In AP Statistics, experimental units are the smallest objects or individuals to which treatments are randomly assigned in an experiment (plants, driveways, twin pairs, people). When the units are human, they're called subjects. Identifying them correctly is the first step in describing any experimental design.
Experimental units are the smallest pieces of the experimental material that actually receive a treatment. Not the things you measure, and not the treatments themselves, but the things the treatments get assigned to. In the released FRQs, the experimental units have been corn plants, rosebushes, sections of driveway, and pairs of identical twins. When the units are people, statisticians call them subjects, but it's the same idea.
Here's the part that trips people up. The experimental unit is defined by what gets the treatment, not by what looks like an individual. If a researcher sprays an entire field with fertilizer, the field is the unit, even though there are thousands of plants in it. In the 2022 acne-drug FRQ, treatments were assigned within pairs of identical twins, so the pair functions as the block while each twin is a unit receiving a treatment. Getting this identification right determines how you count replications, how you describe random assignment, and what conclusions the experiment supports.
Experimental units live at the heart of Unit 3 (Collecting Data), specifically Topic 3.7, and they echo back to Topic 1.1's big idea that numbers only mean something in context. Learning objective AP Stats 3.7.A asks you to interpret the results of a well-designed experiment, and the essential knowledge behind it is built on units. VAR-3.E.2 says random assignment of treatments to experimental units lets you conclude that large observed differences are unlikely to be due to chance (that's statistical significance). VAR-3.E.3 says those significant differences are evidence the treatments caused the effect. And VAR-3.E.4 ties generalization to whether the units are representative of a larger population. In other words, the two biggest conclusions in all of AP Stats experiments, causation and generalization, both hinge on what happened with the experimental units.
Keep studying AP Statistics Unit 1
Random Assignment (Unit 3)
Random assignment is what you do TO experimental units. Randomly assigning treatments to units balances out lurking variables across groups, which is exactly what lets you claim the treatment caused the difference (VAR-3.E.2 and VAR-3.E.3). On FRQs, you describe random assignment in terms of the units, like 'randomly assign 50 of the 100 corn plants to the new fertilizer.'
Treatment (Unit 3)
Treatments and experimental units are the two halves of every design description. The treatment is the condition imposed; the unit is what it's imposed on. The 2019 fungus FRQ had four concentration levels (treatments) applied to insect-infested material (units). Mixing these up is the fastest way to lose points on a design FRQ.
Replication (Unit 3)
Replication means applying each treatment to multiple experimental units, so you count replication in units. One field sprayed with fertilizer is one unit and zero replication, no matter how many plants are in it. That's why identifying the unit correctly comes before judging whether a design is sound.
Sampling Variability and Generalization (Units 1 and 3)
How far your conclusion travels depends on where your units came from. The corn-plant practice question makes this concrete. If all 100 plants came from one Iowa farm, you can only generalize to corn on that farm, because the units aren't representative of a wider population (VAR-3.E.4). Causation comes from random assignment; generalization comes from how the units were selected.
An experimental design question shows up as Question 2 on the FRQ section almost every year (2019 fungus, 2022 acne twins, 2023 concrete driveways, 2026 coffee grounds and rosebushes), and identifying the experimental units is usually the implicit first move. You'll be asked to describe a completely randomized design, explain blocking or matched pairs, or interpret what conclusions are justified. In each case, your answer has to name what the units are and describe assigning treatments to them randomly. Multiple-choice questions test the conclusions side, asking which design feature lets you attribute a statistically significant difference to the treatment (answer: random assignment of treatments to units) and to which population results can be generalized (answer: only the population the units actually represent). The classic trap is a question where treatments are applied to groups, like whole fields or whole classrooms, and you have to recognize the group is the unit, not the individuals inside it.
These aren't rivals, they're nested. 'Subjects' is just the word for experimental units when the units are human. The acne-drug experiment has subjects; the rosebush experiment has experimental units. Where it gets interesting is designs like the 2022 twins FRQ, where each twin is a subject receiving a treatment but the twin pair acts as a block. On the exam, use whichever word matches the units, but know the underlying concept is identical.
Experimental units are the smallest things to which treatments are applied, and when those things are people, they're called subjects.
The unit is defined by what receives the treatment, so if a whole field gets sprayed, the field is one experimental unit no matter how many plants are in it.
Random assignment of treatments to experimental units is what justifies a cause-and-effect conclusion when the difference is statistically significant.
You can only generalize experiment results to a larger population if the experimental units are representative of that population, which usually requires random selection.
Replication is counted in experimental units, so each treatment needs to be applied to multiple units for the design to be valid.
Almost every FRQ design question (2019, 2022, 2023, 2026 all had one) starts with correctly identifying the experimental units.
Experimental units are the smallest objects or individuals to which treatments are applied in an experiment. In released FRQs they've been corn plants, rosebushes, driveway sections, and people (who are called subjects).
Almost. 'Subjects' is simply the term for experimental units when the units are human beings. A rosebush experiment has experimental units; the 2022 acne-drug experiment on twins has subjects.
No, and this is a common trap. The unit is whatever the treatment is applied to, so if a fertilizer is sprayed on entire fields, each field is one experimental unit even though it contains thousands of plants.
Treatments are the conditions the researcher imposes (like four fungus concentrations in the 2019 FRQ), while experimental units are the things those conditions get assigned to. Treatments are done; units are done to.
Not for a valid causal conclusion, since that comes from random assignment of treatments. But random selection of units determines generalization. If all 100 corn plants come from one Iowa farm, the results only generalize to corn on that farm.