Why your school probably shouldn't bother with a formal math attitude survey
We ran them for about four years starting around 2018. District mandate. Every third-grade teacher in our building was expected to administer a validated attitude instrument mid-year and again at the end of the term, then submit the data to the curriculum coordinator for some kind of trend analysis. I'll be honest about what that looked like and what it produced, because most of the people writing about these surveys online haven't actually had to wrestle with seven-year-olds filling out Likert-scale questionnaires about their own anxiety. A math attitude survey for elementary students is essentially a short psychometric instrument designed to measure how young children feel about mathematics—things like enjoyment, confidence, perceived difficulty, avoidance behavior, and whether they see math as something "for smart people" or something anyone can learn. The most commonly referenced tools in K-5 settings are the Mathematics Attitudes Scale (MAS), the Elementary Mathematics Attitude Survey, and various teacher-made inventories adapted from the work of researchers like Baker, LeBlanc, and the National Council of Teachers of Mathematics framework documents. They typically run between 15 and 30 items, use simple smiley-face or agree-disagree formats appropriate for early readers, and aim to produce a composite score that can be tracked across a school year.
What a Math Attitude Survey For Elementary Students Actually Measures (and Doesn't)
The surveys claim to measure attitude. What they actually measure is a child's ability to read the questions, understand what a "sometimes" means compared to "usually," and respond in a way they think the adult wants rather than in a way that reflects their actual feelings. That last point matters more than most administrators realize. Third graders are desperately eager to please. If you ask "Do you like math?" with a straight face, roughly 85 percent of them will say yes regardless of their actual relationship with the subject. The format has to account for this, which is why better instruments use indirect questioning—situational stems like "When your teacher says we're doing math, I feel excited" rather than direct self-evaluation. I found that the items about math anxiety—things like "My hands get sweaty when we have a math test"—actually pulled more honest data than the enjoyment items. Seventh and eighth graders can't pretend they don't have test anxiety, but five-year-olds genuinely don't yet have the interoceptive awareness to connect a physical sensation with an emotional label. You have to read those responses literally. A child checking "sometimes my hands get sweaty" during division practice isn't telling you they're anxious about math as a subject. They're telling you that the specific activity of long division makes them uncomfortable, and that's a curriculum problem, not an attitude problem.
How to administer one without wasting two days of instructional time
Read the items aloud as a group. Don't hand out sheets and expect independent completion until at least fourth grade, and even then you'll get inconsistent results from kids who read below grade level. We did it in small groups of six to eight students in the library, which removed the social pressure of doing it at their desks with peers watching. It took about twenty minutes per session, and you need to schedule two sessions per testing window—one in October and one in April—to get any useful trajectory data. Use a scantron or bubble-sheet format if you're processing more than twenty-five responses by hand. We switched to a simple paper-and-pencil method with a scoring key and spent approximately forty-five minutes per class manually entering scores into a spreadsheet. That added up to roughly six hours of teacher time per grade level across both administrations, and the data quality didn't improve in any measurable way compared to the digital option. Scanning sheets with a standard classroom scanner and using free tools like Google Forms or Paperless Survey cut that down to about fifteen minutes total. The biggest procedural mistake I saw was letting teachers administer the survey during math instruction itself. Kids absorb the context. If you hand out an attitude survey right after a timed multiplication fluency drill where half the class struggled, you're not measuring their general attitude toward math—you're measuring their emotional state at that exact moment. We moved all administrations to occur on designated "brain break" days, completely separate from any math lesson, and the reliability of the data improved noticeably.
Get the Full Details
Scoring and what the numbers mean
Most instruments use a reverse-scoring model where agreement with negative items (e.g., "Math is too hard for me") subtracts from the total, and agreement with positive items adds to it. The range typically runs from around 15 to 75 for a 15-item survey, with a mid-range score near 40 to 45 indicating neutral attitude. Scores below 35 flag potential math avoidance risk. Scores above 55 suggest strong positive orientation but can also indicate a ceiling effect where high-achieving kids just agree with everything regardless of content. Here's the part nobody puts in the administrator's guide: elementary attitude scores have surprisingly low test-retest reliability over short periods. A child scoring 42 in October might score 38 in April without any meaningful change in their actual relationship with math. The standard error of measurement on these instruments is large relative to the score range, which means single-point administrations are essentially noise. You need at least two data points per student per year, and ideally three, to distinguish a real trend from normal variation. A single administration is not defensible for any individual student decision-making, and I've seen it used that way more than once.
A specific problem I ran into and how I fixed it
About halfway through our second year of running the surveys, I noticed that our English language learner population was consistently scoring in the lowest quartile across every item category, including the ones about math enjoyment that shouldn't be language-dependent. The "I like math" items were being misread as "Math is something other people like" rather than a genuine self-report. Their scores looked terrible but when I talked to the kids individually, almost all of them said they liked math class. The survey was failing ELL students because the response anchors—"always," "usually," "sometimes," "never"—carry different interpretive weight for bilingual children who are still calibrating those adverbial distinctions in English. The workaround was straightforward but required an adjustment to the protocol. We added a brief training session where we reviewed each response option with visual supports before administering the survey, using real examples. "Always means every single time. Usually means most of the time but not all the time. Sometimes means a few times. Never means not even one time." We practiced with non-math questions first so they understood the format. This added about eight minutes to the administration but brought ELL scores into a range that actually reflected their attitudes. Without that step, you're collecting data that systematically misrepresents your multilingual learners, which is worse than not collecting it at all.
Where to actually get a survey you can use
The most accessible free options are the adapted versions from the Institute for Learning & Brain Sciences at the University of Washington, which are openly available and specifically designed for early elementary. The National Council of Teachers of Mathematics also publishes a free resource packet that includes age-appropriate attitude measures. For something more comprehensive, the Woodcock-Johnson diagnostic batteries include an attitudes and beliefs scale, though that requires certification to administer and interpret. If you need something ready-to-use tomorrow with minimal setup, the ERIC database has several open-access instruments filed under DES numbers that you can download and print—search for "elementary mathematics attitudes inventory" filtered for documents from 2015 onward. Be careful with surveys you find on commercial teacher-resource websites. Many of them are unvalidated, meaning nobody has established whether they actually measure what they claim to measure. A survey with 500 five-star reviews from other teachers tells you nothing about psychometric soundness. It tells you the formatting is nice. Validated instruments will have published reliability coefficients (Cronbach's alpha typically needs to be above 0.70 for group-level use, above 0.85 for individual decisions) and documented validity studies. If the author doesn't cite those numbers somewhere, treat the tool as a conversation starter, not a diagnostic instrument.
The limitations you need to accept upfront
These surveys cannot tell you why a child has a negative attitude toward math. They can tell you that the attitude exists and roughly how strong it is, but the instrument doesn't include open-ended qualitative components, and even if it did, a seven-year-old's explanation—"because it's boring"—isn't useful diagnostic data. You need structured classroom observation and student interviews to understand the root cause. The survey is a screening tool, not an assessment tool. Treating it as anything more than that is where schools tend to make expensive mistakes. There's also the reactivity problem. Once you tell a class they're going to take a survey about how they feel about math, every subsequent math lesson gets filtered through that awareness. Kids start performing their attitudes rather than experiencing them authentically. In our experience, this effect lasted about two to three weeks after administration before attitudes returned to baseline. That window overlaps with a typical unit of instruction, so if you're using the data to evaluate teaching effectiveness or curriculum quality, factor in that contamination period. The other hard limitation is that attitude data from elementary students correlates weakly with actual math performance. The research literature consistently shows a correlation coefficient in the 0.20 to 0.35 range, which means attitude explains maybe four to twelve percent of the variance in math achievement. The rest is prior knowledge, executive function, instructional quality, home support, and a dozen other factors. Schools that treat attitude surveys as a proxy for predicting which students will struggle mathematically will miss the majority of those students. Use the survey to identify kids who need encouragement and relationship-building, not to identify kids who need academic intervention. Those are two different populations with overlapping but distinct membership.
If you want a more predictive tool for math risk, use a computational fluency assessment or a curriculum-based measurement probe. Those give you actionable data about current skill level. The attitude survey gives you data about how the child feels about the subject, which is valuable for a different set of decisions—parent conversations, program evaluations, and understanding whether your school culture is creating unnecessary math dread. Both matter. They just matter for different reasons.
