Why Histogram Axes and Boundaries Matter
A histogram is meant to show how quantitative observations are distributed across intervals. Its appearance depends not only on the data but also on choices about the horizontal and vertical axes and the bin boundaries. A distorted scale can make ordinary differences look dramatic; a gap or endpoint ambiguity can suggest that data are missing, overlapping, or assigned inconsistently.
A useful check is to ask whether the display tells a reader exactly what values are represented, how the vertical scale is measured, and where each bin begins and ends. The bars should follow one consistent set of intervals, and the axis scale should allow their heights or areas to be compared fairly.
For equal-width bins with frequency on the vertical axis, the height of each bar is the number of observations in that bin. The bars represent the distribution across the intervals, so the vertical axis should ordinarily start at zero. When bin widths differ, bar area—not simply bar height—must represent frequency; the earlier material on histograms with unequal bin widths covers that distinction. In either case, bars representing adjacent intervals share a boundary and touch unless there is a genuine interval with no observations.
Check the Vertical Scale
A vertical axis that starts well above zero can magnify apparent differences between bar heights. The bars are drawn from a baseline, so readers tend to compare the visible bar lengths. If the baseline is quietly moved upward, a small difference in frequency can occupy a large share of the visible plotting area.
For a standard frequency histogram, start the frequency axis at zero and use evenly spaced tick marks. Label the axis with the quantity being counted, such as “Number of observations.” If the vertical axis shows relative frequency or density instead, label that scale clearly and make sure its values and bar areas have the intended meaning.
Worked Example: Diagnosing a Truncated Frequency Axis
Imagine a hypothetical histogram of travel times for 100 bus journeys, grouped into equal-width intervals. Two adjacent bins contain 40 and 45 journeys. The graph labels the vertical axis at 35, 40, 45, and 50, so the bars appear to rise from 35 rather than from zero. What is misleading, and how should the histogram be redrawn?
Identify the distortion. The actual difference between the two bin frequencies is \(45-40=5\) journeys. But the visible bar lengths above the displayed baseline are \(40-35=5\) and \(45-35=10\) units. A reader comparing those visible lengths could conclude that the second bar is twice as tall as the first, even though its frequency is only \(45/40=1.125\) times as large.
Redraw. Keep the same bins and frequencies, but draw both bars from zero. Use a vertical axis labeled “Number of journeys,” with evenly spaced ticks such as 0, 10, 20, 30, 40, and 50. The bars then have heights 40 and 45. The second is slightly taller, matching the actual frequencies.
The horizontal axis must also name the measured quantity and its units—for example, “Travel time (minutes).” The redrawn display preserves the data while removing the visual exaggeration caused by the truncated vertical scale.
A horizontal axis has a different role from a vertical frequency axis. A histogram may show only a selected range of values, but leaving out part of the range can hide observations or the overall extent of the data. If the horizontal scale is deliberately restricted, state the range clearly and indicate that the display is cropped; do not let the reader mistake the visible range for the full range of the data. In an introductory display, showing all the relevant bins is usually the clearest choice.
Worked Example: Repairing an Unclear Horizontal Axis
A hypothetical histogram summarizes the weights of packages, measured in kilograms. Its horizontal axis shows the labels 2, 4, and 6, but there is no axis title, no unit, and no indication of whether these labels mark bin edges or bin centers. The bars cover intervals that appear to run from 1 to 7 kg.
What cannot be determined? A reader cannot tell from the graph alone whether the numbers are kilogram values, whether they mark boundaries or centers, or whether any measurements below 1 kg or above 7 kg were omitted. The displayed range and the measurement units are essential parts of the graph’s meaning.
Redraw. Label the horizontal axis “Package weight (kg).” Mark the bin boundaries consistently at 1, 2, 3, 4, 5, 6, and 7 kg, and show all the bins that cover the observations. Label the vertical axis with the frequency scale and begin it at zero. If the bars cover 1–7 kg only because the graph is intentionally cropped, state that explicitly; otherwise, extend the display to include the full range of relevant observations.
The histogram now communicates both what was measured and the intervals represented by its bars. A reader can interpret the graph without guessing what the axis labels mean.
Bars, Gaps, and Adjacent Bins
Unlike a bar chart for categories, a histogram represents intervals on a continuous number line. If two bins are adjacent, they share a boundary. Their bars should touch: an empty strip between them suggests a gap in the measured values that is not present in the bin definitions. A graphical sliver between adjacent bars can also make a continuous scale look like separate categories.
A gap is appropriate when it represents an actual bin with frequency zero. For example, if the bins are \([0,5)\), \([5,10)\), \([10,15)\), and \([15,20)\), and the \([10,15)\) bin has no observations, the histogram has no bar rising above the baseline over that interval. That empty interval may create a visible gap between positive bars on either side. The gap is meaningful because the zero-frequency bin is part of the scale; it is not merely spacing added between adjacent nonempty bars.
This distinction depends on the bin definitions. If a graph leaves an unlabelled stretch between bars, a reader cannot tell whether it is a real interval with no observations, an omitted range, or a formatting error. Label the bin boundaries or otherwise make the intervals clear. Do not insert gaps between bars for bins that meet at a boundary.
Worked Example: Distinguishing a Real Gap from Bar Spacing
A hypothetical data set records the number of minutes that 12 people waited for a shuttle. The observations are grouped into four equal-width bins with frequencies as follows:
- \([0,5)\): 3 observations
- \([5,10)\): 5 observations
- \([10,15)\): 0 observations
- \([15,20)\): 4 observations
The frequencies sum to \(3+5+0+4=12\), which accounts for all 12 observations. How should the bars be drawn?
The first two bins are adjacent and both have positive frequency, so their bars should touch at 5 minutes. The \([10,15)\) bin is a genuine empty interval, so no bar rises above zero from 10 to 15 minutes. The bar for \([15,20)\) begins at 15 minutes, with height 4. Thus, the empty interval—not extra spacing between adjacent bars—explains the visible gap from 10 to 15 minutes.
The horizontal axis should be labeled “Waiting time (minutes),” with the boundaries 0, 5, 10, 15, and 20 shown or otherwise made unambiguous. This makes the gap interpretable as an empty interval rather than a missing part of the graph.
Make Bin Endpoints Unambiguous
Bin boundaries must assign every observation to exactly one bin. Consider the intervals “0 to 10” and “10 to 20.” If both endpoints are included in both intervals, a value of exactly 10 belongs to two bins. If neither interval includes 10, that value belongs to neither. Either problem can change the frequencies or leave observations unaccounted for.
A standard solution is to use half-open intervals: include the lower endpoint and exclude the upper endpoint. For example, \([0,10)\) includes 0 but not 10, while \([10,20)\) includes 10 but not 20. A final bin can include its upper endpoint if needed to include the maximum recorded value. State the convention or show interval labels that make it clear. This matters particularly for rounded measurements, where multiple observations may equal a boundary exactly.
Worked Example: Fixing Overlapping Endpoint Labels
A hypothetical set of recorded measurements in units is \(9.8,\ 10.0,\ 10.0,\ 19.9,\ 20.0,\ 20.0,\ 29.9\). A draft histogram labels its bins “0–10,” “10–20,” and “20–30,” without saying whether the endpoints are included. How can the bins be defined so the counts are unambiguous?
Use the intervals \([0,10)\), \([10,20)\), and \([20,30]\). The first bin contains 9.8, so its frequency is 1. The second contains both 10.0 values and 19.9, so its frequency is 3. The final bin includes 20.0, 20.0, and 29.9, so its frequency is also 3. The total is \(1+3+3=7\), equal to the number of recorded measurements.
The boundary convention assigns each value of 10.0 to the second bin and each value of 20.0 to the final bin. No value is counted twice or left out. The first two bins meet at 10, and the second and third meet at 20, so their bars should touch where their intervals are adjacent.
A Reliable Redrawing Process
When a histogram seems misleading, separate the underlying data choices from the drawing choices. The goal is not to change the distribution or to make it look more dramatic. It is to make the scale and intervals visible enough that readers can recover what the bars mean.
Identify the variable and its units on the horizontal axis, and identify whether the vertical axis shows frequency, relative frequency, or density.
Look for a truncated frequency axis, uneven tick spacing, a restricted horizontal range, or missing units. Use a clearly labeled scale; a frequency axis should begin at zero.
Confirm that the intervals cover the intended range without overlap or unintended gaps, and decide how boundary values are assigned.
Bars for adjacent bins should touch. Show an empty bin at zero rather than disguising it as unexplained spacing; when widths differ, use the appropriate density scale so area represents frequency.
Why These Checks Matter
A histogram is a visual argument about where observations fall and how many occupy each interval. A misleading axis can exaggerate differences; ambiguous bin endpoints can produce conflicting counts; and unexplained gaps can imply that the data contain a range with no observations. These are not merely cosmetic problems. They affect what readers can reasonably conclude from the display.
The most common mistake is to say, “The bars are separated, so the data have a gap,” without checking the bin boundaries. The correction is to distinguish a real empty interval from a drawing gap: adjacent bins touch, while a zero-frequency interval is represented by a bin with no bar above the baseline. Similarly, a truncated vertical axis does not alter the counts, but it can make their differences look much larger than they are. Check the scale before describing the pattern.
Before interpreting a histogram, verify that the axes have clear labels and units, that the vertical scale is fair, and that the bins have consistent, explicit endpoints. If any of those features is unclear, redraw the display or explain its limitations before describing the distribution.
Check Your Understanding
Use the scale and boundary rules to answer these questions.
- A frequency histogram has bars with counts 24 and 28, but its vertical axis begins at 20. Explain how that choice could mislead a reader and name one repair.
- Bins \([0,5)\), \([5,10)\), and \([10,15)\) have frequencies 2, 4, and 3. Should the bars have gaps between them? Explain.
- If the bin \([10,15)\) has frequency zero while the bins on either side have positive frequencies, what should the histogram show over 10 to 15?
- Why are the labels “0–10” and “10–20” potentially ambiguous for measurements that can equal 10 exactly? Give a boundary convention that resolves the problem.
- A histogram’s horizontal axis is labeled only with numbers, with no variable name or units. What information should be added before interpreting the display?