Mastery Criteria: Where the Percentages in the Manuals Come From
Mastery Criteria: Where the Percentages in the Manuals Come From
You decide this several times a session, usually in about four seconds. The student got most of them right. Do you move on?
Every manualised program you might be running has already answered that for you, in a number. The mastery criteria in reading intervention manuals are specific, confidently printed, and almost never derived from anything. One program wants ninety-five percent accuracy.
One wants seventy percent on new material and ninety on old. One wants three or fewer errors and a perfect quiz. One tells you to advance when "the majority of the students have achieved mastery" and does not publish the number that defines it.
They cannot all be right, because they are not even measuring the same thing.
This is a survey of what the published documents actually specify, who specified it, and what happens when somebody treats the threshold as a question rather than a setting. It is not a recommendation about which number to use. Where the evidence bears on the choice at all, it is thinner than the confidence of the manuals would suggest, and it points in a direction most practitioners will find uncomfortable.
The short version
Every program publishes a mastery criterion and almost none publishes its derivation. Barton specifies about ninety-five percent accuracy. Engelmann's Direct Instruction programs specify seventy percent on first-time material and ninety on material taught earlier. Read Naturally specifies three or fewer errors plus a perfect quiz plus a rate goal. Wilson specifies "a specific mastery level" and does not state it on its public implementation pages.
The unit is not the same across programs, which means the thresholds are not comparable even when the percentages look alike. A percentage of items, a count of errors, a rate per minute and a number of consecutive sessions are four different claims about what learning is.
The federal practice guide for foundational reading skills uses the word "mastery" four times and never attaches a number to it. Its only numeric accuracy threshold is a definition of instructional-level text, written as a fraction in a sentence.
The word "mastery" does not appear anywhere on either of the two NCII tools charts — not on the instructional intervention chart, not on the progress monitoring chart. Every instance of "criterion" on those pages refers to criterion validity.
AET's Code of Ethics contains the word "mastery" zero times, and the words "criterion" and "criteria" zero times. It requires you to establish ways to measure progress at regular intervals and stops there.
Where the criterion has been treated as an experimental variable, it mattered — and the published record is two studies deep. One meta-analysis of mastery learning programs found criterion level significantly predicted effect size, and not in a straight line.
The one threshold in the chain that is genuinely derived is derived from a norm table, which makes it a statement about other students rather than about the skill.
Four programs, four thresholds, and at least three different units
Start with what the publishers themselves put in writing. This is the same four-axis discipline used for assessment instruments and for group size: report what the manual states, name the issuing body, and do not interpolate.
Barton: about ninety-five percent, and the word "true"
The Barton Reading and Spelling System's own FAQ is unusually direct about the threshold. A student "must master the new skill taught in that lesson," and the standard is doing it "easily and with about 95% accuracy, before moving on to the next lesson."
The parent-facing FAQ uses the same figure for a level as a whole, describing the student who can perform the tasks "easily and with at least 95% accuracy (which is true mastery)."
Barton is the only program in this group that attaches an adjective to its number. Ninety-five percent is not merely the threshold; it is "true mastery."
What the published FAQ does not supply is a remediation procedure. It warns about the consequence of advancing early — "if you move on before the student has mastered a new skill, the student will eventually hit a wall" — without specifying what you do with the student who reaches ninety percent and stalls there.
Each level has a post-test, which the materials note can double as a pre-test, and the publisher's description of how the system works sets out the level sequence those post-tests gate.
Wilson: a specific mastery level, unspecified
Wilson Language Training's published implementation guidance describes the decision structure in detail and withholds the number. Within a Step, "the teacher progresses to the next substep when the majority of the students have achieved mastery," and "additional practice is provided to students not reaching the benchmark."
Between Steps, students sit an End-of-Step Assessment that is "both formative and summative," covering word reading accuracy and automaticity, high-frequency words, silent reading, passage fluency and comprehension, and spelling.
Then the operative sentence: students "must reach a specific mastery level," and if they do not, "additional instruction in challenging areas is necessary before proceeding to the next Step."
That is a complete specification of everything except the specification. The number lives in the purchased materials and in certification training, not on the public pages or on the curriculum overview — which is a defensible commercial decision and a real problem for anyone trying to compare programs before buying one.
It is the same structural issue found in the group size manuals, where the ceiling is often published and the reasoning behind it is not.
The public pages are specific about placement, incidentally, where the unit is a percentile rather than a percentage: Group 1 is "between 0–15th percentile" and Group 2 "between 16th and 30th percentile" on word attack and spelling.
Read Naturally: three conditions, and only one of them is a percentage
Read Naturally's published pass requirements for a story are a conjunction of three things. The student "makes three or fewer errors during the pass timing," "answers all the quiz questions for the story correctly," and "meets or exceeds the student's individual reading rate goal."
Notice what is happening in that sentence. The accuracy condition is an error count, not a percentage, which means its stringency depends on passage length. The comprehension condition is one hundred percent. The rate condition is a goal, and a goal is a different kind of object again.
The program's level-and-goal guidance adds a fourth number in a different unit: scores "less than 80 percent correct on the quiz questions on the first try" mean the level should not be raised. Goals are raised "by ten words at a time," levels "by half a year at a time."
Direct Instruction's Corrective Reading: points, and then a rate
SRA's Corrective Reading Decoding guides describe a point system rather than a percentage. Students in Decoding A "are able to earn 20 points per lesson," comprising "as many as 6 points for the word-attack portion of the lesson," "as many as 8 points for doing their Workbook pages well," and "as many as 6 points for Individual Reading Checkouts."
From Lesson 46 the reading is timed, and the published expectation is a rate: students read "at about 30 words in 45 seconds (40 words per minute) in Lessons 46 through 50," rising to "over 60 words per minute near the end of the program."
The publisher's programme overview for the decoding strand describes the same sequence. The guides also reference a schedule of fifteen Mastery Tests for Decoding A, which "indicate remedies for groups or students who have mastered specific reading skills."
The Mastery Test passing criteria themselves are in the purchased guide and were not readable in the publicly posted excerpt. The sample document points to a Mastery Tests section on page 45 and a schedule on page 46, neither of which is included in the public PDF.
If the Decoding A failure rule is load-bearing for a decision you are making, read it in your own copy rather than taking a number from a secondary source.
What a criterion looks like when a reviewer writes it down
The Council for Exceptional Children's own Alert series, describing subskill mastery measurement, states the structure plainly: student mastery of the first subskill "is assessed until the student reaches a pre-selected criterion (e.g., 80% accuracy)," and "reaching that criterion signals the teacher to move on to the next subskill in the hierarchy."
Read that parenthesis again. In a document written to explain the method to practitioners, the single most consequential number in it arrives as an example.
"Pre-selected" is exactly the right word and it is doing a great deal of work. It tells you the criterion is chosen in advance. It does not tell you by whom, or on what basis.
The one place a number is actually argued for
Siegfried Engelmann's own paper on student-program alignment and teaching to mastery is the exception that proves how unusual an argument is. It specifies four criteria, all of them in terms of first-time-correct responses:
Students "should be at least 70 percent correct on anything that is being introduced for the first time."
Students "should be at least 90 percent correct on the parts of the lesson that deal with skills and information introduced earlier in the program sequence."
"For material that is assumed to be mastered, the group should respond perfectly at least 9 out of 10 times."
"At the end of the lesson, all students should be virtually 100 percent firm on all tasks and activities."
It also specifies the consequences. Below seventy percent on new material, "if students perform much below 70 percent, they are not learning the material," and at around fifty percent they "may be at a chance level—guessing at the answers or the steps in the operation."
Below ninety percent on earlier material, "students need more extensive firming and more delayed tests." If students make too many mistakes or are not firm on earlier content, "they are placed too far in the program and should be moved back."
And Engelmann gives reasons. The seventy percent figure is justified on the grounds that a lesson introducing new material is "not composed entirely of material that is new," so the "initial rate of correct responses should not drop below 70 percent." The ninety percent figure "is based on the fact that students must be completely at mastery on earlier-taught material."
Those are rationales, not derivations. No study, dataset or calculation is cited for any of the four numbers, and no rationale at all is offered for the nine-out-of-ten figure. That is not a criticism of Engelmann, who was more explicit about his reasoning than any other publisher in this survey.
It is an observation about the state of the published record: the most carefully argued mastery criterion in widespread use is argued from the structure of a lesson rather than from evidence about retention.
The unit changes what the threshold means, and the units do not convert
Line the specifications up and the percentages turn out to be the least of the differences. Across the documents quoted above, mastery is operationalised as at least five incompatible things:
A percentage of items correct — Barton's ninety-five, Engelmann's seventy and ninety, the CEC Alert's illustrative eighty.
A count of errors — Read Naturally's three or fewer per pass timing, whose stringency varies with passage length.
A rate per unit time — Corrective Reading's forty words per minute by Lesson 50, Read Naturally's individual rate goal.
A proportion of occasions — Engelmann's nine out of ten on assumed-mastered material.
A number of consecutive sessions at some accuracy level, which is the dominant form in behaviour-analytic practice and appears nowhere in the reading program documents quoted here.
A student at ninety percent accuracy on a twenty-item task has made two errors and fails Read Naturally's three-error rule only on a longer passage. A student who is ninety-five percent accurate on one occasion has met Barton's threshold and has told you nothing about the proportion of occasions.
These are not stricter and looser versions of one standard. They are different standards, and this is the same pattern found in intervention dosage, where published frequency and duration turned out to carry four different meanings, and in group size, where the published ceilings came in nine incompatible units.
The behaviour-analytic literature has at least noticed the problem of dimensions. A descriptive analysis of applied research procedures reports that researchers "more often use higher accuracy levels across fewer observations," while practitioners "more often use lower accuracy across more observations" — two dimensions traded against each other, with no common currency for comparing a trade.
The federal guide tells you to teach to mastery and never says what mastery is
The What Works Clearinghouse practice guide on foundational reading skills is the most load-bearing instructional document in this area and it declines the question entirely. The guide's landing page carries its recommendations and evidence levels.
The word "mastery" appears four times in the guide, and not once as a criterion. It appears in the framing note about practices implementable "after students have mastered early reading skills." It appears in the description of repeated reading, in which students "read the same text multiple times for mastery."
It appears in a study description, where a student "demonstrated mastery by responding to a series of multiple-choice questions." And it appears in the panel's explanation of null findings — the panel "attributes the number of studies that found no discernible effects to insufficient mastery of alphabetic skills by beginning readers."
Every instance of the word "criterion" in the guide refers to the guide's own evidence-rating criteria, not to a student's performance.
The guide does contain one numeric accuracy threshold, and its form is instructive. Instructional-level text is defined in prose: students "should be able to read most of the words and grammatical structures, missing no more than one word out of every 10."
The accompanying figure showing frustration, instructional and independent levels carries no numbers at all, only a gradient of "support needed" and "challenge." Elsewhere the guide notes that some tests report "the highest text level a student can read at a predetermined level of accuracy (e.g., 90 percent accuracy)" — again, an example in a parenthesis.
One word in ten is ninety percent. The federal guide's single numeric threshold is a statement about text difficulty, not about when a student has learned something.
That distinction matters more than it sounds. A text-difficulty rule tells you what to put in front of a student today. A mastery criterion tells you to stop teaching something. The guide supplies the first and leaves the second to the program. The Clearinghouse's intervention reports on Reading Mastery, Corrective Reading and Read Naturally rate outcomes and likewise say nothing about advancement thresholds.
NCII's charts do not contain the word at all
The National Center on Intensive Intervention maintains the two charts most practitioners outside a district actually consult — the instructional intervention tools chart and the academic progress monitoring chart. Both were checked this week.
The word "mastery" appears zero times on the instructional intervention tools chart and zero times on the academic progress monitoring chart. On the intervention chart, "criteria" appears three times and every instance refers to the Technical Review Committee's criteria for evaluating study rigour.
On the progress monitoring chart, "criterion" appears six times and every instance refers to criterion validity — the correlation between the tool and an external measure, with thresholds of 0.60 and 0.40 on the lower bound of a confidence interval.
This is not an oversight. The charts rate tools on technical adequacy and on evidence of efficacy. Neither rating has anywhere to put a statement about when a student has finished a skill, because that is a question about a curriculum's internal structure rather than about a tool's properties.
The absence is still worth knowing: if you went to NCII expecting to find out what mastery means, there is nothing there to find, the same way there was nothing there about assistive technology and nothing there about morphology.
What the profession that credentials you says about it
The Association of Educational Therapists Code of Ethics and Standards for the Professional Practice of Educational Therapy, Revision #9, 2025, was read in full for this post. The counts:
"mastery" — zero
"master" — zero
"criterion" — zero
"criteria" — zero
"threshold" — zero
"proficient" — zero
What the Code does say is in section II, on maintaining records and measuring progress. Educational therapists strive to "develop and interpret individual goals and objectives for educational therapy, based upon appropriate assessment procedures," and to "establish ways to measure and determine client progress towards those goals at regular intervals."
Establish ways to measure. Determine progress. At regular intervals. Three obligations, no units, no threshold, no interval length.
This is now the fourth time this pattern has turned up in the same document. The Code names no record format, uses the word "group" in no service sense, specifies no dosage, and now sets no criterion. The consistency is the finding.
AET binds conduct and defers technique — its Fact Sheet describes the role in the same register — and a practitioner looking to the Code for a number will not find one — which means the number you use came from somewhere else, and it is worth knowing where.
Mastery measurement is a different paradigm, not a stricter one
There is a reason the mastery criterion and the progress-monitoring graph feel like they belong to different conversations. They do.
The IRIS Center's information brief on the distinction is the cleanest short statement of it. Mastery measurement "is a valuable tool for evaluating a student's understanding of and proficiency in a single target skill." General outcome measurement "is an effective means of evaluating and tracking student progress across the entire curriculum."
The brief's comparison table shows what mastery measurement does not do: it does not monitor growth across the year, does not offer alternate forms of equal difficulty, does not include standardised administration and scoring, does not predict performance on standardised achievement tests, and does not identify students who need additional instruction.
Its only numeric threshold is an example from a classroom vignette, in which a teacher decides a student has mastered a word type after reading "80% or more of the 'oa' words correctly in one minute." Note that even in an illustrative example, the criterion arrives with two dimensions fused — a percentage and a time limit.
The CEC Alert makes the same split in the same terms. Mastery measurement answers "Has the student learned the skill I have just taught?" General outcome measurement asks whether instruction "led to growth and improvement in the general academic area."
Curriculum-based assessment is "the clearest example of a mastery measurement approach"; curriculum-based measurement is a progress-monitoring system using "test materials that represent an entire curricular" domain. In IEP terms, the Alert notes, mastery measurement "focuses on short-term objectives." IRIS sets out the same two-paradigm split in its progress monitoring module, and a 2016 assessment-vendor paper traces the field's movement between them.
This is the cleanest available answer to a question practitioners ask constantly: why does the four-point rule not tell me whether to move off short vowels? Because it was never built to. The decision rules on a progress-monitoring graph operate on a general outcome measure and answer a question about the adequacy of a program.
The mastery criterion operates on a subskill probe and answers a question about the adequacy of one lesson's teaching. Running both is normal. Expecting either to answer the other's question is the error, and so is assuming the rules governing one have been validated for the other. For which tools do which job, the four progress monitoring systems differ in exactly this respect.
Where eighty percent actually comes from
Trace the eighty percent convention back and it does not originate in reading at all. It originates in the mastery learning literature of the late 1960s — Benjamin Bloom's Learning for Mastery and Fred Keller's Personalized System of Instruction — in which a unit quiz threshold was the mechanism that gated progression through a course.
The most useful surviving document on what those thresholds actually were is the Kulik, Kulik and Bangert-Drowns meta-analysis of mastery learning programs, published in Review of Educational Research in 1990. It matters here for a reason that has nothing to do with reading intervention: it is the only large synthesis that treated the criterion level itself as a variable rather than as a setting.
Two things in it are worth a practitioner's attention.
First, the review's own inclusion rule was that "the criterion for mastery had to be at least 70% correct on quizzes." That floor was the reviewers' construction, not a figure attributed to Bloom or to Keller. The seventy percent that shows up in Engelmann's materials two decades later and the seventy percent in this review's inclusion criteria are not the same claim and are not evidence for each other.
Second, and more substantively, unit mastery level significantly predicted effect size in the review's regression, at p = .016 — one of five study features that did. The mean effects by criterion band were about 0.49 for programs requiring seventy to eighty percent, 0.44 for eighty-one to ninety percent, and 0.64 for ninety-one to one hundred percent.
The authors summarise it as programs "with a high mastery standard were more effective than programs with lower standards," and give the example of programs that "require students to perform at a high level on unit quizzes (e.g., 100% correct)."
The boundary on that finding needs saying in the same breath. This is a 1990 synthesis of mastery learning programs across content areas and age bands, overwhelmingly in general-education course settings, not a study of reading intervention for students with learning differences.
It does not license the claim that a ninety-five percent criterion in a decoding program will outperform an eighty percent one, and nothing in it tells you what to do with a specific student on a Tuesday.
What you may take from it is narrower and still useful: the criterion level is not an inert administrative setting. It was associated with outcome differences in the one literature that bothered to look, and the association was not linear — the middle band had the lowest mean of the three.
Anyone citing "eighty percent" as the standard should know that the nearest thing to evidence on the question found the eighty-one to ninety band performing worst.
What happened when somebody tested the criterion directly
Outside mastery learning, the question has been asked most seriously in applied behaviour analysis, where mastery criteria are written into treatment plans explicitly and where the dominant form is a percentage held across consecutive sessions.
The honest summary of that literature is a sentence from a 2020 descriptive analysis in Behavior Analysis in Practice: "Only 2 studies have systematically evaluated the impact of any dimension of mastery criteria on skill maintenance."
The same paper notes that a large amount of the research base lacks technological descriptions of the criterion used, and lacks maintenance assessment altogether. It also reports that practitioners often adopt criterion levels lower than those found to reliably produce maintenance.
A 2023 Utah State University master's thesis then did the direct experiment on a very small scale, comparing eighty, ninety and one hundred percent criteria across three sessions, with each criterion assigned to a different instruction-following target for each of two three-year-old participants. The results were a mess, and the author says so.
For one participant, the hundred-percent target held at a hundred for two weekly probes before falling to ninety; the ninety-percent target fell to thirty and then zero; the eighty-percent target held at a hundred for two probes before falling to twenty and then zero.
For the second participant, all three targets held at a hundred across all four follow-up probes, so no comparison was possible. The thesis concludes that the results were too variable to support a claim about which criterion maintains best.
Its flat statement about the convention is nonetheless worth quoting: "There is little to no evidence to support the use of an 80% criterion."
Two participants, three targets, one inconclusive result, and that is close to the state of the direct evidence on a number that gates instruction for a very large number of students every day.
Note carefully what this does and does not transfer. The participants were preschoolers learning one-step instruction following, not school-age students learning grapheme-phoneme correspondences, and a maintenance probe on "wave" is not a maintenance probe on a decoding skill.
You may not conclude from this that eighty percent is the wrong criterion for your reading intervention. You may conclude that if someone tells you eighty percent is the established standard, they are describing a convention and not a finding.
Where the practitioners' numbers actually come from
The most interesting claim in this area is a survey finding about origins, and it is the one claim here that could not be verified at source.
The Utah State thesis, citing a 2019 survey by Richling and colleagues, reports that most clinicians' mastery criteria came from supervision, employer policy, or training rather than from research, and that "only 26% have cited reasoning that may have derived from empirical evidence."
That figure is reported here second-hand and should be treated as such. The Wiley-hosted journal page for the related 2019 Journal of Applied Behavior Analysis article could not be retrieved by any automated route available for this post, and the survey itself was read only through the thesis's description of it.
The thesis also describes the common standard inconsistently — its abstract says "80% accuracy across three sessions," its literature review says "80% across one or more consecutive sessions" — which is a reason for caution about the precision of anything quoted from it. If the twenty-six percent figure is load-bearing for something you are writing or presenting, read Richling and colleagues directly before you cite it.
The direction of the finding is consistent with everything else in this survey, and it is the part that does not depend on the number: the criteria in use were transmitted, not derived. That is the same conclusion the record forces about what "the science of reading" names and about what "Orton-Gillingham based" licenses: a phrase or a figure in universal use, with no body adjudicating it.
The one threshold that is genuinely derived, and what it is derived from
There is an exception in this collection, and it rewards attention.
Read Naturally's rate goal is not invented. The program's guidance states that a student's goal is typically reading unpracticed grade-level material at a rate "at least at the 50th percentile of national norms," and names the source: Hasbrouck and Tindal's oral reading fluency norms, published in 2006 as "Oral Reading Fluency: 90 Years of Measurement," with a later update for grades one through six.
So one of the four conditions a student must meet to pass a story has a real empirical pedigree. It is worth being precise about what kind of pedigree it is.
A norm-referenced threshold is a statement about other students, not about the skill. The fiftieth percentile is where the middle of a reference sample fell. It carries no claim that a student reading at that rate has mastered anything, and no claim that one reading below it has not.
It is a comparison, and the published table's own limits — the sample it came from, what it does and does not license — are covered in detail in the oral reading fluency norms, including what Technical Report 1702 does and does not say about interpretation.
The honest reading of Read Naturally's pass rule, then, is that it fuses three kinds of claim into one decision: an error tolerance with no stated basis, a comprehension requirement set at one hundred percent, and a rate target borrowed from a norm table that was never built to answer a mastery question.
That is not a criticism of the program, which is more transparent about its rules than most. It is a description of what a mastery criterion is in practice — a composite, assembled from whatever was available.
When several organisations share a vocabulary
The pattern in this post is now familiar enough on this site to be worth naming in one place, because the same shape keeps turning up wherever a number in daily use is traced to its source.
Program dosage turned out to be a fidelity specification rather than a minimum effective dose, with NCII publishing two different definitions of the word.
Group size turned out to be published by every program in a different unit, and absent from AET's governing documents entirely.
Progress monitoring decision rules turned out to be citable everywhere and attributable almost nowhere, with the federal practice guide rating the practice Low.
Grade equivalent scores turned out to be disowned by the publishers whose tests report them.
The standard error band turned out not to be one object, and not the one practitioners need for comparing two scores.
Evidence ratings turned out to be five systems answering five different questions.
Test norms turned out to have no published shelf life, only a publisher's revision schedule.
Qualification levels turned out to be set by six publishers using six letter schemes.
Math fact fluency turned out to rest on a working-memory claim the federal record labels the opinion of a panel.
Nine posts, nine numbers or labels in universal use, and the same finding each time. The field's working vocabulary is more precise in its form than in its derivation. That is not a reason to abandon the vocabulary. It is a reason to know which parts of it are measurements and which parts are conventions, because the two fail differently when a parent, a school or a colleague presses on them.
A sequence, not a decision
This post will not tell you what criterion to use, and should not. What the published record supports is a sequence of questions whose answers are all facts rather than judgements.
What does your program specify, in its own words and its own unit? Find the sentence. If it is a percentage, note the denominator. If it is an error count, note that its stringency moves with task length. If it is a rate, note the time window.
Is the number in the manual, or did you inherit it? Most practitioners are running at least one criterion that came from a supervisor, a graduate course or a previous employer. Those are legitimate sources of practice and illegitimate sources of citation.
Which paradigm is the number in? A subskill probe criterion and a general outcome measure's decision rule are not interchangeable, and neither has been validated for the other's job.
Does the criterion specify occasions as well as accuracy? A single-occasion threshold tells you nothing about stability. Most reading program criteria are single-occasion.
What does your program say to do on a miss? Several publish the threshold and not the remedy. If yours does, that is the gap you fill with professional judgement, and it is worth writing down what you did and why.
Is the number in a report going to a family or a school? Then say what it is a statement about. "Ninety-five percent accuracy on the lesson's target pattern on one occasion" and "mastered short vowels" are not the same sentence, and only the first is defensible. The session note and progress report conventions matter here, because this is exactly where an unexamined threshold becomes a written claim.
None of that requires you to change your criterion. It requires you to be able to say what it is and where it came from, which is a lower bar and one that a surprising amount of practice does not currently clear.
What is out of scope here
Three things, explicitly.
This is not guidance on whether to continue or discontinue an intervention. That decision has its own literature and its own published rules, it operates on different data, and it is not what a lesson-level mastery criterion is for.
This is not diagnostic guidance. A student's performance against a mastery criterion is information about a lesson, not evidence about a disorder. Nothing in this post supports an inference from a threshold to an identification, and the boundary between what you may do with an evaluation and what you may conclude from one applies in full.
Nothing here is legal advice. Mastery criteria appear in individualised education program objectives, and the rules governing what belongs in those documents are statutory and belong with the student's school team and, where a family needs one, an attorney. This post reports what professional bodies and publishers have written about instruction. It makes no claim about what any law requires of anyone.
Where the published record runs out
The gaps found while writing this, so you do not spend a morning looking for them:
Wilson's End-of-Step mastery level is not on its public implementation pages. The decision structure is described fully and the number is not stated. It is presumably in the purchased materials and in certification training.
Corrective Reading's Mastery Test passing criteria were not in the publicly posted Decoding guide excerpt. The sample PDF points to the section and does not include it.
The Richling and colleagues 2019 survey could not be retrieved directly. Its twenty-six percent figure is reported here only as the Utah State thesis describes it, and the thesis describes the surrounding convention inconsistently.
Barton publishes a threshold and no remediation procedure. What to do with the student who plateaus below ninety-five percent is left to the practitioner.
No program in this survey publishes a derivation. Engelmann publishes rationales, which is the closest anything comes, and cites no study for any of his four figures.
The direct experimental literature on mastery criteria is two studies deep, by the count of the 2020 descriptive analysis, and neither is in reading.
Nobody publishes a conversion between units. There is no authority anywhere that will tell you what Read Naturally's three-error rule is in percentage terms, or whether Engelmann's nine-out-of-ten is more or less demanding than Barton's ninety-five percent.
The federal instructional guides decline the question. The foundational skills practice guide uses "mastery" four times and attaches no criterion to any of them.
If you hold a piece of this that is not public — a Wilson End-of-Step figure from your own certification materials, a Corrective Reading Mastery Test rule from your own guide — that is a genuinely scarce thing in this area, and it is worth writing down somewhere your future self can find it.
Common questions
Is eighty percent the standard mastery criterion?
No body sets it as a standard. It is a convention, and it is the figure that appears as an illustrative example in several explanatory documents — the CEC Alert's "(e.g., 80% accuracy)," the IRIS brief's classroom vignette.
The nearest thing to evidence on criterion level, the 1990 mastery learning meta-analysis, found the eighty-one to ninety percent band with the lowest mean effect of its three bands, and a 2023 thesis states flatly that "there is little to no evidence to support the use of an 80% criterion."
Can I use my program's mastery criterion as a progress-monitoring decision rule?
They are built for different jobs. A mastery criterion applies to a subskill probe and answers whether one lesson's teaching took. A progress-monitoring decision rule applies to a general outcome measure and answers whether a program is working.
The IRIS brief lists what mastery measurement does not do, including monitoring growth across the year and identifying students who need additional instruction. Running both is ordinary practice. Substituting one for the other is not supported by anything published.
Why do two programs with similar percentages disagree about when a student is ready to move on?
Usually because the unit differs rather than the standard. Ninety-five percent of items on one occasion, three or fewer errors on a timed passage, ninety percent of first-time-correct responses across a lesson, and nine perfect performances out of ten are four different measurements. No published source converts between them.
Does AET require a particular mastery criterion?
No. The words "mastery," "criterion," "criteria," "threshold" and "proficient" do not appear in the Code of Ethics and Standards, Revision #9, 2025. What the Code requires is that you "establish ways to measure and determine client progress towards those goals at regular intervals." The method and the threshold are yours, and so is the responsibility for being able to explain them.
Is a higher criterion better?
The honest answer is that one meta-analysis of mastery learning programs found criterion level predicted effect size, with its highest band performing best — and that this was a 1990 synthesis across content areas in general-education settings, not a study of reading intervention for students with learning differences.
It does not license a claim about your program. What it does establish is that the criterion is not an inert setting, which is a reason to choose it deliberately rather than inherit it.
Where should I look for a number my program does not publish?
The purchased teacher's guide and the certification training first, since that is where several publishers keep it. After that, the publisher's own support channel, in writing. A number from a third-party summary, a conference handout or a colleague's recollection is not a citation, and the gap between the published record and common practice in this area is wide enough that the distinction matters.
Related reading
Nearest neighbours on this site: program dosage and what a weekly practice can deliver, what the manuals permit on group size, where progress-monitoring decision rules come from, and the four progress monitoring systems compared.
For the programs themselves, reading intervention programs compared and what "Orton-Gillingham based" actually means. For the roles and the vocabulary around them, start with educational therapist, learning specialist and executive function coach compared.
If this is the work you do
Illuminate publishes this kind of source-checked practitioner material for educational therapists, learning specialists and executive function coaches — the documents read in full, the counts done, the gaps named rather than filled in.
If that is useful to you, join the Illuminate educator list.
Comments