In AP Statistics, replication means having more than one experimental unit in each treatment group. It's one of the four elements of a well-designed experiment (along with comparison, random assignment, and control), and it lets researchers distinguish real treatment effects from chance variability.
Replication has a very specific meaning on the AP Stats exam, and it's probably narrower than what you'd guess. Per the CED (3.5.B), replication means more than one experimental unit in each treatment group. That's it. If you give the new drug to one person and the placebo to one person, you have zero ability to tell whether a difference in outcomes came from the treatment or just from those two individuals being different people. Multiple units per group lets you see how much responses vary within a group, which is the yardstick you need to judge whether differences between groups are real.
Here's the catch that trips people up. In everyday science talk, 'replication' often means repeating the whole experiment to verify the results. That's a real idea (sometimes called replicability or reproducibility), but it is NOT the AP definition. When an FRQ asks you to describe a well-designed experiment, replication means putting several units in each treatment group, not running the study twice.
Replication lives in Topic 3.5 (Introduction to Experimental Design) in Unit 3: Collecting Data, directly supporting learning objective AP Stats 3.5.B, which asks you to describe the elements of a well-designed experiment. The CED lists exactly four: comparison of at least two treatment groups, random assignment, replication, and control of confounding variables. Experimental design FRQs show up constantly (2018 Q4 on ACL surgery recovery and 2021 Q2 on walking and cholesterol both required describing a well-designed experiment), and replication is one of the boxes graders look for. It also sets up Units 6-7, because the inference procedures you'll run on experimental data only make sense when each group has enough units to estimate variability.
Keep studying AP Statistics Unit 3
Randomization (Unit 3)
Random assignment and replication work as a team. Randomization balances out confounding variables across groups, but it only works well when there are enough units to balance. Randomly assigning 2 people can easily produce lopsided groups; randomly assigning 50 usually won't.
Sample Size (Units 3, 6-7)
Replication is why sample size matters in experiments. More units per treatment group means a better estimate of within-group variability, which later (in inference) translates to smaller standard errors and more power to detect a real treatment effect.
Control Group (Unit 3)
Replication applies to every treatment group, including the control. A study with 100 people on the new drug but only 3 in the control group has technically replicated, but weakly. Both groups need enough units for the comparison to mean anything.
Confounding Variable (Unit 3)
Replication helps neutralize confounders indirectly. Individual quirks (age, genetics, motivation) average out across many units in each group, so observed differences between groups are more believable as treatment effects rather than one weird participant.
Multiple-choice questions test whether you know the precise definition. A classic stem describes a study and asks which scenario violates replication (answer: a treatment group with only one experimental unit) or asks why replication is essential (answer: it lets you measure variability and attribute differences to treatments rather than chance). On FRQs, replication shows up inside 'describe a well-designed experiment' prompts like 2018 Q4 (ACL surgery) and 2021 Q2 (walking and cholesterol). To earn full credit, your design should explicitly assign multiple subjects to each treatment via random assignment. The most common point-loser is defining replication as 'repeating the experiment,' which won't earn the design element graders are looking for.
In AP Stats, replication does NOT mean running the whole study again to check the results. That broader scientific idea is replicability or reproducibility. The AP definition is strictly about within-study design, meaning each treatment group contains more than one experimental unit. If an FRQ asks for a well-designed experiment and you write 'replicate by repeating the study,' you've described the wrong concept and likely missed the point.
On the AP exam, replication means having more than one experimental unit in each treatment group, not repeating the entire experiment.
Replication is one of the four elements of a well-designed experiment, alongside comparison, random assignment, and control of confounding variables (AP Stats 3.5.B).
Replication lets you estimate variability within each group, which is what makes it possible to tell whether differences between groups are due to the treatment or just chance.
Random assignment needs replication to work, because balancing confounders across groups only happens reliably when each group has enough units.
A common FRQ mistake is describing replication as 'doing the study again,' which does not earn credit for this design element.
Replication means having more than one experimental unit in each treatment group. It's one of the four elements of a well-designed experiment in Topic 3.5, and it allows researchers to measure variability within groups so they can attribute between-group differences to the treatments.
No, not on the AP exam. The CED defines replication as multiple experimental units per treatment group within a single study. Repeating an entire study is replicability, a different (broader) scientific concept that won't earn you the replication point on an FRQ.
They're related but not identical. Sample size is the total number of units in the study; replication is about having multiple units in each treatment group. A study of 100 people still violates replication if 99 get the treatment and 1 gets the placebo.
With only one unit per group, you can't tell whether a difference in outcomes came from the treatment or from that individual's quirks. Multiple units per group reveal natural within-group variability, which is the baseline for judging whether treatment effects are real.
State explicitly that multiple subjects are randomly assigned to each treatment group, ideally with numbers (for example, 'randomly assign 50 subjects to the walking treatment and 50 to the control'). FRQs like 2021 Q2 reward designs that clearly include all four elements: comparison, random assignment, replication, and control.