Presuming competence and assessing nonspeaking people

Presuming competence is an approach that starts from the assumption that a person can understand and learn, rather than requiring them to show it before they are treated as capable. It is used most often for people who do not speak, whose ability is easy to miss when it is judged by what they say or point to.

The phrase has a documented origin. Anne Donnellan, an autism researcher and advocate, set out the related idea of the "least dangerous assumption" in 1984 and argued for presuming competence, especially but not only for people who do not speak.¹ Many autistic writers use the phrase in the same way, as a rule for how to act toward a person whose ability is not visible (see In the community, below).

The phrase covers two things worth keeping apart. One is a stance about how to treat people: give a person the benefit of the doubt about what they understand, because the cost of assuming too little is higher than the cost of assuming too much. The other is a question about what a particular person does understand, which is what assessment is meant to answer. Most of this entry is about the second.

Standard measures of intelligence and language usually need the person to answer out loud or to point, press or otherwise move to show what they know. When speech and pointing are unreliable, the test may go unfinished, or the score may fall below what the person understands. A 2026 paper on assessing grammar in young autistic children notes that the usual tasks are "inadequate for many autistic children," with low rates of completion.² A 2026 study of minimally verbal autistic adults opens by noting that standard neuropsychological testing "frequently fails to capture the true capacity and full range of abilities" when the motor side of speech is hard.³

One 2025 preprint (not peer reviewed) measured this directly. Thirty-five minimally verbal autistic adolescents and adults, 30 of them with usable eye-tracking data, matched words to pictures by looking rather than pointing. In the reading version of the task, the group's eye-gaze accuracy averaged 78%, below the 97% of non-autistic controls, though the timing and spatial pattern of their looking were close to the controls', and 24 of the 30 (80%) scored significantly above chance. Asked to read the word and point to the right picture, they averaged only about 57%. In a second version, with spoken words instead of written ones, eye gaze and pointing did not differ significantly, and 22 of the 30 finished it.⁴

Motor difficulty is one proposed reason. In a 2026 account written with three minimally speaking autistic adults, apraxia, a difficulty planning and carrying out intended movements, is described as a primary barrier to effective communication, and as something often mistaken for a limit on understanding.⁵ A 2026 study that recorded the activity of facial and neck muscles during speech, in 25 minimally verbal autistic adults and 26 non-autistic adults, found tighter coupling of those muscles in the autistic group, which its authors read as less differentiated control of speech movements.³

The assessment data are often missing from the records. In a US surveillance study of 7,084 eight-year-old children with autism, IQ data were absent for about 39% of them, verbal-status data for about 26%, and adaptive data for about 44%.⁶

Understanding often runs ahead of speech in minimally verbal autistic people, and how wide the gap is varies from person to person. In a cohort of 193 minimally verbal school-age children, most of them autistic and with a mean age of about 7½, a cluster analysis found a subgroup whose understanding of language was strong while their speech production and expressive language were low.⁷ In a much larger sample of 1,579 minimally verbal autistic children and adolescents aged 5 to 18, pooled from two US data repositories, the group understood more than they could say on average, though their understanding was still below the test norms, and the gap to the norms widened with age. Motor skills were the strongest predictor of a wider receptive-expressive gap. Only about a quarter of them, though, showed a significantly better receptive than expressive level as individuals; for most, understanding and production sat closer together than the group average suggests.⁸

Something similar shows up in how language is measured. In 167 autistic children aged 2 to 6, verbal skill, spoken output and how generative the speech was overlapped only partly. Some children with little or no speech had relatively strong structural language, and some children who spoke a great deal used phrases that were not productive.⁹

Reading is another thing a test with a spoken or pointing answer can hide. The eye-gaze study described above found hidden reading ability in adults and adolescents who could not show it by pointing.⁴ A 2024 study of 31 nonspeaking autistic adolescents and adults found they tapped letters faster when the sequence spelled a sentence than when it did not, and paused at the boundaries between words, which its authors read as evidence they could spell; a published response argues the task mostly shows knowledge of spelling conventions rather than reading comprehension, and that some of the participants used partner-supported methods whose authorship is disputed.^10,11 A systematic review of reading-comprehension assessment for minimally and nonspeaking autistic children who use AAC (a preprint, not peer reviewed) found only six studies that met its criteria, so the picture of what this group understands from text is thin.¹²

If a spoken or pointing answer is what fails, one route is to measure the response a person can give. Eye-tracking lets a person answer by where they look. A 2026 study built a passive eye-tracking task for grammar in children aged 2½ to 6 that needs "no overt behavior," and more than 85% of the autistic children in it gave usable data. In the non-autistic children, its measures lined up with a standard test of grammar. The autistic children scored lower than the non-autistic children on the eye-tracking task itself, and they mostly did not complete the standard test, so their eye-tracking scores could not be checked against it in the same way. The task shows that grammar can be measured without a spoken answer; it does not show hidden grammar ability in the children who took it.²

The eye-gaze measure in the 2025 preprint works along the same lines, reading understanding from looking rather than pointing.⁴

A second route is to widen what counts as an answer. Language checklists built for speaking children ask about spoken words. When researchers changed one such checklist, the MacArthur-Bates Communicative Development Inventories, to count words a child produced by any means, including AAC, caregivers reported about 14 more words on average for 15 minimally speaking autistic children, and the mix of word types changed.¹³

A third route is to keep the motor demand in view. The apraxia account argues that understanding and the ability to act on it are separate, and that a test result should be read as what the person could do on that task on that day rather than as a ceiling on what they know.⁵ The paper on muscle activity during speech points the same way, treating motor difficulty as something to measure in its own right.³

No study found as of October 2026 has compared a motor-free or pointing-free assessment with a standard assessment in a controlled trial large enough to show whether it changes a person's diagnosis, placement or support level.

The dispute that sits next to this subject is about who forms the words when a person types or spells with a partner's support, not about whether nonspeaking people can think or communicate. This entry does not settle that question. It states what the controlled tests show and what remains open.

The controlled tests use a message-passing design: the person and the support partner are shown different things, and what the person produces is checked against what each of them knew. A 2014 systematic review for the International Society for Augmentative and Alternative Communication pooled these studies and called the evidence for facilitator control "unequivocal." A 2018 update found no new evidence in 2014 to 2018 that messages delivered through facilitated communication were authored by the person with a disability.^14,15

That is the part of the subject that reaches established: physically supported facilitated communication fails controlled authorship tests. The finding applies to that method. It does not decide the question for the methods that use less contact, which the next section takes up.

Rapid prompting method (RPM) and spelling to communicate (S2C) keep the support partner's hands on a letter board rather than on the person, and they are used widely by people who type to communicate. They make the same claim as facilitated communication, that they reveal communication a person cannot otherwise show, and the question that decides them is the same: who forms the words.

No controlled authorship test of RPM, S2C or the Spellers Method has been found in the peer-reviewed literature as of October 2026. A systematic review published in July 2026 searched for exactly that and found no study eligible to be included. Its authors call for controlled authorship trials and write that the methods "should be discontinued (if in use) or not adopted until independent authorship studies have shown that these methods are indeed valid."¹⁶ One unpublished blinded analysis of RPM, posted as a preprint in 2018, reported no correct responses when the support person was blinded, and it has not appeared in a peer-reviewed journal.¹⁷ The study most cited by supporters is an eye-tracking study in which nine nonspeaking letterboard users looked at most letters about half a second before pointing to them, and its authors argue this makes a cueing account unlikely. It is small, it has not been repeated, and it measures each writer's own looking and pointing rather than who chose the words in ordinary use. A published response argues that the study does not reveal agency in the assisted communication.^18,19

The two sides disagree about what follows from the same facts. ASHA, the professional body for speech-language pathologists, holds that authorship must be verifiable before a method is treated as a person's communication; its 2018 statements call facilitated communication a technique that "should not be used" and rapid prompting "not recommended."^20,21 In comments filed with a US federal committee on August 20, 2026, ASHA repeated that the controlled research has not shown the facilitated individual rather than the facilitator is the author.²² On the other side, a group of researchers and autistic writers argue that the methods deserve fresh study. A May 2026 commentary in Autism Research calls for research on teaching methods "long dismissed because the assistant could influence the text," and a 2022 perspective argues that the dismissal of facilitated communication is "rooted in ableist and outdated approaches."^23,24 The US Interagency Autism Coordinating Committee, in the working draft of its 2026 strategic plan dated July 17, 2026, takes a different starting point, writing that the absence of evidence "should not be represented or considered equivalent to proof of ineffectiveness or evidence of invalidity."²⁵

Both sides start from the same two points. They agree the argument is about authorship, not about whether nonspeaking people can think or communicate, and they agree nonspeaking people should have access to communication. The tests this entry is about, tests that need a spoken or pointing answer, are part of why both sides say many nonspeaking people are underread to begin with.

What would settle the RPM and S2C question is the study that has not been done: a controlled message-passing test of the same kind used on facilitated communication, run by researchers with no ties to the developers, with its questions and outcomes set in advance and any influence by the partner reported. A positive result would need repeating in a second group by an independent team before the method could be treated as validated.¹⁶

The research that would answer these questions does not always include the people the questions are about. A cross-sectional systematic review examined 249 autism language-and-communication articles published in 2023. About half (49%) compared autistic people with non-autistic people, and that share rose to 75% once intervention studies and studies that included people with intellectual disability or who were non- or minimally speaking were set aside. Where the studies predicted a direction, 60% expected autistic participants to do worse and 4% to do better. The studies that did include people with intellectual disability or who were non- or minimally speaking were less likely to use a non-autistic comparison group at all.²⁶

Two examples show what that leaves out. Neuroimaging research often excludes people with intellectual disability or who are minimally verbal; in one 2026 study that deliberately included them, getting a usable scan was linked to older age, higher nonverbal intelligence and fewer ADHD symptoms, which skews who ends up in scan-based research.²⁷ For reading, a review of what is known about comprehension in minimally and nonspeaking autistic children who use AAC (a preprint, not peer reviewed) found only six eligible studies out of 586 screened.¹²

Autistic writers use presuming competence as a rule for how to treat a person whose ability is not on display. B. Martin Allen, writing in Sincerely, Your Autistic Child, calls it "critical" and "a cornerstone of respect," and describes it as an ongoing exercise rather than a single decision.^c1 Lei Wiley-Mydske, in the same collection, gives it as an instruction to parents: "Presume competence. Presume that your child is aware and wants to understand."^c2

The value the writers place on the phrase comes partly from what happens without it. Sol Smith, an autistic professor, describes the battery of tests run during his own diagnosis, including a personality inventory, and writes that the questionnaire relies on "broad word definitions" that it "assumes I will interpret in one way," and that he has "difficulty with that."^c3 The trouble, he writes, is a test built on shared habits of reading that he does not share.

Nonspeaking autistic people describe being assessed as an experience of being read as less capable than they are. Adam J. Morgan, who points on an iPad, writes that he is "very intelligent" and reads at a high level, but that "if you assess me, I would fail."^c4 Ido Kedar, who types to communicate, writes that he "was not able to demonstrate [his] intelligence for many years" despite years of therapy.^c5 Jordyn Zimmerman, who began using an iPad at 18, writes that professionals called her "unintelligent and hopeless" while her schooling consisted of tasks like "Touch your nose. Point to the dime."^c6 Damon Kirsebom, who types independently, writes that non-speaking people are misunderstood and that he is "not a unique case."^c7 Amy Sequenzia, who types, writes that the adults around her "did not presume my competence, not when it came to big decisions about my life."^c8

Several of these writers name presuming competence as the remedy. Kayla Takeuchi, who types to communicate, writes, "My message to everyone is to presume competence in your nonspeaking friends."^c9 Grant Blasko, who uses an iPad with a word-prediction app, describes what communication support opened up, writing that he relies "solely on Augmentative and Alternative Communication."^c10