A zero-order correlation is the straightforward relationship between two variables, measured without accounting for any other variables that might influence them. The term “zero-order” refers to the fact that zero additional variables are being controlled for. It produces a value between -1 and +1, where -1 means a perfect negative relationship, +1 means a perfect positive relationship, and 0 means no linear relationship at all.
If you’ve encountered this term in a research paper, a statistics class, or while reading a study’s results table, you’re essentially looking at the most basic form of correlation. It’s the starting point before any statistical adjustments are made.
How It Works
A zero-order correlation is almost always a Pearson correlation coefficient (represented by the letter “r”). It works by comparing how two variables move together relative to how much they each vary on their own. When one variable tends to be above its average at the same time the other is above its average, the correlation is positive. When one goes up while the other goes down, the correlation is negative.
The result always falls between -1 and +1. A value of 0.85 between height and weight, for example, would indicate a strong positive relationship: taller people tend to weigh more. A value of -0.85 between outdoor temperature and heating bills would mean higher temperatures are strongly associated with lower bills. A value near zero would suggest no consistent linear pattern between the two variables.
For the calculation to be valid, both variables need to be measured on a continuous scale (like height, income, or test scores rather than categories). The relationship between them should be roughly linear, meaning it follows a straight-line pattern rather than a curve. The data should be approximately normally distributed and free of extreme outliers that could distort the result.
Why It’s Called “Zero-Order”
The name makes more sense when you see it alongside its counterparts. In statistics, the “order” of a correlation refers to how many other variables are being held constant. A zero-order correlation controls for nothing. A first-order partial correlation controls for one variable. A second-order partial correlation controls for two, and so on.
Say you’re studying the relationship between hours of exercise and blood pressure. The zero-order correlation gives you the raw association between those two variables. A partial correlation might then control for age, since older people tend to exercise less and have higher blood pressure. If the correlation between exercise and blood pressure shrinks substantially after controlling for age, that tells you age was inflating the original relationship. If it stays about the same, the relationship holds up even after accounting for age.
The zero-order correlation is the baseline. It tells you what the data look like before you start peeling back layers.
Interpreting the Strength of a Correlation
A widely used set of benchmarks, originally proposed by the psychologist Jacob Cohen, categorizes correlation strength as follows:
- Small: r = 0.10
- Medium: r = 0.30
- Large: r = 0.50
These thresholds apply to the absolute value, so -0.30 is the same strength as +0.30, just in the opposite direction. Later researchers have argued these benchmarks are too generous for certain fields. A revised guideline suggests that in individual differences research (personality, cognitive ability, behavior), correlations below 0.20 should be considered small, 0.20 to 0.30 typical, and anything above 0.30 relatively large. Context matters: a correlation of 0.25 between a single personality trait and job performance would be considered meaningful, while the same value in a physics experiment might signal measurement problems.
Where You’ll See It in Research
In published studies, zero-order correlations typically appear in a correlation matrix, a table that shows the relationship between every pair of variables in the dataset. These matrices are symmetric, meaning the correlation between variable A and variable B is the same as between B and A. The diagonal always shows 1.00, since every variable correlates perfectly with itself.
A typical matrix might look like a grid with variable names along both the top and left side, filled with correlation values and sometimes asterisks indicating statistical significance. Researchers include these tables so readers can see the raw relationships before more complex analyses (like regression or structural equation modeling) are reported. When a paper says “zero-order correlations are presented in Table 1,” it’s showing you those unadjusted, pairwise relationships.
The Biggest Limitation: Hidden Third Variables
The most important thing to understand about zero-order correlations is what they can’t tell you. Because no other variables are controlled, the relationship you see might be driven entirely by something you haven’t measured.
The classic example: ice cream sales and shark attacks are positively correlated. But ice cream doesn’t cause shark attacks. Hot weather drives both variables up simultaneously, because people buy more ice cream and swim more often when it’s warm. Temperature is the confounding variable creating a spurious correlation between two things that have no direct connection.
Confounding variables aren’t the only source of misleading results. Sometimes a chain of cause and effect creates the illusion of a direct link. If A causes B, and B causes C, then A and C will be correlated even though A has no direct effect on C. Random sampling error can also produce correlations in a sample that don’t exist in the broader population, especially with small sample sizes. And occasionally, two completely unrelated variables just happen to follow similar patterns by chance.
This is precisely why researchers don’t stop at zero-order correlations. They use partial correlations, regression models, and other techniques to test whether a relationship holds up once competing explanations are accounted for. The zero-order correlation is the first look, not the final word.
Zero-Order Correlation vs. Partial Correlation
The practical difference comes down to one question: are you looking at the raw relationship, or the relationship that remains after removing the influence of other variables?
A zero-order correlation between study time and exam scores might be 0.45. But if you control for prior knowledge of the subject (a partial correlation), the value might drop to 0.20. That would suggest a good chunk of the original association was explained by the fact that students who already knew the material also tended to study more.
A related technique, the part correlation (also called semi-partial correlation), removes the influence of a third variable from only one of the two variables rather than both. This is commonly used in regression analysis to determine how much unique variance a single predictor explains.
In practice, researchers often report all three. The zero-order correlation shows the full picture, the partial correlation shows what survives after adjustment, and comparing the two reveals how much of the original relationship was driven by other factors.