Is Educational Therapy Evidence-Based? What the Record Actually Shows
A referring psychologist asks whether educational therapy is evidence-based. A parent forwards a school's email asking the same thing. You know the work helps. You have the progress graphs. And you still find the question hard to answer in a sentence, which is uncomfortable, because it sounds like it ought to have a one-word answer.
It does not. The question "is educational therapy evidence based" contains at least two separate questions and runs into six different published definitions of the term, and the honest answer depends on which one the person asking has in mind.
Here is the short version, and the rest of this article is the evidence for it. The methods educational therapists use are among the best-evidenced things in education. The service model they are delivered in has not been studied. Those two sentences are both true, they are not in tension, and being able to say both of them plainly is a professional asset rather than an admission.
The defensible claim is that educational therapists deliver methods with strong evidence behind them, in a service model that has not itself been directly tested. The indefensible claim is any sentence that transfers an effect size from the method to the model.
Nothing in this article is legal advice, and nothing in it describes what any law requires of you. Where federal statute is quoted below, it is quoted because it defines a word that gets used about your work, not because it governs your practice.
What the clearinghouses actually hold
Start with the gatekeeper everyone names. The What Works Clearinghouse was created in 2002, in its own words, "to be a central and trusted source of scientific evidence for what works in education." Its intervention reports "look at all of the studies for a particular intervention, review them against evidence standards, and then summarize findings to provide educators with the information they need to make evidence-based decisions."
A keyword search of the full WWC publication corpus on 19 September 2026 returned 649 publications in total. A search of that same corpus for "educational therapy" returned zero. So did "educational therapist." So did "educational therapists."
The search works. "Reading" returns 280 publications. "Tutoring" returns 29. "Dyslexia" returns five, all of them branded reading programs rather than service models: three reports on the Barton Reading and Spelling System, one on the Dyslexia Training Program, and one titled "Unbranded Orton-Gillingham-based Interventions." The bare word "therapy" returns two, and neither is ours: Pivotal Response Training and the Lovaas Model of Applied Behavior Analysis.
Bound that finding precisely, because the precise version is the defensible one. There is no WWC intervention report and no WWC practice guide on educational therapy as a service model. That is not the same as saying nothing in the WWC touches your work, and it is not the same as saying the WWC has reviewed educational therapy and found it wanting. The WWC has not reviewed it at all.
The distinction matters more than it sounds. A clearinghouse rating of "no discernible effects" is a finding. An empty search result is not a finding about educational therapy. It is a finding about what has been submitted, funded and published.
Why the searches come up empty, which is not the reason most people assume
The usual explanation is that the field is young. That is not quite it, and the real explanation is more interesting.
"Educational therapy" names at least four different things
Search ERIC, the federal education research database, for the exact phrase "educational therapy" and you get roughly 340 records. That sounds healthy until you look at them.
Of the records carrying ERIC's own "Educational Therapy" subject descriptor, about three quarters were published before 1990. The term was in heavy use in the 1970s and 1980s and has been in decline ever since. And it was in use for something else.
The British psychodynamic sense. Veerman's 1984 paper describes "developments in educational therapy at the Tavistock Clinic in London," using approaches that draw on "principles of psychotherapy (including countertransference) as well as remedial teaching."
A contemporary German profession. Schubert's 2023 chapter examines "the ongoing process of academization within the field of educational therapy in Germany, particularly in the context of the introduction of university degrees in integrative educational therapy and the psychology of learning." That is a real, distinct, state-recognised occupation and it is not the American one.
The American definitional literature, which is descriptive rather than empirical. Radecki's 1984 article in the Journal of Learning Disabilities defines educational therapy as "an approach to alleviating a wide range of educationally related problems through consideration of the four foci," and is classified by ERIC as Reports - Descriptive.
A trademarked commercial program, which is the one that will trip you up. More on that below.
So the honest sentence is not "there is no research on educational therapy." There are hundreds of records. The honest sentence is that the phrase is ambiguous across decades and continents, and that the American service model specifically has not been the subject of a controlled trial indexed in ERIC.
The field's own literature is invisible to the literature
Here is the part that explains the empty searches better than anything else.
The Association of Educational Therapists publishes a journal. Its public page states that AET "publishes The Educational Therapist twice each year in the spring and fall," and that "this peer-reviewed, professional journal is part of the EBSCO research database used by academic libraries in many colleges and universities."
The journal is real, it has reached its forties in volume number, and its submission policy states plainly that "manuscripts will be peer-reviewed by editorial board members." Note the mechanism rather than the label: that is in-house review by the association's own editorial board, which is a different thing from independent external peer review, and worth describing accurately when someone asks.
More consequentially, the journal is not discoverable. A search of ERIC's source field for "The Educational Therapist" returns no records. A Crossref journal search returns two journals with similar names, both belonging to other organisations in Korea and Singapore, and not AET's. No DOIs, therefore no presence in the citation graph.
A search of ERIC for the exact phrase "Association of Educational Therapists" returns zero records.
That single fact does most of the explanatory work. Whatever case studies, practice papers and outcome observations the field has accumulated over four decades sit in a members-only journal that the standard research databases do not index. They are not absent because nobody wrote them. They are absent because nobody outside the membership can find them.
The one randomised trial with the words in its title is not about you
If you search hard enough you will find a randomised controlled trial with "Educational Therapy" in the title, and somebody will eventually send it to you as proof. Read it before you forward it.
Stebbings and Kline (2020), in the International Journal for Research in Learning Disabilities, "compared the effectiveness of Orton Gillingham (OG) tutoring and National Institute for Learning Development Educational Therapy (NILD). A randomized controlled trial using 27 participants determined whether academic and cognitive outcomes differed between the groups over a 9-month period."
Two things about it.
First, NILD Educational Therapy is the registered, manualised program of the National Institute for Learning Development, a separate organisation from AET with its own training route and its own trademark. A trial of it is a trial of that program. It is not a trial of the practice regulated by AET's Code of Ethics and it does not transfer.
Second, and more to the point, the authors say so themselves. Their own conclusion is that the outcomes support "the efficacy of both approaches which currently lack robust empirical evidence." That is the authors of the one trial telling you that the thing their trial is about has a thin evidence base.
The other quantitative studies in ERIC using the phrase are also NILD studies, and all of them are unpublished doctoral dissertations rather than peer-reviewed articles: Kuschel (2022), which describes itself as "this quantitative, quasi-experimental study"; Entzminger (2013) on reading comprehension outcomes; and Frimpong (2014), which is "a qualitative program evaluation" conducted at a single school. They are legitimate work. They are not what a psychologist means by an evidence base.
Citing that trial as support for your practice is the exact error this article exists to help you avoid.
AET makes no efficacy claim anywhere, and that is more defensible than it sounds
This is the finding that surprised me most, and it holds up.
AET's Fact Sheet is the field's most-quoted single document. It defines the role: "An Educational Therapist is a highly trained professional who works in the academic domain with neurotypical and neurodiverse children, adolescents, and adults who have been diagnosed with or exhibit learning disabilities, exhibit signs of learning differences, or demonstrate difficulties with learning." It sets scope, including the DO NOT list that every practitioner knows.
Checked on 19 September 2026, the text of that page contains not one instance of the words evidence, research, efficacy, proven, effectiveness or outcomes.
The pattern holds across the public site. The parent-facing page says educational therapists "help children and adults fulfill their learning and career potential" by "developing personalized intervention plans and teaching strategies." That is an aspiration, not an outcome claim. There is no studies-show sentence, no effect size and no citation anywhere on it. And there is no research or evidence page on the public site at all.
The history timeline is the quiet confirmation. It records educational therapy being "written into No Child Left Behind" and "included in new regulations for Reauthorization of IDEA." Those are milestones of recognition. Across the whole timeline there is no study, no trial, no evaluation and no evidence review. The profession's own account of its history is a history of institutionalisation, not of evidence accumulation.
You can read that two ways, and the charitable reading is the accurate one: an organisation that has never claimed an evidence base has never overclaimed one either. It is consistent with the scope-first posture we mapped in the practitioner scope map, where AET's definitional documents turn out to describe boundaries far more precisely than they describe results.
Compare that with the credential marketplace covered in our executive function coach certification piece, where the claims outrun the documentation regularly. AET's restraint is a form of honesty, and you can say so out loud.
The training programs make a narrower claim than it first appears
CSUN's MA in Educational Therapy page carries a heading reading "Foundation of Knowledge and Research," under which it says: "At CSUN, we've built our programs on knowledge derived from sound theories and rigorous research in partnership with team members from the Association for Educational Therapists."
Read the grammatical object. The claim is that the training program was built on research. It is not a claim that educational therapy has been shown to work. The same page describes the capstone as an "MA project that focuses on implementing a specific research-based practice and evaluating the impact," which is precisely the right distinction: the components are research-based, and students are taught to apply them.
The CSUN Advanced Study certificate page makes no evidence claim at all. Neither does Notre Dame de Namur's MA page, whose pitch is scarcity and career rather than efficacy. Both appear on AET's approved training programs list, and the list itself is organised around course content and membership eligibility rather than outcomes.
Neither AET-approved university claims an evidence base for educational therapy as an intervention. That is worth knowing before you make one on their behalf.
"Evidence-based" is six different published standards, and they disagree
This is the heart of it. When someone asks whether your work is evidence-based, they are using a term that at least six bodies define differently in writing. Knowing which one they mean is most of the answer.
1. The federal statute, which is looser than almost anyone assumes
"Evidence-based" has a statutory definition, at 20 U.S.C. 7801(21), the definitions section of the Elementary and Secondary Education Act as amended by ESSA. It means an activity, strategy or intervention that
demonstrates a statistically significant effect on improving student outcomes or other relevant outcomes based on (I) strong evidence from at least 1 well-designed and well-implemented experimental study; (II) moderate evidence from at least 1 well-designed and well-implemented quasi-experimental study; or (III) promising evidence from at least 1 well-designed and well-implemented correlational study with statistical controls for selection bias
or, in a fourth category, "demonstrates a rationale based on high-quality research findings or positive evaluation that such activity, strategy, or intervention is likely to improve student outcomes," combined with "ongoing efforts to examine the effects."
Three things practitioners routinely get wrong about this text.
"At least 1" means one. The statute sets no replication requirement, no minimum sample size and no minimum effect size. A single qualifying study is legally sufficient for the top tier.
The statute never uses the word "tier." Tier 1, Tier 2, Tier 3 and "Demonstrates a Rationale" are administrative labels, not statutory ones. The fourth category has no name in law.
The statute contains two different definitions of its own term. For school-improvement funds, subparagraph (B) switches the fourth category off entirely, so "evidence-based" there means the first three only. Which definition applies depends on the funding stream.
This definition exists to decide what schools and districts may spend federal money on. It does not oblige, permit or forbid anything for an independent practitioner, and a private-practice question about what any statute requires belongs with an attorney rather than with this article or with a clearinghouse.
2. The Department's own guidance, which is stricter than the statute
The Department of Education's Non-Regulatory Guidance: Using Evidence to Strengthen Education Investments, dated 16 September 2016, adds conditions the statute does not contain.
For strong evidence, the study "should" also "have a large sample and a multi-site sample" and "have a sample that overlaps with the populations (i.e., the types of students served) AND settings (e.g., rural, urban) proposed to receive the intervention," and must not be "overridden by statistically significant and negative" evidence elsewhere.
The guidance also renames the fourth category "Demonstrates a Rationale" and converts the statute's "rationale based on high-quality research findings" into a requirement for "a well-specified logic model."
So a vendor claiming ESSA Tier 1 may be claiming one RCT of any size at one site, or may be claiming the Department's considerably stricter version, and the label does not tell you which. That is worth knowing whenever a program's marketing lands on your desk, which is the same caution we applied to the reading programs and to the "Orton-Gillingham based" label.
3. The What Works Clearinghouse, which does not rate programs at all
The current standard is the WWC Procedures and Standards Handbook, Version 5.0, WWC 2022008, published August 2022 and revised December 2022. It is the current version, confirmed against the handbooks index.
The familiar vocabulary survives, and it has shifted in a way that matters. Version 5.0 issues "a research rating based on the strength of the research design and its execution" and separately "an effectiveness rating based on the evidence of favorable effects from the intervention." And the research rating attaches to findings:
Based on the WWC's confidence in the strength of the research design and execution of the design, eligible findings from the study will receive one of three research ratings: Meets WWC Standards Without Reservations, Meets WWC Standards With Reservations, or Does Not Meet WWC Standards.
Eligible findings. Not studies, and certainly not programs. A single study can contain findings with different ratings in different outcome domains. "This program meets WWC standards" is not a well-formed sentence, and when you see it in marketing copy you are entitled to ask which finding, in which domain.
Note also what the research rating is about. All three descriptions concern confidence that the intervention caused the observed effect. None of them says the effect was positive, large or educationally meaningful. That is the separate effectiveness rating, which has five categories to the statute's four, including "uncertain effects" for non-significant results and "negative effects" for significant harm.
And the WWC's own Tier 1 is stricter than the statute it borrows the name from. Its criteria require that the analytic sample "includes multiple sites" and "includes 350+ unique individuals." The statute requires neither.
4. NCII, which rates studies and takes voluntary submissions
The National Center on Intensive Intervention's tools charts are the ones practitioners actually open. Their own "About the chart" panel is unusually blunt:
The chart includes information and ratings on the quality of research studies about different interventions, not ratings on the interventions themselves. The ratings are about study quality only. ... All tools that meet the criteria for review are posted on the chart, regardless of results. The chart represents all tools that were reviewed, not those that were "approved."
Two consequences follow, and the second is the one people miss. NCII's tools charts overview states that "products are rated against established criteria and not compared to each other or ranked," and that "the submission process is voluntary." Developer-submitted, voluntarily, and rated against criteria rather than each other.
So absence from the chart means nothing at all. It tells you a developer did not submit, which may reflect cost, capacity or indifference rather than quality. We made the same point about the progress monitoring systems in our comparison of the four, and it is the single most misread feature of the charts.
NCII's rating vocabulary is a fourth system again: its current user guide describes ratings of "convincing evidence (full bubble), partially convincing evidence (half bubble), unconvincing evidence (empty bubble), or data unavailable (dashed line)." Those words map onto neither ESSA's tiers nor WWC's ratings.
5. IDA, which defines the term rather than testing against it
There is a fifth usage, and it is the one closest to daily practice. The International Dyslexia Association's Structured Literacy Defined fact sheet, dated July 2024, describes Structured Literacy as "a comprehensive, integrated approach to reading and writing instruction that directly and systematically builds on and develops oral language while explicitly teaching the structure of written language," and states that educators using "this evidence-based approach" integrate foundational and higher-order skills.
Read the fact sheet's own framing of what it is doing: "In this second sentence, we define Structured Literacy as evidence-based." That is definitional work, done openly. The warrant offered is that Structured Literacy "is built on and grounded in the knowledge gained from decades of research into a number of topics related to literacy and its acquisition."
That is a claim about a converging body of research on how reading works, not a claim of the kind ESSA's definition is built to test. Structured Literacy is an approach rather than an "activity, strategy, or intervention," there is no single trial the claim rests on, and neither the WWC nor NCII rates approaches.
The claim may be entirely defensible on its own terms. It is simply a different kind of claim, and IDA is more transparent about that than most. The same distinction runs through IDA's Knowledge and Practice Standards, which describe what a teacher should know rather than what an outcome study found.
6. The clinical definition, which is about you rather than about a product
The definition a referring psychologist most likely has in mind is not any of the above. The APA Presidential Task Force on Evidence-Based Practice, writing in American Psychologist in 2006, defined it this way:
Evidence-based practice in psychology (EBPP) is the integration of the best available research with clinical expertise in the context of patient characteristics, culture, and preferences.
The Task Force notes that this "closely parallels the definition of evidence-based practice adopted by the Institute of Medicine (2001)," which reads: "Evidence-based practice is the integration of best research evidence with clinical expertise and patient values."
ASHA states the same triad for speech-language pathology, calling evidence-based practice "the process of applying current, best evidence (external and internal scientific evidence), patient perspective, and clinical expertise to make decisions about the care of the individuals you treat." Its practice portal names the three components as clinical expertise, internal and external evidence, and client perspectives.
Note ASHA's phrase "internal ... evidence." That category explicitly admits the clinician's own client data. The statutory definition has no category for it whatsoever.
The two definitions do not merely set different thresholds. They take different grammatical objects. Under ESSA, a program can be evidence-based while sitting unopened in a cupboard: the property belongs to the thing. Under the APA and ASHA formulations, nothing is evidence-based until a clinician does something with a particular client: the property belongs to the decision.
One honest caveat, which you should offer before anyone else raises it. APA's own footnote limits the Task Force's scope to "health services provided by psychologists" and states that "other organizational, community, or educational applications of evidence-based practice by psychologists are outside the scope of this report."
Practitioners who import the three-legged definition into educational therapy are extending it past its stated boundary. That may be entirely reasonable. It should be said out loud rather than assumed.
What is actually evidenced: the components, each with its boundary
Now the constructive half. The methods an educational therapist uses are not short of evidence. Several carry the highest rating the federal clearinghouse issues. What follows states each finding with its boundary in the same breath, because a finding without its boundary is how a defensible claim becomes an indefensible one.
Mathematics intervention: the closest population match in the literature
Assisting Students Struggling with Mathematics: Intervention in the Elementary Grades, WWC 2021006, March 2021, rates all six of its recommendations at the Strong level: systematic instruction, mathematical language, representations, number lines, word problems and timed activities.
Its stated audience and population are the reason this guide matters most to you. It is written for "teachers providing mathematics intervention to students who are struggling," including "special educators, mathematics general education teachers, mathematics specialists, and mathematics coaches," working with "students with or at risk for mathematics disabilities who may also have reading, language, attention, behavior, working memory, or processing-speed difficulties."
That population description is, near enough, a description of an educational therapy caseload. Boundary: the delivery setting throughout is school-based intervention, not private practice, and the recommendations attach to instructional content rather than to who delivers it or where. We built the dyscalculia intervention piece on this guide for exactly that reason.
Reading intervention in grades 4 to 9: the closest provider match
Providing Reading Interventions for Students in Grades 4-9, WWC 2022007, March 2022, rates three of its four recommendations Strong: build decoding of multisyllabic words, provide purposeful fluency-building activities, and routinely use comprehension-building practices. The fourth, on stretch text, is Moderate.
Its named audience is the closest any federal document comes to naming you: "special educators, intervention teachers, reading specialists, reading coaches, and trained volunteers." Boundary: still school-based, still within a multi-tiered system of support, and an educational therapist is not on the list. "Reading specialist" is adjacent, not equivalent.
Foundational reading in K to 3: read the tier on each recommendation, not the guide
Foundational Skills to Support Reading for Understanding in Kindergarten Through 3rd Grade, NCEE 2016-4008, July 2016 and revised December 2019, is the guide most often cited loosely. Read the guide itself rather than a summary of it, because its individual ratings differ sharply:
Develop awareness of the segments of sounds in speech and how they link to letters: Strong
Teach students to decode words, analyze word parts, and write and recognize words: Strong
Ensure that each student reads connected text every day: Moderate
Teach academic language skills, including inferential and narrative language, and vocabulary knowledge: Minimal
Academic language and vocabulary work, a staple of educational therapy sessions, carries the weakest rating in the guide. Do not round it up.
Boundary, and it is a sharp one: the guide's study eligibility criteria required that "at least 50 percent of the sample must be general education and native English speakers." The evidence base was deliberately weighted toward general education students, which is not the population you see.
The tier labels are not constant across guides
This is worth one short section of its own, because it undercuts a lot of casual citation.
The 2021 mathematics guide defines a strong rating as "consistent evidence that the recommended strategies, programs, or practices improve relevant outcomes for a diverse population of students." The 2022 reading guide defines the same word as evidence from "two or more well-designed, well-implemented experimental studies."
Two studies. That is the explicit floor in the more recent of the two documents. A Strong rating is a real signal and it is not a claim of overwhelming evidence, and the 2021 guide says as much from the other direction: "a minimal evidence rating does not indicate that the panel views the recommendation as any less important than other recommendations with a strong or moderate evidence rating."
Writing: where the marketing and the clearinghouse disagree most
Self-Regulated Strategy Development is the best-evidenced thing an educational therapist does for writing, and it is also the clearest example of why the boundary has to travel with the number.
Writing Next, Graham and Perin's 2007 report to Carnegie Corporation of New York, reports that "the average weighted effect size for SRSD studies (1.14) was larger than for non-SRSD studies (0.62)." Graham, McKeown, Kiuhara and Harris (2012) report in the Journal of Educational Psychology that "the self-regulated strategy development model (ES = 1.17)" improved how well students wrote.
Now the two qualifications that almost never travel with those figures.
First, the same 2012 abstract reports strategy instruction at 1.02 and "adding self-regulation to strategy instruction (ES = 0.50)" — the incremental contribution of the self-regulation component is half the size of strategy instruction alone. And Graham and Perin's own caution in Writing Next is that "many of the effect sizes differ only minimally, so readers should be cautious in interpreting the differences in effect strength."
Second, and decisively for your caseload: the WWC reviewed SRSD specifically for students with a specific learning disability in its November 2017 intervention report, and found ten studies meeting its pilot single-case design standards.
In its own words: "No studies meet WWC group design standards. No studies meet pilot single-case design standards without reservations." Together those studies "included 43 children ages 7 to 16 who had a specific learning disability," and the resulting rating was "potentially positive effects."
So SRSD is simultaneously the strongest thing in your writing toolkit and much less firmly established for your actual population than the headline numbers suggest. Both halves belong in the same sentence. We took the same care with the instrument claims in the dysgraphia and writing intervention piece.
One further caution. The SRSD vendor site states that effect sizes range "from 1.0 to 2.55" and "exceed 0.85 in meta-analyses." The 1.17 figure checks out against Graham et al.
The 2.55 could not be traced to any source, and the 0.85 floor sits awkwardly against that same paper's 0.50 for the self-regulation component. Cite the meta-analyses and the WWC report directly. Do not cite a program's own research page for its own effect sizes.
The claim you cannot make, and it is the one everybody makes
Here is where this article stops being reassuring.
The proposition that one-to-one is the most effective instructional format is not supported by the primary literature. It was tested directly, it was expected to win, and it did not.
The National Reading Panel report, NIH Publication 00-4769, published April 2000, examined delivery format inside its phonics meta-analysis. Its subgroup report is worth quoting at length, because the paraphrase always softens it. Across the eight treatments in which students received one-to-one tutoring, the panel wrote:
This was expected to be the most effective form of phonics instruction ... Tutoring produced an effect size of d = 0.57 which was greater than the effect size for small groups, d = 0.43, and for classrooms, d = 0.39. However, none of these effects differed statistically from each other. This evidence falls short in supporting the expectation that tutoring would prove especially effective for teaching phonics.
And the panel pressed on its own result rather than resting on it: the control conditions "did not involve tutoring in the majority (62%) of the cases," which "should have given tutoring an extra advantage. However, it did not."
In the phonemic awareness meta-analysis the comparison ran the other way. Small-group instruction produced effect sizes of 0.81 and 0.83 on reading outcomes, while "effect sizes for children taught individually or in classrooms ranged from d = 0.30 to 0.45." The phonics chapter's summary of that comparison is flat: "small groups were significantly more effective than tutoring or classrooms."
The modern tutoring meta-analysis agrees. Nickow, Oreopoulos and Quan found that for teacher and paraprofessional tutoring, one-to-one showed the largest effect sizes "with ratios above one-to-one statistically similar."
None of this says one-to-one does not work. It says the specific claim that individual delivery outperforms small-group delivery is not what the primary sources found, and that a practitioner who asserts it in a referral conversation is asserting something a psychologist can check in ten minutes.
There is a second, quieter finding in the same report that belongs in an honest account. Systematic phonics "exerted a lower than expected impact on reading growth in low achieving readers (d = 0.15) and disabled readers (d = 0.32)," against a headline of 0.41 to 0.44 overall.
That is on only eight studies of low achievers, and the panel itself noted that "further research is needed to explore how best to remediate their reading difficulties."
Worth adding: the panel's own benchmark, stated in the same report, is that "an effect size of 0.20 is considered small; a moderate effect size is 0.50; an effect size of 0.80 or above is large." By its own scale, its headline phonics finding is below moderate.
That report is now twenty-six years old, which is a limitation worth stating whenever you cite it, and it remains the most rigorous head-to-head test of delivery format in the corpus.
The tutoring literature, which is the closest thing to a model-level evidence base
If anything speaks to the service model rather than the method, it is the tutoring research. It is genuinely encouraging and it does not transfer cleanly, and both halves matter.
Cite the right version. The figure in circulation, 0.37 SD, comes from NBER Working Paper 27476, July 2020, titled "The Impressive Effects of Tutoring on PreK-12 Learning," which carries NBER's standard notice that working papers "have not been peer-reviewed."
The same three authors published the peer-reviewed version in the American Educational Research Journal in 2024 under a notably more restrained title, "The Promise of Tutoring for PreK-12 Learning", doi 10.3102/00028312231208687, reporting "an overall pooled ES of 0.288 SD (SE = 0.029, p < 0.001)." If you quote 0.37 in 2026 you are quoting a superseded working paper.
The published abstract also specifies where the effects concentrate: "ESs tend to be largest for programs that use teachers or paraprofessionals as tutors, are held in earlier grades, occur at least 3 days per week, and are held during school."
Read that list against a private practice and all four conditions are awkward. The authors' own provider categories run teacher, paraprofessional, nonprofessional and parent, and no category corresponds to an independent credentialed clinician.
The provider closest to you, the certified teacher, leads the table partly for a specific reason the authors name: "the advantages of teacher tutoring over paraprofessional tutoring are driven primarily by first-grade interventions," which are "in turn attributable in large part to studies evaluating Reading Recovery."
On frequency the authors are direct: "There is little evidence of once-weekly tutoring sessions generating large effect sizes." State that carefully. It is an absence of evidence within a corpus of school-based programs, not a demonstration that weekly sessions fail, and collapsing the two is the single easiest way to say something the record does not support.
We worked through what published program specifications actually demand, and what a weekly private caseload can deliver against them, in the session frequency and duration piece.
There is also a newer and larger analysis that a 2026 conversation should not omit. Kraft, Schueler and Falken published "What Impacts Should We Expect from Tutoring at Scale? Exploring Meta-Analytic Generalizability" as a working paper and subsequently in Review of Educational Research (doi 10.3102/00346543261446660).
They expanded the corpus substantially and found that "pooled effect sizes from studies with stronger target-equivalence remain meaningful but are only a third to a half as large as those from our full sample," driven by "stark declines in pooled effect sizes as program scale increases."
It would be tempting to read that as good news for small individualised practice. Resist it, because the authors themselves supply the counter-argument: "It is possible, however, that the common practice to use the within-sample standard deviation to estimate effect sizes causes these estimates to be artificially larger for studies of smaller programs serving more targeted and homogenous populations."
A small, homogeneous, carefully-selected caseload is the exact shape they are warning about. Whichever version you cite, cite one and do not blend the working paper's figures with the published ones.
Finally, the definition that is quietly disqualifying. Stanford's National Student Support Accelerator defines high-impact tutoring as instruction with "regular sessions (3+ times per week for 10+ weeks)," "consistent, well-trained tutors," "data-driven instruction," "school-day implementation," and "vetted, structured instructional materials that complement core classroom instruction."
Two of those criteria, school-day implementation and alignment with core classroom instruction, cannot be met by an independent practitioner at all. An educational therapist cannot claim to be delivering high-impact tutoring as NSSA defines it, and should not try.
So what can you actually say?
Here is the practical payoff, in the form of sentences you can use and sentences you should retire.
Defensible, and worth saying confidently:
"The instructional methods I use carry Strong ratings from the Institute of Education Sciences, and I can name the guide and the recommendation."
"Systematic, explicit decoding instruction has the strongest evidence in the field, and it is what I do."
"For a student with a mathematics disability alongside working memory or attention difficulties, the 2021 IES practice guide describes this population directly and rates all six of its recommendations Strong."
"Educational therapy as a delivery model has not been the subject of a controlled trial, and I would not claim otherwise. What I can show you is this client's baseline, this monitoring schedule and this trend line."
"I practise evidence-based in the clinical sense the APA and ASHA use: best available research, my own training and judgement, and this family's priorities, applied to this case."
Retire these:
"Educational therapy is evidence-based." Unqualified, it is a claim about the model, and it is not supported.
"One-to-one is the most effective way to deliver intervention." Directly contradicted by the National Reading Panel.
"Research shows tutoring produces 3 to 15 months of additional learning." This figure is asserted on NSSA's pages without a citation and could not be traced to a named study.
"This program is WWC-approved" or "meets WWC standards." The WWC rates findings, does not approve, and does not endorse.
"It is on the NCII chart, so it is validated." The chart rates study quality, submission is voluntary, and presence is not endorsement.
The pattern here is the same one that runs through the whole credential and scope landscape: the field's vocabulary is shared with bodies that define it more narrowly than practitioners do, and the practitioner who knows the narrower definition is the one who sounds authoritative rather than the one who sounds defensive.
Building your own evidence, which is the part within your control
The clinical definition of evidence-based practice has a component the statutory one does not: your own data. ASHA calls it internal evidence. It is the leg of the stool you can actually strengthen this month.
NCII's framing is the useful one here. Data-based individualization is described as "a research-based process for individualizing and intensifying interventions through the systematic use of assessment data, validated interventions, and research-based adaptation strategies," and NCII states flatly that "DBI is a process, not a specific program or product."
Notice the hedge in that sentence. NCII, a federally funded centre entirely fluent in ESSA's vocabulary, calls its own flagship method "research-based" rather than "evidence-based." That looks deliberate, because a process cannot satisfy a statutory definition written around an activity, strategy or intervention.
It is the closest thing in the literature to a gatekeeper conceding that its own method does not fit the statute while still claiming a research warrant. If that is good enough for NCII, a similar formulation is good enough for your intake paperwork.
Practically, that means four things, none of which requires anyone's permission:
Choose components you can name and cite. If a method appears in an IES practice guide, put the guide and the recommendation number in the treatment plan rather than a generic claim.
Monitor with an instrument whose decision rules are published, and follow them. The four systems and their four different rules are laid out in our progress monitoring comparison.
Document dosage against the published specification, including where you cannot meet it and what you did instead. That is a fidelity record, and it is exactly what the dosage piece sets out. NCII's Taxonomy of Intervention Intensity gives you the seven dimensions to describe it in, and dosage is only one of them.
Keep the assessment boundary clean. Administering is not interpreting, and the scope limits in AET's Code of Ethics and in the qualification levels publishers set are unaffected by any of this. IRIS's DBI module is the free training route if you want the process formally.
None of that makes educational therapy evidence-based in the ESSA sense. All of it makes your individual practice defensible in the clinical sense, which is the sense the referring professional almost certainly means.
Where the published record runs out
Every article on this site ends here, and this one has more gaps than most. These are the things we could not verify, stated as gaps rather than filled in.
AET's EBSCO indexing. AET states its journal "is part of the EBSCO research database." EBSCO's title lists are behind subscription and we could not confirm it independently. Treat it as AET's own claim, attributed.
The WWC's single-study review database. Our zero-result finding covers the 649 intervention reports and practice guides. The separate Reviews of Individual Studies database returned server errors on repeated attempts and could not be searched, so the claim in this article is bounded to intervention reports and practice guides, not to everything the WWC holds.
The "3 to 15 months" tutoring figure. Asserted twice on Stanford NSSA pages, including its research summary, with no citation on either. We could not trace it to a study and we do not repeat it as a finding.
The SRSD range of "1.0 to 2.55." Published on the program's own research page. The upper figure could not be traced to any source.
Whether AET's journal carries an ISSN. None appears on the journal's public page or in the issue we examined. That is an absence from two checks, not a confirmed absence.
NCII's older caveat language. Three widely-quoted sentences from NCII's May 2016 Tools Charts User's Guide are no longer at their published URL, which now returns an error; the document survives only in the Internet Archive. The live pages carry equivalent language and we have quoted the live versions. If you have those older sentences in a slide deck, update the citation.
There is no IES practice guide on tutoring, one-to-one instruction, or intensive intervention as a delivery format. We checked the full current list. Nobody has blessed the format, in either direction.
That last one is the fair summary of this whole article. The absence of evidence about the educational therapy model is not evidence that the model does not work. It is evidence that nobody has funded the study. Those are different sentences and a practitioner who keeps them apart will be trusted more, not less.
Frequently asked questions
Is educational therapy evidence-based?
Not in the sense the federal statute uses, because no controlled trial of the model has been published and the What Works Clearinghouse holds no intervention report or practice guide on it.
It can be practised in an evidence-based way in the clinical sense the APA and ASHA use, by integrating well-evidenced methods, professional judgement and the client's priorities. Which answer is right depends entirely on which definition the person asking is using.
Is there any research on educational therapy at all?
Yes, several hundred records in ERIC, but read them carefully. Most predate 1990 and many use the phrase for a British psychodynamic practice or a German academic profession rather than the American service model. A search of ERIC for "Association of Educational Therapists" returns no records at all, largely because the association's journal is not indexed there.
What about the randomised trial of educational therapy?
The only randomised trial with the phrase in its title is Stebbings and Kline (2020), which compared Orton-Gillingham tutoring with NILD Educational Therapy, a different organisation's trademarked program, with 27 participants. Its own authors write that both approaches "currently lack robust empirical evidence." It is not evidence for AET-model educational therapy.
Does AET claim educational therapy is evidence-based?
No. Checked in September 2026, AET's Fact Sheet contains no instance of the words evidence, research, efficacy, proven, effectiveness or outcomes, and the public site has no research or evidence page. The association defines scope, training and ethics rather than making efficacy claims.
Is one-to-one instruction more effective than small-group?
The primary literature does not show that. The National Reading Panel tested delivery format directly and found tutoring at d = 0.57 against small groups at 0.43 and classrooms at 0.39, with "none of these effects" differing statistically from each other, and found small groups significantly more effective than tutoring for phonemic awareness. Nickow and colleagues found ratios above one-to-one "statistically similar" for teacher and paraprofessional tutoring.
What can I tell a referring psychologist?
That the methods you use carry Strong ratings from named IES practice guides, which you can cite by recommendation; that the service model itself has not been studied and you will not claim it has; and that you monitor with a published instrument and follow its published decision rules, so the client's own data is available for review. That answer is both more honest and more impressive than an unqualified claim.
Where this leaves the profession
A field that publishes carefully, indexes its journal, and runs even one well-designed study of its own model would change this article considerably. Nothing prevents that except funding and habit. Until it happens, the practitioners who do best in these conversations are the ones who know exactly which definition is in play, can cite the guide and the recommendation for what they actually do, and say plainly where the record runs out.
If you want the rest of this series as it is published, and to hear about matched-client referrals through Illuminate, join the educator list.
Comments