A scatter plot displays the relationship between two variables by placing individual data points on a grid with a horizontal (x) axis and a vertical (y) axis. Each dot represents one observation, and the overall pattern of dots tells you whether the two variables are related, how strongly, and in what direction. Reading one is straightforward once you know what to look for.
What the Axes Tell You
The x-axis (horizontal) holds the explanatory variable, sometimes called the independent variable. This is the factor you think might influence the other. The y-axis (vertical) holds the response variable, or the outcome you’re measuring. For example, if a scatter plot shows hours of exercise per week on the x-axis and resting heart rate on the y-axis, the chart is asking: “As exercise increases, what happens to heart rate?”
Before interpreting any pattern, read both axis labels and check the scale. A scatter plot where the y-axis starts at 90 instead of 0 can make small differences look dramatic. Knowing the units (pounds, dollars, years) also matters because a “big” change on one scale might be trivial on another.
Direction: Positive, Negative, or None
The first thing to identify is which way the dots trend as you move from left to right.
- Positive correlation: The dots rise from lower-left to upper-right. As one variable increases, the other tends to increase too. Think height and weight: taller people generally weigh more.
- Negative correlation: The dots fall from upper-left to lower-right. As one variable increases, the other tends to decrease. More hours of TV per day, for instance, might correspond with lower test scores.
- No correlation: The dots look scattered randomly with no clear upward or downward trend. The two variables aren’t related in a predictable way.
Strength: Tight Clusters vs. Loose Clouds
Direction alone doesn’t tell you how reliable the relationship is. Strength does. A strong relationship means the dots cluster tightly around an imaginary line running through them. A weak relationship means the dots are spread out, forming more of a cloud than a band. If you can barely tell which direction the dots are trending because they’re so dispersed, the relationship is weak even if it technically tilts one way.
Many scatter plots include a correlation coefficient, labeled “r,” that puts a number on this. The value ranges from -1 to +1. An r of +1 means every dot falls perfectly on an upward line; -1 means every dot falls perfectly on a downward line; 0 means no linear relationship at all. In practice, an r above 0.7 or below -0.7 is generally considered strong. Values between 0.3 and 0.7 (or -0.3 and -0.7) are moderate, and anything closer to zero is weak. These thresholds vary slightly across fields, but they give you a reliable rule of thumb.
When you see an r value reported alongside a scatter plot, you’ll often also see a p-value. If the p-value is below 0.05, the relationship is considered statistically significant, meaning it’s unlikely to have appeared by random chance alone. A smaller p-value (like 0.001) means even stronger confidence. But statistical significance doesn’t automatically mean the relationship is large or important. A correlation of 0.31 can be statistically significant with a large enough sample while still being a weak relationship in practical terms.
Shape: Linear vs. Curved
Not every relationship follows a straight line. Some scatter plots show dots that curve upward, dip into a U-shape, or follow an S-pattern. These are called curvilinear relationships, and they’re important to recognize because a straight trend line would completely miss the real story.
Imagine plotting the relationship between stress level and job performance. At low stress, performance is low. As stress increases to a moderate level, performance rises. But past a certain point, more stress causes performance to drop again. The dots form an inverted U. If you only checked whether the overall trend was “positive” or “negative,” you’d conclude there’s no relationship at all, when in reality the relationship is strong but curved. Whenever the dots seem to bend, a simple correlation coefficient won’t capture what’s going on.
The Line of Best Fit
Many scatter plots include a straight line drawn through the data, called a line of best fit or trend line. This line is positioned to be as close to all the data points as possible, minimizing the overall distance between each dot and the line. It serves two purposes: it summarizes the general trend at a glance, and it lets you make predictions.
To use it for prediction, find a value on the x-axis, trace up to the line, then read across to the y-axis. If a scatter plot of advertising spending vs. revenue has a trend line, you could estimate the revenue you’d expect at a given spending level. Predictions within the range of existing data are more reliable than predictions that extend far beyond it. A trend line based on data from $10,000 to $100,000 in ad spend may not hold at $500,000.
Spotting Outliers
An outlier is a data point that sits far away from the general pattern. It might appear well above or below the cluster of dots, or off to one side where no other points exist. Outliers matter because they can shift the trend line and distort your interpretation of the relationship.
The impact depends on where the outlier sits. A point that falls far from the trend in the y-direction (vertically) but sits in the middle of the x-range is less likely to change the slope of the trend line significantly. In one analysis, removing such a point only shifted the slope from 5.12 to 5.04. But a point that sits far from the group in both the x and y directions can pull the trend line dramatically. In another analysis, a single influential outlier changed the slope from 5.12 to 3.32, nearly cutting it in half. When you see an obvious outlier on a scatter plot, mentally ask how different the trend line would look without it.
Clusters Within the Data
Sometimes a scatter plot doesn’t show one smooth pattern but instead has two or three distinct groupings of dots, each forming its own cluster. This usually signals that the data contains sub-populations with different characteristics. For example, a scatter plot of height vs. weight for a mixed group of children and adults would likely show two separate clusters rather than one continuous band.
When you notice distinct clusters, the overall trend line becomes misleading because it tries to average across groups that behave differently. The real insight comes from analyzing each cluster separately. Color-coded dots or different marker shapes are common ways that charts distinguish these groups, so always check the legend.
Correlation Does Not Mean Causation
This is the single most important rule for interpreting any scatter plot. A strong correlation between two variables does not prove that one causes the other. There is a negative correlation between the number of children a woman has and her life expectancy, but having fewer children doesn’t directly cause a longer life. A third factor, such as access to healthcare, independently influences both: better healthcare means more available birth control and longer lives.
The classic example is the positive correlation between ice cream sales and drownings at the beach. Ice cream obviously doesn’t cause drowning. Hot weather drives both variables up simultaneously. These hidden third factors are called lurking variables, and they’re present in far more scatter plots than most people realize. Even a strong dietary correlation, like the link between low saturated fat intake and reduced heart disease, could partly reflect genetics that make some people both less interested in fatty food and less prone to heart problems.
No matter how tight the dots cluster around a trend line, the scatter plot alone cannot establish that changing one variable will change the other. That requires a controlled experiment. When you read a scatter plot, describe what you see as a correlation: “These two things tend to move together.” Stop there unless experimental evidence supports a causal claim.