Identifying Threats to Internal Validity: Why Your Causal Claims Matter in ABA
When you introduce a new behavior intervention and a client improves, you want to be confident that your intervention caused the change—not a coincidence, a developmental shift, or something else happening in the client’s life. That confidence rests on understanding threats to internal validity.
Internal validity is the ability to attribute a behavior change to your intervention rather than to competing explanations. If you can’t rule out those alternatives, you can’t credibly claim you’ve found a functional relation.
This matters in clinical practice because misidentifying what caused a behavior change can lead you to continue ineffective interventions, withhold treatments that actually work, or recommend approaches that don’t fit your client’s real needs.
This article is for practicing BCBAs, supervisors, and clinically minded RBTs who want to strengthen their ability to detect and control threats to internal validity. We’ll cover what threats are, why they matter ethically, and how to spot and manage them using straightforward tools and documentation strategies.
What Threats to Internal Validity Really Are
Internal validity answers a straightforward question: Did my intervention cause the behavior change I’m seeing, or is something else responsible?
More formally, internal validity is the degree to which you can confidently attribute a change in the dependent variable (the behavior you’re measuring) to the independent variable (your intervention). To make that causal claim, three conditions must be met: the intervention must come before the behavior change, the intervention and behavior change must move together, and other plausible explanations must be ruled out.
A threat to internal validity is any alternative explanation for the behavior change you observe. If you claim the intervention caused the change, a threat is any other reason why that change might have happened.
These threats fall into several categories:
- History refers to external events—a schedule change, a new family member, or a medication adjustment—occurring alongside your intervention.
- Maturation is natural developmental change: a toddler becoming more coordinated, or a teenager’s impulse control improving with age.
- Testing effects occur when repeated measurement itself changes behavior.
- Instrumentation happens when your measurement method changes—a different observer, or a shift in how you define the behavior.
- Regression to the mean describes the tendency for extreme scores to move back toward average over time.
- Selection bias emerges when groups have pre-existing differences unrelated to the intervention.
- Attrition occurs when participants drop out, potentially biasing results.
- Diffusion or contamination happens when clients receive additional interventions that affect the outcome.
- Experimenter effects and placebo effects occur when clinician expectations or client beliefs—rather than the intervention itself—drive change.
Here’s a key distinction: internal validity is about causation within your case. External validity—whether your findings apply to other clients, settings, or behaviors—is separate. You can have strong internal validity and still not know if your results generalize.
Another common confusion: measurement error (unreliable instruments) differs from instrumentation threats (a change in your measurement method over time). Measurement error affects data clarity; an instrumentation threat produces an artificial shift in the data itself.
Why Identifying Threats Matters in Your Daily Practice
The stakes of getting this wrong are real.
If you misattribute a developmental gain to your intervention, you might continue an unnecessary program, consuming resources and clinician time. If you fail to notice a history threat—say, a new medication coinciding with behavior improvement—you might credit your intervention when the medication did the work.
More broadly, identifying threats protects client safety and dignity. When you can’t confidently claim your intervention caused the change, you have an ethical obligation to be transparent with caregivers and your team.
It’s the difference between saying, “The behavior improved, and we can’t yet rule out natural maturation,” and saying, “The intervention caused this change,” when the evidence doesn’t support it.
Identifying and managing threats also builds professional credibility. Clinicians, supervisors, and families trust practitioners who acknowledge the limits of their evidence. This is especially important in complex cases where multiple factors might be at play.
Key Features of Internal Validity Threats
A threat to internal validity typically has these hallmarks:
- It co-occurs with your intervention and could explain the behavior change.
- It is often time-linked: something that happens before, during, or after the intervention that creates an alternative explanation.
- It can be systematic or harder to predict.
Some threats are more likely in long-term studies. Maturation and instrumentation drift are sneaky culprits when baselines extend for many weeks. Other threats are more common in non-randomized comparisons: if you’re comparing a volunteer group to a non-volunteer group, selection bias is a major concern.
You can often spot a potential threat by examining your data graph or session notes. If behavior improves sharply before your intervention starts, that’s a red flag for history. If improvement is gradual and began weeks before intervention, maturation is suspect. If you changed how you measure behavior midway through, an instrumentation threat is present.
When and How to Identify Threats in Your Work
Threat analysis happens at critical decision points.
During baseline planning and design selection: Before choosing a reversal design or multiple baseline, ask: What threats are most likely here? Is there a history threat (the client is about to start school)? Is maturation likely (the child is in a developmental period where skills typically emerge)? Your answer should inform which design you choose.
When you see unexpected changes: If behavior improves before intervention begins, or if improvement stalls after intervention starts, investigate. Document what else was happening in the client’s life.
Before making treatment recommendations: Don’t recommend an intervention as “effective” based on a single case with uncontrolled threats. Replicate across settings or behaviors. Check your data using different conditions.
Practical documentation checkpoints include:
- A concurrent-events log tracking schedule changes, medication updates, and environmental shifts with dates
- Procedural fidelity logs showing your intervention was delivered as planned
- Notes on any measurement changes in your session records
Real-World Examples in ABA Practice
History threat in action: A child’s disruptive classroom behavior drops sharply the week after summer break begins. You’ve just started a new token system, and the data look great—until you realize the daily routine changed dramatically when school ended. Fewer transitions, more free time, and a different environment could easily explain the improvement.
To rule out this history threat, document the environmental changes, look for similar improvements in settings without the token system, and consider replicating the intervention later when the routine changes again.
Maturation threat in action: Over six months with no formal intervention, a toddler steadily improves at dressing herself. Parents are pleased, and it’s tempting to credit the informal coaching—but natural motor and cognitive development might be the real driver.
To detect maturation, look at whether improvement is gradual and smooth (typical of maturation) rather than tied to a specific intervention phase. Age-matched comparisons and staggered-baseline designs help demonstrate that change tracks with the intervention, not just time passing.
Understanding Threats Beyond the Clinic
The logic of threats to internal validity applies everywhere.
Imagine a workplace introduces new productivity software. Productivity rises—but the improvement coincides with end-of-year bonuses. Did the software cause the increase, or did the bonuses? That’s a history threat in a business setting.
Similarly, a community health program promotes walking and average daily steps increase over spring. But seasonal weather improved at the same time. Separating the program’s effect from the season’s effect requires documentation, comparison groups, or staggered rollouts.
These examples show that threat identification is a cornerstone of any credible causal claim.
Common Mistakes and How to Avoid Them
One persistent mistake is treating correlation as causation. You see a timeline: intervention starts, behavior improves. But correlation does not prove causation—you must check whether a threat offers a better explanation.
A second mistake is failing to document concurrent events. If you don’t record that the client started a new medication or switched classrooms, you can’t later ask whether that event explains the behavior change.
A third mistake is overlooking slow baseline trends. When behavior drifts gradually over weeks, it’s easy to miss. Regression to the mean or maturation often looks like a gentle slope, not a sudden shift.
To protect yourself, use a simple three-step check:
- Check timing. Did behavior change after the intervention started, or before? Is the change sudden or gradual? Does timing align with the intervention phase or some other event?
- Plan for replication. Don’t make a causal claim based on one instance. Replicate across settings, behaviors, or clients.
- Document concurrent events and fidelity. Keep a running log of environmental changes, medication adjustments, and schedule shifts. Pair this with fidelity data showing your intervention was delivered as planned.
The Ethics of Uncertainty
Here’s the hard truth: sometimes you collect good data with solid measurement and a reasonable design, and you still can’t rule out all threats. That’s not failure—it’s honesty about the limits of evidence.
When threats exist or uncertainty remains, you have an ethical responsibility to share that with caregivers and your team. You might say: “The behavior has improved since we started the intervention. However, we also changed the classroom routine last month, so we can’t yet say with certainty that the intervention caused the change. Here’s what we’re doing next to clarify.”
This transparency respects caregivers’ right to informed consent and builds trust.
Document the steps you’re taking to reduce uncertainty. If you’re extending the baseline, note why. If you’re planning a replication, record that intention. These records show professional rigor and support you if questions arise later.
Practice: Can You Spot the Threat?
Test your understanding with a few scenarios.
Scenario 1: A client’s self-injurious behavior decreases immediately after you introduce a token system. But the decrease actually began two days before implementation. Which threat is most likely?
This points to history: some event occurred just before the intervention. The temporal sequence is wrong—the intervention didn’t precede the change.
Scenario 2: Over a six-month baseline, a child’s language use increases steadily each week before any formal intervention. What threat best explains this?
Maturation is most plausible. The gradual, time-linked improvement suggests natural developmental progression.
Scenario 3: Two participants receive your intervention, but one also enrolls in speech therapy outside your control. Differences in outcomes could be due to what threat?
Diffusion or co-intervention: the extra treatment creates an uncontrolled variable.
These examples show how threat analysis is a practical skill any clinician can develop.
Bringing It All Together: Your Toolkit for Internal Validity
Identifying and managing threats doesn’t require fancy statistics. It requires three key practices:
Stable measurement: Use consistent, reliable methods. Maintain high interobserver agreement. If you change how you measure, note it prominently and interpret data carefully around that change.
Design choices: Choose a design that provides built-in protection against common threats. A multiple baseline across settings helps rule out history threats because the same historical event would have to occur simultaneously in multiple unrelated environments—unlikely.
Replication: Replicate your intervention across behaviors, settings, or clients. Each replication showing the same effect strengthens your causal claim and makes alternative explanations less plausible.
When you commit to identifying threats, you’re committing to honest, evidence-based practice. You’re saying to clients and caregivers: “I will not claim I know what caused this change unless I have good reason to believe it.”
That’s the foundation of professional credibility in ABA.
Review a current case in your practice. Look at your baseline and intervention data. Ask yourself: What threats could explain the change I’m seeing? What concurrent events happened? Did your measurement method stay the same?
Document your answers. If threats are present, plan how you’ll replicate or strengthen the design to rule them out. This reflective work is how you build internal validity into your daily practice, one case at a time.
