Error Correction in Reading Intervention: Three Procedures
You will make this decision more often than any other decision in a session. A student reads a word wrong. What do you do in the next two seconds?
Error correction in reading intervention is the most frequent procedural choice a practitioner makes, and it is the one the published record specifies least. Three of the documents on your shelf give three different defaults. Two of them are assessment manuals that tell you to supply the word and score it wrong.
One is a federal practice guide that tells you not to supply the word at all. And the federal taxonomy that rates intervention intensity counts how many corrections a student receives per session without ever asking what a correction is.
This post compares what each published source actually specifies, and attributes every requirement to the body that published it.
The short version
There are three distinct published procedures, and they are not variants of one idea.
Supply and score. Wait three seconds, give the student the word, mark it incorrect. This is the rule in both major curriculum-based measurement systems, and it is a measurement rule, not a teaching rule.
Model and drill. Supply or prompt the word, have the student repeat it, then reread the sentence containing it, often a set number of times. This is phrase drill and its relatives.
Withhold and prompt self-correction. Do not simply tell the student the word. Pause, let them try, have them reread the sentence, and ask whether it made sense. This is what the federal practice guide recommends, with one named exception.
No published body adjudicates between the three. The clearinghouse that rates the programs counts the corrections and does not describe them.
Three seconds, then tell them: the rule the two measurement systems share
Start with the one rule in this area that is genuinely settled, because two independent publishers state it identically.
The University of Oregon's DIBELS 8th Edition Administration and Scoring Guide gives one acceptable prompt for Oral Reading Fluency and it is a hesitation prompt: "If student hesitates for 3 seconds on a word, give the correct word, and mark the word as incorrect."
The guide repeats the structure across subtests. Word Reading Fluency adds "point to the next word, and say: Keep going." Letter Naming Fluency says "score the letter as incorrect, provide the correct letter." Phoneme Segmentation Fluency says "give the next word, and score the word as incorrect."
Acadience Learning's published rule for ORF is the same sentence in different words: "If a student hesitates or struggles with a word for 3 seconds, tell the student the word and mark the word as incorrect." Its Wait Rule is stated plainly as "On ORF, 3 seconds."
The two systems also agree on the surrounding scoring rules. A word initially mispronounced but self-corrected within three seconds is scored correct in both. Repetitions and rereadings are ignored by both. Inserted words are not counted by either. Omitted words are errors in both.
Two publishers, two manuals, one procedure: three seconds, supply the word, score it wrong. This is the only part of error correction the field has actually standardised, and it is a rule about measuring, not about teaching.
That agreement is worth noting because these same two publishers do not agree everywhere. On nonsense word fluency they score the identical student response one and zero, which is the subject of the pseudoword decoding comparison already published here. On the three-second hesitation rule they converge completely.
What the measurement rule is for, and what it is not for
Both guides are explicit that these are administration rules. The hesitation prompt exists so a one-minute timed sample keeps moving and produces a comparable number. It is not presented as instruction, and neither publisher claims it teaches the word.
This matters because the procedure migrates. A practitioner who administers oral reading fluency probes weekly has the three-second supply rule in muscle memory, and it is the procedure most readily to hand when a student stumbles during teaching. Nothing in either manual authorises that transfer, and nothing in either manual forbids it. The manuals are silent on the question, because it is not their question.
If you use a timed measure to track progress, the progress monitoring decision rules already published here cover what the resulting number licenses, and the oral reading fluency norms post covers what the published table does and does not say.
Read Naturally puts the same rule inside a program, and then hands the marking to the student
Read Naturally is the clearest published case of the measurement rule sitting inside an intervention program, because the program is built around timings.
The Read Naturally teacher's manual and the company's knowledge base on conducting timings both give the rule in one sentence: "If a student hesitates for three seconds, tell the word to the student and count the word as an error."
The surrounding scoring rules match the measurement systems exactly. A mispronounced word or a dropped ending is an error. A substitution is an error even when the substitute is a synonym. A self-correction counts as correct. Repetitions are not errors. Insertions are not errors.
What Read Naturally adds is a step the assessment manuals do not have. During the cold timing the student, not the teacher, marks the errors: in the software the student clicks each word read incorrectly and the program plays the correct pronunciation, and in print versions the student underlines the words while the teacher models them.
The company's own blog post on the subject, Own Those Mistakes, argues the point directly, saying students "should be in charge of clicking on or underlining them," with the teacher pointing out errors at first and withdrawing as the student learns to hear them.
The missed words then get extra attention during practice, and once a story is passed the student may be asked to read the underlined words in isolation.
So Read Naturally's answer to the question is not simply "tell them." It is: supply it in the timing, score it wrong, make the student mark it, and have the student practise it. That is a four-part procedure, and the three-second supply is only the first part.
One thing the manual does not do is give a separate rule for misreads during practice. It gives rules for timings in detail and says comparatively little about what happens to a misread word during the practice step itself.
The federal practice guide recommends the opposite default
Now the source that disagrees. The IES and What Works Clearinghouse practice guide Foundational Skills to Support Reading for Understanding in Kindergarten Through 3rd Grade contains a correction procedure in Recommendation 4, and its default is not to supply the word.
The guide's wording, on pages 34 and 35, is this. When a student makes a word-reading error on a word he or she should be able to read, "pause so the student can correct the error; provide support if needed." It continues: "Rather than simply telling the student the correct word, have students reread the sentence in which the misread word appears."
For a student who cannot find the error alone, the guide says to "read the sentence(s) exactly as the student did, including the error," then ask "Did that make sense?" or "Did that sound right?" It closes with an instruction to fade: "Use these scaffolds less frequently as students begin to independently self-monitor and self-correct their errors."
The federal guide's default and the assessment manuals' default are opposites. One says supply the word in three seconds. The other says do not simply tell the student the word.
Two details in that passage are easy to miss and both change what it licenses.
The first is the scope condition. The procedure is written for "a word he or she should be able to read." It is not a general rule for every misread word, and the guide does not present it as one.
The second is the exception branch, which appears later in the same recommendation: "When students stop on a word that is very challenging or irregular, tell them the word and have them repeat the word and reread the sentence."
So the guide does contain a supply-the-word procedure. It is conditioned on the word being irregular or very challenging, and it is followed by a repetition and a reread rather than by a score.
Read together, Recommendation 4 is a two-branch procedure keyed to the word, not to the clock. Which branch you take depends on whether the student should have been able to decode it.
The guide also publishes a script, and it is a meaning check
The same recommendation carries a worked example, labelled Example 4.3, for a self-monitoring activity the guide calls the "Fix It" game. The steps are published in full: the teacher reads sentences, some containing a word that does not make sense; students say whether each sentence makes sense or sounds right, explain why not, and then fix it.
The guide prints the dialogue. The teacher reads "The bus stepped at the corner," asks "Does that make sense?", the student answers no, the teacher asks why not, the student says "A bus can't step," and the teacher says "Fix it!"
It is worth being straightforward about what that script is. The prompts "Did that make sense?" and "Did that sound right?" are meaning and syntax checks, and practitioners trained in structured literacy will recognise the family they come from.
The guide pairs them with explicit decoding instruction elsewhere in the same document and scopes them to words the student should already be able to read. Reporting the procedure accurately means reporting both the prompts and the scope condition, and leaving the argument about cueing systems to the people having it.
"Corrective feedback" appears zero times in the guide that recommends feedback thirty times
The guide's own vocabulary is the sharpest thing in this area, and it is countable.
The foundational skills guide extracts to 58,915 words. Counted as whole words on whitespace-normalised text, the phrase "corrective feedback" appears zero times. So does "miscue." So does "miscues." So does "immediate feedback." The bare word "corrective" appears twelve times and every one of the twelve is the program name Corrective Reading, appearing in the evidence tables as a studied intervention.
The phrase "error correction" appears exactly once in the whole guide, and it is not in a recommendation. It sits inside a description of a study the panel reviewed: a tutor "provided training on the passage, focusing on error correction, repeated reading with modeling, and repeated readings."
Against those zeros, "feedback" appears thirty times, "model" twenty-one, "errors" eleven, "prompt" eleven, "modeling" ten, "prompts" six, "prompting" five, "self-correct" twice and "self-correction" once.
The federal guide that tells you to provide feedback thirty times never once uses the field's standard name for the thing it is describing.
The thirty "feedback" mentions are instructive on their own. They are overwhelmingly of the form "model strategies, scaffold, and provide feedback," "provide practice with feedback," "through modeling and feedback," "with teacher feedback provided later." The guide mandates the activity in general terms and specifies the mechanism once, in Recommendation 4, in the passage quoted above.
One of the thirty is a warning rather than a recommendation, and it is worth having: the guide cautions against students reading text "without feedback, as it can lead them to practise" ineffectively. So the guide's position is that feedback matters and that unmonitored practice is a risk, while its procedural specification runs to about four sentences.
A methodological note, because it nearly produced a false finding here. Extracted with layout preserved, "corrective feedback" would have been reported as zero even if present, because a line break between the two words defeats a phrase search. The count above was re-run on whitespace-normalised text and the zero held. Any absence claim built on a multi-word phrase in a PDF needs that second pass.
Two measurement manuals never use the phrase either
The same count on the assessment side produces a result nobody would predict from how practitioners talk.
The DIBELS 8 administration and scoring guide runs to 41,205 words. In it, "error correction" appears zero times, "corrective feedback" zero times, and the bare word "corrective" zero times. What it does contain is eighteen separate rules keyed to the phrase "3 seconds," fifty-one uses of "prompt," fifty-nine of "discontinue," nine of "self-correct" and seventeen hits on the stem "hesitat."
The document that specifies the three-second supply rule more precisely than any other source in this post never calls it error correction. It calls it a hesitation prompt, and it files it under administration.
That is the pattern this site keeps finding, in a new place: the vocabulary practitioners use to discuss a procedure is not the vocabulary the governing documents use to specify it. The same problem shows up in mastery criteria, where four programs set thresholds in at least three non-convertible units, and in decodable text criteria, where the defining term has no agreed method behind it.
Phrase drill is the one fully specified procedure, and no program publisher owns it
If you want a correction procedure written out step by step with a number attached, there is one, and it did not come from a program publisher.
The New York State RtI Technical Assistance Center publishes the phrase drill procedure as a discrete intervention. The steps as published are these.
Select a passage the student can read with 85 to 95 percent accuracy.
Have the student read it while the teacher marks misread words on a separate copy.
Point to each marked word and read it aloud.
Have the student read each marked word aloud.
Have the student read each sentence containing a marked word aloud three times.
Where a sentence holds more than one marked word, practise each word alone first, then read the sentence three times.
Afterwards, check progress with an oral reading fluency probe.
The page attributes the procedure to Brown-Chidsey, Bronaugh and McGraw's 2009 RTI handbook and lists Begeny, Daly and Valleley (2006) and O'Shea, Munson and O'Shea (1984) among the research, while not stating that those studies describe this exact procedure.
So the most completely specified error correction procedure in common use is published by a state technical assistance centre, sourced to a trade handbook, with a repetition count that is a round number. That is not a criticism of phrase drill. It is a statement about where procedural detail in this field actually lives, and it is not in the program manuals.
The comparative evidence for it is real but narrow. The Meadows Center's synthesis of interventions for improving oral reading fluency reports that in Begeny and colleagues' four-condition comparison, the phrase-drill-with-error-correction condition produced the highest words correct per minute and the lowest errors per minute of the conditions tested.
That was one study of one student. The same synthesis records a separate study in which participants showed positive fluency outcomes during repeated reading without a model and without error correction.
What NCII's own briefs say the correction is: six of seventeen say anything at all
The National Center on Intensive Intervention publishes academic intervention taxonomy briefs for seventeen named programs, eight in reading and nine in mathematics. Each brief rates the program against the Taxonomy of Intervention Intensity, and one of its seven dimensions is Dosage, defined as "the number of opportunities a student has to respond or practice and receive corrective feedback."
All seventeen briefs were downloaded and counted for this post. The phrase "corrective feedback" appears three to seven times per brief. Every brief carries it at least three times in boilerplate alone: once in the checklist question that opens the document and once or twice in the Dosage definition.
Strip the boilerplate and the picture is layered rather than empty, so it is worth setting out in three steps.
Five of the seventeen say nothing whatever about corrective feedback beyond the definition: Fusion, Leveled Literacy Intervention, ROOTS, Read Naturally Live and Visualizing and Verbalizing.
Seven more say only that it happens. Fraction Face-off! and Pirate Math both say "independent practice is timed, and corrective feedback is provided for incorrect responses." Number Rockets says "corrective feedback is provided for incorrect responses on each independent practice sheet." Hot Math says tutors "provide corrective feedback to repair misconceptions and errors in solution strategies."
Spring Math says the student gets "multiple opportunities to practice solving with immediate corrective feedback until the student is accurate," and lists corrective feedback among the components of explicit instruction. The two Lexia briefs, Core5 and PowerUp, go slightly further and describe a response: students who struggle online "receive immediate corrective feedback and a scaffolding system for support that includes fewer stimuli, more structure, and more explicit instruction."
Six describe a mechanic — something a person or a material actually does at the moment of the error. Only two of those name a prompt or a technique. Here is the whole set, quoted.
Sound Partners is the only brief with a verbatim prompt. For new and harder activities, "the tutor instructions include reminders for correction prompts," and the published example is "Is there an /s/ at the end? Try it again."
Seeing Stars names two techniques, with a worked exhibit for each. The first is "Socratic questioning of student errors to promote self-correction," in which the teacher asks questions that help the student compare the response to the stimulus. The second is "Miscalling for Monitoring," where "the teacher makes an error and prompts the student to take the teacher's role and correct the error."
PACT describes correction as a property of the materials rather than of the teacher: "Scratch-off answer sheets are provided to teams for immediate feedback on accuracy." A correct scratch reveals a star; an incorrect one does not, and the team returns to the question.
Fraction Face-off! specifies it inside a timed flash card activity: "when a student makes an error, that student explains the correct solution strategy and corrects the error as seconds elapse."
The two multiplication briefs say only that "the teacher monitors students and intervenes to provide immediate feedback."
Seeing Stars is therefore the only reading program in the whole set whose brief names its correction techniques as techniques, and one of them inverts the roles entirely by having the adult make the mistake. Sound Partners is the only brief in the set that prints a sentence a tutor would actually say.
So of the eight reading briefs: three describe a mechanic, two more describe a software scaffolding response, and three say nothing beyond the definition. The programs whose briefs are most specific about the correction are, with the exception of Sound Partners and Seeing Stars, the mathematics programs and the software.
Bound this the way it should be bound. These are summary briefs, not manuals. An absence in a brief is an absence in NCII's published summary of a program, not proof that the program's own training materials lack a correction procedure. Several of these programs are only available to trained practitioners, and the program comparison post sets out which ones.
The taxonomy counts the corrections and never asks what they are
This is the finding that organises everything above.
Across all seventeen briefs the taxonomy's Comprehensiveness dimension is rated against named explicit-instruction principles, and each brief prints those principles as its own sub-headings. Extracted across the whole set they come to twenty-eight distinct sub-headings, which reduce to about ten underlying principles.
Those are: explanations or directions in clear, simple language; ensuring the necessary background knowledge and skills; modelling efficient strategies rather than expecting discovery; providing adequate practice opportunities; practice that generates many correct responses; systematic cumulative review; judicious review with mixed problem types; gradually fading support; priming relevant background knowledge; and strategic integration of new with existing knowledge.
Not one of those sub-headings is about the correction. The word "correct" appears in several of them, and in every case it refers to the student generating correct responses or executing a strategy correctly, never to the adult's response to an error.
Where a brief does describe a correction, as Seeing Stars and Sound Partners do, it appears underneath a heading about practice or about clear directions, because the framework offers nowhere else to put it.
So corrective feedback enters NCII's taxonomy exactly once, as the unit in which Dosage is counted, and never as something a program can be rated on. The framework asks how many opportunities to receive corrective feedback a student gets per session, and never asks what the feedback is.
This is the same structure as the finding in the NCII tools chart ratings post, where the rating columns are strongest on the dimensions where a tool is least distinctive. Here the framework is precise about the quantity and silent about the thing being quantified.
And the quantity is reported in units that do not convert
Having counted the corrections, the briefs report the count in incompatible units. Taken from the Dosage section of each brief, with the group size each one assumes:
Fusion, two students: an estimated 114.33 opportunities per student, and at five students, 92.13.
Pirate Math, four students: 106.88.
Fraction Face-off!, four students: 97.14.
Hot Math, four students: 93.75.
Number Rockets, four students: 88.23.
Spring Math, no more than six students: 76.
Sound Partners, one student: 73.
ROOTS, two students: 58, and at five students, 49.
Leveled Literacy Intervention, three students: 53.
The two multiplication briefs, three students: 44 each.
Seeing Stars and Visualizing and Verbalizing, five students: 30 each.
Lexia PowerUp, one student online and two offline: 29.
Lexia Core5, one student online and two offline: 26, computed from three randomly selected lessons.
PACT, fifteen students: "nine opportunities to respond and receive corrective feedback during each class period."
Read Naturally Live: "an estimated 15 opportunities to respond and receive corrective feedback per completed lesson."
Most of those figures are per session. PACT's is per class period. Read Naturally's is per completed lesson, and a lesson in that program spans several sittings. Core5 defines a lesson as one online unit plus one corresponding face-to-face lesson, with a different group size assumed for each half.
A number like 106.88 and a number like 15 look comparable and are not. One is a per-session estimate at a group size of four, the other is a per-lesson estimate in a program where a lesson takes days.
The two briefs that publish figures at more than one group size behave exactly as intuition predicts. Fusion falls from 114.33 at two students to 92.13 at five. ROOTS falls from 58 at two to 49 at five. Within a program, a bigger group means fewer corrections each.
Across programs it runs the other way. Pirate Math at four students is credited with more opportunities per student than Sound Partners at one-to-one, and Spring Math at up to six students with more than Leveled Literacy Intervention at three. So group size predicts correction density inside a program and not between programs, which is worth knowing before a brief's number is used to argue for a smaller group size.
Wilson's published materials describe charting, not prompting
A structured literacy program makes a useful contrast case, because the expectation is that it would specify this tightly.
Wilson Language Training's public page on implementing the Wilson Reading System sets out the ten-part lesson plan, the roughly ninety-minute duration, the two-lessons-per-week minimum and the mastery gates at Wordlist Charting and Dictation. It describes what happens to errors in aggregate — "the errors identified during the Wordlist Charting help to pinpoint trouble spots and patterns" — and it gives no in-the-moment correction procedure, no prompt hierarchy and no wait time.
The publisher's free materials information booklet runs to about 11,270 words and reproduces pages from the student notebook. Counted as whole words, it contains "error correction" zero times, "corrective feedback" zero times and "mistake" zero times.
The plural "errors" appears five times. The closest thing to a procedure anywhere in it is a charting instruction: have the student read fifteen words aloud, "record all errors on the Wordlist Chart," then "discuss and correct all errors with the student."
That is a post-hoc procedure, not an in-the-moment one, and it is a different kind of answer from a three-second rule. The booklet's thirteen references to tapping describe the decoding mechanic a student is taught to use, including the detail that for welded sounds "the student uses three fingers tapping together at the same time."
The bound on this is the same as for the NCII briefs, and it is a real bound. This is an absence in a sales and materials document and on a public implementation page, not proof of an absence in a curriculum that is delivered through accredited training. The same caution applies to every program in the Orton-Gillingham lineage post.
What the research review found: no clear recommendation
The honest answer to "which of these three procedures works better" is that the research does not currently support a clear recommendation, and the most recent review says so in its abstract.
Grønli, Walgermo, McTigue and Uppstad's critical review of teachers' feedback on oral reading, in Educational Psychology Review 36:121 (2024), DOI 10.1007/s10648-024-09957-z, analysed twenty-four empirical studies of teacher feedback in K-5 settings published between 1995 and 2022.
Its opening statement of the problem is unhedged: "research has yet to provide clear recommendations for effective classroom feedback." Its summary of the result is that findings "reveal significant variability in effective feedback types and content, making clear-cut generalizations challenging."
The review's coding rule is worth knowing because it sets the bar it was measuring against. Where seventy percent or more of the outcomes in a domain pointed the same way, the authors coded that direction as the effect; where fewer than seventy percent agreed, they coded it as "no clear effect." They also report that their coding matched the original studies' own presentations of their results with no discrepancies.
On performance feedback the review concludes that the "varied findings from these 14 studies make it challenging to draw clear conclusions." On feedback content it finds that "the strength and direction of these effects are inconsistent across studies, and further high-quality research is required to establish clear patterns." On motivational outcomes, from four studies: "No clear conclusions can be drawn."
The review also notes a vocabulary problem in the research literature itself. Because terminology was inconsistent across researchers, the authors had to treat "reading difficulties," "at risk," "low-achieving" and "poor readers" as pragmatically synonymous and code them all as "struggling readers."
The finding that needs the most careful handling
Within that review is a sentence practitioners should know about and nobody should over-read. Summarising its studies on feedback content, the review reports that some struggling readers "performed best with no feedback," citing three studies.
State clearly what that does not license. It is not evidence that withholding correction is a better procedure, it is not a finding about any named program, and no decision about an individual student follows from it.
It is three studies inside a twenty-four-study review whose headline conclusion is that the direction of effects varies and the evidence is insufficient for clear patterns. The same review reports positive effects for performance feedback combined with goal setting, and for meaning-based feedback on comprehension.
What it does establish is that "more correction is better" is not a safe default, and that a practitioner who has been correcting every error on principle is not following a settled finding.
One large trial did compare feedback types directly
The most recent direct comparison of feedback types is a software trial, which brings its own boundary.
Strik and colleagues report in Technology, Knowledge and Learning 31(3):1741 (2025), DOI 10.1007/s10758-025-09860-8, free in full text at PubMed Central, on 525 first-grade pupils across 44 schools, with 263 receiving explicit feedback and 262 implicit feedback from a Dutch reading tutor built on automatic speech recognition.
Both conditions produced significant positive effects on reading accuracy. Explicit feedback was additionally more beneficial for accuracy gains in reading sentences, and the authors conclude that it "may be more beneficial in early reading interventions."
They also report a trade-off that matters for anyone tracking words correct per minute: improvements in accuracy were often accompanied by slower reading speed.
The boundaries are substantial and belong in the same sentence as the result. The feedback was delivered by software rather than by a person, in Dutch rather than English, to first graders rather than across a range, and in a general classroom rather than an intensive intervention caseload. It is the largest direct comparison available and it is not a study of what a practitioner should say.
A smaller study worth knowing for its design is Wu, Stratton and Gadke's group reading intervention with individualised error correction in the Journal of Special Education Apprenticeship 12(3), 51-68 (2023), DOI 10.58729/2167-3454.1181.
It built a small-group package in which the error correction procedure was individualised per student rather than fixed for the group, and reported strong effects for each of three middle school students with disabilities in an ABAB single-subject design. Three students is three students. The design point — that error correction is a candidate for individualisation rather than a package default — is the transferable part.
What AET's Code of Ethics says about any of this
Once again, and this has now been the finding on every assessment and intervention post published here, the governing document of the profession is silent on the post's subject.
The Association of Educational Therapists' Code of Ethics contains no clause on error correction, no clause on corrective feedback and no clause governing how a therapist responds to a student's mistakes. Nothing in it addresses the conduct of teaching within a session at that level of detail.
What it does contain is two clauses that bear on the choice indirectly. Section Two, I.E requires members to "select and use appropriate instructional methods, curricula, materials, and other resources" to meet each client's needs. Section Two, I.G requires them to "create safe and effective learning environments which contribute to the fulfillment of needs."
Alongside those sit the competence clauses, and they are the operative ones here. Section One, I.B requires members to "provide professional services only within the boundaries of their competence." Section One, II.A requires them to "provide only those professional services for which they have been adequately trained."
Read together with everything above, the ethical position is reasonably clear even though no clause names the procedure. If a program's correction procedure is part of a trained method, the competence clauses point toward using the procedure you were trained in rather than substituting one from a different source. The Code does not say that in those words, and this is a reading of the clauses rather than a quotation of them.
A sequence, not a decision
This site does not tell practitioners which procedure to use, and this post will not either. What follows is the order in which the published sources answer the question, which is a different thing.
Ask what you are doing at that moment. If you are administering a timed measure, the manual's hesitation prompt is the procedure, and it is not optional: three seconds, supply, score it incorrect. Departing from it changes what the number means.
If you are teaching, ask what the program specifies. If you are trained in a program with a published correction procedure, that procedure is the answer, and the AET competence clauses point the same way.
If the program does not specify one, ask what kind of word it was. The federal guide's branch point is whether the student should have been able to read it. A decodable word the student has been taught gets the pause-and-reread branch; an irregular or very challenging word gets told, repeated and reread in context.
Decide in advance whether the correction is scored. The measurement systems score the supplied word wrong. Read Naturally scores it wrong and then practises it. Phrase drill does not score it at all. These are different purposes and mixing them produces a number that means nothing.
Write down which procedure you used. If the procedure varies by session and nothing records which one was used, a progress graph cannot be interpreted. What the codes require of your records is covered in the session notes post.
Do not treat correction density as a dosage proxy. The NCII figures above show per-student correction estimates running backwards from group size, so a brief's number cannot be used to argue that a smaller group delivers more correction.
When several organisations share a vocabulary
Everything above is one more instance of the pattern this site keeps documenting, and the earlier posts are the evidence that it is a pattern rather than a coincidence.
The repeated reading protocol post already covers the error rules inside the repeated reading protocols themselves, including the finding that "error correction" appears zero times in the IES grades 4 to 9 practice guide summary. That section is not restated here. This post takes the cross-program comparison of the correction procedure instead, and the two sit alongside each other.
The mastery criteria post found four programs setting thresholds in at least three units that do not convert. The decodable text post found the field's most-used adjective with no agreed method behind it.
The pseudoword decoding post found two publishers scoring the same response one and zero. The NCII chart ratings post found a rating system strongest where a tool is least distinctive. The group size post found no two programs stating group size in the same unit.
If you are working out where your own role sits among the people who do this work, the role comparison pillar is the place to start.
What is out of scope here
Three things, named explicitly so nobody mistakes their absence for a finding.
This post does not say which correction procedure is more effective. The most recent review of the research says that question is not currently answerable with a clear recommendation, and inventing an answer would be worse than reporting the absence.
It does not address any legal or regulatory obligation. Nothing here concerns what a school district, a state or federal law requires of anyone, and nothing here is legal advice. Those questions belong with an attorney who practises in your state.
It does not describe the correction procedures taught inside accredited program trainings that are not published. Where a program's materials are restricted to trained practitioners, the only honest statement is what the publisher says in public, and that is what is reported above.
Where the published record runs out
The gaps are specific, and listing them is more useful than filling them.
No published body adjudicates between the three procedures. The clearinghouses rate programs on evidence of effect, not on correction procedure. No practice guide compares the procedures against each other. There is no standard a practitioner can point to.
The federal guide's procedure is four sentences long. Recommendation 4 gives a branch point, a prompt pair and a fade instruction. It gives no wait time, no number of rereads, no rule for a second error in the same sentence, and no guidance on what to do when the pause produces nothing.
Wilson's and most other publishers' in-session procedures are not publicly readable. Eleven of seventeen NCII briefs describe no correction mechanic at all, and the public implementation pages for the major structured literacy programs describe lesson architecture rather than prompts. Whether the manuals specify more is not something this engine can verify from outside the training.
NCII's opportunity counts are estimates in incompatible units with no stated method. The briefs say an estimate was computed from sampled lessons and give a figure to two decimal places. They do not publish the counting rule, so the figures cannot be recomputed or compared across briefs with confidence.
The research base is twenty-four K-5 studies with inconsistent terminology. The most recent review had to treat four different descriptions of struggling readers as synonyms to code them at all, and found fewer than seventy percent agreement in most domains it examined.
Nothing published addresses the caseload this site is written for. Every source above assumes a classroom, a school tutoring program or a research protocol. None of them is about a once- or twice-weekly private session, and the transfer is an inference rather than a finding.
Frequently asked questions
Should I tell a student the word or let them work it out?
The published sources disagree, and which one applies depends on what you are doing. During a timed measure, both DIBELS 8 and Acadience require you to supply the word after three seconds and score it incorrect.
During instruction, the IES foundational skills guide says not to simply tell the student the correct word when it is a word they should be able to read, and to have them reread the sentence instead; it does say to tell them the word when it is very challenging or irregular.
Is "three seconds, then tell them" an instructional rule?
It is a measurement rule in both major curriculum-based measurement systems, where it exists so a timed sample stays comparable. Read Naturally uses the same rule inside its timings and adds a student-marking and practice step afterwards. No source in this post presents the three-second supply as a teaching procedure on its own.
Does the What Works Clearinghouse use the term "corrective feedback"?
Not in the foundational skills practice guide. In 58,915 words that guide uses "feedback" thirty times and the phrase "corrective feedback" zero times, and the phrase "error correction" appears exactly once, inside a description of a reviewed study rather than in a recommendation.
What does NCII's taxonomy say about what a correction should be?
Nothing. Corrective feedback appears in the taxonomy only as the unit in which the Dosage dimension is counted. None of the twenty-eight explicit-instruction sub-headings used across the seventeen briefs is about the adult's response to an error.
Which programs publish an actual correction prompt?
Of the seventeen programs with NCII taxonomy briefs, six describe any correction mechanic and eleven do not. Sound Partners publishes a verbatim prompt. Seeing Stars names two techniques, Socratic questioning and Miscalling for Monitoring. PACT locates the feedback in self-revealing materials. Fraction Face-off! and the two multiplication briefs describe it inside a timed activity.
Does AET require a particular correction procedure?
No. Its Code of Ethics contains no clause on error correction or corrective feedback. The clauses that bear on the question are the requirement to select appropriate instructional methods and the competence clauses requiring members to work within their training and to provide only services for which they have been adequately trained.
Is there evidence that correcting every error is best?
No, and the most recent review is explicit that the research does not yet support clear recommendations about feedback type or content. That review also reports three studies in which some struggling readers performed best with no feedback, which is not a basis for withholding correction and is a reason not to treat maximal correction as a settled default.
How should I record which procedure I used?
Whatever your note format, record the procedure alongside the data, because a progress graph produced under a changing correction procedure cannot be interpreted. No published code requires this; it follows from the fact that the measurement systems score a supplied word as an error and the instructional procedures do not score it at all.
Join the Illuminate educator network
Illuminate Educational Therapy Group publishes this kind of source-level comparison for practitioners: educational therapists, learning specialists and executive function coaches who need to know what a published procedure actually specifies before they rely on it.
If that is the work you do, join the Illuminate educator list and new practitioner resources will reach you as they publish.


Comments