gay," "homosexual, or "bisexual" categories and/or report having engaged in same-sex sexual activities have been (and are) at significantly greater risk for suicidality - and related problems - than are males who do not self-identify in these ways. There are also a number of reasons why males would not perceive themselves in this way. Among these, the most significant would be their non-recognition of same-sex sexual desires, their recognition of only insignificant same-sex sexual desires that does not interfere with a heterosexual identification, and not having related sexually with same-sex individuals, at least not as an adult.

Much scripting could be devoted to the deconstruction of the Savin-Williams (2001) study. This would especially apply to the possible deceptions contained therein as illustrated by the author who reports that, if the definition of "sexual minority" is "broadly defined" (without mentioning just how "broad" the definition has become), the differences in suicidality will then become minimal. That is, on the basis of his study results, Savin-Williams emphasises in the abstract that "sexual-minority youth [are] only slightly more likely than heterosexual youths to report a suicide attempt" (p. 1), meaning here "a true suicide attempt." In the paper, Savin-Williams notes that "sexual-minority youths" are only at much greater risk for reporting "false attempts" compared to 100% heterosexual youths. Given the previously reported "true suicide attempt" results for males, however, this statement would be false, or only half true, as in applying only for females, and only if the definition of "sexual-minority" is radically changed. Given that "sexual-minority" males (in Study 2) account for 83 percent (5/6) of the "true suicide attempters" and for 100 percent (3/3) of the "life-threatening" attempts, it is incorrect to conclude that "sexual minority" male youth are only "slightly more at risk" for a real suicide attempt than their 100% heterosexual counterparts. In fact, the opposite applies, or is "suggested" given that the Chi Square for "sexual-minority males versus 100 percent heterosexual males" is only close to statistical significance for true suicide attempts and attempts associated with having received medical attention: X2 = 3.4, p = 0.063 and X2 = 3.5, p = 0.060, respectively. Interestingly, Savin-Williams (2001) reports the first Chi Squared to be "c2 (1, N = 114) = < .10" (p. 988), thus somewhat removing the value from its 'close to statistical significance' status that is nonetheless 0.095 for the related Odd Ratio: 0.92 < 6.2 < 41.4. Furthermore, Savin-Williams does not report that his sample size for males (n = 114), and related counts for "true suicide attempts" (5/53 for 'sexual minority' males; 1/63 for 100 percent heterosexual) would preclude statistical significance until "true suicide attempts counts" would reach "6" (instead of 5) for 'sexual minority' males. That is, it is only when 'sexual minority' males would have a Risk Ratio of almost "7" (6.9), this being a risk ratio much higher than the ones produced in many large population studies (e.g Garofalo, 1998; Fergusson et al. 1999), that statistical significance could occur given the small sample size for males in Study 2. This fact therefore severely caveats Savin-Williams statement: "No significant differences were found among gender/sexual orientation groups in true suicide attempts." Interestingly enough, the Monitor on Psychology (American Psychological Association) did not detect this problem with the study and reported: "In the other study, Savin-Williams reported suicide attempts in 126 young sexual-minority young people and 140 young heterosexual men and women. Again, while sexual-minority men and women were far more likely to report suicide attempts than heterosexual subjects, the two groups showed similar rates of true suicide attempts" (DeAngelis, 2002).

A few of the many problems with the Savin-William (2001) study are as follows:

1. Savin-Williams speaks about the "Kinsey 1s and 2s" or "Kinsey 5s and 6s" in his Study 2 sample as being actual Kinsey ratings of sexual orientation (p. 989), without stating the vast difference in methodologies used. This omission creates the illusion that a Kinsey determination of sexual orientation was carried out. References are also not given to justify the use of the Kinsey scale in a manner not used by Kinsey et al. (1948, 1953). The only reference given in association with the use of the scale is the 1948 Kinsey et al. study (p. 988). A copy of the questionnaire that was made available by Savin-Williams contains the statement "Sexual Orientation: (check one)," followed by eight categories, the first seven apparently being equivalent to a Kinsey Scale determination of "sexual orientation," with the last one being an "other" category.

2. Savin-Williams presents the case that asking a question such as "Have you attempted suicide?" apparently creates interpretation problems. From this he concludes: "Whatever is being measured, sexual minority youths appear more inclined than other adolescents to reply in the affirmative [to the "attempted suicide" question] when simplistic suicide attempt research instruments are used" (p. 989). Such conclusions, however, are only made possible by, for example, knowing about the Bagley and Tremblay (1997) study but not mentioned or referenced in his paper. The Bagley and Tremblay (1997) study had been referenced in the bibliography of a draft of his paper that was made available, and he had also referenced the study in a paper published at the beginning of 2001: Savin-Williams (2001a: 9, 11). These facts raise the question of whether it was deliberately omitted, possibly because it was a study with methodology that Savin-Williams is advocating, but also because it had reported that homosexually oriented males accounted for 62.5% of suicide attempters as highlighted in the study abstract. Referencing this study would have certainly challenged many of Savin-Williams assertions about suicidality research associated with sexual orientation measures. The omission therefore permitted Savin-Williams to presents his study as being somewhat innovative with respect to separating out "true attempts" from "false attempts," but this is an illusion made possible by not citing the Bagley and Tremblay (1997) study. Their methodology did separate "true attempts" from "self-harm". If Savin-Williams had researched the subject of separating false attempters from true attempters, he would have discovered that the methodology used by Bagley and Tremblay (1997) was noted and complimented by Moscicki (1989). In her discussion of the problems with the concept of "attempting suicide" as sometimes being acts of self-harm that are not related to an intent to kill oneself, she complimented the pioneering work of Bagley and Ramsay in this regard. (Ramsay and Bagley, 1985; Bagley and Ramsay, 1985). Their research was conducted in the early 1980s and was referenced in the Bagley and Tremblay (1997) study.

3. Savin-Williams begins his critiques of Youth Risk Behavior Studies as follows: "A common research procedure is to ask only one or two cursory or perfunctory questions about suicidality (see Table 1). For example, the YRBS poses two questions - whether a suicide attempt has occurred 'within the past 12 months' and whether the attempt has resulted in injury (Garofalo et al., 1998)." This commentary, however, illustrates that Savin-Williams did not seek out the YRBS questionnaire that was used in the data set studied by Garofalo et al. (1998): the 1995 Massachusetts Youth Risk Behavior Survey. In that survey (and all the relevant information has been available on the Internet since 1996), the suicidality questions asked were as follows (with most results are given at - Specifying "In the past 12 months," students were asked: 1) if they seriously considered suicide, 2) if they planned an attempt, 3) if they attempted suicide, and 4) if an attempt "resulted in an injury, poisoning, or overdose that had to be treated by a doctor or nurse." In addition, but the results for this question are not reported on the cited web page, are the results for yet another suicidality question that has been standard fare in YRBS as designed by the CDC. This is Question # 29 on the Massachusetts YRBS Questionnaire located at - : "29. During the past 12 months, how many times did you actually attempt suicide?" Possible answers: "a. 0 times, b. 1 time, c. 2 or 3 times, d. 4 or 5 times, e. 6 or more times." Therefore, Savin-Williams is in "great error" with respect to what he asserts about YRBS suicidality questions. There are five questions as opposed to the two that he asserts to only exist, for a 150 percent magnitude of error. Readers, however, may believe what is asserted by Savin-Williams, likely because they would be assuming that he is informed on the subject. It is true, however, that Garofalo et al. (1998) only reported on the results for two variables, but that is not a problem in the survey instrument that Savin-Williams claimed it to be. In addition, Savin-Williams did not report that Youth Risk Behavior Surveys have been tested for their reliability by Brener et al. (1995), choosing instead to emphasise yet another problem: the apparent lack of reliability testing for these survey instruments (p. 989). The Savin-Williams questions on "suicidality" were also remarkably similar to the YRBS questions that he criticised on the basis of limited knowledge about these questions. In his study there was a difference in the order of the questions. Savin-Williams placed the "Have you attempted suicide?" question first, followed by four questions. The first was for those who did not report attempting suicide, followed by a request for the number of times suicide was attempted, one's age when the suicide attempt(s) occurred, and the factors that were associated with the suicide attempt(s). These questions were followed by

"Based on the following scale, circle the number that corresponds to this attempt:

One objective in suicidality research is to explore variables that help in the identification of "at risk" youth. There is little doubt that male youth self-identifying or self-labelling their homosexuality in some way are significantly more "at risk" for suicidality than are "other students," or self-identified "heterosexual" students. This is true for those who self-identify as "gay," "homosexual," or "bisexual" and/or the ones reporting being homosexually active and willing to acknowledge this to a researcher in interview situations or on a questionnaire. Results from non-random community samples and from more random population samples, including the ones that have produced results while controlling for many possible confounding variables, have repeatedly reported similar results. It would seem, however, according to Savin-Williams, that the "sexual orientation" identifying variables traditionally used should be changed, maybe to something resembling the 'Kinsey Scale' that he used (misused?) in an apparent effort to challenge the idea that sexual-minority youth are at higher risk for suicide problems. But then, the definition of 'sexual minority' has been so dramatically changed by Savin-Williams that it is not the original "at risk" population that is being commented on.

A major assumption being made by Savin-Williams (and others using the same results) is that gay, lesbian, and bisexual identified youth apparently report more false attempts than heterosexual youth. From this it is concluded that these same GLB youth are much less at risk for suicidality than previously believed on the basis of past research work. This is highlighted, for example, in the title of November 26, 2001 article in USA Today by Marilyn Elias: "LGBT Teen Suicide Less Than Thought, Report Says," and Savin-Williams is widely quoted in the article that begins with:

"Gay and lesbian teenagers are only slightly more likely than heterosexual kids to attempt suicide, contrary to past studies that suggest gay youths have about triple the rate of trying suicide, says a Cornell University psychologist in a controversial report due next month."
The following citation from the same article illustrates how the expanded 50%-of-study-population definition of "sexual minority" is equated with the concept of "gay youth," even in Savin-Williams' mind:
"The other study of 266 college men and women found that gay youths were not significantly more likely than straight classmates to have tried to take their own lives. Again, the homosexual students were more likely to report "attempts" that further questioning revealed as thoughts. 'They're trying to communicate that they do have difficult lives,' Savin-Williams says. 'But most gay kids are healthy and resilient.' Poorly designed studies that exaggerate their suicide risk 'pathologize gay youth, and that's not fair to them,' he says."
For Savin-Williams, there is an equation between GLB youth "attempting suicide" and "pathology," but is it true that GLB youth that attempt suicide are "pathological"? The concern here does not seem to be for GLB youth that are "at risk" because, without doubt, telling them that they are "pathological" for having attempted suicide will not be helpful. The opposite is more likely. Equating their "suicide attempt" with "pathology" or being "mentally disordered" will not help "at risk" GLB youth. Harmful outcomes may also result if they are inferred or deemed not to be healthy nor resilient, as Savin-Williams does when he states: "most gay kids are healthy and resilient" [apparently because they have not attempted suicide]. In addition to this, what would the most "at risk" GLB youth be thinking when reading such statements? Could the following be what is implied?
Reporting that you have attempted suicide (and many of you are liars in this respect, as my research shows, and making sure that this fact becomes public knowledge) has been causing an even more serious problem. All because of you attempting suicide, your pathology, and your lack of resilience, all the other GLB youth are made to look bad because they are now believed to be pathological like you! This is all so unfair to healthy and resilient gay youth!
Savin-Williams major concern does not appear to be for "at risk" GLB youth that he seeks to pathologise and stigmatise, but for the GLB youth who do not attempt suicide. "Not attempting suicide" seems to underlie his definition of resilience and health. For Savin-Williams, it is these healthy and resilient GLB youth who are being treated so unfairly, all apparently because of the results of "poorly designed studies," meaning that they are methodologically flawed. Many youth suicide studies are not perfect, but, as with YRBS, they are not what Savin-Williams has claimed by seriously misrepresenting what actually were the suicidality questions asked in the best of these studies. The Savin-Williams' assertion that GLB youth lie more than heterosexual youth when, in studies, they report having attempting suicide, is also an assertion not supported by his data. Even more serious is his lack of reporting that 'sexual minority' males accounted for 83.3% of all "true male suicide attempters, choosing instead to report that they were not significantly more at risk (p. 988). This assertion was also only made possible by not mentioning that his sample size for males precluded any statistically significant differences if the higher relative risk factor for a "true suicide attempt" had not reached "7".

Finally, when "at risk" GLB youth do report having attempted suicide, will it be helpful or harmful to them if professionals uncritically accept the Savin-Williams' assertions, as they are reported in his paper, or as they were reported in the media, and behave accordingly? What could be the responses to GLB youth reporting having attempting suicide if, for example, school counsellors and mental health professionals are 'educated' to believe that they are more likely to lie about having attempted suicide?

2. Russell ST, Joyner K (2001). Adolescent sexual orientation and suicide risk: evidence from a national study. American Journal of Public Health, 91(8): 1276-1281. Supplementary information from: Russell ST (2002, In Press). Substance use and abuse and mental health among sexual minority youth: Evidence from Add Health. In: Allen Omoto, Howard Kurtzman, Eds. Recent Research on Sexual Orientation, Mental Health, and Substance Use. A digital copy of the paper was supplied by Dr. Russell.

Borowski IW, Ireland M, Resnick MD (2001). Adolescent suicide attempts: risk and protective factors. Pediatrics, 107(3): 485-93.

The Russell ST, Joyner K (2001) and Borowski et al., (2001) studies both reported results based on 'sexual orientation' in the National Longitudinal Study of Adolescent Health (Add Health) survey. The survey was administered in two time periods called "Wave 1" and "Wave 2" and "Time 1" and "Time 2," respectively, in both studies. The study sample was representative of the adolescent population in Grades 7 to 12 who are living at home, and a total of 13,130 students answered an in-home interview administered via a laptop computer in 1995-96. A total of 13,110 students completed both Wave 1 and Wave 2 questionnaires. The two surveys were separated by an average time period of 11.0 months with a 95% Confidence Interval ranging from 7.6 months to 14.3 months (Borowski et al., 2001: 485-6), but Russell (2002) gives the time interval to be 18 months, with a time range not given. The former is tentatively accepted to be correct.

At Wave 1: The questions asked to apparently determine one "sexual orientation" were (Russell and Joyner, 2001; Russell, 2002, Add Health Data Codebook, Wave 1):

Have you ever had a romantic attraction to a female?
Have you ever had a romantic attraction to a male?

"In the last 18 months, have you had a romantic relationship with anyone?" and "What is their sex?"

At Wave 2: The "sexual orientation" questions asked were (Russell, 2002; Add Health Data Codebook, Wave 2):
Since the last survey,
Have you ever had a romantic attraction to a female?
Have you ever had a romantic attraction to a male?

"In the last 18 months, have you had a romantic relationship with anyone?" and "What is their sex?"

The "suicidality" questions asked at Wave 1 was (Russell and Joyner, 2001: 1277; Add Health Data Codebook, Wave 1):
"seriously thinking about suicide in the past 12 months"

"During the past 12 months, did you ever seriously think about committing suicide?" (From Codebook)

"number of times suicide was actually attempted in the past 12 months"

"During the past 12 months, how many times did you actually attempt suicide?" (From Codebook)

At Wave 2: The "suicidality" question asked was (Borowski et al., 2001: 486; Codebook, Wave 2):
"During the past 12 months, did you actually attempt suicide?"

"During the past 12 months, how many times did you actually attempt suicide?" (From Codebook)

In The Russell and Joyner (2001) study, there are problems with the information presented. Male and female 'sexual minority' adolescents were determined by adding all males reporting same-sex attractions and/or relationships, and doing the same for females, the authors noting a "0.05%" overlap in each group. This appears to be correct for females given the data (Table B-1), but incorrect for males. For females the overlap would be 0.4%, but an overlap of 0.0% applies for males.
Table B-1 - National Longitudinal Study of Adolescent Health: 
The 'Sexual Minority' Adolescents Samples in Wave 1
Russell and Joyner (2001)
% Percent
Same-sex Romantic
Same-Sex Romantic
Total: Same-sex Romantic
Attraction and Relationships

Russell and Joyner (2002: 1277-8, Table 1)

"5% of boys and girls reported both same-sex romantic attraction and relationship" (p. 1277)

The problem with these overlaps, or lack of overlap, may have implication with respect to what the word "romantic" meant to some of the study subjects. For females, the 0.4% overlap means that 80% of females (1.6/2.0) reporting "same-sex romantic relationship(s)" in the last 18 months did not report having ever experienced "same-sex romantic attraction." For males, the data indicates that all males (100%) reporting "same-sex romantic relationship(s)" in the last 18 months did not report having ever experienced "same-sex romantic attraction." If, however, the "0.5%" overlap applies for males, as it is reported by the authors (this would mean that reported numbers are in error), there would still be 54.5 percent of males (0.6/1.1) reporting "same-sex romantic relationship(s)" in the last 18 months who do not report ever having experienced "same-sex romantic attraction."

At issue is the problem with adolescents reporting same-sex relationships. That is, what would a researcher think if 100 adolescent females had reported having a "same-sex romantic relationship" but that only 20 of them (20%) report having ever experienced "same-sex romantic attractions"? The situation would be worse for males, if 100 percent of the ones reporting "same-sex romantic relationships" were reporting never having experienced "same-sex romantic attractions." This problem is serious, possibly indicating that most adolescents reporting "same-sex romantic relationships" did not understand the meaning of the word "romantic." If so, the problem indicated is one of "validity" for the variable. However, if the adolescent did understand the meaning of "romantic," this would mean that all boys (or most boys), and 80 percent of girls who are involved in "same-sex romantic relationships" are not in these relationships because they are romantically attracted to the other person. If so, the only option left is the proposition that it is the other same-sex individuals in the relationship who has "romantic feelings," thus making it possible to describe these relationships being "romantic" in nature. In this case, the question is, Why are most adolescents reporting "same-sex romantic relationships" doing in these relationships given that he have never experienced same sex romantic attractions? Would/Could they be users/abusers of the individual who has romantic feelings for them? Another troubling question related to these data is the fact that most researchers would also assume that adolescents who report being in "same-sex romantic relationships" would likely be self-labelling as gay, lesbian, or bisexual, as opposed to the ones reporting only "same-sex-romantic attraction."

Russell and Joyner (2001) used the combined categories noted above to separate out a group of adolescents that apparently have a "sexual orientation" (as suggested in the paper's title) that is different from the others with whom they will be compared. The variables to be used in their statistical analysis for males are given in Table B-2. The Risk Ratio were not supplied by the authors and were estimated via the conversion of percentages given into counts, and using a 2X2 Table calculation.

Table B-2 - National Longitudinal Study of Adolescent Health: Incidences
For All Males with Same-Sex "Romantic Attraction / Relationship"
versus All Others in Wave 1 (Russell and Joyner, 2001)
Categories à
Adolescent Males: Grades 7 to 12 (N = 5.686)
RAR: 8.4%
Incidence % / Mean M
No Same-Sex*
RAR: 91.6%
Incidence % / Mean M
Suicidal Thoughts (12 months)
1.6 a
Suicide Attempt(s) (12 months)
2.5 a
S. Attempt by Family Member
2.7 a
Suicide Attempt by Friend
1.4 b
1.2 b
Alcohol Abuse
2.7 (M)
1.6 (M)
6.02 (M)
5.10 (M)
2.18 (M)
2.18 (M)
*Same-Sex Romantic Attraction and/or Relationship (Same-Sex RAR): 8.4% (n = 453, Weighted). No Same-Sex Romantic Attraction and/or Relationship (No Same-Sex RAR): 91.6% (n = 5,233, Weighted).

**Risk Ratios are not given. Calculated estimates by Pierre Tremblay. Statistical Significance: a p < 0.000, b p < 0.01, c p < 0.05

***A t-test of significance is not given for difference in the Means (M).

Russell and Joyner (2001) reported the results of two logistic regression analyses for males in which the dependent variables are "serious suicidal thoughts" and "attempting suicide" (reported to have occurred in the past 12 months) obtained from the Wave 1 survey. The result of the logistic regression analyses (Predictor Odd Ratios) are given in Table B-3.
Table B-3 - National Longitudinal Study of Adolescent Health
Regression Analyses For All Males: Wave 1 Data
(Russell and Joyner, 2001)
Categories à
Adolescent Males: Grades 7 to 12 (N = 5.686)
Suicidal Thoughts
ORs, 95% CI
Suicide Attempt(s)
ORs, 95% CI
Logistic Regression #1
Same-Sex Romantic AR
Logistic Regression #2
Same-Sex Romantic AR
S. Attempt by Family Member
Suicide Attempt by Friend
Alcohol Abuse
Logistic Regression #1: Only males reporting Same-Sex Attraction and/or Relationships in Model.

Logistic Regression #2: All listed variables in the model.

Logistic Regression #1,2: Controlling for race/ethnicity (Black, Asian, Hispanic), for age, parental education, poverty status (on welfare or not), and both biological parents being in the home.

The Russell and Joyner (2001) regression analysis produced a predictor Odd Ratio of 2.4 (Table B-3) for all males, and this Odd Ratio is lower than those produced for three categories of males (White, Black, Hispanic) by Borowski et al. (2001) in their logistic regression analysis (Table B-4). In the latter study, however, the "attempted suicide" variable is from the Wave 2 survey, meaning that it represents the relative suicidality situation existing about a year later, and their same-sex oriented group only included adolescents reporting "same-sex romantic attraction." The Odd Ratio for the three groups of males based on 'race/ethnicity' (2.7, 2.9, 3.4) and, even though their independent variables from the Wave 1 survey are greater in number and often different than the ones used by Russell and Joyner (2001), these results suggest that the odds or risk for attempting suicide by males reporting same-sex attraction increased over a one-year period.
Table B-4 - National Longitudinal Study of Adolescent Health
Logistic Regression Analysis for Males Attempting Suicide
(Borowski et al., 2001)
Race/Ethnic Groups à
/ Variables
Adolescent Males: Grades 7 to 12
White ORs*
Black ORs*
Hispanic ORs*
Same-Sex Romantic Attraction
2.7 b
3.4 c
2.9 c
Suicidal behaviour: friend
11.3 a
4.8 b
Suicidal behaviour: family member
4.5 b
33.7 a
Easy family access to guns
2.5 c
3.0 c
Suicide Attempt at Wave 1
19.5 a
28.7 a
63.6 a
Mental health treatment
5.1 a
Somatic Symptoms
36.4 a
Poor general health
7.1 a
Weight dissatisfaction
Repeated a grade
School problems
4.4 a
7.2 b
6.9 a
Skipping school
2.9 b
Skipped a grade
3.7 b
Violence victimisation
6.0 a
6.6 a
3.3 a
Weapon carrying
3.9 a
13.9 a
5.5 a
Violence perpetration
4.4 a
4.6 a
4.3 a
Alcohol use
6.3 a
6.2 b
10.8 a
Marijuana use
6.8 a
5.9 a
2.9 c
Other illicit drug use
3.6 a
7.8 a
*Logistic Regression Analysis done separately for the three race/ethnic groups.
Odd ratios reported and the Statistical Significance: a p < 0.000, b p < 0.01, c p < 0.05
Controlling for age, both biological parents being in the home, and poverty status (on welfare or not).
Attempting Suicide in the past 12 months reported in the Time 2 survey. All Independent Variables are from the Time 1 survey.
The results from both studies, however, are based on "romantic" measures for "sexual orientation," while the results in most published homosexuality related youth studies of suicidality have used 'sexual orientation" measures based on gay, lesbian, bisexual self-identification and/or labelling, and/or reports of having engaged in same-sex sexual activity. This would mean that results from the studies in question are not comparable with other studies, thus producing an interesting question. Why did the professionals who designed the Add Health study chose to solicit "sexual orientation" information that could not produce results that are comparable with results from all other studies?

The problems are many with both studies of the Add Health data and the results indicate that "attempting suicide" for 'someone' is being explored, but who are these apparently "same-sex" oriented adolescent males? We only know that all males in the Borowski et al. (2001) study, and most males in the Russell and Joyner (2001) study were reporting "same-sex romantic attraction" in the Wave 1 survey. It would therefore certainly be interesting to also know how many of the males reporting "same-sex attraction" in the Wave 1 survey were also reporting the "same sex attraction" having occurred in the past year (since the Wave 1 survey) when asked the question in the Wave 2 survey. Doing this is possible given that the subjects had identification numbers and that Russell (2002, In Press) reports relevant data (Table B-5).

Table B-5 - Romantic Attractions Reported By Males
Russell (2002, In Press)
Rows: Wave 1 (counts), Columns: Wave 2 (counts)

Attractions to:





Same- / 


Neither sex
Same- / Both-sex
Attempted Suicide (Past 12 months) in "Romantic Attraction" Category for Two Waves (in Counts)
"Same-Sex / Both-Sex" means "Same-Sex Attraction."
Table B-5 Results:

Out of 485 males (7.1% of males) reporting lifetime "Same-Sex / Both-Sex" Romantic Attraction in the Wave 1 survey, only 16.1 percent of them (78/485) will be in the "Same-Sex / Both-Sex" Attraction category in Wave 2 because they experienced "Same-Sex / Both-Sex" Romantic Attraction since the Wave 1 survey (Mean = 11 months later: 95%CI: 7.6-14.3 months). That is, 83.9 percent of the males who reported ever having had same-sex romantic attraction in the Wave 1 survey would not be reporting having had such a response to a male in the upcoming year.

Out of 294 males (4.4% of males) reporting "Same-Sex" Romantic Attraction in the Wave 2 survey ("Same Sex" Romantic Attraction that had occurred since the Wave 1 survey: Mean = 11 months, 95% CI: 7.6-14.3 months), only 26.5% (78/294) of these males had reported lifetime "Same-Sex" Attraction in the Wave 1 survey. Most males reporting "Same-Sex" Attraction in Wave 2 survey, or 73.5 percent of males (78/294) had never experienced such a response before, meaning up to the time the Wave 1 survey was done. Furthermore, 59.5% of these males (175/29) were reporting only lifetime "Other-Sex" Romantic Attractions in Wave 1 survey.

The results for "Same-Sex" Romantic Relationships are similar and are not given here.

An important conclusion stemming from the information presented is that "same-sex romantic attraction" and/or "same-sex romantic relationships" should not be used as a measure of "sexual orientation," even if they may produce results indicating that the category is at risk. If used, they should only be used in association with others measures that use words well known to mean "a same-sex sexual orientation" in North America such as gay, lesbian, and bisexual. Information about same-sex sexual activity should also be solicited. Researchers could, however, produced results including all adolescents who reported same-sex romantic attractions in the Wave 1 and Wave 2 surveys to see how the results compare or if, as a group, they would be at greater risk for suicidality than males always only reporting opposite-sex attractions.

3. Cochran, Susan & Mays, Vickie (2000). Lifetime prevalence of suicide symptoms and affective disorders among men reporting same-sex sexual partners: results from NHANES III. American Journal of Public Health, 90(4), 573-78.

The NHANES III (National Health and Nutrition Examination Survey) "is a periodic population based health survey conducted by the National Center for Health Statistics of the Centers for Disease Control and Prevention" (p. 574). "S. D. Cochran conceived and designed the study, conducted the analysis, and co-wrote the manuscript. V. M. Mays cowrote the manuscript and contributed to the interpretation of study findings" (p. 577).

Research Highlights

A subsample of the NHANES III sample (N = 3648 males, 17- to 39-years-old) were interviewed in their homes. They were "assessed for both prevalence of affective disorders and related symptoms and genders of their lifetime sexual partners." Only 2.2 percent of males (defined to be homosexually oriented) acknowledged having "any male sex partners in their lifetime." This group was reported to have a lifetime suicide attempt incidence of 19.3%, compared to a 3.6% incidence for males who reporting having had only female sexual partners, and compared to a 0.5% incidence for males reporting having had no sexual partners.

For homosexually oriented males, 98% of the ones reporting a suicide attempt were 17- to 29-years-old, but heterosexually active males who had attempted suicide did not have such a skewed distribution. Their incidence for having attempted suicide was unrelated to the two age categories: 17 to 29, and 30 to 39 years.

The homosexually oriented males (any male sex partner in their lifetime) would therefore be 5.4 times (19.3 vs. 3.6%) more likely to have attempted suicide than males reporting having had only female sex partners. The reported logistic regression predictor is: OR, 95% CI: 2.21<5.36<12.98, controlling for age, ethnicity, and income. The greater odds for recurrent Major Depression and for "any affective disorder" for homosexually oriented males compared to males reporting only females sex partners was 3.5 times (12.2 vs 3.5%) with a predictor OR, CI 95%: 0.93<3.64<14.2, and 2.4 times (21.5 vs 8.8%) with a predictor OR, CI 95%: 0.90<2.55<7.24). Both ORs were not statistically significant (Cochran & Mays, 2000: 576, Tables 2 & 3). (Relevant information is given in Table B-6.) It is also reported that homosexually oriented males had significantly higher composite scores for suicide-related symptoms.

Table B-6 - NHANES III Survey: Cochran & Mays (2000) Study Results
Males with Same-Sex, Opposite-Sex, and No-Sex Partners
Age = 17 to 39 Years
Category à
/ Age Group
N = 78
Only Opposite-
Sex Partners
N = 3,214
N = 211
Age = 17-39 Years

Attempted Suicide %

15/78 = 19.3%
116/3,214 = 3.6%
Suicide Attempters

(Estimate for "n")

15.05 = 15
115.7 = 116

Same-Sex versus Opposite-Sex (Logistic Regression Analysis) - OR, 95% CI: 2.7<6.4<15.2
Same Analysis, controlling for age, race/ethnicity, and family income: - OR, 95% CI: 2.2<5.4<13.0
Cochran & Mays (2000: 576, Table 3)
Same-Sex versus Opposite-Sex - Estimated OR, 2X2 Table: 3.5<6.4<11.5 - X2 = 48.6, p = 0.0000

The authors note that almost all suicide attempters (98%) in the male group reporting same-sex partners are below the age of 30 years, with little said to explain their relatively great absence in the 30- to 39-year-old category. This anomaly may well be related to the AIDS factor given that its greatest effect (related illness and deaths) occurred in the 30- to 39-year-old category and during the sampling/study years: 1988-94. The authors did not address this issue in spite of the greatest difference in "suicide attempt" incidence is not being between males reporting "Same-Sex Partners" compared to males reporting "Only Opposite-Sex Partners" (19.3% vs. 3.6%: a "5.4 times" factor). The greatest difference is between males reporting "Same-Sex Partners" compared to males reporting in the 17-29 year-old category versus those in the 30-39 year-old categories: 31.3% (14.7/47) vs. 0.97% (.3/31), a "32.3-times" factor difference. (The fraction of an attempter is related to weighting.) Related calculations were done using 2 X 2 Table located at - 14.7 attempters / 32.3 non-attempters (age = 17 to 29 years) versus 0.3 attempters / 30.7 non-attempters (age = 30 to 39 years): Risk Ratio = .90<32.3<1166.6; Odd Ratio = 1.2<46.6<1789.0 (X2 = 11.05, p = 0.0008). These figures are based on the weighted data given by Cochran & Mays (2000) and this information also made possible the "estimated" re-calculation of "suicide attempt" data for males in the 17 to 29 and 30 to 39 Years category (Table B-7).

The Odd Ratios for males reporting "Same-Sex Partners" compared to males reporting "Only Opposite-Sex Partners" in the 17 to 29 Years category (OR, 95% CI: 6.2<12.1<23.6; X2 = 83.3, p = 0.0000), and in the 30 to 39 Years category (.007<.26<9.6; X2 = .61, p = 0.43) are radically different. In fact, they are more like opposites, suggesting that the two groups should not have been combined (Table B-7).

Table B-7 - NHANES III Survey: Cochran & Mays (2000) Study Results
Males with Same-Sex, Opposite-Sex, and No-Sex Partners
Age = 17 to 29 & 30 to 39 Years
Category à
/ Age Groups
N = 78
Only Opposite-
Sex Partners
N = 3214
No Sex
N = 211
Age = 17-29 Years
Attempted Suicide %
14.7/47 = 31.3%
47 = 60.7 % of males
in Sexual Category
60/1,655 = 3.6%
1,655 = 51.5% of males
in Sexual Category
1/202 = 0.50%
202 = 95.7% of males
in Sexual Category
Suicide Attempters
(n) Estimated
(98% of Attempters
in Sexual Category)
59.9 = 60
51.5% of Attempters 
in Sexual Category)
(100% of Attempters 
in Sexual Category)
Age = 30-39 Years

Attempted Suicide %

0.3/31 = 0.97%
31 = 39.3 % of males
in Sexual Category
56/1,559 = 3.6%
1,559 = 48.5% of males
in Sexual Category
0/9 = 0.0%
9 = 4.2% of males
in Sexual Category
Suicide Attempters

(n) Estimated

(2% of Attempters
in Sexual Category)
56.1 = 56
48.5% of Attempters
in Sexual Category)
(0.0% of Attempters 
in Sexual Category)
Percentage of males in each age category for each sexual category, and numbers in each sexual category is given by Cochran & Mays (2000: 575, Table 1)
"Same Sex Partners" versus "Only Opposite Sex Partners," 17 to 29 Years of Age - OR, 95% CI: 6.2<12.1<23.6; X2 = 83.3, p = 0.0000
"Same Sex Partners" versus "Only Opposite Sex Partners," 30 to 39 Years of Age - OR, 95% CI: .007<.26<9.6; X2 = .61, p = 0.43 - or reversed, "Only Opposite Sex Partners" versus "Same Sex Partners," 30 to 39 Years of Age - OR, 95% CI: .10<3.8<140.4; X2 = .61, p = 0.43
"Same Sex Partners" (Age = 17-29 Years) versus "Same Sex Partners" (Age = 30-39 Years) - OR, 95% CI: 1.2<46.6<1789.0; X2 = 11.05, p = 0.0008
In Table B-8, the results are given for males reporting "Same-Sex Partners" compared to males reporting "Only Opposite-Sex Partners" combined with males reporting "no Sex Partners in the 17 to 29 Years category (OR, 95% CI: 6.9<13.4<26.1; X2 = 94.079, p = 0.0000). This Odds Ratio applies for a sample of males ranging in age from 17 to 29 years, with a mean of 23 years and it is close to the Odds Ratios (OR, 95% CI: 3.4<14.4<61.4; X2 = 22.1, p = 0.0000) produced in the Bagley & Tremblay (1997) study of males ranging in age from 18 to 27 years, with a mean of 22.7 years.
Table B-8 - NHANES III Survey: Cochran & Mays (2000) Study Results
Males with Same-Sex Partners, All Other Males - Age = 17 to 29
Category à
/ Age Groups
N = 78
Males with Only Opposite-Sex Partners
And Males with No Sex Partner (all Other Males) - N = 3214 + 211 = 3,525
Age = 17-29 Years
Attempted Suicide %
14.7/47 = 31.3%
47 = 60.7 % of males
in Sexual Category
61/1,857 = 3.3%
1,857 = 52.7% of males in 
Sexual Category
Suicide Attempters
(n) Estimated
(98% of Attempters
in Sexual Category)
(52.1% of Attempters in 
Sexual Category)
"Same Sex Partners" versus "Only Opposite Sex Partners," 17 to 29 Years of Age - OR, 95% CI: 6.9<13.4<26.1; X2 = 94.079, p = 0.0000 After the weighting/controlling statistical processes, the 31 males reporting "Same-Sex Partners" in the 30 to 39 years of age category have only 2% of attempters (0.3 of an attempter). This would be equal to a lifetime suicide attempt incidence of 1%, compared to an estimated 3.6% incidence for males reporting "Only Opposite-Sex Partners" in the same age group. Therefore, the latter have been about 4-times more likely to have attempted suicide, compared to the males reporting "Same-Sex Partners." This "1%" also means that the older (30-39 years) segment of males reporting "Same-Sex Partners" had a suicide attempt incidence about 31 times lower that their younger counterparts. However, this highly anomalous situation was not highlighted nor addressed by the authors. One important implication would be that the older homosexually oriented males of a particular decade had almost stopped attempting suicide, and were also at much lesser risk for attempting suicide than heterosexual males. This possibility, however, is unlikely given the consistent "higher risk" for attempting suicide results for homosexually oriented males studied in samples dating back to 1960s, producing data stemming back to the 1930s. Therefore, the "representation" likelihood of the older part of the 'homosexual' male sample (30- to 39-year-old) is in question. The authors not only ignored this, but they also avoided discussing this serious problem occurring in data to be statistically analysed.

It would seem, given the count distribution, that the authors chose to place all male together for analysis to produce an OR that would be much lower compared to the OR that would have resulted if the younger males has been separated for analysis. The reasons for this could/should have been explained by the authors, first by noting what the OR would be for both age groups analyses separately (as given in Table B-3). Then by explaining why the regression analyses results would be produced only for the two age groups (spanning a 23-year period) analysed together. In suicidology, such age ranges are anomalous. As a rule, suicide rates are given for age ranges of 5 years (standard practice), and sometimes for 10-year.

It is also most interesting that Cochran and Mays (2000) cited the male homosexuality related suicidality results of the American Herell et al. (1999) study and the male-female homosexuality results from the New Zealand Fergusson et al. (1999) study. This observation is made because both studies also referenced the Canadian Bagley and Tremblay (1997) study and that Cochran and Mays (2000) did not reference the Canadian study that they would have known about, and the demographic results for the same study sample published in 1998. (Both studies are referenced in PubMed & Medline.) Had they done this, however, they may then have had to caveat their results (The ORs of 5 to 6) in the direction of a significant underestimate strongly suggested by the Bagley and Tremblay (1997, 1998) study results that included noted "underestimating" caveats for the NHANES III type of studies. Instead, Cochran & Mays (2000) essentially selected data that would produce an OR that would replicate those of the two cited studies. Had they cited the Bagley and Tremblay (1997, 1998) studies, however, they would maybe have had to report the results for homosexual males compared to all other males in the 17 to 29 year of age category, noting that the results are a replication of the Calgary study results for a group of males who had approximately the same mean age as the younger part of their sample. For the sample chosen for analysis, the ages ranged from 17 to 39 years, with a mean age of about 27 that is quite different from that in the two cited studies. The New Zealand study was to the age of 21 years for males and females, and Herell et al. (1999) had studied middle-age men.

It is therefore possible, given Susan Cochran's publicly manifested concern that homosexually oriented males are not be perceived to be pathological on the basis of their "OR = 5 to 6" suicidality results (Freiberg, 2000), that she may not have wanted to mention the existence of study results indicating that homosexually oriented male youth may actually be at much higher risk for attempting suicide, and that they likely account for more than half of all male youth suicide attempters. This Bagley and Tremblay (1997) result cannot be missed because it was placed in the abstract, as was the replication nature of the study, thus highlighting the fact that these are not first-time maybe anomalous results. "Homosexually oriented males accounted for 62.5% of suicide attempters. These findings, which indicate that homosexual and bisexual males are 13.9 times more at risk for a serious suicide attempt, are consonant with previous findings" (Bagley and Tremblay, 1997: 24).

Finally, the Cochran and Mays (2000) incidence for a lifetime suicide attempt(s) for American homosexually oriented males ranging in age from 17 to 29 years is 31.3 percent (Table B-8). This estimate replicates the results obtained from nonrandom community samples of gay and bisexual male youth (average = 30%), and the results for gay and bisexual adolescent males produced in the Youth Risk Behavior Surveys (28 to 33%).

Previous Section
Top of Page
Email to Richard RamsayEmail to Pierre Tremblay