A histogram of student scores can make a group look concentrated in one range. But each bar combines individual values into an interval. The choice of interval boundaries affects what can be seen, and a published grouped table may not contain enough information to reconstruct the original scores.

NIST describes histograms as counts grouped into class intervals and distinguishes counts from other normalizations. The example below uses invented scores and equal-width intervals. It is not an estimate of any institution's assessment results or a recommendation for a universal bin width.

Start with the same eight observations

Suppose eight fictional scores are 40, 41, 42, 43, 57, 58, 59 and 60. First use intervals that include the lower boundary and exclude the upper boundary, written as [lower, upper). Under that rule, a score of 60 belongs in the interval starting at 60, not the one ending at 60.

Interval Count
[40, 50) 4
[50, 60) 3
[60, 70) 1

These are the same eight students throughout: 4 + 3 + 1 = 8. The middle interval contains no scores from 50 through 56, but its bar still has height three because it contains 57, 58 and 59. A reader cannot see the empty portion within that bar from its height alone.

Now use wider, 20-point intervals aligned at 40: [40, 60) contains seven observations and [60, 80) contains one. The seven-person bar combines the lower cluster of four and upper cluster of three. It does not prove that the scores are evenly distributed inside the interval. The apparent shape changed even though no student's score changed.

Change one design decision at a time

Width is not the only choice. Shifting the boundary can move a value into a neighboring bar even when the width remains fixed. The interval convention also matters at exact boundaries, as the score of 60 shows. State the boundary rule and check that every observation is counted once.

Compare counts with counts, or proportions with proportions, and keep the denominator visible. If the eight observations are a sample of a larger cohort, the counts describe the sample unless the report supplies a justified method for generalizing. Missing or suppressed observations cannot be silently allocated to a bar.

Unequal-width bins need extra care: a wider interval has more room to accumulate observations. If a display uses density rather than raw counts, both height and width affect the represented proportion. Check the axis label and methodology before comparing bar heights across unequal intervals. The simple table above deliberately uses equal widths within each display.

Do not reconstruct what the table does not contain

If only the wide counts seven and one were published, many different sets of eight scores could produce them. The table alone cannot tell a reader the exact seven scores, the median or the size of the gap. It also cannot show whether the gap is meaningful in a wider population. A grouped distribution is useful evidence, but it does not retain every detail of the raw data.

Where privacy rules prevent publication of individual values, a more detailed but appropriately protected summary may be available. Ask for interval definitions, counts, missing values and the applicable disclosure policy. Do not demand identifiable records or infer an individual's result from a small group.

Use the chart for the intended question

If a student is comparing the distribution of class marks, first check whether the same course, assessment, year and scoring scale are being compared. Different bins can make two charts look more or less similar, but matching bins cannot make different assessments equivalent. A chart's source and population remain essential.

A reproducible reading note records the boundaries, widths, denominator and any missing observations. When feasible, inspect more than one sensible grouping and compare it with the underlying data. This makes the visualization a way to ask better questions rather than an apparent answer that conceals the choices used to draw it.