top of page

When to Discontinue Reading Intervention: What the Record Says

Writer: Andrea Chernin
Andrea Chernin
Oct 24
24 min read

When to Discontinue Reading Intervention: What the Record Says

The short version


If you are trying to work out when to discontinue reading intervention for a student on your caseload, the published record will not settle it for you, and it is worth knowing exactly where it stops being helpful. Three different decisions get called "when to stop."


The programs answer one of them. The federal record mentions the second once, in a sentence that begins "In our experience." Nobody answers the third except the ethics code of the profession that credentials you, and what it tells you to weigh is mostly not about reading.


  • The programs tell you when the material runs out, not when the student is finished. Barton's own FAQ recommends taking an elementary student "all the way through the end of Level 8" and marks the endpoint of the system as the point where "there are no more rules to teach." That is a statement about the curriculum.

  • One publisher in this group attaches numbers to the end of each level, and calls them "Outcomes." McGraw Hill's Corrective Reading brochure gives Decoding A as "Read at 60 words per minute at about a 2.0 grade level with 98 percent accuracy," and three further levels at 90, 130 and 150 words per minute. The word "exit" appears nowhere in it.

  • The string "exit" occurs exactly once in the 2009 federal reading intervention guide, inside the word "complexity." It occurs zero times in the 2022 guide for grades 4 through 9, zero times on NCII's data-based individualization page, and zero times in NCII's decision-rules handout.

  • NCII's handout gives nine possible actions and the best one is "increase the goal." None of the nine is to stop.

  • So the de facto rule is "the student met benchmark" — and the publisher of the benchmark says a student right at it has roughly a seventy to eighty-five percent chance of meeting the next one. The student you are most likely to exit is the one for whom the benchmark predicts least well.

  • AET's Code of Ethics is the one document in this landscape that squarely addresses ending the work, and three of the four things it tells you to weigh are the relationship, the fit and the contract.

Three different decisions get called "when to stop"


A parent asks how much longer this will take. A school asks whether the student can come off the intervention list. You ask yourself, somewhere around the eighteenth session, whether what you are doing is still the best use of the hour.


Those are three questions, not one.


The first is has this student finished the program? That is a question about a curriculum, and the publisher of the curriculum is the right authority on it. It has an answer.


The second is does this student still need intervention? That is a question about the student, and it is the one every search engine result assumes you are asking. It has almost no published answer at all.


The third is should this engagement end? That is a question about a professional relationship with a family who is paying you, and it is the only one of the three that a code of ethics addresses.


Confusing the first with the second is the common error, and it runs in both directions. A student who completes Level 8 of a program has completed Level 8 of a program. A student who hits a benchmark on a one-minute measure has hit a benchmark on a one-minute measure. Neither event is a finding about whether the underlying difficulty has resolved, and no document in this article claims otherwise.

What the programs actually publish, and only one of them is a number


The programs in general use among educational therapists publish an end to their own material and, with one exception, nothing about the student.


Barton: the endpoint is a property of the curriculum


Barton Reading and Spelling System. Its FAQ page, read on 24 October 2026, says a tutor working through a school year "will need at least Levels 1, 2, and 3," recommends taking an elementary-school student "all the way through the end of Level 8," and describes Levels 9 and 10 as ideally taught before a child enters high school.


On the endpoint of the system it says "there are no more rules to teach." For advancement inside the system it specifies about ninety-five percent accuracy, which is a mastery criterion rather than an exit criterion, and which this site has written about separately.


Read that list again and notice what is absent. There is a recommended stopping point, and it is defined by the student's school level and by the exhaustion of the content. It is not defined by a measurement.


Corrective Reading: the only numbers in the set


Corrective Reading. McGraw Hill's overview brochure is the only document surveyed here that puts a number on the end of a level, and it labels those numbers "Outcomes":


"Outcomes: Read at 60 words per minute at about a 2.0 grade level with 98 percent accuracy." Decoding B1 is given as 90 words per minute at about a 3.0 to 3.5 grade level, B2 as 130 words per minute at about 4.0 to 4.5, and C as 150 words per minute at about a 7.0 grade level, each with 98 percent accuracy.

Three things about that. First, it is an outcome claim, not a discharge rule: the brochure says what a student who finishes the level is expected to be able to do, not what you should do next. Second, the word "exit" does not appear in the brochure at all, and neither does "discontinue."


Third, two of the three quantities are grade equivalents, which are the one score type that the test publishers' own documentation warns practitioners away from, as this site has set out at length.


This is also a correction of something written on this site. An earlier post described these figures as Corrective Reading's "exit criteria." They are not labelled that way by the publisher, and the distinction is the whole subject of this article, so it is worth fixing in print rather than quietly: they are published outcomes for a level of a program.


Wilson: nothing public on sufficiency


Wilson Reading System. Wilson's materials specify what a certified instructor must do and how often, and this site has covered its implementation requirements and the time it asks of you. No publicly readable Wilson page located for this article states a criterion for deciding that a student no longer requires the program, as distinct from having completed its Steps. That absence is reported as an absence, scoped to what is public.

Read Naturally's threshold is set by you, and the manual says why


The clearest published statement of how one of these thresholds is actually chosen comes from Read Naturally, and it is unusually candid.


From the Read Live teacher's manual, verbatim:


"In addition to selecting a level and series, the teacher must set an individualized, developmentally appropriate goal reading rate for each student. This Words Correct Per Minute (wcpm) goal is the number of words a student must read correctly in one minute to pass a story. The goal needs to be sufficiently high to require the student to practice reading the story many times before achieving the goal rate, but not so high that the student becomes frustrated."

So the threshold is set by the practitioner, it governs passing a story rather than finishing an intervention, and the design criterion the publisher gives for choosing it is motivational: high enough to force repeated practice, low enough to avoid frustration.


That is a defensible and sensible way to set a practice target. It is not a measurement of sufficiency, and the manual does not pretend it is. Anyone converting a Read Naturally goal rate into evidence that a student is ready to stop is using the number for something its publisher never built it to do.


The wider problem with treating a rate figure as a destination is the subject of this site's piece on oral reading fluency norms.

The word "exit" is almost absent from the federal record


Each of the documents below was downloaded and searched directly for this article on 24 October 2026. The counts are of the literal strings in the extracted text.


  • Assisting Students Struggling with Reading: Response to Intervention and Multi-Tier Intervention in the Primary Grades (NCEE 2009-4045), about 24,500 words: the string "exit" appears once, and it is inside the word "complexity" in the appendix. As a word, it does not appear.

  • In the same guide, "discontinue" appears twice and neither is about a student. One is "regression discontinuity," a study design. The other instructs teachers to discontinue a phonemic awareness activity when a student reaches 100 percent accuracy on all of its items.

  • Providing Reading Interventions for Students in Grades 4–9 (WWC 2022007): "exit" zero, "discontinue" zero, "dismiss" zero, "benchmark" zero, "criteria" zero, "no longer" zero, "duration" zero.

  • All six occurrences of "stop" in that guide are instructional — "stop at periods," "stop periodically to ask the group to think about whether they are understanding the text." It tells you four things to do and never how long to do them for.

  • NCII's data-based individualization page: "exit" zero, "discontinue" zero, "fade" zero. The published process is a five-step loop — validated program, progress monitoring, diagnostic assessment, intervention adaptation, progress monitoring — and the loop has no exit.

  • NCII's academic intervention tools chart: "exit" zero.

  • The Acadience Reading K–6 benchmark goals handout: 581 occurrences of "benchmark" and zero of "exit."


Set those together and the picture is not that the field disagrees about when to stop. It is that the question is barely posed.

The one federal recommendation that covers it is rated at the lowest tier


There is one place in the federal reading record where the decision is named as a decision, and it is Recommendation 4 of the 2009 guide:


"Monitor the progress of tier 2 students at least once a month. Use these data to determine whether students still require intervention. For those students still making insufficient progress, school-wide teams should design a tier 3 intervention plan."

Read the second sentence closely. It assigns the decision and specifies nothing about how to make it. The recommendation's own how-to section is where the only operational instruction appears, and it is a participial clause in a list: schools should establish a schedule to assess tier 2 students at least monthly, "reassigning students who have met benchmarks," graphing progress reliably, and regrouping students who need continued support.


That clause is, as far as this search found, the whole of the federal instruction on exiting a reading intervention. It outsources the decision to whatever benchmark your screening system publishes.


And the recommendation carrying it is rated at the bottom of the scale. The 2009 PDF labels it Low; the current What Works Clearinghouse site renders the same tier as Minimal Evidence.


Three other practical details sit in the same recommendation and are worth having: monitoring "should occur at least eight times during the school year"; the panel suggests regrouping tier 2 students approximately every six weeks; and the guide states that "little evidence demonstrates that weekly measures are superior to monthly ones."

And the sentence that names the decision begins "In our experience"


The evidence paragraph under Recommendation 4 is the most quotable thing in this whole area, and it says less than practitioners assume:


"The panel rated the level of evidence as low. Only three studies of tier 2 interventions that met WWC standards or that met standards with reservations included a weekly progress monitoring or unit mastery component. However, neither of the studies evaluated progress monitoring as an independent variable. Thus, no inferences can be drawn about its effectiveness based on the research reviewed."

There is an internal slip in that passage worth noting in passing: it says "only three studies" and then "neither of the studies," and its footnote names three citations. The arithmetic does not resolve, which is a small thing, but it is in the document.


What follows is the part that matters:


"Despite the lack of evidence supporting use of progress monitoring, the panel decided to recommend this practice for students in tier 2 interventions. In our experience, progress monitoring data are used to determine students' response to tier 2 instruction and to inform instructional decisions in two ways."

The second of those two ways, a sentence later, is this: "It also identifies students who no longer require tier 2 instruction."


So the single place in the federal reading record that says a student may no longer need the intervention is a sentence prefaced "In our experience," written immediately after the guide states that no inferences can be drawn about the practice from the research reviewed.


That is the panel being scrupulous about its own warrant, and it deserves to be read as such rather than as a weakness. But it is panel experience, and it is the strongest thing there is.


This is the third time a post on this site has found the field's most-repeated operational rule sitting in the federal record as the opinion of a panel. The math fact fluency claim was the second. The pattern is now predictable enough to check for.

NCII's decision framework has nine outcomes and none of them is stop


NCII's Decision Rules for Analyzing Academic Progress Monitoring Data handout, dated 2024, is about 690 words long and describes three methods: four-point analysis, trend line analysis, and median of the last three data points. Each method has three possible readings of the graph, so the handout specifies nine actions in total.


Here is the complete set of things it tells you to do:


  • If the data are above the goal line: increase the goal.

  • If the data are below the goal line: consider adapting the intervention, and for two of the three methods, assess fidelity.

  • If the data straddle the goal line: keep collecting data and continue to monitor progress.


That is all nine. The best available outcome in the framework — the student is outperforming the goal line on every recent point — produces a harder goal. There is no state in which the intervention ends.


This site's post on progress monitoring decision rules traces where the four-point rule came from and what happened when it was simulated. The finding here is narrower and structural: whatever the rules' accuracy, the framework they belong to has no terminal state, so a practitioner who runs it faithfully will never be told by it to stop.

So the benchmark is doing the work


Put the two previous sections together and you get the actual mechanism. The federal instruction is to reassign students who have met benchmarks. The decision framework never reaches a stop. So in practice the exit decision is made by a benchmark published by a screening or progress-monitoring vendor, and inherits whatever that benchmark was built to do.


It is worth knowing exactly what it was built to do. From the Acadience Reading K–6 benchmark goals handout, dated 11 March 2021:


"Acadience Reading benchmarks are empirically derived, criterion-referenced target scores that represent adequate reading skill for a particular grade and time of year." And: "A benchmark indicates a level of skill at which students are likely to score above the 40th percentile on any high-quality reading assessment and achieve the next Acadience Reading benchmark or reading outcome."

The external criterion was the Group Reading and Diagnostic Evaluation, with its 40th percentile treated as the indicator of adequate skill, in a study run in the 2009 to 2010 school year across thirteen elementary and middle schools in five states.


DIBELS 8th Edition frames its own cut scores the same way, as levels of support rather than as thresholds of sufficiency. Its legend defines the minimum score for core support as the point where "about 80% of students who score at or above the 40th percentile rank on criterion measure fall in this range or above."


So a benchmark is a conditional probability about a later test. It is not a statement that a difficulty has resolved, and both publishers say so in their own words. The related problem of reading a single score as if it were exact is the subject of this site's piece on the standard error of measurement, and the related problem of comparing systems is covered in progress monitoring tools compared.

The student you are most likely to exit is the one the benchmark predicts least well


This is the sharpest thing in the article and it is arithmetic anyone can check in the publisher's own handout.


Acadience publishes the likelihood attached to each band. For students at or above benchmark, "the overall likelihood of achieving subsequent reading benchmarks is approximately 80% to 90%." Its Table 1 splits that: Above Benchmark is given as 90 to 99 percent, and At Benchmark as 70 to 85 percent.


And the handout states the gradient explicitly:


"Within this range, the likelihood of achieving subsequent benchmarks is lower for students whose scores are right at the benchmark and increases as scores increase above the benchmark." And: "A score at or above the benchmark indicates that the odds are in the student's favor of achieving the next benchmark, but it is not a guarantee. For example, if students at or above the benchmark have an 85% chance of meeting the next benchmark, that means that 15% of students in the At or Above Benchmark range may not achieve the subsequent benchmark."

Now consider who actually gets exited. Not the student well above benchmark, who was never in your intervention. The student who just crossed it. That student sits in the part of the band the publisher says predicts least well, with a figure that may be as low as seventy percent, and the handout adds that "some students who achieve scores at or above the benchmark may still need supplemental support."


None of that makes the benchmark a bad instrument. It makes it an instrument whose published accuracy is lowest exactly where the exit decision is taken, which is a fact about the decision rather than about the test.

Do different exit criteria agree? Somebody checked


They did, in 2008, and the answer is more useful than a yes or a no.


Barth, Stuebing, Anthony, Denton, Mathes, Fletcher and Francis examined an existing database of 399 first-grade students against the cut-points, measures and methods then cited for identifying inadequate responders, computing a series of 543 two-by-two measures of association across 808 total comparisons. From the abstract:


"The results indicate that agreement is generally poor and that different methods tend to identify different students as inadequate responders, although agreement for identifying adequate responders is higher. Approaches to the assessment of responder status must use multiple criteria and avoid formulaic decision making."

Read the clause most people skip. Agreement for identifying adequate responders is higher. The paper also says why: "At 1.0 and 1.5 SD, the agreement for adequate responders is higher, but agreement for inadequate responders is generally poor," because overall agreement rises with the base rate of adequate responders while kappa stays low. The kappa values in its comparison tables sit in the low-to-middle 0.4s with overall agreement around 0.68 to 0.77.


The boundary matters and must travel with the finding. This is one database of 399 first graders, published in 2008, addressing the identification of learning disabilities rather than the termination of a service, and it is a study of agreement between criteria rather than of whether any criterion is correct. You may not take from it that a particular exit rule is accurate.


What you may take from it is the authors' own instruction, which is squarely about how to decide: use multiple criteria, and avoid formulaic decision making. That is the opposite of reading one number off one graph.

What the follow-up research shows about the months after you stop


The question behind the exit decision is not really whether the student has met a threshold. It is whether the gain will hold once the sessions end. There is a meta-analysis on exactly that.


Suggate (2016), in the Journal of Learning Disabilities 49(1), 77 to 96, assembled 71 intervention-control groups with both posttest and follow-up data, N = 8,161 at posttest, with a mean follow-up interval of 11.17 months. The headline result:


"The posttest effect sizes indicated effects (dw = 0.37) that decreased to follow-up (dw = 0.22). Overall, comprehension and phonemic awareness interventions showed good maintenance of effect that transferred to nontargeted skills, whereas phonics and fluency interventions, and those for preschool and kindergarten children, tended not to."

Three moderator patterns in it are directly relevant to a private caseload.


Three moderators that bear on a private caseload


Age at intervention ran the opposite way to the received wisdom. Effect sizes fell from 0.34 to 0.12 for preschool and kindergarten children, from 0.40 to 0.26 for grades 1 and 2, and rose from 0.35 to 0.43 for children in grades 3 to 4.


The author's own summary: the interventions "were more than 3.5 times more effective for older children at follow-up." That bears on the age band covered in this site's post on reading intervention for older students.


Intervention in addition to typical instruction held up better than intervention in place of it. The reported contrast is dw = 0.44 against 0.26 at posttest, and 0.31 against 0.19 at follow-up. Private educational therapy is almost always additive rather than substitutive, which puts it on the better side of that comparison.


The approach whose effects were least well maintained is the approach most of this field delivers. Phonics and fluency work lost more effect to follow-up than comprehension and phonemic awareness work did, and fluency interventions in particular, in the author's words, "did not result in good transfer to reading comprehension, with benefits being more confined to targeted decoding and reading skills at follow-up."


Now the boundary, in the same breath, because the discipline matters more than the headline. These are between-study moderator comparisons across 71 contrasts, not randomised comparisons of one approach against another inside a single trial.


The author's own verb is "tended." The finding does not license telling a family that structured literacy work does not last, and it does not license choosing a comprehension program over a decoding program for a student who cannot decode.


What it supports is a narrower and more useful claim: the maintenance of a gain is a separate empirical question from the production of it, and the published maintenance picture is weakest for exactly the kind of work an educational therapist is most likely to be doing. That is a reason to plan the end of an engagement deliberately, not a reason to doubt the work.

The booster finding, and what it does not license


One line in the same paper is counterintuitive enough that it should be reported and then handled carefully. Among the moderators, "offering a booster intervention was associated with lower effect sizes." The author offers the hedge himself: "One possibility for this counterintuitive finding is that more of the same may not work."


That is a single moderator in a between-studies analysis, and studies that built in a booster were plausibly studies anticipating fragile gains in the first place. It is not evidence that returning for a second round of work is harmful, and nobody should tell a family that.


It is, though, the only published signal in this area about going back for more, and it points the same way as the author's hedge: if a student needs a second engagement, the interesting question is whether it should be the same work again or different work. That is the question this site's pieces on treatment planning and program comparison are for.

What AET's Code of Ethics says, and this is the one decision it does address


Four consecutive posts on this site have found AET's Code of Ethics silent on the method question they were asking about. On records it contains "documentation" zero times. On grouping it never uses "group" in a service sense. On thresholds it contains "mastery" zero times and "criterion" zero times.


On stopping, it speaks. That is the reversal worth noticing.


The Code of Ethics and Standards for the Professional Practice of Educational Therapy, Revision #9, 2025 runs to roughly 2,980 words. Searched on 24 October 2026 it contains "exit" zero times, "discontinue" zero times, "dismiss" zero times, "no longer" zero times, "criteria" zero times, "benchmark" zero times, "mastery" zero times and "proficient" zero times. It uses "terminate" or "termination" four times, across three clauses.


The substantive one sits in Section Two, under the standard on baseline data, as its third sub-item — the clause this site conventionally points at as II.I.N.3. Verbatim:


"terminate services upon consideration of the student's best interests, goodness of fit between the skill set of the educational therapist and the needs of the client, any contractual agreement, and/or the established educational therapy goals. It is recommended that the final termination session(s) be held in the manner in which sessions have typically been conducted throughout the educational therapy process to allow time for any final assessments and for closure between student/client and educational therapist."

Count the considerations. There are four: best interests, goodness of fit between your skills and the client's needs, any contractual agreement, and the established goals. Only the fourth is about the student's learning. One is a judgement about welfare, one is a judgement about you, and one is a document.


The two clauses that put the family in the decision


Two further clauses complete the picture. Section Two, II.D requires seeking parents' and guardians' perspective and expertise "in planning, conducting, and evaluating services, as well as determining optimum time for termination of services to clients." Section Two, III.B requires consulting the family "in evaluating the appropriateness, initiation, continuation and/or termination of related services."


So AET's answer to "when should this stop" is procedural and relational rather than psychometric: weigh four things, one of which is your own fit, and decide it with the family rather than for them.


For a practitioner who has been looking for a number, that reads at first like an evasion. It is better understood as the code answering the third question in this article rather than the second — and it is the only document here that answers any of them squarely.


Two smaller observations. The guidance on how to conduct the final sessions is prefaced "It is recommended," so it is advisory where the rest of the clause is not. And the clause's placement is odd: the only standard about ending the work is filed as a sub-item of the standard about keeping accurate baseline data.


The adjacent sub-item, II.I.N.1, requires establishing ways to measure client progress "towards those goals at regular intervals," which is the nearest the code comes to the measurement question and still specifies no unit. This site's post on session notes and progress reports covers the documentation half of the same clause.

One code in this landscape does specify it in terms of need


For contrast, the American Psychological Association's Ethical Principles of Psychologists and Code of Conduct states a stopping criterion directly. Standard 10.10(a), Terminating Therapy:


"Psychologists terminate therapy when it becomes reasonably clear that the client/patient no longer needs the service, is not likely to benefit, or is being harmed by continued service."

Three grounds, and the striking thing is that two of them are about a student who has not improved. "Is not likely to benefit" and "is being harmed by continued service" are exit criteria for a case that is going badly. AET's four considerations contain no equivalent of either, which is a real difference between the two documents and not a defect in either.


Standard 10.10(c) adds that, except where precluded by the client's actions or a third-party payor's, "prior to termination psychologists provide pretermination counseling and suggest alternative service providers as appropriate." Set beside AET's recommended final session held in the usual manner to allow for closure, the two codes are asking for similar conduct and different reasoning.


A sourcing limitation, stated because it is load-bearing. The apa.org ethics pages serve a security challenge to automated clients, so this text was read from a university-hosted reproduction of the code as published in American Psychologist in December 2002, volume 57, number 12, pages 1060 to 1073.


The code carries later amendments, so if a standard number is load-bearing for you, confirm the current numbering at apa.org directly. This is the same limitation this site stated in print when writing about informed consent across five codes.

When several organisations share a vocabulary


The word "criteria" is doing different work in every document above, and the pattern is the one this site keeps finding. The program's criterion is a property of its scope and sequence. The screener's criterion is a conditional probability about a later test. The decision rule's criterion is a comparison with a goal line somebody drew. The code's criterion is a judgement about a relationship.


It is the same structure this site has found seven times already:



Checking whether two organisations using one word mean one thing is the cheapest reliable way to find out something true about this field.

A sequence, not a decision


What follows is not a rule for when to stop, because no document in this article supports one. It is the order in which the published material can actually be used, and the point of writing it as a sequence is that each step answers a question the next one depends on.


  1. Name which of the three decisions you are making. Has the student finished the program, does the student still need intervention, or should this engagement end? Write down which one. The evidence available differs completely between them.

  2. Find out what your program actually says, and in what unit. Levels remaining, lessons remaining, an outcome statement, a pass rate for a story. Note whether the figure is about the student or about the material.

  3. Separate the program's endpoint from the student's need. Finishing Level 8 is an event in a curriculum. Treat it as the completion of a plan, which is what it is.

  4. If you are using a benchmark, read its publisher's own probability statement. At benchmark is not the same band as above benchmark, and the publisher prints the difference.

  5. Use more than one criterion, because the only study that checked says to. A benchmark plus a curriculum-based measure plus a work sample plus the teacher's report will disagree, and the disagreement is information.

  6. Ask what the next eleven months look like without you. That is the interval the follow-up research actually covers, and the question of whether a gain holds is separate from whether it was produced.

  7. Plan the ending as part of the work, not as its cancellation. AET recommends a final session conducted in the usual manner, with time for final assessments and closure. That costs one session and is the only thing in this whole area that every document surveyed agrees about.

  8. Decide it with the family, and write down the reasoning. The code requires the first. Your own notes are the only record that will exist of the second.

What is out of scope here


This article is about what programs, clearinghouses, publishers and professional bodies have published about ending an intervention. Several adjacent questions are governed by law, regulation or contract rather than by any of those documents, and they are deliberately not answered here.


Exiting a school-based intervention and ending a private engagement are different acts with different authorities. Out of scope here: whether a student may be removed from a school's intervention or multi-tiered support programme, and what a district's own exit protocol requires.


Also out of scope: how special education eligibility or an individualised education programme is determined or ended, what a state's reading or literacy legislation specifies about promotion, retention, exemptions or exit, and what any contract or insurer requires of you. All of that is statute, regulation, district policy and contract.


Where a state publishes a reading plan with criteria in it, that is published policy, and policy is not the same thing as a professional obligation on you.


Nothing here is legal advice. For anything in the previous paragraph, the people to ask are your state education agency, the student's school team, and an attorney who practises in your state.

Where the published record runs out


Stated plainly, because the gaps are the most useful part of this subject.


  • No program surveyed publishes a criterion for deciding that a student no longer needs it, as distinct from having completed its material. That is reported as an absence across the publicly readable pages located for this article, not as a claim about every document any of these publishers has ever issued.

  • Wilson's public pages were searched and no such criterion was found. Scoped to what is public on 24 October 2026.

  • No peer-reviewed study was located that directly compares exit decision rules on the outcome that matters — whether students exited by rule A hold their gains better than students exited by rule B. Barth and colleagues compared criteria against each other for agreement, not against an outcome, and Suggate's meta-analysis reports what happened after interventions ended without comparing how the decision to end them was made.

  • The follow-up literature reports group averages at a mean of about eleven months. It says nothing about an individual student, and nothing about intervals longer than that for most of its contrasts.

  • The federal instruction is one participial clause in a recommendation rated at the lowest evidence tier, and the sentence naming the decision is prefaced "In our experience."

  • Four domains have no exit guidance in the sources checked at all: written expression, spelling, mathematics intervention and executive function coaching. The reading side of this question is thin; those are empty. This site's posts on written expression assessment and spelling assessment cover what is measured, not when to stop measuring it.

  • The source universe here is small, and that is a property of the subject. This article cites fourteen external sources across twelve domains, below the thirty-domain mark this site has reached on broader topics.

  • The whole readable universe is small too: two federal practice guides, two NCII documents, two benchmark publishers' handouts, four program publishers, two professional codes and two research papers. Padding it would have meant citing other people's blog posts about these documents instead of the documents.

Common questions


Is there a published rule for when to discontinue reading intervention?


No. The nearest thing in the federal record is a single clause in Recommendation 4 of the 2009 practice guide instructing schools to reassign students who have met benchmarks, and that recommendation is rated Low in the guide and Minimal on the current What Works Clearinghouse site. NCII's decision-rules handout specifies nine possible actions and none of them is to stop.


Does finishing a program mean the student is finished?


It means the student has finished the program. Barton's FAQ marks the endpoint of its system as the point where "there are no more rules to teach," which is a statement about the curriculum. Corrective Reading publishes numeric outcomes for each level and labels them outcomes rather than exit criteria. No publisher surveyed here claims that completing its material establishes that a student no longer needs intervention.


If a student hits benchmark, can I stop?


You can, and that is what the federal guidance's one clause points at. Know what you are relying on when you do. Acadience publishes a likelihood for each band, and puts students at or above benchmark at roughly 80 to 90 percent likelihood of meeting the next one.


It then splits that figure: 90 to 99 percent above benchmark, and 70 to 85 percent at benchmark. It states that the likelihood is lowest for students whose scores are right at the benchmark, and says in terms that a score at or above benchmark "is not a guarantee."


Do the different exit criteria agree with each other?


Partly, and more for exit than for escalation. Barth and colleagues, using 399 first graders and 808 comparisons, found that agreement among response criteria is generally poor for identifying inadequate responders but higher for identifying adequate responders. Their own recommendation is to use multiple criteria and avoid formulaic decision making. That finding is about identifying learning disabilities rather than terminating a service, and it is one database from 2008.


Will the gains hold after I stop?


On average, partly, and the published picture varies by approach. Suggate's meta-analysis of 71 intervention-control groups found effects of dw = 0.37 at posttest falling to dw = 0.22 at a mean follow-up of 11.17 months, with comprehension and phonemic awareness interventions maintaining better than phonics and fluency interventions. Those are between-study comparisons rather than head-to-head trials, so they do not license a claim about any particular program or student.


What does AET actually require?


Its Code of Ethics Revision #9, 2025 tells you to terminate services upon consideration of the student's best interests, the goodness of fit between your skill set and the client's needs, any contractual agreement, and the established educational therapy goals.


It also requires you to seek the family's perspective in determining the optimum time for termination, and to consult the family in evaluating continuation or termination of related services. It recommends holding the final sessions in the manner sessions have usually been conducted, with time for final assessments and closure. It sets no measurement, no threshold and no interval.

Related reading


If this piece was useful, the posts nearest to it are the ones on progress monitoring decision rules, which traces where the four-point rule came from; mastery criteria, which asks who set the percentages the manuals advance students on; and reading program group size, which does the same job for how many students the manuals permit in the room.


Further out, what is educational therapy maps the scope of the role, and is educational therapy evidence-based sets out what the record does and does not hold about the model itself.

If this is the work you do


Illuminate Educational Therapy Group runs a referral network for educational therapists, learning specialists and executive function coaches. If you want the next piece of work like this when it is published, and to be considered for student referrals, you can join the educator list.

Recent Posts

See All
Reading Program Group Size: What the Manuals Permit

Wilson names four students and six with a skilled teacher. Barton caps it at three and says you will not save any time. The 2022 federal guide counts group size 116 times and recommends nothing. AET's

 
 
 

Comments


bottom of page