Written and maintained by CASRAI Editorial Board
Last updated
A manipulated variable is the variable a researcher deliberately changes, or sets to different levels, to see what effect the change has on an outcome. In an experiment it is the same thing as the independent variable. This guide gives the definition, the terminology map (manipulated variable, independent variable, factor), a worked example with the arithmetic shown, and short Excel, R and Python snippets.
What is the definition of a manipulated variable?
The definition of a manipulated variable is simple: it is the variable the experimenter controls and changes on purpose. OpenStax Psychology 2e puts it this way: “An independent variable is manipulated or controlled by the experimenter.” The outcome the researcher measures to see how much effect the manipulation had is the dependent variable (also called the measured or response variable).
So the meaning of manipulated variable has two parts. First, the researcher chooses the levels (for example, no practice session versus a 30-minute practice session). Second, the researcher decides which participants receive which level. If a variable is only observed as it already exists in people, such as age or an existing habit, it is measured, not manipulated.
Manipulated variable vs independent variable: the same thing?
In an experiment, yes. “Manipulated variable” and “independent variable” name the same role; the first describes what the researcher does to it, the second describes how it behaves in the design. Engineering and industrial statistics often use a third word for it. The NIST/SEMATECH e-Handbook of Statistical Methods refers to the primary variables that affect the responses of interest as “factors”.
| Role in the study | Common names | What the researcher does |
|---|---|---|
| Variable that is changed to see its effect | Manipulated variable, independent variable, factor | Sets the levels and assigns participants or units to them |
| Variable that is observed as the outcome | Measured variable, dependent variable, response | Records it and compares it across levels |
For the related pair used in non-experimental work, see explanatory variable vs response variable, and for the full list of variable roles see types of variables.
Define manipulated variables: levels, groups and conditions
A manipulated variable always has at least two levels, because an effect can only be seen by comparison. The levels are usually delivered as groups. OpenStax describes the standard arrangement: the experimental group “gets the experimental manipulation” and the control group does not.
How to manipulate the independent variable
The phrase “manipulate independent variable” describes the step of setting its levels and giving them to participants. OpenStax notes that with random assignment, “all participants have an equal chance of being assigned to either group”. Random assignment is what lets the researcher say the groups were alike before the manipulation. The difference between assigning people to groups and drawing a sample is covered in random assignment vs random sampling.
Worked example with real numbers
This is an illustrative example with made-up data, not a real study. Ten participants are randomly assigned to two groups of five. The manipulated variable is a 30-minute practice session before a memory test, with two levels: no practice (control) and practice. The measured variable is the number of items recalled out of 20.
| Condition | Scores | Sum | n | Mean |
|---|---|---|---|---|
| Control (no practice) | 12, 14, 11, 15, 13 | 65 | 5 | 65 / 5 = 13 |
| Practice (30 minutes) | 16, 18, 15, 19, 17 | 85 | 5 | 85 / 5 = 17 |
Arithmetic: 12 + 14 + 11 + 15 + 13 = 65; 16 + 18 + 15 + 19 + 17 = 85. The difference in group means is 17 − 13 = 4 items.
Meaning of the result: manipulated vs controlled and measured variables
The 4-item difference is the observed effect of the manipulation in this sample. Reading guidance:
| What you see | How to read it |
|---|---|
| Group means differ by 4 (17 vs 13) | The practice group recalled 4 more items on average; whether this is more than chance variation needs a significance test or confidence interval |
| Participants were randomly assigned | The groups can be treated as comparable before the manipulation, so the difference points to the manipulation |
| Participants chose their own condition | The variable was selected, not manipulated, and the difference may reflect who chose practice |
Three variable roles are easy to mix up. The manipulated variable is changed by the researcher. The measured variable is the outcome. A controlled variable is held constant for everyone (here, the same test, the same time limit and the same room), so it cannot explain the difference.
Assumptions and common mistakes
- Calling an observed variable manipulated. OpenStax explains that in correlational research some other factor, a confounding variable, may be “actually causing the systematic movement in our variables of interest.” See confounding variable and lurking variable.
- No control group. With only one level, there is nothing to compare against.
- Changing several things at once. If the practice group also had a different room or tester, you cannot tell which change produced the effect.
- Confusing a manipulated variable with a mediator. A mediator sits between the manipulation and the outcome; see mediator variable.
- Skipping random assignment. Without it, the groups may differ before the manipulation.
How to report it (APA example)
A generic sentence for the example above: “Participants in the practice condition recalled more items (M = 17.00) than participants in the control condition (M = 13.00).” Report the manipulated variable by name, its levels, how participants were assigned, the measured variable, and the group means with a test result.
How to calculate it in Excel, R or Python
Excel
Put the condition label in column A and the score in column B (rows 2 to 11). Then:
=AVERAGEIF(A2:A11,"control",B2:B11) returns 13
=AVERAGEIF(A2:A11,"practice",B2:B11) returns 17To assign conditions at random, put =RAND() next to each participant, sort by that column, and give the first five rows “control” and the last five “practice”.
R
set.seed(1)
assigned = sample(rep(c("control","practice"), each = 5)) # random assignment of 10 people
score = c(12,14,11,15,13, 16,18,15,19,17)
cond = rep(c("control","practice"), each = 5)
tapply(score, cond, mean) # control 13, practice 17Python
import random, statistics
ids = list(range(1, 11))
random.seed(1); random.shuffle(ids)
control_ids, practice_ids = ids[:5], ids[5:] # random assignment
control = [12, 14, 11, 15, 13]
practice = [16, 18, 15, 19, 17]
print(statistics.mean(control), statistics.mean(practice)) # 13 17FAQ
What is the definition of manipulated variable?
It is the variable the researcher deliberately changes or sets to different levels to test its effect on an outcome. In an experiment it is the independent variable.
Is a manipulated variable the same as an independent variable?
In an experiment, yes. In non-experimental studies the predictor is not manipulated, so it is better called an explanatory or predictor variable.
Can a study have more than one manipulated variable?
Yes. Designs with several factors are common; each factor has its own levels. Keep all other conditions constant so each effect can be separated.
What is the difference between a manipulated variable and a controlled variable?
The researcher changes a manipulated variable and holds a controlled variable constant.
What is a manipulated variable called in engineering?
The NIST/SEMATECH e-Handbook refers to the primary variables that affect the responses of interest as “factors”.
Sources
- OpenStax, Psychology 2e, section 2.3 “Analyzing Findings”: https://openstax.org/books/psychology-2e/pages/2-3-analyzing-findings (checked 2026-10-29).
- NIST/SEMATECH e-Handbook of Statistical Methods, section 5.1.1: https://www.itl.nist.gov/div898/handbook/pri/section1/pri12.htm (checked 2026-10-29).
