Two school groups can have identical-looking box plots even though one group’s marks stretch across the middle while the other’s marks split into low and high clusters. Is one of the graphs wrong? Not necessarily. A box plot compresses the data into a few summary positions. The shape between those positions can disappear.
A box-and-whisker plot typically communicates the median, lower and upper quartiles, and a definition of whiskers and possibly outliers. It is valuable for comparing location and spread, but it does not show every individual value or reliably reveal how many peaks or gaps a distribution contains. A dot plot, histogram or the actual observations may tell a different part of the story.
- Read what is present: median, quartiles, spread, whiskers and any marked outliers.
- Ask what is absent: individual values, local clusters, empty score ranges and often the sample count.
- Check the convention: quartiles and whiskers can be computed differently by different tools.
- Use a second view: a dot plot or appropriately chosen histogram can reveal gaps.
- Do not infer: a single summary shape proves every student has similar results.
A hand-checked example: twelve imaginary pupils in each group
The following are fictional numbers for learning, not actual children, school performance or tuition outcomes. Both groups have 12 results on a 0–100 point practice scale. Each list is already sorted.
Group A: 10, 20, 20, 30, 40, 45, 55, 60, 70, 80, 80, 90.
Group B: 10, 20, 20, 30, 30, 30, 70, 70, 70, 80, 80, 90.
Group A has scores at 40, 45, 55 and 60. Group B has no score anywhere between 30 and 70. A reader looking at the sorted lists can see that immediately. The remarkable part is that their simple box plots can still be identical.
Calculate the five-number summaries yourself
Step 1: minimum and maximum
For each group, the minimum is 10 and the maximum is 90. Both groups cover the same numerical range, 90 − 10 = 80 points. A wide range tells us the extremes differ, but not where the middle students are gathered.
Step 2: median
Each list contains 12 values, so its median is the mean of the sixth and seventh values. For Group A, (45 + 55) ÷ 2 = 50. For Group B, (30 + 70) ÷ 2 = 50. The common median does not mean there is an actual score of 50: neither group contains it.
Step 3: lower and upper quartiles
For this exercise use the median-of-halves convention: split the 12 sorted values into the lower six and upper six, then find the median of each half. In Group A, the lower-half middle values are 20 and 30, giving Q1 = 25. The upper-half middle values are 70 and 80, giving Q3 = 75. In Group B, those same positions are 20 and 30, and 70 and 80; thus Q1 is also 25 and Q3 also 75.
The interquartile range for both is Q3 − Q1 = 75 − 25 = 50 points. With the common 1.5 × IQR rule, the inner fences are 25 − 75 = −50 and 75 + 75 = 150. All observed scores lie inside, so a conventional Tukey-style box plot with whiskers to extreme non-outlier values would extend from 10 to 90 for both.
Their summaries match exactly: minimum 10, Q1 25, median 50, Q3 75, maximum 90. Under this stated convention their box-and-whisker marks can coincide even though the individual score patterns differ.
Did You Know? Even the mean is identical here
Add Group A’s twelve values: they total 600 points, giving a mean of 50. Group B also totals 600, for the same mean of 50. We have deliberately constructed an example in which the mean, median, quartiles, range and whiskers all agree, yet the pattern in the middle is very different.
Group B clusters its lower half at or below 30 and its upper half at or above 70, leaving an empty central interval. Group A includes values in that interval. The fact that a mean and box plot agree is not a guarantee that the distributions have the same detailed shape.
Why a box plot loses this information
A box plot records the positions of selected order statistics, not a complete map of the observations. Twelve students are represented by the values at just a few boundaries. Values can move inside many intervals without changing those boundaries. There is no mechanism in a basic box plot to draw a separate symbol for each interior observation.
A dot plot would show one dot per pupil, with repeated values stacked. Group B’s empty central span would then become evident. A histogram might show the split too, though the effect depends on the bin widths and boundaries. A narrow cluster can disappear into an overly broad bin; this is another reason to inspect design choices.
The US National Institute of Standards and Technology discusses box-plot construction using quartiles and fences in its Engineering Statistics Handbook. The method shows spread and potential outliers, not a full reconstruction of every raw value.
A quick diagnostic to find the real misconception
Show a learner only the five-number summaries: 10, 25, 50, 75, 90 for both groups. Ask, ‘Must both sets of twelve scores be identical?’ The correct answer is no. If the pupil says yes, they have confused a summary with the full dataset. Now reveal the two sorted lists and ask them to point to the empty interval in Group B.
A second prompt: ‘Can the five-number summary tell us that no one scored exactly 50?’ No. It tells us the median is 50, which can be an average of two middle scores. It does not identify each pupil’s mark.
Progressive practice: from five values to a real explanation
Stage 1 — locate without drawing
Use the simpler five values 1, 2, 3, 4, 5. Ask students to identify the median, count observations and describe what a single number like 3 omits. This teaches that a summary is selective before introducing quartiles.
Stage 2 — calculate our twelve-score example
Students work in pairs to derive the five-number summary for Group A and Group B independently. Each must label the chosen quartile convention and show which sorted positions were used. Exchange sheets and check all five figures.
Stage 3 — build two visualisations
Draw matching box plots, then draw one dot per actual score on the same 0–100 horizontal scale. Compare what each view communicates. Ask students to find the range 31–69 in Group B and explain why the dots, but not the box, make the gap obvious.
Stage 4 — transfer to a new context
Imagine two classes recording reading durations rather than test marks. Both have the same median and quartiles. Ask for a responsible reporting sentence: ‘The middle 50% spans the same interval, but more detail is needed to know whether the reading durations form one cluster or several.’ This is a more defensible conclusion than ‘The classes are identical.’
Parents and teachers: what the chart can and cannot guide
- Students: check the raw values or dot plot before making claims about peaks, gaps and subgroups.
- Teachers: report group sizes and the quartile convention; do not explain differences in ability solely from one box plot.
- Parents: ask what data a chart summarises and whether a few individuals or an invisible cluster need different support.
- Project writers: pair a box plot with an appropriate detailed visualisation where distribution shape is central to the question.
Important limits: quartiles, outliers and causes
Spreadsheet packages and textbooks do not always compute quartiles by the same definition. Some interpolate between positions rather than splitting the data into two halves. Two groups that match under one convention may not match under another, so describe the method. Whiskers can also mean the extreme observed values or the most extreme non-outliers, depending on the plotting design.
A clustered plot alone does not tell you why clusters appear. Different revision habits, question types, attendance or marking practices might be hypotheses, not established explanations. Do not attribute causes to students without evidence. Also protect student privacy when sharing individual-level dot plots.
Frequently asked questions
If the box plots match, can the class averages differ?
Yes. Our constructed pair also has the same mean, but matching five-number summaries do not generally force equal means. The mean uses every value and can change as interior observations change.
Is a histogram always better than a box plot?
No. A box plot is compact and useful for comparing many groups. A histogram or dot plot is useful when shape or a gap matters. Select the visualisation that answers the actual question, sometimes using both.
Could the median be a score no student received?
Yes. With an even number of observations, the median is commonly calculated as the average of the two central values. Both examples have median 50 without a score of 50 in either group.
What to explore next
Learn why averaging class medians can misrepresent a whole-school median and why a correlation number can miss a curved pattern. Visit the Core Aim learning hub, Education Hub and the Clementi Mathematics teaching reference for progressive understanding and independent checks.
The core aim of studying a box plot is not to distrust a useful graph. It is to know precisely which questions it answers—and which questions still need the underlying data.