Learning Loss and Recovery: What the Data Says About Math and Science

Every couple of years, in schools across the country, a small group of students are pulled from their classes to take a test with results they’ll never see. Nothing goes home in a backpack. Their parents and teachers won’t see the results, and neither will their principal. Yet what those students do over the next hour or so becomes one of the most closely watched measures of how American students are learning.

The test is the National Assessment of Educational Progress, or NAEP, better known as the Nation’s Report Card. It has been testing students in grades four, eight, and twelve since 1969, with the National Center for Education Statistics running the assessment and the National Assessment Governing Board setting its policy.

The way students are selected is what makes the results trustworthy. NAEP doesn’t test everyone. It starts by randomly selecting schools, with larger schools more likely to be chosen, and it groups schools by location, demographics, and past achievement so the sample reflects the whole. Then, inside each selected school, roughly thirty students per grade per subject are chosen at random. Taking the test is voluntary for students. Each student answers only a fraction of the questions, for up to ninety minutes, and it’s only when thousands of students’ responses are combined that a full picture emerges. Results are reported for the nation, for states, and for some large urban districts, never for individual students or schools.

Because the design stays consistent from one year to the next, NAEP lets us follow student achievement across decades. And lately, what it shows is worth taking a look at.

A Long View of Math and Science

In 2024, fourth-grade math scores improved, but eighth-grade scores were flat compared to 2022. That sounds like a plateau until you remember that 2022 had brought a historic eight-point drop in eighth-grade math. Students stopped losing ground, but they haven’t yet won it back. By twelfth grade, the average math score sat below where it was in 2005 (NAEP2024).

Science tells a similar story. The 2024 eighth-grade assessment showed a four-point drop from 2019, the first time the average score had fallen since the current version of the test began in 2009. The decline showed up in physical science, life science, and Earth and space sciences alike. Only 31 percent of students reached NAEP Proficient, a level that signals solid command of challenging material and is a higher bar than “on grade level.” Another 38 percent scored below NAEP Basic (NAGB Science).

One detail in the science data stayed with me, and I think it will matter to anyone who writes or edits instructional materials. Students reported doing scientific inquiry activities less often in class. Science is a subject you learn by investigating, so fewer opportunities to ask questions and test ideas may be part of the explanation.

Not a Sudden Turn, but a Long Slide

It would be easy to read these numbers as a recent event. The researchers behind the Education Scorecard, a collaboration among Harvard, Stanford, and Dartmouth, noticed something surprising. They trace what they call a “learning recession” back to 2013, when student progress in math and reading stalled and then began to decline.

The same researchers remind us that it didn’t always go this way. From the early 1990s through 2013, math and reading skills improved dramatically, by more than two grade levels in math. Schools have done this before, which gives the current moment a different feel. The question isn’t whether improvement is possible, but where it’s happening now and why.

The picture is uneven. Scores for the lowest-performing students are at historic lows, and the Scorecard describes a U-shaped pattern in recent gains. The highest-income and lowest-income districts improved the most, while middle-income districts, which received little federal aid, improved the least (Harvard Graduate School of Education).

Where Students are Gaining Ground

The most encouraging stories share a common thread: someone made a deliberate, sustained decision to support students who were behind.

The Scorecard found that student achievement grew more in districts that invested more in academic interventions like tutoring and enrichment programs. Washington, DC, offers a good example. Between 2022 and 2024, it ranked first for gains in both math and reading, after its math results had been near the bottom of the list. A 2025 Stanford study found that DC students who received tutoring were more likely to attend school and showed greater academic growth. Louisiana is one of only two states performing above its 2019 level in math. And across the country, researchers identified 108 districts improving substantially faster than demographically similar neighbors, which suggests that a district’s circumstances don’t have to decide its outcome. (DCPS).

Attendance deserves a mention too. In 2024–2025, 23 percent of students were chronically absent, up from 15 percent in earlier years. The researchers estimate that recovery would have been meaningfully stronger at every income level if attendance had held steady. Even the best lesson can’t reach an empty seat (Harvard Graduate School of Education).

There’s a caution in the story as well. Much of the improvement in the highest-poverty districts appears to have been driven by federal relief funding, and without it, the average high-poverty district would have stayed where it was in 2022. That funding is gone now. Whatever comes next will depend on what schools and states decide to sustain.

What It Means for the Work We Do

At Contentra, we spend our days developing the materials students and teachers use, so we can’t help asking what this data means for the page and the screen. Do our math resources support tutoring and targeted help, or do they assume one pace for everyone? Do our science resources give students real chances to investigate and explain? And do they work for learners across the whole range of readiness, including those furthest behind?

Our team helps clients with math and science content development, editorial support, and standards alignment. Whether a project is brand new or a refresh of existing content, we begin with what students should be able to do and work backward from there.

The numbers show a hard stretch, but they also show places where it’s turning around. The work now is to understand what those places are doing and help it spread.

Stay Connected with Contentra

Contentra shares expert perspectives on curriculum development, educational publishing, instructional design, project management, and academic alignment.

Join education leaders, publishers, and curriculum specialists who subscribe to receive new insights as soon as they’re published.