Autism ableism seen through research abstract contents (Autism AbSTRACt): a mixed-methods analysis of language in NIH-funded genetic and genomic autism research

Abstract

In recent years, genetic and genomic autism research has come under increasing scrutiny, moving to the center of debates about ableism, neurodiversity, autism acceptance, and the future of research and care. At the same time, both autism research and genetics and genomics research have, as fields, begun to reckon with the significance of the language researchers use in the course of their work and the harmful ideas that may thereby be reinforced. Although the language of research cannot be assumed to straightforwardly correspond to individual researchers’ beliefs, the presence of widespread ableist language may indicate structural and institutionalized ableism, including ableist assumptions at the foundations of research. We conducted a mixed-methods analysis of 166 genetic and genomic autism research projects funded by the US National Institutes of Health (NIH), in order to understand the prevalence of potentially ableist discourse, language, and stigmatizing language about autistic people. We found that such discourse and language was ubiquitous across our sample, including a discourse of prevention. This study lends empirical evidence to current debates about language in autism research. Evaluating language can prompt researchers and institutions to reflect on how they conceptualize, design, discuss, and pursue their work.

Lay Abstract

Genetic research about autism is controversial. Researchers are starting to think more carefully about the words they use to talk about autism and the way they do their research. Past research has found that researchers sometimes write about autism in ableist ways. This means that they write about autistic people as though they are less important than non-autistic people. We looked at the way genetics researchers have written about autism in the paperwork for their research. We found that they often write about autistic people in an ableist way. We think that researchers should think carefully about the way they write about autistic people, and how they plan and do their research.

Introduction

In recent years, genetic and genomic autism research has come under increasing scrutiny, moving to the center of debates about neurodiversity, autism acceptance, and the future of research and care. Psychiatric and neurodevelopmental genetic research is especially ethically fraught because of its historical ties to eugenics, prompting current concerns over its use to guide reproductive decision-making. Several large-scale and high-profile autism genetics projects have attracted serious criticism from autistic communities. The pausing of recruitment for, and eventual cancellation of, the Spectrum 10K study in the UK after vociferous opposition encapsulated the disjuncture between some researchers and parts of the community: criticisms included concerns about consent to participate (especially where consent is sought from caregivers rather than autistic individuals); data privacy and the potential misuse of data; the possibility that the development of polygenic risk scores could facilitate pre-natal screening, embryo selection, and/or selective abortion; and the past work and words of prominent researchers associated with the study (Pring, 2021) (Spectrum 10k, 2025). In the US, the Autistic Self-Advocacy Network (ASAN) released a statement on Genetic Research and Autism in 2022, highlighting similar issues as well as the possibility of gene editing, issues around genetic technologies, and questions about who has control of and power over research. Although it recognizes the potential for positive outcomes (including greater understanding and treatment of co-occurring conditions), the ASAN does not support research in autism genetics and recommends a series of safeguards if such research is to go forward (ASAN, 2022).

Though there are a multitude of views across autistic and autism communities – including, for some, support for exploring the biology of autism in order to find a “cure” – studies in the US and UK have consistently found that overall autistic people strongly favor research designed to help autistic people flourish, including research into co-occurring conditions, lifespan issues, and effective service provision (Frazier, 2018) (Roche, 2021) (Pellicano, 2014). The contemporary research landscape, however, does not align with these priorities. Recent reviews of research funding in the US have shown that less than 9% of NIH funding allocated for autism research is applied to services (Cervantes, 2021). In contrast, in 2017–2019 32.59% of federal funding was awarded to research on the biology of autism and 22.87% on treatments for autism (Harris, 2021). Furthermore, McDonald & Scudder (2023) have highlighted structural and organizational barriers to the funding of the kinds of research that autistic people want (McDonald & Scudder, 2023).

Many critiques of genetic and genomic autism research hinge on the idea that such research has grown out of ableist assumptions about autism. Ableism is “a network of beliefs, processes and practices that produces a particular kind of self and body (the corporeal standard) that is projected as the perfect, species-typical and therefore essential and fully human. Disability then is cast as a diminished state of being human” (Campbell, 2001). Assuming a “species-typical” person as an ideal leads to the ableist idea that disabled people are inferior to non-disabled people. This perpetuates assumptions that autistic children and adults require interventions in order to meet societal expectations, rather than vice versa, and ignores the ways that societies disable autistic individuals (see e.g. Kapp, 2011). While it is important to note that not all autistic people consider themselves to be disabled, it is equally important to recognize that one does not have to be disabled to be subject to ableism, since ableism is often located in the beliefs and actions of others. Autism ableism includes pervasive negative stereotypes about what life is like for autistic people, and the consequent assumption that any given person’s life would be better were they not autistic. As Campbell writes, “a chief feature of an ableist viewpoint is that impairment or disability (irrespective of ‘type’) is inherently negative and should the opportunity present itself, be ameliorated, cured or indeed eliminated” (Campbell, 2009). It is crucial that those working in medical research understand that ableism can dictate the characteristics that research seeks to ameliorate, cure, or eliminate. Simultaneously, it is crucial to understand ableism in genetic and genomic autism research in historical and social context.

Autism genetics and ableism in historical and social context

From the earliest identification of autism as a condition, its history has been rife with the stereotyping, alienation, harm, and abuse of autistic people. At the root of these harms has been the ableist idea that autistic people are deficient compared to non-autistic people and that science should orient itself towards “correcting” autistic “deficiencies.” Apprehension towards scientific research amongst autistic and autism communities makes sense in relation to this history, and it is important for researchers to understand where their own work sits in relation to the troubling genealogy of autism research. Professionals working in genetics should be especially attuned to the dangers of ideas about some kinds of people being inferior to others given how deeply their field is entwined with the broader history of eugenics (Jackson, 2023), including the forced institutionalization and sterilization of tens of thousands of people in the USA (Reilly, 2015). New technologies lead community members to fear that disability could simply be “edited out,” leaving disabled people further stigmatized and society all the poorer for its decreased heterogeneity (Cokley, 2017). In this context, it is important to examine the ways that researchers discuss the conditions they research and the assumptions that might lay unexamined at the root of their work.

Language in research

While autistic communities have voiced opposition to autism genetics, researchers in these areas have begun to reckon with the significance of the language they use in the course of their work and the harmful ideas that may thereby be reinforced as their research informs policies, practices, and attitudes with real-world consequences for autistic people. With regard to autism, Bottema-Beutel et al have drawn on autistic scholarship and disability studies to describe the ways that ableism manifests in communication about autism research, identifying a series of potentially ableist discourses and language (Bottema-Beutel, 2021). Simultaneously, in genetics and genomics Cho et al note that since human genetics focuses on identifying and explaining difference, “reassessment of language is particularly important. […] language and concepts of difference can be not only inaccurate but also harmful” (Cho et al 2023). With specific regard to disability, Cho et al argue that “Changing the language used in human genetics to describe disability is a critical step toward undermining structural disability discrimination in the field. […]. Genetics professionals must also seek to uncover biases in how groups are defined in the first place” (Cho et al, 2023).

In response to Cho et al’s call for genetics professionals to “continually reassess the words we use to describe human difference” (Cho et al, 2023), we used recommendations made by Bottema-Beutel et al as a framework to explore ableist rhetoric in NIH-funded genetic and genomic autism research. Our analysis of 166 research projects offers empirical evidence of ableism within genetic and genomic autism research, which should prompt reflection, discussion, and change.

Analyses of the language of published research have been used to identify common and contrasting patterns within and between disciplines (Bunton, 2005), between countries (郭清河, 2017), and over time (Scalfani, 2017). Clarification of the uses of language can prompt researchers and research institutions to reflect on the way they pursue and discuss their work. Relevant examples from this field include an analysis of shifts over time in the use of population descriptors including “ancestry,” “ethnicity,” and “race” in the society journal The American Journal of Human Genetics (Byeon, 2021). Our own institution, the National Human Genome Research Institute, is inconsistent in its use of terms: the “Talking Glossary of Genetics” entry for autism was written by CG and strives not to include stigmatizing terms (National Human Genome Research Institute, 2025), whereas the older “Genetic Disorders” entry on autism does (National Human Genome Research Institute, 2019). Analysis of research documentation is less common, despite the importance of the process of applying for funding for scientific research. Writing abstracts and grant applications is considered a core skill for researchers, who typically receive specific training from their home institution as well as learning through practical experience. There is evidence that applicants increasingly use “hype” (subjective or emotional language) to write persuasive applications (Millar, 2022). “Selling science,” Vinkenburg et al. argue, is a “tacit criterion” during research funding decision-making processes. Their analysis exposed bias in the evaluation of grant applications and showed that the presence of such biases in the management of scientific research has serious implications for scientific quality (Vinkenburg et al, 2021). Influenced by such work, we set out to explore the language of genetic and genomic autism research using the titles, abstract text and public health relevance statements for projects awarded funding by the NIH. Research documentation is subject to both formal requirements and explicit and implicit norms. Prospective and ongoing research is a space of uncertainty, promise, and persuasion, where prospective grantees feel the requirement to “sell their science” within the confines of strict word counts. For these reasons, the language of research cannot be taken to straightforwardly correspond to individual researchers’ beliefs and behavior. Nevertheless, the language of research can embody and perpetuate harmful ideas. In addition to seeking to understand the overall prevalence of potentially ableist language and discourse in NIH-funded genetic and genomic autism research, we hypothesized that prevalence may differ across different types of research as identified by NIH Spending Category.

Methods

Search Strategy

We identified projects for potential inclusion in this study using the Advanced Search functionality of the NIH Research Portfolio Online Reporting Tools Expenditures and Results (RePORTER) module, an electronic tool that allows users to search and export data from a publicly-accessible repository of NIH-funded research projects. We used trial and error to identify search terms that would retrieve relevant projects – that is, projects that both explicitly presented themselves as making a contribution to autism research and had a sufficient focus on genetics or genomics or made significant use of genetic and genomic technologies. Our final search returned 13136 entries in NIH RePORTER for projects between the years 1985 and 2022 (the last full calendar year at the time of data collection) whose titles, abstract text, or public health relevance statements included the words “autism”, “autistic,” or “Asperger” in addition to at least one of an extensive list of genetics and genomics key words and autism-associated genetic conditions (see Supplementary Materials).

Data Extraction

Search results were downloaded from NIH RePORTER as a .csv file and wrangled in Excel to produce the final sample. Data from the sample was exported to Microsoft Word and imported to Nvivo for coding. Coding data was exported from Nvivo and joined with the sample dataset in Excel for analysis.

Project Selection

Changes to the abstract text within the same multi-year grant period (most often five years for NIH grants) have historically been rare, so we chose to focus on the earliest year of any studied grant. In cases where entries for multi-year or otherwise related projects contained identical abstract text, we excluded all but the earliest entry into NIH RePORTER (n=7710, 59% of the search results). Non-identical abstract text from multi-year projects displayed a very high degree of similarity, and so we also excluded all but the earliest entry from the remaining multi-year projects (n=3490, 27% of the search results). To check the validity of this exclusion decision, we selected ten grants at random from our large sample, and verified that the text did not change by even one word from year 1 to the latest project year within the same award, if there were multiple years. In order to generate a sample representative of the larger dataset across 15 characteristics including Organization Type, Grant Type, Fiscal Year, and which of NIH’s 27 Institutes and Centers (NIH, 2024) administrated the study (see Supplementary Materials), each project was indexed and assigned a weighted probability of selection calculated based on the percentage of projects in the dataset with which it shared each of those 15 characteristics. Numerical corrections were applied such that the cumulative probabilities for the dataset totaled 1. Numbers between 0 and 1 were randomly generated using the RAND function in Excel and matched to cumulative probabilities in the dataset to populate the pilot sample and final sample. Pilot results prompted an additional check for project relevance for the final sample, after the study team reflected that many of the projects in the pilot sample were not sufficiently autism-focused and some merely mentioned autism in passing. Therefore, we added an additional check for project relevance to autism to the inclusion criteria: for the final dataset, we excluded projects with fewer than four uses of the words: “autism,” “autistic,” or “Asperger” (n=910, 47% of pilot dataset). We found that this did not affect representativeness across the 15 characteristics described above. The final dataset contained 1936 studies.

Coding

We conducted a pilot study wherein the lead author assessed the feasibility of the study by using Nvivo to single-code the title, abstract text, and public health relevance statements of 100 projects for the potentially ableist discourses and language identified by Bottema-Beutel et al. (2021). Bottema-Beutel et al. identified 18 categories of ableist language and discourse in autism research. They differentiate between “Potentially ableist terms” (a list of specific words and phrases, further subdivided into “Patronizing language” and “Medicalized / deficit-based language”), and “Potentially ableist discourses” (ways of discussing autism not relegated to the use of particular terms, that reflect and/or contribute to dehumanization, oppression, or marginalization of autistic people). The pilot initially employed deductive coding using this pre-defined list. During the course of pilot coding a number of additions were made to the codebook, and the coding approach evolved from merely deductive to a blended or abductive approach, one which might include iterations of a pre-defined codebook but which also allows codes to be developed directly from the data (Skjott Linneberg et al, 2019).

The pilot analysis identified this need to make additions and changes to the codebook (and to remain open to further changes), since not all of the apparent ableism within the data could be captured within the deductive framework. During the pilot, the discourse of prevention and the terms “Mental Retardation” and “Patient” were added. Prevention discourse presented autism as something that scientific research should seek to prevent from occurring. “Mental retardation is an outdated and offensive term for learning and developmental disabilities, one which was replaced by “intellectual disability” across all US federal laws and regulations under Rosa’s Law in 2010. “Patient” refers to an individual seeking or receiving medical treatment for their condition(/s), which is not true of all autistic people (or, indeed, of all participants in biomedical research). Three terms were also added under a new category of “potentially stigmatizing terms”: “Debilitating”, “Devastating”, and “Severe”. These are all terms that, in context, implied that autism is straightforwardly and entirely negative for individuals, families, and communities. The item “psychopathology” was also broadened to include other terms with the same root (neuropathy, pathology, etc.). The abductive approach taken during the final study led to the addition of one final term: “disease”. Table 1 represents the final codebook. In Table 1, we also include suggested alternatives from Bottema-Beutel et al. (2021) and from our own wider work, as one objection we hear from researchers is that they do not have other suitable options; Natri et al (2023) and Bottema-Beutel et al (2023) have established that inclusive language and scientific accuracy are compatible (Natri et al, 2023) (Bottema-Beutel et al, 2023). 66% of the projects in the pilot sample were coded as containing potentially ableist discourse, language, and stigmatizing language.

Table 1: Codebook.

Asterisked definitions and suggested alternatives are directly from Bottema-Beutel et al. (2021)

Code	Definition	Example	Suggested alternative
Potentially ableist discourse
Burden	Burden of autism	“a rapidly growing medical and social burden” (ID:128, 2017)	Describe specific effects on specific people e.g. autistic individuals, caregivers. Consider whether you are advocating for minimizing the allocation of resources to support autistic people or advocating for ensuring that their needs are met.
Cost	*Discussions about economic impacts of autism that situate costs in the existence of autistic people themselves, or compare the costs to those of potentially fatal diseases / conditions such as cancer or stroke	“the long-term medical and educational costs for children with ASD approach $13 billion annually” (ID:103, 2017)	*Discussions about economic impacts of autism that situate costs in society’s systemic failure to accommodate autistic people and that recognize the people most affected by oppression to this failure are autistic people themselves (not “taxpayers)
Cure	*Cure / recovery / “optimal outcome” rhetoric	“help design more specific and effective therapeutics to prevent, treat and even reverse ASD” (ID:146, 2017)	*Discussions focusing on quality-of-life outcomes that prioritize what autistic people want for themselves
Deficits	*Interpretations of all group differences between autistic and nonautistic groups as evidence of autistic deficits	“characterized by deficits in attention, communication and social interaction” (ID:103, 2017)	*Interpretations of group differences that consider the possibility that autistic people may have relative strengths over nonautistic people or that differences between groups are value-neutral unless actively demonstrated otherwise
Epidemic	*Describing autism as an epidemic	“Recent reports have indicated a dramatic rise in the incidence of Autism spectrum disorders” (ID:010, 2013)	*Autism as increasingly recognized / diagnosed
Passing	*Prioritizing “passing” as nonautistic at the expense of mental health and wellbeing	NOT FOUND IN SAMPLE	*Prioritizing mental health and well-being, which can include embracing autistic identities
Prevention Added following pilot	Describing prevention or eradication of autism as a research goal, or as an ultimate goal towards which research contributes	“Little is known concerning the genetic determinants of autism pathophysiology, limiting treatment and prevention strategies” (ID:108, 2007)	-
Puzzle	*Describing autism as a puzzle	NOT FOUND IN SAMPLE	*Autism as part of neurodiversity
Potentially ableist language: patronizing language
Behavior	*Challenging behavior / disruptive behavior / problem behavior	“abnormal behavior, such as irritability.” (ID:104, 2018)	*Meltdown (when uncontrollable behavior), stimming (when relevant), specific description of the behavior (e.g., self-injurious or aggressive behavior)
Person-first language	*e.g. Person with autism	“This grant uses biological samples from children with autism” (ID:116, 2020)	*Identity-first language; “on the autism spectrum”
Special interests	*Special interests	NOT FOUND IN SAMPLE	*Areas of interest or areas of expertise, focused, intense, or passionate interests
Special needs	*Special needs	NOT FOUND IN SAMPLE	*Description of specific needs and disabilities
Potentially ableist language: medicalizing language
Comorbid	*Comorbid; comorbidity	“Aggression is a frequent comorbidity in autism spectrum disorders (ASD)” (ID:029, 2017)	*Co-occurring
Healthy controls	*Healthy controls / normative sample	“an independent neuroimaging cohort of ASD, SCZ cases and healthy controls” (ID:112, 2014)	*Nonautistic (if determined via screening), neurotypical (if determined via extensive screening ruling out most forms of neurodivergence), comparison group (with description of relevant group characteristics)
Disease Added during final coding	Describing autism as a disease	“This study will utilize a mouse model of the disease” (ID:109, 2016)	Condition
High-Low	*High/low functioning; high/low severity or support needs	“part of an ongoing clinical trial in high and low functioning children with ASDs” (ID:030, 2017)	*Describe specific strengths and needs, and acknowledgment that the level of support needs likely varies across domains (e.g., requires substantial support to participate in unstructured recreation activities, but minimal support to complete academic work)
Mental retardation Added during pilot	Mental retardation	“Human mutations in CASK lead to X-linked mental retardation with some of the patients falling within ASD phenotype” (ID:174, 2015)	Intellectual disability; cognitive disability
Patients Added following pilot	Autism patient; assumes that autism is an illness and that autistic research participants are seeking treatment when they may not be	“studies have demonstrated that both human patients with ASD and rodent models of ASD can acquire trace conditioning similarly to controls” (ID:100, 2021)	Person, participant
-Pathy Amended during pilot	*Psychopathology, neuropathology, pathological, etc. in relation to autism	“Little is known concerning the genetic determinants of autism pathophysiology” (ID:108, 2007)	*If referring to a condition: Neurodevelopmental conditions, neuropsychiatric conditions, developmental disabilities, mental illnesses (or specific mental health condition). If referring to a process: etiology, cause, mechanism, development, difference
Risk	*“At risk” for ASD	”Familial risk for ASD alters connectivity in developing brain” (ID: 103, 2017)	*Increased likelihood / chance of autism
Suffering	*Suffering from autism	“Autism spectrum disorders affect over 1% of the population and cause suffering for patients and their families” (ID:110, 2013)	*Impact, effect
Symptoms	*Autism symptoms	“Research is urgently needed to identify markers that precede symptom onset” (ID:101, 2016)	*Specific autistic characteristics, features, or traits
Treatment	*Treatment for autism, treating autism	“allowing the development of targeted drug treatment strategies” (ID:116, 2020)	*Support, services, educational strategies (where applicable)
Potentially stigmatizing language
Debilitating Added during pilot	Describing autism and/or autistic characteristics as debilitating	“a common and profoundly debilitating psychological disorder” (ID:001, 2008)	Describe specific strengths and needs, and acknowledge that the level of support needs likely varies across domains
Devastating Added during pilot	Describing autism and/or autistic characteristics as debilitating	“a prevalent and devastating neuropsychiatric condition” (ID:141, 2016)	Describe specific strengths and needs, and acknowledge that the level of support needs likely varies across domains
Severe Added during pilot	Describing autism or autistic characteristics as causing “severe” difficulties in and of themselves	“their severe communication deficits” (ID:169, 2019)	Describe specific strengths and needs, and acknowledge that the level of support needs likely varies across domains

Given the highly nuanced contextual nature of the items in our codebook, we elected to hand-code rather than use computational tools to identify words and phrases; therefore, our sample size reflects a balance between the resource available to us and the resource required to code meaningfully: close reading was necessary to ensure that the coders identified only cases where the language and discourses in the codebook were being employed specifically in relation to autism. We generated a sample of 205 entries (20% of the dataset). During coding, 39 entries were identified as not being sufficiently relevant to genetics or genomics and were excluded, leaving a final sample of 166 projects.

The title, abstract text, and public health relevance statement from each entry in the sample were assigned an identification number between 1 and 166, exported into individual Microsoft Word Documents, and uploaded into NVivo where they were triple-coded by three coders with expertise in history and philosophy of science, biology, psychology, and genetics. Each coder exported their coding data into Microsoft Excel to be compared and integrated into a single dataset. After the first round of coding, there was very high (87%) agreement across all three coders. At this stage, “disease” was added to the codebook and all three coders revisited the sample to code for the use of this term in relation to autism. Remaining disagreements over coding were resolved through group discussion across the study team.

Although every individual instance of potentially ableist discourse, language, and stigmatizing language was coded within each project, the coders agreed that, given the style and tone of scientific abstracts and the propensity of applicants to repeat sentences word-for-word at the start and end of an abstract and/or in the public health relevance statement, the presence of multiple uses of the same term within a project was not more meaningful than the presence of the same term only once. To reflect this and avoid inflated counts, the coders produced a final dataset where coding data was transformed into a series of binary variables indicating the mere presence or absence of each code within each project.

Community Involvement Statement

Members of the autistic community were not directly involved with this study, but several autistic people were consulted in the early stages of the study and later presentations on the findings.

Ethics statement

This work was designated non-human subjects research by the appropriate Institutional Review Board.

Results

Quantitative Findings

Overall, potentially ableist discourse, language, and stigmatizing language were ubiquitous across our sample. 97% of projects coded contained at least one instance (Figure 1). Potentially ableist language was present in more than 90% of projects, potentially ableist discourse was present in more than 50% of projects, and potentially stigmatizing language was present in more than 30% of projects. Five items were present in more than a third of the projects in our sample (Figure 2): (1) using deficit discourse, which presents autistic people’s differences as deficiencies compared to nonautistic people (43.37%); (2) discussing “treatment” for autism (42.17%), presenting autism as always requiring medical intervention; (3) using the language of pathology to describe autism (36.14%); (4) using “risk” language about autism (34.94%), suggesting that autism is inherently dangerous; and (5) using person-first language (e.g., “person with autism”) as opposed to identity-first language (e.g., “autistic person”) (33.73%). We note that this last figure is lower than the 65% of journal abstracts found to do so in a 2024 study of person-first and identity-first language in autism research journals (Zajic & Gudknecht, 2024). However, Zajic and Gudknecht were able to categorize all abstracts in their study as using person-first language, identity-first language, or both, whereas we reflected that identity-first language was vanishingly rare across our sample: many of the 66.27% that did not use person-first language simply did not refer to autistic people at all.

Figure 1:

Number of codes identified in sampled projects

Red lines indicate quartiles.

Figure 2:

Prevalence of individual codes across sample projects

For reporting purposes, the NIH categorizes research projects as belonging to one or more NIH Spending Categories, which identify the type of research taking place. This enabled a comparison of the prevalence of potentially ableist discourse, language, and stigmatizing language across different kinds of research. NIH Spending Categories for each project were contained within the dataset we downloaded from NIH RePORTER and so we were able to group projects by Spending Category within Excel and count the number of codes present within each Spending Category. Spending Categories are not mutually exclusive and thus the total number of projects in Figure 3 is greater than the total number of projects in Figure 1. The average number of items (i.e., the number of different types of potentially ableist discourse, language, and stigmatizing language) per project was consistent across all of the spending categories in our sample, falling within a very narrow range of between 2 and 4 per project; therefore, we were not able to confirm our hypothesis that the prevalence of ableist language and discourse would be different in different types of genetic and genomic autism research. However, it is perhaps notable that projects in the “Prevention” spending category had the highest average (4.24 different items per project, Figure 3). A similar study with a larger sample size might yield more conclusive results.

Figure 3:

Average number of codes present per project across NIH Spending Categories

Medicalizing language was the most common category of items (Figure 3); this is perhaps unsurprising in a biomedical context where professionals are trained to view people through the medical gaze (Foucault, 1973). Although it is the case that this category contained the most individual items (11), this difference alone cannot account for its overwhelming presence compared to Ableist discourse (8 items), Patronizing language (4 items), and Stigmatizing language (3 items).

Qualitative Findings

Close reading of our coding terms in context and in-depth discussion across the research team informed the identification of seven themes in addition to the specific language and discourses included in the quantitative analysis. These findings are tentative and we present them here in order to inform future discussion of and research into the ways that autism is presented and understood in genetic and genomic research. Table 2 lists the seven identified themes, along with a brief explanation and examples from our sample.

Table 2: Emergent themes.

ID numbers correspond to projects in our sample.

1. Autism as disastrous Researchers often presented autism as a disaster, something inherently negative and a tragic source of great suffering. The language in these examples is emotive and unscientific.	Examples: “this disastrous disorder” (ID:038, 1998) “this debilitating human condition” (ID:067, 2011)
2. Lack of treatments for autism as the research problem Researchers often defined lack of treatments as the problem that their research contributed to solving. Specifically, many researchers saw a need for treatments for autism itself and for specific autistic traits (rather than treatments for co-occurring conditions) – and framed their research as ultimately helping to meet that need. Though some autistic people want treatments, overall this is not a priority for autistic communities and indeed many autistic people oppose the seeking of treatments or cures for autism and autistic traits.	Examples: “This research addresses one of the greatest public health needs in mental health today which is finding safer and more effective medical treatments for autism.” (ID:022, 2007) “The long-term goal of our research is to advance the science behind the pharmacologic treatment of autism and related disorders. This is especially important since there are very few treatments that have been shown to be effective for this devastating disorder.” (ID:022, 2007)
3. Drawing equivalence between autism and serious or life-threatening conditions like cancer Sometimes projects for research with potentially wide-ranging applications included drawing equivalence between autism and serious or life-threatening conditions like cancer in a manner that could be interpreted as presenting cancer and autism as equally negative conditions. Whilst there may be shared biological mechanisms relating to both conditions, researchers should be careful not to appear to draw equivalence between autism, an inherent part of neurodiversity across the human population, and conditions like cancer that are life-threatening.	Examples: “Epigenetic mechanisms act at the interface between genetics and environment and are an important determinant in disease risk to complex human diseases such as cancer or autism.” (ID:162, 2006) “this work will elucidate epigenetic processes essential for preventing cancer, for ensuring fertility, and for normal brain function, presenting an unprecedented opportunity to identify convergence between cancer, infertility, and autism that may impact how these diseases are treated in the future.” (ID:143, 2014)
4. National and/or international scale Prevalence, increasing diagnosis, and estimated costs were often emphasized at the national or even international scale as researchers sought to invoke the supposed size and scale of autism as a financial and social “problem.” This implies that autism is a “problem” because supporting autistic people can be expensive, rather than addressing the difficulties that autistic people themselves might face.	Examples: “According to the Centers for Disease Control and Prevention (CDC), about 2.25% of children in the U.S. have been diagnosed with Autism Spectrum Disorder (ASD). The economic burden of ASD in the U.S. for 2015 has been estimated to be $268 billion in direct medical, direct non-medical, and productivity costs.” (ID:077, 2019) “Given the high prevalence of autism and the immense, lifelong adverse impact it has on affected individuals and their families, the present lack of any effective treatment or preventive intervention based on the biology of the disorder has produced an burgeoning, very expensive public health burden on the United States and worldwide horizon.” (ID:092, 2013)
5. Medical model of disability Several projects recognized select difficulties faced by many autistic people, but framed these in terms of the characteristics of autism, in opposition to a narrow view of human flourishing. This focus on the ostensible differences and failures of the autistic person, rather than emphasizing how society fails to meet the autistic person’s needs, reflects the medical model of disability, which has been heavily critiqued (see e.g. Wasserman, 2022).	Examples: “Repetitive behaviors and rigid, inflexible thinking are cardinal symptoms of the disorder and prevent many otherwise high functioning people from being productive members of main stream society.” (ID:026, 2010) “Autism spectrum disorders (ASD) and schizophrenia often result in marked social deficits, which manifest as significantly limited social supports and relationships, difficulty maintaining employment, and overall deficits in social skills.” (ID:141, 2016)
6. Dehumanization and objectification Although autism is exclusive to humans and its characteristics are often defined interpersonally (for example, differences in social interaction), we saw the adjective “autistic” and the diagnosis “ASD” applied to other entities including organoids, post-mortem brains, and whole families. Occasionally, the presence of an autistic person was not invoked at all: rather, researchers discussed autism, autism symptoms, or “autistic brains.”	Examples: “This application aims at identifying fundamental mechanisms of regulation of gene expression at early stages of human brain development in cellular model (brain organoids) from normal individuals and autism spectrum disorders (ASD).” (ID:035, 2018, our emphasis) “normal organoids and ASD organoids” (ID:035, 2018)
7. Preventing autism The presence of a discourse of prevention in 10.84% of the projects in our sample is concerning. As noted above, a fundamental worry among many autistic people is the possibility that the pursuit of genetic knowledge and use of novel technologies will promote reproductive decisions and interventions with eugenic consequences. This fear is informed by the historical abuse of autistic people (see above) and recent decreases in the number of babies born with Down Syndrome in countries offering prenatal screening and access to abortion (Lou, 2018) (Shang, 2020).	Examples: Primary prevention of these disorders is of great public health importance, calling for the identification of modifiable risk factors, such as environmental chemical exposures.” (ID:128, 2017) “Therefore, this research is relevant to part of NIH’s mission, since it develops the basic knowledge that will lead to improved methods for preventing and treating autism spectrum disorders.” (ID:109, 2016)

Discussion

Potentially ableist discourse, language, and stigmatizing language are not unique to research environments, and researchers learn from and rely on an ecosystem of resources which themselves reinforce the ableist terminology. The most recent edition of the widely-used Diagnostic and Statistical Manual of Mental Disorders (DSM V-TR, 2022), for instance, frames autism in terms of deficits and abnormality (“abnormalities in eye contact and body language or deficits in understanding and use of gestures”) and uses many other of the terms and discourses identified by Bottema-Beutel et al. (2021). It is also probable that in many cases researchers writing about their work simply echo the language used in the funding calls to which they are responding, and possible that researchers feel pressure to “hype up” their work to increase their chance of being awarded funding. Our empirical finding that potentially ableist discourse, language, and stigmatizing language were ubiquitous across projects for NIH-funded genetic and genomic autism research over the last 30 years should, therefore, provoke reflection and critical discussion within and beyond genetic and genomic autism research.

Most projects in our sample (83%) contained multiple items of potentially ableist discourse, language, and stigmatizing language (mean: 3.08; median: 3.00); 61% contained three or more items; 33% contained four or more items; 20% contained five or more items; and 8% contained six or more items (Figure 1). This indicates that what our coding captured was not simply a series of isolated usages, but rather a more general state of affairs where researchers rely on multiple interacting potentially ableist ideas to describe their work. This is further demonstrated by the themes that emerged from our qualitative analysis based on the study team’s discussion of the sample and reflections on reading and coding the projects.

The most common specific items in our analysis (deficits; treatment; -pathy; risk) suggest that within NIH-funded genetic and genomic autism research from 1985 to 2022 autism has been constructed as a mostly or entirely negative condition, one that is inherently pathological and requires treatment, and one that poses a dangerous “risk” that scientific research must work to reduce. Alongside the seven themes that emerged through our qualitative analysis – most notably the discourse of prevention – this externalizing, medicalizing, and pathologizing language emphasizes deficits and envisions a world where there are fewer autistic people, and where people are “less autistic.” We assert this is inconsistent with ethical and equitable scientific research.

Four terms from Bottema-Beutel et al.’s list were not present in our sample: the discourse of “passing” (appearing to be nonautistic), the idea of autism as a “puzzle” to be figured out, and the terms “special interests” and “special needs”. There was only one mention of “problem,” “challenging,” or “abnormal” behavior in our sample. The coders reflected that the absence of most of these terms (with the exception of “puzzle” discourse) was likely explained by the lack of any kind of discussion or description of autistic people’s everyday life or specific discussion of their interests, behaviors, or needs.

More generally, both the absence of four terms and discourses identified by Bottema-Beutel et al. and the presence of terms and discourses not identified by their research suggests that, overall and in general, NIH-funded genetic and genomic autism research employs its own particular narrative about autism, one that is often ableist and stigmatizing. The presence of a discourse of prevention is especially troubling given well-known concerns among many autistic communities about the potential for genetics and genomics to be used for eugenic purposes.

Recommendations

We recommend that those working in genetic and genomic autism research reconsider the way they describe and discuss autism and, at the least, use the more inclusive terms suggested by Bottema-Beutel et al. (2021) and ourselves (Table 1). More broadly, in a context of criticism and mistrust, there should also be reflection on the assumptions at the foundation of research projects, which should be clearly explicated in research documentation along with the presumptive benefit to relevant communities. This would require engagement with autistic communities and an openness to changing the research one does, how one conducts research, and how one communicates about research. Recent research has shown a serious lack of public and patient involvement and engagement work in autism genetics, noting that “If this lack of engagement is not addressed, it may result in study disruption, loss of trust between researchers and the intended beneficiaries, and ultimately bias results (Kaljusto et al, 2025).”

We believe that further research seeking to understand the language and discourse of genetic and genomic autism research and/or ableist language and ideas in research documentation would be valuable, because it provides an empirical bases for ongoing discussions about anti-ableism and inclusion. Future research could utilize computational methods to process a large sample (an approach also used in Kaljusto, 2025) though we believe close reading will always be an important component of research into the use of highly nuanced and context-dependent terminology. Cross comparisons across different types of research (within and beyond genetics and genomics) and internationally could also provide a useful evidence base: ableism manifests differently in different places, and working to further inclusion in scientific research requires an understanding of and sensitivity to those particularities. Finally, studies using document analysis to understand ableism must be complemented by qualitative work with researchers themselves in order to understand what researchers themselves think (which may indeed be different to what they write in grant applications or published work) and to paint a full picture of how ableist ideas operate in research. In this way, a range of empirical evidence can be made available to highlight unexamined assumptions and norms in research culture, provoking individual reflection and furnishing discussions about policy and practice.

Botha and Cage have shown that autism researchers rely on an “amorphous” and ill-defined idea of autism, and that narratives in autism research are reflective of “core ideological disagreements […], such as whether researchers consider autism to be an intrinsic barrier to a good life, and whether researchers prioritize research which tackles ‘autism’ versus barriers to societal inclusion for autistic people” (Botha & Cage, 2022). In autism genetics, there is a need for further research into the beliefs and attitudes of individual genetics and genomics researchers towards autism and autistic people in order to understand how they conceptualize autism and where they fall across the range of research priorities, and in order to understand the potential barriers to and the support and resources required for more inclusive – and less ableist – research and communication.

Limitations

The foremost limitation of this study is the relatively small sample size. This did not allow us to compare results over time, across NIH institutes or specialties, or across different kinds of research. This study did not include studies which applied unsuccessfully for NIH funding, which were funded by other means, or which took place outside the US. It is also theoretically possible that abstract text had changes over time within the same award, and our choice to use the earliest year might have affected the results. However, such abstract changes within award terms have been quite rare before 2025, and our validation check did not find any. Changes in NIH policy by the current US administration mean that future analyses would have to account for a much higher frequency of abstract revisions than ever in NIH history.

The nature of the coding process also had limitations: firstly, the items identified by Bottema-Beutel et al (2021) were intended as a guide for researchers and not as a research tool (though they have been utilized in at least one other study looking at inclusion in genetic and genomic autism research (Kaljusto et al, 2025)). Nevertheless, since their work is rooted in the work of autistic scholars and informed by disability studies and discourse analysis, we believe that it was the best available starting point for our codebook. Another possible limitation in our coding process is, that since we set out to understand the prevalence of potentially ableist language and discourse, we did not actively look for or record examples of anti-ableist or neuro-affirming language and discourse. The coding team noticed and discussed some examples of inclusive language that appeared in our sample, but reflected that these instances were few and far between.

Supplementary Material

Supplemental material for this article is available online.