Field notes · Attention

Careless mistakes in schoolwork: what 102 worksheet answers showed about attention

One 4th grade student, 9 worksheets, two weeks. A score alone would say the student struggles with some skills. How the work was done says something else.

The short answer

For this student, the mistakes looked more like lapses in attention than gaps in knowledge. 29 of 33 wrong answers were on skills the student had already answered correctly, and accuracy fell from 77% at the start of a sheet to 51% in the middle, then came back to 76% at the end.

One child, two weeks, 102 answers: a pattern to act on, not a diagnosis.

Answers right, by part of the worksheetOne student, 9 worksheets, 102 answers, November 5 to 17, 2026. Each part is about a third of the answers, so each bar rests on a few dozen.
Start
77%
Middle
51%
End
76%
See the numbers
Part of the sheetAnswered right
Start77%
Middle51%
End76%

Data: the app's answer log, read with Harvv

For this student, focus seems to fade in the middle of a sheet, then come back near the end. Same student, same worksheets.

I run Harvv, which records how people use a web page: pauses, tab switches, hovers and clicks, with no names and no typed text. I put it on a learning app a 4th grade student uses for worksheets and read it next to the app's own answer log.

Are careless mistakes a sign of not knowing the material?

Not here. 29 of 33 misses were on skills the student had already answered correctly. That looks like a slip more than a gap: the student had done it before, and didn't do it every time.

The harder questions point the same way, on a small count. The student got 78% right on questions above grade level and 64% right on questions at or below it.

Answers right, by how hard the question wasSame student and worksheets. The harder group is 27 answers, so it rests on a small count.
Above grade level21 of 27
78%
At or below grade level48 of 75
64%
See the numbers
Question levelRightAskedShare right
Above grade level212778%
At or below grade level487564%

Data: the app's answer log, read with Harvv

If this were about ability, the harder questions would be where the slips are. The student did at least as well on them, so easy questions may be where attention slips.

Do careless mistakes come from rushing?

Not at the start of the question. On reading questions the student missed, 42 seconds passed before the first tap, against 26 seconds on the ones answered right. After that, the choice took about 2 seconds either way.

Time before the first tap on a questionReading questions, same student. After the first tap, the choice took about 2 seconds either way.
Questions missed
42 seconds
Questions answered right
26 seconds
See the numbers
OutcomeBefore the first tap
Questions missed42 seconds
Questions answered right26 seconds

Data: the app's answer log, read with Harvv

On reading questions, the wrong answers were slow to start. Right or wrong, the choice was quick. A slow start could also mean the question was hard for the student. The pauses and tab switches are why I read it as drifting.

7 of 15 visits to the app had a long pause, with a median of 66 seconds, and in 5 of 15 the student left the tab and came back. The app asks "Still there?" after 90 seconds with no activity, and it asked 15 times in 9 worksheets. A pause is only time with no input, so some of it could be working a problem out on paper. Leaving the tab is harder to read that way.

Why didn't the student ask for help?

The app has two help buttons, "I don't get it" and "Show me how". At times the pointer sat over both without clicking, and 30 of 33 misses came with no request for help. It looks like the student thinks about asking, then guesses instead. Maybe being stuck feels like failing.

It doesn't look like wanting to stop, either. The app offered 9 breaks and the student turned down 7 of them.

How can you check this in your own student's work?

None of this needs a tracking tool to start. Three counts on a stack of finished worksheets show whether the same pattern is there.

  1. Mark where each miss sits. First third, middle or last third of the sheet. Misses that bunch in one part may point at timing, or at the kind of question that sits there.
  2. Mark whether the skill was right before. A miss on something answered correctly last week is a different problem from a miss on something new.
  3. Compare hard and easy. If the harder questions go as well as the easy ones, more practice on the easy ones may not be the fix.

What paper can't show is the time: how long before the student starts, and where they stop. That part came from the app's log and from Harvv.

What are we changing to reduce careless mistakes?

Three changes to the worksheets, each tied to one of the patterns above. They are being tried now and I don't have results yet, so read them as a plan for one student and not as advice. If the middle of the sheet doesn't improve, that counts against the attention reading.

  1. Lead with the hard stuff. More challenge questions mixed in, and the hardest work moved into the first third, before the mid-sheet dip.
  2. Break the middle. Shorter focus blocks, so the dip lands on a pause instead of a question.
  3. Make asking normal. After a miss, the app shows why the pick was wrong and gives one more like it. Help comes to the student, so nobody has to admit to being stuck.

How this was measured

How this was written. The numbers come from Harvv's own data. I drafted this post with an AI assistant, checked every number against the query that produced it, and edited the final text.

See where people pause, leave and come back on your own pages.

Install Harvv on a site or web app and ask it where visitors get stuck. One script, 72 event types across 10 categories, no recordings or form values. Free for 3 sites, no card.