Skip to content

Universal Early Literacy Screening

Principle · All principles
Evidence · 39 claims (9 for, 30 mixed) · 18 studies (11 associational, 2 quant-synthesis, 2 design, 1 causal, 1 review, 1 theoretical), q1–q3 · 0 of 18 report an effect size · 38 claims rest on one study

Conditional relationship

This page owns finding, early and for every child, who may be at risk of reading difficulty, and what happens next. In the first years of school (kindergarten to grade 3, child, pre-k-12), giving every learner a brief, validated literacy screener at set points in the year (commonly the beginning, middle and end), and acting on each result within weeks (a closer diagnostic look for anyone flagged, targeted reading instruction, more frequent progress monitoring, a rescreen), is expected to identify more children whose early reading skills (phonological-awareness, letter knowledge, decoding, reading-fluency) are behind, sooner, than waiting for teacher referral or for a failed grade 3 test, and so to give more of them instruction while the gap is smaller. Screening is a measurement step: any benefit to children comes from the instruction it triggers, not from the screen itself.

The relationship is conditional on four things: the screener predicts later reading on the measures that matter locally, with known error; the cut score is read as a flag for a closer look, not as a diagnosis or a label; flagged children actually receive added instruction and are rescreened; and the screen measures reading risk rather than something else the child brings (proficiency in the language of the test, unfamiliarity with the format).

What the wiki's evidence tests, and what it does not. No claim in this wiki compares children who were screened with children who were not, or screening three times a year with once or twice; the earlier page's prescription (three times a year in K–3, plus screening for characteristics of dyslexia at the end of kindergarten and the start of grades 1 and 2) comes from a policy toolkit, not from a study. What the wiki holds is evidence on how well screeners work as measures, almost all from one research programme (WestEd's analyses of Massachusetts screening data, 2023–2025, several reports) plus one peer-reviewed validity study of DIBELS in urban kindergartens (Rouse and Fantuzzo 2006) and a vendor's accuracy study (NWEA MAP Growth). Read against this model:

Group differences in who is flagged are gaps in circumstances, treatment and possibly measurement, not causes. The Massachusetts reports find children from low-income families, English learners, children receiving special education services, and Black and Hispanic children flagged more often and more repeatedly than others (Students from low income, English learner, special education, and several racial/ethnic groups are more likely to be repeatedly identified as significantly below benchmark, with low income students 2.5 times as likely [~W]; Low income, English learner, special education, Black, and Hispanic students were more likely to be classified significantly below benchmark at each time period and more than once [~W]; Students from low income backgrounds, English learners, students receiving special education services, and Black and Hispanic students were more often identified as at significant risk [~W]). Income, language status and race do not make children less able to learn to read. The same reports show what these gaps track: schools with more low-income students, lower teacher retention and lower attendance flag more of their children (School characteristics—percentage of low income students, teacher retention, and attendance—are associated with students' likelihood of being identified as at significant risk in early literacy screening [~W]), which is a difference in the instruction and resources children have had. Part of the gap may also be the measure: 90 percent of Massachusetts English learners were screened only in English (English-only screening is the norm for English learners in Massachusetts: 90 percent were assessed only in English across all three screening periods in 2023/24 [~W]), so an English screener partly measures how much English a child has learned so far, not only their reading risk. The design consequence is the same either way: a group flagged more often needs more instruction and better-matched measures, never a lower expectation or a label.

How the converted siblings sit around this: Formative Assessment owns using evidence of learning to adapt teaching within lessons; screening is the coarse, whole-population version of that loop, run a few times a year. Phonics owns the instruction most screening flags call for in kindergarten to grade 2. Flexible Grouping owns forming temporary similar-need groups from recent evidence, which is how a flag usually turns into instruction. Validity, reliability and bias in classroom assessment, Standardized Test Fairness and Bias and Criterion- and Norm-Referenced Testing own the measurement questions this page applies to screeners.

Default design, while the relationship is untested

The earlier page's policy (universal screening three times a year in K–3 with progress monitoring, screening for characteristics of dyslexia at the end of kindergarten and the start of grades 1 and 2, diagnosis only by a licensed professional), kept as a concrete default and extended into an arc. Each step says what evidence, if any, stands behind it; numbers are untested proposals unless a claim is named.

  1. Choose a screener for its accuracy on the outcome that matters locally, and keep it. Prefer one with published classification accuracy (sensitivity, specificity, area under the curve) against the reading measure the school cares about, for each grade and term. One vendor's study reports accuracy of at least 0.8 on these indices for most grades and terms, lower in kindergarten to grade 2 (Recommended MAP Growth screening cut scores yield sensitivity, specificity, and lower-bound AUC of at least 0.8 for most grades and terms [+M]; Classification accuracy of the screening cut scores is lower for Grades K-2 than Grades 3-8, attributed to a 12-36 month lapse between MAP Growth and criterion measures [~M]; the vendor's own study). Do not switch screeners without re-reading the new one's cut scores, because risk rates are not comparable across screeners (above).
  2. Screen every child near the start of the year, then at midyear and the end. Screen as soon as school begins so that instruction can start; a midyear screen is worth keeping because those on track by then mostly stay on track and those still flagged need a change (Most students at risk at BOY remain at risk later, but students who meet benchmark by midyear tend to stay on track [+W]). The three-times-a-year schedule itself is policy, not a tested comparison; Kansas schools that adopted it largely met it (Data-based decision making is the least-met full-implementation criterion while universal screening is the most-met [~W]).
  3. Measure the skills that predict reading at each grade. In kindergarten, letter naming and phonemic awareness (segmenting sounds); from grade 1, word reading and oral reading fluency. Letter naming fluency was the strongest kindergarten predictor in one cohort, and nonsense-word fluency added nothing once the other subtests were known (Kindergarten DIBELS subtests predict first-grade instructional reading, with Letter Naming Fluency the strongest predictor [+M]; Nonsense Word Fluency adds no unique predictive information about first-grade reading when considered simultaneously with other DIBELS subtests [~M]; one urban kindergarten cohort). At the end of grade 1, a child's level of oral reading fluency predicted later comprehension better than their rate of growth (End-of-Grade-1 oral reading fluency status predicts later reading comprehension better than growth slope, though slope added to first-grade screening accuracy in one study [~M]).
  4. Treat a flag as a question, and look closer within two weeks. For every child below the cut score, give a short diagnostic assessment of the component skills (sounds, letters, decoding, fluency, vocabulary) and gather the teacher's observations before deciding on instruction. A screen only sorts; children just either side of a cut score showed no difference in later state-test performance in one exploratory analysis (Exploratory analysis found no observed differences in later MCAS performance between students just below and just above screening risk thresholds [~W]), so treat children near the cut as borderline in both directions. The two-week window is a proposal.
  5. Give flagged children added instruction matched to the diagnosis, and monitor it. Small-group instruction in the skill the diagnostic points to, in addition to core reading instruction, with a short progress check every two to three weeks; change the instruction when the checks do not move. One phonological-awareness programme screened children every three weeks to target its lessons (PA Lessons implementation included testing students with the Phonological Awareness Screener for Intervention every three weeks to target instruction [~W]; an implementation detail, not a tested component). The instruction itself belongs to Phonics and Flexible Grouping. Frequency and group size are untested proposals.
  6. Do not let a "met benchmark" result clear a child the teacher is worried about. Because screeners miss many children who later struggle (above), keep a route for teacher referral and recheck any child whose classroom reading does not match a passing screen.
  7. Screen English learners in a way that separates language from reading risk. Where a screener exists in the child's home language, screen in both languages, and read an English-only result alongside the child's English proficiency; do not place an English learner in reading intervention on an English screener alone. Practice in Massachusetts differed by programme type: children in dual-language programmes were more often screened in Spanish (Screening language practices are related to English language education program type: SEI students were mostly screened in English while DLBE students were more often screened in Spanish or alternately [~W]). The rule is an untested proposal.
  8. Screen for characteristics of dyslexia where policy requires it, and refer for diagnosis. At the end of kindergarten and the start of grades 1 and 2, add the subtests the state names (often rapid naming and phonological tasks) and refer children whose pattern persists despite good instruction for an evaluation by a licensed professional; a screen is not a diagnosis. This is the toolkit's policy; untested here.
  9. Review the whole system every year, by school and group. Tabulate how many children are flagged, how many get added instruction within weeks, and how many reach benchmark by the next screen, by school and by group; where one school or group is flagged more often, look at what instruction and resources its children have had and at the language of the screen before looking at the children. Kansas schools met the screening criterion far more often than the one about using the data to decide instruction (Data-based decision making is the least-met full-implementation criterion while universal screening is the most-met [~W]). The review cycle is a proposal.
  10. Plan the arc across the year and the grades. A default, all untested proposals:
    • First two to three weeks of the year. Screen everyone (5–10 minutes per child); diagnostics for those flagged within two weeks; added instruction starts by week four.
    • Autumn to midyear. Flagged children get 20–30 minutes of added small-group instruction three to five days a week; a progress check every two to three weeks; regroup on the checks.
    • Midyear screen. Children now at benchmark leave added instruction but are watched at the next check; children still flagged get a change (more time, a smaller group, a different focus) or a fuller evaluation.
    • End-of-year screen. Results go to next year's teacher with the instruction tried; a child flagged at the end of two consecutive years is referred for evaluation.
    • Across grades. Act most intensively in kindergarten and grade 1, where risk persists less from year to year (Risk status tends to persist: 75 percent of students significantly below benchmark at BOY remained so at EOY, and earlier grades showed lower year-to-year persistence [+W]).

When a condition is not met: if flagged children cannot receive added instruction, screening only labels them and should be paired with a plan for the instruction before it starts; if the screener's accuracy is unknown, treat its flags as provisional and weigh teacher observation more.

Fitting the design to a situation

The three facts that most change the decision:

  • What happens after a flag. Screening without added instruction changes nothing for children. If unstated, ask: who will teach the flagged children, when, for how long, and how soon after the screen?
  • Which screener, against which outcome. Accuracy differs by screener, grade and outcome, and risk rates are not comparable across screeners. If unstated, ask: which screener, what are its published sensitivity and specificity for this grade, and what later reading measure does the school care about?
  • Who the children are and which language they are learning to read in. An English screener partly measures English for English learners, and some groups are flagged far more often for reasons of circumstance. If unstated, ask: how many English learners, in which programme, and is a home-language screener available?
If the brief says… Then change… Basis
Kindergarten (child, age 5–6) Screen letter naming and phonemic awareness; skip nonsense-word fluency if the other subtests are given; expect lower accuracy than in later grades Kindergarten DIBELS subtests predict first-grade instructional reading, with Letter Naming Fluency the strongest predictor [+M]; Classification accuracy of the screening cut scores is lower for Grades K-2 than Grades 3-8, attributed to a 12-36 month lapse between MAP Growth and criterion measures [~M]
Grade 2 or 3, or the screener is used to predict a state test Use the screen to find children at risk, not to clear the rest; follow up children the teacher is concerned about even if they passed The predictive value of grade 3 screening benchmarks for MCAS ELA proficiency varies widely: 15 to 69 percent of students classified as on track did not reach MCAS proficiency, and students flagged at risk rarely met proficiency [~W]
Many English learners, or a dual-language programme Screen in the home language where a screener exists, as well as in English; read English results alongside English proficiency; do not place in reading intervention on an English screen alone English-only screening is the norm for English learners in Massachusetts: 90 percent were assessed only in English across all three screening periods in 2023/24 [~W] (describes practice; the effect of home-language screening is untested here)
A school serving many children from low-income families, or with high teacher turnover and low attendance Expect more children flagged; budget added instruction and staff for that number before screening; read the rate as a measure of the school's resources and the children's prior opportunities, not of the children School characteristics—percentage of low income students, teacher retention, and attendance—are associated with students' likelihood of being identified as at significant risk in early literacy screening [~W]
Results show some groups (by income, language, race or special education status) flagged more often Check the screen's language and format for those children, check what instruction they have had, and add instruction; never lower expectations or use the flag as a label Students from low income, English learner, special education, and several racial/ethnic groups are more likely to be repeatedly identified as significantly below benchmark, with low income students 2.5 times as likely [~W]; The likelihood of being identified as significantly below benchmark in early literacy screening increases as historically underserved background characteristics intersect [~W] (gaps reflecting circumstances, treatment and possibly measurement, not causes)
The district uses several screeners, or is switching Do not compare risk rates across screeners or years of a switch; re-read cut scores against the local outcome; report which screener each figure comes from Agreement between K–3 screening assessments on risk classification varied from 47 percent to 85 percent depending on the assessment combination [~W]; Screening assessment benchmarks indicating significant risk shift up and down on the MCAS scale within and across grades, so students with the same skills may be classified differently at different times [~W]
The screener's norms were updated, or a percentile cut point is carried over Re-check the cut score against grade-level expectations rather than keeping the old percentile Static percentile screening cut points carried across norm sets can under-identify students who remain below grade level [~W] (an illustrative argument)
Many children sit just above or below the cut score Treat a band around the cut as borderline: a diagnostic look and a recheck rather than an automatic placement either way Exploratory analysis found no observed differences in later MCAS performance between students just below and just above screening risk thresholds [~W]
Scarce staff time for screening and follow-up Screen everyone briefly; spend the scarce time on diagnostics and added instruction for those flagged, not on extra screening rounds untested proposal
Only one screen a year is possible Screen at the start of the year so instruction can begin; rely on progress checks for flagged children in place of a midyear screen untested proposal; the midyear evidence is in Most students at risk at BOY remain at risk later, but students who meet benchmark by midyear tend to stay on track [+W]
State law requires screening for characteristics of dyslexia Add the required subtests at the named points; refer persistent cases for evaluation by a licensed professional; tell families a screen is not a diagnosis untested here; from the policy toolkit in Key Sources
The brief names its own screener, benchmarks, risk categories or MTSS routine Keep them; check only that each flag is followed by a diagnostic look and added instruction within weeks, and that results are reviewed by school and group untested proposal
Screening data are sensitive (special education status, family language, risk labels) Share flags only with the staff who teach the child and with families; do not display risk colours or levels to children; keep group-level reports anonymous untested proposal
Older learners (grade 4 and above, or adults learning to read) Use a screener validated for that age; this page's evidence is K–3 untested here
A high-stakes decision is proposed on the screen (retention, placement in special education) Do not decide on a screen alone; use a full evaluation and the child's response to instruction Literacy screener benchmarks predict which grade 3 students will not meet MCAS ELA expectations, but not necessarily who will meet them [~W]; Screening assessments vary in design, administration, and risk definitions such that there is no truly common definition of risk across assessments [~W]

Ten of the fifteen rows cite a claim; every cited claim is about how screeners classify children (accuracy, agreement, cut scores, who is flagged), and none tests whether screening, or any response to it, improves reading. The other five rows are untested proposals.

Observation, state and explanation

What can be observed: each child's screener scores and classification at each time point, which screener and in which language, the diagnostic results for those flagged, the instruction they then received (when, how long, what focus), their progress checks, and their later reading results. "At risk", "dyslexic", "a struggling reader" and "on track" are inferences from those observations, and the first three are easily turned into labels.

Response Plausible explanations Observation that could change the interpretation
A child is flagged at the start of the year A gap in early reading skills; little prior instruction in letters and sounds; the screen was in a language the child is still learning; unfamiliarity with timed tasks; a bad day A diagnostic of component skills; a home-language screen; the teacher's observation of the child reading; a recheck after two weeks of instruction.
A flagged child stays flagged after a term of added instruction The instruction did not target the skill the diagnostic showed; too little of it was delivered; a persistent difficulty (such as dyslexia) that needs more intensive help; the screen keeps measuring English proficiency Check attendance at and content of the added instruction; compare progress checks on taught and untaught skills; refer for evaluation if instruction was sound and sustained.
A child passes the screen but struggles in class The screener misses this child (screeners miss many later strugglers); the difficulty is in comprehension or vocabulary that the early screen does not measure A teacher-referral check with a diagnostic; a comprehension or vocabulary measure.
One school or group is flagged far more often Fewer resources and less instruction so far; teacher turnover; attendance; the screen's language; a screener with a different cut score Compare by screener and language of screening; look at the school's instruction, staffing and attendance; compare progress after added instruction.
Risk rates change sharply between years Real change in children's skills or in instruction; a change of screener; benchmarks or norms that moved Check whether the screener, its version or its norms changed before reading the change as progress.

Evidence and qualifications

One research programme on one state's screening data. WestEd's analyses of Massachusetts early literacy screening (Lemke and colleagues, 2023–2025, several reports and briefs) supply most of the claims here: descriptive and correlational analyses of district screening data from tens of thousands of children, coded q2, associational, no effect sizes. They are one line of evidence, so their markers stay at W. They show how screeners classify and how classifications persist; they say nothing about the effect of screening or of what schools did with the results.

Peer-reviewed validity evidence for one screener. Kindergarten DIBELS subtests predict first-grade instructional reading, with Letter Naming Fluency the strongest predictor [+M], Kindergarten DIBELS subtests significantly predict first-grade TerraNova vocabulary, language, and reading outcomes [+M] and Nonsense Word Fluency adds no unique predictive information about first-grade reading when considered simultaneously with other DIBELS subtests [~M]: Rouse and Fantuzzo (2006), one urban kindergarten cohort of about 280 children, q3, associational. Prediction of later scores, not classification accuracy at a cut score. End-of-Grade-1 oral reading fluency status predicts later reading comprehension better than growth slope, though slope added to first-grade screening accuracy in one study [~M]: a review's report of two studies that point different ways on growth slope.

A vendor's accuracy study. Recommended MAP Growth screening cut scores yield sensitivity, specificity, and lower-bound AUC of at least 0.8 for most grades and terms [+M] and Classification accuracy of the screening cut scores is lower for Grades K-2 than Grades 3-8, attributed to a 12-36 month lapse between MAP Growth and criterion measures [~M]: the publisher's own study, for one screener. Static percentile screening cut points carried across norm sets can under-identify students who remain below grade level [~W]: an illustration in a white paper.

Screening inside a tiered system. Universal screening data from full-implementation schools show average fall-to-spring gains in reading and math, largest in kindergarten [+W] and Universal screening data from a small sample of full-implementation schools show fall-to-spring benchmark gains in reading and most math grades, largest in kindergarten, but behavior data show losses and the data are inconclusive [~W]: one evaluation of Kansas's multi-tier system; fall-to-spring changes in screening results without a comparison group, which the report itself says cannot be read as causal. Data-based decision making is the least-met full-implementation criterion while universal screening is the most-met [~W]: schools screened far more reliably than they used the results.

Learners, screeners and outcomes differ across these, so do not rank them. The strongest evidence here is about measurement: screeners flag children who will struggle better than they clear those who will not, disagree with each other, and flag some groups far more often.

The evidence by question

Further evidence, not yet read against this model

Claims bearing on screening that this page does not use in the model, read only for their scope.

Objective and learner-valued goal

The designer's objective is usually that every child reads at grade level by the end of grade 3, and, on the way, that no child with a reading difficulty goes unnoticed for long. Screening serves that objective only through the instruction it triggers. Elicit separately what children and families value: to read the books their friends read, not to be pulled out of class in front of peers, not to be labelled, to keep their home language valued. Families may also want a diagnosis (for example of dyslexia) that a screen cannot give. Agreement and divergence are both observations: a family that fears a label may decline intervention a child needs, and a child who values staying with classmates may resist a pull-out group that a push-in arrangement could replace.

What would revise this model?

The expectation should weaken if comparisons of schools or cohorts that screen universally against those that rely on referral show no difference in how early children receive added instruction or in later reading; if screening does not change who gets instruction (flagged children receive nothing more); or if a screener's accuracy for the local outcome turns out too low to add anything to teacher judgement. It should change if home-language or untimed screening substantially changes which English learners are flagged, in which case English-only screening is mostly measuring English. If three screens a year identify no more children in time than one or two, the schedule is a cost rather than a benefit. Do not protect the model by calling every disappointing result "poor implementation" after the fact.

Accurate identification, timely added instruction, reading at grade level and children's own experience of being screened are separate claims. The present evidence establishes no schedule, cut score or screener.

Claims

Examples

Key Sources

  • Comprehensive Early Literacy Policy: Fundamental Principles. (2024). ExcelinEd Policy Toolkit. https://www.excelined.org
  • Rouse, H. L., & Fantuzzo, J. W. (2006). Validity of the Dynamic Indicators for Basic Early Literacy Skills as an indicator of early literacy for urban kindergarten children. School Psychology Review, 35(3), 341-355. doi:10.1080/02796015.2006.12087971
Page metadata
Field Value
type principle
id universal-early-literacy-screening-thrice-yearly
title Universal Early Literacy Screening
description In kindergarten to grade 3, briefly screening every child for early reading risk at set points in the year and following each flag within weeks with a diagnostic look and added instruction is expected to find children who need help sooner than referral; the wiki holds no test of screening itself, and its evidence (mostly one Massachusetts research programme) shows screeners flag later strugglers better than they clear the rest, disagree with each other, and flag children from low-income families, English learners and Black and Hispanic children more often, gaps that reflect circumstances, treatment and possibly the measure, not the children.
status review
canonical True
generated by claude/unspecified on 2026-10-19
trust unverified — no one has confirmed this page's content
sources 2 — comprehensive-early-literacy-policy-fundamental-principles-2024, rouse-2006

Frontmatter is the page's machine-readable record — see the Schema & Guide.