What Is Correlational Research? Finding Patterns That Are Actually Real

In this post:

• Why stock market technical analysts see patterns that aren’t there, and what that reveals about a trap behavioral scientists work hard to avoid
• What correlational research actually is, and how scientists use statistics to separate real relationships from noise
• The single most important rule in correlational research: correlation is not causation
• How researchers go beyond simple correlations to analyze relationships between different types of variables—from continuous measures to categorical comparisons

There’s a certain kind of stock market investor who spends hours staring at charts. They study the jagged lines of price movements the way a fortune teller studies tea leaves, looking for patterns with odd names. They draw trend lines, calculate moving averages, and speak confidently about “support levels” and “resistance zones.” They do this because they want to make money and they believe that how a stock’s price has moved in the past can predict its future.

The investment approach that relies on a stock’s past performance is called technical analysis. In his classic book A Random Walk Down Wall Street, economist Burton G. Malkiel describes the theories these analysts construct. One pattern supposedly signals that a stock is about to rise. Another warns of decline. The charts are a kind of language, and the analyst becomes fluent at reading the story written in the data.

There’s just one problem: evidence suggests this approach doesn’t work.

Study after study has shown that the movements of stock prices are largely random—what economists call a “random walk.” The patterns that technical analysts see are, largely, what people see in clouds and constellations: shapes imposed by a pattern-hungry brain eager to make sense of noise.

The past does not reliably predict the future, at least not in the stock market.

People See Patterns Everywhere

Technical analysts are intelligent, highly trained, and convinced of their methods. So why do they see patterns that aren’t really there?

The answer lies in how the human mind works. People are, by nature, correlation-detecting machines. Our brains evolved to find relationships between things: the rustle in the grass that precedes the predator, the cloud formations that signal rain, the facial expressions that reveal the person across from you is angry. Detecting patterns served our ancestors well. The cost of seeing a pattern that wasn’t there (a false positive) was usually small—you hid from a rustle that turned out to be wind. But the cost of missing a pattern that was there (a false negative) could be death.

Thanks to this inheritance, people are wired to find patterns, even when they don’t exist. Drawing associations between the things we experience is a natural part of being human. But this natural tendency bumps into a critical question when we use science to investigate human behavior: how can we tell whether the relationships we observe are real?

How Science Separates Signal from Noise

The difference between a technical analyst and a behavioral scientist isn’t that one looks for patterns and the other doesn’t. The difference is that scientists have developed methods to test whether the patterns they see are real or just noise. These methods are the essence of correlational research.

What Is Correlational Research?

Unlike stock prices, many things in the world are associated with each other in reliable and measurable ways. People with more education tend to earn more money. People who experience high anxiety tend to experience high depression too. Older people, on average, report less emotional distress than younger people. These aren’t artifacts of a pattern-hungry brain. They’re real relationships that repeatedly emerge in scientific studies.

To establish these kinds of patterns, behavioral scientists follow a process. First, they measure the variables they are interested in. Then, they calculate a statistic—typically Pearson’s r—that quantifies both the direction and strength of the relationship. They test whether the correlation exceeds what you’d expect by chance alone, report effect sizes that indicate whether a relationship is small, medium, or large, and replicate their findings across different samples to see if the pattern holds.

This doesn’t mean scientists are always right. But it does mean they have tools for separating the signal from noise.

Correlation Is Not Causation

Finding a correlation between two variables doesn’t mean one causes the other. This is perhaps the most important lesson in correlational research, and one of the most frequently misunderstood.

Consider the correlation between ice cream sales and drowning deaths. Both increase in summer. Does ice cream cause drowning? Obviously not. A third variable—warm weather—is the cause of both.

Or consider the correlation between education and income. People with more education tend to earn more money. But does education cause higher income? Maybe. Or maybe people from wealthier families are more likely to pursue education and more likely to earn high incomes later. The correlation alone can’t tell us.

This limitation doesn’t make correlational research useless. It makes it the starting point. Correlations tell us where to look. Experiments, which you’ll learn about in Chapter 7, tell us whether one thing actually causes another.

What You’ll Learn in Chapter 5

Chapter 5 of Research in the Cloud teaches you the scientific way to find patterns in human behavior—and, just as importantly, how to avoid fooling yourself the way technical analysts do.

You’ll start with the fundamentals: what a correlation coefficient is, how to interpret its direction and strength, and how to tell if it is statistically significant. From there, you’ll learn to read correlation matrices, or tables that show how multiple variables relate to each other at once.

After covering the basics, you’ll explore how to detect associations between different types of variables. While many correlational examples are between variables treated as continuous, researchers sometimes want to compare categories—men versus women, employed people versus unemployed, one experimental condition versus another. You’ll learn how to analyze associations between categorical and continuous variables (using t-tests) and between two categorical variables (using chi-square tests), and you will do it using real data from hundreds of participants, examining the relationships between anxiety, depression, sleep, trauma, age, and income.

Finally, the chapter will guide you through a sample project investigating how different moral values are correlated with people’s decisions in the moral dilemma you used back in Chapter 3. After practicing each step of a correlational study, you’ll have the opportunity to design and conduct your own correlational study: forming a research question, collecting data, analyzing the relationships, and interpreting what you find.

By the end of the chapter, you’ll understand not just how to calculate correlations but why they matter in behavioral research.

This post is part of a series exploring the chapters of Research in the Cloud: An Introduction to Modern Methods in Behavioral Science by Aaron Moss, Jonathan Robinson, and Leib Litman.

Ready to learn how to find real patterns in behavioral data?

Read Chapter 5 for free! Research in the Cloud teaches you how to detect, measure, and interpret correlations in behavioral research.