Your child's assignments come back with high marks all quarter. Homework is turned in on time. The teacher's comments are positive. Then the test comes back and the score is a full letter grade lower — sometimes two.
Most parents' first instinct is "test anxiety." Sometimes that's real. More often, it isn't — and treating the gap like a nerves problem when it's actually a skill problem sends you down the wrong road for weeks. This post lays out the three real causes of the pattern, how to tell which one is happening, and what closes the gap.
The three real causes
1. The homework was completed with help
This is the most common one and the most uncomfortable to say out loud. "Help" includes: a parent explaining every problem, a tutor who walks through it step by step, a friend group text where one kid does it and the rest copy, and increasingly, an AI chatbot that produces the answer on request. The student turns in correct work. The student learned very little from producing it. The test is the first place the shortcut can't reach.
The "help" doesn't have to be dramatic. Even a parent sitting next to a kid and saying "try 3 there" is doing a lot of the thinking. The kid nods, writes 3, and moves on. Tomorrow morning they can't explain why 3.
2. Homework was done with the book open
The second-most-common cause. The homework problems mirror a worked example on the previous page, so the student copies the pattern without understanding the underlying method. This is called recognition, and it feels a lot like learning — the answers come out right, so the student thinks they've got it.
The test hits them with the same math in a different wrapper, and recognition fails. They can't pattern-match to a worked example that isn't in front of them, and they don't have the underlying rule that would let them build the solution from scratch. That's called recall, and it's what tests measure.
3. The test format is genuinely harder than the homework format
This one is real, and it's smaller than parents usually think. Timed conditions, no partial credit, more mixed-topic questions in one sitting — the test asks for the same knowledge under harder access conditions. If a student has genuinely learned the material, format shifts the score down by five or ten points, not thirty. A larger gap almost always points to one of the first two causes underneath.
Test anxiety is real, and worth taking seriously when it appears. But it is the third suspect, not the first.
How to figure out which cause is yours
The "same problem, different context" check
Pick a problem from a recent homework assignment that came back correct. Copy it out, change the numbers slightly, and hand it to them tomorrow morning at breakfast — without the textbook, without the notebook, without their phone.
Three outcomes tell you three different things:
- They solve it easily. The homework score reflected real learning. The test-day issue is probably format or nerves.
- They start okay and stall halfway through. They half-learned the method — the classic recognition-not-recall pattern. Cause #2.
- They can't start. The original assignment wasn't produced by them under their own power. Cause #1.
This one check takes ten minutes and tells you more than a parent-teacher conference will.
The "cover the page" check
For a math kid: sit next to them the next time they do homework. Ask them to attempt each problem with the textbook and notes closed. When they get stuck, then open the book. If they can only start problems with the book open, they're pattern-matching, not solving.
Look at the actual wrong answers on the test
The test paper is the most honest document in the whole system. Group the wrong answers into three piles:
- Blank or barely started. They didn't know how to begin. That's a setup problem — they don't understand what the problem is asking them to do.
- Right approach, arithmetic wrong. They knew the method but couldn't execute cleanly — often because the method was still shaky and eating up mental bandwidth.
- Wrong approach. They confidently used a method that doesn't apply to that problem — often the method from the most recent worked example, applied by pattern-matching.
The pile that's biggest tells you the story.
What doesn't work
More homework. If the homework is being finished with help or with the book open, adding more of it just adds more hours of the same broken practice. Volume isn't the issue; the mode is.
The-night-before test prep. Cramming reviews facts. It doesn't build the fluency that timed test conditions require. If the material isn't already close to automatic by test week, one late night doesn't get it there.
Blaming the test or the teacher. Almost always emotionally satisfying, almost never useful. The tests aren't going away and the format won't change. What can change is how your student prepares.
What actually closes the gap
Practice under test-like conditions, not homework-like conditions
The single biggest change that helps: for the last few days before a test, have them do at least one full problem set with no notes, no textbook, and no help. If they get stuck, they stay stuck and move on to the next problem. Then, and only then, they check their work. This reveals which problems they can actually do on their own.
The explain-back method
After they finish a problem, ask them to explain out loud why they did each step, as if teaching a younger sibling. Anywhere they hand-wave ("you just do this") is a gap. Explaining is much harder than solving, and it exposes the difference between recognition and understanding fast.
Spaced practice
A student who can solve linear equations on Tuesday might have forgotten how by the following Friday. Coming back to old material every few days, briefly, keeps it available. Cramming it once, then not seeing it for two weeks, is why so much of what students "learn" evaporates before the unit test.
The uncomfortable version of this
If the "same problem, different context" check goes badly — and they can barely start the problem — you're looking at a cause that isn't really about tests. Two other posts on this site cover the two versions of what's underneath: signs your kid is struggling in math (if they've been quietly falling behind on the concepts) and how to tell if your kid is using ChatGPT for homework (if the homework is being produced by something that isn't them).
Either way, the fix is the same: change the mode of practice so the student actually has to do the work themselves, and back-fill the specific gaps that emerge when they try.
An alternative worth trying
We built Lernos because this exact pattern — homework grades disconnected from test scores — was happening in our family. It's an AI tutor that never gives a student the answer. It asks questions until they work it out, which forces recall instead of recognition. Because the tutor never shows the completed solution, the student has to produce each step themselves rather than copy one from a worked example. Parents get a dashboard showing which topics have been genuinely worked through and which are still shaky.
If your child's homework and test scores don't line up and you want to see the alternative, this is the page to read next: See how it works →