The short version
These numbers exist because of a law. 2023 Wisconsin Act 20 requires every public school district in the state to screen students in 4K through third grade for early literacy skills, and the Department of Public Instruction selected Pearson’s aimswebPlus as the statewide screener. Beginning in 2025–26, screening is required in fall and spring for 4K through third grade, and at midyear for 5K through third — which is why the 4K line in the chart below has a gap in it. Four-year-olds are screened twice a year, not three times. (2024–25 used a different, reduced schedule; see below.)
Statewide screening began the year before, in 2024–25, and Madison has results from that year too. What changed is the cadence: a 2024 amendment, Act 192, reduced the number of required administrations for 2024–25 only, so 2025–26 is the first year run on the full statutory schedule. The two years still shouldn’t be lined up against each other, for reasons that have nothing to do with missing data, covered in full below — including where each year’s numbers came from: the 2025–26 figures were obtained directly from the district, since DPI has not yet published a statewide report for that year. Five things stand out in the 2025–26 results.
-
Benchmark rates fell in every grade except third.
4K dropped from 82% to 66%. Kindergarten fell from 76% to 60%. First grade slid from 73% to 65%. Second grade held roughly steady, 69% to 68%. Third grade sat at exactly 72% at all three windows — flat, not lower.
-
More is expected of students as the year goes on, and the bar rises to match.
aimswebPlus is norm-referenced against seasonal benchmarks, so the score needed to clear the bar rises from fall to winter to spring, and the subtests themselves shift from letter names and sounds toward decoding and passage reading. A student can start the year at benchmark, score higher at every testing that follows, and still finish the year below benchmark — not because they lost ground, but because their pace didn’t keep up with a rising expectation. Whether that’s the main story behind the numbers below is an inference, not a measurement: proving it would take student-level matched scores, the actual seasonal cut scores, and like-for-like measures across windows.
-
The younger the grade, the bigger the drop.
The 16-point declines in 4K and kindergarten are twice first grade’s and far larger than second grade’s, which barely moved. That gradient is the largest pattern in the data — though, as with the rising bar above, the data alone can’t say whether it reflects something distinctive about the early grades or simply that norms, subtests, and measurement precision all shift fastest at those ages.
-
Whatever the cause, spring is where the school year lands.
By May, 40% of kindergarteners and 35% of first graders were below their grade’s aimswebPlus benchmark — a screener cut score, not Wisconsin’s official measure of grade-level reading, which for third grade is the separate Forward Exam. And under Act 20, a student isn’t identified only in spring: a below-25th-percentile result in 5K through third grade starts a diagnostic assessment and personal reading plan the first time it happens, whether that’s fall, winter, or spring. Spring is simply the most complete picture on this page — every grade has been measured three times by then (twice for 4K) — not the only moment that matters under the law.
-
The gaps are large everywhere, and every one of them widened.
In spring, White students were at benchmark at rates of 83% to 89% in every grade. Black students ranged from 41% to 53%, Hispanic and Latino students from 33% to 57%, English learners from 30% to 50%, and students with IEPs from 33% to 42%. Checking every grade against Black, Hispanic, economically disadvantaged, and IEP students — twenty comparisons in all — the gap between White students and each group grew from fall to spring in every single one, sharpest in the youngest grades. The kindergarten gap between White and Hispanic students grew from 42 points in fall to 50 points in spring — the widest single gap anywhere in the data.
The grade-by-grade slide
Read the change column below from top to bottom and the shape is unmistakable: the decline shrinks as the grades get older, until it disappears entirely.
White students’ rates hold up across the year — third grade actually improved, from 87% in fall to 89% in spring. Almost every other group’s did not. The result is that whatever is decreasing early-grade rates does so unevenly, and gaps that were already wide in September were wider in May.
Students with IEPs
The most consistent decline in the file. Every grade ended lower than it began: 4K from 54% to 34%, kindergarten from 49% to 33%, first grade from 49% to 42%, second from 45% to 37%, third from 42% to 39%. No grade finished above 42%.
English learners
A sharp fall in kindergarten, from 58% to 30%, but nearly flat in grades one through three — 58% to 49%, 50% to 48%, and 51% to 50%. The kindergarten drop tracks the point in the year when the screener’s language demands increase, which is worth knowing before drawing conclusions about instruction.
Asian students
The quiet exception to the pattern that high-performing groups stayed high. They fell from 89% to 75% in kindergarten and from 79% to 60% in 4K — declines closer to the district’s than to White students’. Whatever is decreasing early-grade scores is not simply a function of where a group started.
What the screener actually is
aimswebPlus, from NCS Pearson, is a PreK–12 reading and math screening and progress-monitoring system. Wisconsin’s Department of Administration selected it through a public bid, DPI contracted with Pearson in July 2024, and the early literacy and reading measures are provided to Wisconsin schools and districts at no cost. Madison, like every other public district in the state, did not choose this instrument.
The screener also measures different aspects of literacy at each grade. For Act 20 compliance in grades 2 and 3, only two measures count: Vocabulary and Oral Reading Fluency. In kindergarten, the Early Literacy Composite combines two of the six required measures; in first grade, three of seven. So “68% at benchmark in Grade 2” and “72% at benchmark in Grade 3” aren’t two scores on one ruler — they’re two grades being measured on different things. Comparisons within a single grade, across the school year, are more meaningful than comparisons across grade levels.
Screening is standardized. Diagnosis isn’t.
Act 20 has two stages, and only the first is uniform. Every district screens on aimswebPlus — that’s the fall/winter/spring benchmark data on this page. The second stage, a diagnostic assessment for students below the 25th percentile, is a local choice: Pearson’s own guidance for Wisconsin says diagnostic measures are “what you decide, locally, to use.” Madison’s 2024–25 filing names its diagnostic tool as FastBridge and Star by Renaissance Learning — not aimswebPlus — for every school and grade that reported one. A student’s screener and their diagnostic assessment can be, and in Madison’s case are, two different products.
That split is common statewide. Of the 429 Wisconsin districts that reported a diagnostic tool for 2024–25, aimswebPlus was the choice for most, but not all:
Of 430 Wisconsin districts and independent charter operators in the 2024–25 Act 20 report, 429 named at least one diagnostic tool; one reported none. Many districts named more than one — most often aimswebPlus paired with FastBridge/Star or with i-Ready — so each is counted once per tool named and the column does not sum to 429.
Why 2024–25 and 2025–26 don’t line up
Madison has a prior year of screener results, published in DPI’s 2025 Act 20 annual report. It shouldn’t be compared with 2025–26, for three reasons. Pearson re-normed aimswebPlus for 2025–26, replacing a 2015 reference sample, and DPI says results from before and after that change aren’t directly comparable. The schedule differed, too: 2024–25 began with the midyear window instead of fall, so 4K was screened just once that year, in spring, and 5K through third grade were screened twice, at midyear and in spring — not the fall-and-spring, three-times-a-year schedule described above, which took effect in 2025–26. And the two years aren’t even the same kind of source: the 2024–25 figures below are DPI’s own statewide compilation, while the 2025–26 figures on this page came directly from Madison, since DPI hasn’t yet published a 2025–26 report to check them against.
Madison’s 2024–25 numbers
DPI’s Act 20 annual report doesn’t publish a benchmark rate. It publishes one figure per grade: the share of students below the 25th percentile — the statutory threshold, detailed further down this page, that triggers a diagnostic assessment and personal reading plan in 5K through third grade. Madison’s district-wide figures for 2024–25, the first year of statewide screening (also viewable on an interactive map of DPI’s published data):
† DPI’s published file redacts Madison’s first-grade count directly. The figure above is estimated from the number of first graders who began a personal reading plan (1,098 of 1,879 enrolled, or 58.4%) — a close stand-in everywhere else in the file, where the personal-reading-plan count sits within a point of the below-25th-percentile count.
This is not the benchmark rate. A percentile ranks a student against other test-takers; a benchmark is different — it’s the score Pearson has found predicts roughly an 80% likelihood that a student will be reading at grade level by the end of the year. Publicly available Pearson technical documentation puts that cut near the 45th percentile for most reading measures, and closer to the 35th percentile for the early-literacy measures used in the youngest grades. The 25th percentile, by contrast, is just the bottom quarter of test-takers — a lower, simpler line that Act 20 uses because a statute needs one, not because it was calibrated to predict future reading success the way Pearson’s benchmark was. “At or above benchmark,” the number used everywhere else on this page, and “at or above the 25th percentile,” the only number DPI compiles statewide, are answering different questions. There is no published conversion between them. Setting the 2024–25 at-risk rate beside the 2025–26 benchmark rate would not show a year-over-year trend — it would show the distance between two different rulers, applied in different years, under different norms. Both problems compound rather than cancel.
How widely is it used?
Pearson does not publish adoption counts, and no reliable public tally of aimswebPlus states, districts, or schools exists — so any specific number should be treated with suspicion. What can be verified is narrower:
- Wisconsin is a single-screener state. Every public school district and independent charter school in Wisconsin screens 4K through third grade on aimswebPlus under one statewide contract — that mandate covers the screener only, not the diagnostic tool used afterward; see above.
- Elsewhere, aimswebPlus generally appears as one option among several on state-approved screener lists rather than as a statewide mandate — Oregon and Arizona both list it among approved universal screening tools, alongside Acadience, DIBELS, mCLASS, and others.
- The broader context is that early literacy and dyslexia screening mandates are now close to universal: 39 states had adopted such a requirement before California’s took effect in 2025–26. The mandates are widespread; the instrument varies by state and often by district.
What “below benchmark” does and doesn’t mean
Everything on this page is a benchmark rate. Act 20 does not run on benchmark rates. The statute turns on a different threshold — the 25th percentile — and applies it differently by grade:
Act 20 “at-risk” determination
| Grade | How a student is identified |
|---|
| 4K | Below the 25th percentile on both required spring subtests. No student is identified from fall results. No diagnostic assessment or personal reading plan is required. |
|---|
| 5K & Grade 1 | Below the 25th percentile on the specified composite. |
|---|
| Grades 2–3 | Below the 25th percentile on oral reading fluency. |
|---|
For 5K through third grade, a student below that line must receive a diagnostic assessment — including a family history survey — and a personal reading plan, with notification to families and ongoing progress reporting. So behind every percentage on this page sits a set of individual obligations. But the two thresholds are not the same, and the number of students “below benchmark” is not the number legally identified as at risk. Nor can a page of district percentages establish whether Madison met its Act 20 duties: compliance is documented student by student, not in aggregate.
Context this data doesn’t contain
Madison purchased a new early literacy curriculum in 2022, following a 2021 task force report. These 2025–26 results fall in the fourth year of that adoption — which makes them a natural baseline and a poor before-and-after. No aimswebPlus series exists from before the purchase, because the statewide screener was not in use until 2024–25; and that one prior year was scored against different norms, so it cannot serve as a baseline either.
Where the data is thin
An honest read includes what the file cannot tell us. American Indian/Alaska Native and Native Hawaiian/Pacific Islander results are masked at every grade and every window, so those students are absent from this analysis entirely. Several 4K and kindergarten cells for English learners and Advanced Learners are masked as well.
A handful of published cells also look like reporting errors rather than findings. Winter first-grade results for Black students show 47% alongside a count of 47, where the count should be closer to 150. Spring 4K results for English learners show 39% alongside a count of 9. Anyone citing those specific cells should confirm them with the district first.
The question the data raises
If the early-grade declines are largely an artifact of an advancing bar, the spring snapshot is still the most complete one available — and it says that between a quarter and two-fifths of Madison’s youngest students finished the year below their grade’s benchmark, with that share climbing past half for Black, Hispanic, English-learning, and IEP students in most grades. That is a benchmark rate, not a legal determination; a student can be identified as at-risk under Act 20 from any window’s result, not only spring’s.
Screening data cannot say why. It can only say where to look, and it is pointing at kindergarten and first grade, where the floor drops out fastest and the gaps open widest.
Independent review
Three AI systems reviewed this page twice, a few weeks apart, against the same question: are the observations and conclusions correct, using the linked data and the Act 20 requirements? The first round’s dissent (from Perplexity) prompted several of the corrections already folded into the sections above — treating the rising-bar reading as an inference, separating “below benchmark” from the statutory 25th-percentile threshold, and noting that compliance can’t be judged from aggregate rates. This second round checked the revised page. Gemini and Grok found no remaining issues, including specifically validating the two additions made after round one — Madison’s diagnostic-tool choice and the district-direct provenance of the 2025–26 figures. Perplexity found three new, narrower problems, all corrected below.
Gemini · round 2
“Yes, the observations and conclusions presented in the linked report are correct.”
Gemini confirmed the statutory framework point by point — the aimswebPlus mandate, the twice-yearly/three-times-yearly cadence, and Act 192’s one-year reduction that makes 2025–26 the first full-schedule year — and called out the screening-versus-diagnosis distinction specifically: “the observation that Madison utilizes FastBridge and Star for diagnostics, rather than aimswebPlus, aligns perfectly with the flexibilities granted to districts under the law.”
On the data, it re-verified the fall-to-spring changes by grade, the widening kindergarten White–Hispanic gap, and the IEP and English-learner subgroup trends against the tables. It described the seasonal-norming mechanism as “a highly accurate psychometric conclusion” and endorsed the page’s own caveat that proving it would take student-level matched scores — the inference is sound, not the same as proof.
Grok · round 2
“Correct based on the linked district data and Act 20 requirements, with appropriate caveats already noted in the report itself.”
Grok specifically confirmed the provenance framing added after round one: that the 2025–26 figures come directly from the district because DPI has not yet published a statewide report for that year, and that comparing 2024–25 to 2025–26 is invalid on both re-norming and schedule grounds. It re-checked the fall-to-spring figures, stable enrollment counts, and spring demographic cross-tabs against the tables and found them consistent.
It credited the page for not overclaiming Act 20 compliance, since compliance “turns on individual diagnostics/plans/progress monitoring” rather than aggregate rates, and for correctly distinguishing the district’s reported “at or above benchmark” metric from the statute’s exact below-25th-percentile threshold.
Perplexity · round 2, dissenting in part
“Mostly accurate… but several conclusions overreach what these cross-sectional benchmark percentages can establish.”
Perplexity confirmed the schedule, the arithmetic, and the subgroup gaps, then raised three new, more specific objections than round one.
- The age gradient isn’t necessarily an instructional finding. Calling it “the single most important pattern” implied a distinctive early-grade trend, when the same shape could come from differences in measures, norms, or administration across grades rather than anything happening in classrooms.
- “Not yet reading at the level the screener expects” needs a narrower referent. Below-benchmark means below the aimswebPlus cut score for that grade’s composite — not below Wisconsin’s official measure of grade-level reading, which for third grade is the separate Forward Exam.
- “The spring picture is the one that counts” misstates Act 20. A 5K–3 student is identified as at-risk the first time any window’s result falls below the 25th percentile — fall and winter results trigger a personal reading plan just as spring’s do. The law doesn’t treat spring as uniquely dispositive, even though it is the most complete snapshot available.
What changed as a result. The age-gradient finding now says “the largest pattern in the data” rather than the single most important one, with an explicit note that the data can’t distinguish an early-grade instructional story from a measurement one. The spring-focused finding is rewritten twice — here and in the closing section — to say a below-benchmark result is a screener cut score rather than a claim about grade-level reading, to link to the Forward Exam as Wisconsin’s actual grade-level measure, and to state plainly that Act 20 identification happens at any window, not spring exclusively. And a new sentence in the screener-description section flags that the benchmark composite differs by grade, so rates should be compared within a grade across time more confidently than across grades at a point in time.