Archives of Sexual Behavior

, Volume 39, Issue 2, pp 304–316

The DSM Diagnostic Criteria for Pedophilia

Original Paper

DOI: 10.1007/s10508-009-9536-0

Cite this article as:
Blanchard, R. Arch Sex Behav (2010) 39: 304. doi:10.1007/s10508-009-9536-0


This paper contains the author’s report on pedophilia, submitted on June 2, 2008, to the work group charged with revising the diagnoses concerning sexual and gender identity disorders for the fifth edition of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM). The author reviews the previously published criticisms and empirical research concerning the diagnostic criteria for pedophilia and presents criticism and relevant research of his own. The review shows that the DSM diagnostic criteria for pedophilia have repeatedly been criticized as unsatisfactory on logical or conceptual grounds, and that published empirical studies on the reliability and validity of these criteria have produced ambiguous results. It therefore seems that the current (i.e., DSM-IV-TR) diagnostic criteria need to be examined with an openness to major changes in the DSM-V.


DSM-V Hebephilia Paraphilia Pedophilia Pedohebephilia Penile plethysmography Sexual offending 


On June 2, 2008, I submitted a report on pedophilia to the work group charged with revising the diagnoses concerning sexual and gender identity disorders for the fifth edition of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM). That report is reproduced in the remainder of this paper, beginning in the next section. I have made no changes to the original text, except to update the references where possible.

The original report included my proposal for a revised set of diagnostic criteria. In the year since I submitted my report, these diagnostic criteria have been extensively modified and—in my view—improved by input from the Paraphilias Subworkgroup of the Sexual and Gender Identity Disorders Work Group and from official Advisors to the Paraphilias Subworkgroup. Thus, the diagnostic criteria presented later in this paper are substantially different from the diagnostic criteria currently being considered by the Paraphilias Subworkgroup, and they are almost certainly different from the criteria that will eventually be approved by the DSM-V Task Force and the Board of Trustees of the American Psychiatric Association. I have included them because they were part of my original report, and because they help to document the evolution of the diagnostic criteria that will eventually form part of the DSM-V.

Report on Pedophilia

According to the DSM-IV-TR (American Psychiatric Association, 2000), “The paraphilic focus of Pedophilia involves sexual activity with a prepubescent child” (p. 571). The DSM diagnostic criteria for pedophilia have repeatedly been criticized as unsatisfactory on logical or conceptual grounds, and published empirical studies on the reliability and validity of these criteria have been interpreted by their authors as reinforcing that conclusion. According to Marshall (1997), the diagnostic utility of the DSM diagnostic criteria is so low that these criteria are virtually ignored by clinicians as well as researchers. Marshall’s observations are presumably based on his experience in Canadian settings, and it is possible that American clinicians are necessarily forced to make greater use of DSM diagnostic criteria for legal or administrative purposes, whether they regard these criteria as useful or not. O’Donohue, Regev, and Hagstrom (2000), however, writing about DSM-IV (American Psychiatric Association, 1994) from an American perspective, endorsed Marshall’s (1997) observations regarding the practical irrelevance of DSM criteria. It therefore seems that the DSM-IV-TR diagnostic criteria need to be examined with an openness to major changes in the DSM-V.

In this paper, I review the previously published criticisms and empirical research concerning the diagnostic criteria for pedophilia, present criticism and relevant research of my own, propose a revised set of diagnostic criteria for the consideration of the Sexual and Gender Identity Disorders Work Group, and explain the rationale for the wording that I propose. I naturally make frequent reference to the diagnostic criteria for pedophilia in the DSM-III (American Psychiatric Association, 1980), DSM-III-R (American Psychiatric Association, 1987), DSM-IV, and DSM-IV-TR. These criteria are reproduced in the Appendix to this paper.

History and Overview of the Diagnostic Criteria

DSM-III had only one key diagnostic criterion, Criterion A, which concerned signs and symptoms of pedophilia. From DSM-III-R onward, there have been two key diagnostic criteria. Criterion A still concerned signs and symptoms. Criterion B concerned distress and impairment. Both criteria had to be satisfied to diagnose the disorder of pedophilia.

In DSM-III, Criterion A included acts and fantasies involving sexual interference with children. Sexual acts were clearly conceptualized as signs of pedophilia.

In DSM-III-R, sexual acts were removed from Criterion A, leaving sexual urges and fantasies about children as the designated symptoms. Sexual acts were inserted into the newly formulated Criterion B, which states, “The person has acted on these urges, or is markedly distressed by them.” The grouping of sexual acts with psychological distress in a clinical significance criterion implies that sexual acts are de facto evidence of psychosocial impairment.1 In other words, the role of sexual acts was changed from signaling that pedophilia is present to signaling that it is clinically significant.

In DSM-IV, sexual acts were reinstated in Criterion A as signs of pedophilia. Sexual acts were still mentioned in Criterion B, not as de facto evidence of impairment, but as one of the signs and symptoms of pedophilia that might (or might not) result in distress or impairment. This meaning, intended or not, is implied by the wording of Criterion B: “The fantasies, sexual urges, or behaviors cause clinically significant distress or impairment in social, occupational, or other important areas of functioning.”

In DSM-IV-TR, the wording of Criterion A remained identical to that in DSM-IV. The wording of Criterion B, however, was changed back to resemble that of Criterion B in DSM-III-R: “The person has acted on these sexual urges, or the sexual urges or fantasies cause marked distress or interpersonal difficulty.” Thus, in DSM-IV-TR, the datum, sexual acts, has been used in two different ways. In Criterion A, it is evidence that the patient is pedophilic. In Criterion B, it is evidence that the patient’s pedophilia is materially affecting his or her functioning in society. In other words, sexual acts simultaneously indicate that pedophilia is present and that it is causing problems.

A useful way to conceptualize the diagnostic criteria in DSM-IV-TR is the following: There is one sufficient condition for diagnosing pedophilia—a history of sexual acts involving children. That is sufficient because sexual acts satisfy the signs/symptoms criterion and the distress/impairment criterion. There are no necessary conditions for diagnosing pedophilia. Either fantasies or urges can be used to make the diagnosis, provided they are accompanied by marked distress or interpersonal difficulty.

Prior Logical and Conceptual Criticism

Role of Sexual Acts in the Diagnostic Criteria


First and Frances (2008) have recommended that Criterion A for all paraphilias be restored to its DSM-III-R wording, that is, that sexual acts or behaviors should be removed from it. Although First and Frances write about paraphilias in general, their major examples are pedophilia and rape (which is not a paraphilia per se). When they make their argument against the inclusion of sexual acts in Criterion A, they use the example of rape:

The addition of “or behaviors” [to Criterion A in DSM-IV] led some forensic evaluators to conclude that sexual offenders might qualify as having a mental disorder based only on their having committed sexual offenses (e.g., rape)…. The revised Criterion A wording has sometimes been used to justify making a paraphilia diagnosis based solely on a history of repeated acts of sexual violence, which is then argued as satisfying the statutory mandate for the presence of a “mental abnormality”…. Defining paraphilia based on acts alone blurs the distinction between mental disorder and ordinary criminality. (p. 1240)


First and Frances’s argument against diagnosing paraphilia from sexual offenses seems reasonable if not compelling when the clinical issue is diagnosing paraphilia—they do not say what paraphilia—from multiple episodes of rape. It breaks down when the clinical issue is that of diagnosing pedophilia.

In clinical practice, the patient’s history of sexual offenses against children is often the only basis for making a diagnosis of pedophilia. It is well established that self-report alone cannot be used to diagnose pedophilia in offenders against children (see, e.g., Kingston, Firestone, Moulden, & Bradford, 2007; Marshall, 1997; O’Donohue & Letourneau, 1993; O’Donohue et al., 2000; Wormith, 1983). Men whose histories of sexual offending against children are so extensive that they cannot plausibly be explained by anything besides pedophilia may nonetheless deny that they have a sexual preference for children or else claim that they had “a problem” in the past but that their sexual feelings for children have now disappeared.2

The widespread clinical opinion that self-report is unreliable in pedophiles has been reinforced by laboratory studies. In these studies, sexual interest in children was measured with phallometric testing, a procedure for assessing erotic interests in male adults and adolescents. In this procedure, the examinee’s penile blood volume is monitored while he is presented with a standardized set of laboratory stimuli depicting a variety of potentially erotic activities or objects. The examinee’s penile blood volume increases (i.e., degrees of penile erection) are taken as an index of his relative attraction to the different classes of stimuli. When phallometric testing is used to measure erotic age-preference, the laboratory stimuli include visual and auditory representations of children and adults.

In a series of studies in my laboratory, my predecessor, Kurt Freund, M.D., D.Sc., and I specifically studied men who had committed sexual offenses against children but who claimed that they were sexually attracted only to adults (Blanchard, Klassen, Dickey, Kuban, & Blak, 2001; Blanchard et al., 2006; Freund & Blanchard, 1989; Freund & Watson, 1991). One example will suffice. Blanchard et al. (2001) studied 59 men who had charges, convictions, or credible accusations of illegal sexual behavior involving three or more unrelated (male or female) children under the age of 12, no charges (etc.) involving persons age 15 or older, and no charges involving related persons of any age. These patients stated in interview that they felt a greater sexual attraction to females age 17 and older than to any other class of person. The self-report of the majority was directly contradicted by their laboratory results. On phallometric testing, 61% produced substantially greater penile tumescence to audiovisual depictions of children than to depictions of adults. When the same phallometric test and diagnostic cutting score were applied to 27 sex offenders who had extensive histories of sexual activity with (consenting or nonconsenting) females age 17 and older, only 1 (4%) was classified as pedophilic.

Although phallometric testing can sometimes be useful, especially when conducted in laboratories that calculate and adjust their diagnostic cutting scores to maintain high specificity, it is not widely available. Because of the general unavailability of phallometric testing (or alternative laboratory tests) and because of the general unreliability of self-report in pedophiles, repeated sexual acts involving children are practically indispensable as a diagnostic sign of pedophilia. The use of sexual acts as de facto evidence of psychosocial impairment is a somewhat different matter that should be considered separately.

Paradoxical Effects of the Distress/Impairment Criterion


The attempt to separate the diagnostic criteria for pedophilia (and other paraphilias) into signs and symptoms (Criterion A) vs. distress or impairment (Criterion B) has not been accompanied by an appropriate adjustment to terminology. This has led to the unsatisfactory result that it is necessary to be distressed or impaired by a paraphilia in order to have a paraphilia. The problem has been partially patched over in DSM-IV-TR by substituting societal judgments about impairment for the patient’s. Thus, a man who has an erotic preference for children and who engages children sexually in real life is a pedophile, regardless of his feelings about his situation, because sexual acts with children count as impairment. This solution has not, however, been adequate in the eyes of all critics. Green (2002) wrote of the DSM-IV-TR:

So what then of the pedophile who does not act on the fantasies or urges with a child? Where does the DSM leave us? In Wonderland. If a person does not act on the fantasies or urges of pedophilia, he is not a pedophile. A person not distressed over the urges or fantasies and who just repeatedly masturbates to them has no disorder. But a person who is not distressed over them and has sexual contact with a child does have a mental disorder. (p. 470)


One solution to this diagnostic conundrum might be applied to the paraphilias in general. The DSM-V could distinguish between paraphilias and paraphilic disorders. A patient who satisfied the signs and symptoms criterion (Criterion A in DSM-IV-TR) would be ascertained—not diagnosed—as having a paraphilia. A patient who satisfied the signs/symptoms criterion and the distress/impairment criterion (Criterion B in DSM-IV-TR) would be diagnosed as having a paraphilic disorder. This solution should be especially useful to researchers. It would prevent a paraphilia from becoming invisible to clinical science just because it lacks any secondary effect of disturbing the individual or others.

The hypothetical patient conjured by Green (2002) represents a particularly challenging test of this conceptualization. It is therefore worthwhile re-examining Green’s example in more detail. Suppose there exists a pedophilic man whose sexual interest is solely directed at children. His masturbation fantasies exclusively concern children, and he feels no self-disgust after ejaculation. He feels no dissatisfaction with his pedophilic orientation in general and he has no wish to be otherwise. He feels sexual “urges” toward children, but he has never approached a child sexually, and there is no possibility that he would ever do so. He does not even participate in sexual offenses against children at second hand by accessing child pornography.

According to the distinction I proposed earlier, the hypothetical patient has a paraphilia but not a paraphilic disorder. The professional-acceptance test of the proposed terminology is this: How many clinicians would be comfortable with the conclusion that this man has no disorder? The answer is probably: Not many. What prevents this from posing a serious practical problem is that few real patients are likely to match the profile of the hypothetical patient. Such a combination of behaviors and attitudes, in real life, would be very rare. How could one experience a lifetime of sexual “urges,” which are never satisfied, with no sense of frustration? If the absence of any real-life gratification causes no distress at all, can one really say there was an “urge” in the first place?

In conclusion, the proposed terminology identifies Green’s hypothetical patient as a pedophile whereas the DSM-IV-TR does not. The distinction between paraphilias and paraphilic disorders may actually be more compatible with the separation of diagnostic criteria into signs/symptoms and distress/impairment than is the current DSM language.3

The Meanings of Recurrent and Intense


O’Donohue et al. (2000) criticized various aspects of Criterion A in DSM-IV, which reads, “Over a period of at least 6 months, recurrent, intense sexually arousing fantasies, sexual urges, or behaviors involving sexual activity with a prepubescent child or children.” Criterion A has the identical wording in DSM-IV-TR, so their comments would apply to the current DSM as well.

As an overall evaluation, O’Donohue et al. state that Criterion A “seems too vague and thus precludes the clinician from assessment without making inferences…Because each clinician might draw different inferences, the reliability, and thereby the validity, of the criterion is reduced” (p. 99). They note that there is no definition of recurrent (beyond “more than once”) or of intense. In other words, the DSM specifies inherently quantitative indicators but does not specify the critical threshold quantities. A similar objection was raised by Marshall (1997), who wrote, “It would improve things if future diagnostic manuals were to specify what ‘recurrent’ means with respect most particularly to behavior, but also for fantasies and urges” (p. 154).


The language criticized by Marshall and O’Donohue et al. was introduced in DSM-III-R. One way of addressing this criticism involves returning to the model of an earlier DSM.

In some ways, the approach to quantifying pedophilic feelings in DSM-III was more elegant than in the later editions. Criterion A of DSM-III reads, “The act or fantasy of engaging in sexual activity with prepubertal children is a repeatedly preferred or exclusive method of achieving sexual excitement.” If one makes the reasonable assumption that “preferred” means “preferred over adults,” then the criterion can be interpreted to mean that a pedophile is someone who is more attracted to children than to adults. That notion can readily be applied to self-report. Patients who are willing and able to describe their erotic preferences at all can almost certainly say whether their sexual feelings for children are greater than, less than, or approximately equal to their feelings for adults. The notion can just as easily be applied to phallometric testing or to any other method for laboratory measurement of sexual response that might be devised in the future (e.g., fMRI).4 Even the most primitive laboratory quantifications, if they are clinically usable at all, will allow the practitioner to determine whether patients’ sexual responses to children are greater than, less than, or approximately equal to their responses to adults. If one wants to minimize false positive results, one can limit the ascertainment of pedophilia to those examinees who respond substantially more to children than to adults (e.g., Blanchard et al., 2001).

The foregoing approach could not be applied to the patient’s sexual history, that is, one could not reliably ascertain patients’ erotic age-preferences by calculating whether the number of children they have engaged sexually is greater than, less than, or equal to the number of adults. The variables of sexual experiences with children and sexual experiences with adults are influenced by too many factors besides the patient’s preferences: (a) Sexual interaction with consenting adults is legal in most jurisdictions, whereas sexual interaction with children is a criminal offense, whether the children are consenting or not. (b) Opportunities to meet adults and to be alone with adults in privacy are much greater than opportunities to meet (unrelated) children and to be alone with them. (c) Social pressures would tend to push pedophiles to experiment sexually with adults in hopes of finding them acceptable sexual partners, whereas social mores would tend to discourage anyone, pedophilic or not, from experimenting sexually with children. (d) Law and social norms would encourage pedophiles to make use of adults as “second-best” sexual outlets in place of children, but these factors (in contemporary society, anyway) discourage the use of children as substitutes for adults.

For the foregoing reasons, some other approach must be used to make inferences about erotic age-preference from sexual history data. I discuss this matter in a later section.

The Domain of Relevant Behaviors


O’Donohue et al. raise a more subtle problem regarding the seemingly clear term behaviors in Criterion A of DSM-IV and DSM-IV-TR:

Another question pertaining to the first criterion is what characterizes “behavior”? If a person chooses to work as a school bus driver because it fulfills a sexual desire to be around children, is that choice considered a behavior that is sufficient to fulfill the criterion? Suppose that the driver has not actually touched a child in an inappropriate manner, but is clearly behaving because of his or her sexual attraction. Does that constitute a behavior that is sufficient to meet this criterion? Another noncontact behavior, for example, might be purchasing child pornography. Would that constitute a behavior that is sufficient to meet the criterion? Should clinicians be assessing microresponses, such as staring at children, in order to assess for pedophilia? Could this constitute relevant behavior for the diagnosis? Again, because the criterion is unclear, it becomes difficult for clinicians to reliably diagnose this disorder. (p. 100)


O’Donohue et al.’s comments are not without merit and their examples are not unrealistic. My laboratory has, in fact, received referrals from group homes for mentally retarded persons when a patient’s intense staring at children alerted staff to possible pedophilia. In practice, however, behaviors such as staring or arranging to be in the company of children are not feasible as primary signs of pedophilia (although they might contribute to a clinician’s confidence in his or her diagnosis).

The acquisition of child pornography is another matter. Analysis of data from my laboratory has shown that child pornography use may actually be a stronger indicator of pedophilia than is sexual offending against children (Seto, Cantor, & Blanchard, 2006; see also Blanchard et al., 2007). Another behavior that should be considered in the next revision of the DSM is a patient’s sexual chat and/or attempts to arrange a meeting with a police officer posing as a child on the Internet.

Duration of Signs and Symptoms


O’Donohue et al. (2000) question whether the DSM-IV (and DSM-IV-TR) Criterion A requirement that signs and symptoms have persisted for 6 months is justified:

the characteristic of the fantasies, urges, or behaviors recurring over a 6-month period is problematic. The inclusion of a minimal temporal criterion is understandable in order to refer to something that has some temporal stability. What is less clear is why 6 months?…. According to Dohrenwend and Dohrenwend (1965), temporal stability of symptoms is essential because valid diagnoses must rule out the possibility of transient stressors (such as combative conditions) mimicking the symptoms of a disorder (PTSD). This is not a concern regarding pedophilia. There are no transient stressors that can account for this disorder in the short-term. (p. 101)


I agree with this criticism, including the last sentence. There is evidence that the probability of pedophilia increases with the number of victims (see below), but I do not know of evidence that the probably of pedophilia relates to the time interval between victims. It is possible that certain acute situations (e.g., manic episodes, drug or alcohol binges) might cause a person to approach several children within the space of a few days, or to approach two or more children (e.g., siblings or playmates) in a single episode. Generally speaking, however, these exceptional situations are easy to identify.

Number of Sexual Acts Involving Children


It is now widely accepted that not all child molesters are pedophiles, and not all pedophiles are child molesters (e.g., Konopasky & Konopasky, 2000; Seto, 2002). The existence of pedophiles who never approach a child sexually poses a problem for the distress/impairment criterion. The existence of persons who have engaged children sexually but do not prefer children poses a problem for the signs/symptoms criterion. The solution to the signs/symptoms criterion involves the answer to this: How does one use information about sexual acts with children to decide which child molesters are probably pedophiles and which are not?

O’Donohue et al. touch upon the problem of deciding which child molesters are pedophiles in a few places. In their first mention of this matter, they observe the following:

In the DSM-III-R, only urges and fantasies were relevant for satisfying the first criterion. In the DSM-IV these or behaviors can satisfy this criterion. This could be viewed as a positive change as it allows the clinician to rely on overt phenomena to make this diagnosis. However, it could also be problematic. The basic question is whether there are two kinds of cases that should remain distinguished. The first kind of case is represented by an inclination, propensity, or motivation—an underlying diathesis. The second kind of case is represented by the presence of disordered behavior, which may or may not be related to the diathesis. (pp. 100–101)

I understand this to mean that the personological characteristic underlying a specific act of child molestation could be either pedophilia or something different, such as antisociality plus opportunity. In a later passage they seem to suggest, although not in these words, that the qualitative distinction might be made on quantitative grounds:

We must ask what the best alternative might be to increase the specificity of the DSM diagnostic criteria. One possibility is specifying a number of occurrences during a time period and using it as a cutoff. (p. 101)


O’Donohue et al.’s use of the phrase, “during a time period,” is puzzling, given their previous criticism of the 6-month requirement. Other than that, their suggestion that a numeric cutoff be applied to number of sexual acts (or number of sexual victims) in ascertaining pedophilia accords with empirical data.

As previously stated, Blanchard et al. (2001) found that 61% of men with sexual offenses against three or more children produced substantially greater penile tumescence to audiovisual depictions of children than to depictions of adults. This test result was found for 42% of men with offenses against two children and 30% of men with offenses against one child. Thus, there clearly is a correlation between the number of sexual offenses against children and the presence of pedophilia, even among men who deny any sexual interest in children.

For reasons already explained, pedophilia cannot be ascertained from patients’ numbers of sexual encounters with children relative to their numbers of encounters with adults. It is necessary to consider the absolute number of sexual encounters with children. The results of Blanchard et al. (2001) show that absolute cutoff scores matter, at least up to three known offenses. The problem of having to choose the best cutoff value may therefore be unavoidable.

Quantitative Threshold for Sexual Acts


O’Donohue et al. (2000) make a rather radical suggestion about the number of sexual acts with children needed to diagnose a disorder:

As an alternative to viewing pedophilia as a trait, it can be viewed as a behavioral disorder. As such, a single behavior of a sexual nature would be sufficient to categorize someone as having pedophilia response disorder. The extent to which the behavior(s) persist would be subsumed under the subcategory of a single occurrence that is acute in its course, or under a more chronic condition. A single instance of sexual behavior with a child should be sufficient to label someone as having a disorder. We argue that a single incidence would be sufficient on three grounds: (a) from epidemiological data, one incidence places the adult in an infrequent subgrouping; (b) it is the only nonarbitrary demarcation—none clearly would be inappropriate; and (c) one incident can cause significant harm to a child and an adult. (p. 103, emphasis in original)


I have two points to make in response to this. The first is that O’Donohue et al.’s proposal to reconceptualize pedophilia as a behavioral disorder and rename it pedophilia response disorder has nothing to do with my proposal to distinguish between paraphilias and paraphilic disorders. I regard paraphilias, including pedophilia, as erotic preferences or orientations that inhere in the individual and that have some existence independent of specific, observable actions.

The second is simply that a large proportion of persons who have offended sexually against a child are not pedophiles. Their erotic preference is for physically mature adults, and their sexual behavior with children is caused by some other motivational state or circumstance. Labeling them as persons with “pedophilia response disorder” is merely restating that they have offended against a child.

I have to agree with O’Donohue et al. that “1” is a unique number and that one sexual offense against a child places a person in a statistically infrequent (and suspicious) category. I do not, however, think it follows from that that “1” is the optimum cutoff for ascertaining pedophilia.

Clinical Significance, Distress, and Impairment


O’Donohue et al. have two different criticisms about the references to “distress” and “impairment” in DSM-IV, which would also apply to the references to “distress” and “interpersonal difficulty” in DSM-IV-TR. The first criticism is that the clinical significance criterion is badly worded; the second criticism is that it is not needed at all. The first criticism is relatively minor: “It is unclear about what constitutes ‘clinically significant’ distress. Does clinically significant stress need to result in a stress-related Axis I diagnosis—or is the standard weaker?” (pp. 101–102). Their second criticism goes to the heart of the distress/impairment criterion for pedophilia.

Moreover, what constitutes impairment in social functioning? A person should be considered impaired by the mere fact of having sexual fantasies, urges, or behaviors targeting children instead of people their own age. Given that Criterion A is met, it could be construed that Criterion B is always met. It does not seem possible for a person sexually interested in children not to be socially impaired in some way because societal norms dictate that it is abnormal for a person to be sexually interested in children.

In addition, why does a person need to be distressed by the fact that he or she is attracted to children in order for the diagnosis of pedophilia to be made? By the mere fact that an attraction exists, the diagnosis of pedophilia is warranted…. We recommend that Criterion B be removed from the DSM diagnostic criteria of pedophilia. (p. 102)

Marshall (1997) is in essential agreement with O’Donohue et al. about the superfluity of the clinical significance criterion: “That pedophilia should be diagnosed only if it causes significant distress or impairment of functioning seems an odd caveat to add to the diagnostic criteria” (p. 154).


Both Green (2002) and O’Donohue et al. call special attention to the “contented pedophile” (O’Donohue et al., 2000, p. 104), although Green would solve the problem by taking pedophilia out of the DSM, whereas O’Donohue et al. would solve the problem by taking Criterion B out of the DSM. The classification of ego-syntonic, euthymic, chaste pedophiles may be viewed as a psychiatric example of the generalization that “hard cases make bad law.” As I have already indicated, I doubt that such cases are common, compared with the numbers who are distressed by their pedophilia or else are comfortable enough with it and are prepared to interact with a child when the opportunity presents. As I also indicated, I think that ascertaining such hypothetical cases as pedophilic without diagnosing them as having a pedophilic disorder would be a reasonable compromise.

I have no suggestions for quantifying “marked distress” or “interpersonal difficulty” or for determining the threshold values that would trigger the application of the distress/impairment criterion. If “interpersonal difficulty” simply means that patients are sexually attracted to children rather than to adults, then it is redundant with their ascertainment as pedophiles, as O’Donohue et al. sensibly point out.

Definition of Pedophilia and Age of Erotic Objects


The DSM-IV-TR follows the traditional definition of pedophilia as sexual interest in, or sexual activity with, prepubescent children. This definition, if taken literally, would exclude from diagnosis a sizable proportion of patients whose strongest sexual feelings are for physically immature persons. The existence of patients whose erotic interest centers on pubescent rather than prepubescent children has been recognized for decades. Glueck (1955) coined the term hebephiles to refer to them. Despite the familiarity of this phenomenon to experienced clinicians, few have criticized the DSM for ignoring it. One notable exception is Marshall (1997), who wrote of the DSM-IV:

The age specified for the child identified as the object of the pedophile’s fantasies, urges, or behavior also presents problems. Whereas there must be some cutoff age, defining pedophilia as an attraction to or involvement with prepubescent children, and defining pubescence as typically age 13 years, seems arbitrary…. Also, a significant number of offenders molest victims who are postpubescent but still quite young. Does this mean that these offenders do not have a mental disorder when those who molest younger children do? (p. 154)


Marshall’s criticism is reinforced by the findings of a recent study from my laboratory (Blanchard et al., 2009; see also Blanchard, 2009). We began this study by reviewing developmental research from the field of pediatric endocrinology, which showed that (contemporary) pubescent children are generally those from age 11 or 12 years to about 14 or 15; prepubescent children are those who are younger. We then reviewed data on the typical ages of victims of child sexual abuse, which yielded the following information.

The modal age of victims of sexual offenses in the United States is 14 years (Snyder, 2000, Fig. 1; Vuocolo, 1969, p. 77); therefore, the modal age of victims falls within the time-frame of puberty. In anonymous surveys of social organizations of persons who acknowledge having an erotic interest in children, attraction to children of pubescent ages is more frequently reported than is attraction to those of prepubescent ages (e.g., Bernard, 1975; Wilson & Cox, 1983). In samples of sexual offenders recruited from clinics and correctional facilities, men whose offense histories or assessment results suggest erotic interest in pubescents sometimes outnumber those whose data suggest erotic interest in prepubescent children (e.g., Cantor et al., 2004; Gebhard, Gagnon, Pomeroy, & Christenson, 1965; Studer, Aylwin, Clelland, Reddon, & Frenzel, 2002). The foregoing findings are consistent with the results of large-scale surveys that sampled individuals from the general population and included questions regarding sexual experiences with older persons when the respondent was underage. These results suggest that a substantial proportion of respondents who had had such experiences reported ages at occurrence that fall within the normal time-frame of puberty (Boney-McCoy & Finkelhor, 1995; Briere & Elliott, 2003; Finkelhor, Ormrod, Turner, & Hamby, 2005). The available data therefore indicate that hebephilia may be as great a clinical problem as pedophilia.

Blanchard et al. (2009) also reviewed studies demonstrating the utility of specifying a hebephilic group for research purposes. These studies have compared pedophilic, hebephilic, and teleiophilic5 men on a variety of dependent measures. The results have shown hebephiles to be intermediate between pedophiles and teleiophiles with regard to IQ (Blanchard et al., 2007; Cantor et al., 2004), completed education (Blanchard et al., 2007), school grade failure and special education placement (Cantor et al., 2006), head injuries before age 13 (Blanchard et al., 2003), left-handedness (Blanchard et al., 2007; Cantor et al., 2005), and stature (Cantor et al., 2007).

The main goal of Blanchard et al. (2009) was to validate the concept of hebephilia by examining the agreement between self-reported sexual interests and objectively recorded penile responses in the laboratory. The participants were 881 men who were referred for clinical assessment because of paraphilic, criminal, or otherwise problematic sexual behavior. Within-group comparisons showed that men who verbally reported maximum sexual attraction to pubescent children had greater penile responses to depictions of pubescent children than to depictions of younger or older persons. Between-groups comparisons showed that penile responding distinguished such men from those who reported maximum attraction to prepubescent children and from those who reported maximum attraction to fully grown persons. These results indicated that hebephilia exists as a discriminable erotic age-preference.

The implication of the foregoing study is that the DSM-V should recognize the clinical and scientific importance of patients preferentially attracted to children who have entered puberty but are still physically quite immature. This would systematize what is already happening unsystematically. Levenson (2004, p. 360) has noted that practitioners evaluating patients for civil commitment under sexually violent predator statutes typically diagnose such patients with “Paraphilia NOS (Hebephilia).”

Studies of Reliability and Validity

Research by Kingston et al. (2007)


Kingston et al. (2007) studied adult men who had been convicted of hands-on sexual offences against an unrelated male or female child who was under the age of 16 at the time of the offence. The patients were assessed at a university teaching hospital in Ottawa, Ontario between 1982 and 1992. If police records indicated that a patient had ever offended against an adult or against a family member, he was excluded from the analysis. The patients were diagnosed by psychiatrists as pedophilic or not pedophilic according to DSM-III or DSM-III-R criteria (hereafter, DSM pedophiles and DSM nonpedophiles, respectively). The psychiatrists had access to previous medical charts and police reports, including diagnostic history, previous psychological assessment, psychosocial history, and criminal history.

After their clinical psychiatric diagnosis, the patients were tested in the hospital’s phallometric laboratory. Phallometric test results were obtained for 82 DSM pedophiles and 75 DSM nonpedophiles. The patient’s penile responses were used to compute a Pedophile Index, which was the highest response to a child divided by the highest response to an adult. Thus, scores greater than 1.0 would indicate a pedophilic preference. The authors also computed a Pedophile Assault Index, which was the highest response to depictions of violent or coercive interactions with children divided by the highest response to depictions of cooperative or enthusiastic children. It appears that depictions of sexual interaction with adults were not used in computing the Pedophile Assault Index.

The mean scores of both groups on the Pedophile Index were greater than 1.0, which is not very surprising, given that both groups had histories of sexual offenses against children. Nevertheless, the DSM pedophiles had significantly higher scores (more arousal to children) than the DSM nonpedophiles. In contrast, the mean scores of both groups on the Pedophile Assault Index were less than 1.0 (i.e., sadistic behavior toward children was less arousing than nonsadistic behavior), and the means of the two groups were virtually identical.

In an additional manipulation, Kingston et al. computed a Phallometric Deviance Index, by combining the Pedophile Index and the Pedophile Assault Index. It is unclear what the interpretation of this measure is supposed to be, since one component of it concerns arousal to children vs. adults, whereas the other component of it concerns arousal to coerced/mistreated children vs. cooperative children. It is also unclear why the authors would choose to combine measures that had already been shown to behave differently. Using this derived variable, they obtained the unsurprising result that “There was no significant relationship between individuals diagnosed as pedophilic according to the DSM criteria and individuals classified as pedophilic according to [the Phallometric Deviance Index]” (p. 431).


The significant result obtained with the Pedophile Index supports the validity of DSM-based psychiatric diagnosis in a general sort of way. It would be misleading to attempt any kind of effect size analysis here, because there was no comparison group of men who lacked a history of offenses against children. The application of these findings to the DSM-IV-TR diagnostic criteria is imprecise in any event, since the psychiatric diagnoses were made according to DSM-III or DSM-III-R criteria.

Research by Levenson (2004)


Levenson (2004) studied 277 male, adult, competent, convicted sex offenders in Florida prisons who received a face-to-face evaluation by psychologists or psychiatrists for sexually violent predator civil commitment between July 1, 2000 and June 30, 2001. Consistent with statutory language, these subjects were examined by more than one forensic evaluator.

DSM diagnoses were made according to DSM-IV-TR criteria. The diagnoses were coded dichotomously (yes/no) and included the diagnoses most commonly considered: Pedophilia, Sexual Sadism, Exhibitionism, Paraphilia NOS, Antisocial Personality Disorder, Personality Disorder NOS, Other Personality Disorder, Substance Disorder, and Other Major Mental Illness.

The kappa reliability coefficient for pedophilia was .65. Levenson considered this value to be merely “fair,” considering the serious consequence of civil commitment following incarceration. The kappa reliability coefficients for sadism, exhibitionism, and paraphilia NOS were even lower, ranging from .30 to .47. Levenson noted that it would have been useful to analyze the DSM-IV criteria for each diagnosis to determine if particular criteria were more or less reliable than others, but these data were not available to her.


Levenson’s finding for the reliability of the DSM diagnosis of pedophilia is not as bad as one might have feared, given the very negative assessment by Marshall (1997). Furthermore, Packard and Levenson (2006) reanalyzed Levenson’s (2004) data using alternative measures of inter-rater reliability and concluded that the reliabilities of DSM paraphilia diagnoses (including pedophilia) are generally better than indicated by a sole reliance on the kappa statistic.

Packard and Levenson found that the prevalence-adjusted bias-adjusted kappa (PABAK; Byrt, Bishop, & Carlin, 1993) for the diagnosis of pedophilia was .70, and they pointed out that this value would be considered a “substantial” level of inter-rater agreement by some statisticians (Landis & Koch, 1977). Other statistics of present interest were various proportions of agreement. Among all cases, the evaluators agreed on the presence or absence of pedophilia 85% of the time. Among cases who received at least one negative diagnosis, the evaluators agreed on a negative diagnosis 80% of the time (proportion of negative agreement). Among cases who received at least one positive diagnosis, the evaluators agreed on a positive diagnosis 62% of the time (proportion of positive agreement).6 In summary, the reliability of the DSM-IV-TR diagnostic criteria for pedophilia, as re-assessed by Packard and Levenson, could be seen as acceptable.

On the other hand, the diagnostic assessments may not have been truly independent in all cases, because the second evaluator might have been aware of the opinions of the first (Packard & Levenson, 2006). This could have had the effect of inflating the agreement between raters. On balance, therefore, one may conservatively conclude that Levenson’s (2004) data indicate that there is still room for improvement in the reliability of the DSM diagnosis of pedophilia.

Proposed Diagnostic Criteria for DSM-V

General Considerations

In proposing a revised set of diagnostic criteria for DSM-V, I have attempted to combine the best features from previous versions of the DSM with new features suggested by the criticism and research reviewed above. The proposed criteria incorporate the formal structure of DSM-III-R and the concept of preference from DSM-III. The proposed criteria also enlarge the boundaries of diagnosis to include hebephilia, while preserving “classic” pedophilia as a specifiable subtype. As in DSM-IV-TR, repeated sexual acts involving children indicate both that pedophilia is present and that it represents a disorder. Thus, the arrangement of diagnostic elements into Criterion A and Criterion B does not constitute a complete separation of signs and symptoms from distress and impairment.

The addition of the word “Disorder” to the condition is meant as a reminder that people who meet Criterion A but not Criterion B can still be designated as pedophiles, for purposes like research. It is unclear what, if anything, would be lost by excluding such persons from the diagnosis of mental disorder, since, by definition, these hypothetical individuals would not wish to change, would not distress themselves, and would not harm anyone else. The proposed criteria are given in Table 1.
Table 1

Proposed diagnostic criteria for Pedohebephilic Disorder

A. The person is equally or more attracted sexually to children under the age of 15 than to physically mature adults, as indicated by self-report, laboratory testing, or behavior.

B. The person is distressed or impaired by these attractions, or the person has sought sexual stimulation from children under 15 on three or more separate occasions.

C. The person is at least age 16 years and at least 5 years older than the child or children in Criterion A.

Specify if:

 Sexually Attracted to Children Younger than 11 (Pedophilic Type)

 Sexually Attracted to Children Age 11–14 (Hebephilic Type)

 Sexually Attracted to Both (Pedohebephilic Type)

Specify if:

 Sexually Attracted to Males

 Sexually Attracted to Females

 Sexually Attracted to Both

Commentary on the Proposed Criteria

Number of Sexual Acts

The most difficult challenge in improving the objectivity (and potential reliability) of the diagnostic criteria is choosing a minimum value for the number of separate sexual episodes involving children for Criterion B. This is, in practice, a problem for sexual offenders with one or few known child victims, who deny any erotic interest in children, and who have undergone no laboratory testing to assess their erotic age-preference. These are the persons for whom the sexual behaviors clause of Criterion B completely determines the diagnosis.

There can be no perfect cutoff point. A higher threshold value necessarily increases the number of false negative diagnoses; a lower threshold value necessarily increases the number of false positive diagnoses. This trade-off is inherent. There are two further problems complicating the problem: (a) There is no “gold standard” to use in any statistical study of cutoff scores for nonadmitting patients, and (b) even if there were a gold standard, a purely statistical solution to establishing the cutoff score would ignore the relative harm to the patient of a false positive diagnosis and the potential harm to society of a false negative diagnosis.

I have suggested a threshold value of three victims. I believe, on the basis of my laboratory’s experience, that this cutoff would bias the diagnostic criteria toward making false negative diagnoses rather than false positive diagnoses. In other words, it will tend to err on the side of underdiagnosing pedohebephilia.

A clinician assessing an individual patient can always recommend to the courts or to children’s protective agencies that the patient be prohibited from unsupervised access to children, on the grounds that the patient has already demonstrated a propensity to behave inappropriately with children and therefore represents a risk for further offending. In other words, the psychiatric diagnosis of pedohebephilia would not always be needed for the protection of society. On the other hand, a false positive diagnosis of pedohebephilia could do irreparable harm to the patient.

Real Children, Virtual Children, and False Children

A substantial and still increasing number of patients are referred for clinical assessment of erotic age-preference because of Internet-related offenses: Downloading child pornography, and conducting sexual chat with police officers posing as children or arranging rendezvous with police officers posing as children. I recommend that, for diagnostic purposes, photographed children and impersonated children be treated the same as real children. The validity of child pornography use as an indictor of pedohebephilia has already been demonstrated (Seto et al., 2006; see also Blanchard et al., 2007). The erotic orientation of an adult patient who chooses to flirt on the Internet with a real 12-year-old is probably the same as that of a patient who flirts with a police officer impersonating a 12-year-old (although this has not been empirically demonstrated, to my knowledge).

Laboratory Tests for Pedohebephilia

The reference to “laboratory testing” in Criterion A is not meant to refer solely to existing diagnostic tests (e.g., phallometric testing). It is also meant to include any diagnostic tests for pedohebephilia that might be developed in the future.

It is well within the range of possibility that clinical diagnostic fMRI tests for pedohebephilia will be developed within the next several years. These could use experimental designs and stimuli similar to those currently used for phallometric tests. The subject would be shown a standardized set of nude images of male and female children and adults. Instead of evaluating the patient’s penile responses, the clinician would evaluate the patient’s brain responses to male and female children and adults. (The brain regions that activate during sexual arousal have already been established by fMRI studies.) Studies using fMRI technology have already demonstrated that homosexual and heterosexual teleiophiles can be accurately differentiated according to brain activity during exposure to erotic photographs of adult men and women (e.g., Safron et al., 2007). It has not yet been investigated whether fMRI can differentiate accurately between pedohebephiles and teleiophiles. However, a few fMRI studies of pedophiles have already been published (Schiffer et al., 2008a, 2008b; Walter et al., 2007), and it is virtually certain that a diagnostic test for nonadmitting child molesters will be attempted in the near future.


In the DSM definition of mental disorder (e.g., DSM-IV-TR, p. xxxi), “an important loss of freedom” (presumably including imprisonment) is listed along with other sequelae that make a behavioral or psychological syndrome clinically significant: present distress (e.g., a painful symptom), disability (i.e., impairment in one or more important areas of functioning), and an increased risk of suffering death, pain, or disability. Since sexual acts against children are serious criminal offenses, they are closely associated with criminal conviction and incarceration (loss of freedom).


It should be noted that these offenders have little objective motivation to be truthful and many good reasons to dissemble. Offenders are not necessarily rewarded for being truthful about pedophilic impulses; they might experience even more severe consequences of their actions if they acknowledge being pedophiles. Furthermore, some common treatment options are not really attractive, from the patient’s point of view. Many clinicians have turned to “relapse-prevention” treatment of pedophiles, which means, in essence, teaching pedophiles to control themselves. This may well be the best option relative to further offending and incarceration, but a life of sexual denial would hardly be viewed by most people as desirable in an absolute sense. The same considerations apply to treatment with sex-drive-reducing medication.


When the distinction between paraphilias and paraphilic disorders is applied to other anomalous erotic behaviors, it will tend to correlate with a distinction between low severity vs. high severity, or benign vs. malignant. For example, a man or woman with masochistic interests in light spanking or verbal abuse from a safe, consensual partner is less likely to experience distress or impairment than a person with strong masochistic interests that cause serious injury or risk of death. Since real-life examples of mild and harmless masochism, mild and harmless sadism, mild and harmless fetishism, and so on, are relatively common, the paraphilia/paraphilic disorder distinction may seem more intuitive when applied to these other interests than when applied to pedophilia.


In a study conducted after this report was submitted, Blanchard et al. (in press) demonstrated that the notion of preference not only can, but probably must, be applied in the interpretation of phallometric test results.


The term teleiophilia (Blanchard et al., 2000) denotes the erotic preference for persons between the ages of physical maturity and physical decline.


Packard and Levenson interpreted the difference in magnitude between the proportions of negative and positive agreement to suggest that “the evaluators were applying stringent criteria for inclusion in a diagnosis, with a preference given for eliminating false positives in favor of potentially allowing a greater proportion of false negatives” (p. 9).



The author is a member of the DSM-V Workgroup on Sexual and Gender Identity Disorders. He wishes to thank James M. Cantor and Kenneth J. Zucker for their input regarding the distinction between paraphilias and paraphilic disorders. Reprinted with permission from the Diagnostic and Statistical Manual of Mental Disorders V Workgroup Reports (Copyright 2009), American Psychiatric Association.

Copyright information

© American Psychiatric Association 2009

Authors and Affiliations

  1. 1.Kurt Freund Laboratory, Law and Mental Health ProgramCentre for Addiction and Mental HealthTorontoCanada
  2. 2.Department of PsychiatryUniversity of TorontoTorontoCanada

Personalised recommendations