Repeated Reading Protocols: What Each One Specifies
Updated: 3 days ago
You and the specialist down the hall can both say you use repeated reading and be running two procedures that share almost nothing.
A repeated reading protocol can fix the number of rereads or leave it open; it can set a words-correct-per-minute criterion or name no rate at all; it can correct an error the second it happens or hold the correction back until the passage is finished. The label has been stable for fifty years. The procedure underneath it has never been.
That matters more for you than it does for a classroom teacher. A teacher inherits a program with a manual. You are usually assembling the procedure yourself, from a practice guide, a published study, a product you own and whatever the last training told you — and those four sources do not agree on the three decisions that define the thing.
This is a side-by-side account of what each published version actually specifies: how many rereads, to what criterion, in what text, with which error-correction rule. Everything here is quoted from the document that says it, with the document linked. Where a document declines to specify something, that is reported as a finding rather than filled in.
Two protocols can both be "repeated reading" and differ on the number of rereads, the criterion, the text level and the error rule. All four are the procedure.
Out of scope, deliberately. Nothing here is about statutory obligation, eligibility, state read-plan policy or what any law requires of a school. Those questions belong with an attorney and with your state education agency, and a credential does not change that. This is about what the published instruments and guides say, and about where the published record stops.
The short version
The National Reading Panel reported that repeated reading studies "either required a set number of repetitions (as few as one and as many as seven) or required students to practice repetition for some amount of time or until some fluency criteria were reached." That spread was already in the literature in 2000.
In the original procedure, the number of rereads was an outcome, not an instruction. In the current federal guidance for grades 4–9, it is an instruction: three to four, each time with a different purpose.
Across 31 studies reviewed in 2016, fluency criteria ranged from 30 to 210 correct words per minute, with error ceilings between two and ten per minute where one was set at all.
The strongest recent federal evidence statement about fluency never mentions a rate, a criterion, a timer or an error. The research-derived classroom protocol is built almost entirely out of those four things.
The best-known commercial implementation leaves the number of practice readings to the student's own judgement and gates the passage on four criteria, one of which is "three or fewer errors."
One randomised comparison removed the repetition from a repeated reading program and reported the stripped version was "as effective if not more so."
What "repeated reading" meant in the report that consolidated it
The National Reading Panel's fluency chapter is the document most people are gesturing at when they say repeated reading is evidence-based. It is worth reading for what it concedes. Its background section opens:
"There is common agreement that fluency develops from reading practice. What researchers have not yet agreed upon is what form such practice should take to be most effective."
The panel then groups repeated reading with neurological impress, radio reading and paired reading as members of one family — guided oral reading practice — rather than treating it as a single named technique. And when it describes how the studies in its own analysis were run, it says this:
"Repeated reading studies either required a set number of repetitions (as few as one and as many as seven) or required students to practice repetition for some amount of time or until some fluency criteria were reached."
One to seven. Or a clock. Or a criterion. That is not a protocol; it is a category. The word "criterion" does not appear anywhere in the chapter, and "corrective feedback" appears exactly once — in a report that is routinely cited as the evidentiary basis for a procedure whose modern versions are built on both.
The panel's summary of findings is where fluency entered the standard five-component list, and it is the reason every program brochure now carries the word. It is not where the procedure was specified.
The original procedure: the number of rereads was a result, not a rule
Here is the part that almost never survives into a training deck. In the founding studies, you did not tell the student how many times to reread. You set a rate, and the number of rereadings was what you measured.
Kostewicz, Kubina, Selfridge and Gallagher's review in Reading Improvement — the single most useful document in this literature for a practitioner, and the one nobody cites — reconstructs the original criteria. Two findings from it are worth memorising.
First, Dahl's 1974 criterion of 100 words per minute was not derived from a norm table. The review reports it came from curriculum-based criteria of 35 to 50 words per minute — a figure built for a different purpose and then raised.
Second, in Samuels's original work the review reports that the number of test readings needed to reach criterion fell from seven on the first passage to three by the fifth. The student was not instructed to read three times. Three is what it eventually took.
If the number of rereads is an outcome, a protocol that fixes it has converted your progress measure into a procedural step. You can do that. You just cannot then read it as progress.
This is the quiet inversion at the centre of the whole topic, and it explains why two practitioners citing the same literature can run incompatible sessions in good faith.
The research-derived classroom protocol: three components and two error rules
The most specified version available free is Therrien and Kubina's, published in Intervention in School and Clinic 41(3), 156–160 (DOI 10.1177/10534512060410030501; record). It defines repeated reading as a procedure in which "a student rereads a passage until meeting a criterion level," and it names three essential instructional components:
Passages should be read aloud to a competent tutor. The article is blunt about why: monitoring oral reading and giving feedback "is directly tied to program success," and the meta-analysis behind it found interventions run by adults or well-trained peer tutors were on average three times more effective.
Corrective feedback should be provided — and the rule has two branches, which is the detail that gets lost. If the student hesitates on a word for three seconds or omits it, the correction is immediate: supply the word, have the student repeat it, carry on. If the student mispronounces a word, the correction is delayed — supply it after the reading is complete but before the reread.
Passages should be read until a performance criterion is reached, with performance feedback on speed and accuracy after every single reading.
Its other specifications are equally concrete. Passages at the student's instructional level, defined as 85% to 95% word accuracy, of 100 to 200 words, readable in one to two minutes. Sessions of 10 to 20 minutes, three to five times a week.
A reread ceiling of up to four times per session. And a self-correcting rule for passage difficulty: if the student consistently needs extended rereading to hit the criterion, move to easier passages; if the criterion falls in a few readings, move to harder ones.
The criterion examples it gives are grade-referenced — 94 correct words per minute for second grade, 114 for third — and they are taken from a 1992 compilation. If you are going to borrow a rate from a table, it is worth knowing exactly what that table is and is not, which is the whole subject of our piece on what oral reading fluency norms do not say.
The underlying meta-analysis — Therrien's, in Remedial and Special Education 25(4), 252–261 — covered 18 studies published between 1977 and June 2001 with students aged 5 to 18, and reported mean effect sizes of 0.75 for fluency and 0.73 for comprehension (summary record). Those are the numbers the field quotes. They describe a specified procedure, not the category.
The 2022 federal guide: three to four rereads, a purpose, and no rate at all
Now put that beside the strongest current federal evidence statement about fluency. The IES practice guide Providing Reading Interventions for Students in Grades 4–9 (WWC 2022007) rates its fluency recommendation Strong on 33 studies. Here is what Recommendation 2 actually tells you to do, verbatim from the guide summary:
"The panel recommends having students reread the same passage a total of 3–4 times, each time with a different purpose."
The purposes it lists are reading at an appropriate pace and with expression, answering questions, identifying words the student does not know, and reflecting on what was learned or why the group is reading the passage.
On text selection it says to "choose short, content-rich passages at the students' instructional level that include multisyllabic words, vowel and consonant sounds and combinations, or vocabulary the students have previously been taught." On prosody it gets specific in a way it never does about rate: pause at commas, stop at periods, raise or lower the voice at a question mark, show emotion at an exclamation point.
And then there is what is not in it. Running word counts across the 22-page summary:
"criterion" — 0. "criteria" — 0.
"corrective feedback" — 0. "error correction" — 0. "errors" — 0.
"timer" — 0. "timed" — 0. "goal rate" — 0.
"words per minute" — 0. "words correct" — 0.
The word "rate" never appears as a word at all — 0 whole-word matches in the document.
"prosody" — 4. "instructional level" — 1.
The only two appearances of "minute" in the whole summary are a three-to-five-minute topic introduction and a two-to-four-minute video clip, and both sit inside the comprehension recommendation, not the fluency one.
The strongest federal evidence statement about fluency specifies a count and no rate. The best-specified research protocol specifies a rate and caps the count. They are not two versions of one instruction.
Fluency, in this guide, is "the ability to read text accurately, with ease, expression, and appropriate pacing." Pacing, not speed. The guide also tells you to counterbalance the procedure: "Reading a wide range of texts counterbalances the limitations of repeatedly reading the same brief passage." That is a published guide naming a limitation of the practice it is recommending, in the recommendation itself.
One more thing in the summary that rewards a second look. Its evidence grid reports Recommendation 2 as a statistically significant positive effect on oral passage reading fluency and an indeterminate effect on silent passage reading fluency. The practice has an evidence tier and a domain where it does not show. We went through that grid row by row in what the guides actually recommend for older students.
The 2016 federal guide: a glossary definition and a Moderate rating
The companion guide for the earlier grades, Foundational Skills to Support Reading for Understanding in Kindergarten Through 3rd Grade (NCEE 2016-4008, panel chaired by Barbara Foorman), houses repeated reading inside Recommendation 4 — "Ensure that each student reads connected text every day to support reading accuracy, fluency, and comprehension." Its glossary definition is as thin as a definition can be:
"Repeated reading refers to the instructional practice of having students practice rereading the same text as a way to support the development of oral reading fluency."
No count. No criterion. No rate. The phrase "fluency criterion" appears nowhere in the guide, "timer" appears nowhere, and the only appearance of "words per minute" is inside a footnote describing one study's eligibility screen.
What it does specify is a pairing. It tells you to assign "both repeated reading — in which they read the same text multiple times for mastery — and wide reading — in which they read many different texts."
And it gives a rationale for the repetition that has nothing to do with speed: "In repeated reading, students are less likely to practice incorrect word reading or to guess unknown words."
Its activity list for fluency practice is broader than most practitioners' working definition: individual oral reading with support, individual oral reading into a recording device with teacher feedback given later, partner reading, choral reading in small groups, simultaneous teacher-and-student reading, and individual oral reading with computerised devices provided the text is read at a pace appropriate to the student.
And its recommendation carries a Moderate rating, not a Strong one. Recommendations 2 and 3 of that guide are Strong; Recommendation 1 is Minimal; the connected-text recommendation that contains repeated reading sits in the middle column. The same practice is inside a Strong recommendation for grades 4–9 and a Moderate one for kindergarten through third grade, because the recommendations are not the same recommendation.
Two federal guides, two definitions of instructional level
This one is small and it will change a session. Both guides tell you to use instructional-level text. They do not mean the same band.
The 2016 K–3 guide says a student reading instructional-level text "should be able to read most of the words and grammatical structures, missing no more than one word out of every 10" — a 90% accuracy floor. Therrien and Kubina specify 85% to 95% word accuracy. An 87%-accuracy passage is instructional level under one specification and frustration level under the other.
The K–3 guide adds a conditional nobody else makes: text should be at the student's independent level when they read without feedback, and at the instructional level when they read with feedback. So the right text for a reread depends on whether you are in the room listening — which, for a once-weekly private caseload with practice assigned between sessions, is the decision that actually bites.
Where accuracy percentages and advancement thresholds come from in the first place is its own subject, and we took it apart in where the percentages in the manuals come from.
The criterion: thirty to two hundred and ten words correct per minute
If you set a rate criterion, the literature will not tell you which one. It will tell you the range.
The 2016 Reading Improvement review screened for repeated reading studies that used a fixed fluency criterion and found 29 qualifying articles containing 31 studies across 15 journals. Its headline numbers:
Goal rates ranged from 30 to 210 correct words per minute.
Where studies set an error ceiling as well, the ceiling ranged from 2 to 10 incorrect words per minute.
Some studies counted correct words only; others counted total words read, correct and incorrect together.
Inside that, the criteria came from three different places, and the review's most useful contribution is separating them:
Researcher- or teacher-imposed rates, in 10 studies. Student-specific or school-wide figures, ranging from 75 to 200 words. One school used a single rate for every student in the building.
Normative or grade-level rates, in nine studies. Criteria lifted from published compilations or from the measured rate of a proficient reader, varying from 30 to 180 depending on grade level or reading age.
Behavioural fluency rates, in six studies. Criteria taken from the precision-teaching literature rather than from reading norms at all.
Two procedural details in that review are worth stealing. One study required the student to meet criterion on three consecutive sessions before moving to a new passage — a rule that converts a single reading into a stability judgement. And the early studies reported accuracy or word-recognition errors but did not incorporate either into the criterion; the error ceiling is a later addition, not part of the original design.
A score at a criterion is still a score with a standard error attached, which is the one thing nobody prints beside a goal rate. We wrote about what that band actually means, and about the difference between a norm, a benchmark and a cut point.
What a synthesis says about the number of rereads — and what it rests on
Stevens, Walker and Vaughn synthesised fluency intervention research from 2001 to 2014 for K–5 students with learning disabilities — 19 studies, full text free (DOI 10.1177/0022219416638028). Their conclusion is quotable and they attach its limits in the same abstract:
"Findings suggest that RR remains the most effective intervention for improving reading fluency for students with LD. Limitations include sample size, only three group design studies, and infrequent use of standardized measures."
On the components, they report that "providing a model of fluent reading and performance feedback, using easier level text, setting a performance criterion, and practicing RR with peers also contributed to improved outcomes." The five studies in their first table used rereading "ranging two to four times, without modeling by a more proficient reader" — so two is in the published range as well as seven.
Then the question everyone actually asks: does more rereading help? The synthesis answers it from one comparison, and says so. Ardoin and colleagues ran three rereads against six with four students, one of whom had a learning disability.
Mean gains for that student were 43.6 correct words per minute at three rereads and 60.0 at six. On generalisation to a high-content-overlap passage, the same-day gains were 15.4 and 19.6, and the one-week gains 34.8 and 39.2. The synthesis's reading of it:
"While increasing the number of RRs may immediately affect performance, gains may not generalize to unfamiliar but similar texts."
That is the whole evidence base for doubling your reread count, and it is one student in one study. Treat anyone who quotes a number of rereads as settled with the scepticism you would bring to any single-subject finding.
Two further component findings in that synthesis are more load-bearing than the count. Setting a performance criterion was associated with greater gains than rereading with no criterion — but using a higher criterion "does not differentially effect the decrement of WCPM over time; as such, gains may not be sustainable over time."
And one study compared systematic error correction alone with error correction plus rereading: the combined conditions outperformed correction alone, and correction plus rereading of previously read material beat correction plus rereading of new text.
The honest caveat the authors give applies to every list of components in this article: "it's difficult to identify the efficacy of individual components within combined intervention packages."
On generalisation to unfamiliar passages the synthesis is explicit: "Only two studies examined generalization of skill to unfamiliar passages, and the results are inconclusive."
The one trial that took the repetition out
This is the finding most likely to be new to you, and it is in the same research group's own work.
Therrien, Kirk and Woods-Groves compared a reading fluency intervention built on repeated reading plus question generation against a modified version of the same program with the passage repetition removed — 30 students in grades 3 through 5, 50 sessions over four months, random assignment to condition (Remedial and Special Education 33(5), 309–319; record, DOI 10.1177/0741932511410360). Two reported findings:
"Regardless of condition, all students made gains in oral reading fluency on independent passages."
"The modified RAAC program without passage repetition appeared to be as effective if not more so."
State clearly what that does and does not license. It is one trial of one program with thirty students, and it does not show that repeated reading is ineffective — the synthesis above, drawing on nineteen studies, points the other way on fluency outcomes.
What it does show is that in at least one well-specified package, the repetition was not demonstrably the active ingredient. A separate study in the Stevens synthesis found three rereads outperformed fifteen minutes of continuous reading against a no-treatment comparison, with medium-to-large effects on standardised measures. Both results are in the record. Neither settles it.
What you may not draw from this: that your own student will do as well without the repetition. That is a claim about an individual from a group comparison, and it is not available.
What the commercial implementation publishes, and what it leaves to the student
Take the best-known product in this space. The clearest public specification of Read Naturally's core strategy is not on the vendor's site — it is in the What Works Clearinghouse intervention report for beginning reading, which describes the repeated reading component like this:
"Students engage in 1-minute practice readings to build their mastery of the passage. Once students feel they can achieve their reading speed goal, they alert the teacher."
Read that second sentence again. The number of practice readings is set by the student's own judgement. Not by the protocol, not by a reread cap, not by a timer — by the student deciding they are ready. The procedural gate is at the other end: the teacher then runs a "pass timing" evaluated against four criteria, and the report lists all four.
The student reaches the goal rate.
The student makes three or fewer errors.
The passage is read with appropriate phrasing.
The comprehension questions are answered correctly.
Miss any of them and the student practises more and the pass timing runs again. That is a criterion-based protocol in the original Dahl and Samuels sense — the rereads are an outcome — wrapped around a product.
The vendor's own published step list is consistent with this and quantifies less than you would expect. The cold timing measures how many words the student reads correctly in one minute on an unpractised story, and the teacher conducts the first three timings before deciding whether the student self-times.
The practice step says the student reads aloud quietly, without audio, "several times." The pass step requires the goal rate, appropriate expression and "three or fewer errors." The page gives no read-along repetition count and no numeric goal rate; it refers the reader to the teacher's manual. The strategy page names the three components — teacher modelling, repeated reading, progress monitoring — and stops there.
Even the one-minute timing page in the user guide does not define an error on the page where you count them. It says "as the student reads, count his or her errors," tells you what to do when the bell sounds, and sends you elsewhere for what an error is.
None of that is a criticism of the program. It is the ordinary condition of commercial material: the manual holds the specification and the manual is not public. It is a reason to open the manual you own before assuming your implementation matches the one in the study.
The partner version: two readings and no stated criterion
A second commercial shape worth naming, because plenty of practitioners run something like it. The publisher's overview of the Six-Minute Solution describes a partner fluency routine: teachers assess instructional reading levels and place students in fluency pairs, each session contains two one-minute readings of the same passage — partner one reads, then they switch and partner two reads it — and the listening partner marks errors and the stopping point.
What the overview does not state: the pairing rule, the error-correction procedure itself (it says the listening partner "does error correction procedure" and leaves it there), and any criterion for moving to a new passage. Two readings, no criterion, correction unspecified. That is a fourth distinct protocol, and it is the one furthest from the research-derived version.
Group size and delivery format turn out to be specified very differently across published programs, which we mapped in what the manuals permit on group size.
What NCII's brief adds — and who supplied it
The National Center on Intensive Intervention publishes per-program taxonomy briefs that answer the dosage question commercial pages usually dodge. The Read Naturally Live brief is the most quantified document available on this program outside its manual:
Group size: up to six students per adult in grades 1–5; up to eight per adult from grade 6 to adult.
Frequency and duration: 3 to 5 days a week, 20 to 30 minutes a session, for a minimum of 8 to 12 weeks.
Minimum dose: 24 stories, the equivalent of one sequenced or phonics level.
Opportunities to respond: "Each student has an estimated 15 opportunities to respond and receive corrective feedback per completed lesson."
That last figure is the one worth carrying. NCII defines dosage as "the number of opportunities a student has to respond or practice and receive corrective feedback," which makes fifteen per lesson a directly comparable number across programs — and the brief series lets you compare it.
But read the provenance line in the brief, because it governs everything above it:
"The evidence comes from the intervention's vendor or developer. It is accurate as reported to the National Center on Intensive Intervention (NCII); it was not independently verified by NCII."
So the most specified public account of this program's dosage is the vendor's account, published in a federal centre's template. That is better than nothing by a wide margin and it is not independent verification. The taxonomy of intervention intensity itself, and the academic intervention tools chart, are the right places to go next if you want to put a second program beside it.
The clearinghouse record on the best-known program
Since Read Naturally is the implementation most practitioners have actually seen, its clearinghouse record is worth stating plainly rather than leaving as an impression. The July 2013 beginning reading report, built on five studies that met WWC standards, concluded:
"Read Naturally® was found to have potentially positive effects on general reading achievement, mixed effects on reading fluency, and no discernible effects on alphabetics and comprehension for beginning readers."
The improvement indices behind those words:
General reading achievement +10, across 2 studies and 126 students, extent of evidence small.
Reading fluency +7, across 4 studies and 440 students, extent medium to large.
Alphabetics +2, across 2 studies and 264 students.
Comprehension 0, with a range running from −16 to +9, across 4 studies and 439 students.
The adolescent literacy report from March 2013 reported potentially positive effects on general literacy achievement for adolescent readers.
A fluency program rated mixed on fluency and no discernible effect on comprehension is not a damning finding — four studies and a 2013 search window are a thin basis either way — but it is a long way from how the practice is usually described in a training.
What an evidence badge means, and what each rating system is actually rating, is the subject of five systems and five questions; the general problem of a clearinghouse record aging under a product that keeps shipping is in what actually makes a record too old.
What AET's Code of Ethics says about any of this
Nothing. And that is the finding, not an omission on our part.
We ran word counts across the AET Code of Ethics, Revision #9, 2025:
"fluency" — 0.
"repeated reading" — 0.
"fidelity" — 0. "evidence-based" — 0.
"reading" — 1, and it is in the phrase "reading of professional literature."
"protocol" — 1, and it is about copyright attribution.
The Code does name intervention design as part of the role: it describes an educational therapist as skilled in, among other things, "development and implementation of appropriate intervention programs." And it sets one requirement that bears directly on borrowing a protocol wholesale. Educational therapists are to "understand that educational therapy services are based on the unique needs of each individual client and conduct an initial educational assessment/screening to inform their intervention plan."
The competence clauses in the instruction and assessment section do the rest of the work: select and use appropriate instruments "recognizing their limitations with respect to reliability, validity, and bias," and use only those for which you have been adequately trained. Those are the clauses that apply when you decide whether your version of the procedure is the version that was studied.
One citation note we keep repeating because it matters: the Code numbers its provisions by section heading and letter and prints no dotted references, so a dotted citation is a convention for pointing at a clause rather than a quotation from the document. The AET fact sheet is the companion document for scope questions.
Where this leaves a once-weekly private caseload
Every specification above was written for a setting that is not yours. The grades 4–9 guide names its audience as special educators, general education teachers, intervention teachers, reading specialists, reading coaches and trained volunteers — educational therapists are not on the list, and neither is private practice.
The group sizes in the NCII brief are classroom group sizes. The frequency figures, three to five sessions a week in every source that gives one, assume a school timetable.
Three consequences follow, and none of them is a reason to abandon the procedure.
A once-weekly session cannot deliver the frequency any source specifies. Therrien and Kubina say three to five times a week; NCII's brief says three to five days a week; the 2016 guide says connected text every day. If repetition happens once a week and not between sessions, you are running something the literature has not described — say so in the progress report rather than cite an effect size.
The criterion decision is yours and it should be written down. Pick the rate, record where it came from, and record the error ceiling if you set one. "Reads to criterion" in a session note means nothing to the next practitioner who reads it. What belongs in a note at all is its own question, and we worked through it in what the codes actually require of session notes.
The thing you measure should not be the thing you also prescribe. If you fix the number of rereads, your rereads are no longer telling you anything. Track them, and let the criterion do the stopping.
Who is qualified to run which procedure, and where that boundary sits between an educational therapist, a learning specialist and a coach, is the question the whole site is built around: start with the role comparison and then what educational therapy actually is.
A decision sequence, not a decision
We will not tell you how many times to have a student reread a passage. Nobody who has read the primary sources would. What you can do is make the four decisions deliberately and in the right order.
Decide whether you are running a count protocol or a criterion protocol. Those are the two families and they do different jobs. A count protocol is simpler to deliver, is what the grades 4–9 guide recommends, and makes rereads useless as data. A criterion protocol is the original design, keeps rereads informative, and needs you to have chosen a rate.
If criterion: choose the rate and say where it came from. Teacher-imposed, grade-referenced or behavioural — the 2016 review's three families. Any of them is defensible. An unrecorded one is not.
Choose the text band against an accuracy figure, not a feel. 85% to 95% if you are following Therrien and Kubina; no more than one error in ten if you are following the K–3 guide. And decide whether you are present with feedback, because the 2016 guide makes the right band depend on that.
Write down your error-correction rule before the session, including the delay. Three seconds and immediate for a hesitation or omission; after the reading and before the reread for a mispronunciation, if you are using the specified version. Any rule is better than an implicit one, because an implicit rule changes when you are tired.
Then add the thing every federal guide asks for and no product delivers: wide reading beside the repetition. Both guides say it. One of them says it inside the recommendation it is making.
Where this sits next to what is already published here
Oral reading fluency norms: what the table does not say — where a grade-referenced rate comes from and what it cannot bear.
Reading intervention for older students — the grades 4–9 guide read recommendation by recommendation, including the evidence grid.
Mastery criteria: where the percentages in the manuals come from — the companion question for advancement thresholds.
Progress monitoring decision rules — what to do with the graph the criterion produces.
When to discontinue reading intervention — the other end of the same sequence.
Math fact fluency intervention — the identical automaticity argument in a different subject, with a different documentary record.
Reading intervention programs compared and Orton-Gillingham based: what the label means — when a program name does and does not specify a procedure.
Intervention session frequency and duration — four meanings of a dosage figure.
The science of reading: what the phrase names — the umbrella this whole literature now gets filed under.
Is educational therapy evidence-based? — the prior question about the profession rather than the procedure.
Morphology instruction programs and Phonological awareness assessment — adjacent instructional and assessment territory.
Dyslexia definition compared — five bodies, five definitions, same inconsistency problem one level up.
The error rules inside these protocols are one program's answer to a question every program answers differently. Error correction procedures across the published programs sets the three published defaults side by side, including the finding that two assessment manuals and the federal practice guide prescribe opposite responses to the same misread word.
Every protocol above is specified in rereads, accuracy and rate, and none of them scores how the reading sounds. Prosody and the fluency rubrics sets out the four rubrics that do, the incompatible scales they use, and why a prosody rating belongs in a report as description rather than as a second independent outcome.
Where the published record runs out
Named honestly, because a list of what could not be verified is worth more than a confident sentence.
No document anywhere specifies an optimal number of rereads. The word "optimal" does not appear in the National Reading Panel's fluency chapter. The 3–4 figure in the 2022 guide is the panel's recommendation, presented as such, not a dose-response finding.
The dose-response evidence is one study with four students. Ardoin and colleagues' three-versus-six comparison is the entire basis for "more rereads, faster reading" in the 2016 synthesis, and the synthesis says so.
Generalisation to unfamiliar text is unresolved. Two studies in the synthesis examined it and the results were called inconclusive.
Commercial manuals are not public. Read Naturally's published step list defers to the teacher's manual for read-along counts and goal rates; the Six-Minute Solution overview defers on error correction and gives no criterion. Neither can be verified from outside.
NCII's dosage figures are vendor-reported. The brief says so in its own text. They are the best public numbers available and they are not independently verified.
No figure here was verified for private, once-weekly delivery. Every frequency, group size and duration in this article comes from a school setting.
We could not reach the full text of the 2004 meta-analysis. Its study count and effect sizes are reported here from a summary record rather than from the article itself, and the transfer-versus-same-passage split that is sometimes quoted from it is not reproduced here for that reason.
Frequently asked questions
How many times should a student reread a passage?
There is no verified optimal number. The IES practice guide for grades 4–9 recommends three to four rereads, each with a different purpose, as a panel recommendation. Therrien and Kubina's protocol caps it at four per session and otherwise lets a performance criterion decide. The National Reading Panel reported published studies using anywhere from one to seven. If you want the rereads to function as data, do not fix the number.
Does repeated reading need a words-correct-per-minute criterion?
Not under current federal guidance. The grades 4–9 practice guide summary contains no criterion, no rate and no error count at all, and its recommendation is rated Strong on 33 studies. The research-derived protocol does require one, and a 2016 synthesis found setting a criterion was associated with greater gains than rereading with no criterion. Both are defensible; the mistake is running one and reporting the other.
What criterion rate should I use?
That is your decision and the literature will not make it for you. Across 31 studies, criteria ran from 30 to 210 correct words per minute, drawn from three different sources — teacher-set, grade-referenced and precision-teaching rates. Record which family yours came from. Note also that a 2016 synthesis found a higher criterion did not protect against decline over time.
What is the correct error-correction rule?
The most specified published version has two branches. If the student hesitates for three seconds or omits a word, supply it immediately and have them repeat it. If the student mispronounces a word, supply it after the reading is finished but before the reread.
Performance feedback on speed and accuracy follows every reading. Commercial programs frequently do not publish their rule at all, which is a reason to open the manual rather than to improvise.
Is the repetition itself the part that works?
Not established. One randomised comparison of thirty students removed passage repetition from a repeated reading program and reported the stripped version was as effective if not more so. A separate study found three rereads outperformed fifteen minutes of continuous reading against a no-treatment comparison.
The 2016 synthesis is explicit that it is difficult to identify the efficacy of individual components inside combined packages. Nothing here licenses a claim about how your particular student will respond.
Can I tell a family that repeated reading is evidence-based?
You can say, accurately, that the IES practice guide for grades 4–9 rates purposeful fluency-building activities including repeated reading as Strong on 33 studies, and that a 2016 synthesis of 19 studies concluded it remains the most effective intervention for improving reading fluency for students with learning disabilities while naming its own limitations.
What you cannot do is attach that evidence to a procedure that differs from the studied one on the count, the criterion, the text band or the error rule — and you cannot promise an individual result. If the program you use carries a clearinghouse rating, quote the rating and the domain it applies to, not the headline.
Join the Illuminate educator network
We write one of these a day for practitioners — educational therapists, learning specialists and executive-function coaches — and every claim in them is traced to the body or publisher that made it. If reading a protocol against its own primary sources is the kind of work you want in your inbox, join the Illuminate educator list.




Comments