Written and maintained by CASRAI Editorial Board
Last updated
The maturation threat to internal validity is the risk that participants change simply because time passes, and that this natural change is mistaken for the effect of a treatment or intervention. Harris and colleagues list it among the threats to quasi-experimental studies as “naturally occurring changes over time” that “could be confused with a treatment effect”. This guide defines the threat, separates it from neighbouring threats, works through a numeric example of how a comparison group exposes it, and explains how to rule it out in group and single-case designs.
What is the maturation threat? A definition in one paragraph
Maturation is a threat to internal validity that arises when participants change gradually over the course of a study, even in the absence of any treatment, and the researcher attributes that change to the intervention. Petursdottir and Carr describe it as changes in participant behaviour that occur gradually over time, even in the absence of treatment. “Maturation” covers more than growing up: children develop, patients recover or deteriorate, learners gain skill, and participants become tired or bored. Any process that moves the outcome along with the calendar can mask or mimic an effect. For the broader framework, see the guide to types of validity in research.
Threats to internal validity: where maturation sits
Internal validity concerns whether an observed change in the outcome can be credited to the manipulated variable rather than to something else. Maturation is one of several rival explanations. It is most dangerous when the study measures the same people at an early and a later time point and has nothing to compare them against. The comparison internal vs external validity shows why this threat is about causal inference inside the study, not about generalising beyond it.
Maturation versus history, testing and regression to the mean
These threats are often listed together because each produces a before-after change without a treatment. They differ in the process that creates the change.
| Threat | What produces the change | Typical sign |
|---|---|---|
| Maturation | Gradual change in the participants themselves as time passes | Outcome drifts steadily, including in people who receive no treatment |
| History | An external event occurring between measurements | Change begins at or after a specific event |
| Testing | Taking the pretest alters later scores | Scores improve on repeat measurement even without treatment |
| Regression to the mean | Extreme scores tend to move toward the average on re-measurement | Participants selected for extreme scores move toward the mean |
Petursdottir and Carr note that ruling out maturation requires design considerations similar to those used against history threats. The difference is conceptual: history is an outside event, maturation is a process inside the participant. See also the dictionary entries for regression to the mean and confounding variable.
Worked example: a comparison group estimates maturation
This example uses invented, illustrative numbers. It is not data from a real study. Suppose a programme aims to reduce a symptom score (higher means worse) over twelve weeks, and the researcher measures a treatment group and a comparison group before and after.
| Group | Pretest mean | Posttest mean | Change (post – pre) |
|---|---|---|---|
| Treatment | 40 | 31 | 31 – 40 = -9 |
| Comparison (no treatment) | 41 | 37 | 37 – 41 = -4 |
In a one-group pretest-posttest design, only the first row exists. The researcher would see a drop of 9 points and might credit the whole drop to the programme. The comparison row shows that people who received nothing also improved by 4 points over the same twelve weeks. That 4-point change is an estimate of the combined natural drift in the outcome, which includes maturation (and, in practice, other time-linked influences such as history and testing that affect both groups alike).
The change attributable to the programme is the difference between the two changes:
(-9) – (-4) = -9 + 4 = -5 points.
So 5 of the 9 points, not all 9, can be credited to the treatment under the assumption that both groups would have drifted equally. Reporting the raw 9-point drop would overstate the effect by 4 points, or 80 percent (4 / 5 = 0.80).
How to read the result
| What you see | What it suggests |
|---|---|
| Comparison group changes in the same direction as the treatment group | Part of the treatment-group change reflects time-linked drift, so maturation is plausible |
| Comparison group is flat; treatment group changes | Maturation is a less likely explanation for the treatment-group change |
| Comparison group changes more than the treatment group | The intervention may be ineffective or harmful relative to natural change |
| No comparison group exists | Maturation cannot be separated from the effect; the design cannot rule it out |
Assumptions and common mistakes
- Assuming the groups mature at the same rate. The subtraction above works only if the comparison group is a fair stand-in for how the treatment group would have changed without treatment. If the groups differ in age, severity or readiness, they may drift at different rates, which is a selection-by-maturation problem.
- Treating a short study as immune. Naturally occurring change can occur within a single study period; the longer the interval between measurements, the more room for it.
- Using a one-group pretest-posttest design for slow-changing outcomes. Without a comparison condition the design cannot separate change from treatment.
- Confusing maturation with history. Name the process you suspect: a gradual internal change, or an external event.
The control group guide explains how comparison conditions are built, and the comparison of experimental and quasi-experimental designs shows which designs leave maturation unaddressed.
How to rule out maturation
In group designs
Add a comparison group measured at the same times as the treatment group, as in the worked example. Assigning participants at random to the two groups makes it more reasonable to expect equal drift, because chance rather than choice distributes age, severity and readiness across groups.
In single-case designs
Petursdottir and Carr describe the safeguards used in applied behaviour analysis: run each condition until the data are stable, repeat condition changes, and use reversal or multiple-baseline designs. In a multiple-baseline design, the effect replicates across participants or settings and no change appears before treatment begins, which is difficult to explain by gradual maturation. They add that in multiple-baseline designs across participants, shared outside influences are effectively ruled out only when data are collected concurrently for all participants. They also note that some maturational influences may have an abrupt rather than gradual onset, which is why repeated condition changes matter.
How to report it (APA style example)
State the comparison group’s change alongside the treatment group’s change and report the difference. A sentence following this pattern, using the illustrative numbers above, would read: “Symptom scores fell by 9 points in the treatment group and by 4 points in the comparison group, a net difference of 5 points; the comparison group’s change indicates the extent of change expected over the same period without the intervention.” Report your test statistic, degrees of freedom, p value and confidence interval according to the APA edition your journal requires.
How to calculate the net change in Excel, R or Python
Maturation is a design threat, not a statistic, so there is no formula to compute it directly. What you can compute is the net change after subtracting the comparison group’s change. Each snippet reproduces the worked example.
Excel
Cells: B2 = 40, C2 = 31 (treatment pre, post); B3 = 41, C3 = 37 (comparison pre, post) Net change: =(C2-B2)-(C3-B3) returns -5
R
pre <- c(treatment = 40, comparison = 41) post <- c(treatment = 31, comparison = 37) change <- post - pre unname(change["treatment"] - change["comparison"]) # -5
Python
pre = {"treatment": 40, "comparison": 41}
post = {"treatment": 31, "comparison": 37}
change = {g: post[g] - pre[g] for g in pre}
print(change["treatment"] - change["comparison"]) # -5
Frequently asked questions
What is the maturation threat to internal validity?
It is the possibility that natural change in participants over time, not the treatment, explains the observed outcome. Harris and colleagues and Petursdottir and Carr, both listed under Sources, describe it in these terms.
What is an example of a maturation threat?
A twelve-week programme for a symptom that tends to ease on its own appears to work because scores fall, but a no-treatment comparison group also improves. The worked example above shows 4 of the 9 points of improvement appearing without treatment.
Does a control group eliminate the maturation threat?
A comparison group measured on the same schedule lets you estimate the natural change and subtract it, provided the groups would have changed at similar rates. Random assignment supports that assumption.
How is maturation different from history?
Maturation is gradual change within participants; history is an outside event between measurements. Both are addressed by similar design features.
Sources
- Harris AD, et al. The Use and Interpretation of Quasi-Experimental Studies in Medical Informatics. Journal of the American Medical Informatics Association, 2006. https://pmc.ncbi.nlm.nih.gov/articles/PMC1380192/ Date checked: 2026-10-29.
- Petursdottir AI, Carr JE. Applying the Taxonomy of Validity Threats from Mainstream Research Design to Single-Case Experiments in Applied Behavior Analysis. Behavior Analysis in Practice, 2018;11(3):228-240. https://pmc.ncbi.nlm.nih.gov/articles/PMC6182849/ Date checked: 2026-10-29.
