The research
There are already AAC products. These are the gaps the published research shows in them, and how each gap shaped myEZvoice, EZspeak and EZbot. Every item is sourced and says how far it was checked.
We did not set out to build a tablet.
First
Our first goal was a device for ages 12 and up, for those who like Evan had outgrown the tablet: its visual cues, its structure and its tedious navigation. That became myEZvoice.
Then
The more research we did, the more we saw a real need for younger children to have more modern visual cues and a different structure: sentence starters, rooms that hold the words for each daily activity, and pictures in those rooms that open boxes of words with the verbs along the side. That became EZspeak.
And
EZbot, designed to help the child, to help parents use the board at home, and to help us measure when a child is advancing, so the board can grow with them by skill and by age.
The three we lean on most. Each was read in full: a CDC report, the Cochrane review, and the field’s own journal.
The need
1 in 31
eight-year-olds identified with autism in 2022, up from 1 in 68 in 2010.
CDC MMWR, 17 April 2025 Read in full · The studies
Read it carefully: these are children already identified by a provider, and CDC changed how it counts with the 2018 data, so the years are not a like-for-like count.
The gap
2 trials
The Cochrane review found only 2 randomized trials (154 children) for minimally verbal autistic children and rated the evidence “very low.” In the one AAC trial, “gains were not maintained 10 months later.”
Brignell et al. 2018, Cochrane Read in full · The studies
Read it carefully: a review of randomized trials, not of every study. Our own searches found no study that follows today’s AAC for 3 years or more.
The design
“Simply replicate”
Many AAC apps “simply replicate the approaches to representation and organization of vocabulary items seen with traditional AAC systems.” And phones “are socially valued; they are free of the stigma sometimes associated with the use of assistive technologies.”
McNaughton & Light 2013, pp. 108 and 113 Read in full · The studies
Read it carefully: written in 2013, about the apps of 2013. A view from the field’s leading researchers, not a measurement.
Each one links to the studies behind it, says how far its source was checked, and what we built for it.
The studies · Brignell et al. 2018, Cochrane Read in full; Romski & Sevcik 1996 Official record
BuiltEZspeak: EZbot counts the words a child uses without help and proposes the next level; an adult approves. Progress inside the app, not a study over years.
The studies · Holyfield et al. 2017 Official abstract
BuiltmyEZvoice: made specifically for ages 12 and up.
The studies · Apple App Store records, 12 Oct 2026 Official record
BuiltmyEZvoice: designed for a teenager’s own phone first: 9 to 12 large pictures to a screen.
The studies · Drager et al. 2003 Read in full
BuiltEZspeak: new visual cues drawn for it, and rooms whose objects open boxes of words with the verbs along the side.
The studies · McNaughton & Light 2013 Read in full
BuiltBoth: started from the research and from Evan, not from an existing board.
The studies · Johnson et al. 2006 Read in full
BuiltEZspeak: EZbot’s “Show me” builds a sentence on the real buttons so a parent sees how the board is used, and the parent chooses how much EZbot helps. Not on the phone yet.
The studies · AAC journal 2025 Read in full
BuiltmyEZvoice: whole everyday phrases on one button: one tap says the sentence.
The studies · our searches; the modeling studies used people (O’Neill et al. 2018; Sennott et al. 2016) Official abstract
A study we could runEZspeak: EZbot is software that models on the board. No study has tested that yet; EZ Voice could be the one to run it.
The studies · App Store records for LAMP Words for Life and Proloquo2Go Official record; Tobii Dynavox TD Snap page; our searches, 12 Oct 2026
In testingBoth: the tablet has the dialect setting for Mexican, Puerto Rican, Dominican and Cuban Spanish; on the phone, Cuban and Mexican Spanish are in testing. Every Spanish line is reviewed by native speakers before it is final.
Evan had outgrown the tablet: its visual cues, its structure, and the many screens it took to say something. We wanted to know what the research says about AAC for middle and high school students.
The research centered on young children. Researchers say teenagers’ needs “are quite different” and “more research is urgently needed.” Teenagers who use AAC need the right words, and how classmates see them matters.
myEZvoice is made specifically for ages 12 and up. The school day holds grade 7 to 12 classes, each with its own phrases, and the Sentence Builder has a High School topic. Every picture is the app’s own art.
“Much of augmentative and alternative communication (AAC) research for individuals with autism spectrum disorder has focused on young children.” For teens and adults, “more research is urgently needed.”
Limit: about autism; the study count behind it is not in the abstract.
Holyfield et al. 2017 Official abstract
Interventions for adolescents who use AAC should address “language needs, particularly vocabulary.”
Limit: guidance, not a test of a vocabulary.
Smith 2015 Official abstract
Of 136 adolescents, “girls were more positive than boys; and familiarity with peers with disabilities was associated with more positive attitudes.” Of 115 children aged 11 to 13, attitudes were more favourable when the device had voice output.
Limit: typically developing raters watching videos.
Beck et al. 2010 Official abstract · Lilienfeld & Alant 2001 Official abstract
One participant: “Many AAC apps feel like they’re made for kids or students, and it feels infantilizing.”
Limit: a preprint, 12 adults aged 18 to 44, not teens.
Martin & Nagalakshmi Read in full
Teenagers carry a phone, not a tablet. We looked for AAC designed for the phone.
Nearly every teen has a smartphone, and researchers say phones are free of the stigma of assistive devices. But the leading AAC apps on iPhone are the same app as on the iPad, and families use the iPad first and the phone second.
myEZvoice is designed for the phone first, for a teenager’s own phone: 9 to 12 large pictures to a screen.
US teens 13 to 17: 95% in fall 2024, the same in 2023 and 2022, up from 73% in 2014 to 2015. 1,391 teens surveyed.
Limit: all teens, not teens who use AAC; “have or have access to” at home.
Pew Research Center 2024, pp. 12 and 19 Read in full
“Mobile technologies are socially valued; they are free of the stigma sometimes associated with the use of assistive technologies.” A parent of a teen who uses AAC: “Kids using an iPad have a device that says, ‘I’m cool.’”
Limit: the authors citing another paper, and one parent’s view about the iPad.
McNaughton & Light 2013, p. 108 Read in full
Apple’s records list Proloquo2Go, TouchChat HD and LAMP Words for Life as one app for both; TD Snap is iPad only on iOS. In a 2012 survey of 232 respondents, “90% use an iPad for communication” and “over 25% also use an iPhone or iPod touch.”
Limit: the survey is a convenience sample from one app maker.
Apple App Store records Official record · AssistiveWare survey 2012 Read in full
A 17-year-old with autism had trouble activating the speech on his iPod-based device. It was first read as a motor problem and a reason to “abandon this technology”; a behavioral intervention instead “proved effective.”
Limit: one teenager, on an iPod touch running a tablet-style app.
Kagohara et al. 2010 Official abstract
On the tablet, every sentence is built a word at a time, and a teenager’s conversation does not wait.
Talking with AAC can be ten times slower than speech, and whole stored phrases made conversation faster and made the speaker seem more competent.
myEZvoice puts whole everyday phrases on one button: one tap says the whole sentence. It still has a Sentence Builder for speech therapy.
Speech runs 150 to 250 words a minute; aided AAC is “often as low as 15-25 words per minute.”
Limit: a 2025 paper citing a 2013 textbook, which was not itself checked.
AAC journal 2025 Read in full
Compared with a word construction system, “Conversational rate and perceived communicative competence were both higher when the UBD was used.”
Limit: the participants are not stated in the abstract.
Todman et al. 2008 Official abstract
We asked whether the visual cues and the structure Evan grew up with were built on research. This is where the research turned us toward the younger children.
Researchers found AAC displays were built on adults’ thinking and said they should be redesigned. Drawn scenes did better than grids with young children, and pictures of real places are now called an evidence-based practice for young children. Which display design works best is still an open question.
EZspeak, for ages 4 and up: new visual cues drawn for it; rooms that hold the words for each daily activity; objects in the rooms that open boxes of words with the verbs along the side; and sentence starters down the left, so a child begins with a whole idea. Every room opens in levels.
“The current generation of augmentative and alternative communication (AAC) technologies is largely based on conceptual models of adults who are not disabled.” “AAC technologies should be redesigned to reduce learning demands.”
Limit: the children studied were typically developing 2-year-olds.
Drager et al. 2003, p. 298 Read in full
After four sessions, children using a drawn scene of a house found 4.1 of 12 words on average; with the two grid layouts, 1.6 and 1.9.
Limit: typically developing 2-year-olds, not children who use AAC.
Drager et al. 2003, p. 304 Read in full
12 of 14 studies met all quality indicators, with positive results for 37 of 42 participants aged 3 to 8. Visual scene displays “can be considered an EBP” (evidence-based practice) for young children with ASD/IDD.
Limit: ages 3 to 8; 11 of the 12 interventions were run by researchers.
Patenaude et al. 2025 Read in full
“Future research is required to determine whether the display variables discussed influence functional and authentic communication outcomes.” Every study it covers “presented AAC displays on computer screens.”
Limit: a review of display features, not a test of any product.
Light, Wilkinson et al. 2019 Read in full
A board only works if the people around the child use it too, at school and at home.
Clinicians rated the adults as the top reasons devices are dropped: partners who think they understand without the device, families not trained, partners who do not model. Adults modeling on the device is “highly effective.”
EZbot, on the tablet. Tap “Show me” and EZbot builds a sentence step by step, each step glowing green, with sound or quietly, so a parent sees how the board is used. The parent chooses how much EZbot helps the child: Off, Glow only, or Step-by-step. A tap on EZbot also brings the words of the room and link the child is in, with the verbs beside them.
Clinicians rated partners believing they understand the person without the device (77.75% of cases), “The family is not trained to operate and/or program the system” (64.52%) and “Communication partners do not model system for the user” (63.78%).
Limit: clinicians’ estimates, not measured causes.
Johnson et al. 2006, Table 4, p. 92 Read in full
Aided AAC input was “highly effective across participants of various ages, disabilities, and language skills,” across 26 single-case designs (88 participants) and 2 group designs (103).
Limit: the models were people, usually combined with other strategies.
O’Neill et al. 2018 Official abstract
Packages that include partner modeling led to “meaningful linguistic gains” in communication turns, vocabulary, multi-symbol turns and word endings, across 9 single-case studies (31 participants) and a group study (63).
Limit: the modelers were communication partners, not software.
Sennott et al. 2016 Official abstract
A state-of-the-science review covers AAC technologies that “(e) teach interaction strategies to communication partners.”
Limit: what the review reports on these technologies is in the body, not yet read.
Light, McNaughton & Caron 2019 Official abstract
A family lives with a device for years. We looked for studies that follow children that long.
No study found follows today’s AAC devices or apps for 3 years or more. The best modern trial stopped at 9 months, and gains in one trial were gone 10 months later. Clinicians estimated only 39.35% of clients succeed beyond a year, and devices are dropped most in the first year and after 5 years.
A board that grows with the child, on the tablet. EZbot counts the words a child uses without help and proposes the next level; an adult approves it, and nothing already on the board moves. Home rooms count toward growing; school subjects and the Sentence Builder count toward grades. This measures progress inside the app; it is not a study over years.
The Cochrane review found only 2 randomized trials (154 children aged 32 months to 11 years), rated the evidence “very low” quality, and in the PECS trial “gains were not maintained 10 months later.” The longest outcome study found ran 2 years with 13 boys and young men aged about 6 to 20, and was published in 1996.
Limit: “no study found” means our searches; a university library search could still find one.
Brignell et al. 2018, Cochrane Read in full · Romski & Sevcik 1996 Official record · ages from Barker et al. 2019 Read in full
61 minimally verbal autistic children aged 5 to 8: “The study duration was 9 months (36 weeks).”
Limit: the device was added to a 2-hour-a-week communication program, so it was not tested alone.
Almirall, Kasari et al. 2016 Read in full
In a survey of AAC clinicians, 28.68% of clients were estimated to have abandoned AAC.
Limit: clinicians’ estimates from 2006, all ages, not measured outcomes.
Johnson et al. 2006, p. 90 Read in full
29.3% of devices were “completely abandoned.” Abandonment rates “were highest during the first year and after 5 years of use.”
Limit: a 1993 survey of 227 adults across all assistive technology, not AAC or children.
Phillips & Zhao 1993 Official abstract
People ask: there are already products, so why is this any different? We looked at what changed once AAC moved onto phones and tablets.
AAC came to the iPhone in April 2009. Four years later, leading researchers wrote that most AAC apps were not research-based and simply replicated the older systems, and passed on a warning about developers following the status quo.
We started from the research and from Evan, not from an existing board. myEZvoice is not a tablet board made smaller for a phone, and EZspeak is not a grid of small buttons.
“Most of the current AAC apps available for the iPad and other mobile technologies are not research-based; many simply replicate the approaches to representation and organization of vocabulary items seen with traditional AAC systems” (p. 113). Developers “may be tempted to follow the status quo, simply reiterating existing features in apps, rather than forging new ground” (p. 112).
Limit: describes apps as of 2013; it does not measure what has changed since.
McNaughton & Light 2013, pp. 112 and 113 Read in full
Proloquo2Go was first released on the App Store on 21 April 2009, for iPhone and iPod touch, before the iPad existed.
Limit: a first release date, not a measure of design.
Apple App Store record Official record
To know how many children and teenagers this is for.
Identified autism has risen to 1 in 31 eight-year-olds, and about 30% of autistic children are estimated to stay minimally verbal.
Both products are for nonverbal and minimally verbal children, teenagers and young adults: EZspeak for ages 4 and up, myEZvoice for ages 12 and up.
CDC estimates for 8-year-olds: about 1 in 150 for 2000 data, 1 in 31 for 2022 data.
Limit: a new case definition began with 2018 data, and the sites are not nationally representative.
CDC MMWR, 17 April 2025 Read in full
“It is currently estimated that about 30% of children with autism spectrum disorder remain minimally verbal, even after receiving years of interventions and a range of educational opportunities.” A Swedish study of 165 children recruited at ages 4 to 6 found 15% nonverbal and 10% minimally verbal 2 years later, after intervention.
Limit: definitions of minimally verbal vary across studies.
Tager-Flusberg & Kasari 2013 Read in full · Norrelgen et al. 2015 Official abstract
Mexican, Puerto Rican, Dominican and Cuban Spanish use different words for the same thing. We wanted to know whether the AAC in use today lets a child speak the Spanish of their own family.
Of 18 AAC products checked, 9 confirm Spanish. The Spanish they name is Spain, US, Latin American or Mexican. None offers Puerto Rican, Dominican or Cuban Spanish, though voices for all three exist. Research with Spanish-speaking families matches the vocabulary to the family’s dialect by hand.
Spanish in both products. On the tablet, Spanish words and the dialect setting for Mexican, Puerto Rican, Dominican and Cuban Spanish are live. On the phone, Cuban and Mexican Spanish are in testing in the newest build (13 Oct 2026), with Latin American, Puerto Rican, Dominican and Spain Spanish to follow. Recorded Cuban and Mexican voices are in progress, and every Spanish line is reviewed by native speakers before it is final.
LAMP Words for Life lists “English (US), German, Spanish (Mexico), and French (Canada) vocabulary files.” TD Snap Core First offers “Spanish (LA/MX)”. Proloquo2Go lets you “select a language and region”; its Spanish has been described as Castilian and North American.
Limit: a “Spanish (Mexico)” vocabulary does not show which accent the voice has.
LAMP Words for Life, App Store Official record · Proloquo2Go, App Store Official record · Tobii Dynavox TD Snap page Official record
18 product families checked, 9 with Spanish confirmed from the maker or the App Store. The varieties named: Spain or Castilian, US or North American, Latin American, Mexican.
Limit: “not found” means our searches; some makers’ sites could not be opened, and only the US App Store was checked.
Our searches, 12 Oct 2026 (brain: research_notes/Spanish AAC dialects)
Microsoft’s speech service lists Spanish (Puerto Rico), Spanish (Dominican Republic) and Spanish (Cuba), two voices each.
Limit: adult voices; the only Spanish child voice listed is Mexican.
Microsoft speech service language list Official record
Children used “iPad minis with the Proloquo2Go communication app,” and “vocabulary matched the family’s dialect and vocabulary preferences.” The families came from Mexico, Cuba and Venezuela.
Limit: a small study; the matching was done by the researchers, not built into the app.
Pak et al. 2025 Read in full
“The options were Mexican or Spain Spanish, and none of them sounded like anyone he knew.” A respondent: “Las voces en español, son de España y no latinas” (the Spanish voices are from Spain, not Latin America).
Limit: a preprint, not yet peer reviewed.
Weinberg et al. 2026 Read in full
“The roughly 40 million people of Mexican origin in the U.S. represented 57% of the nation’s Hispanic population in 2024. At 6.1 million, people of Puerto Rican origin are the next largest group.” Cubans, Salvadorans, Dominicans, Guatemalans, Colombians, Hondurans, Venezuelans and Ecuadorians each number 1 million or more.
Limit: origin is self-described, and not every Hispanic family speaks Spanish at home (68% did in 2024).
Pew Research Center, 22 Oct 2025 Read in full
How far each item was checked: Read in full the whole paper or record was read. Official abstract checked against the official abstract (PubMed, Crossref or ERIC). Official record checked against the official catalogue or store record.
Checked 12 October 2026. “No study found” means our searches did not turn one up; it does not prove none exists. Product facts are the products’ own, as they run today. Product names are their owners’ trademarks.
The research