Why Pattern Form Matters
A scatterplot can show that two quantitative variables tend to vary together, but not every pattern has the same shape. Some point clouds follow an overall straight path; others bend, flatten, or turn. Distinguishing these forms matters because a straight-line description assumes that changes in one variable are associated with a roughly steady rate of change in the other. A curved pattern can behave quite differently.
This tutorial focuses on describing the form of a scatterplot pattern. It does not require calculating a correlation or fitting a line. The first step is to look at the point cloud as a whole and ask whether its main path is approximately straight or visibly curved. A pattern may also have no clear overall form, or may contain distinct clusters; those possibilities should not be forced into either label.
What “Linear” Means Here
In a straight-line pattern, the typical change in the vertical variable for a given change in the horizontal variable is roughly steady over the range shown. For example, if \(x\) increases by the same amount repeatedly, the corresponding changes in the typical \(y\)-value are approximately similar. The points need not be equally spaced, and individual observations can sit above or below the general path.
A mathematical straight line can be written as \(y=a+bx\), where \(a\) and \(b\) are constants. This expression helps explain the idea of a steady rate: if \(x\) increases by an amount \(\Delta x\), then \(y\) changes by \(b\Delta x\), the same amount whenever \(\Delta x\) is the same. In a real scatterplot, the data are rarely exact points on a line, so the question is whether a straight path is a useful description of the overall form—not whether the equation fits every observation perfectly.
A curved pattern has a rate of change that varies across the plot. It may rise steeply at first and then level off, or fall and then turn upward. Curvature does not necessarily mean that the variables have no relationship. It means that a straight-line description misses an important part of the pattern’s shape.
Worked Example: A Roughly Steady Increase
Suppose a hypothetical greenhouse records the number of hours \(x\) that a lamp is used and the average height \(y\) of a group of seedlings, in centimeters. The observations are \((1, 5), (2, 7), (3, 9), (4, 11)\). Describe the form suggested by these points.
Compare equal steps in \(x\). Each step increases lamp-use time by 1 hour. The corresponding height changes are \(7-5=2\), \(9-7=2\), and \(11-9=2\) centimeters.
Describe the form. These particular points lie exactly on a straight line: height increases by 2 centimeters for each additional hour in this small hypothetical data set. For real observations, small departures from this steady pattern would still be consistent with an approximately linear form.
Stay within the evidence. The plot’s form can be described as approximately linear and increasing. This description does not establish that lamp use caused the height changes, nor does it say whether a straight-line model is suitable beyond the observed range.
How to Notice Curvature
A practical way to notice curvature is to follow the point cloud from left to right and ask whether its direction of travel changes. Does it become flatter or steeper? Does it bend upward or downward? Does it turn around? These questions concern the shape of the central path, not the distance of points from that path.
When data are listed in increasing order of \(x\), successive changes can make the pattern easier to see. If equal increases in \(x\) are accompanied by changes in \(y\) that grow or shrink systematically, the rate is changing. That is a clue to curvature. A scatterplot remains the main visual tool: the numerical differences support the description but do not replace checking the overall point pattern.
Worked Example: An Increase That Levels Off
A hypothetical materials lab records heating time \(x\), in minutes, and temperature \(y\), in degrees Celsius. The data are \((0, 20), (1, 38), (2, 50), (3, 57), (4, 61)\). Decide whether a straight-line description captures the pattern’s form.
Calculate changes over equal time steps. The increases in temperature are \(38-20=18\), \(50-38=12\), \(57-50=7\), and \(61-57=4\) degrees Celsius. The time increases by 1 minute at each step, but the temperature increases become smaller.
Interpret the changing rate. The values keep rising, while the rise becomes less steep. The pattern therefore bends and levels off rather than following a roughly steady straight path. A suitable description is “increasing with a curve that flattens as heating time increases.”
Check the scope. These five hypothetical observations suggest curvature over the recorded interval. They do not prove that temperatures will continue to level off in the same way at times not observed.
A Short Justification Using Successive Changes
Why do changing increments provide evidence against a straight-line pattern? Consider points whose \(x\)-values are equally spaced by an amount \(h\). If the points follow the straight-line form \(y=a+bx\), then at any two consecutive \(x\)-values, the change in \(y\) is:
The constants \(a\), \(b\), and \(h\) do not change from one step to the next. Therefore, the change in \(y\) is the same for every equal step in \(x\). If the observed changes instead grow or shrink in a persistent way, the points do not follow one exact straight line. With real data, there is scatter, so a few unequal changes do not by themselves rule out an approximately linear pattern. The useful clue is a systematic change in the rate that is also visible in the overall scatterplot.
Worked Example: A Pattern That Turns
A hypothetical motion sensor records an object’s horizontal position \(x\), in seconds after observation begins, and its measured height \(y\), in meters. The pairs are \((0, 9), (1, 4), (2, 1), (3, 0), (4, 1), (5, 4), (6, 9)\). Describe the form.
Follow the values in order. The height falls from 9 to 0 meters over the first three seconds, then rises from 0 to 9 meters over the next three seconds. The successive changes are \(-5, -3, -1, +1, +3,\) and \(+5\) meters per one-second step.
Describe the bend. The pattern decreases, reaches a low point near \(x=3\) seconds, and then increases. It is curved, with a clear turn, rather than one straight-line trend across the full interval. Calling it merely “decreasing” would miss the rise after the low point; calling it “increasing” would miss the initial fall.
Use an appropriately bounded statement. These hypothetical data have a U-shaped pattern over the observed times. The description does not claim that the same shape continues outside the recorded interval.
Describing Form Carefully
A useful form description says what the point cloud does as \(x\) increases. For a curved pattern, name the bend rather than forcing it into the language of a single straight trend. For example, a pattern might “increase steeply at first and then level off,” “decrease and then turn upward,” or “rise while becoming progressively steeper.” Include the measured variables and units when they help make the statement clear.
A descriptive statement should be no stronger than the plot supports. If the pattern only appears to bend slightly, say “some curvature is visible” rather than claiming a dramatic curve. If the points are too irregular to identify a consistent shape, say the form is unclear. A curve drawn through a few selected points can be misleading; attend to the overall path, not an imagined line that passes through every observation.
Form is only one feature of a scatterplot. Direction tells whether the general pattern rises or falls as \(x\) increases; strength concerns how closely points follow the pattern. These descriptions answer different questions. A curved pattern can rise overall, fall overall, or turn, and the presence of curvature alone does not tell how closely the points follow that curve.
| Visible pattern | Form-focused description |
|---|---|
| Points follow a roughly straight rising path | Approximately linear; increasing |
| Points follow a roughly straight falling path | Approximately linear; decreasing |
| Points rise, but the path gradually flattens | Curved; increasing at a decreasing rate |
| Points fall and then rise | Curved; decreases to a low point, then increases |
| Points show no consistent overall path | No clear linear or curved form |
Why the Distinction Matters
A straight-line summary can hide important behavior when the actual pattern bends. In the heating example, one overall line would not express that temperature rises more slowly later in the observed interval. In the motion example, a single straight direction would fail to capture the change from decreasing to increasing. Describing form first helps prevent a convenient straight-line summary from being mistaken for a faithful account of the data.
Recognizing curvature also helps keep conclusions within the data’s scope. A visible bend across the observed values does not automatically justify predicting far beyond them. Nor does a curved pattern alone identify a physical explanation. The plot describes how the recorded variables vary together; explanations require additional information about how the data were collected and what mechanisms may be involved.
Common Mistakes
- Calling every rising pattern linear. A plot can rise while becoming flatter or steeper. Check whether the path is roughly straight, not just whether it goes upward overall.
- Treating small differences as proof of curvature. Real observations vary. Look for a persistent bend in the overall point cloud rather than relying on one unequal pair of changes.
- Describing a turning pattern with only one direction. “Decreasing” omits the rise after a low point. Describe both the initial movement and the turn.
- Assuming a curve means a strong relationship. Form and strength are different features. Curvature describes shape; it does not by itself describe how closely points follow that shape.
- Extending the pattern beyond the plot. A bend observed over a limited range does not establish what happens outside that range.
Check Your Understanding
For each question, focus on the form of the pattern rather than calculating a correlation.
- For equal increases in \(x\), the typical \(y\)-values increase by about 3 units each time. What form would this support, and why?
- A scatterplot rises from left to right but becomes progressively flatter. Explain why “approximately linear and increasing” may not adequately describe its form.
- Successive \(y\)-changes over equal \(x\)-steps are \(2, 3, 2, 10\). Is this alone enough to establish curvature? Explain what else you would inspect.
- A pattern falls to a minimum and then rises. Write a form-focused description that captures both parts.
- Why can a curved pattern not be described adequately by saying only that it is increasing?