Did you know? A pupil can get four questions wrong but record six mistakes. That is possible when some wrong answers have more than one contributing cause. The error log is doing two jobs at once: counting unsuccessful questions and identifying what needs teaching.
A multi-cause learning error log is a record that gives each question or task one stable ID, records its outcome, and allows several clearly evidenced error tags on the same response. Keep the number of distinct wrong questions separate from the number of tagged problems. Then use the evidence to choose what to practise next.
This small distinction changes the conversation at home. Instead of ‘You made six mistakes again,’ a parent or teacher can ask, ‘Which thinking steps produced the four wrong answers, and what would help you correct them?’
The educational question is not only how many answers were wrong
Traditional correction can become a row of red crosses followed by a row of rewritten answers. That sometimes fixes the page without fixing the learner’s decision. An error log is useful when it helps identify a specific incorrect move — for example, copying a negative number without its sign, choosing the wrong unit, or assuming a condition that the question never supplied.
But a single answer may contain several such moves. When the record uses one row for every error tag, the pupil can accidentally count one question as two or three separate wrong questions. The resulting statistics become confusing, and the next practice plan may focus on the wrong priority.
A fully worked example: eight invented practice questions
Imagine a fictional learner, Mira, completes eight short mathematics and science reasoning tasks. She answers Q1, Q3, Q6 and Q8 correctly. Her Q2, Q4, Q5 and Q7 need correction. This example is entirely illustrative, not an actual student’s performance.
- Q2: the pupil copies −3 as 3, then treats a subtraction as addition. Tag: transcription and sign operation.
- Q4: a distance calculation is numerically correct but the answer is labelled cm instead of m. Tag: unit selection.
- Q5: the equation is set up with the wrong unknown quantity, and a fraction is later simplified incorrectly. Tag: modelling and fraction arithmetic.
- Q7: a graph label is read as a measurement rather than a category name. Tag: graph interpretation.
Count the unique incorrect question IDs: Q2, Q4, Q5 and Q7. That is 4 incorrect questions out of 8. The correct-question rate for this one practice set is 4 ÷ 8 × 100 = 50%. This rate tells you about the questions, not how many causes might explain the errors.
Next count the tags: Q2 contributes 2, Q4 contributes 1, Q5 contributes 2 and Q7 contributes 1. The log contains 6 tagged error instances. The number 6 is not ‘six wrong questions’. It is the number of labels provisionally attached to four wrong responses.
A reliable summary states both: ‘Four of eight responses were incorrect; review identified six error tags across those four responses.’ That sentence is longer than ‘six mistakes’ but far more useful.
Why the denominator matters
If the learner has answered eight questions, dividing six tags by eight questions gives 0.75 tags per question. That is an average tag count, not an error rate of 75% and certainly not a test score. Tags can overlap and depend on how finely someone defines categories. Use meaningful labels rather than creating a dramatic percentage.
One diagnostic tag is not necessarily a separate cause. If a child’s model of a problem is wrong, later arithmetic may also go wrong simply because it was applied to the wrong expression. Label what you can observe, distinguish primary from downstream errors, and test the explanation before assigning a practice prescription.
Build an error log that resists double-counting
A good log can be a page of handwriting or a simple spreadsheet. The structure matters much more than the software.
- One stable question ID: the question appears once in the question-level results, even when several issues are found.
- Outcome: correct, incorrect or not attempted, with explicit local scoring rules.
- Evidence: the exact line, diagram, statement or method showing the problem.
- Error tags: one or more provisional labels, such as units, reading, method, sign or representation.
- First causal move: the earliest unsupported decision that needs checking.
- Next practice: one short task that tests whether the underlying idea has changed.
- Recheck date and result: evidence of whether the learner can apply the skill again without a prompt.
When a learner writes ‘careless mistake’ for everything, the log has stopped being diagnostic. ‘I copied −3 as +3 on the second line’ is more helpful because it names an observable event and a potential action. Not every error is carelessness: a wrong sign can also expose uncertain number-line reasoning.
Keep the original response. A correction written after receiving the solution is not proof of independent mastery. Distinguish ‘fixed with explanation’ from ‘solved correctly without help on a fresh question’. These are both useful states but not interchangeable.
Worked correction: distinguish symptom and root cause
Return to Q2. The intended expression is 8 − (−3), whose value is 11. Mira initially transcribes it as 8 − 3 and obtains 5. If the copy was wrong, it is premature to claim the pupil does not understand subtraction of negative numbers. First ask her to explain and solve the unaltered expression while looking at it.
If she correctly says subtracting a negative increases the value and calculates 11, the immediate target may be a copying and checking routine. If she still says 5 while reading the original accurately, there is evidence for a conceptual sign-operation gap. If both happen, two different short practices may be justified. The point is to test, not diagnose from a label alone.
For Q5, suppose the learner writes the equation for three equal packets plus 2 loose items total 17 as 3(x + 2) = 17. The correct model, if x is the number in one packet, is 3x + 2 = 17, so x = 5 items per packet. The initial misplacement of brackets is a modelling error. A later fraction mistake on the incorrect equation cannot by itself tell us whether fractions would have been a problem under a correct model. Offer a separate fraction item to find out.
Progressive practice: correction becomes independent thinking
Stage 1: guided diagnosis
Show two wrong examples: 6 + (−4) = 10 and a length of 5 m labelled 5 cm. Ask which observable decision needs attention. Guide the pupil to identify the sign handling in the first and unit choice in the second, rather than simply saying both are ‘maths mistakes’.
Stage 2: count questions and tags separately
Give five tasks. Three are wrong: one has two tags, two each have one. Ask for the number of incorrect tasks (3) and total tags (4). Then ask why 4 ÷ 5 is not automatically an 80% wrong-answer rate. The learner should say that a task may carry more than one tag.
Stage 3: use new evidence to revise a diagnosis
Initially label Q2 as transcription and sign operation. After a fresh sign question is answered correctly with a written explanation, ask whether both diagnoses remain equally plausible. The learner should revise the log rather than defend an earlier guess. Good investigation is allowed to change its mind.
Stage 4: design a transfer test
Choose the most consequential verified gap, teach the underlying idea, then return a week later with two unfamiliar problems that use the same principle in a different context. Log the unassisted method, not only the answer. For sign handling, use a temperature change and an algebraic expression; for model setup, vary the grouping language while preserving the relationship.
A sensible weekly review, not an endless spreadsheet
One weekly 10-minute review can be enough for a short practice programme: choose two recurring tags, inspect actual worked lines, practise each with feedback, and test each independently later. A log with hundreds of labels but no follow-up has become bookkeeping rather than learning.
The Education Endowment Foundation’s feedback guidance emphasises actionable feedback, attention to misconceptions and opportunities for pupils to use the feedback. An error log supports these principles only when it leads to a clear next learning action; it is not a magic improvement tool by itself.
What parents, pupils and teachers can do
- Parents: ask for one specific wrong step and the next helpful practice; avoid turning tag totals into labels about a child’s ability.
- Pupils: retain original working and check which corrections you can reproduce from memory on a different question.
- Teachers and tutors: agree on a small, consistent tagging vocabulary and discuss uncertain classifications instead of pretending the labels are perfectly objective.
- Everyone: use private, respectful records and assess improvement over comparable tasks, not raw tag totals taken from different difficulty levels.
For a neighbouring assessment question, read how to handle a non-applicable rubric criterion fairly. For a technical example of a specific incorrect step, see how to debug a Python slice boundary. Both show why checking the rule matters more than attaching a vague error label.
What observable progress looks like
A learner improves when they can spot the earliest unsupported move, propose a corrected explanation, avoid double-counting questions, and complete a related unfamiliar task without hints. Another useful signal is fewer repeated conceptual errors across comparable practice sets. Falling tag counts alone are not enough because the marker may be using fewer labels or the tasks may have become easier.
Frequently asked questions and limitations
Should each incorrect question receive exactly one tag?
No. A response can show several independently evidenced weaknesses. Allow multiple tags, but count unique question IDs separately and avoid treating every downstream slip as an unrelated cause.
Do six tags mean the student made six separate errors?
They mean the log has six labels or instances. Depending on how they were assigned, some may describe the same causal chain. Report what the tags mean and how they were derived.
Is a blank answer incorrect or not attempted?
Record it according to the assessment’s scoring rules, but keep ‘not attempted’ visible in the diagnostic log when appropriate. Lack of an answer does not by itself prove a particular misconception.
How can we know an intervention worked?
Test the same underlying concept with fresh, comparable tasks after a suitable interval and without the original scaffold. Even then, the result is local evidence of improvement, not a guarantee about future examinations.
The deeper aim of education
An error is useful evidence when it guides a better explanation and a better next attempt. That principle connects with the eduKate Education Hub and the small-group diagnostic philosophy of our Secondary 1 Mathematics reference.
The aim is not to collect the greatest number of mistakes. It is to turn a child’s actual thinking into a clearer lesson, then give that child the chance to show the idea works independently.