Why the Axes Matter
A scatterplot shows paired values as points, but the points cannot be interpreted properly without reading the axes. The axis labels tell you what the horizontal and vertical positions represent; the units tell you how to measure those positions; and the displayed ranges determine how much of each variable’s scale is visible. These choices affect the visual impression of the pattern.
For example, the same small increase in a measured quantity may look modest on a graph that displays a broad range and striking on one that displays only a narrow range. Likewise, changing a graph’s width or height can make a pattern appear flatter or steeper. Those visual changes do not change the recorded values. Reading a scatterplot well therefore means attending to both its points and the scale on which they are drawn.
Read the Labels, Units, and Tick Marks
Begin by checking which variable is on each axis. The horizontal axis is often called the \(x\)-axis and the vertical axis the \(y\)-axis, but those names do not tell you what the variables mean. A horizontal axis labelled “study time (hours)” and a vertical axis labelled “quiz score (points)” describes a different context from axes labelled “distance (kilometres)” and “fuel used (litres).”
Next, read the numbers at the tick marks. If adjacent ticks are labelled 10, 20, and 30, the scale increases by 10 units per interval. If they are labelled 0, 5, and 10, it increases by 5 units per interval. The physical spacing between tick marks is not enough to determine the scale: the numerical labels do that. When the axis uses a regular linear scale, equal distances on the page represent equal increments in the variable. If a scale is not regular, or if part of it is omitted, that feature should be clearly indicated and taken into account.
The displayed range is not necessarily the full set of possible values, nor does it necessarily start at zero. A vertical axis might show scores from 70 to 100 even if scores below 70 are possible. A horizontal axis might begin at 0 or at a positive value. Note the first and last displayed values, and check whether zero is included. A truncated axis is an axis whose displayed range omits zero or another substantial part of its scale. Truncation is not automatically invalid, but it can magnify the apparent size of changes.
Worked Example: Interpreting an Axis Scale
A hypothetical scatterplot has “walking time to clinic (minutes)” on the horizontal axis and “resting pulse (beats per minute)” on the vertical axis. The horizontal ticks are 0, 5, 10, 15, and 20; the vertical ticks are 60, 70, 80, 90, and 100. A plotted point is at \(x=10\), \(y=80\). Interpret its position and the displayed ranges.
The point represents a hypothetical observation with a walking time of 10 minutes and a resting pulse of 80 beats per minute. The displayed horizontal range is 0 to 20 minutes, with 5 minutes between adjacent ticks. The displayed vertical range is 60 to 100 beats per minute, with 10 beats per minute between adjacent ticks.
The vertical scale does not begin at zero. Thus, a visual comparison of vertical distances in this graph concerns differences within the 60-to-100 range, not differences on a scale extending from zero. The point’s pulse value is 20 beats per minute above the bottom of the displayed range, not 80 units above it on the page’s displayed scale.
How Ranges Affect Visual Emphasis
A graph allocates a fixed amount of space to the values within each displayed range. If an axis shows a narrow interval, a given data difference occupies a larger fraction of that space than it would on an axis showing a broad interval. This can make small differences look prominent. A broad range has the opposite effect: it can compress the same differences into a smaller part of the plot.
Consider a hypothetical set of measurements whose values lie between 80 and 100. If the vertical axis runs from 0 to 100, that interval uses only one fifth of the vertical scale. If the axis runs from 80 to 100, the same interval uses the full vertical scale. The data have not changed; the framing has. A reader who sees only the second graph may overstate the size of the observed variation unless the axis labels and range are checked.
Worked Example: A Truncated Vertical Axis
Suppose two hypothetical scatterplots show exactly the same four paired measurements. Their vertical values are 82, 86, 90, and 94 points. One plot displays a vertical range from 0 to 100 points, and the other displays a range from 80 to 100 points. Compare how much of the vertical plotting space the observed values occupy.
On the plot ranging from 0 to 100, the observed values span \(94-82=12\) points. The full displayed range is \(100-0=100\) points, so the data span \(12/100=0.12\), or 12% of the vertical scale.
On the plot ranging from 80 to 100, the same observations still span 12 points, but the displayed range is \(100-80=20\) points. They now span \(12/20=0.60\), or 60% of the vertical scale. The plotted vertical variation therefore takes up five times as much of the available vertical space, since \(0.60/0.12=5\).
The second graph can make the variation look much more dramatic. That does not mean it is necessarily misleading: a narrow range may be useful when small differences matter. But readers should be told what range is shown and should describe the difference in points, not just by its apparent size on the page.
Why Plot Dimensions Change Apparent Steepness
The shape seen on a page depends not only on the axis ranges but also on the width and height of the plotting area. A wide, short plotting area spreads horizontal distances while compressing vertical ones, making a rising pattern look flatter. A narrow, tall plotting area does the reverse. If either axis range changes, that also changes how many pixels represent one unit of that variable.
The following calculation makes the effect explicit for a straight-line pattern. Let \(R_x\) and \(R_y\) be the displayed ranges (maximum minus minimum) on the horizontal and vertical axes. Let \(W\) and \(H\) be the plot’s width and height in the same screen units, such as pixels. If the data pattern has slope \(m\), then a horizontal change \(\Delta x\) corresponds to a vertical change \(m\Delta x\). On the page, each horizontal data unit occupies \(W/R_x\) pixels and each vertical data unit occupies \(H/R_y\) pixels. Dividing the vertical pixel change by the horizontal pixel change gives the apparent slope magnitude:
The \(\Delta x\) cancels, so the plotted steepness depends on the data slope and the graph’s dimensions and displayed ranges. The expression is about appearance on the page; it is not a new measurement of the relationship in the data. In particular, comparing apparent angles across differently scaled plots is unreliable unless the axis ranges and plot dimensions are also considered.
Worked Example: One Pattern in a Wide and a Tall Plot
A hypothetical set of points lies on the line \(y=2x\), where \(x\) is measured in hours and \(y\) in points. Consider two plots with the same displayed ranges: \(x\) from 0 to 30 hours and \(y\) from 0 to 60 points. The first plotting area is 600 pixels wide and 150 pixels tall. The second is 240 pixels wide and 480 pixels tall. Compare the apparent steepness.
For both plots, the data slope is \(m=2\), \(R_x=30\), and \(R_y=60\). In the wide plot, the apparent slope magnitude is
In the tall plot, it is
The data points and axis ranges are identical, yet the line appears eight times as steep by this screen-based measure, because \(2/0.25=8\). The wide plot makes the pattern look comparatively flat; the tall plot makes it look steep. Neither appearance changes the values or the line \(y=2x\).
A Careful Reading Routine
Before describing a scatterplot’s visual pattern, use a short routine. This does not require calculating a statistic; it helps ensure that your description matches what the graph actually displays.
Read both axis labels and state what a one-unit change means in context.
Identify the smallest and largest displayed values on each axis, and notice whether zero is included.
Use the numerical labels to determine how much each tick-to-tick interval represents.
Notice whether the graph is unusually wide or tall, and avoid comparing apparent angles across plots with different scales or dimensions.
Say, for example, “the values span 12 points,” rather than relying only on how much space they occupy in the graph.
Common Mistakes and Why This Matters
A common mistake is to infer the size of a change from how large it looks on the page. For example, a reader might see a point pattern rise sharply in a plot with a narrow vertical range and conclude that the measured changes are large. That conclusion does not follow from appearance alone. Check the axis values, calculate differences in the stated units, and compare them with the context.
Another mistake is to assume that every axis begins at zero or that equally long plotted distances always mean equal numerical increments. Neither assumption is safe without reading the labels and scale. A graph may use a nonzero starting point, unequal tick intervals, or an indicated break. If a feature of the scale is unclear, do not guess what the plotted distances mean.
Finally, do not confuse a visual change with a change in the data. Re-drawing the same observations with a different range or plot shape changes their appearance, not their measured values. This distinction matters when interpreting one graph and especially when comparing graphs: two patterns that look different may be displayed on different scales, while two patterns that look similar may cover very different ranges in their original units.
Check Your Understanding
Answer each question using the axis information as well as the plotted pattern.
- A scatterplot’s horizontal axis is labelled “dose (milligrams)” and its vertical axis “response (units).” What information does each label provide, and why are the units important?
- An axis shows ticks at 20, 30, 40, and 50. What is the tick interval, and what displayed range do these ticks indicate?
- The same values from 82 to 94 are shown on vertical axes ranging from 0 to 100 and from 80 to 100. What percentage of each displayed range do the values span?
- Why can a rising pattern appear flatter in a wide, short plot than in a tall, narrow plot, even when the plotted data and displayed axis ranges are the same?
- A point pattern looks dramatic, but the vertical axis begins at a nonzero value. What should you check before describing the size of the observed differences?