2.1. Differential prevalence rates as potential underdiagnosis in females

Empirical data show that males are more likely to be diagnosed with ADHD than females. For example, the US Centers for Disease Control and Prevention (CDC) report a prevalence of 15 percent in boys versus 8 percent in girls,Footnote ² based on data from the 2022 National Survey of Children’s Health (Danielson et al. Reference Danielson, Claussen and Bitsko2024). This depiction of ADHD affecting twice as many boys as girls is a common one. Among the proposed explanations of this difference are genetic factors linked to sex chromosomes. Empirical evidence in this regard, however, is inconclusive at best (Martin Reference Martin2024). Moreover, reported prevalence rates and gender ratios differ substantially depending on sampling methods and target populations, ranging from 1:1.8 to 1:16 (Stibbe et al. Reference Stibbe, Huang, Paucke, Ulke and Strauss2020). Gender differences are more pronounced in studies on clinical (already diagnosed) versus general populations, and the gap closes if adult populations are considered. Biological explanations of the latter trend also focus on the role of sex hormones. For instance, it is argued that female sex hormones make a late onset of ADHD more likely, or that they make it more likely for ADHD to persist into adulthood. Again, empirical results in this regard are mixed, and much remains to be learned about the effects of sex hormones on ADHD presentation (Nussbaum Reference Nussbaum2012; Stibbe et al. Reference Stibbe, Huang, Paucke, Ulke and Strauss2020; Martin Reference Martin2024; Kooij et al. Reference Kooij, Maxime and Agnew-Blais2025).

While biological factors thus might have a causal role in the differential prevalence rates of ADHD, the link is neither established nor well understood. In light of the variance of reported gender ratios, as well as the steep increase in ADHD diagnosis among females in recent years, many ADHD researchers now point to the role of diagnostic criteria and practices. The differential prevalence rates are now often portrayed as an outcome of gender bias affecting diagnosis and research, which is, moreover, connected to a general tradition of androcentricity and sexism in medicine and psychiatry (e.g., Hinshaw et al. Reference Hinshaw, Nguyen, O’Grady and Rosenthal2022; Martin Reference Martin2024; Morgan Reference Morgan2024; Hartnett et al. Reference Hartnett, MacHale and Duffy2025; Kooji et al. Reference Kooij, Maxime and Agnew-Blais2025).

In general, the notion of gender bias refers to shortcomings in dealing with gender as a variable. Common instances of gender bias include androcentricity/gynocentricity (focusing on one gender where more than one gender is affected), overgeneralization from such data to all genders, gender insensitivity (e.g., failing to analyze gender differences in aggregated data), or double standards in the treatment of identical phenomena based on gender (e.g., evaluating the same behavioral symptoms differently in boys versus girls). Overspecificity often shows up in gendering phenomena that are as such gender-neutral, like germ cells or bacteria (cf. Eichler Reference Eichler1988; Bueter Reference Bueter2024). These traditional lists of kinds of gender bias are, however, not exhaustive, as there are other ways to fail to account for gender diversity. Important to add is a category of sex/gender binarism, referring to problematic assumptions that the only relevant categories are male/female or masculine/feminine (e.g., Krieger Reference Krieger2020). This can (but does not have to) include cisgenderism, that is, an identification of gender with sex assigned at birth, and it may be connected to categories of sexual orientation (heteronormativity) (Ansara and Hegarty Reference Ansara and Hegarty2014).

Regarding ADHD, a central concern is that the diagnostic criteria are centred on male symptoms, thereby leading to underdiagnosis in females, whose symptom patterns are said to diverge from these. The Diagnostic and Statistical Manual of Mental Disorders (DSM) classifies ADHD as a neurodevelopmental disorder, characterized as “persistent pattern of inattention and/or hyperactivity-impulsivity that interferes with functioning or development.” Nine specific symptoms are listed for inattention and hyperactivity/impulsivity each, of which one needs six or more to be present before the age of twelve in at least two different settings (e.g., at home and at school), persisting for at least six months. These symptoms should not be better accounted for by another psychiatric disorder, and a diagnosis requires symptoms to negatively impact academic, social, and/or occupational functioning (APA 2013, 59f.).

The age criterion was raised considerably in contrast to the DSM-IV (seven years) yet is still critiqued as leading to underdiagnosis in females, insofar as these are more likely to have a later onset of ADHD (e.g., Moffitt et al. Reference Moffitt, Houts and Asherson2015; Breda et al. Reference Breda, Rohde and Menezes2021). Moreover, it is often argued that the symptoms associated with hyperactivity/impulsivity (e.g., leaving seat in the classroom, running or climbing in inappropriate situations) are taken to present “typical” ADHD. This ADHD prototype is more common in boys, whereas girls present more frequently with symptoms of inattention (e.g., forgetfulness, difficulties in organizing tasks). While the evidence on gender-specific symptom profiles is somewhat mixed, a recent meta-analysis supports this pattern (Loyer Carbonneau et al. Reference Loyer Carbonneau, Demers, Bigras and Guay2021). Hinshaw et al. (Reference Hinshaw, Nguyen, O’Grady and Rosenthal2022) report similar results, as well as that ADHD in males is often associated with externalizing symptoms (such as being argumentative or breaking things), whereas females with ADHD are more likely to display internalizing symptoms (such as anxiety or low self-esteem).Footnote ³ Martin (Reference Martin2024) concludes that the diagnostic criteria are not well-suited to female populations, as they are focused on the presentation of ADHD in males.

Even if the diagnostic criteria were adequate for female populations, however, another decisive factor has to do with diagnostic practices. ADHD diagnosis in children requires referral by an adult (usually a parent or teacher), as well as an interpretation of behaviors as potential symptoms. This is relevant because externalizing behaviors and hyperactivity/ impulsivity-related symptoms tend to be more disruptive in social and educational settings and are easier to observe. Problems with attention, distraction, and anxiety, by contrast, may be very difficult for the affected individual but less problematic in terms of managing a classroom. Teacher referral is therefore said to be more likely for ADHD cases with a presentation that is predominantly found in boys. Furthermore, interpretation and referral by parents and teachers is affected by lay conceptions of ADHD, which still often cast ADHD as a disorder of boyhood. This further exacerbates referral bias. For instance, teachers and parents who were presented with vignettes of symptomatic childhood behavior rated their likelihood to recommend or seek psychiatric evaluation lower when the (identical) cases were presented as portraying girls, not boys (Young et al. Reference Young, Adamo and Björk Ásgeirsdóttir2020). Many women, moreover, report struggles and experiences of dismissal within the health care system (e.g., Morgan Reference Morgan2024; Holden and Kobayashi-Wood Reference Holden and Kobayashi-Wood2025).

In addition, girls and women diagnosed with ADHD do not get the same treatment as diagnosed males. The standard treatment for ADHD is pharmacotherapy, predominantly consisting in the use of pharmaceutical stimulants such as methylphenidate (Ritalin). A systematic literature review by Kok et al. (Reference Kok, Groen, Fuermaier and Tucha2020) concludes that females with ADHD are less often prescribed pharmaceutical treatments than their male counterparts. This difference was most pronounced for younger patients, with adult women closing the gap toward males in terms of using pharmaceutical treatments. In fact, a more recent report shows that the number of stimulant prescriptions increased by 92 percent for female patients versus 36 percent in male populations from 2012 to 2022 in the United States (IQVIA 2024, 12). The medication of female ADHD, to put it somewhat cynically, is a growth market.

While this increase in prescription rates might be interpreted as overcoming double standards in the treatment of ADHD, it is problematic in terms of insufficient knowledge about gender differences regarding the efficacy and safety of the respective drugs. As Kok et al. (Reference Kok, Groen, Fuermaier and Tucha2020) point out, there is evidence pointing toward differences regarding how effective standard ADHD medications are, but too little research to draw general conclusions. A recent review on research needs regarding female ADHD similarly points out the existence of blind spots in the evidence base (Kooji et al. Reference Kooij, Maxime and Agnew-Blais2025). It underlines the need for a better understanding of the interactions of specific female phenomena (e.g., hormonal fluctuations during the menstrual cycle, pregnancy, or menopausal transition) with both ADHD presentation and treatments. As they conclude, “ADHD in girls and women has long been under-recognised, under-researched, and under-treated, and research was limited to cisgender women” (ibid., 7).

A great deal more research (funding) is needed in this area. This also concerns aspects of treatment beyond pharmaceutical drugs; for instance, the review by Kooji et al. (Reference Kooij, Maxime and Agnew-Blais2025) points out that women with ADHD are more likely than their neurotypical counterparts to suffer from sexual abuse, and to identify as transgender and/or nonheterosexual. As Hartnett et al. (Reference Hartnett, MacHale and Duffy2025) state, there is a great need for gender sensitive mental health care that adequately responds to the effects of gender norms on mental illness and is aware of gendered experiences of violence. They call for trauma-sensitive approaches that respond to overlapping and interacting variables of social oppression, such as poverty, ethnicity, or gender.Footnote ⁴ While the authors here draw attention to the role of gendered oppression in mental health (care), they also treat gender as well as sex as binary. As Kooji et al. (Reference Kooij, Maxime and Agnew-Blais2025) note, an exclusive focus on cisgendered women is common in the literature. This omission is problematic, especially because populations with a diagnosis of ADHD or autism spectrum disorder (ASD) display significantly higher rates of gender variance than undiagnosed populations (7.59 times higher for ASD and 6.64 times higher for ADHD [Strang et al. Reference Strang, Kenworthy and Dominska2014]). A more recent systematic review of the (still sparse) literature on this connection by Goetz and Adams (Reference Goetz and Adams2024) supports the finding of higher rates of transgender and/or gender diverse individuals among people diagnosed with ADHD.Footnote ⁵

In summary, there are clear indicators for gender bias such as androcentricity in the research on ADHD and overgeneralization of data from male study populations, which leads to gendered ignorance. Androcentricity also affects the diagnosis of ADHD by rendering the prototypical case that of disruptive boyhood, which leads to referral bias and potentially biased interpretation of symptoms in clinical contexts. Studies, moreover, show double standards in prescription rates given similar symptoms and levels of impairment.Footnote ⁶ Finally, while this article focuses on ADHD in male versus female populations in line with the current debate, it should be noted that much of the relevant literature displays gender binarism, even though the data show an interesting connection between prevalence rates and gender diversity in a broader sense.

It can be concluded that females are underdiagnosed relative to males in a systematic and unfounded way. This gender bias, as has been said above, is often said to harm girls and women. A missed or delayed diagnosis hinders access to treatment as well self-understanding and social acceptance, thereby negatively affecting self-esteem, well-being, and functionality in social and occupational contexts. Many women describe receiving a diagnosis in very positive terms, such as being revelatory, providing relief from guilt, and increasing self-compassion as well as social support (e.g., Babinski and Libsack Reference Babinski and Libsack2025; Holden and Koboyashi-Wood Reference Holden and Kobayashi-Wood2025). Yet, as I will argue in the next section, aiming at diagnostic parity by increasing diagnoses in females sets the male prevalence rates as a standard to reach—a standard that is deeply problematic, as it likely involves substantial overdiagnosis.

2.2. Differential prevalence rates as potential overdiagnosis in males

Calls for more research into the impacts of sex and gender on symptom presentation, pathological pathways, diagnostic practices, and treatments certainly deserve support. However, the framing of gender bias in female ADHD in terms of underdiagnosis suggests that overcoming such bias consists in diagnosing more females. In this spirit, it has been proposed to create sex-specific thresholds for diagnosis. Lowering the number of symptoms required for an ADHD diagnosis in females would increase the likelihood of females being diagnosed, including those whose symptoms are currently “subthreshold” (Martin Reference Martin2024, 306). This, however, raises the danger of overmedicalization, that is, of treating nonpathological problems and struggles in life as medical ones, as well as of overdiagnosis, that is, the diagnosis of ever milder cases with low benefits of such diagnosis (cf. Hofmann Reference Hofmann2022).

ADHD is notoriously controversial in this regard, with many scholars arguing that ADHD is mis- or overdiagnosed in male populations (e.g., Conrad Reference Conrad2006). One reason for worry is the steep increase in cases. A recent survey from the United States reports that 11.4 percent of children aged three to seventeen have ever been diagnosed with ADHD, with boys of seventeen years having the highest rate of 23 percent (Danielson et al. Reference Danielson, Claussen and Bitsko2024). Xu et al. (Reference Xu, Strathearn, Liu, Yang and Bao2018) found that overall prevalence in the United States rose from 6.1 percent in 1997/1998 to 10.2 percent in 2015/2016. In addition, an increasingly higher proportion of diagnosed people receive medication (IQVIA 2024).

It might be the case that this increase in diagnoses picks up on an actual increase in cases or corrects for former underdiagnosis. Important to note in this regard is that prevalence rates correlate with a host of sociodemographic factors beyond gender, such as socioeconomic status, ethnicity, or geographical location (Xu et al. Reference Xu, Strathearn, Liu, Yang and Bao2018; Danielson et al. Reference Danielson, Claussen and Bitsko2024). For instance, risk factors for ADHD diagnoses include living in an area with a high density of psychiatrists, having insurance coverage for mental health services, or being among the younger children in one’s class (Layton et al. Reference Layton, Barnett, Hicks and Jena2018). A review by Kazda et al. (Reference Kazda, Bell, Thomas, McGeechan, Sims and Barratt2021) supports the hypothesis of overdiagnosis of ADHD based on findings that the increase in diagnoses means an inclusion of milder cases, as well as of diminishing benefits of medication in these milder cases.

Psychiatry as a field is prone to diagnostic inflation, many have argued (e.g., Frances Reference Frances2013; Haslam Reference Haslam2016). One reason for this is the syndromal approach to classification and diagnosis of mental disorders. In absence of etiological explanations and unequivocal pathophysiological markers, mental disorders are classified using their symptoms. As the various revisions of the DSM show, such lists of symptoms are malleable and lack definitive cutoff points, and the DSM continues to be highly controversial (e.g., Tabb Reference Tabb2015). Arguably, the diagnostic criteria for ADHD have become less stringent throughout their various iterations (Fabiano and Haslam Reference Fabiano and Haslam2020). Moreover, symptoms often require interpretation of behavior (e.g., what constitutes “excessive” talking or running around), which at least in the case of ADHD in children is mediated by teacher and parent expectations.

Other reasons for diagnostic inflation include sociocultural factors. For example, the emphasis on children’s academic success may increase, and psychiatric diagnoses have lost some of their stigma. Being diagnosed with ADHD may no longer be a source of shame but rather a reprieve from such shame. Many scholars have observed an increasing “psychiatrization of society,” not just in terms of rising diagnoses and utilization of healthcare services (Beeker et al. Reference Beeker, Mills and Bhugra2021) but also regarding an increasing use of psychiatric and psychotherapeutic vocabulary in everyday contexts (Almagro and Isern-Mas Reference Almagro and Isern-Mas2025). Such psychiatrization is not simply pushed onto the public by experts or big pharma. Rather, it is often fought for by patient advocacy groups aiming to raise awareness and resources (see the example of “ADDitude” mentioned above). The idea of adult ADHD was popularized by several very successful self-help books (cf., Conrad Reference Conrad2005, Reference Conrad2006), and social media accelerate these phenomena (Chevalier Reference Chevalier2024).

ADHD may be a perfect example of what Ian Hacking (Reference Hacking, Sperber, Premack and Premack1996) calls “looping effects” in the classification of human kinds. Humans are not like, for example, chemical elements, in so far as they react to being classified or labeled a certain way. In particular, they react by assigning those labels a certain meaning. Psychiatric diagnoses, according to Hacking, can create distinct ways of being that become central for identity and community. This, in turn, over the long run may change the presentation of the original classificatory object, such as the phenomenon of ADHD (Chevalier Reference Chevalier2024). Instructive in this regard are interviews by Jønsson (Reference Jønsson2025) with self-diagnosed individuals. Even if they are denied the official diagnosis, some insist on “having it” but experiencing “untraditional” symptoms. Many laypeople push for diagnosis and for more lenient (or inclusive) criteria.

A caveat is in place here: Suffering is always subjective, and the suffering behind seeking out a diagnosis for oneself or one’s children is, by default, real. The question at hand is about the best way to respond to this suffering. Diagnosis comes with benefits, as mentioned previously—but it also involves risks and harms. One might object here that at least for adult women seeking out the diagnosis, this should be their autonomous decision. Yet to be autonomous, such decisions must be well-informed regarding potential risks and benefits, so it is worth exploring the downsides.

To start with, diagnosis and treatment require resources. People seeking out a diagnosis must undergo an evaluative process that can be uncomfortable and difficult to access. Private clinics can be pricey, and in a system with public insurance waiting times are often very long. Moreover, diagnosis is the stepping stone to treatment. Again, this may require one to find the right therapist and the means to afford them. In the case of ADHD, diagnosis will often result in the prescription of medication such as methylphenidate. Like all medications, this comes with the risk of side effects. For example, these include insomnia, loss of appetite, facial tics, cardiovascular issues, or slowing growth in children. Stimulants can be habit forming and discontinuation may lead to withdrawal symptoms. In general, results on these medications’ efficacy are less convincing than their popularity may suggest.

Regarding the use methylphenidate in (mostly male) children and adolescents, a recent Cochrane review (Storebø et al. Reference Storebø, Storm and Ribeiro2023) concludes that the medication has moderate effects on teacher-rated ADHD symptoms and may improve general behavior, but it does not seem to improve quality of life. The certainty of the evidence is described as very low, meaning that the authors “have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect” (ibid.). The systematic review included 212 studies comparing methylphenidate with a placebo or no intervention group. Forty-one percent of these studies were partly or fully funded by the pharmaceutical industry. Of the 212 studies, 191 were deemed to have a high risk of bias. In terms of improving academic outcomes of children, a meta-analysis from 2013 concluded that methylphenidate improves “seatwork productivity,” but results concerning academic performance were mixed. In other words, stimulant medication helps children to sit still and complete the tasks expected of them, but it does not necessarily help with accuracy and the development of academic skills (Prasad et al. Reference Prasad, Brogan, Mulvaney, Grainge, Stanton and Sayal2013).

The situation is similar regarding the efficacy of methylphenidate in treating adults with ADHD. A Cochrane review on extended-release methylphenidate included twenty-four studies (90 percent partly or fully industry funded), of which twenty were classified as displaying a high risk of bias. Effect sizes on ADHD symptoms and quality of life were small, and the certainty of evidence was rated as very low. Most study participants were male, and there is no gender-specific analysis of the results (Boesen et al. Reference Boesen, Sand Paludan-Müller, Gøtzsche and Juhl Jørgensen2022).

(Over-)diagnosis may therefore lead to overtreatment with drugs that have questionable harm-benefit profiles. This comes with financial burdens for individuals and public health care systems. A common worry is that the steep increase in psychiatric diagnoses overloads the capacities of current healthcare systems and threatens access for the most severe cases. In addition, (over)diagnosis can come with social costs. Stigma of psychiatric diagnoses is generally well-established and relates to different levels from public stigma (e.g., common prejudices), self-stigma (internalizing such prejudices), and courtesy stigma (extending to family members) (e.g., Rössler Reference Rössler2016). Public stigma can lead to experiences of discrimination, for example, within the healthcare system (Thornicroft et al. Reference Thornicroft, Rose and Kassam2007) or regarding employment (Østerud Reference Østerud2023). Regarding ADHD particularly, Mueller et al. (Reference Mueller, Fuermaier, Koerts and Tucha2012) report evidence for stigma across these different levels. For instance, a diagnosis might create negative effects because of lowering teacher expectations. As this review notes, however, presence and degrees of such stigma vary with variables such as age, gender, or ethnicity (cf. also Visser et al. Reference Visser, Peters and Luman2024). A major source of ADHD stigma, importantly, are public perceptions of ADHD as “not real,” which can be promoted by trends toward more diagnoses—as well as debates on overdiagnosis.

In relation to self-stigma, being diagnosed with a mental order can also come with an emotional toll. Many authors have pointed out how a psychiatric diagnosis (including ADHD) affects people’s views of themselves in ways that are not necessarily only positive (e.g., Tekin Reference Tekin2011; Mueller et al. Reference Mueller, Fuermaier, Koerts and Tucha2012; Visser et al. Reference Visser, Peters and Luman2024; Veldmejer et al. Reference Veldmeijer, Terlouw, Boonstra and van Os2025). A psychiatric diagnosis affects people’s self-narratives and how they make sense of their experiences. While a diagnosis can enable one to develop a self-narrative that fits with dominant frameworks of psychological distress or behavioral difficulties—something that is often perceived as a relief—it can also be an obstacle to alternative narratives beyond biomedical conceptions of mental disorder (Hassall Reference Hassall2024; Nielsen Reference Nielsen2025).

Foulkes and Andrews (Reference Foulkes and Andrews2023) suggest that this leads to feedback effects, in which a narrative of one-self as mentally ill turns into a self-fulfilling prophecy. For instance, this may happen if I start to avoid challenging situations because I think of myself as incapable or as pathologically anxious, thereby depriving myself of learning opportunities and ultimately fueling my anxiety. Given the lens of a psychiatric diagnosis, my interpretation of unpleasant experiences in terms of symptoms may also reenforce these, thereby making me suffer more. In consequence, increased mental health awareness may lead to a prevalence inflation of mental illness—not just because people start to adopt psychiatric terminology but also because this shift makes them feel worse.

Considering all this, it is ill-advised to take the rates of ADHD diagnosis in male populations at face value. There are good reasons to be wary of systematic overdiagnosis and overtreatment, such as diagnostic inflation and the financial interests at play. Diagnosis can come with harms and benefits. Whether the harms outweigh the benefits (constituting overdiagnosis) is a complex question because it involves many variables and may depend on the individual in question as well as their social context. While overdiagnosis in male populations thus cannot be proved here, it does seem an important possibility to consider, especially as diagnosis and treatment of male populations can also be said to be affected by bias. It might seem intuitive to view this as a situation in which bias leads to underdiagnosis in females and overdiagnosis in males, with the truth lying somewhere in the middle. Yet this does not follow: Underdiagnosis in females relative to males could still represent overdiagnosis in absolute terms. Gender bias would then lead to fewer diagnoses but also (accidentally) bring us closer to the truth. A definition of bias through a link to truth is therefore not helpful in this context, as I will further discuss in the next section.

3.1. The notion of bias

“Bias” is ambiguous in that it is used to refer to a variety of phenomena.Footnote ⁷ In the preceding discussion, the notion of bias plays a prominent role at several points. For example, it shows up in the description of relative underdiagnosis of female ADHD as resulting from gender bias, which can affect study designs, data analysis, research agendas, or interpretations of behavior through, for example, stereotypes or androcentricity. It is also prominent in the assessment of Ritalin’s efficacy. As pointed out, many of the respective studies have been described as exhibiting a high risk of bias. This is based on standardized assessment tools by the Cochrane foundation such as the “RoB 2” (Sterne et al. Reference Sterne, Savović and Page2019), which check for problems in areas such as placebo-control, outcome measures, or reporting of results. In a nutshell, these tools presume certain methodological standards and identify deviations from these, which are thought to lead to an erroneous estimation of a treatment’s effects. For example, selective reporting of favorable results or the lack of an adequate control group can put a drug into overly positive light.

This way of conceptualizing bias is firmly rooted in epidemiological methodology and the evidence-based medicine movement, which aims to create and identify the best possible evidence through standardized trial designs and quality criteria. Diverging from the respective methodological standards is described as bias, insofar as this deviation introduces systematic errors, understood as systematic deviation from the truth. For example, the Dictionary of Epidemiology defines “bias” as “systematic deviation of results or inferences from the truth” as well as “[p]rocesses leading to such deviation” (Porta Reference Porta2014, 21). Such deviations can result from different mechanisms (such as prejudice, cognitive heuristics, value ladenness, or methodological flaws), yet their alleged unifying feature is the systematic divergence from an impartial outcome (such as the truth).

In Bueter (Reference Bueter2022), I describe this as a narrow or “ontological” notion of bias and argue that while this is a straightforward conceptualization of “bias” in contexts such as statistical testing of drug efficacy (where there is something like the true effect size that we want our results to match), it may not be suitable in other contexts such as, for example, qualitative research. Yet, understanding bias as deviation from truth has become very dominant especially in biomedical contexts (and beyond).

In response, one might prescribe a more restrictive use of the term “bias”; that is, to use it in terms of systematic deviation from truth but only in contexts in which this is appropriate. Because the notion of bias is, however, useful (and commonly used) to flag epistemic shortcomings in general, in Bueter (Reference Bueter2022) I propose a wider notion of bias termed “epistemic,” which focuses on processes rather than outcomes. Epistemic bias refers to instances in which we have good reasons to believe that this research could have been (done) systematically better (ibid., 311.) For example, this may be the case if research diverges from standard methodology or if the deliberation of such standards is dominated by financial interests or an unquestioned value consensus (drawing on Longino Reference Longino1990). It does include cases of, for example, insufficient blinding in drug testing, but it breaks the general link between bias and truth.

In a nutshell, one might characterize bias via the systematic deviation from a given standard on both the ontological and the epistemic account. The epistemic account, however, does not require this standard to be neutral or impartial. Accordingly, it also allows for the relevant standards to evolve over time and includes consideration of how standards are established. Systematic deviations from either current standards or meta-level norms for the deliberation of such standards, as argued in Bueter (Reference Bueter2022), constitute bias because they give us good reasons to think that the respective research could have been done better. For instance, methodological and institutional standards for drug testing as well as publishing results have been extensively debated and made increasingly explicit in recent decades. Systematic deviation from current guidelines can accordingly be considered as bias. Yet standards as such may also be problematic, and they need to be open to change alongside scientific advances. For this reason, Bueter (Reference Bueter2022) requires the setting of standards in a process that lives up to Longino’s (Reference Longino1990) criteria for social objectivity.

Disadvantages of the epistemic notion are that it makes bias relative to current standards and thereby context sensitive, as well as that it leaves more room for interpretation of whether some research practices or results are biased. An advantage is that the epistemic notion can include clear cases of statistical bias, but also characterizes processes and outcomes as biased that are not directly truth-linked, such as ignoring important aspects in agenda setting or disregarding critical inputs when it comes to the development of methodological standards. The ontological and epistemological notion of bias can also come apart; for example, a failure to disclose financial interests in research results would qualify as bias on the epistemological approach (as it violates current standards) but not the ontological one (as it may or may not lead to a deviation from truth) (Bueter Reference Bueter2022, 312).

In the remainder of this section, I will argue that an epistemic approach is better suited to conceptualize gender bias in ADHD diagnoses than an ontological one. The problem with the latter is, firstly, that we lack knowledge on what the correct prevalence rates are and whether we are systematically over- or undershooting or even accidentally matching them. Secondly, a deeper problem is that there may not be a correct rate in the sense of “the true” prevalence of ADHD.

3.2. Against the presumption of a true prevalence rate

The idea of diagnostic bias as systematic deviation from the truth presumes that there is a stable, discrete disease entity occurring at specific rates to be matched, which encounters ontological, conceptual, and epistemic challenges. In this section, I will present four reasons to be skeptical of this assumption: the reification (1), performativity (2), and value-ladenness (3) of psychiatric diagnoses as well as the underlying understanding of diagnosis (4).

(1) The first problem concerns the assumption of ADHD as a discrete disease entity. Above, the precarious state of psychiatric classification has already been mentioned, that is, the fact that the definition and diagnosis of mental disorders is not based on underlying causal factors and processes but on surface level symptoms. More specifically, it uses polythetic lists of potential symptoms, a certain number of which together are taken as sufficient, but usually no individual item is necessary to make a diagnosis. These symptoms require interpretation, and the concrete threshold (how many symptoms are needed) is based on expert consensus and irreducible to empirical data. Critics point out that this results in patients with very different symptom profiles sharing a diagnosis, that symptoms often are not unique for individual diagnoses, and that overlaps between diagnoses are very common. Therefore, current classifications of mental disorders can be argued to fail in picking out distinct phenomena (cf. Tabb Reference Tabb2015).

At the same time, these diagnostic criteria shape the study populations in psychiatric research, furthering their reification but not necessarily validity (Hyman Reference Hyman2010). Reification describes a fallacious inference from the existence of a concept to the existence of a thing to which the concept refers. Te Meerman et al. (Reference Te Meerman, Freedman and Batstra2022) identify several ways in which ADHD is reified, such as linguistic choices that present ADHD as disease entity or problematic background assumptions of genetic reductionism. While the diagnostic criteria remain at the level of describing hyperactive or inattentive behavior, they are treated in medical research and practice as referring to an underlying dysfunction, thereby perpetuating the view of ADHD as disease entity. Playing into this reification, moreover, is a general categorical approach to the classification of mental disorders as people either having “it” or not, versus thinking about phenomena such as ADHD in terms of behavioral patterns present on a continuous spectrum (cf. Kotov et al. Reference Kotov, Krueger and Watson2017). As Susan Hawthorne (Reference Hawthorne2010) has argued, a premature reification of the construct “ADHD” should be avoided given these problems yet is furthered by interacting social and scientific factors. For instance, this can be seen in an institutionalization of a dichotomized view of either having ADHD or not—and, consequently, access to accommodations, treatment, and so forth.

To be clear, these issues do not imply that ADHD is “not real.” The question of whether the construct of ADHD picks out a discrete disease entity is related to the general debate on the ontology and epistemology of psychiatric classification. Worth pointing out in this regard is Zachar’s model of psychiatric kinds as practical kinds (Zachar Reference Zachar2000, Reference Zachar2002, Reference Zachar2014). Rooted in a pragmatist approach to classification, Zachar characterizes psychiatric kinds as neither natural kinds nor purely arbitrary or socially constructed.Footnote ⁸ Practical kinds aim to represent stable patterns in ways that are useful with regards to both research and clinical practice, such as, for example, enabling reliable diagnosis, prognosis, and treatment selection (Kendler et al. Reference Kendler, Zachar and Craver2011). This utility here is not just an effect of inherent features of a kind, Zachar argues, but also of external and relational factors (such as, e.g., the availability of treatments or the effects of labeling) (Zachar Reference Zachar2002, 221). Practical kinds, moreover, can have fuzzy boundaries and are well-suited to a dimensional versus categorical understanding (Zachar Reference Zachar2000).

(2) Much more could be said in this regard but let this suffice here to establish reasonable doubt about whether the diagnostic concept of ADHD picks out a clearly demarcated phenomenon. In addition, it is also questionable whether the disease entity in question is stable or, in Ian Hacking’s (Reference Hacking2007) terminology, presents a moving target. The problem here is not that the targeted phenomenon (ADHD) is on the rise (e.g., due to environmental factors) but that the very concept of ADHD, as a classification of kinds of people, plays a performative role in the sense of introducing feedback loops. As mentioned in section 2.2, the existence of this very concept may lead people to interpret their experiences in certain ways. As Foulkes and Andrews (Reference Foulkes and Andrews2023) argue, the normalization of psychiatric diagnoses may lead to a rise not just in diagnostic rates but also in the very phenomena to which they refer. In that case, then, diagnosing many people with ADHD makes it true that many people have ADHD. With self-fulfilling claims such as these, it is problematic to define bias as a deviation from the truth because this truth is not independent from claims being made about it.

(3) Another problem stems from the contested nature of the notion of mental disorder. In philosophy of medicine, a long-standing debate concerns the question of whether “disease” and “disorder” can be defined using descriptive criteria (naturalism) or necessarily involve value-laden assessments of desirability (normativism). Within (philosophy of) psychiatry, the most prominent account is a hybrid position, which aims to integrate descriptive and normative criteria: Wakefield’s (Reference Wakefield1992) account of mental disorder as harmful dysfunction.

According to this, both a dysfunction and harmful consequences are necessary conditions for mental disorders. As Wakefield argues, dysfunction alone (the descriptive criterion) is insufficient for disorder status because the demands of current environments can differ significantly from those constituting the original selective pressure during the evolution of relevant functions. What may have been beneficial for our ancestors (e.g., high levels of anxiety or aggression) might be problematic in modern societies (or vice versa). Harm (the normative criterion) is insufficient on its own because we need to prevent the arbitrary declaration of socially unwanted behaviors as disorders (see Cooper Reference Cooper2002 for a discussion).

If we grant this, it directly undermines the idea of mental disorders as stable, discrete disease entities. Even if there were identifiable dysfunctions underlying phenomena like ADHD (e.g., a dysfunction in the regulation of dopamine), whether this makes for a disorder depends on the resulting harm. Such harm is context dependent, relating to the social expectations of a person’s environment and their relative ability to meet these expectations. ADHD-like traits might be appreciated or celebrated in one context but considered deeply problematic in a society that values certain types of academic achievement very highly (cf. Hawthorne Reference Hawthorne2010) or is deeply structured by certain types of chrononormativity, that is, temporal norms about punctuality, patience, or life events (cf. Nielsen Reference Nielsen2017).

This can lead to a situation in which cases of ADHD rise not because more children (or adult women) develop ADHD-like traits but because the normative expectations become more demanding, with the range of behavior considered normal shrinking correspondingly. As Hawthorne (Reference Hawthorne2010) argues, increased ADHD diagnosis can lead to “institutionalized intolerance”: Once identified as a medical problem, the expectation will be to treat ADHD-like behavior—offering accommodations but at the same time thereby placing such behavior outside of the realm of “normal” or “acceptable.”

We could think of this in terms of a true prevalence rate relative to a particular time and set of social expectations and values. However, this is not how the issue is presented in most of the literature; that is, rising ADHD rates are usually discussed as absolute phenomenon, not as more people struggling relative to changing societal expectations. Even more problematic, again, is the fact that there may not even be one underlying dysfunction in any categorical sense. It will then not only be the case that we diagnose milder and milder versions of a certain dysfunction because its harmful consequences are exacerbated by the social environment. Rather, what is considered as dysfunctional itself depends on setting thresholds for continuously distributed traits. One could, of course, further conditionalize claims about ADHD as dependent on certain social expections and certain thresholds; yet this would still leave the problem of the instability of such claims.

(4) Finally, thinking of bias in terms of diagnostic rates deviating from the truth implicitly presumes a view of diagnosis as aimed at representation, that is, at establishing whether individuals have ADHD or not and, based on this, getting the actual prevalence rate right. An alternative view of diagnosis would emphasize its role in clinical and public health contexts, and thereby its relation to decision making. This is exemplified in the discussions about overdiagnosis mentioned above. Nielsen (Reference Nielsen2025, 6) characterizes overdiagnosis in psychiatry in terms of “excessive diagnoses that turn out to be unbeneficial to those who receive them.” Overdiagnosis is not misdiagnosis, that is, it does not refer to cases in which a diagnosis is made despite diagnostic criteria not being satisfied, but rather to “instances of distress that live up to the formal criteria for diagnosis, but will not benefit from treatment.” The problem with overdiagnosis, then, is not a misrepresentation of reality but that its harms outweigh its benefits.

This harm-benefit relation depends on a variety of contextual factors, such as the availability of safe and effective treatments, the degree of stigmatization, and the long-term effects of self-narratives. Beyond the individual level, harms and benefits of diagnoses relate to the distribution and depletion of public resources. Harm-benefit profiles will, moreover, be affected by how harms and benefits are conceptualized and measured, which can be value-laden in itself. For example, is the benefit of pharmaceutical treatment evaluated through teacher-rated improvements in behavior, or rather through children’s self-reported well-being?

In light of these questions, it seems more adequate to think of diagnosis as a tool rather than a measure of reality. Its goal, or function, is not so much to enable establishing the correct rate of disease entities as it is to ameliorate suffering. The net effect of diagnosis for an individual, that is, should be beneficial. This is especially so in cases of psychiatric diagnoses, where the inference from construct to entity can be criticized as problematic reification based on the arguments given previously. Understanding diagnosis as a tool in particular squares well with an understanding of psychiatric kinds as practical rather than natural (see above). Yet it does not require these arguments to be successful: Even if we are diagnosing stable and delineated disease entities, the function of diagnosis clearly involves ethical and political dimensions. Ultimately, diagnosis should help people—even if diagnosis is, of course, also an epistemic tool in the sense of aiming to understand and explain certain experiences.

3.3. Bias as systematic deviation from current standards and norms of critical discourse

These four problems show why a definition of bias in psychiatric diagnosis through a relation to truth is not helpful and calls for an alternative. In section 3.1, the epistemic notion of bias has been introduced, which understands bias in terms of deviation not from impartial outcomes but from standard procedures, as well as of problems in the procedures for setting such standards (Bueter Reference Bueter2022). Based on this, bias is constituted by deviations beyond chance from at least one of the following:

1. (a) Current methods and standards

   We can identify bias where there is a systematic deviation from current best practice in research. This includes methodological guidelines and institutional standards that have been agreed upon in the scientific community. For example, indicators of known forms of gender bias such as androcentricism and andronormativity in the research processes constitute such deviations from contemporary best practice because norms for gender-sensitive medical research were discussed already since the 1970s, which resulted in the NIH 1992 Revitalization Act supporting and formalizing these demands (cf. Bueter Reference Bueter2017). In fact, discussion of underdiagnosis of females with ADHD, coupled with calls for more research, go back to the 1980s/1990s (e.g., Berry et al. Reference Berry, Shaywitz and Shaywitz1985; Gaub and Carlson Reference Gaub and Carlson1997). Androcentricity is also present in the research on ADHD medication (see above). In addition, this research has been evaluated as substandard in relation to methodological standards for drug testing by the Cochrane foundation.

2. (b) Deliberative procedures for setting standards

   The second level concerns the process through which such standards are established. As mentioned previously, this draws on Longino’s (Reference Longino1990) conditions for social objectivity, which are aimed at a transformative criticial discourse rather than impartiality. These conditions include the presence of venues for criticism, uptake of criticism, shared standards for criticism (themselves the outcome of critical deliberation), and tempered equality of intellectual autority in a diverse scientific community. For instance, it would then constitute bias if certain critical voices have been excluded or unfairly dismissed in the deliberation of current standards.

   In the case of ADHD and other psychiatric diagnoses, the process of DSM-revisions is of particular interest in this regard. How are the concrete procedures for changing diagnostic criteria structured? Who is included at what point, and are critiques of gender bias accounted for? For instance, in recent years there have been critiques with regard to female underrepresentation and industry ties among members of the expert panels (e.g., Cosgrove and Krimsky Reference Cosgrove and Krimsky2012), as well as increasing calls for a more participatory approach to represent patient voices (e.g., Knox Reference Knox2022). To identify bias, we would thus inquire into what critical perspectives exist and to what extent they have received uptake.

   This level is important because methodological and institutional standards are always in flux (and necessarily so). For instance, research that involves male and female subjects and performs a gender-sensitive data analysis would accord with current standards. Yet, as mentioned in section 2.1, such a gender binarism can also be criticized as a form of gender bias. It is thus important to look beyond current standards to the processes of their deliberation and implementation. For instance, there have been critical voices identifying and criticizing gender binarism and cisgenderism in the relevant literature (e.g., Ansara and Hegarty Reference Ansara and Hegarty2014; Krieger Reference Krieger2020). If these are not included, or their arguments not given a fair uptake, this would likewise constitute bias.

Such an understanding of bias can illuminate the case of relative underdiagnosis of ADHD in women as epistemically problematic without construing it as undershooting the true prevalence rate. Underdiagnosis here constitutes bias not because it diverges from the truth but because it results from research practices that fail to implement current standards for gender-sensitive science (e.g., using androcentric study populations and diagnostic criteria). Likewise, research on pharmaceutical treatments of ADHD in male populations displays bias in failing to implement current standards for trial design and reporting. On the meta-level, current standards may be characterized as biased, if they do not result from a rigorous and inclusive critical process (e.g., if DSM-revisions fail to include or attend to critical voices regarding the role of gender in diagnostic criteria or are overly pervaded by financial conflicts of interest).

The epistemic approach thereby provides a way to conceptualize bias independent of truth as well as a two-level approach to the identification of bias. Avoiding gender bias, accordingly, would require researchers to live up to current standards, and to have an inclusive debate about such standards.