Controlled Experiments¶

Overview¶

A controlled experiment is a structured research design in which the researcher actively manipulates one or more independent variables to observe their effects on a dependent variable, while holding other factors constant. Controlled experiments are fundamental in scientific research because they can establish causality by controlling extraneous variables.

Key Characteristics¶

Direct manipulation of variables by the researcher.
Random assignment of subjects to different experimental groups (e.g., treatment and control groups).
Controls for confounding factors, enhancing the study's internal validity.
The investigator—not the subject—decides who goes into which group.

Structure of a Controlled Experiment¶

A typical controlled experiment follows this structure:

Population
    │
    ▼
Random Assignment
    ├──────────────────┐
    ▼                  ▼
Treatment Group    Control Group
(receives            (receives
 intervention)        placebo / standard)
    │                  │
    ▼                  ▼
Measure Outcome    Measure Outcome
    │                  │
    └──────┬───────────┘
           ▼
    Compare Results

Control Group vs. Experimental Group¶

The control group receives either no treatment, a placebo, or a standard treatment. The experimental (treatment) group receives the intervention being tested. By comparing outcomes between the two groups, researchers can isolate the effect of the treatment from other influences.

Placebo and Placebo Effect¶

A placebo is an inert treatment (e.g., a sugar pill) that looks identical to the real treatment. It is given to the control group to account for the placebo effect—the phenomenon where patients experience real improvements simply because they believe they are receiving treatment. Without a placebo control, it would be impossible to distinguish genuine treatment effects from psychological effects of receiving care.

Why Controlled Experiments Minimize Confounding¶

In observational studies, confounding is a major concern because the researcher has no control over group assignment. In a controlled experiment, randomization ensures that potential confounders are distributed roughly equally across groups. This means any observed difference in outcomes can be attributed to the treatment rather than to pre-existing differences between groups.

	Controlled Experiment	Observational Study
Who chooses groups?	Investigators	Subjects
Confounding	Minimized	Many potential confounders
Causation	Can establish	Can only identify association

Example: Enriched vs. Deprived Environments¶

UC Berkeley psychology researchers divided rats into two groups. Rats raised in enriched environments with diverse experiences developed heavier, denser brain cortices suited for higher-order activity and greater resilience, compared to rats raised in deprived environments. This is a controlled experiment because the researchers assigned rats to the two conditions and manipulated the environment.

Limitations of Controlled Experiments¶

While controlled experiments are the gold standard for causal inference, they have practical limitations:

Ethical constraints: It may be unethical to assign humans to harmful conditions (e.g., forcing subjects to smoke).
Artificiality: Laboratory conditions may not reflect real-world behavior.
Cost and time: Large-scale experiments with human subjects can be extremely expensive and time-consuming.
Generalizability: Results obtained under tightly controlled conditions may not generalize to broader populations.

When controlled experiments are not feasible, researchers fall back on observational studies combined with statistical adjustment techniques to approximate causal conclusions.

Key Takeaways¶

Controlled experiments allow researchers to establish cause-and-effect relationships by manipulating variables and using random assignment.
The control group and placebo are essential for isolating the treatment effect.
Randomization distributes potential confounders evenly across groups, minimizing bias.
When ethical or practical constraints prevent experimentation, observational methods with appropriate statistical controls are used instead.