Deconstructing “Sexual Deviance”: Identifying and Empirically Examining Assumptions about “Deviant” Sexual Fantasy in the DSM

ABSTRACT

We identify and examine three assumptions underpinning “sexual deviance” in the Diagnostic and Statistical Manual (DSM) of Mental Disorders: (1) the “sexual deviant” – often, “the male sex offender” – prefers “deviant,” and has limited (if any) “non-deviant,” sexual fantasies; (2) this differentiates them from the non-sexual-/non-offending “norm”; (3) preferred fantasies are “deviant” or “non-deviant.” Adult volunteers (N = 279; equal numbers of sexual offending [SO], non-sexual offending [NSO] and non-offending [NO] men) provided anonymous descriptions of their favorite sexual thought and responses to a revised Wilson Sex Fantasy Questionnaire during a wider computerized survey of 6,289 men from prison and the community. Latent class analysis identified five types of favorite sexual thought; vaginal/oral sex with 1+ woman was commonest for SO men and the WSFQ findings supported this – challenging the first assumption. Both SO and NO men were over-represented for thought types considered “deviant” by the DSM – tempering the second assumption – although SO men were over-represented for thoughts involving children specifically. All thought types were multidimensional; none included solely elements considered “deviant” by the DSM – contesting the third assumption. Notions of the “sexual deviant” as “different”/“other” may underpin these assumptions, potentially negatively impacting research, therapy and understanding sexual crime.

Introduction

Socially disapproved sexual desires, fantasies and practices were initially known as “perversions,” and more recently, as “sexual deviance” or “paraphilias.” Initially understood in moral, legal and theological terms, the rising popularity and authority of psychiatry from 1850 onwards led to “perversions” being constructed as a medical or psychological problem (De Block & Adriaens, 2013). In 1886, Krafft-Ebing (1965), a German-Austrian psychiatrist, produced the first detailed biomedical account of “perversions” in the book, Psychopathia Sexualis. He argued that “perversions” were diseases of the “sexual instinct,” involving deriving sexual pleasure from imagining and fantasizing about non-reproductive acts (De Block & Adriaens, 2013). The later Diagnostic and Statistical Manual (DSM) of Mental Disorders reflected similar ideas. “Sexual deviations” in DSM-II were acts outside of “coitus” with the “opposite sex,” or “coitus … under bizarre circumstances” (American Psychiatric Association [APA], 1968, p. 44), and the introduction of diagnostic criteria in DSM-III indicated this could include fantasies or acts (American Psychiatric Association [APA], 1980).

Numerous papers have offered general conceptual critiques of the DSM’s diagnoses of “sexual deviance” or “paraphilia,” often using prior research to support their critique (e.g., Joyal, 2018, 2021; Moser, 2016; Moser & Kleinplatz, 2006). What we offer here is different in two regards. Firstly, we interrogate, in particular, how “deviant” sexual fantasy, and the people who might experience these, are represented within “sexual deviance” in the DSM. Further, we undertake this interrogation through a process of “deconstruction.” Developed by the French philosopher, Jacques Derrida (Derrida, 1978, 1981), “deconstruction” involves analyzing the language and constructs in a text to identify the underlying assumptions within it, with particular attention given to the dualistic structure of the language, that is, the binary oppositions within the text. Deconstruction enables a questioning of the elements that might affect the text’s meaning but that might not be immediately visible within it, illuminating ambiguities and suppressed meanings. Secondly, we report research that enables us to empirically evaluate the assumptions we identify.

Deconstructing “Sexual Deviance” in the DSM

We undertook close and repeated readings of the past and current editions of the DSM, focusing on the sections on “sexual deviance” or “paraphilias,” to identify presuppositions and dualisms about “deviant” sexual fantasies and the people who might experience them. Here, we present the three key assumptions about “deviant” sexual fantasies that we identified during this analysis and a consideration of who is cast in the role of “the sexual deviant.” Recognizing that these assumptions will have been shaped over time by historical, clinical, cultural, political and other factors, we also briefly highlight highly influential works and ideas that are likely to have contributed to the representation of these assumptions within the text.

Firstly, an enduring idea in the DSM is that “sexual deviance” is a defining aspect of the individual’s sexual life. In particular, the “deviant” person is constructed as either exclusively engaging in “deviant” sexual fantasies or having a distinct preference for them. DSM-III to DSM-IV-TR indicated “deviant” sexual fantasies could be either exclusive or preferred (APA, 1980, 1987, 1994, 2000). For example, the DSM-III diagnostic criteria frequently described a given “paraphilia” as, “a repeatedly preferred or exclusive method of achieving sexual excitement” (APA, 1980, pp. 268–275). However, in DSM-5, the latest version of the DSM, there was a move away from “deviant” sexual fantasies being exclusive or preferred, to a focus on the latter, with “paraphilic disorders” being separated into “anomalous activity preferences” or “anomalous target preferences” (American Psychiatric Association [APA], 2013, p. 685). When “deviant” sexual fantasies are the exclusive means of achieving sexual arousal, this suggests an absence of “non-deviant” sexual fantasies, whereas when “deviant” sexual fantasies are preferred, this suggests the presence of “non-deviant” sexual fantasies, albeit more limited. Thus, in the DSM, the “deviant” person prefers “deviant” sexual fantasies and has limited (if any) “non-deviant” fantasies.

This assumption echoes earlier, highly influential ideas. Freud’s early work suggested that exclusively experiencing particular sexual fantasies was pathological (De Block & Adriaens, 2013). And later, McGuire et al.’s (1965) paper on “sexual deviations,” building on earlier ideas by Binet (1888) and Norman (1892), argued that repeated fantasy-masturbation pairings led to a sexual preference for “deviant” fantasies and limited or no other fantasies.

Norms are another deeply entrenched notion within the DSM. “Sexual deviance” is constructed as “abnormal”; that is, a deviation from socially accepted (DSM-I; American Psychiatric Association [APA], 1952), “natural”/reproductive (DSM-II; APA, 1968) or statistical (DSM-III; APA, 1980) norms. In DSM-I, for example, the category of “sexual deviation” was placed under the broader nosology of “sociopathic personality disturbance,” with the individual being, “ill primarily in terms of society and of conformity with the prevailing cultural milieu” (APA, 1952, p. 38) – emphasizing the deviation from socially accepted norms. However, from DSM-III onwards there was a greater focus on statistical norms, and in particular, “paraphilic” sexual fantasies were described as “unusual” or “bizarre” from DSM-III to DSM-IV-TR (APA, 1980, 1987, 1994, 2000) and “anomalous” in DSM-5 (APA, 2013). Thus, a second assumption in the DSM is that the “deviant” person’s preferred sexual fantasies differentiate the “deviant” person from the “norm.”

This assumption reflects a long, wider tradition. Whilst Freud, and to some extent Kinsey, argued that the statistical norm was a perverse norm, many others have constructed “sexual deviance” as unusual, and deviating from divine, “natural,” socially accepted or statistical norms (De Block & Adriaens, 2013). For example, Johann Heinroth’s leading handbook on “mental disorders” in 1818 treated such disorders as sins (deviating from divine norms), while Krafft-Ebing and Kraepelin argued that the different forms of “sexual deviance” were biologically abnormal (deviating from “natural” norms; De Block & Adriaens, 2013).

A related set of dualistic opposites in the DSM concerns the notion of “sexual deviance” itself. What constitutes “sexual deviance” is juxtaposed with “non-deviance.” For example, in DSM-III, “The Paraphilias are characterized by arousal in response to sexual objects or situations that are not part of normative arousal-activity patterns” (APA, 1980, p. 279, emphases added). Thus, “paraphilia” is defined and has meaning only in relation to its opposite, and in so doing, the DSM highlights the greater value given to “non-deviance.” Similarly, in DSM-5, “paraphilia” is defined as, “any intense and persistent sexual interest other than sexual interest in genital stimulation or preparatory fondling with phenotypically normal, physically mature, consenting human partners” or “any sexual interest greater than or equal to normophilic sexual interests” (APA, 2013, p. 685, emphases added). DSM-5 continues by explaining that these interests can be determined by a person’s sexual fantasies, urges or behaviors (p. 686). This indicates a third assumption in the DSM: that preferred sexual fantasies can or should be separated into “deviant” and “non-deviant.”

This assumption is not unique to psychiatry; it has long been assumed that “deviance” can or should be separated from “non-deviance” and demarcating these boundaries has been of enduring interest to theologians, philosophers, and physicians, too (De Block & Adriaens, 2013). Where these boundaries have been drawn has varied across time and place, in line with the values and norms of a given society at the time (Bhugra et al., 2010; De Block & Adriaens, 2013).

In terms of who is cast in the role of “the sexual deviant,” this is almost always a male (APA, 1980, 1987, 1994, 2000, 2013), and often, someone who is likely to, or has, committed a sexual crime. Firstly, forms of “sexual deviance” that are sexual crimes are foregrounded in the DSM. For example, DSM-5 states,

These disorders have traditionally been selected for specific listing and assignment of explicit diagnostic criteria in DSM for two main reasons: […] some of them entail actions for their satisfaction that, because of their noxiousness or potential harm to others, are classed as criminal offenses. (APA, 2013, p. 685)

Second, sexual crime is framed as a common outcome of “sexual deviance.” For example, DSM-III-R states that, “Paraphilias involving another person, particularly Voyeurism, Exhibitionism, Frotteurism, Pedophilia, and Sexual Sadism, often lead to arrest and incarceration.” (APA, 1987, p. 281, emphases added). Further, in DSM-IV-TR, “sexual deviance” is constructed as pathological only when it leads to sexual crime or other consequences: “Fantasies, behaviors, or objects are paraphilic only when they lead to clinically significant distress or impairment (e.g., […] lead to legal complications […])” (APA, 2000, p. 568, emphases added). Indeed, the DSM often frames fantasies or urges as driving behavior – they are described as being “acted out” or “acted on” (APA, 2000, 2013; De Block & Adriaens, 2013). Third, “the sexual deviant” is constructed as someone who has already committed a sexual crime. For example, DSM-III-R to DSM-IV-TR state that individuals “with Exhibitionism, Pedophilia, and Voyeurism make up the majority of apprehended sex offenders” (APA, 1987, p. 281, 1994, p. 523, 2000, p. 566). In DSM-IV-TR and DSM-5 it is also suggested that committing recurrent sexual crimes is alone sufficient to determine a “paraphilia” (APA, 2013; De Block & Adriaens, 2013). For example, Criterion A of the DSM-5 diagnostic criteria states, “Over a period of at least 6 months, recurrent and intense sexual arousal from […], as manifested by fantasies, urges, or behaviors” (APA, 2013, pp. 686–702, emphases added). Thus, behavior alone – such as sexual crime – can be used to determine the presence of sexual arousal. This is reiterated more plainly elsewhere in DSM-5; for example, “Recurrent voyeuristic behavior constitutes sufficient support for voyeurism (by fulfilling Criterion A) and simultaneously demonstrates that this paraphilically motivated behavior is causing harm to others (by fulfilling Criterion B).” (APA, 2013, p. 687). As such, this creates a tautology, where recurrent sexual crimes are constructed as evidence of sexual arousal to “deviant” themes simply because the individual has committed recurrent sexual crimes. Overall, then, an underlying assumption within the DSM appears to be that it is often “the male sex offender” who is being invoked as “the sexual deviant.”

The idea of fantasies driving behavior echoes MacCulloch et al.’s (1983) earlier highly influential proposition that 13 men in a psychiatric hospital had experienced repetitive sadistic masturbatory fantasies that led to “increasingly dangerous in vivo ‘try-outs’ of their fantasies” (p. 20), including sadistic sexual assaults and rape. Moreover, the increasing links to sexual crime in the DSM’s representation of the “sexual deviant” may also be partly a product of the composition of the DSM-5 Paraphilias Subworkgroup Panel, which typically comprised people specializing in the assessment, management and treatment of people who commit sexual crimes, rather than sexologists (Beech et al., 2016).

In sum, the analysis of the DSM presented in this section shows that “the sexual deviant” – often constructed as “the male sex offender” – is assumed to prefer “deviant” sexual fantasies and have limited (if any) “non-deviant” fantasies, and that these preferred sexual fantasies differentiate the “male sex offender” from the “norm.” Here, we might infer that the “norm” would be people, typically men, who do not commit sexual crimes (and who are simultaneously implied as not having “deviant” sexual fantasies). Further, the DSM assumes that these preferred sexual fantasies can or should be separated into “deviant” or “non-deviant.”

Research on the Preferred Sexual Fantasies of Sexual Offending Men

It is important to empirically interrogate these identified assumptions about “sexual deviance.” A DSM diagnosis of a “mental illness” is likely to affect how individuals see themselves and others treat them. Further, more broadly, by virtue of the DSM being positioned as an authoritative text on “mental illness,” professionals and the public alike may look to the DSM for guidance on what constitutes “sexual deviance” – but without questioning the assumptions that underlie the text. The DSM also influences legal decisions about civil (involuntary) commitment for “sexually violent predators” (First & Halon, 2008) and how research is conducted and interpreted.

Whilst the assumptions about “sexual deviance” that we have identified in the DSM are also reflected in psychological and psychiatric theories of sexual crime (Lussier & Cale, 2016; Ward et al., 2006), there is very limited empirical evaluation of these claims. Many studies have examined the sexual fantasies of sexual offending men (e.g., Allan et al., 2007; Baić et al., 2019; Gee et al., 2004; Hudson et al., 2002; Marshall et al., 1991; Sheldon & Howitt, 2008; Stinson & Becker, 2008; Swaffer et al., 2000; Woodworth et al., 2013), sometimes with a comparison sample of non-sexual offending and/or community (although, not necessarily, without a history of sexual crimes) men (Bartels et al., 2019; Baumgartner et al., 2002; Cortoni & Marshall, 2001; Curnoe & Langevin, 2002; Daleiden et al., 1998; Gallant & Wormith, 1986; Langevin et al., 1998; Looman, 1995; O’Donohue et al., 1997; Ronis et al., 2022). However, research on the preferred sexual fantasies of sexual offending men, and research that directly compares these preferred sexual fantasies to those in men who do not commit sexual crimes, is lacking. This is important given the DSM’s emphasis on a preference for, rather than simply the presence of, “sexual deviance.”

To our knowledge, there has been only one study of the preferred sexual fantasies of sexual offending men to date. Rokach (1990) and her colleagues (Rokach et al., 1988) asked Canadian sexual offending men, non-sexual offending men and women, and student men and women to describe their favorite sexual fantasies and then analyzed these descriptions using content analysis. They found that sexual offending and student men’s most preferred sexual fantasies were “conventional heterosexual” (vaginal penetration, oral sex, sexual touching, “courtship rituals”), while non-sexual offending men preferred both “conventional heterosexual” and “exploratory” (sex with more than one partner, outside the home, anal penetration) themes. This suggests that the sexual offending men did not prefer “deviant” sexual fantasies and that their preferred fantasies were similar to men who had not committed sexual crimes. However, the comparison groups were not screened for sexual crimes (a common issue in sexual offending research; Turner-Moore & Waterman, 2017) and inferential testing was limited. In addition, the content analysis of the preferred sexual fantasies focused on pre-selected elements (partners, acts, objects, locale) that were grouped conceptually by the researchers, rather than comprehensive inductive coding and multivariate modeling of the fantasies. This not only impedes the conclusions that we can draw about the content of sexual offending men’s preferred sexual fantasies but also whether these preferred fantasies can be neatly separated into “deviant” and “non-deviant.”

The Sexual Thoughts Project (STP)

The STP is a large, anonymous, cross-sectional survey of men’s sexual thoughts and experiences. In the STP, we conceptualized a sexual thought as a thought with sexual content (i.e., content typically seen as sexual within a culture) and/or that the individual finds sexually arousing. We conceptualized a sexual fantasy as a sexual thought with particular experiential and functional features – a pleasant, engaging, elaborated mental picture that creates, maintains or intensifies sexual arousal; only some sexual thoughts possess these features, so we explored sexual thoughts more broadly. In our earlier paper, we examined the relationship between having sexual thoughts involving children or coercing others and engaging in sex offending behaviors (Turner-Moore & Waterman, 2017). In this paper, we focus on examining participants’ favorite sexual thoughts to empirically evaluate the identified assumptions about “sexual deviance” in the DSM.

As part of the STP, we collected participants’ written accounts of their favorite sexual thought. We systematically coded the favorite sexual thoughts of men convicted for sexual offenses and comparison samples of men convicted for non-sexual offenses or who had no convictions (first screening these comparison groups for sexual crimes). Then, for the first time, we statistically modeled the content of these preferred sexual thoughts and examined the associations between thought type and group. We report these analyses with the aim of evaluating whether: (1) sexual offending men prefer “deviant” sexual thoughts and experience limited (if any) “non-deviant” sexual thoughts; (2) these preferred “deviant” sexual thoughts differentiate them from men who do not commit sexual crimes; and (3) these preferred sexual thoughts can be clearly categorized as “deviant” or “non-deviant.” Additionally, we present relevant findings from a revised version of the Wilson Sex Fantasy Questionnaire (Wilson, 1978) to contextualize the findings for favorite sexual thoughts.

Method

Participants

Participants in the STP were 6289 men aged 18–90, comprising 208 inmates in an English medium-security prison and 6081 community men. We focus on a subsample of 279 men, comprising the 93 datasets for men imprisoned for 1+ sexual offenses (sexual offending group/“SO group”) and comparison samples of equal size to the SO group, including 93 men imprisoned for 1+ non-sexual offenses and screened for sexual offenses (non-sexual offending group/“NSO group”) and 93 UK community men screened for offenses (except minor non-sexual summary offenses, e.g., speed limit offenses; non-offending group/“NO group”).

The SO group was recruited by distributing information leaflets to inmates, holding a research information session (28/31 attendees participated) and approaching men on the “Vulnerable Prisoner” wings (72/86 participated; 11 refused, 3 excluded due to learning disabilities or difficulty communicating in English). Of the 100 datasets from the SO group, 7 datasets were largely incomplete and excluded, producing a final sample of 93 SO men. The NSO group was recruited by distributing information leaflets and approaching men on the main prison wings (108/114 participated; 4 refused, 2 excluded due to poor reading ability). Of these 108 datasets, 1 dataset was excluded after screening for self-reported sexual convictions or experiences of sexual offending. For the present study, 93 NSO men were selected from the 107 based on most complete datasets. Participation for both the SO and NSO groups attracted the same pay as prison work or education sessions. Community men were recruited via posters/advertisements in sports centers, businesses and libraries, articles in the British press, and registration of the study website with major search engines and a directory of World Wide Web links. Initial news articles led to international press coverage, internet articles and posts on other websites and forums. There were 13,720 first-time “hits” to the website; 6005 men participated (predominately from Britain, the Republic of Ireland, Italy, and the USA). Additionally, 130 postal alternatives were requested; 76 (58%) were returned. Participants were not paid. For the present study, 93 NO men from the UK were selected from a pool of 305 UK NO men (out of 1819 total community datasets available at the point of selection) for comparison with the 93 SO men. Each NO and SO participant was individually matched on key demographic variables: 73 participants on age ± 5 years, education ± 2 years and sex of adult sexual partners (female/male/both), and where multiple matches were available, relationship status, ethnicity and religiosity; broader education and age criteria were required for the remaining 20 participants.

In terms of the demographics of the three groups, the NSO group (M_age = 27.81 ± 6.73) was younger than the SO (M_age = 42.80 ± 14.12) and NO (M_age = 42.41 ± 13.70) groups. All groups differed on years of education, notably so for the NSO group (SO M_education = 12.51 ± 3.49; NSO M_education = 10.58 ± 3.33; NO M_education = 13.80 ± 2.84). All groups commonly had female sexual partners (SO 89.0%; NSO 96.5%; NO 87.6%; no sexual intercourse, SO = 4, NO = 2). Roughly equal proportions of SO men had completed (34.4%), were currently (31.1%) or had never participated in (34.4%) an intervention for sexual offending. Table 1 in Turner-Moore and Waterman’s (2017) article provides further information on the characteristics of these groups.

Measures

Computerized Interview for Sexual Thoughts (CIST)

The CIST, designed by the authors, comprised between 48 and 247 open- and closed-ended questions, depending on participants’ responses and the corresponding question branching. The questions covered: less sensitive socio-demographic questions, favorite sexual thought, sexual thoughts involving children, coercive sexual thoughts, and sensitive socio-demographic questions (e.g., current and prior convictions). Participants were asked to describe their favorite sexual thought in their own words and in detail, with a prompt for further information if they provided <20 words. Next, participants were asked if there were other people in the thought, and if present, how many, and details for up to four of them (sex, age, relationship to participant, description of the “type of person” that they are in the thought); participants without people in the thought were asked to describe the objects/things instead.

Wilson Sex Fantasy Questionnaire (WSFQ)

The WSFQ (Wilson, 2010; Wilson, 1978) comprises 40 sexual fantasies, with four factors of 10 items: Intimate (e.g., “Kissing passionately”), Exploratory (e.g., “Participating in an orgy”), Impersonal (e.g., “Intercourse with an anonymous stranger”) and Sadomasochistic (e.g., “Whipping or spanking someone”). The reliability and validity of the WSFQ have been demonstrated previously (e.g., Bartels et al., 2019; Baumgartner et al., 2002). Minor revisions were made to 21 items to: simplify; explain or update terms; or replace making love, intercourse and having sex with the standardized term sex. The number, meaning and ordering of items were retained. In the original WSFQ, participants rated each item on a 0–5 frequency scale for five contexts: daytime fantasies, fantasies during intercourse or masturbation, dreams while asleep, past behavior and desired behavior; however, the first three contexts are often highly intercorrelated so only the daytime fantasy ratings are used when scoring (Gallant & Wormith, 1986; Wilson, 1978; Wilson & Lang, 1981). In the STP, these first three contexts were collapsed into a single thought context (“Think about it”) – consistent with our focus on sexual thoughts, and the sexual behavior (“Have done in reality”) and sexual desire (“Would like to do in reality”) contexts were retained. The revised WSFQ is available as Online Supplemental Material (OSM). The validity of the revised WSFQ was not tested in the present study; Cronbach’s alphas are reported in Table 1.

Table 1. Means (standard deviations), ANOVAs and internal reliability for the WSFQ thought context by group.

	Group
	SO (n= 59)			NSO (n = 74)			NO (n = 71)
Subscale/total	M	(SD)	α	M	(SD)	α	M	(SD)	α	F	df	ω^2
Intimate	26.49a	(13.22)	.91	32.19b	(12.95)	.90	35.94b	(9.82)	.84	9.99*	(2, 201)	.08
Exploratory	9.98a	(8.61)	.79	14.27b	(7.90)	.72	16.20b	(8.44)	.69	9.34*	(2, 201)	.08
Impersonal	11.41a	(8.38)	.73	16.38b	(8.88)	.75	20.21c	(8.52)	.68	16.84*	(2, 201)	.13
Sadomasochistic	3.80a	(5.77)	.82	8.65b	(9.83)	.86	11.01b	(8.50)	.76	17.92*	(2, 133)	.10
Scale total	51.67a	(31.48)	.94	71.54b	(34.22)	.93	83.37c	(26.25)	.87	17.18*	(2, 201)	.14

Higher means indicate greater thought frequency. Means in the same row that do not share subscripts differed at p < .05 at post-hoc testing. See Text S4 in the OSM for how the tests of difference were conducted. Cronbach’s alpha was calculated before imputing missing values. SO = Sexual offending; NSO = Non-sexual offending; NO = Non-offending. * p < .05 with a multistage Bonferroni correction

Computerized Inventory of Sexual Experiences (CISE)

The CISE, designed by the authors, comprised between 22 and 410 primarily closed-ended questions, depending on participants’ responses and the corresponding question branching. The questions covered: consensual sexual activity with adults, non-consensual sexual activity with adults, sexual activity with children, masturbation, viewing sexually explicit or abusive media, and childhood victimization experiences. These questions included participants’ sexual offending experiences (including unconvicted offenses).

Balanced Inventory of Desirable Responding (BIDR-6)

The BIDR-6 (Paulhus, 1998) comprised 40 statements on everyday events; agreement to these is indicated on a 7-point Likert scale. Relevant here is the Impression Management subscale (IM; self-presentation tailored to an audience), comprising 10 true-keyed and 10 false-keyed items. Item 34 (“I never read sexy books or magazines”) was omitted from the scoring to avoid conflation with sexual variables (Meston et al., 1998). Following Stöber et al. (2002), all scale responses were counted. Adjusted scores were computed for ≤5/40 missing values; data with >5 missing values were not analyzed (Paulhus, 1998). The reliability and validity of the BIDR have been demonstrated previously (Paulhus, 1984, 1991, 1998). In the present study, Cronbach’s alphas for IM were high (Turner-Moore & Waterman, 2017).

Procedure

The University of Leeds and Her Majesty’s Prison Service provided ethical approval. Participants completed the CIST, a two-minute picture-matching distracter task and three counterbalanced measures: the WSFQ, CISE and BIDR. In prison, one or two men participated per session, positioned at either end of a private room. A standardized written and verbal briefing provided assurances that their decision to participate or not, to respond to a question or not, or what response was given, would not affect their prison sentence, treatment or parole. Most men participated via a secure laptop mounted with a custom computer program (two chose paper-based alternatives); participation signified consent. The researcher sat outside the room to provide privacy or assistance as necessary. Most community men participated using a custom application on the University of Leeds website (postal alternatives were available). Following a standardized written briefing, participants were required to confirm they were a man, 18+, and able to complete the study in private, before typing a generic password to indicate consent. All participants were assured of anonymity and confidentiality, and that sexual thoughts are a common activity that can be varied and diverse.

Analyses

Type of Sexual or Non-sexual Offense

To provide more specific follow-up analyses, recognizing that people who commit crimes are not homogenous, we classified the SO and NSO men by type of offense. This was based on their current and prior convictions (as reported in the CIST), and for the SO men, also their sexual experiences (as reported in the CISE). For SO men: sexual offenses against children (SOC; n = 46) included 1+ sexual offenses against a person <16 years and no sexual offenses against adults; sexual offenses against adults (SOA, n = 26) included 1+ sexual offenses against a person ≥16 years and no sexual offenses against children; and SOC+SOA (n = 12) included both types of sexual offenses. For NSO men: violent (vNSO; n = 51) included 1+ offense involving actual/attempted/threatened physical injury against people (fatal/non-fatal), violent public disorder offenses, robbery, possession of firearms with intent; and nonviolent (nvNSO; n= 20) included 1+ offense involving burglary, theft, handling stolen goods, dishonesty/deception offenses, drug offenses, nonviolent offenses against justice, nonviolent driving and motoring offenses, other summary offenses, and no convictions for a violent offense. Some men were unclassified: their offenses were ambiguous (could be violent or nonviolent; n = 20) or data were incomplete (n = 9 SO; n = 2 NSO).

Favorite Sexual Thoughts

A predominantly inductive content analysis (Turner-Moore & Waterman, 2017), formalized into a Coding Manual, was applied to the favorite sexual thought descriptions. This comprised 20 coding variables (e.g., type of sexual acts, respondent’s emotional states) that were single- or multiple-selection (i.e., the coder selected only one code for the variable or marked each code within the variable as Present/Absent; see Table S1 in the OSM); consequently, 112 sub-variables were coded per account. A stratified random sample of 15% (N = 42) of descriptions were independently second coded; 62% of the 112 sub-variables showed excellent or good kappa values (Turner-Moore & Waterman, 2017). Most disagreements reflected minor coding violations by the second coder, so the first coder’s codes were retained. The coding framework’s development, and the coding, were undertaken blind to group membership.

Missing data were minimized by substituting Absent or Unable to determine codes with information from the quantitative data, including the number, sex, age and relationship of the other people in the thought. Using the participant’s age, we re-coded the age of the adult actors as younger, the same age or older than the participant. Composite variables were computed, including count variables (e.g., the number of different sexual acts coded Present) and higher-order variables (e.g., combining codes from Respondent’s states and Other actors’ states to produce the variable Interpersonal context).

Latent class analysis in LatentGOLD 4.0 identified types of favorite sexual thought. To reduce the number of variables for analysis, those least likely to discriminate between latent classes (i.e., where ≥80% of descriptions were assigned the same code, or were divided over one valid code and an Unable to determine code, for a variable) were excluded. Low-frequency codes were combined into higher-order codes and missing value codes (e.g., Not applicable/Unable to determine) were merged. We requested models for one to six latent classes; the Bayesian Information Criterion (BIC), a likelihood-based measure of model fit recommended for non-nested models (Kass & Raftery, 1995), determined the optimum number of latent classes (i.e., types of favorite sexual thought). To produce the simplest and most discriminatory model, we iteratively excluded the variable with the least effect on each latent class (as indicated by R² for each variable within the lowest-BIC model) and then re-computed and assessed the models. If there were high bivariate residuals (substantially >1), one of the variables in the pair was deleted or another latent class/es was added (Magidson & Vermunt, 2003); deletion was chosen if the variable was redundant (e.g., for the variables, Other actors’ desire and enjoyment and Number of different states for other actors, the latter can be partly inferred from the former).

Using the latent class probabilities, each participant was assigned to the latent class (i.e., type of sexual thought) he was most likely to have (Magidson & Vermunt, 2003). A fine-grained analysis of each thought type followed in SPSS v.15/20; we disaggregated all variables from the final model and conducted exact chi-squares (Everitt, 1992) to examine the disaggregated, and previously excluded, variables by thought type. For multiple-selection variables, an exact chi-square was calculated for each code and a Bonferroni correction applied (Agresti & Liu, 1999). The adjusted standardized residuals indicated which cells were responsible for a significant chi-square value (Everitt, 1992). Lastly, we reviewed the participant accounts assigned to each latent class to identify illustrative extracts (participant extracts are provided verbatim, preserving any grammatic or spelling errors).

WSFQ

Ninety-one participants had 1+ missing values; 74 did not answer all Thought context items (an apparent misunderstanding of the task instructions; participants often provided a response for only one context – Thought, Behavior or Desire – per item); 13 omitted the task or exited the study; 4 datasets were lost to technical difficulties. There was no significant association between group and viewing but not answering 1+ items for the Thought context (%_SO = 34.4, %_NSO = 25.8, %_NO = 19.4), χ²(2, 279) = 5.44, p = .071, V = .14. Missing values were imputed for ≤3/40 missing values for the Thought context (n = 16) using the participant’s corresponding subscale mean (Wilson, personal communication, October 16, 2008); participants with >3 missing values were removed. Outliers (cases values with a z-score greater than 3.29, p < .001) were reduced to 3.29 standard deviations from the mean (Field, 2013).

Results

Favorite Sexual Thoughts

The latent class analysis was undertaken with 16 variables (see Table S2 in the OSM), resulting in a five-class model (see Text and Table S3 in the OSM) of five variables (see Table 2). Follow-up analyses examining disaggregated and excluded variables by thought type, and the qualitative accounts, elaborated the nature of each type.

Table 2. Conditional and latent class probabilities for favorite sexual thought.

	Class
	1	2	3	4	5
“Quasi-paraphilic” behaviors^a
Any	0.034	0.994	0.001	0.138	0.096
None	0.953	0.005	0.034	0.863	0.898
Missing	0.013	0.001	0.965	0.001	0.006
“Paraphilic” behaviors^b
Any	0.001	0.336	0.032	0.734	0.579
None	0.998	0.663	0.007	0.265	0.418
Missing	0.001	0.001	0.961	0.001	0.003
Means of compliance
Any	0.000	0.846	0.063	0.128	0.002
None	1.000	0.024	0.502	0.871	0.744
Missing	0.000	0.130	0.435	0.001	0.254
Number of different sexual acts^c
1	0.101	0.199	0.077	0.254	0.069
2	0.158	0.314	0.121	0.400	0.109
3+	0.070	0.139	0.054	0.177	0.048
Missing	0.671	0.348	0.748	0.169	0.774
Interpersonal context^d
No states	0.258	0.210	0.188	0.002	0.003
Self-oriented	0.223	0.268	0.127	0.132	0.644
Other-oriented	0.105	0.105	0.247	0.001	0.128
Mutual	0.410	0.365	0.221	0.844	0.019
Missing	0.004	0.052	0.217	0.021	0.206
Prevalence of each class	0.540	0.145	0.121	0.104	0.090

N = 264.

^{a “}Quasi-paraphilic” behaviors includes 16–17 year-old actors, Social voyeurism (observation of knowledgeable others who are naked or engaged in sexual activity), Social exhibitionism (uninvolved background actors), Causing psychological suffering or physical pain, Receiving psychological suffering or physical pain.

^{b “}Paraphilic” behaviors includes 0–15 year-old actors, Voyeurism, Exhibitionism, Frotteurism, Causing psychological suffering or physical pain with sadism, Receiving psychological suffering or physical pain with masochism, Fetishism, Cross-dressing, Partialism, Other “paraphilia.”

^cCount variable for Type of sexual contact.

^dComposite variable based on whether any states were reported for the participant only (Self-oriented), the other actor/s only (Other-oriented), or both (Mutual).

The commonest type (54.0% of the total sample) described “having sex,” “making love,” oral sex and/or penile-vaginal penetration with a woman they usually knew (i.e., their partner or another known younger woman). Sometimes, the men described attraction toward the woman and mutual desire and enjoyment. Occasionally, they described mutual feelings of love, as this extract illustrates: “[…] to be alone on a paradise island with my ex wife.falling in love all over again.making passionate love on the beach.” (SOC, 54 years old). Although the sexual acts were usually culturally conventional, they often included more than one woman (labeled Vaginal/oral sex with 1+ woman). For example, “performing oral sex on my partner, an extremely vivid picture of her pussy in my mind […] Sometimes another woman will enter the room and start playing with both of us. […]” (nvNSO, 33).

The second commonest thought type (14.5%) depicted almost universally consensual scenes of dominance and submission (labeled Dominance and submission), sometimes with mild bondage (e.g., handcuffs), discipline (e.g., caning) and fetishistic elements (e.g., uniforms). Sadomasochism was not a strong feature; that is, pain or humiliation was used in gaining compliance within the scene (instrumental), rather than for its own sake (expressive). Typically, the men used restraints and mild physical force with an individual woman, or less frequently, a woman during group sex with women and/or men, and engaged in varied sexual acts, particularly non-genital contact (e.g., breasts, buttocks), oral sex and vaginal penetration. Some men described mutual desire and enjoyment:

I own a sex shop […] She tells me of her fantasies of mild bondage that she has had all her life but never fulfilled with her husband. […] I […] tie her, spread eagled, to the bed with silk scarves. I then blindfold her and start using one of the vibrators she bought all over her body, finally thrusting it into her and making her climax. […] We then pleasure each other orally to a climax […] (SOC, 42)

Others, however, omitted their and/or their sexual partner’s emotional states. Although most men described being dominant, a substantial minority described being submissive, and a few switched between these:

[…] it is about being dominated by them with cuffs and whip, […] short dress or mini skirt the top being just a bra or a low cut item that shows plenty of cleavage, also knee high boots come into it aswell as holdups (stockings). […] [or] where i dominate and she dresses up in maid, nurse ect and i act out as the superior and have sex in many different positions at my say […] (NO, 27)

The third commonest type (12.1%) incorporated short and/or vague accounts that had to be coded Unable to determine for most variables (labeled Ambiguous). The descriptions suggested underspecified versions of the other thought types. For example, “i think about past occurances. past sexual experiences, with past lovers, times we had sex or sex related activities” (NO, 18). The most frequent detail was sex with two or more younger, unknown, sexually aroused women, such as, “threesomes and dirty sex [Following question for the other actors’ temperaments:] horny gentle” (vNSO, 23).

The final two types included either a fetish or sexual activity with a child; these were brought together statistically within the same thought type owing to a description of mutual participation (10.4%; Fetish or child: Mutual) or only the men’s emotions (9.0%; Fetish or child: Self). The fetishes usually entailed a woman wearing particular materials (e.g., PVC, latex, lace) or items of clothing (e.g., stockings, suspenders, high heels). The children were pre- or post-pubescent (usually girls), rather than an adult role-playing a child.

In the Mutual variation, the man sometimes described attraction to the person, and the people took turns to initiate and lead varied sexual acts (kissing/hugging/clothed contact, non-genital and genital contact, oral sex, vaginal and anal penetration) in a home setting:

[…] The first is being cross dressed in a bar and meeting a man who then seduces me and takes me home. The other is the reverse. I am in a bar and seduce and take home a beautiful transvestite. In both cases, the physical side contains kissing, oral and anal sex while the cross dressed one of the couple remains dressed in, at least, lingerie. (NO, 29)

The man often knew the person, describing mutual sexual desire, and his own enjoyment. For example, in one particularly explicit account, a 43-year-old man in the SOC group recounted a sexual thought involving clothed contact, genital contact, oral sex, and vaginal and anal penetration with two or three girls, aged 6–12, who were described as friends. The account detailed sexual turn-taking between him and the girls, and between the girls, and multiple orgasms experienced by those involved.

The Self variation described only the man’s desire and enjoyment, and focused on the fetishized material, clothing or activity:

[…] thigh high boots and pvc rubber and any thing to do with that i love water sports and going to fetish clubs asphixiation nights also i am a cross dresser and this plays a big part in my sex life and thoughts (nvNSO, 32)

or sexual acts with an unknown female child/children:

To have two 10y girls [later described as strangers] in bed doing sex acts on each other with me in the room. I like girls at the age of 10 and it would be nice to see two girls having sex at 10y old. (SO – target unclear, 27)

Cross-tabulation of thought type by group (see Table 3) and subgroup (see Table 4) showed that all groups and subgroups most favored Vaginal/oral sex with 1+ woman. All three groups reported all five thought types (Table 3), though this was not the case for all subgroups (Table 4). There were small significant associations between thought type and group, χ²(8, 264) = 23.20, p = .003, V = .21, or subgroup, χ²(20, 235) = 35.23, p = .020, V = .19. The SOC+SOA subgroup reported Fetish or child: Mutual more than expected statistically. The vNSO and SOA subgroups reported Vaginal/oral sex with 1+ woman more than expected statistically. The NO group reported Dominance and submission and Fetish or Child: Self more than expected statistically. A one-way ANOVA indicated no significant differences in IM scores by thought type, F(4, 232) = 1.41, p = .23, ω² = .007).

Table 3. Descriptive statistics and adjusted residuals for favorite sexual thought by group.

	Group
	SO			NSO			NO
Thought type or element	%	n	(AR)	%	n	(AR)	%	n	(AR)
Thought type
Vaginal/oral sex with 1+ women	61.9	52	(0.7)	68.1	62	(2.3)	46.1	41	(−3.0)
Dominance & submission	7.1	6	(−2.3)	11.0	10	(−1.1)	24.7	22	(3.4)
Ambiguous	13.1	11	(0.3)	12.1	11	(0.0)	11.2	10	(−0.3)
Fetish or child: Mutual	13.1	11	(2.3)	3.3	3	(−1.9)	6.7	6	(−0.4)
Fetish or child: Self	4.8	4	(−1.0)	5.5	5	(−0.8)	11.2	10	(1.8)
Children
Yes	10.7	9	(2.4)	2.2	2	(−1.8)	4.5	4	(−0.6)
No	82.1	69	(−3.0)	94.5	86	(1.7)	93.3	83	(1.2)
Ambiguous	7.1	6	(1.7)	3.3	3	(−0.5)	2.2	2	(−1.1)
Other person’s lack of desire
Yes	3.6	3		1.1	1		2.2	2
No	96.4	81		98.9	90		97.8	87
Other person’s lack of enjoyment
Yes	1.2	1		1.1	1		1.1	1
No	98.8	83		98.9	90		98.9	88

SO = Sexual offending; NSO = Non-sexual offending; NO = Non-offending; AR = Adjusted residual. ARs shown for significant chi-squares only.

Table 4. Descriptive statistics and adjusted residuals for favorite sexual thought by subgroup.

	Subgroup
	SOC			SOA			SOC+SOA			vNSO			nvNSO			NO
Thought type	%	n	(AR)	%	n	(AR)	%	n	(AR)	%	n	(AR)	%	n	(AR)	%	n	(AR)
Vaginal/oral sex with 1+ women	51.3	20	(−1.0)	80.8	21	(2.5)	45.5	5	(−0.9)	72.0	36	(2.2)	70.0	14	(1.1)	46.1	41	(−3.0)
Dominance & submission	12.8	5	(−0.5)	0.0	0	(−2.3)	9.1	1	(−0.6)	12.0	6	(−0.7)	10.0	2	(−0.7)	24.7	22	(3.1)
Ambiguous	12.8	5	(0.3)	11.5	3	(0.0)	18.2	2	(0.7)	8.0	4	(−0.9)	15.0	3	(0.5)	11.2	10	(−0.1)
Fetish or child: Mutual	15.4	6	(1.7)	7.7	2	(−0.2)	27.3	3	(2.3)	6.0	3	(−0.7)	0.0	0	(−1.4)	6.7	6	(−0.8)
Fetish or child: Self	7.7	3	(0.4)	0.0	0	(−1.4)	0.0	0	(−0.9)	2.0	1	(−1.4)	5.0	1	(−0.3)	11.2	10	(2.4)

SOC = Sexual offenses against children; SOA = Sexual offenses against adults; SOC+SOA = Sexual offenses against children and adults; vNSO = Violent non-sexual offenses, nvNSO = nonviolent non-sexual offenses, NO = No offenses; AR = Adjusted residual

We then specifically examined thoughts involving children or a person described as not wanting or enjoying the sexual acts (see Table 3). Only a small proportion of each group (10.7% or fewer) described children. There was a small significant association with group, χ²(4, 264) = 9.53, p = .044, V = .13; SO men reported these more than expected statistically, all of whom were in the SOC or SOC+SOA subgroups. A very small proportion of each group (3.6% or fewer) described a person not wanting or enjoying the sexual acts; there was no significant association between another person’s lack of sexual desire and group, χ²(2, 264) = 1.20, p = .452, V = .07, or lack of enjoyment and group, χ²(2, 264) = 0.003, p > .999, V = .004.

WSFQ

The most frequent item for all groups and subgroups was “Having sex with a loved partner.” All groups and subgroups reported experiencing all four subscales, and for all groups and subgroups, Intimate was most common, followed by Impersonal, Exploratory, and finally, Sadomasochistic (see Tables 1 and 5). Kendall’s tau correlations between IM and the WSFQ subscales or total by group with a multistage Bonferroni correction per group indicated no significant correlations. Kendall’s tau correlations between IM and the WSFQ subscales or total by subgroup indicated only one significant correlation (τ = −.33 between IM and the Intimate subscale for vNSO men, p < .05 with a multistage Bonferroni correction per subgroup).

Table 5. Medians (interquartile range) and Kruskal-Wallis tests for the WSFQ thought context by subgroup.

	Subgroup
	SOC (n = 35)		SOA (n = 17)		vNSO (n = 40)		nvNSO (n = 17)		NO (n = 71)
Subscale/total	Mdn	(IQR)	Mdn	(IQR)	Mdn	(IQR)	Mdn	(IQR)	Mdn	(IQR)	H	df	r
Intimate	28.00	(15.00)	28.00	(19.00)	37.00a	(23.00)	35.00a	(15.50)	37.00a	(15.00)	18.29*	(4)	.23
Exploratory	8.00	(11.00)	7.00	(10.00)	13.00a	(13.75)	17.00ab	(9.00)	15.00ab	(11.00)	26.32*	(4)	.30
Impersonal	11.00	(13.00)	11.00	(9.50)	14.00ac	(12.00)	19.00ab	(11.50)	20.00ab	(11.00)	37.73*	(4)	.35
Sadomasochistic	0.00	(5.00)	0.00	(7.00)	4.00	(17.75)	7.00a	(13.44)	10.00ab	(13.00)	31.78*	(4)	.34
Scale total	49.00	(38.44)	55.00	(37.11)	73.00ab	(56.50)	79.00ab	(38.83)	81.00ab	(41.00)	35.54*	(4)	.38

Higher medians indicate greater thought frequency. SOC = Sexual offenses against children; SOA = Sexual offenses against adults; vNSO = Violent non-sexual offenses; nvNSO = nonviolent non-sexual offenses; NO = No offenses; IQR = Interquartile range. The SOC+SOA subgroup was excluded as n = 6. Significant differences of p < .05 with step-down follow-up analyses:

^asignificantly different from SOC subgroup;

^bsignificantly different from SOA subgroup;

^csignificantly different from NO group. See Text S4 in the OSM for how the tests of difference were conducted.

* p < .05 with a multistage Bonferroni correction.

Discussion

We aimed to identify and empirically examine assumptions about “deviant” sexual fantasy within the DSM’s categories of “sexual deviance” or “paraphilia.” Here, we discuss our empirical findings in relation to each of the three assumptions identified in the Introduction, consider explanations for why these assumptions endure, and draw out the implications of the findings.

Do SO Men Prefer “Deviant” Sexual Fantasies and Have Limited (If Any) “Non-deviant” Fantasies?

The commonest favorite sexual thought for SO men, and all SO subgroups, was Vaginal/oral sex with 1+ woman, typically comprising oral sex and penile-vaginal penetration with their woman partner or another known woman, and often, more than one woman. The WSFQ examined thought frequency, rather than preference, although the findings were consistent. For SO men, and all SO subgroups, the most frequent item was “having sex with a loved partner” and the most frequent subscale was Intimate (including sex with their partner or another known person, giving/receiving oral sex).

As identified in the Introduction, “sexual deviance” or “paraphilia” is not well-defined in the DSM. Arguably, however, the most detailed, and certainly the most current, definition is the one given in DSM-5: “any intense and persistent sexual interest other than sexual interest in genital stimulation or preparatory fondling with phenotypically normal, physically mature, consenting human partners” (APA, 2013, p. 685). The Vaginal/oral sex with 1+ woman and Intimate thought findings are not consistent with the DSM-5ʹs definition of “paraphilia” (APA, 2013), and thus, the findings contest the assumption that SO men prefer “deviant” sexual fantasies and have limited (if any) “non-deviant” fantasies.

Our findings for favorite sexual thought are consistent with Rokach’s earlier study (1990; Rokach et al., 1988). Rokach and her colleagues found that SO men’s most preferred sexual fantasy was “conventional heterosexual” (vaginal penetration, oral sex, sexual touching, “courtship rituals”). Further, our finding for the WSFQ is a robust finding: the WSFQ’s Intimate subscale is most frequent for SO men from a range of industrialized countries and settings (Allan et al., 2007; Bartels et al., 2019; Baumgartner et al., 2002; Cortoni & Marshall, 2001; Gallant & Wormith, 1986; Gannon et al., 2012; Sahota & Chesterman, 1998). Studies using other fantasy checklists are consistent too (Daleiden et al., 1998; O’Donohue et al., 1997; Sheldon & Howitt, 2008); for example, Sheldon and Howitt found that British SOC men most frequently fantasized about consensual acts with an adult woman, including vaginal penetration, giving/receiving oral sex, and genital touching.

Does a Preference for “Deviant” Sexual Fantasies Differentiate SO Men from Men Who Do Not Commit Sexual Crimes?

Less commonly preferred sexual thoughts included Dominance and submission, Fetish or child: Mutual, and Fetish or child: Self. These thought types could be, or certainly would be, consistent with the DSM-5ʹs definition of “paraphilia” (APA, 2013).

However, none of these preferred “deviant” sexual thoughts were unique to SO men. Further, whilst SO men, particularly the SOC+SOA subgroup, were statistically over-represented for the Fetish or child: Mutual type, NO men reported the Fetish or child: Self and Dominance and submission types more than expected statistically. Similarly, preferred thoughts involving children specifically were not unique to SO men, and whilst SO men described these more than expected statistically, this represented only 11% of the SO group and the effect size was small. Lastly, preferred sexual thoughts describing a person as not wanting or enjoying the sexual acts were not significantly associated with group.

Together, these findings temper the assumption that a preference for “deviant” sexual fantasies differentiates SO men from men who do not commit sexual crimes. Rokach (1990) and her colleagues’ (Rokach et al., 1988) analysis of favorite sexual fantasies was less comprehensive; however, their findings were broadly consistent. Preferred sexual fantasies involving sadism, masochism, fetishes or children were reported by a minority of SO, NSO or student men, and for preferred sexual fantasies involving children, whilst the largest minority was in the SO group, this represented only 8% of SO men.

Can Preferred Sexual Fantasies Be Clearly Separated into “Deviant” and “Non-deviant”?

Our findings demonstrated that men’s favorite sexual thoughts are multidimensional; that is, they comprise multiple elements. Each thought type identified through the latent class analysis was demarcated by multiple variables (see Table 2), and the follow-up analyses, and participant extracts, illustrated how multiple elements (e.g., people, sexual practices, emotions) comprised and combined within a thought type. Our findings also showed that none of the thought types solely contained elements that met DSM-5ʹs definition of “paraphilia”; for example, where descriptions of the self, others, interactions between people, sexual practices, and settings within the thought type were all “deviant.”

These findings contest the assumption that sexual fantasies can be clearly separated into “deviant” or “non-deviant.” Instead, these findings suggest that sexual thoughts and fantasies comprise a set of elements on a continuum of “non-deviance” to “deviance.”

Explaining the Longevity of the DSM’s Assumptions about “Sexual Deviance”

There are several reasons why these assumptions about “sexual deviance” are likely to have endured over time. Firstly, as identified in the Introduction, these assumptions are not unique to the DSM; rather, they build on other highly influential – but usually not rigorously tested – ideas. Thus, these assumptions have been shaped over time by influential works and ideas, and related historical, clinical, cultural, political and other factors, leading to them becoming “taken for granted” (but not necessarily empirically evaluated) forms of knowledge.

Second, how we have researched sexual fantasies, and men who sexually offend, is likely to have been influenced by, and subsequently reinforced, these assumptions. For example, if researchers believe that sexual fantasies can or should be separated into “deviant” or “non-deviant,” this leads to creating simplified, often unidimensional, checklist items of ostensibly “deviant” or “non-deviant” fantasies (e.g., “cross-dressing,” “kissing”), which then facilitates the subsequent identification of corresponding “deviance” and “non-deviance” fantasy factors (e.g., Daleiden et al., 1998). This, in turn, serves to reinforce the bifurcation of fantasies, rather than examining and acknowledging the potentially complex, multidimensional nature of sexual fantasies.

Another approach that we commonly take to researching sexual fantasies is to focus on differences between men who sexually offend and men who do not. A focus on difference is common to mainstream science and differences can be important to identify. However, a sole focus on difference can lead to obscuring important findings. For example, Cortoni and Marshall (2001), using the WSFQ, reported that SO men experienced Sadomasochistic fantasies significantly more frequently in adolescence than NSO men in adolescence, concluding that SO men were “preoccupied with sexual themes of control and mastery” (p. 34); however, the authors did not highlight that the most frequent factor overall for SO men in their study was Intimate fantasies. Thus, a focus on difference served to reinforce the assumption that SO men prefer “deviant” sexual fantasies, whereas additionally discussing the overall pattern of sexual fantasy for SO men would have placed those relative differences into context.

A third potential explanation for the endurance of the DSM’s assumptions about “deviant” sexual fantasy is that we can be suspicious or distrustful of findings that refute these assumptions – even when, as outlined earlier, there is consistency in findings across different sexual offending populations and settings, and using different methods. For example, Langevin et al. (1998) reported that SO, NSO and NO men were more likely to fantasize about “non-deviant” than “deviant” acts with women. The authors subsequently suggested that the SO men might be trying to “block out” their “deviant” fantasies, impress the researcher and/or describe fantasies that they wished they experienced. Thus, the findings for SO men were questioned, while the findings for men without sexual offense histories were not. Whilst some social desirability is likely, we found limited relationships between fantasy reports and socially desirable responding, similar to some other studies of SO men (Gallant & Wormith, 1986; Langevin et al., 1998). Further, Stevens et al. (2015) found that only 11% – 14% of the variance in SOC men’s WSFQ scores was explained by socially desirable responding. Therefore, socially desirable responding might play a smaller part in the findings than we are perhaps inclined to believe.

Drawing together these three explanations, one potential overarching explanation for why the DSM’s assumptions about “sexual deviance” endure is that “we” (researchers, clinicians, practitioners, policymakers) have constructed – and continue to construct – the “sexual deviant” or “male sex offender” as “other”; that is, there is a strong belief that “we” must be different from “them.” This belief appears evident not only in the DSM, but also, in earlier influential works and ideas, the approach we take to our research, and how we analyze and interpret our findings. “Othering” of people who commit sexual crimes is not unique to psychiatry; however, constructing “the sex offender” as “the perverted outsider” inhibits us from acknowledging our own potential to commit sexual crimes and that “sex offenders” are typically members of our own families and communities (Spencer, 2009).

Implications of the Findings

The findings for the first two assumptions invite us to question our construction of “the male sex offender.” We should acknowledge that some SO men may prefer sexual thoughts with “non-deviant” elements – and that, even when SO men prefer sexual thoughts with “deviant” elements, NO men can prefer thoughts with these elements, too. Further, we might consider why we have been resistant to accepting these findings to date, and whether this reflects, at least in part, our desire to define ourselves as “normal” and SO men as “other.” It is important to emphasize that these findings in no way detract from the seriousness of the crimes that SO men have committed; rather, the findings help to direct research into other factors that might shape offending for many SO men.

The findings for the third assumption invite us to question our conceptualization of sexual thoughts and the meaning and utility of the term “sexual deviance.” If we continue to approach sexual thoughts as a simple dichotomy, labeling a sexual thought as “deviant” based on a sole “deviant” element, it can impose an artificial boundary between thoughts that accentuates difference, obscuring “non-deviant” elements within the ostensibly “deviant” thought. This labeling of the thought (and thus, often, the individual experiencing it) as “deviant” may have a therapeutically detrimental impact on how the individual perceives themselves, affect the quality and focus of the therapeutic relationship, and how professionals, family, friends and the wider public respond to that individual (cf., Peternelj-Taylor, 2004). Acknowledging the multidimensionality of sexual thoughts also enables practitioners, in collaboration with the client, to consider the implications of the different thought elements, holistically, for the client’s life and behavior. For example, the potential personal and social benefits and negative impacts for that particular client of having a thought with these collective elements at this point in their life, whether or not they would like to enact a thought with these collective elements, and if so, for thoughts involving adults, whether they could do so ethically (e.g., individuals wishing to enact the Dominance and submission type could follow kink community guidelines; Williams et al., 2014). Based on these considerations, the various elements within the thought might be less or more “risky” for a particular individual. In addition, if we see sexual thoughts as comprising a set of elements, this encourages us to explore the varied elements within the thought during therapy and additionally acknowledge “non-deviant” elements within the thought – potentially useful therapeutic targets. Therefore, resisting the simple binary and universal classification of the thought as “deviant” vs. “non-deviant,” or concomitant assumptions of “problematic” vs. “unproblematic,” promotes an idiographic consideration of the various thought elements and any issues with them for that particular client and their context.

Lastly, the findings provide an opportunity for us to consider whether “sexual deviance” or ”paraphilia” are terms we wish to continue using. Arguably, these terms construct sexual thought elements in individual terms – that is, as related to mental illness, or possibly, immorality – and, as discussed above, this might impact on how an individual sees themselves, what options they believe are available to them, and how others respond to them. Instead, we could describe sexual thought elements in social terms – that is, as socially questionable or unacceptable – and thus, construct sexual thoughts as a set of elements that are on a continuum of social transgression. This change in terminology would help to move away from constructions of the “mentally ill” or immoral “deviant” who is perhaps not in control of their own actions, and instead, move us toward recognizing that individuals might have sexual thoughts with socially transgressive elements and that they make choices about whether to act on those thoughts or not.

Study Limitations

Participants were usually White with female sexual partners and all were UK residents. Additionally, the SO men were in a medium-security prison and preferred violent sexual fantasies might be more common in particularly violent SO men, who are typically in higher-security prisons or hospital settings (Baker & White, 2002). Thus, further research in broader (and for some subgroups, larger) samples is needed.

SOA and vNSO men were statistically overrepresented for Vaginal/oral sex with 1+ woman, and SOA and nvNSO men indicated a reduced range of preferred thoughts with socially transgressive elements (no SOA men described the Dominance and submission or Fetish or child: Self types; no nvNSO men described the Fetish or child: Mutual type). When compared to the community men, these findings are surprising. Possibly, previously enjoyed sexual thoughts about acts that could be sexual offenses, or that seem related to them, take on a new and more sinister significance in prison, reducing the enjoyment and personal acceptability of the sexual thought, leading to it being temporarily suppressed, permanently changed, or, if still held, then not reported. SOC men are lowest in a status hierarchy among prison inmates (Ricciardelli & Moir, 2013), and thus, might not experience the same cognitive dissonance. Further research should compare the preferred sexual thoughts of offending men in prison and the community (previously convicted and unconvicted) to determine how incarceration may impact sexual thoughts.

Recognizing the multidimensionality of sexual thoughts, participants could have had different thoughts in mind when endorsing the mainly unidimensional WSFQ items. For example, given that oral sex could occur in four of the five types of favorite sexual thought, participants could have had varied thoughts in mind when endorsing the WSFQ item “receiving oral sex,” including oral sex with children. We partly mitigated this limitation by comparing the WSFQ findings to the multidimensional modeling of favorite sexual thoughts (showing that the most frequent WSFQ item/subscale aligned with the commonest favorite thought type); however, researchers using solely fantasy checklists should exercise caution when interpreting their findings.

As with any nomothetic approach, the favorite sexual thought types summarize the most commonly co-occurring elements within each thought type, and thus, there was individual variation in the nature of the types.

Lastly, the data for the STP were collected between 2005–2007, and possibly, the findings for favorite sexual thoughts might be different now. However, in a recent survey of 4,175 U.S. adults, the four most common favorite sexual fantasy themes were multi-partner sex (including “threesomes” and group sex); Bondage, Discipline, Dominance, Submission, Sadism and Masochism (BDSM); variations in sexual acts (e.g., oral or anal sex), settings (e.g., on the beach) and situations (e.g., trying new sex toys); and “taboo and forbidden sex” (e.g., fetishism or voyeurism; Lehmiller, 2018). Given that these themes were also evident in our analysis of favorite sexual thoughts, this suggests that the findings presented here still hold currency in the present day.

Conclusions

Our findings challenge the assumption that “sex offenders” prefer “deviant” sexual fantasies and have limited (if any) “non-deviant” fantasies; temper the assumption that a preference for “deviant” sexual fantasies differentiates SO men from men who do not commit sexual crimes; and contest the assumption that sexual fantasies can be clearly separated into “deviant” or “non-deviant.” We propose that notions of the “sexual deviant” or “sex offender” as “other” may, at least in part, underpin the continuing endurance of these assumptions, and that these assumptions potentially negatively impact how we do research, therapeutic alliances and/or outcomes, and our understanding of sexual crime. We recommend that: researchers highlight and discuss overall patterns in their findings for each group, in addition to relative differences; researchers and practitioners resist the temptation to classify sexual thoughts as a simple binary phenomenon, rather than as a collection of elements on a continuum; and that we all consider the benefits of framing sexual thoughts in social terms – that is, as having socially transgressive elements – rather than in the pathological, individual terms of “sexual deviance” or ”paraphilia.”

Further research is needed into the personal and social impacts of thoughts containing socially transgressive elements. In the DSM, these thoughts are closely linked to notions of presumed dangerousness; that is, that they can cause harm to self (constructed as distress or impairment of functioning) or harm to others (constructed as fantasies or urges that might be “acted out” or “acted on”; APA, 2000, 2013). We need to interrogate, empirically, whether these thoughts and fantasies, and the elements therein, are indeed “dangerous,” and if so, how, to whom and in what circumstances. Although some studies have begun to explore this (e.g., Ahlers et al., 2011; Bailey et al., 2016; Brown et al., 2020; Turner-Moore & Waterman, 2017), much more research is needed. Crucially, without evidence for the “dangerousness” of fantasies with socially transgressive elements, a DSM diagnosis of “paraphilic disorder,” informed by a preference for “deviant” sexual fantasies, becomes simply a means to regulate socially undesirable aspects of a person’s inner world.

Acknowledgments

We would like to thank Paul Smith for his input during the early stages of this project, Dave Horton for software development, and Robert West for his assistance with the latent class analysis. This research was supported by a doctoral scholarship from the University of Leeds.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Data Availability Statement

Permission was not sought from participants at the time of data collection to share their data publicly, and as the data are anonymous, retrospective permission cannot be sought.

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00224499.2022.2109568

Additional information

Funding

This work was supported by a University of Leeds doctoral scholarship