Pseudoword Decoding Tests: Four Units, No Conversion
Updated: 22 hours ago
The short version
A pseudoword decoding test asks a student to read something that has no correct pronunciation to look up. That is the whole point of the task, and it is also the whole problem with it: somebody has to decide what counts as right, and the people who publish these measures have decided differently.
Five instruments in common use on a private caseload report nonword reading in four incompatible units. One reports items correct with no clock. One reports items correct in forty-five seconds. Two report correct letter sounds per minute, which is not a word at all.
And the two that look most alike — the two one-minute nonsense word fluency measures sitting on the federal tools charts — score the identical student response differently. A child who says the sounds and then blends them earns a whole-word point from one publisher and nothing from the other.
None of that is a defect in any of the instruments. Each rule follows from what its publisher built the measure to do. But a score is only interpretable against the rule that produced it, and nothing on a score report tells you which rule that was.
Here is what each measure actually counts, where the published record is explicit, and where it runs out.
What makes a task a pseudoword task, and what disqualifies one
The family has one defining feature. The student is shown a printed letter string that is pronounceable under English spelling conventions and is not a word in English, and is asked to read it aloud. Riverside's own description of Word Attack in the Woodcock-Johnson V puts it plainly: on later items "the examinee reads printed, phonically regular nonsense words (i.e., pseudowords) aloud."
Two things disqualify a task that otherwise looks similar, and both of them get miscounted in practice.
The first is print. A nonword repetition task is oral. The examiner says a nonsense syllable sequence and the student repeats it. No letters are involved, so nothing about decoding is being measured.
This is not a hypothetical confusion: the Woodcock-Johnson V contains a test called Nonsense Word Repetition, and Riverside's technical documentation records that it was previously called Nonword Repetition and "was renamed Nonsense Word Repetition to more accurately describe the measured construct."
So one battery now carries two tests with the phrase "nonsense word" in the name, one of which involves no print whatsoever. If you are reading a report and see a low "nonsense word" score, check which one it was.
The second is the real-word boundary. FAST's earlyReading Nonsense Words subtest states its own criterion on the National Center on Intensive Intervention's progress monitoring chart: none of its pseudowords "were included in a Standard English dictionary." That is a curation rule, and it is not the only one — more on that below.
Four units, and none of them convert
This is the finding that matters most for a caseload, because it is the one that quietly breaks a comparison.
The achievement batteries report an untimed count of items read correctly, converted to a standard score. WIAT-4 Pseudoword Decoding, KTEA-3 Nonsense Word Decoding, Woodcock-Johnson V Word Attack and WRMT-III Word Attack all work this way. The number behind the standard score is a number of items.
The TOWRE-2 Phonemic Decoding Efficiency subtest reports items correct under a clock. PRO-ED's own description is specific: the subtest "measures the number of pronounceable nonwords presented in vertical lists that an individual can accurately decode within 45 seconds." Forty-five seconds, vertical lists, four alternate forms lettered A through D.
The curriculum-based measures report correct letter sounds per minute. DIBELS 8th Edition Nonsense Word Fluency and Acadience Reading Nonsense Word Fluency both do this, and both also report a second whole-word score alongside it.
So the unit is items, or items per forty-five seconds, or letter sounds per minute, or whole words per minute. A standard score of 85 on one and a benchmark colour band on another are not two readings of the same quantity. There is no published conversion between them and no publisher offers one.
A student can be below benchmark on correct letter sounds and within the average range on a standard score from an achievement battery, with no error anywhere. The two measures are counting different things.
If you have read what we published on why two scores disagree on reading comprehension, this is the same structure one level down in the system.
The same response, scored two different ways
Here is the sharpest case in the whole area, and it is in print at both publishers.
The University of Oregon's DIBELS 8th Edition Administration and Scoring Guide states the rule for the whole-word score directly. Nonsense Word Fluency "provides two scores: the sum of correct letter sounds (CLS) and the sum of words read or recoded correctly (WRC)."
And then, without qualification: "Words read correctly, whether sounded out initially or not, receive 1 point each for WRC." The guide's own worked example makes it concrete. A student shown "rab" who says "/r/ … /a/ … rab" scores three for correct letter sounds and one for words read correctly.
Acadience Learning's published description of its Nonsense Word Fluency measure defines the equivalent score the other way. Whole Words Read is "the number of make-believe words read correctly as a whole word, one time and only one time, without first being sounded out."
Read those two sentences next to each other. Sound out and then blend, and DIBELS 8 gives you a whole-word point. Do exactly the same thing on Acadience, and you do not get one.
Two one-minute measures of the same construct, with nearly the same name, score the identical student response one and zero.
NCII's own chart entry for DIBELS 8 confirms the rule rather than leaving it to the manual: on that page, "students receive credit for blending nonsense words whether they sound them out first or not."
Neither publisher is wrong. DIBELS is measuring whether the string can be produced as a word at all; Acadience is measuring whether it can be produced as a word without the intermediate step, which is a claim about automaticity rather than accuracy. That is a real distinction and worth having.
The problem is purely practical: a practitioner who inherits a Whole Words Read figure from a school, or who switches between the two systems mid-caseload, is comparing numbers that were produced under different rules. The same hazard we described for progress monitoring systems applies here with a sharper edge, because the scoring difference is invisible in the number.
Word Attack is not only a nonword test
Riverside's description of the Woodcock-Johnson V Word Attack test has two clauses, and practitioners usually quote the second one only.
"On the early items, the examinee must produce the sounds for single letters and letter combinations. On later items, the examinee reads printed, phonically regular nonsense words (i.e., pseudowords) aloud."
For a young student who does not reach the later items, a Word Attack score may be built substantially or entirely out of isolated letter and letter-combination sounds.
The task name has not changed and the standard score looks like every other standard score, but the behaviour sampled is different. If you are writing up a kindergarten or early first grade evaluation, that is worth knowing before you describe the score as a measure of nonword reading.
What the publisher's own technical document says about the construct
The Woodcock-Johnson V Technical Abstract classifies Word Attack under two broad CHC abilities, Grw (reading and writing) and Ga (auditory processing). Its own factor analyses do not support the second one strongly.
In the Horn no-g broad CHC model, Word Attack's loadings on Ga across the six age groups are 0.14, 0.26, 0.23, 0.29, 0.13 and 0.18, for a median of 0.21.
In the Carroll hierarchical model reported in the next table, the same loadings are 0.22, 0.22, 0.24, 0.30, 0.17 and 0.17, median 0.22. Against that, Word Attack's loadings on Grw run from 0.60 to 0.75 in the first model and 0.60 to 0.71 in the second.
So in the publisher's own cross-validation samples, across two competing factor models and all six age bands, the test labelled as part auditory processing behaves almost entirely as a reading and writing measure. That is not a criticism of the instrument. It is a caution about the inference: a low Word Attack score is not evidence of a phonological processing weakness, and the technical manual is the document that says so.
The reliability runs lowest exactly where screening happens
The same abstract reports Word Attack's internal consistency reliability by age: 0.80 at ages four to five, then 0.94, 0.91, 0.87, 0.85 and 0.88, with a median of 0.88.
The lowest figure sits on the youngest band, which is also the band where the question "is this child on track with decoding" gets asked most often. The practical consequence is a wider confidence interval, which is the subject of what standard error of measurement actually tells you.
Four batteries, four age bands, one letter
The four untimed nonword subtests inside the major achievement batteries are easy to treat as interchangeable. On the four factual axes — what the publisher says it measures, who supplied the normative data, the age band, and the qualification level required to buy it — they are not.
WIAT-4 Pseudoword Decoding. Pearson lists the battery at publication date 2020, age range 4:0 to 50:11, qualification level B.
KTEA-3 Nonsense Word Decoding. Pearson lists publication date 2014, age range 4:0 to 25:11, qualification level B. The battery also offers norm-referenced error analysis and two parallel forms.
WRMT-III Word Attack. Pearson lists publication date 2011, age range 4:6 to 79:11, qualification level B, with within-item-level error analysis on Word Attack specifically.
Woodcock-Johnson V Word Attack. Riverside's WJ IV to WJ V comparison document confirms Word Attack carries forward into the fifth edition, which Riverside began shipping in 2025.
TOWRE-2 Phonemic Decoding Efficiency. PRO-ED lists Test Level B, ages 6:0 through 24:11, five to ten minutes, copyright 2012. Pearson, which also sells it, lists publication date 2012, age range 6:0 to 24:11, qualification level B.
Two things are worth pulling out of that list.
The first is that the TOWRE-2 is the one case in this cluster where the two sellers agree. That is not the pattern we have found elsewhere: the vocabulary instruments post documented the same test carrying three different qualification letters at three different sellers, and the qualification levels post established that a letter is a purchasing rule for one storefront rather than an industry scale.
Here, for once, B means B at both. Do not generalise from that — check the storefront you are actually ordering from.
The second is the WRMT-III's norms. Pearson's own page records that they were "conducted from July 2009 through June 2010 on a nationally representative sample of 3,360 individuals." Those norms are now around sixteen years old.
Whether that makes a score too old is not a question with a ten-year rule behind it, as we found when we went looking for one in what actually makes a score too old — but the date belongs in the report, and it belongs in your decision about which battery to buy.
The nonword lists are curated artefacts, and one criterion is taste
A pseudoword list is not a sample of anything. It is built, item by item, under rules the publisher chooses, and FAST's chart submission is the most candid published account of what those rules look like.
A word bank of 378 consonant-vowel-consonant or vowel-consonant pseudowords was created. Items beginning ci- or ce-, and items ending in -l, -q, -r or -y, were excluded. None of the items appear in a Standard English dictionary.
The first fifteen to twenty items of each progress monitoring form carry unique initial sounds in random order; the remaining thirty to thirty-five are drawn randomly from the bank, for fifty items per form, arranged in ten rows of five. The bank now runs to a thousand items.
And one more exclusion, stated in FAST's own words on the chart: "All inappropriate sounding words were not included in the word bank."
A nonword list has a taste filter. That is obviously the right call, and it is also a reminder that the item pool is a designed object rather than a draw from the language.
The scope of the pool matters for interpretation. Acadience's published description restricts its items to VC and CVC strings, and the measure is administered from the middle of kindergarten to the beginning of second grade.
A measure built from VC and CVC items cannot tell you anything about whether a student can handle a consonant blend, a vowel team or a multisyllabic string — which is the territory the programs in our syllable division post spend their middle levels on. The instrument is not failing. It was never asked the question.
The federal practice guide names the task once
We have found this pattern enough times now that it is worth stating as a method rather than a surprise: take the governing document, extract the text, and count.
The IES and What Works Clearinghouse practice guide Foundational Skills to Support Reading for Understanding in Kindergarten Through 3rd Grade is the federal document that practitioners cite for decoding instruction. It extracts to roughly 58,900 words. Counted as whole words, it contains:
"pseudoword" and "pseudowords": zero times each.
"nonsense word" and "nonsense words": zero times each.
"word attack": zero times.
"nonwords": exactly once.
Against those, "phonics" appears 88 times, "decodable" 36 times and "decoding" 32 times.
The single appearance of "nonwords" is in the guide's own table of constructs, under the heading Word reading, in the column listing how the construct is measured: "Reading aloud lists of words or nonwords." That is it. The guide that recommends the instruction eighty-eight times over names the task the field uses to measure it once, in a definitional table, and never names a single instrument.
This is the same shape as the finding behind our post on decodable text criteria: the federal record mandates an outcome and leaves the mechanism to the market. It is not a scandal. It is simply the reason that when you want to know what a nonword score means, the publisher's manual is the only document that will tell you.
And the American Educational Therapy Association's Code of Ethics, which runs to roughly 2,980 words, contains "decoding" zero times, "phonics" zero times, "pseudoword" zero times and "nonword" zero times.
It uses "assessment" seven times. This is now the seventh consecutive post in which the Code turns out to be silent on the subject at hand, which is a fact about the genre of the document rather than a failing of it — a code of ethics governs conduct, not method. It does bind you here, and we come back to how below.
What the research actually tested, and what it found
The standard objection to nonword measures in practitioner conversation is that they penalise good readers, who will read flarm as "farm" because their lexical knowledge gets in the way. It is a reasonable-sounding worry. Somebody tested it.
Castles, Polito, Pritchard, Anandakumar and Coltheart, writing in the Australian Journal of Learning Difficulties in 2018 (doi:10.1080/19404158.2018.1549088), examined the errors made by sixty-four Year 2 children reading twenty monosyllabic nonwords aloud.
The children came from three primary schools in Melbourne, with a mean age of eight years zero months and a range of six years eleven months to eight years nine months. Across the group the mean error count was 8.67 of twenty, with a standard deviation of 6.02. One child made no errors and one erred on all twenty items.
The authors split the group at the median raw score on the Word Identification test of the Woodcock Reading Mastery Tests and compared error types. The stronger word readers were significantly less likely to produce a lexicalisation error than the weaker readers, F(1, 61) = 4.88, p = .031.
The stronger readers' predominant error was producing another, incorrect nonword highly similar to the target — a mean of 72 per cent of their total errors. The authors conclude that nonword reading measures are a valid index of phonics knowledge and that these tests do not disadvantage children who are already reading words well.
That is a direct answer to the field's most common objection, and it points the other way from the folk wisdom. It also has to be read for what it is. Sixty-four children, one year group, one twenty-item nonword set, one Australian sample, one median split.
It is evidence that the lexicalisation worry is overstated in that population on that task. It is not evidence about any individual student in front of you, and it does not validate any particular published instrument.
The same paper supplies a useful piece of context on the policy side: England's Phonics Screening Check, which uses pseudowords, had a pass rate of 58 per cent when it was introduced in 2012, rising to 82 per cent by 2018 on the Department for Education's figures. The authors read that as improved compliance in delivering systematic phonics instruction as much as improved phonics knowledge, which is a distinction worth keeping.
Where a nonword screen stops working
The clearinghouse data on this is specific, and it is the kind of number the charts publish and almost nobody reads.
NCII's academic screening chart carries an entry titled DIBELS 8th Edition Nonsense Word Fluency Correct Letter Sounds — note that the rated tool is the letter-sound score, not the whole-word score.
For fall administration, with the criterion measure being the DIBELS Next composite at kindergarten and the Iowa Assessment Total Reading score at grades one through three, and a cut point at the twentieth percentile on the criterion, the submitted area under the curve values are:
Kindergarten: 0.82, with a ninety-five per cent confidence interval of 0.76 to 0.87.
Grade 1: 0.79, confidence interval 0.71 to 0.88.
Grade 2: 0.83, confidence interval 0.77 to 0.90.
Grade 3: 0.66, confidence interval 0.51 to 0.82.
All four samples are from spring 2019. The grade three figure is the one to sit with. An area under the curve of 0.50 is chance, and the lower bound of the grade three interval is 0.51.
By third grade, on the publisher's own submitted data, a nonword screen's ability to sort students by later reading risk is no better than a coin toss at the bottom of its confidence interval.
That is not a surprise once you think about what the measure is for. Nonword reading discriminates well while the alphabetic principle is still being acquired and poorly once most students have acquired it, because there is no variance left to detect.
This is the floor-and-ceiling story that runs through screening tools generally, and it has a direct operational consequence: a nonword screen is a good instrument for a six-year-old and a weak one for a nine-year-old, and the decision about which to use is about the age band rather than about the brand.
For comparison, the iSTEEP Nonsense Word Fluency entry on the same chart reports a grade one area under the curve of 0.91 with an interval of 0.84 to 0.98, a base rate of 0.22, sensitivity of 0.80, specificity of 0.90 and positive predictive power of 0.70.
Different sample, different criterion, and not comparable to the DIBELS figures as though they were two readings of one quantity — but it does show that a well-placed nonword screen in the right grade can perform.
Nonword measures live on both charts, which is itself informative
Nonword measures appear on NCII's screening chart — DIBELS 8th Edition Nonsense Word Fluency Correct Letter Sounds, iSTEEP Nonsense Word Fluency and Star CBM Expressive Nonsense Words — and on the progress monitoring chart, where FAST earlyReading Nonsense Words sits.
That dual listing is a genuine difference from the construct in our rapid automatized naming post, which is a screening domain and not a progress monitoring one, for the straightforward reason that nobody who publishes a RAN measure claims you can teach it. Nonword decoding is a teachable skill, it has alternate forms built for repeated administration, and that is why it is on both charts.
If you are choosing something to re-administer every two weeks in a private practice, that matters more than any of the accuracy statistics above, and it is the reason the TOWRE-2's four equivalent forms and FAST's fifty-item forms exist at all. The forms question is the one we worked through in decision rules for progress monitoring.
One caution on the criterion measures
Star CBM's Expressive Nonsense Words entry reports classification accuracy against DIBELS Nonsense Word Fluency as the criterion. That is a nonword measure validated against another nonword measure.
It is a legitimate and common design, and it tells you about agreement between two instruments rather than about either one's relationship to later reading. Read it as convergent evidence and not as predictive evidence, which is the general point we made about what the chart bubbles are actually claiming.
Can a pseudoword decoding test identify dyslexia? No.
It cannot, and no publisher in this group claims it can.
A nonword score is a sample of behaviour on a specific task under a specific scoring rule. It is one input to a diagnostic process conducted by somebody qualified to conduct one, and on a private educational therapy caseload that is usually not you.
This is the same boundary we set out in detail in what you may do with a psychoeducational report, and the boundary does not move because the task happens to be the one most strongly associated with decoding in the research literature.
Two further limits follow from the material above rather than from the ethics.
A low nonword score is not evidence of a phonological processing weakness, because the Woodcock-Johnson V's own factor analyses show its nonword test loading at around 0.21 on auditory processing. And a low whole-word nonsense word fluency score is not evidence that a student cannot decode, because on Acadience's rule a student who decodes every item correctly by sounding out and blending scores zero on that measure by design.
What a low score does license you to do
Plenty, and this is the part the caveats can obscure.
Choose where in a sequence to start instruction. A nonword task with VC and CVC items tells you whether single-syllable short-vowel decoding is secure. That is a scope-and-sequence placement question, and it is squarely inside practice.
Separate decoding from word recognition. The TOWRE-2 pairs Phonemic Decoding Efficiency with Sight Word Efficiency precisely so the two can be read against each other. A student strong on one and weak on the other is telling you something a composite would hide.
Monitor the thing you are teaching. If your instruction is phonics instruction, a nonword measure with alternate forms is a closer match to the taught content than a passage measure. The dosage and interval questions are the ones in mastery criteria for reading intervention.
Notice when the measure has stopped being informative. When a student reaches ceiling on VC and CVC nonwords, the measure is done and you need a different one. That is a planning signal, not a plateau.
Write it down precisely. "Forty-two correct letter sounds per minute, DIBELS 8 Nonsense Word Fluency, fall benchmark form" is a defensible session note. "Low nonword score" is not, for the reasons in session notes and progress reports.
What AET's Code actually binds you to here
The Code of Ethics is silent on decoding and silent on nonwords, as the word counts above show. It is not silent on what you may do with a test.
The relevant provisions are the ones about working inside your competence and seeking interpretation from related-field professionals, which we have unpacked against the Fact Sheet's own wording in the practitioner scope map. The Fact Sheet's instruction not to administer cognitive, intelligence or psychological tests carries its parenthetical — "unless otherwise qualified to do so" — and that parenthetical is doing real work for anyone who holds a separate credential.
Two practical notes that sit underneath the ethics rather than inside it. The qualification level attached to a battery is a purchasing rule set by a storefront and is not a statement about what your credential authorises; the two questions are separate and only one of them is answered at the checkout.
And if you are administering any of these remotely, the permission comes from the publisher's own telepractice terms, test by test, which is the territory of what publishers permit for remote administration. The same applies to any departure from standard administration, which is bounded by the manual rather than by your judgement — see what the assessment manual permits.
Nothing here is legal advice. Questions about what a statute or a state agency requires of you belong with an attorney licensed in your state.
A decision sequence, not a decision
We do not tell you which instrument to pick for a described student, for the reasons set out in our instrument comparison framework. We can give you the order in which the questions settle.
What is the score for? Placement in a sequence, repeated monitoring, or a figure for a report that someone else will read. These point at different families and the families do not substitute.
What age band? Below roughly second grade, a one-minute curriculum-based nonword measure is doing its best work. Above it, the discrimination data above says to be careful, and an untimed battery subtest is the more defensible choice for a report.
Timed or untimed? If the referral question is about accuracy, a clock adds noise. If it is about automaticity, the clock is the measure.
Which scoring rule? If you are reporting or inheriting a whole-word nonsense word fluency figure, establish which system produced it before you compare it to anything.
How many forms do you need? Repeated administration needs alternate forms. One form administered six times measures memory as much as decoding.
Can you buy it, and may you use it? The storefront letter and your own scope are two separate gates, in that order.
What will the report say? Write the unit, the form, the date and the administration conditions, or the number will not be interpretable in six months, including by you.
Where this sits next to the other assessment posts
This post owns one question: what a printed-nonword task can and cannot show. The neighbouring pieces own the others, and it is worth saying which is which.
The qualification levels post owns the letters at the till. The dyslexia screening post owns what a screener is and is not. The spelling assessment post owns the production side of the same orthographic knowledge.
The phonological awareness post owns the pre-print constructs this task is sometimes wrongly used to infer. The oral reading fluency post owns rate in connected text, and the grade equivalent scores post owns the derived score that does the most damage on a report.
Further out: what counts as evidence-based across six bodies, the reading intervention programs these scores feed instructional decisions into, repeated reading protocols where the unit problem recurs, the science of reading as a phrase, and when to discontinue an intervention, which is the decision these measures are most often asked to support.
If you are newer to the field, the role distinctions are in our pillar on the educational therapist, the learning specialist and the executive function coach.
The same pattern runs on the orthographic side of word reading, and our post on orthographic processing assessment is the companion to this one. The Word Attack factor-loading problem described above has a direct parallel: an independent re-analysis of the WIAT-4's own correlation matrices found that its orthographic subtests do not dissociate from the decoding subtests at any age band, which makes a low orthographic composite in a poor word reader the same difficulty measured a second way rather than an explanation of it.
Real-word phonics surveys and inventories sit alongside this, and they answer a different question: which phonics elements a student can read, element by element, rather than how fast a nonword is decoded. Phonics screeners and phonics inventories are compared separately, including why NCII rates the timed measures and files the untimed inventories on a different, unrated list.
Where the published record runs out
Six things we could not establish from the public record, stated plainly so you know what is missing rather than inferring that it is settled.
The acceptable-pronunciation lists. Every untimed nonword subtest must publish a set of acceptable responses for items with more than one defensible pronunciation, because that is the only way the task can be scored reliably. Those lists live inside the examiner's manuals, which are not public documents. We can tell you the rule exists. We cannot tell you how any two publishers' lists compare, and that comparison would be the most useful thing in this area.
The TOWRE-2 manual erratum. PRO-ED's product page carries a notice headed "Important Update for TOWRE-2 Users: TOWRE-2 Examiner's Manual Page 42 Errata." We have not been able to read the erratum itself, so we cannot tell you what it corrects. If you administer the TOWRE-2, that is a two-minute question for PRO-ED and worth asking.
The kit contents. PRO-ED and Pearson both sell a TOWRE-2 complete kit and their published component lists are not identical. Confirm what is in the box with the seller before ordering, especially if the alternate forms are the reason you are buying it.
The WJ V Word Attack item structure. We have the published description of the task and the technical abstract's factor and reliability data. The number of items at which the test transitions from isolated letter sounds to whole pseudowords is not in the public material.
Whether DIBELS 8 and Acadience whole-word scores have ever been equated. Given the scoring difference documented above, somebody may have published a crosswalk. We did not find one, and neither publisher offers one.
The current England Phonics Screening Check figures. The 58 and 82 per cent pass rates above are the Department for Education figures as reported in a 2018 paper. We have not verified the position since.
Frequently asked questions
What is a pseudoword decoding test?
It is a task in which a student reads aloud printed letter strings that follow English spelling conventions but are not English words, such as "rab" or "vit". Because the items are not real words, the student cannot recognise them from memory and has to apply letter-sound knowledge. Common examples are WIAT-4 Pseudoword Decoding, KTEA-3 Nonsense Word Decoding, Woodcock-Johnson V Word Attack, WRMT-III Word Attack and the TOWRE-2 Phonemic Decoding Efficiency subtest.
Is nonsense word fluency the same as a pseudoword decoding subtest?
No. Nonsense word fluency measures are one-minute curriculum-based measures that report correct letter sounds per minute alongside a whole-word count. Pseudoword decoding subtests inside achievement batteries are untimed and report items correct as a standard score. The units do not convert and no publisher offers a conversion between them.
Do DIBELS and Acadience score nonsense word fluency the same way?
Not on the whole-word score. The University of Oregon's DIBELS 8 guide awards a words-read-correctly point for a word read correctly "whether sounded out initially or not". Acadience's published rule credits a whole word only when it is read as a whole word "without first being sounded out". The identical student response therefore scores one point on DIBELS 8 and none on Acadience.
Can a pseudoword decoding test diagnose dyslexia?
No. It samples behaviour on one task under one scoring rule and is one input to a diagnostic process carried out by a qualified professional. On most private educational therapy caseloads that professional is not the educational therapist, and the AET Fact Sheet's instruction on administering cognitive, intelligence and psychological tests carries its own parenthetical for anyone separately qualified.
Does a low nonword score mean a phonological processing problem?
Not on the available evidence. Riverside classifies Woodcock-Johnson V Word Attack under both reading-writing and auditory processing (achievement battery CHC classifications), but its own technical abstract reports Word Attack loading at a median of 0.21 to 0.22 on auditory processing.
Against that it loads 0.60 to 0.75 on reading and writing, across two factor models and all six age bands. The score is a reading measure, and the technical abstract is the document that says so.
At what age do nonword screens stop discriminating well?
The published classification accuracy data suggests the decline is steep by third grade. NCII's chart entry for DIBELS 8 Nonsense Word Fluency Correct Letter Sounds reports a fall area under the curve of 0.82 at kindergarten, 0.79 at grade one, 0.83 at grade two and 0.66 at grade three, with a lower confidence bound of 0.51 at grade three against a chance value of 0.50.
Do nonword measures penalise strong readers?
The one study that tested the claim directly found the opposite. Castles and colleagues, examining sixty-four Year 2 children reading twenty nonwords, found stronger word readers significantly less likely to make a lexicalisation error than weaker readers, and concluded that nonword measures are a valid index of phonics knowledge. That is one small sample in one year group, not a general law.
Which nonword measure should I buy for a private practice?
That depends on what the score is for, the age band, whether you need a timed or untimed measure, how many alternate forms repeated administration requires, and both the storefront's qualification rule and your own scope. The sequence above works through those in order. We do not recommend an instrument for a described student.
Join the Illuminate educator network
If this is the kind of thing you want more of — the published record read closely, the units checked, and the caveat given rather than buried — we write for educational therapists, learning specialists and executive function coaches, and we send the new work out as it lands.




Comments