What happened
A new preprint submitted on September 2, 2026, reports on “Math with Matt,” an intelligent tutoring system for seventh-grade algebra that added emotionally supportive, mindful feedback to ordinary cognitive tutoring. The authors found that the tutoring system as a whole reduced executive state math anxiety and improved math learning; the mindfulness version did not clearly outperform the cognitive-only version on every outcome, but students using it reached similar learning with less time and fewer hint requests, and rated the agent as more supportive. The paper is available from arXiv.
The study is timely because it fits a broader pattern in recent learning science: math anxiety is not just a feeling students bring into the room. It changes how they allocate attention, how much working memory remains for the problem, how quickly they avoid practice, and how accurately they judge whether they are ready for a test.
A 2026 review in the British Journal of Educational Psychology summarizes the state of the field: math anxiety is related to general anxiety and test anxiety, but it is still a specific problem; it is usually negatively associated with math performance; and the relationship is often reciprocal. Anxiety can hurt performance, while weak performance and repeated failure can feed later anxiety.
Why it matters before tests
The practical implication is that test preparation should not be designed as a last-minute stress simulation. High-pressure drilling may look rigorous, but for anxious learners it can consume the same working-memory resources they need for multi-step reasoning. A better target is productive familiarity: students should repeatedly experience retrieving, checking, explaining, and correcting math under low stakes before the real assessment.
Recent evidence also points to the importance of knowing which part of math anxiety is being triggered. A 2025 study of Brazilian first graders in Frontiers in Education found that higher math anxiety was associated with weaker performance in number production and comprehension, but not equally across every math skill. That finding is a warning against broad labels such as “bad at math.” A learner may be anxious around symbols, notation, timed recall, word problems, or evaluation, and each calls for a different practice design.
Confidence is another part of readiness. A classroom study of 12- to 13-year-old mathematics students, posted by Loughborough University, used six low-stakes formative assessments over two months. Students rated how sure they were on each item, then compared confidence with accuracy. Across the sequence, their calibration broadly improved, and students’ comments about the practice were strongly positive. This matters because anxious students can misread a hard practice moment as proof that they are unprepared, while overconfident students can stop studying too early.
The role of worked examples
Worked examples should not be treated as shortcuts. For anxious or low-prior-knowledge learners, they can lower unnecessary cognitive load so attention goes to the structure of the solution instead of panic about the next step. A 2026 tutoring-system study on interactive worked examples, available from arXiv, found that guided and “buggy” examples outperformed passive examples in a logic problem-solving tutor. Guided examples were especially useful for lower-prior-knowledge learners, while bug-fixing examples helped stronger learners explore and revise.
That pattern gives teachers and learners a practical sequence before tests: start with fully worked examples, move to partially completed examples, then ask learners to find and fix common errors, and only then shift to independent mixed practice. The goal is not to remove challenge. The goal is to make challenge interpretable.
Practical learning conclusion
- Use low-stakes retrieval early. Short quizzes, blank-page recalls, flash checks, and quick problem starts should happen before grades are attached. Retrieval should feel routine enough that the test is not the first time a student has had to produce an answer from memory.
- Pair retrieval with feedback quickly. Waiting too long to check answers lets anxiety fill the gap with guesswork about ability. Immediate or same-day correction turns errors into information.
- Use worked examples before independent problem sets. For newer or anxious learners, begin with a solved model, label the purpose of each step, and then fade support gradually.
- Calibrate confidence item by item. Ask students to mark whether they are unsure, somewhat sure, or very sure before checking. The most useful study list is not simply “what I got wrong,” but “what I got wrong while feeling sure” and “what I got right while guessing.”
- Separate practice pressure from test pressure. Timed work has a place, but it should come after accuracy, explanation, and confidence calibration have been built under low stakes.
The news is not that math anxiety can be solved by one app, one mindset prompt, or one pre-test pep talk. The stronger conclusion is more concrete: before tests, practice should reduce avoidable cognitive load, give students repeated low-stakes retrieval, and teach them to judge confidence from evidence rather than fear.