Norm-Referenced vs Criterion-Referenced: Not Two Tests
The short version
If you have ever tried to settle whether a tool is a norm-referenced or a criterion-referenced assessment, you were asking a question the measurement standards do not answer, because they do not classify tests that way at all. They classify interpretations. The same raw performance, from one sitting, can be read both ways, and on several of the instruments already in your cupboard it is read both ways on the same printout.
That is not a technicality for a methods seminar. It decides which sentences you are entitled to write in a progress summary, and it explains why two numbers from the same administration can seem to disagree.
Here is what the record actually says, instrument by instrument.
The standards define two interpretations, not two kinds of test
The 2014 Standards for Educational and Psychological Testing — released open access in 2021 by AERA, APA and NCME, and catalogued on AERA's own standards page, so you can read the whole thing free — puts both terms in its glossary as kinds of score interpretation.
A criterion-referenced score interpretation is defined there as the meaning of a score "indicating the individual's or group's level of performance in relationship to some defined criterion domain," with cut scores and expectancy tables given as examples. The NCME glossary carries the same two entries, worded the same way, and cross-references each to the other.
Notice what is missing from both definitions. Neither one is a statement about the instrument. Each is a statement about what a reader does with a number.
The chapter on scores and scales then says so directly, on page 96:
"Some scale scores are developed primarily to support norm-referenced interpretations; others support criterion-referenced interpretations. In practice, however, there is not always a sharp distinction. Both criterion-referenced and norm-referenced scales may be developed and used with the same test scores if appropriate methods are used to validate each type of interpretation."
The same page adds that interpretations built on cut scores "may likewise be either criterion referenced or norm referenced," and works through a case where ranking candidates to fill a fixed number of places makes a pass-fail cut score norm referenced despite looking absolute.
The one obligation the standards do attach
Standard 5.5 is short and worth knowing by heart, because it is the thing you can hold a publisher to:
"When raw scores or scale scores are designed for criterion-referenced interpretation, including the classification of examinees into separate categories, the rationale for recommended score interpretations should be explained clearly."
So when a report hands you a proficiency band, a mastery flag or a risk category, the rationale for that classification is something the test's documentation owes you. If you cannot find it, that absence is itself a finding, and it is a fair thing to note in your own write-up.
A test publisher's own glossary says a test can be both
This is not only an academic position. Riverside Insights publishes a glossary of testing, measurement and statistical terms for its customers, and its entry for norm-referenced test ends with two sentences that settle the matter from the vendor side:
"Tests can be considered both norm-referenced and criterion-referenced. It depends only on the use and interpretation of the scores."
Its criterion-referenced interpretation entry makes the same point from the other direction, noting that "at times, criterion-referenced interpretations are also made about tests that provide norm-referenced information."
The glossary is also careful about a word that gets used loosely in staff meetings. Norms, it says, "represent typical performance, not necessarily a desired level of performance," and it contrasts norms with standards explicitly. Typical is not the same as adequate, and a score at the median is not thereby a score that is good enough for anything in particular.
The WJ V prints both, from one administration
The clearest working example sits in a battery many practitioners already own. Riverside's assessment service bulletin on the W score and the Relative Proficiency Index sets out four levels of interpretive information, and two of them are the two frames in this article.
Level 3 is proficiency. The bulletin says it "provides criterion-referenced scores," and lists the Relative Proficiency Index, the instructional and developmental zones derived from it, and the cognitive-academic language proficiency level.
Level 4 is relative standing in a group. Those, it says, "are norm-referenced scores, such as standard scores (SS) and percentile ranks (PR), that describe an examinee's relative standing, or rank order, compared to their age or grade peers."
One sitting. One raw score. Two frames, printed side by side, and the bulletin is explicit that Level 4 scores "indicate ordinal position in a group" rather than the quality of the performance.
Riverside's score report interpretation guide confirms that this is not a theoretical pairing but what actually prints: a generated report "can include W Difference Scores, relative proficiency indexes (RPIs), age and grade equivalents" and instructional zones, "along with indications for above- and below-average normative performance." Both frames, one page.
What the RPI actually is
The RPI is defined in the same bulletin as "a criterion-referenced score that describes the probability of a person's success on a task similar to the tasks used in the assessment at the level of difficulty that 90% of average grade or age peers can manage."
Read that definition slowly and you will see the hinge of this whole article. The RPI is a criterion-referenced score whose criterion is calibrated against what average peers can do. The bulletin says as much, and treats it as a selling point rather than an embarrassment:
"The advantage of the RPI over other types of criterion-referenced scores is that the criterion items and the person's proficiency level are derived from norms—real scores from real people."
If you have been teaching yourself that criterion-referenced means "no comparison with other children," the publisher of one of the most widely used achievement batteries disagrees with you in print. The two frames are about what the comparison is for, not about whether a comparison happened.
The standard error band printed beside those standard scores belongs to the Level 4 frame and not to the Level 3 one, which is one reason the two can look like they are telling different stories.
The KTEA-3's error analysis looks criterion-referenced and is not
Here is the correction most likely to be useful, because the practice it describes is extremely common and the mental model behind it is wrong.
Pearson's page on KTEA-3 error analysis describes a procedure that reports a student's performance on specific skill categories — literal against inferential comprehension, item and section types in phonological processing, error categories in computation — and labels them as a weakness, average, or a strength. It reads exactly like a criterion-referenced skills profile, and it is the part of the output practitioners most often lift straight into a plan.
Pearson says the method "resembles other criterion-referenced assessment procedures" and then explains how it differs. A criterion-referenced procedure would set a mastery cut-off. The KTEA-3 deliberately does not, and instead compares the number of errors a student made against the average number of errors made by the norm sample.
So the categories are norm-based. "Weakness" there means an unusually high error count relative to the standardisation sample, not a judgement that a skill is unmastered against any defined standard. The two statements license different sentences, and only one of them is supported.
This matters when the profile is being used to pick a starting point, because a skill can be entirely unmastered and still produce an average error count if most children that age have not mastered it either.
The same inventory, published twice, in two frames
The cleanest demonstration that the frame is a choice rather than a property is an instrument that Curriculum Associates publishes in two parallel editions.
The BRIGANCE Inventory of Early Development criterion-referenced edition describes criterion-referenced assessments as measuring "an individual's performance on a specified set of skills over time," and its overview says the inventory "can be administered to students of any chronological age."
The IED III Standardized edition of the same inventory "provides standardized scores across key developmental domains," contains fifty-five norm-referenced assessments comparing a child with same-age peers, and produces age equivalents, standard scores and percentile ranks for children from birth to age seven.
The frame is prominent enough to sit in the product name. Curriculum Associates' BRIGANCE product line page announces the "BRIGANCE Criterion-Referenced Inventory of Early Development 4," so the criterion-referenced line has moved to a fourth edition while the standardized edition remains at its third.
The item content overlaps heavily. What changes with the frame is the population the instrument can speak about at all. The criterion-referenced edition travels up the age range with a student whose skills sit well below their chronological age; the standardized edition stops at seven, because beyond that there is no reference group for the comparison it exists to make.
That is the practical lesson hiding in a catalogue listing. A norm-referenced interpretation is only available where norms exist for the person in front of you, which is also the reason an out-of-date normative sample damages one frame and leaves the other alone.
The criterion in a criterion-referenced cut score is very often a percentile
The screening tools are where this nesting becomes impossible to ignore, and both of the big reading screeners document it themselves.
Acadience Learning's benchmark goals handout calls its benchmarks "empirically derived, criterion-referenced target scores that represent adequate reading skill for a particular grade and time of year." Four lines later it says what the target is calibrated to:
"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 development section names the external criterion as the Group Reading and Diagnostic Evaluation, and states that the fortieth percentile on that measure "was used as an indicator that the students had adequate early literacy and/or reading skills for their grade."
The study ran in the 2009–2010 school year across thirteen schools in five states, with 3,816 participants in grades K–6 — though only a subset of 1,306 sat the GRADE, which is the comparison the criterion rests on.
The DIBELS 8th Edition benchmark goals define their colour bands the same way. The blue goal is the range where "nearly all students in this range score at or above the 40th percentile rank on criterion measure."
The green range captures about eighty per cent of students at or above that percentile, the yellow range about eighty per cent of those below it, and the red range about eighty per cent of those below the twentieth percentile. The University of Oregon's own research brief documents the receiver-operating-characteristic analyses these were derived from.
So a benchmark on either instrument is a criterion-referenced classification resting on a norm-referenced anchor. Nobody is being evasive about it; both documents state it on the page that gives you the numbers. But a practitioner who reports "at benchmark" as though it carried no peer comparison is reporting something the handout does not say.
Which is also why these numbers do not convert
The two instruments use the same percentile anchor on different criterion measures, which is part of why their cut points sit at different places and why moving a student between systems mid-year breaks the graph. The same problem turns up in compiled oral reading fluency norms built by averaging tables from instruments that disagree about how many passages make a score.
The curriculum-based end of the shelf claims something narrower
There is a third label in common use, and the publishers who could reach for "criterion-referenced" mostly do not.
Wilson Language Training describes the Wilson Assessment of Decoding and Encoding as "a curriculum-based measure which specifically assesses a student's decoding and encoding (spelling) skills," correlated to the word structures taught in the Wilson Reading System. The page talks about mastery and about guiding instruction. It never uses the phrase criterion-referenced, and it makes no claim about norms either way.
Wilson's own assessment guidance places the WADE as a post-test "administered as a posttest at the end of instruction or at the end of a school year" to evaluate "mastery of the curriculum." Its End-of-Step assessments are described as both formative and summative, with a specific mastery level required to advance.
That is a precise and modest claim, and it is the right one. A curriculum-based measure tells you whether a student has learned what this program taught. It does not purport to locate them in a population, and it cannot, because its content universe is one publisher's sequence rather than a sampled domain.
The Wilson Reading System program page is consistent about this, describing the program as based "on phonological-coding research and Orton-Gillingham principles" and specifying instruction and grouping rather than any psychometric frame.
Where Wilson does require norm-referenced testing, it is for a different purpose entirely: its Level I certification practicum requires pre- and post-testing "with two norm-referenced tests, one being the Word Identification and Spelling Test," which is a credentialing requirement rather than a claim about the WADE.
Contrast easyCBM, which sits in the same curriculum-based family and does publish norms: its 2014 norming technical report stratified a sample drawn from four regions of the country against federal school data. Same family of measure, different frame available, because somebody did the norming work.
Mastery measurement is a third paradigm, and this article is not it
The threshold question — what counts as mastered, and where the number came from — is a separate argument with its own literature, and this site has already taken it apart at length. That post makes the case that mastery measurement is a different paradigm rather than a stricter one, and works through where the familiar eighty and ninety per cent thresholds actually come from across four published programs.
Rather than restate it here, the boundary is worth naming: that article is about where a criterion comes from. This one is about what kind of statement a score licenses once you have one. If you are trying to choose a threshold, read that one.
What each frame actually licenses you to write
Strip away the vocabulary and the practical difference is a difference in the sentence you are entitled to put in a document somebody else will read.
A norm-referenced interpretation licenses a statement of position. This student's performance sits at this point in a distribution of same-age peers, with the precision the standard error allows. It does not tell you what the student can do.
A criterion-referenced interpretation licenses a statement of content or classification. This student can or cannot perform tasks of this kind, or falls in this category against a defined domain. It does not tell you whether that is unusual.
A curriculum-based measure licenses a statement about this program. This student has or has not mastered what was taught in these lessons. It travels no further than the sequence it was built on.
The errors worth avoiding all involve borrowing a licence from the wrong frame. Reporting a percentile as though it described a skill. Reporting a benchmark status as though no peer comparison were involved. Reporting a program post-test as evidence about a population. Reading an error-analysis "weakness" as a statement that a skill is unmastered.
None of the three frames supports a diagnostic statement. If the question is whether a student has a disability, the answer here is a flat no: no score interpretation of any kind identifies a condition, and the question of what you may do with a psychoeducational report is a scope question rather than a psychometric one.
The Association of Educational Therapists' Code of Ethics directs members to seek interpretation from related-field professionals, and its fact sheet sets out the scope of the role.
When several organisations share a vocabulary
The pattern in this article is the one that keeps recurring across this field, and the posts that have documented it are worth reading together.
"Aligned" is a claim nobody adjudicates, which is the finding in the piece on the IDA Knowledge and Practice Standards. "Orton-Gillingham based" is a label no organisation governs, per the post on that phrase. "The science of reading" names different things to different bodies.
A grade equivalent is printed by publishers who tell you not to use it. "Supervised hours" means six different things to six credentialing bodies. A qualification letter is one storefront's purchasing rule rather than an industry scale. "Evidence-based" means five different things to five rating systems.
Norm-referenced and criterion-referenced belong on that list, with one difference worth noting. Unlike most of the others, this pair is carefully defined by a governing document. The confusion is not that nobody defined the words. It is that the definitions are about interpretations and the field uses them as though they were about products.
A sequence, not a decision
When a score lands in front of you and you need to know what you may say about it, the order that works is this.
Identify which frame the number belongs to. A percentile, standard score or stanine is a position. A band, category, benchmark status or mastery flag is a classification. A program post-test is a curriculum statement.
For any classification, find the rationale. Standard 5.5 says the documentation should explain it clearly. Look for what the cut score was derived from and against which external measure.
Check whether the criterion is itself a percentile. On the reading screeners it usually is, and if so your classification carries a peer comparison you should not describe as absent.
Check that norms exist for this person. Age range, and how old the normative sample is. Where they do not, the norm-referenced frame is simply unavailable.
Write the sentence the frame licenses, and no more. Then attach the precision, because a position without a confidence band overstates what a single administration established.
If you are re-testing to show change, the interval and practice-effect rules in the manual govern whether a second score means anything at all.
That is a sequence for reading a number honestly. It is not a decision rule for which instrument to use, and it is deliberately not one.
What is out of scope here
This article reports what measurement standards and test publishers have published about two kinds of score interpretation. It does not tell you which instrument to administer, and it does not interpret anybody's scores.
Nothing here is legal advice. Eligibility determinations, entitlement to services and the evidentiary weight a particular score carries in a school process are not measurement questions, and they belong with the relevant agency and, where a family's rights are at stake, with an attorney.
Questions about what you are qualified to administer or interpret belong with your own credentialing body and, for the purchasing rules, with the publisher — PRO-ED's qualifications policy defines its lowest tier as "There are no special qualifications to purchase these products," which is a purchasing rule and plainly not a competence standard.
Where the published record runs out
Several things a practitioner would reasonably want to know are not in the published documents, and it is worth being plain about which.
No publisher states, for a mixed battery, which of its printed scores are intended for which frame. The WJ V bulletin is the exception that proves the rule: it maps its score types onto interpretive levels explicitly, and no equivalent mapping was found for the other achievement batteries on the shelf.
PRO-ED's current product pages would not serve their instrument descriptions to an automated request on this run, so the descriptors it applies to the CTOPP-2, TOWRE-2 and GORT-5 could not be re-verified here. If a specific descriptor is load-bearing for a report you are writing, load the product page yourself or ask the publisher.
The rationale Standard 5.5 calls for is frequently not findable without the technical manual, and the technical manual is frequently not the volume that ships in the kit.
The National Center on Intensive Intervention's tools charts rate instruments on technical properties without classifying them by interpretive frame at all, so the chart cannot answer this question for you.
Nobody has published a crosswalk showing, for the instruments an independent practitioner is most likely to own, which frames each supports. That is the document this article is a first approximation of, and it is assembled from publishers' own pages rather than from any single authority.
Frequently asked questions
Is the WIAT-4 norm-referenced or criterion-referenced?
That question has no answer as posed, because the 2014 Standards classify interpretations rather than tests, and state that both kinds of scale "may be developed and used with the same test scores." The useful question is which frame a particular printed score belongs to. A standard score or percentile is a norm-referenced interpretation; a band, category or mastery classification is a criterion-referenced one.
Does criterion-referenced mean the score is not compared with other children?
No, and this is the most common misconception in the area. Riverside's own glossary says tests "can be considered both norm-referenced and criterion-referenced," and the Woodcock-Johnson Relative Proficiency Index is defined as a criterion-referenced score calibrated to the difficulty level "that 90% of average grade or age peers can manage." Many criterion-referenced cut scores are anchored to a percentile on an external measure.
What is the criterion in a reading screener's benchmark?
On both major screeners it is a percentile rank on an external measure. Acadience Reading's benchmarks indicate a level at which students are "likely to score above the 40th percentile on any high-quality reading assessment," with the Group Reading and Diagnostic Evaluation as the external criterion in the development study. DIBELS 8th Edition defines its bands against the fortieth and twentieth percentile ranks on a criterion measure.
Is the KTEA-3 error analysis a criterion-referenced procedure?
Not as Pearson describes it. Pearson says the method "resembles other criterion-referenced assessment procedures" but differs from them: rather than setting a mastery cut-off, it compares a student's error count against the average error count of the norm sample. The resulting weakness, average and strength labels are therefore norm-based, and should not be reported as statements that a skill is unmastered.
Can one instrument be published in both frames?
Yes, and the BRIGANCE Inventory of Early Development is published that way. The criterion-referenced edition measures "an individual's performance on a specified set of skills over time" and can be administered at any chronological age; the standardized edition contains fifty-five norm-referenced assessments and produces standard scores and percentile ranks for children from birth to seven.
Where can I read the measurement standards themselves?
The 2014 Standards for Educational and Psychological Testing were released open access by AERA, APA and NCME in 2021, and the full text is downloadable at no cost. The treatment of the two frames is in the chapter on scores, scales, norms, score linking and cut scores, with the key passage on page 96 and the governing requirement at Standard 5.5.
Does a curriculum-based measure give me a criterion-referenced score?
Not necessarily, and the publishers are careful here. Wilson describes the WADE as "a curriculum-based measure" assessing skills "correlated to the word structures taught in Wilson Reading System," without claiming either frame. A curriculum-based measure supports a statement about what this program taught, which travels no further than that program's sequence unless the publisher has also done norming work, as easyCBM has.
Related reading
If this piece was useful, the companion posts on the standard error band, grade equivalent scores and outdated test norms take apart three more numbers that appear on reports without explanation. For the wider picture of how the roles differ, start with educational therapist, learning specialist and executive function coach compared and the practitioner scope map.
On instrument selection by construct, see the posts on dyslexia screening tools, phonological awareness assessment, pseudoword decoding measures and vocabulary assessment instruments.
Join the Illuminate educator network
Illuminate Educational Therapy Group writes for the practitioners doing this work — educational therapists, learning specialists and executive function coaches. If you would like these posts as they publish, along with the occasional note on what the published record does and does not support, join the educator list.



Comments