top of page

Basal and Ceiling Rules: What the Manuals Actually Say

Writer: Andrea Chernin
Andrea Chernin
5 days ago
22 min read

Basal and Ceiling Rules: What the Manuals Actually Say


You start the Word Attack subtest at the suggested item for a nine-year-old. He gets four right, then six wrong. You stop, count, and write down a raw score. Then you turn to the next test in the same book, and the rule has changed.


That is not carelessness on your part. Basal and ceiling rules are the least discussed and most consequential thing an examiner does, and the publishers do not define them the same way — not between publishers, not between batteries, and in at least one case not even between tests sitting in the same spiral-bound test book.


The rule decides how many items the student sees. The items the student sees decide the raw score. The raw score decides the standard score, the percentile, the confidence band and whatever the band is used for.


So a rule nobody writes about is sitting underneath every number on your report.


This article is about the stopping rules themselves — start points, basals, ceilings, reversals, discontinue rules and what an unadministered item is worth. It is not about what a permitted change to administration looks like, which is Test Accommodations: What the Assessment Manual Permits, or about the error around the score once you have it, which is Standard Error of Measurement: What That Band Means.

The governing volume does not contain the word


Start where you would expect an authority. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) is the field's governing document, it is 240 pages, and since 2021 it has been open access. The full text was downloaded and searched on 16 November 2026.


Counted as whole words on the flow-order extraction of roughly 133,000 words:


  • "basal" — zero.

  • "discontinue" — zero.

  • "reverse rule" — zero.

  • "start point" — zero.

  • "ceiling" — exactly once, and it is not the administration rule. It is "floor or ceiling effects", in a passage about what goes wrong when individual scores are combined into an aggregate index.

  • "stopping rule" — exactly once. It is about adaptive testing: "When tests are administered adaptively, test length (the number of items administered to each examinee) is determined by stopping rules, which may be based on a fixed number of test questions or may be based on a desired level of score precision."

  • "routing" — twice, both describing multistage adaptive testing.

  • "raw score" — fifteen times.


So the volume that governs testing in the United States has nothing to say about the procedure that produces the raw score, under any of the names a practitioner would use for it.


Except in one place, and the exception is the finding


There is a single occurrence of "starting point" in the whole book, and it is worth quoting in full because of what it is scoped to. Standard 4.3:


"Test developers should document the rationale and supporting evidence for the administration, scoring, and reporting rules used in computer-adaptive, multistage-adaptive, or other tests delivered using computer algorithms to select items. This documentation should include procedures used in selecting items or sets of items for administration, in determining the starting point and termination conditions for the test, in scoring the test, and in controlling item exposure."

Read that against what you do on a Tuesday. A documented rationale and supporting evidence for the starting point and the termination conditions are required — when a computer selects the items. When a human examiner selects them, by applying a printed rule about consecutive correct and incorrect responses, the Standards ask for nothing.


That is not a drafting oversight, and it is worth being fair about why. The volume is explicit elsewhere about where the responsibility sits. Standard 6.1: "Test administrators should follow carefully the standardized procedures for administration and scoring specified by the test developer and any instructions from the test user."


Administration procedures are delegated to the developer, wholesale, and the Standards then say nothing about what the developer must publish or justify about them.


The effect is a gap with a precise shape. Where an algorithm picks the items, the rationale must be documented and evidenced. Where you pick them, the rule must only be followed.

Which leaves the question of what the developers actually specify. They specify a great deal, and they do not agree.

One publisher, one battery, four different regimes


The Woodcock-Johnson is the clearest case, because Riverside publishes its own examiner training workbooks and they set the rules out as a single figure.


Start with the publisher's own statement of what a basal and a ceiling are for, because it is the clearest rationale any publisher in this comparison gives:


"The purpose of basal and ceiling requirements is to limit the number of items administered but still be able to estimate, with high probability, the score that would have been obtained if all items were administered."

That sentence is doing something easy to miss. The raw score produced by a basal-and-ceiling administration is an estimate of the score the student would have got on the whole test. It is not a count of what the student did. It is a count of what the student did, plus a credited block of items nobody administered, chosen by a rule.


Now the rules. On the WJ IV Tests of Achievement, per Figure 1 of the workbook:


  • Basal of 6 consecutive correct, ceiling of 6 consecutive incorrect — Letter-Word Identification, Spelling, Passage Comprehension, Calculation, Word Attack, Number Matrices, Editing, Spelling of Sounds, Science, Social Studies and Humanities. Eleven tests.

  • Basal of 5 consecutive correct, ceiling of 5 consecutive incorrect — Applied Problems. One test.

  • No basal or ceiling at all; follow continuation instructions or administer a pre-selected block — Writing Samples, Oral Reading, Reading Recall.

  • Start at Item 1 and stop at the time limit — Sentence Reading Fluency, Math Facts Fluency, Sentence Writing Fluency, Word Reading Fluency.

  • And Reading Vocabulary, which has two subtests, applies basal and ceiling rules to each subtest separately, at five and five.


Four regimes and two different consecutive counts, inside one test book. The workbook does not treat this as an edge case; it tells you to check before every test: "Because the basal and ceiling criteria are not the same for each test and subtest, review the criteria before testing."


And the sibling battery adds a third count


The Tests of Oral Language workbook has the same structure and a different set of numbers. Six and six for Picture Vocabulary, Oral Comprehension, Sound Blending and Sound Awareness. Five and five for Segmentation. Four and four for Sentence Repetition. Continuation instructions for Understanding Directions. Start-at-item-one-and-stop-at-the-time-limit for Rapid Picture Naming and Retrieval Fluency.


So within one publisher's own co-normed family, the criterion for "the student has reliably stopped being able to do this" is four consecutive items on one test, five on another and six on another. Nothing in either workbook explains why Sentence Repetition needs four and Segmentation needs five.


Three further WJ IV rules worth knowing, because they are not intuitive


There can be more than one basal, and the rule is to take the highest. The workbook's own practice exercises make the point: "When there appears to be more than one basal, use the highest-numbered one."


If the criterion is never met, the end of the test becomes the criterion. If no basal is established after testing backward, "Item 1 serves as the basal." If no ceiling is reached, "the examiner needs to continue testing until the last item is administered. The last item serves as the ceiling."


The page, not the item, can determine where you stop. Best practice is to test by complete pages when there is stimulus material on the examinee's side of the book, so "if a ceiling is reached in the middle of a page, testing should continue to complete the page."


And if a new basal appears while you are finishing that page, testing continues until a new ceiling is reached and the page is completed. A ceiling is therefore partly a property of the book's layout.


For a test like Oral Reading, which has no basal at all, the continuation instruction is a simple futility rule: testing began with items 1 to 5, continued through items 6 to 10, and "may be discontinued if the examinee scores 5 or fewer points on Items 6 through 10."

Pearson solved the same problem with a different object


Pearson's achievement batteries do not run on basals. They run on grade-based start points, reverse rules, discontinue rules and item sets — and Pearson's own published documentation explains the substitution directly.


The clearest public statement is in Pearson's WIAT-III Frequently Asked Questions, which contains a question a practitioner would actually ask: "Why were item sets used instead of basal and discontinue rules on the Reading Comprehension subtest?" The answer:


"The Reading Comprehension subtest was created using item sets that were designed to assess an adequate sample of the student's reading achievement, while attempting to minimize the number of reading passages required. The improved scoring procedures for the Reading Comprehension subtest takes into account the difficulty level of the item sets, so it is not necessary to administer additional items to get a more accurate score."

Set that beside Riverside's rationale and the difference is not cosmetic. Riverside estimates the whole-test score by crediting the easy items it did not give. Pearson declines to estimate a whole-test score at all, and instead scores a fixed set against its own difficulty. Two publishers, two theories of what a raw score is, both defensible, neither adjudicated by anyone.


The same document specifies the rest of the architecture, and three details matter for a caseload.


A reversal is capped. "No, you can only reverse three times. Do not go back more than three start points from the grade-appropriate start point." And if the student still cannot do the reversed set, "The student's standard score will be based on his or her performance on the last reversed item set."


A discontinue rule is a separate object from a reversal, and the FAQ gives the cleanest formulation of the pair anywhere in this comparison: "it's best to think of the reversal as finding the start point, and then discontinue as finding the stop point." For subtests with a discontinue rule of four consecutive wrong answers, four consecutive wrong answers ends the subtest.


Choosing to start low costs you a reversal. Where you have data supporting a student several levels below grade placement, the FAQ permits starting one level below the grade-appropriate start point — "HOWEVER," it continues, you may then apply the reverse rule "a maximum of two more times, as opposed to three more times if they had begun at the grade 5 start point."


That last one is the detail worth carrying away, because it is the one that catches people. A kindly decision at the start of the subtest — begin a grade lower, protect the rapport — silently removes one of the three floors beneath the student.

Note also what the architecture does to a figure practitioners rely on. PRO-ED publishes a testing time for the GORT-5: NU of 20 to 30 minutes. Pearson's WIAT-4 product page and its Q-interactive flyer both give completion time as "Varies by grade level and number of subtests administered."


When the stopping rule decides how many items the student sees, the publisher cannot put a number on the clock. It is a small thing, and it is the architecture showing through the catalogue copy.

The test that changed its own rule twice, and said why


If the preceding sections make the stopping rule sound like a technicality, this one settles it. The Gray Oral Reading Tests has published three different basal-and-ceiling regimes in roughly a decade, and the first change happened because the rule in the manual was producing the wrong number.


On 21 June 2013, PRO-ED's Director of Research and Development, Elizabeth A. Allen, sent GORT-5 users a letter with replacement pages for pages 8 to 10 of the manual. It is reproduced, on PRO-ED letterhead, as a distributor's sample of the replacement pages. It opens:


"Recently, several GORT-5 users have reported that the application of the basal rule for the Comprehension subtest printed in the manual inflates the scores of some students. After extensive review of the data and scoring procedures, we have verified that this is, in fact, the case.

During the process of reviewing the scoring procedures, we discovered that the instructions provided in the manual differed from those that were used to prepare the normative tables and establish the test's reliability and validity."

Read that second sentence twice. The rule printed in the manual was not the rule used to build the norms. A student's score was being compared against a distribution generated by a different stopping procedure from the one the examiner had just applied.


PRO-ED then did the thing that makes this citable rather than merely alarming: it rescored the standardization sample both ways and published the comparison.


"We then rescored the standardization data using the instructions in the manual and compared those scores with those actually used during the development of the GORT-5. Surprisingly, we found that in 94% of the sample the difference between the scorings was equal to or less than 1 standard error of measurement (i.e., 1 scaled score point). For the remaining 6%, however, the Comprehension score was indeed inflated."

So the publisher's own answer to "how much does the stopping rule move the score" is: for 94 of every 100 students, within one standard error of measurement; for the other six, enough to inflate the reported comprehension score.


That is the only quantification of this effect located anywhere for this article, it comes from the publisher, and it is expressed in the one unit that makes it interpretable — the standard error of measurement, which is where this library has already set out why a score is a band rather than a point.


The replacement rule was a fourth architecture


The corrected pages do something none of the other instruments here does: they make one subtest's basal govern a different subtest's scoring.


"Although the GORT-5 yields scores for Rate, Accuracy, Fluency, and Comprehension, basals and ceilings are computed only for Fluency. The Fluency basal is established when the student scores 9 or 10 points on two consecutive stories. The ceiling occurs when the student's Fluency score is 2 or less on two consecutive stories."

Comprehension gets no stopping rule of its own. "No basal or ceiling rules are applied to Comprehension. Instead, every story that is administered in the process of obtaining a Fluency basal and ceiling is also scored for Comprehension" — and in calculating the Comprehension raw score you "add 5 points for each unadministered story below the Fluency entry point or basal."


The warning attached to it is the sternest sentence any publisher writes on this subject: "Never administer stories below the Fluency basal or above the Fluency ceiling for any reason. Doing so will result in invalid scores."


Two further provisions are worth knowing because they contradict a widespread assumption. If no Fluency basal is obtained on the first two stories, testing continues until a ceiling is reached or Story 16 is administered, and only then does the examiner return to the entry point and work backwards.


And: "Some students will not obtain a basal for Fluency even when Story 1 has been administered. Other students will not obtain a ceiling for Fluency even when Story 16 has been administered. These situations do not invalidate the test."


And then it changed back


The current edition is the GORT-5: NU, the normative update. Among the features PRO-ED lists as new on its own product page, alongside additional data collected to update the original 2008–2010 normative sample:


"Reinstated separate basal and ceiling rules for Fluency and Comprehension"

So the sequence, entirely from the publisher's own documents, is: separate rules for the two subtests; a single Fluency-derived rule after the 2013 correction; separate rules again in the normative update. Three regimes, one construct, one publisher, about a decade. Nothing about oral reading changed in that time. What changed was the rule for deciding when to stop, and each time it changed, the raw score it produces changed with it.


One small thing on that same product page is worth noticing on its own account. The new-features list also mentions "extensive studies of the test's floors, ceilings, and item gradients" — where "ceilings" means the psychometric property, the point at which a test stops discriminating among able students, and not the administration rule discussed two lines earlier.


The word does double duty on a single page, which is a fair warning about reading it quickly anywhere else. The one use of "ceiling" in the Standards is this second sense too.

The curriculum-based measures threw the whole vocabulary out


For contrast, take the measures a practitioner is most likely to be using weekly. The DIBELS 8th Edition Administration and Scoring Guide runs to roughly 41,000 words and is the most procedurally specific document in this comparison. Searched on 16 November 2026:


  • "basal" — zero. "ceiling" — zero. "start point" — zero. "reverse" — zero. "item set" — zero.

  • "discontinue" — 49. "discontinue rule" — 20. "gating" — 22. "raw score" — 35.


A timed one-minute measure has no use for a basal, because the time limit is the stopping rule. What DIBELS 8 has instead is a second kind of stopping rule that none of the norm-referenced batteries has at all, and its consequences are larger.


"Each DIBELS subtest has a specific discontinue rule. An assessment should only be discontinued if the specified conditions have been met, or if the administration is irrevocably interrupted (e.g., a fire drill occurs)... For some subtests at certain time points, not only is that subtest discontinued, but benchmark assessment is also discontinued altogether."

That is the gating rule, and the guide states its purpose plainly: the rules "prevent unnecessary and excessive testing for the most vulnerable learners by giving educators the option to stop the administration of benchmark assessments based on student performance."


The justification is empirical and the guide gives it: the rules "were derived from a national field trial that indicated students who scored 0 for the indicated assessments in the periods specified above were extremely unlikely to get any items correct on the remaining subtests."


The scoring consequence is in a table, and it is the sharpest contrast in this article. In fall of kindergarten, if Phoneme Segmentation Fluency is discontinued, the examiner does not administer Nonsense Word Fluency or Word Reading Fluency — and the instruction is: "Enter 0 for PSF. Do not enter scores for the remaining subtests." In winter of kindergarten, if Nonsense Word Fluency is discontinued, Word Reading Fluency is not administered.


Discretion is preserved and the guide says so: "examiners have the option of administering the remaining subtests based on professional judgment."


The same measure, a different stopping rule, at the other publisher


The comparison worth making is with Acadience, because these two publishers measure the same constructs under nearly the same subtest names and have already been shown to score the same student response differently.


They stop differently too. Acadience's published Nonsense Word Fluency page gives the rule as a single line: "Discontinue Rule: No correct letter sounds in the first row."


That is a third species of stopping rule and it is not a count of anything. It is not a number of consecutive errors, and it is not a time limit.


It is a position on the page — the student who produces nothing in the first row has produced enough information to stop, and the boundary is drawn by the layout of the probe rather than by a tally. The same page specifies a three-second wait before supplying a letter sound or a word.

The question underneath all of it: what is an unadministered item worth?


Every stopping rule creates items the student never saw, and every publisher has to decide what those items are worth. Laid out, the answers are irreconcilable.


  • On the WJ IV, an unadministered item below the basal is worth a point. You add one point for every item below the basal on most tests. The student is credited with items they were never shown, on the publisher's stated theory that this estimates the whole-test score.

  • On the GORT-5 after the 2013 correction, an unadministered story below the basal is worth full credit on Rate, Accuracy and Fluency, and five points on Comprehension. Above the ceiling, the same story is worth zero.

  • On DIBELS 8, a subtest stopped by a discontinue rule is entered as zero, and the subtests that follow it are not entered at all. They do not score zero in the composite; they are absent from it.

  • On the Pearson item-set subtests, the question does not arise, because there is no unadministered-but-credited block. The set is the measurement, and its difficulty is in the scoring.


Credited as correct, credited as full marks, entered as zero, absent from the composite, or structurally impossible. That is five answers to one question, and the question is "what happened in the part of the test you did not give."


This is the pattern this library keeps running into, and it is worth naming rather than rediscovering each time. When several organisations share a vocabulary, check whether they share a definition.

The letters in a qualification level turn out to be a purchasing rule for one storefront rather than an industry scale.


A full bubble on the NCII tools charts means a different coefficient in every column.


A mastery criterion is specified by every program and derived by almost none, and a reading level is four incompatible scales with no conversion anybody endorses.


"Basal" is the same finding one level further down, in the machinery that produces the number all of those are built on.

What this does and does not license you to say


Three boundaries, because this is territory where it is easy to overstate.


It does not mean the scores are wrong. A basal-and-ceiling administration produces a raw score the publisher's own norms were built on, provided the rule you applied is the rule the norms were built with.


The GORT-5 episode is notable precisely because it is the documented case where those two came apart, and the publisher found and corrected it. Every other instrument discussed here has published nothing of the kind, and "nothing published" is not evidence of a problem.


It does not license comparing raw scores or item counts across instruments. A WJ IV raw score includes credited items; a Pearson item-set score does not; a DIBELS score is a one-minute count. The fact that two numbers are both called raw scores does not make them the same kind of object, which is the same caution norm-referenced and criterion-referenced scores require.


And it does not make a misapplied rule into a diagnostic finding. If you suspect a ceiling was applied early on a report that crossed your desk, what you have is a question about administration, not a conclusion about the student. What you may do with somebody else's protocol is set out in Psychoeducational Evaluation for Educational Therapists, and the short version is that you may ask.

What the profession that credentials you says about any of this


Nothing, which by now is the expected answer.


The AET Code of Ethics, Revision #9 of 2025, does not contain the words "basal" or "ceiling", and it contains no clause about test administration or scoring procedure at all.


What it does contain is the pair of clauses that put this inside the role anyway. II.I.B requires members to "select and use appropriate assessment instruments, recognizing their limitations with respect to reliability, validity, and bias." II.I.C requires them to "use only those assessment instruments for which they have been adequately trained."


Those two clauses are the whole professional obligation here, and they are more demanding than they look. A stopping rule you have not read is a limitation you cannot recognise, and a battery whose four stopping regimes you have not checked is not one you have been adequately trained on.


This is the tenth consecutive post in this library to find the Code silent on the method question and specific about the competence question, and the pattern is now predictable enough to state as a rule: AET regulates conduct, not procedure, and procedure is where the number comes from.


One limitation, stated because it bears on the claim. aetonline.org now sits behind a bot-protection interstitial that refused three separate direct downloads of the Code this run, so the document was read through a fetching service rather than word-counted locally as it has been on previous runs.


The absence of "basal" and "ceiling" is reported on that basis. If a count of any other term in the Code is load-bearing for you, open the PDF yourself.

A sequence, not a decision


1. Before you administer anything, read the stopping rule for the specific subtest. Not the battery — the subtest. The WJ IV is the proof that a single test book can carry four regimes, and its own publisher's training material tells you to check before every test.


2. Record the start point and the stopping rule you applied, not just the raw score. This is the cheapest insurance in the whole process. "Word Attack, started item 15, basal at 15–20, ceiling 31–36, raw 24" is reconstructable by anyone. "Word Attack, raw 24" is not. Whatever your session notes format, the protocol is the place this belongs.


3. Treat a reversal as a budget. On a Pearson item-set subtest you have three, and starting a grade low spends one before you begin. Decide deliberately whether to spend it on rapport.


4. When a report crosses your desk, check whether the basal and ceiling are recorded on it. Many protocols show them and many summary reports do not.


If the summary gives you a standard score and no administration detail, you are being asked to rely on a rule you cannot see — which matters most for exactly the students at the edges of the distribution, where grade equivalents and extrapolated scores are least trustworthy anyway.


5. Check the edition, and check whether the rule changed with it. The GORT-5 to GORT-5: NU transition changed the stopping rule, which means two protocols from the same test, a few years apart, were not scored on the same procedure. Edition drift is already a reason to re-check test norms; add the stopping rule to the list of things an edition can change.


6. If you are re-testing, remember that the stopping rule and the form interact. A student who reached a different ceiling on a second administration has not necessarily changed; the start point may have differed. What publishers do and do not say about re-testing intervals and alternate forms is in Re-test Intervals and Practice Effects.


7. Do not tell a family that a stopping rule invalidated their child's score. You cannot know that from a report, almost certainly it did not, and the one publisher-verified instance of a stopping rule inflating a score affected six per cent of a standardization sample by about one scaled score point for most of them. State what you can see and ask about what you cannot.

What is out of scope here


Nothing in this article addresses whether a particular protocol was correctly administered, whether a score is valid for an eligibility decision, or what any statute, regulation, district policy or insurer requires of an evaluation. Those are questions for the evaluating professional, your state education agency and, where entitlement is in question, an attorney licensed in your state. Nothing here is legal advice.


It also does not address what you may change about an administration — that is an accommodation or a deviation, and it has its own publisher-by-publisher answer. It does not address what you may do with the test materials themselves, which is test security.


And it does not address which instrument to choose for a presenting problem; the comparisons for spelling, pseudoword decoding, phonological awareness and reading comprehension each give a decision sequence rather than a decision.

Where the published record runs out


As of 16 November 2026, the following could not be established, and saying so is more useful than filling it in.


  • No study comparing stopping rules was located. Nothing establishes that six consecutive items is better than five, or that a basal-credited raw score estimates a whole-test score more accurately than an item-set score. The search was targeted rather than exhaustive, and absence from a search is not a verdict — but it is consistent with the Standards requiring no rationale for a hand-applied rule.

  • No publisher locatable for this article explains its particular number. Riverside states the purpose of a basal and never says why Sentence Repetition needs four consecutive items and Segmentation needs five.

  • Whether the WIAT-4's digital platform applies the start-point and reverse rules itself, or merely records the examiner applying them, is not published on the material checked. It matters more than it sounds: Standard 4.3's documentation requirement is triggered by a test "delivered using computer algorithms to select items", so a platform that selects items may fall inside a requirement the paper edition of the same test falls outside. That is an open question here, not a finding, and the place to ask it is the publisher.

  • The GORT-5: NU's reinstated rules were not read. PRO-ED's product page states that separate basal and ceiling rules for Fluency and Comprehension were reinstated; the rules themselves are in the examiner's manual, which is not public. What they now say, and whether they match the pre-2013 wording or a third formulation, could not be determined.

  • The 2013 GORT-5 errata was read from a distributor's reproduction. It is on PRO-ED letterhead, signed by a named PRO-ED officer, with a PRO-ED contact address, and its content is corroborated by the current product page's reference to reinstating what it replaced. It could not be found on PRO-ED's own current website, which is unsurprising for a thirteen-year-old manual correction but is stated here for completeness.

  • The Pearson architecture was verified from a WIAT-III document. Pearson's clearest public explanation of item sets, reverse rules and discontinue rules is a WIAT-III FAQ. The WIAT-4's own equivalent documentation is in its manuals rather than on a public page, so whether every rule carried over unchanged to the fourth edition was not established.

  • The AET Code could not be word-counted locally this run, for the reason given above.

  • Whether any state, district or funder requires administration detail on a submitted report is not addressed. That is policy and contract, and it varies.

Frequently asked questions


What is the difference between a basal and a start point?


They answer different questions. A start point is where you begin, and on most batteries it is suggested by the student's age or grade. A basal is a level of demonstrated success — a specified number of consecutive correct items — that licenses you to credit everything below it without administering it.


You can have a start point without a basal, which is what a timed curriculum-based measure has, and you can fail to establish a basal from your start point, in which case you test backwards. On Pearson's item-set subtests the FAQ's own formulation is the useful one: the reversal finds the start point and the discontinue rule finds the stop point.


Is a ceiling the same thing as a ceiling effect?


No, and the two senses sit on the same pages. An administration ceiling is a stopping rule — a specified number of consecutive incorrect responses that ends the subtest. A ceiling effect is a psychometric property of a test whose hardest items are too easy to discriminate among able test takers.


The Standards for Educational and Psychological Testing uses the word exactly once, in the second sense. PRO-ED's GORT-5: NU product page uses both senses within a few lines of each other. When you meet the word in a technical manual, check which one is meant.


Do unadministered items below the basal count as correct?


On the Woodcock-Johnson, yes — you add a point for each item below the basal on most tests, and the publisher's stated rationale is that this estimates the score the student would have obtained on the full test. On the GORT-5 after its 2013 correction, stories below the Fluency basal receive full credit for Rate, Accuracy and Fluency and five points each for Comprehension.


On DIBELS 8 the question does not arise in that form: a subtest stopped by a discontinue rule is entered as zero and subsequent subtests are left out of the composite entirely. So the answer is publisher-specific and it is not safe to generalise from the battery you know best.


Has a basal rule ever actually changed somebody's score?


Yes, and the publisher documented it. In June 2013 PRO-ED wrote to GORT-5 users to say that the Comprehension basal rule printed in the manual inflated some students' scores, and that the manual's instructions "differed from those that were used to prepare the normative tables".


Rescoring the standardization sample both ways, PRO-ED reported that for 94% of the sample the two scorings differed by no more than one standard error of measurement, and that for the remaining 6% the Comprehension score was inflated. Replacement manual pages were issued.


Can I start a subtest lower than the manual says to protect a student's confidence?


Pearson's FAQ permits it where you have data or case history supporting a placement several levels below grade — and tells you what it costs: you may then apply the reverse rule a maximum of two more times rather than three.


Treat it as a trade you are making deliberately rather than a kindness with no consequence, and record the start point you used. On a battery with a conventional basal rule, starting lower generally costs administration time rather than measurement, because the basal will be established and the items below it credited either way.


Does the examiner's stopping rule affect the confidence band on the report?


Not directly, and it is worth being precise. The confidence band is built from the test's reliability, and it describes the error around the score you obtained. The stopping rule determines which score you obtained.


The two are separate machinery, and the reason the GORT-5 errata is quotable is that PRO-ED expressed the size of a stopping-rule change in units of standard error of measurement, which is the only published bridge between them located for this article. What the band itself is and is not is set out in Standard Error of Measurement.

If this is the kind of detail your practice runs on, the Illuminate educator list gets these comparisons as they publish — what the manuals specify, what the clearinghouses rate, and where the published record stops. Join the educator list.

Recent Posts

See All
Standard Error of Measurement: What That Band Means

Every score report prints a confidence interval and almost no practitioner acts on it. What the standard error of measurement is, why the band's width is a software setting rather than a property of t

 
 
 

Comments


bottom of page