Longitudinal speaker criteria: Speakers were classified as longitudinal if they had at least 3 statements, spread across more than one calendar month, and at least 30 days between their first and last statement. 7 speakers (41.2%) met these criteria, accounting for 83.5% of all statements (n = 101).

Group comparison: A Welch’s two-sample t-test was used to compare mean statement scores between groups, given unequal sample sizes and differences in variance. Each statement was treated as the unit of analysis.

Variance: Variance is the square of the standard deviation and reflects the average squared distance of each score from the speaker’s mean. Higher variance indicates greater inconsistency across statements. A variance of 0 means every statement received the same score. For context, a variance of 1.0 corresponds to an SD of 1.0, meaning scores typically deviate about 1 point from the mean.

Temporal split: Statements were divided into two equal periods at the median statement date. This reflects the actual distribution of discourse activity, which was front-loaded in 2025, rather than imposing an arbitrary calendar cutoff. Each group (all officials, longitudinal, less frequent) uses its own group-specific median date.

Consistency: To assess how consistently individual speakers framed disability-related issues across their statements, we calculated the mean score, standard deviation, range, and total number of disability-related statements across 2025 by speaker. Variability in discourse was operationalized as the within-speaker standard deviation of scores. Consistency categories were derived empirically from the distribution of standard deviations across longitudinal speakers. Quartile cut points were used to classify speakers into four internally comparative categories: (1) Very Consistent (SD at or below the 25th percentile), (2) Consistent (between the 25th and 50th percentiles), (3) Variable (between the 50th and 75th percentiles), and (4) Highly Variable (above the 75th percentile). This empirically anchored approach situates each speaker’s variability relative to the distribution of standard deviations observed among longitudinal speakers.

Directional shift: OLS regression of statement score on statement date. Slope classified using empirical quartiles into Strong Negative, Mild Negative, Mild Positive, and Strong Positive Shift. No stable category was included as no speaker demonstrated a slope near zero.

Autism detection: Statements were classified as autism-related if the terms “autism” or “autistic” appeared in any of five data fields: (1) Quote — the verbatim statement made by the official; (2) Source — the outlet or document where the statement was recorded; (3) Context — the setting or circumstances in which the statement was made; (4) Analysis — AI’s initial analysis of the content and framing of the quote; and (5) Scoring Explanation — AI’s explanation for why the statement received its specific statement score. This broad search ensures statements are captured even when autism is referenced in the context or scoring rationale rather than in the direct quote itself.

Word clouds: Word frequencies were computed exclusively from the verbatim text of each official’s public statement — not from analytical or scoring commentary. Text was lowercased, punctuation and numbers were removed, and a comprehensive stopword list was applied to exclude common English function words, filler words, and proper names. Words appearing fewer than twice across the relevant statement subset were excluded. Word size reflects relative frequency within each subset; word color reflects the mean DDM score of statements in which that word appeared, using the same four-color scale as the rest of the report. The final button filters to statements scored 1 (Dehumanizing), allowing direct inspection of the language most associated with the most harmful discourse.

Detection method: R-word instances were identified using the regular expression pattern r[*e]ta?r?d(ed|s|ing)? applied to speaker text only. Reporter text was stripped prior to counting using pattern matching on “Reporter:” prefixes. All r-word instances are associated with a Statement Score of 1.

Contextual note: The context in which the term appeared differed across speakers. In two instances, the term was used in a derogatory manner directed toward two individuals (Governor Tim Walz and a social media user). In contrast, one usage occurred within a historical and institutional reference, specifically describing the “Wassaic Home for the Retarded.” In this instance, the term was used to reference the official name of an institution at the time, rather than as a descriptor applied directly to individuals.

Education and disability policy were frequent and consistent topics throughout 2025, and these education-focused statements received substantially higher average scores. However, in contrast to the forceful autism messaging, the more concerning aspect of special education discourse is what is left unsaid. In March 2025, President Trump announced that HHS “will be handling special needs… programs” and that these programs “will be taken out of the Department of Education” (Diament, 2025). Despite the significance of this proposed policy shift, neither the White House, Secretary Kennedy, nor Secretary McMahon have provided details about how this transfer would be implemented. Secretary McMahon’s public remarks devote comparatively little attention to how IDEA would be enforced, how oversight would function, or how schools and federal agencies would be held accountable if disability rights protections are violated.

In discussing special education and IDEA, Secretary McMahon frequently relied on non-committal and future-oriented language that leaves key structural and administrative questions unresolved. Rather than outlining a defined enforcement plan for IDEA and oversight of special education programs, she often speaks in terms of possibilities: IDEA “more than likely” resting in HHS (Bash, 2025), “we’ve not made a decision yet about the programs with children with special needs from where it will go,” (Garrett, 2025), “if” the Department of Education moves programs (Bash, 2025; Nelson, 2025), or funding continuing “whichever agency” ultimately distributes it (Gelpieryn, 2025). In her Senate confirmation hearing, when asked directly who would enforce IDEA in the absence of the Department of Education, she responded that she would “like to understand more and look into” the issue and was “not sure” whether the law might be better served elsewhere (Nomination of Linda McMahon, 2025). Even in later interviews, assurances focus almost exclusively on funding continuity, “the funding will continue to flow,” (Nelson, 2025), “we’re not cutting any of the IDEA funding,” (Arundel, 2025b), “that should be the least concern,” (Suter, 2025), while leaving largely unaddressed how compliance, dispute resolution, or federal oversight would function under a decentralized structure.

Secretary McMahon’s language consistently foregrounds that funding will continue, yet offers limited clarity about who will monitor rights, enforce guarantees, or resolve interstate disparities. The use of “if,” “more than likely,” “we’re looking at,” and “we haven’t yet made a decision” underscores that the institutional architecture for protecting disability rights under the IDEA is unsettled after the first year of this administration. The result is a messaging pattern that repeatedly promises continued funding while remaining noncommittal and leaving unclear how that funding would be distributed or how IDEA’s federal guarantees would be enforced. This pattern could be interpreted as special education and protections under IDEA are not currently a priority.

McMahon Coding: All public statements by Secretary McMahon (n = 17) were analyzed using a coding scheme that examined the presence and/or absence of three categories: Continued Funding, U.S. Department of Education Restructuring, and Civil Rights and Accountability. These themes reflect core questions and concerns that emerged in public education discourse among parent, education, and disability-focused organizations following the announcement that the Department of Education would be shut down, as well as subsequent cuts to the Department’s special education staff in fall 2025 (Autistic Self-Advocacy Network, 2025; Dilworth, 2025; Disability Rights Education and Defense Fund, 2025; Fabian, 2025; Long, 2025; National Center for Learning Disabilities, 2025; National Parents Union, 2025; National Education Association, 2026). Our coding system is defined as follows: Continued Funding captures assurances that financial support protected under the IDEA would remain intact. U.S. Department of Education Restructuring captures discussion of special education moving outside the department’s control. Civil Rights and Accountability captures references to legal protections, enforcement, and equal access for students with disabilities. Statements that did not align with these themes were coded as Other. Statements could align with more than one category.

Policy Analysis: Education bill framing was coded using the DDM rubric, with Score 1–2 treated as harmful, Score 3 as deficit-framing, and Score 4 as humanizing.