Did you know that a table can show how many pupils have reached a score without saying how many pupils got each score? A cumulative-frequency table looks like a list of totals, but each new total contains all the pupils in the earlier rows. That makes it useful for answering ‘how many up to this point?’ and easy to misread when the question is ‘how many in this group?’
Cumulative frequency is a running total of observations up to a stated boundary. To recover the frequency of a particular interval, subtract the preceding cumulative total from the current cumulative total. The first frequency is the first cumulative count, provided the first row starts from zero observations. Keep the interval boundaries and the population definition unchanged.
Start with the question, not the subtraction
Imagine a teacher collects how many minutes 20 pupils spend reading a story after school. The table is organised into non-overlapping time bands: 0 to under 10 minutes; 10 to under 20; 20 to under 30; and 30 to under 40. The running totals at the end of those bands are 4, 11, 17 and 20.
These are invented teaching figures, not measurements from a real class. Assume each pupil gives one non-negative time, every response falls below 40 minutes, and the boundaries ‘under’ exclude the upper endpoint. Therefore a time of exactly 10 minutes belongs in the second band, not the first. The final cumulative count of 20 tells us the sample size, not the size of the last band.
The complete worked example: unwrap the totals
- First band, 0–under 10 min: cumulative total 4. Frequency = 4 pupils.
- Second band, 10–under 20 min: cumulative total 11. Frequency = 11 − 4 = 7 pupils.
- Third band, 20–under 30 min: cumulative total 17. Frequency = 17 − 11 = 6 pupils.
- Fourth band, 30–under 40 min: cumulative total 20. Frequency = 20 − 17 = 3 pupils.
- Independent check: 4 + 7 + 6 + 3 = 20 pupils, matching the last cumulative total.
The largest band is 10–under 20 minutes, with seven pupils. It would be wrong to claim that 20 pupils read for 30–under 40 minutes. Twenty is the number accumulated across all bands. Nor can we find the exact average reading time from this grouped table: individual reading times are not available. Using band midpoints would produce an estimate, not a recovered exact mean.
Why subtracting adjacent rows works
A running total is built by addition. At the end of the second band, the 11 pupils consist of the four in the first band plus seven in the second. So 11 − 4 removes everyone already counted, leaving only those newly included. More generally, frequency in band i = cumulative total at band i − cumulative total at band i−1. We may take the preceding total as zero before the first band if the table begins at zero.
The operation also provides a powerful diagnostic. If a cumulative total falls from 17 to 15 as the boundary moves upward, the reported values cannot be a valid non-decreasing count of the same observations. Check the original table, categories and transcription; do not quietly replace the count with a number that ‘looks right’. A flat step, such as 17 followed by 17, is possible: it means zero new observations in that interval.
A quick learner diagnostic
Ask a pupil: ‘The cumulative numbers are 3, 9 and 12. How many pupils are in the second group, and how many are in the first two groups combined?’ The correct answers are six (9 − 3) and nine respectively. An answer of nine to both questions reveals that the pupil has mistaken a total so far for a group count. Ask the pupil to draw three boxes and add one box at a time before teaching an algorithm.
What can go wrong even when the arithmetic is right
- Overlapping interval labels: if the bands say 0–10 and 10–20 without clarifying where exactly 10 goes, pupils cannot allocate boundary values consistently.
- Unequal band widths: subtracting still recovers counts, but comparing frequency densities or drawing a histogram needs extra care; raw bar heights cannot be treated as interchangeable when class widths differ.
- Different populations: two totals from different classes, days or samples should not be subtracted as though they were successive counts of the same cohort.
- Cumulative percentages: subtracting two cumulative percentages yields a percentage-point share, not necessarily a pupil count until the sample size is known.
- Unrecorded observations: the final cumulative total counts only observations that actually enter the table. Do not silently assume that the class register equals the number of valid responses.
Progressive practice: from a model to independence
Stage 1 — guided reconstruction
Start with the three totals 2, 5 and 10 on a card. With a parent or teacher, shade a strip of two squares, extend it to five, then extend it to ten. Name the new squares at each step: 2, 3 and 5. Check 2 + 3 + 5 = 10. Physical grouping makes the subtraction visible.
Stage 2 — a small change to the table
Now use running totals 6, 6, 14 and 19. Without a worked solution, calculate 6, 0, 8 and 5. Explain why zero in the second band is valid. Ask what would be wrong if the third cumulative value were 4 instead of 14. The learner should say the total would decrease for a count of the same observations.
Stage 3 — new context, independent transfer
A librarian reports cumulative books borrowed by the end of four consecutive weeks as 9, 15, 22 and 30. Find how many were borrowed during each individual week: 9, 6, 7 and 8 books. Then check the sum is 30 books, assuming the record counts new borrowings cumulatively without removals or resets. Ask the pupil to state this assumption rather than merely reciting the subtraction.
How parents and teachers can observe progress
Give one unseen four-row table a few days later. A useful success measure is not speed alone: the learner identifies the running-total heading, labels the recovered values as counts within intervals, subtracts adjacent entries without a prompt, and verifies their sum against the last total. Track which of these checks still needs reminding; progress means the prompts can gradually disappear.
Teachers can include a second question asking what the table cannot reveal. Parents can ask a child to invent their own four-group dataset, build cumulative totals, hand it to another family member and see whether the original counts can be recovered. A learner who can build and reverse the table understands the mechanism, not just the worksheet.
Where the method stops
The subtraction method assumes that each successive cumulative row includes the previous observations and adds those in one new, non-overlapping interval. It will not retrieve each individual reading time from grouped data, establish why a pupil chose that time, or correct missing data automatically. If the table’s headings describe ‘more than’ counts instead of ‘less than’ counts, the order and interpretation must be examined before applying the same recipe.
Frequently asked questions
Do I always subtract from the row above?
For increasing, less-than cumulative totals over successive non-overlapping intervals, yes; the first interval starts from a previous total of zero. For descending ‘more than’ totals, differences may be taken in the reverse direction. Read the heading first.
Can a band contain zero pupils?
Yes. Two identical adjacent cumulative counts mean no additional observations were included between their boundaries. Zero is data here, not a missing entry.
Does the last cumulative number equal the entire class size?
Only if all class members contributed a valid observation and the table covers every observation. Otherwise it equals the number actually represented. Check what was counted.
Helpful routes through eduKate
For another data-quality check, see how repeated headers can inflate survey counts. For wider mathematical reasoning, explore how mathematics learning develops from number sense into explanation, the eduKate Education Hub, and the diagnosis-first tutorial approach.
For definitions of ordinary and cumulative frequency, consult OpenStax Introductory Statistics, section 1.3. The pupil counts and examples in this guide are illustrative.
A cumulative table is not hiding information on purpose. It is answering a different question. Once a learner knows how to move between a running total and a single-group frequency—and checks the boundaries and population—the same numbers become much easier to explain.