A re-analysis of data from Sullins, Rosik, and Santero (2021): are sexual orientation change efforts (SOCE) helpful for those who don't change?

Participants

The participants in this study were recruited through an online survey administered pursuant to the doctoral dissertation of Paul Santero (2011) at Southern California Seminary, which, as Sullins et al. (2021) have noted, “contains a more complete description of the survey methods, administration, and question wording.” (p. 3). Sullins et al. (2021) also reported that participants were “contacted through religious organizations and therapist networks who offered services including talk therapy, retreats, and support groups that serve this population” (pp. 3-4). The original participants included 25 men from countries outside the United States, but those men were not included in the analyses here or in Sullins et al. (2021). Respondents were distributed across the United States: west coast (24.0%), mountain west (27.2%), southwest (6.4%), south (8.8%), northeast (4.8%), east coast (14.4%), central (3.2%), with one response “having lived all over the USA” with no missing data for that information. The percentages here for location differ from those reported in Sullins et al. (2021) because two variables concerning residence were merged into one variable to eliminate apparent missing data. For example, in one case, the primary variable was not answered but the second variable answer was “Utah”, which was counted as “mountain west.” No women were among the 125 participants studied here, although 8 women had participated originally. Most of the sample was white (91%) and 73% had a college education or higher, while 58% reported a household income above $50,000. With respect to religious attendance, 88% reported weekly or more often.

Measurement

For the items and scales used in this report, Appendix A lists the mean, median, standard deviation, minimum, maximum, and total number of non-missing cases. All data used here were obtained from the data set provided by Sullins et al. (2021). The data set did not include any measures of social desirability response bias; however, a more detailed discussion of different forms of social desirability measurement is available elsewhere (Schumm, 2015).

Scales

Generally, multi-item scales are preferable to single item measures, at the very least because of the potential for increased measurement reliability in terms of internal consistency. There were three questions that concerned sexual behavior, kissing, and daydreaming about sex for both heterosexuality and homosexuality as well as prior to SOCE and during/after SOCE, a total of 12 items. Items about desiring emotional intimacy were added to each of the four scales to see if an additional, but possibly unrelated item, would improve reliability. Adequate-to-fair Cronbach alphas were obtained as follows: prior same-sex sexuality, 0.68; prior heterosexuality, 0.76; post same-sex sexuality, 0.68; and post heterosexuality, 0.79. Adding the fourth item changed the reliabilities, respectively, to 0.58 (a decrease), 0.76 (no change), 0.76 (an increase), and 0.82 (an increase).

Overall helpfulness of SOCE

There were 10 items that asked how helpful different forms of SOCE had been, with answers from not at all/none, slightly, moderately, markedly, and extremely. The forms of SOCE listed included psychiatrist, psychologist, social worker, mental health counselor, pastoral counselor, religious peer group, nonreligious peer group, weekend retreat, personal study, and mentoring. Some subjects used only one form of SOCE while others used multiple forms. To create an overall measure for each client, regardless of how many forms they had used or experienced, the highest rating given to any of the ten forms of SOCE was used to create one overall measure of SOCE helpfulness. Taking the opposite approach, using the worst results to create a scale, would have overlooked forms of SOCE that had been effective for the subjects – someone might try a lot of things to find help, but even if some don’t help, the one(s) that did help, would hopefully be those that mattered the most. For the 123 subjects who answered at least one of the helpfulness questions, the percentage results were none (1.6%), slightly (8.9%), moderately (11.4%), markedly (19.5%), and extremely (58.5%), with the values coded from 1 to 5, respectively. The mean score for overall helpfulness was 4.24 (SD = 1.07) with a median of 5.0. The item was significantly skewed, -1.27 (SE of skew = 0.22) with the distribution of responses more strongly favorable than unfavorable. Furthermore, a trinary measure was created by recoding the none/slightly, moderately/markedly, and extremely responses into three groups, which was intended to provide a type of control group (SOCE was not effective, as perhaps no treatment control group would have been) along with two levels of relative SOCE effectiveness. The results for this new item reflect what Sullins et al. (2021) concluded – that some subjects were not benefitted by SOCE but that the majority did seem to report beneficial outcomes.

Typology of sexual orientation before SOCE

Sexual orientation is usually expressed or described in terms of attraction, behavior, and identity. Would it be possible to develop a typology of sexual orientation of subjects prior to SOCE? Using median splits for sexual attraction, same-sex sexual behavior, and sexual identity, a typology of the three aspects of sexual orientation prior to SOCE was created. Six of the eight possible combinations were obtained for 123 subjects: low behavior, low attraction, and low identity (7.3%); low behavior, high attraction, low identity (30.9%); low behavior, high attraction, high identity (21.1%); high behavior, low attraction, low identity (4.1%); high behavior, high attraction, low identity (15.4%); and high behavior, high attraction, high identity (21.1%). A one-sample Chi-squared test (df = 5) of 36.2 indicated that the types were not equally distributed (p < .001).

Changes in marital status

Marital status was assessed for the time prior to SOCE and after/during SOCE. Sullins et al. (2021) reported that the men in their study were more likely to be single than in the United States in general but less likely to be divorced; their study seemed to assume that all marriages were heterosexual. Some subjects were single, married, or divorced/widowed before SOCE; some got married from the single state; some were divorced from the married status after/during SOCE. A new item was created with four categories: single (before and later), married (before and later), got married (changed from single to married), and divorced/widowed (either stayed in that category or changed to that category). The percentages of subjects (N = 125) in each category, respectively, were 52.8%, 26.4%, 15.2%, and 5.6%.

Maximum number of SOCE sessions

Respondents reported the number of SOCE sessions for each of the types of SOCE listed, which included sessions with a psychiatrist, a psychologist, a social worker, a mental health counselor, a pastoral counselor, with an ex-gay support group, and with a nonreligious support group. To create a summary variable, the maximum number of sessions for any of the types of SOCE was noted. In other words, if pastoral counseling involved 30 sessions and mental health counseling involved 50 sessions, the coding would be for 50 sessions. For two subjects, none of the seven types of counseling sessions were used. Thus, the breakdown for maximum number of sessions for the most used type of counseling was none (1.6%), 1-10 (4.8%), 11-25 (8.8%), 26-50 (16.0%), 51-100 (29.6%), 101-200 (12.0%), and more than 200 (27.2%). The logic was that the most used type was probably the one that subjects felt was most useful and therefore those number of sessions probably were most related to any favorable outcomes. Another measure of the maximum number of sessions was created by crediting the variable with the upper limit of each category and crediting ratings of over 200 sessions with a score of 250. While that approach might seem to overestimate the number of sessions, it is counterbalanced by the fact that many SOCE participants had sessions in multiple therapeutic approaches.

Analysis

Paired samples t-tests, linear regression, Chi-squared tests, and repeated measures analyses of variance will be used to evaluate these research questions, using SPSS version 26 (RRID: SCR 002865). An open access alternative is JASP (RRID: SCR 015823); however, numerous alternative free website sources for statistical analysis have been listed elsewhere (Schumm et al., 2021).

First question: effect sizes in terms of Cohen’s d

Sullins et al. (2021) reported nonparametric tests to compare client conditions before and after SOCE. Here, t-tests with Cohen’s d as a measure of effect size are reported. While all the variables used, including the scales, had significantly non-normal distributions, t-tests are generally robust with respect to such violations. However, the main point of the analyses here was to provide a widely recognized measure of effect size that also considered the correlated nature of the variables being compared to avoid underestimating the true effect sizes involved. Cohen (1992) indicated that an effect size of 0.50 or greater would be “visible to the naked eye of a careful observer” (p. 156), though Funder and Ozer (2019) noted that even small effect sizes (d = 0.20) were important (also see VanVoorhis & Morgan, 2007). Results are presented in Table 1.

It appears that the effect sizes detected range from 0.24 to 0.94, falling into the small to large effect size range. Same-sex sexual behavior decreased by more than twice as much as the increase in different-sex behavior (0.56 versus 0.24). Of the three main components of sexual orientation, attraction appeared to decrease the most (0.94) while same-sex sexual behavior (0.56) and same-sex sexual identity (0.60) appeared to decrease by about the same amount. Desired heterosexual emotional intimacy increased by about the same amount (0.68) as desired same-sex emotional intimacy decreased (.63). By Cohen’s (1992) definition, same-sex sexual attraction decreased to a large extent (≥ .80) while the other two aspects decreased by medium size effects (≥ .50). By Cohen’s definition, all three effects would be observable to a trained naked eye.

Second question: pre-SOCE conditions

Sullins et al. (2021) compared SOCE scores before and during/after but did not assess a typology of pre-SOCE sexual orientation. Tables 2-7 display the results obtained for apparent changes in four key aspects of sexual orientation: attraction, identity, same-sex behavior, and different-sex behavior.

Tables 2-7 predicting basic changes as a function of a typology of six pre-SOCE conditions.

In Table 2, the low/low/low group was small, and these subjects reported higher levels of heterosexual behavior than homosexual sexual behavior prior to SOCE. Homosexual behavior was very low before SOCE. Both sexual attraction and sexual identity were, on average, in the lower bisexual to heterosexual range. Though not significant with a two-tailed test (p < .10) due to the small sample size, decreases in homosexual sexual attraction and identity were in Cohen’s medium range (both 0.67). If one were to assume that SOCE would only decrease same-sex factors, such that a one-tailed test would be appropriate, then the results for sexual attraction and sexual identity would be statistically significant (p < .05).

In Table 3, pre-SOCE sexual attraction scores were high while same-sex sexual behavior was lower and lower than different-sex sexual behavior; sexual identity fell between the levels for sexual attraction and sexual behavior. Sexual identity changed little with SOCE for this group while sexual attraction changed substantially (d = 0.78, p < .001); the next largest effect size was for heterosexual behavior increasing (d = 0.33, p < .06), a small-to-medium effect that was not quite significant at the. 05 level. These subjects were probably more concerned about their levels of same-sex sexual attraction than anything else and, if so, the results suggest that their expectations were met, at least to some extent.

In Table 4, the subjects initially were high on same-sex attraction and identity but lower on same-sex sexual behavior. While same-sex sexual behavior changed little with SOCE, changes in both attraction and identity were well above the limit for large effect sizes (≥ .80) per Cohen (1992) with 0.95 for attraction and 1.07 for identity, both significant (p < .001). Changes in either same-sex or different-sex sexual behavior were not significant (p < .05), although their effect sizes were in the small-to-medium range (0.25 to 0.37).

In Table 5, the subjects before SOCE were lower on attraction and identity (being in a low bisexual range) but higher on same-sex sexual behavior. With only five cases, only the changes in sexual behavior were significant or near significant, even though the effect sizes were well over the limit for large effect sizes per Cohen (1992), at a remarkable 4.07 for homosexual behavior and 1.10 for heterosexual sexual behavior. Effect sizes for attraction and identity were, respectively, nearly large (0.73) and in the small-to-medium range (0.35), though not significant statistically. It is possible that these subjects entered SOCE with more concern with their sexual behavior than with their attractions or identities; hence, the SOCE may have focused more on behavior than the other factors.

In Table 6, the subjects began higher on same-sex sexual attraction than on identity or behavior; however, in terms of change, effect sizes were very large for both same-sex sexual attraction (1.36) and for gay sexual behavior (1.82), both significant at p < .001. Smaller changes occurred for heterosexual behavior (0.28) and for identity (0.72, p < .01). Notably, same-sex sexual behavior was at a higher level (d = 1.84) than different-sex behavior before SOCE but was lower than heterosexual behavior during/after SOCE (d = 0.17).

In Table 7, subjects entered SOCE with higher scores on all three aspects of homosexual sexuality. While changes in heterosexual sexual behavior were minimal (0.16), there were very large changes in sexual attraction (1.11, p < .001), sexual identity (1.18, p < .001), and in sexual behavior (1.75, p < .001). While homosexual behavior remained higher than heterosexual behavior at both time points, it declined from an effect size advantage of 3.88 to only 0.34, less than a tenth of the initial effect size difference.

Third question: initial conditions as main effects on SOCE helpfulness

Including prior sexual attraction, sexual identity, and same-sex sexual behavior as independent variables led to a non-significant overall degree of explained variance. After deleting sexual attraction as one of the independent variables, the adjusted R-squared was 0.04, with F(2, 199) = 3.52 (p = .033), with significant or near-significant results for the independent variables, prior same-sex sexual behavior (b = .17, p = .06) and sexual identity (b = -.19, p < .04). This would indicate that prior conditions had relatively little impact on therapeutic outcomes as reported by the subjects in this study, although there was a slight tendency, for higher levels of prior same-sex behavior (controlling for sexual identity) and lower levels of prior same-sex sexual identity (controlling for same-sex behavior) to predict slightly higher levels of reported therapeutic helpfulness.

Fourth question: effects of marital status

Sullins et al. (2021) also indicated that marital status played an important role in SOCE outcomes but I wanted to test that hypothesis more carefully, using a repeated-measures design and establishing a more detailed breakdown of marital status, including four categories: never married (N = 66), married (N = 33), got married or engaged (N = 19), and was divorced/widowed and stayed divorced/widowed or got divorced (N = 7).

The primary concern was whether there would be an interaction effect between apparent change in sexual orientation variables and this new version of marital status. For sexual identity, the interaction effect was not significant (p = .393). The main effect of marital status was not significant (p < .07). For sexual attraction, the interaction effect was not significant (p = .246) while the main effect of marital status was significant (p < .01). For same-sex sexual behavior, the interaction effect was not significant (p = .07) while the main effect of marital status was significant (p = .01). However, in terms of heterosexual sexual behavior, the main effect of time was less significant (p = .04) than the interaction term (p < .001) while the main effect of marital status was also very significant (p < .001). Therefore, the detailed results for heterosexual sexual behavior are presented in more detail (Table 8).

Fifth question: predicting helpfulness ascribed to most helpful SOCE from maximum sessions

Sullins et al. (2021) did not appear to have developed or used the two measures featured with our fifth research question. The curve estimation program under regression in SPSS was used to predict the perceived helpfulness of SOCE from the two measures of the maximum number of SOCE sessions. For the ordinal measure of sessions, the linear and quadratic model featured F(2, 120) = 4.51, p < .02 with an adjusted R-squared of 0.054. The linear standardized beta coefficient was 1.00 (p = .01) while the quadratic beta was -.84 (p < .03). The pattern suggested that maximum overall helpfulness peaked between 51-100 and 101-200 sessions. For the ratio measures of sessions, the linear and quadratic model featured F(2, 120) = 5.00 (p < .01), with an adjusted R-squared of 0.062. The linear standardized beta coefficient was 1.32 (p = .003) while the quadratic beta coefficient was -1.21 (p = .006). The pattern suggested that the maximum helpfulness peaked at about 160 sessions. For both measures, the stronger increase in helpfulness occurred between no sessions and the maximum while the decrease in helpfulness was not as strong after the maximum.

Sixth question: repeated measures analyses using a trinary measure of client-rated SOCE helpfulness

As noted in the introduction, it remains an open question if better SOCE ratings for providers will reflect change in sexual orientation. Perhaps, ratings will increase with increased changes in the direction of heterosexuality; perhaps, the reverse might occur. There are three groups with respect to client-rated helpfulness: none or slightly (N = 13), moderately or markedly (N = 38), and extremely (N = 71), labeled respectively as A, B, and C in the Tables 9 through 12.

For sexual attraction, the mean scores were as in the following Table 9.

For same-sex sexual behavior, the mean scores were as in the following Table 10.

For heterosexual sexual behavior, the mean scores were as in the following Table 11.

For sexual identity, the mean scores were as in the following Table 12.

For heterosexual sexual behavior, changes from SOCE were small and not significant statistically. However, for the three same-sex sexuality factors, the main effects over time and the interaction between time and the three levels of reported helpfulness were always significant (p < .01). For group A, the reported changes were consistently in a pro-gay direction, suggesting that, on average, changes in that direction were not perceived as helpful in terms of SOCE effectiveness. For groups B and C, greater changes in a heterosexual direction were perceived as more helpful. Both of these results suggest that SOCE’s effectiveness was judged on the basis of its association with desired changes in a heterosexual direction, even though changes in the direction of heterosexual behavior were smaller. For group B, changes tended to be from small-to-large; for group C, changes tended to be large, in terms of Cohen’s (1992) criteria.

Question 7: typology of change and reported helpfulness of SOCE based on sexual attraction and sexual identity before SOCE and after/during SOCE

A typology was created of the main types of change reported by participants crossed with a five-category measure of reported helpfulness of SOCE. The results are shown in Table 13.

For all categories, a Chi-square test (df = 56) = 90.96, p = .002, indicating differences among the several categories. Comparing the three subgroup counts, a Chi-square (df = 8) = 36.99, p < .001, r = .45, p < .001. Comparing the two largest subgroups, a Chi-square (df = 4) = 26.38, p < .001, r = .48, p < .001. A comparison of the two smallest subgroups was not statistically significant. A comparison of the largest subgroup with the smallest subgroup yielded a chi-square (df = 4) = 22.76, p < .001, r = .43, p < .001. For this last comparison Cramer’s V = 0.24, which represents a large effect size, as calculated using a website [https://www.statology.org/effect-size-chi-square/]. Notably, while those men who became more gay in general (91.7%) reported at least slight or better degrees of helpfulness of SOCE, a small percentage (8.3%) of that group of men reported that SOCE was not helpful at all.

Question 8: will reports of therapy outcome change as a function of recency of or time since SOCE?

Any time a respondent reports on past events, there may be recall bias or distortions of memory. Social desirability may contaminate reports of past or current events. Here the question was whether the time elapsed since having experienced SOCE would be associated with ratings of SOCE, which might be expected if recall bias was substantial (Table 14).

Results indicate that neither recall bias nor recency bias are predominant; the negative correlation suggests a slight trend in favor of recency bias. It is possible that both biases are operant but canceled each other out.

Ninth research question: what role does congruence between sexual attraction and sexual identity play in SOCE?

Before SOCE, SSA scores (mean, 5.74; SD, 1.10) were significantly greater (t₁₂₄ = 6.00, p < .001) than scores for SSI (mean, 4.80; SD, 1.76) with the scores correlated, r = .465, p < .001. After SOCE, SSA scores (mean, 4.14; SD, 1.76) were significantly greater (t₁₂₃ = 5.62, p < .001) than scores for SSI (mean, 3.60; SD, 1.95) with the scores correlated, r = .839, p < .001. After SOCE, SSA and SSI scores were closer and more highly correlated.

SSASSI congruence were positive when SSA > SSI and negative when SSA < SSI. Before SOCE, the scores ranged between -4 and 6, with a mean of 0.936, a median of zero, and SD = 1.74; after SOCE, the scores ranged between -1 and 4, with a mean of 0.540, a median of zero, and SD = 1.07. Before SOCE, 57.6% of the participants had scores of zero; after SOCE, it was 65.6%. Comparing congruence before versus after SOCE, t₁₂₃ = 2.95, p = .004, with a mean of 0.895 (SD = 1.69) before and a mean of 0.540 (SD = 1.07) after, indicating that congruence scores decreased after, although remaining highly correlated, r = 0.611 (p < .001).

To investigate whether or how congruence scores might matter, four subgroups of the total sample were created: no exact (SSI = SSA) congruence before or after SOCE, congruence before but not after, congruence after but not before, and congruence both before and after. Chi-square tests were performed for each of the four groups, crossing our trinary outcome variable versus post-SOCE sexual orientation identity split at bisexual to heterosexual orientation versus the more homosexual sexual identity orientations. None of the tests were significant except for the before and after congruence group. For that group, all the participants who reported no to only slight effectiveness, 100% (N = 11) were homosexually oriented. For those who reported moderate effectiveness, 22.7% (N = 5) were more heterosexually oriented compared to 77.3% (N = 17) who were homosexually oriented. For those who reported extremely effective, the respective percentages were 85.7% (N = 24) and 14.3% (N = 4). The chi-square test (df = 2) = 31.76, p < .001, with r = -0.697 (p < .001). Another way of reporting the results would be that 34.4% (11/32) of those who had stronger and congruent gay identities after SOCE indicated that SOCE was of only slight or no effectiveness compared to 82.8% (24/29) of those who had stronger and congruent more heterosexual identities indicated that their SOCE was extremely effective.

Tenth research question: positives and harms for the youngest participants, ages 18-25

There were 18 participants between the ages of 18 and 25 at the time of the survey. Of those 18, 55.6% were currently in SOCE, 22.2% had finished SOCE within the past year, 5.6% within one to two years, and 16.7% within the past two to five years. In terms of maximum sessions for the longest type of SOCE, 27.8% had been involved in a maximum of 50 sessions, 33.3% for a maximum of 100 sessions, and 38.9% for a maximum of 200 or more sessions. Thus, it is likely that some of these participants had begun SOCE before age 18 (Tables 15, 16).

For all four outcomes used here, results were significant (p < .01) and indicated that positive effects of SOCE exceeded negative effects or harms, as reported by the young participants themselves.

A supplementary analysis that was not planned along with the others was to evaluate the relative predictive strength of positive versus negative outcomes with respect to the reported helpfulness of SOCE, as shown in Table 17. In general, positive changes seemed to impact reported helpfulness more than negative changes did. The strongest effect occurred for self-esteem, which also featured the strongest impact of a positive change. Although the results for suicidality were limited by a large amount of missing data, the positive and negative changes were equal in magnitude and the negative change’s impact was the largest for all five outcomes, suggesting that SOCE providers need to be acutely aware of and continue to check on participants’ feelings and intentions in that area and also with respect to changes in depression.

Limitations

This report has many of the same limitations as Sullins et al. (2021). The sample is nonrandom, purposive, did not include women, and the data are over a decade old. The sample included men who were, as Sullins et al. (2021) stated, most often “white, affluent, well-educated, highly religious, and overrepresented the Mormon faith” (p. 14). In addition, our analyses used parametric statistics, but the assumption of normality was violated for most of our variables. However, many of our effect sizes are large enough that even if they were positively biased, they would usually still be of small-to-medium size. In general, more favorable results have been found for SOCE when the subjects were eager for treatment, highly religious, and married or planning to get married, as in this sample; less favorable, even harmful results have been found for SOCE when the subjects were pressured into experiencing SOCE, were not religious, were not married, and were high on same-sex sexual orientation identity at the time of the survey about past SOCE attempts. It is unlikely that these different populations will experience SOCE in the same way; results from one group should not be generalized to the other groups. Sullins et al. (2021) have presented other limitations that would also apply to this research.

Opportunities for independent analysis

Data sets held by Bondy (2021) and Pela and Sutton (2021), if released by their authors, might provide similar opportunities for independent analysis. Compared to Sullins et al. (2021), Bondy’s data features a larger sample size (N = 156) and includes more data on SSASSI congruence, extrinsic motivations to try SOCE, and client reported sources of SSA (including 44 reports of child sexual abuse; would those reports correlate with initial levels of SSA?). Bondy found that extrinsic motivations for SOCE predicted poorer ratings of SOCE experience, a result like other areas of counseling (e.g., Schumm & Denton (1979). Pela and Sutton’s (2021) data could be explored by age group to see how the youngest (age 18-25) responded to SOCE, by level of religiosity (did those with more frequent church attendance or citing religion as a reason to change report the most change?), or by level of desire for marriage (did a higher level of desire for marriage seem to affect the reported changes?). Even though the data indicated no average decline in mental health with SOCE, it might be useful to create a typology of those whose mental health did deteriorate, those whose remained similar, and those whose increased substantially and investigate if that typology would be associated with different SOCE motivations or outcomes.

Challenging identity theories

From this author’s perspective, there may be at least two “elephants in the living room.” First, there seems to be an assumption by some scholars that if you experience any degree, slight to large, of SSA, then you must adopt a gay, lesbian, or bisexual identity (SSI), assuming (1) there should be absolute congruence or else one risks becomes inauthentic as a human being and (2) that determination is not up to the individual but to a variety of others wielding some sort of social influence. It seems that for some, SOCE is in part about disagreeing with, breaking, or reframing those assumptions (Bondy, 2021). Some religious groups may interpret SSA as a “temptation” rather than an intrinsic part of oneself, relying on the religious premises that everyone (even Jesus) has been tempted to do unusual, strange, even wrong things but that only verbal or nonverbal behavior involves moral culpability (Sutton, 2019). As Karten & Wade (2010) noted, “there are men who are sexually attracted to other men but do not identify as gay; rather, they experience their homosexual orientation and behavior as at odds with who they really are” (p. 86). Furthermore, as the number of persons identifying as nonheterosexual has increased in recent decades (Pellicane & Ciesla, 2022), perhaps a greater number of individuals might describe themselves as mostly heterosexual or somewhat bisexual – and might find SOCE more relevant to their needs than those who would identify as exclusively gay or lesbian.

Basing one’s identity around a particular type of issue might be counterproductive for some, even if it provided self-justification or social justification for others. As an extreme example, believing in a flat earth might provide a person with an identity of sorts and a community of fellow believers, but is basing an identity on an incorrect belief helpful in the long run? An identity might inhibit a person’s perception of their ability to change when presented with more accurate facts. There is some evidence that a stronger identity as gay or lesbian may be associated with SOCE being ineffective, even counterproductive (Karten & Wade, 2010). The data here weakly suggested the same outcome, with a stronger LGB identity predicting less chance of change in attraction or behavior. My hypothesis here is that encouraging persons to assume nearly any identity (gay, Muslim, Christian, transgender, evangelical, tea party, professional, etc.) may be motivated for sociological or political purposes as much or more than for the personal or psychological benefit of those persons. For example, if a person is same-sex attracted but does not want to identify as gay/lesbian, what right would I have to force him (her) to do so or to try to prohibit him (her) from doing so? Should it not be his (her) choice on how to identify? For an SSA person to identify accordingly, may not necessarily be in their best interest (Pela & Sutton, 2021, p. 63; Diamond, 2008). While I think it’s fair to discuss the possible or likely advantages and disadvantages of any given identification, I don’t think it would be right for me to impose my desires about the other person’s potential identifications on them, even if it would make me feel better if they identified in a way I might prefer. Moreover, I am not sure the other person should assume a right to demand that I accept their identification, in part because many of the labels we might attach to ourselves can change over short or long times for a variety of both anticipated and unanticipated reasons.

The second elephant is something both sides in the debate over SOCE seem to overlook. Lesbian, gay, and bisexual behavior and identity can have many positive advantages, especially from a near-term perspective (Schumm, 2018, pp. 31-46; Schumm, 2020). As one gay youth said, “We have all the fun!” (Russell, 2003, p. 1253) or as a famous actress, Cynthia Nixon, said, “I’ve been straight and I’ve been gay, and gay is better” (Schumm, 2018, p. 39, citing from Diamond & Rosky, 2016, p. 382). A recent study found that lesbian and gay persons, especially those married, had higher socioeconomic status than heterosexual persons (Elwood et al., 2020), at least in California. In my view, these advantages should be openly discussed in SOCE; it may well be that some SOCE subjects may realize that they enjoy and want the advantages of a nonheterosexual lifestyle and should feel free to continue or move in that direction, without apology. Given such numerous advantages, LGB persons may experience guilt from being “over benefitted”, a prediction from social exchange theory, guilt that might account for internalized homophobia or lowered psychosocial health rather than or in addition to sexual minority stress.

It may also be that the difficulty found in SOCE is due in part to a failure to recognize and consider such advantages, as well as any disadvantages. Even if one believes that SSA is genetic or inborn, it might still be possible that the many advantages of being lesbian, gay, or bisexual work to reinforce those identities and behaviors over time. The advantages might lead a person to become LGB even without SSA as was found for some of the children of lesbian mothers, as detailed elsewhere (Schumm, 2004). In other words, beneficial experiences may reinforce attraction, behavior, and identity. At the same time, there may be some persons who feel that some negatives, perhaps especially long-term risks, of a gay, lesbian, or bisexual lifestyle might outweigh the positives. In general, my view is that when a person weighs short-term versus long-term advantages and disadvantages of a decision, change will be more difficult when the former have many positives and the latter (whether advantages or disadvantages) may not accrue for years, even decades. Even if change of sexual orientation were not possible, such subjects may find professional assistance helpful in evaluating their situation and understanding their own (accurate or inaccurate) perceptions of advantages and disadvantages, from both a short-term and a long-term perspective.

Asymmetric standards

One possible interpretation of this research is that because SOCE did not lead to all participants attaining a completely heterosexual outcome in terms of attraction, behavior, or identity it must have failed. Such a conclusion may reveal implicit bias in terms of standards used for evaluating “success” in therapy. Using some raw data drawn up in only a few minutes, let’s suppose that a therapist was running a marriage therapy program, using the scores from one spouse to assess change. The data used here used five 1’s, four 2’s, four 3’s, four 4’s, three 5’s, and four 6’s at pre-test, with the following sets of scores for pretest 1’s (7, 2, 3, 4, 5), 2’s (1, 1, 5, 5), 3’s (6, 6, 5, 4), 4’s (5, 3, 5, 6), 5’s (6, 5, 6), and 6’s (7, 6, 6, 7), so anyone is welcome to replicate the analyses. There were 24 spouses assessed at pre-test and post-test on an item whose score ranged between 1 and 7, with higher scores indicating greater marital satisfaction. With the data used, the pre-test mean was 3.33 (SD = 1.79) and the post-test mean was 4.83 (SD = 1.74). The mean difference was 1.50 and the standard deviation of the difference was 1.69. Depending on which website calculator is used, Cohen’s d was between 0.84 and 0.89, a large effect size. The results were very significant, t(23) = 4.23 (p < .001). Using a Wilcoxin signed-ranks test, z = 3.45, p = .001, so the results would be similar using either parametric or nonparametric statistics. It is likely that most therapists would consider the results impressive, both substantial in effect size and very significant statistically.

However, SOCE critics could argue that in the raw data only one spouse changed from a 1 to a 7 and only three ended up at a 7 while three scored lower at post-test and three more were unchanged at post-test with many (n = 11) changing by only one or two points in a positive direction (so that the majority of the clients (n = 14) either did not change at all or only changed a “little”. One might claim that of the 24 clients, seven were divorced during or after the program, which might be taken as failure, harm, or success (Moxley et al., 1987). Both explanations of the results are technically correct.

While most scientists would present the first set of results and claim “success”, SOCE critics are more likely to take issue with the results by focusing on the second set of results, assuming the outcome measured was sexual orientation (i.e., few (n = 3) clients became completely heterosexual (and most - two of three - of them started as “mostly” heterosexual so their change was small), only one changed from completely gay to completely heterosexual, and most remained more or less bisexual (i.e., started out as bisexual and ended up as bisexual), while three became “more” gay. SOCE critics would probably conclude from the data that SOCE was not effective, despite the “impressive” first set of results. Furthermore, SOCE critics could argue that the program was “harmful” because some clients got “worse”, some did not change at all, and for all the time and expense lost to the participants, a majority got worse or got little benefit from the program. Perhaps marital therapy should be banned, given such poor results! It is also interesting that recent research has found a number of interventions (other than SOCE) to be ineffective (Williams et al., 2020), even more harmful than effective, and yet we are unaware of calls for their termination by major professional organizations, or at least not with the same fervor as for SOCE.

We believe that the same standards should be used for SOCE as for other types of therapeutic interventions rather than carving out a special set of standards for SOCE not used elsewhere for evaluating therapeutic interventions. In other words, we do not think it’s logically coherent to apply different standards statistically just because the outcome measure is different. In other words, SOCE critics are apt to use a double standard or special pleading when evaluating SOCE results.