An intervening variable sits between a cause and its effect, explaining the process through which one variable influences another. Rather than being an outside factor, it is part of the causal chain itself. If exercise reduces anxiety, and the reason is that exercise increases certain brain chemicals that calm the nervous system, those brain chemicals are the intervening variable. They answer the question “why” or “how” the relationship works.
How Intervening Variables Fit in a Causal Chain
Every cause-and-effect relationship has at least two pieces: something that causes a change (the independent variable) and something that changes as a result (the dependent variable). An intervening variable adds a middle step. The independent variable causes a shift in the intervening variable, and that shift then causes the change in the dependent variable.
Take a concrete example. Suppose a researcher finds that students who attend tutoring sessions earn higher exam scores. The tutoring is the independent variable and exam scores are the dependent variable. But tutoring doesn’t magically raise scores. It works because students gain a better understanding of the material. That improved understanding is the intervening variable: tutoring increases understanding, and understanding increases scores. Without it, you know that tutoring helps but not why.
In research diagrams, this chain is drawn as a series of arrows. Tutoring points to understanding, and understanding points to exam scores. The arrow from tutoring directly to exam scores may still exist, but it typically shrinks once the intervening variable is accounted for. If it disappears entirely, the intervening variable fully explains the relationship.
Intervening Variable vs. Mediator Variable
You’ll often see the term “mediator variable” used interchangeably with “intervening variable,” and in most practical contexts they mean the same thing. Both describe a variable that transmits the causal effect of one variable to another. In published research, “mediator” tends to be the more common term, especially when researchers are running statistical tests. “Intervening variable” is the broader conceptual label, often used in textbooks and theory discussions. If your professor or a paper uses one term, you can safely treat it as equivalent to the other.
How It Differs From a Confounding Variable
This distinction trips up a lot of people, but it matters. A confounding variable is an outside factor that creates a false or misleading link between two variables. An intervening variable is part of the real causal pathway between them.
Here’s a useful example from clinical research. Suppose a study examines whether a sedative medication impairs cognitive test performance. Sleepiness is associated with both the medication and with poorer test scores. That might look like a confounding variable at first glance, but it isn’t. Sleepiness lies directly in the path between the medication and poor performance: the medication causes sleepiness, and sleepiness causes worse test scores. It’s an intervening variable, not a confounder, and researchers should not statistically adjust for it. Doing so would actually hide the real mechanism.
A confounding variable, by contrast, sits outside the causal chain. It independently affects both the cause and the outcome, creating a misleading association. If ice cream sales and drowning deaths both rise in summer, temperature is the confounder. It influences both variables separately rather than connecting them in a chain.
How It Differs From a Moderating Variable
A moderating variable answers a different question entirely. While an intervening variable explains how or why one thing leads to another, a moderating variable explains when or for whom the relationship is stronger or weaker.
Say a therapy program reduces depression symptoms. An intervening variable might be improved coping skills: the therapy teaches coping, and better coping reduces symptoms. A moderating variable might be severity of depression at the start of treatment. If the therapy works well for people with moderate depression but not for people with severe depression, severity is a moderator. It doesn’t transmit the effect; it changes how strong the effect is across different groups.
The key distinction is that an intervening variable specifies a causal sequence (X causes M, which causes Y), while a moderating variable does not sit in any causal chain. It simply tells you that the relationship between X and Y looks different depending on the level of the moderator.
How Researchers Test for Intervening Variables
Identifying an intervening variable isn’t just a matter of logic. Researchers use a formal statistical process called mediation analysis. The most widely known approach comes from a 1986 framework that lays out a series of conditions to check.
First, the independent variable must be related to the dependent variable. There needs to be a relationship to explain. Second, the independent variable must be related to the proposed intervening variable. If the cause doesn’t actually affect the middle step, it can’t work through it. Third, the intervening variable must predict the dependent variable even after accounting for the independent variable’s direct influence. This confirms the middle step has its own effect on the outcome. Finally, the direct link between the independent and dependent variables should weaken once the intervening variable is included in the analysis. If it shrinks to near zero, the intervening variable fully explains the relationship (called full mediation). If it shrinks but remains significant, the intervening variable partially explains it (partial mediation).
More recent research has found that this classic step-by-step method has low statistical power, meaning it often fails to detect real intervening effects. A better approach, now widely recommended, is to directly test two things: whether the independent variable significantly affects the intervening variable, and whether the intervening variable significantly affects the dependent variable. If both links are statistically significant, there is evidence of mediation. Researchers then calculate confidence intervals using methods like bootstrapping to confirm the indirect effect is real.
How Intervening Variables Appear in Path Diagrams
In visual models, intervening variables show up as a middle box or circle sitting between the independent and dependent variables. Measured variables (things you can directly observe and record, like test scores or hours of sleep) are drawn as rectangles. Variables you can’t directly observe (like motivation or self-esteem) are drawn as ovals or circles.
Arrows represent the direction of influence. A single arrow from the independent variable to the intervening variable is labeled “path a.” A single arrow from the intervening variable to the dependent variable is labeled “path b.” The product of these two paths (a multiplied by b) gives you the indirect effect, which is the portion of the total effect that flows through the intervening variable. A separate arrow may run directly from the independent variable to the dependent variable, labeled “path c prime.” This direct path represents whatever effect remains after the intervening variable is accounted for.
If the indirect effect (a times b) is large and the direct path (c prime) is small or zero, the intervening variable is doing most of the explanatory work. This visual framework makes it easier to see whether a proposed mechanism genuinely accounts for the relationship you’re studying, or whether something else is going on.