Can Participants Serve as Their Own Controls in a Research Study?

Can You Compare Participants With Themselves Instead of Recruiting a Separate Control Group?

Suppose you want to determine whether a new learning intervention improves students' performance. Rather than assigning one group to the intervention and another to a control condition, you measure the same students before and after they receive it.

Each student now has two observations. One represents performance before the intervention, and the other represents performance afterward. Does this mean the students have served as their own controls?

In a broad descriptive sense, yes. Their earlier observations provide a reference for their later outcomes. But whether this constitutes a credible controlled comparison depends on what the researcher wants to estimate and what might have changed between measurements.

The distinction matters because within-person comparisons can remove certain differences between participants while remaining vulnerable to changes over time, previous exposure, and other influences unrelated to the intervention.

Yes, but a Within-Person Comparison Is Not Automatically a Valid Causal Control

How Participants Can Function as Their Own Comparison Group

What Does It Mean to Serve as Your Own Control?

In a conventional between-subjects comparison, outcomes are compared between different individuals or groups. In a within-person comparison, observations from the same individual under different conditions or periods are contrasted.

For example, a researcher might compare a participant's reaction time under two interface designs, blood pressure during different treatment periods, or writing performance before and after an instructional intervention.

The participant provides information about more than one condition. This reduces the need to estimate the comparison entirely from other individuals who may differ in important characteristics.

However, the participant cannot ordinarily experience both treatment conditions at exactly the same time under identical circumstances. The researcher must therefore consider whether differences in timing, sequence, and exposure could affect the outcome.

Which Research Designs Use Participants as Their Own Controls?

Several designs use within-person comparisons, but they differ in how the conditions are assigned, when outcomes are measured, and what causal assumptions are required.

These designs should not be treated as interchangeable. In particular, an uncontrolled pretest-posttest comparison is not methodologically equivalent to a randomized crossover experiment simply because both collect repeated observations from the same people.

What Participant Differences Does Self-Control Remove?

Within-person comparisons can eliminate confounding from characteristics that remain constant across the relevant observation periods.

These might include relatively stable genetic characteristics, early-life experiences, or enduring individual attributes. In educational research, stable aspects of a student's prior preparation or background may be accounted for through within-person comparison, although knowledge and ability themselves can change.

Consider two students who differ substantially in baseline writing proficiency. If each student completes comparable tasks under two feedback conditions, a within-person analysis can focus on changes or differences within each student rather than relying entirely on comparisons between the two students.

This can improve statistical efficiency when repeated observations from the same participant are strongly correlated and the design's assumptions are satisfied.

Nevertheless, self-control does not eliminate time-varying confounding. Motivation, health, workload, practice, and exposure to other interventions may change between observations.

Why Is a Before-and-After Comparison Not Always Enough?

A pretest-posttest design compares outcomes measured before an intervention with outcomes measured afterward. The difficulty is that time and intervention exposure change together.

Suppose students improve their research-writing scores over a semester after using a new feedback platform. They may also have attended lectures, completed additional assignments, received faculty feedback, and gained experience through repeated practice.

The observed improvement cannot automatically be attributed to the platform.

Other threats include maturation, historical events, testing effects, changes in measurement instruments, and regression to the mean.

Researchers should therefore distinguish the descriptive question of whether participants changed from the causal question of whether the intervention produced that change.

How Does a Randomized Crossover Trial Strengthen the Comparison?

In a crossover trial, participants receive different treatments or conditions during separate periods, often in a randomized sequence.

For example, some participants may receive Treatment A followed by Treatment B, while others receive Treatment B followed by Treatment A.

Randomizing the sequence helps distribute certain period-related influences across treatment orders. Each participant contributes observations under both conditions, allowing treatment comparisons to account for stable participant differences.

However, randomization does not eliminate carryover, period effects, or treatment-by-period interactions. These must be considered during design and analysis.

Crossover designs are most suitable when the condition is sufficiently stable, outcomes respond within an appropriate period, and treatment effects are reversible or otherwise compatible with repeated treatment exposure.

When an intervention causes lasting learning or an irreversible physiological change, the limitations of crossover designs for irreversible interventions become especially important.

What Are Order Effects?

Order effects occur when a participant's response depends partly on the sequence in which conditions are experienced.

Practice is a common example. Participants may perform better during a second task because they have become familiar with the procedure, regardless of which experimental condition is presented.

Fatigue can produce the opposite pattern. Participants may respond more slowly or less accurately during later conditions because of sustained effort.

Expectations and adaptation can also influence responses.

Researchers may use randomization, counterbalancing of condition order, suitable rest intervals, or other design features to reduce or evaluate these effects.

The effectiveness of these strategies depends on the mechanism involved. Counterbalancing does not necessarily remove a persistent effect of the first intervention.

What Is a Carryover Effect?

A carryover effect occurs when exposure to an earlier condition continues to influence outcomes during a later condition.

For example, a participant who learns a new problem-solving strategy during the first experimental condition may continue using it during the second condition.

In pharmacological research, a treatment may remain biologically active after administration has stopped.

Carryover can distort the interpretation of treatment differences because outcomes during the second period may reflect both the current and previous interventions.

A washout period may reduce carryover when the effect dissipates over time. However, the appropriate washout duration depends on the intervention and outcome.

For learned skills, behavioral changes, or irreversible treatments, waiting may not restore the original condition.

Can Participants Serve as Their Own Controls Without an Intervention?

Yes. Some observational designs compare different exposure periods within the same individual.

A case-crossover study is particularly relevant when investigating whether a transient exposure is associated with an acute event. The individual's exposure during a period shortly before the event is compared with exposure during appropriately selected control periods.

A self-controlled case series examines event incidence during defined exposure periods and comparison periods among individuals who experienced the event.

These approaches can control for stable individual characteristics, but they have distinct assumptions. Exposure trends, age effects, event-dependent exposure, and event-dependent observation may threaten validity.

Researchers should select the specific self-controlled design according to the exposure process, outcome, and intended causal contrast rather than treating self-control as a complete research design in itself.

When Is a Within-Person Comparison Statistically Efficient?

Repeated observations from the same individual are often correlated. When that correlation is positive, comparing paired observations can reduce variability associated with stable differences between participants.

For two measurements, the variance of the within-person difference depends on the variability of each measurement and their covariance.

The calculation illustrates why positive within-person correlation can improve precision. It does not imply that repeated-measures designs are always more efficient. Carryover, missing observations, additional study periods, and the chosen outcome model can offset or complicate this advantage.

How Should Repeated Observations Be Analyzed?

Because observations from the same participant are not independent, analysis should account for their within-person association.

For simple paired measurements, a paired-samples analysis may be appropriate when its assumptions are satisfied.

More complex designs may require repeated-measures models, mixed-effects models, generalized estimating equations, conditional regression, or methods tailored to specific self-controlled designs.

The appropriate analysis depends on the outcome distribution, number of periods, treatment assignment, missingness, and the scientific question.

Researchers should not analyze repeated observations as if each represented an independent participant. Doing so can produce incorrect standard errors and misleading inferences.

When Should Participants Not Serve as Their Own Controls?

Within-person control may be unsuitable when the intervention permanently changes the participant, the underlying condition changes rapidly or unpredictably, or repeated measurement substantially alters the outcome.

It may also be inappropriate when outcomes can occur only once, treatment sequences cannot be ethically assigned, or substantial dropout is expected between conditions.

For example, comparing a student's performance before and after completing an entire degree program does not create a reversible treatment comparison. The student cannot return to the same pre-program state to experience a genuinely alternative educational pathway.

In such situations, a parallel-group or other appropriately designed comparison may be more defensible.

Comparing Two Feedback Interfaces Within the Same Students

Consider five illustrative participants from the study:

The design allows each student's response under Interface A to be compared with their response under Interface B. Stable differences in general academic ability are therefore less likely to distort the comparison.

However, the researchers still need to investigate whether task order, practice, fatigue, or learning from the first feedback interface influenced performance under the second.

They should analyze the complete sample using a method that accounts for pairing and relevant design factors. If the first interface teaches strategies that persist into the second task, the treatment contrast may be contaminated by carryover.

The central advantage is not simply that fewer participants may be required. It is that the design can estimate a within-person difference while controlling for stable participant characteristics, provided the conditions are sufficiently comparable and sequence effects are appropriately addressed.

Misconceptions About Using Participants as Their Own Controls

Deciding Whether Self-Control Is Appropriate for Your Research Question

Begin by identifying the outcome and asking whether it can be observed meaningfully under more than one condition within the same participant.

Then determine whether exposure to the first condition changes the participant in ways that could affect later observations. This question is particularly important in educational interventions because learning often persists beyond the period in which it occurs.

The broader decision between between-subjects and within-subjects comparisons should account for feasibility, causal validity, statistical efficiency, and the nature of the intervention.

Do not choose self-control solely because recruiting a separate comparison group is difficult. The design must still produce an interpretable contrast.

Before Using Participants as Their Own Controls, Check These Conditions

Questions About Self-Controlled and Within-Subjects Designs

Self-Control Removes Some Differences, Not Every Alternative Explanation

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.