Does Your Question Require Evidence About Why Something Happened Rather Than Merely Whether It Happened?

Do You Need to Know Whether Something Happened or Why It Happened?

Suppose a university introduces an AI-supported tutoring program. After one semester, students assigned to the program achieve higher examination scores than students receiving conventional instruction.

A carefully designed study might establish that the tutoring program improved performance. But the result leaves another question unanswered: Why did the program produce that improvement?

Perhaps students received more individualized practice. Perhaps immediate feedback helped them recognize errors. Alternatively, the program may have encouraged students to spend more time studying, regardless of the instructional quality of its feedback.

These explanations are not equivalent. Each suggests a different process and may have different implications for improving or transferring the intervention.

The distinction between establishing whether something happened and explaining why it happened is fundamental to research design. Both questions can be valuable, but they do not necessarily require the same evidence.

Why Questions Require Explanatory Evidence Beyond Establishing an Outcome

How Questions About Occurrence Differ From Questions About Explanation

What Does It Mean to Ask Whether Something Happened?

A whether question asks researchers to establish the occurrence, presence, or existence of a phenomenon, relationship, change, or effect.

For example:

• Did students' examination scores improve during the semester?
• Did teachers adopt the new instructional platform?
• Was student participation higher after the policy change?
• Did the intervention improve performance compared with conventional instruction?

These questions appear similar because they ask whether something occurred. However, they involve different evidentiary demands.

Determining whether scores increased requires comparable measurements across time. Determining whether teachers adopted a platform requires a defensible definition and measurement of adoption.

Determining whether an intervention caused higher performance requires more. It involves estimating what would have happened under an appropriate alternative condition.

Therefore, a whether question is not necessarily descriptive. Some whether questions require causal inference, while others can be answered through observation or measurement alone.

What Does It Mean to Ask Why Something Happened?

A why question seeks an explanation of an observed phenomenon. It asks researchers to identify factors, conditions, processes, or causes that account for an outcome.

For example:

• Why did students' examination scores improve?
• Why did some teachers adopt the platform while others did not?
• Why did participation increase after the policy change?
• Why did the intervention improve performance in one institution but not another?

These questions may require evidence about different kinds of explanation.

A researcher investigating why teachers adopted a platform might examine perceived usefulness, institutional expectations, technical support, and professional judgments.

A researcher investigating why an intervention improved learning might examine the activities and cognitive processes connecting the intervention to the outcome.

A researcher investigating why effects differed between institutions might examine contextual conditions or differences in implementation.

The word "why" does not prescribe a particular method. It signals the need to clarify what kind of explanation the researcher intends to develop or test.

How Do Descriptive, Causal-Effect, and Explanatory Questions Differ?

One important distinction is between documenting an outcome, estimating a causal effect, and explaining the processes responsible for it.

These purposes can overlap within one study. However, the fact that a study answers one question does not mean it has answered the others.

A randomized experiment may estimate an intervention's effect without identifying the process responsible for it. A detailed qualitative investigation may explain how participants experienced a process without estimating its average causal effect.

Neither contribution should be dismissed merely because it addresses a different inferential purpose.

Why Does Establishing an Effect Not Automatically Explain Its Cause?

In causal inference, researchers often distinguish between estimating the effect of an intervention and investigating the mechanism through which the effect occurs.

Suppose random assignment establishes that an AI-supported tutoring program improves mathematics performance compared with conventional instruction.

The study may provide credible evidence that assignment to the program caused an average improvement under the observed conditions.

But the intervention may contain several components: adaptive exercises, immediate feedback, additional practice opportunities, and progress monitoring.

The overall effect does not reveal which components were responsible, whether they interacted, or which processes connected them to the outcome.

To answer those questions, researchers need additional evidence or designs capable of distinguishing competing explanations.

For example, a factorial experiment might manipulate intervention components separately when such manipulation is feasible and theoretically meaningful. A process evaluation might examine how participants interacted with the intervention and how those interactions related to the outcomes.

The distinction is important because establishing effectiveness and explaining the source of effectiveness are related but separate inferential achievements.

Are All Why Questions Causal Questions?

Not necessarily. In ordinary research language, "why" can refer to several kinds of inquiry.

Some why questions ask about causal processes. Others concern participants' reasons, intentions, interpretations, or the theoretical meaning of an event.

For example, asking why a teacher selected a particular instructional strategy might mean investigating the teacher's stated professional reasoning. Asking why the strategy improved learning concerns its causal consequences.

These questions may be related, but the evidence needed to answer them differs.

Interviews may provide direct evidence about participants' accounts of their decisions. However, reported reasons should not automatically be treated as complete explanations of behavior or as proof that particular conditions caused an outcome.

Researchers should therefore clarify whether the intended explanation concerns subjective meaning, reasons for action, causal effects, generative mechanisms, or some combination of these.

What Evidence Is Needed to Explain a Causal Mechanism?

Mechanism-focused explanations examine the processes connecting proposed causes with outcomes.

Relevant evidence may include temporal sequences, intermediate outcomes, observed behavior, participant accounts, institutional records, and comparisons that help distinguish alternative pathways.

Suppose a researcher proposes that automated feedback improves writing because students become more capable of identifying weaknesses in their arguments.

Evidence might include feedback records, successive drafts, measures of students' ability to evaluate arguments, and observations of revision practices.

However, simply observing that students who revise more also achieve higher scores would not establish the proposed mechanism. Motivation, prior ability, instructor support, or other factors could influence both revision behavior and performance.

Statistical mediation analysis may contribute to mechanism investigation, but causal mediation claims require assumptions that extend beyond the significance of an indirect effect.

Qualitative process tracing may examine the sequence of events and evaluate whether observations are consistent with competing causal explanations.

The appropriate strategy depends on the mechanism evidence required by the research question.

What If the Explanation Concerns Context Rather Than a Specific Mechanism?

Sometimes researchers are less interested in the internal process through which an intervention operates than in the circumstances that make its effects possible or variable.

For example, an educational program might produce different outcomes in institutions with different levels of instructional support.

The researcher may ask whether institutional support modifies the program's effect or whether particular organizational arrangements help explain differences in implementation.

These questions require contextual evidence and an appropriate comparison across conditions.

However, observing that high-support institutions have better outcomes does not establish that support caused the difference. Institutions may differ in student characteristics, resources, leadership practices, or other relevant factors.

Researchers should therefore distinguish contextual description from causal explanation.

When the intended conclusion concerns why findings vary across settings, evidence about context may be necessary alongside effect estimates or process observations.

Can Implementation Evidence Help Explain Why an Intervention Did Not Work?

Yes. Implementation evidence can help distinguish an intervention that was delivered as intended from one that was incompletely adopted, substantially modified, or inconsistently received.

Suppose an intervention evaluation finds no meaningful improvement in student learning.

If implementation records show that most participants received only a small fraction of the intended activities, the findings may indicate that the intervention was not delivered sufficiently to evaluate its intended operation under those conditions.

Nevertheless, poor implementation does not prove that the intervention would have been effective if delivered differently.

Likewise, high implementation fidelity combined with disappointing outcomes does not establish that every underlying theoretical assumption was incorrect.

Implementation evidence clarifies what occurred during delivery and can support explanations for observed results. Stronger causal claims may require additional comparison, measurement, or experimental manipulation.

This is why evidence about implementation is often valuable when interpreting an intervention's apparent success or failure.

Can Quantitative Research Answer Why Questions?

Yes. Quantitative research can investigate explanatory hypotheses using experimental manipulation, longitudinal data, causal models, mediation analysis, and other methods appropriate to the causal question.

For example, a researcher may experimentally vary feedback timing to examine whether immediate feedback produces better learning outcomes than delayed feedback.

A longitudinal study might investigate whether changes in a proposed intermediate variable precede changes in the outcome.

However, numerical relationships do not become causal explanations merely because they are statistically significant.

Researchers must consider temporal ordering, confounding, measurement validity, model assumptions, and alternative explanations.

A quantitative explanation can be credible when the research design supports the relevant causal interpretation. It need not include qualitative interviews simply because the research question begins with "why."

Can Qualitative Research Answer Why Questions?

Yes. Qualitative research can investigate reasons, interpretations, decisions, organizational processes, and causal sequences.

For example, interviews and observations might reveal how teachers interpreted a new assessment policy and why they changed particular instructional practices.

Case-oriented approaches may investigate the sequence of events through which an institutional decision produced subsequent changes.

The explanatory strength of qualitative evidence depends on the research question, case selection, quality of observations, and treatment of alternative explanations.

Researchers should avoid assuming that collecting participants' explanations automatically establishes the causes of their behavior. Accounts may be incomplete, retrospective, or influenced by how questions are asked.

Nevertheless, well-designed qualitative inquiry can provide explanatory evidence that would be difficult to obtain from outcome measurements alone.

When Are Mixed Methods Useful for Answering Both Questions?

Mixed methods may be appropriate when the study genuinely requires different but complementary forms of evidence.

For example, a randomized evaluation could estimate whether an intervention improved learning, while interviews and observations investigate how participants engaged with its activities.

An explanatory sequential design might use quantitative results to identify patterns requiring further investigation, followed by qualitative inquiry designed to examine plausible explanations.

However, the qualitative component must be capable of contributing meaningful explanatory evidence. Interviews conducted only to obtain favorable participant quotations do not establish why an intervention worked.

Similarly, adding a quantitative survey to a qualitative study does not automatically strengthen causal inference.

The decision to combine methods should follow the specific evidentiary requirements of the research question rather than an assumption that two methods are necessarily better than one.

Why Must Competing Explanations Be Considered?

A proposed explanation becomes more credible when researchers examine whether alternative explanations could also account for the observations.

Suppose student participation increases after a university introduces an attendance incentive.

The researcher might hypothesize that the incentive increased participation by making attendance more rewarding.

But participation might also have increased because instructors introduced more engaging activities, course requirements changed, or the student population differed from the previous semester.

A study that documents the increase without examining these possibilities may establish a descriptive pattern but not the proposed causal explanation.

Different designs address competing explanations in different ways. Randomization can reduce baseline confounding in expectation. Quasi-experimental approaches rely on design-specific assumptions. Process tracing evaluates whether observed evidence supports one causal account more strongly than others.

The aim is not necessarily to eliminate every imaginable explanation. It is to address the plausible alternatives that materially threaten the intended conclusion.

Does Explaining Why Something Happened Require Knowing Every Cause?

No. Many outcomes arise through multiple interacting conditions and processes. A research study may provide a useful explanation without identifying every contributing factor.

For example, a study might establish that a particular feedback strategy improved students' writing by encouraging substantive revision under specified conditions. Other influences, including prior knowledge and instructor support, may also contribute to writing development.

The explanation can remain valuable if the researcher clearly defines its scope and acknowledges relevant limitations.

However, researchers should avoid presenting one supported contributing process as the complete explanation of a complex outcome.

Explanatory completeness is not an all-or-nothing property. The necessary level of explanation depends on the question, theoretical framework, available evidence, and intended claim.

When Is Answering Whether Enough?

A study does not become inadequate simply because it does not explain every observed result.

Some decisions primarily require knowing whether an intervention produces a sufficiently important effect. A credible estimate of that effect may be valuable even when the mechanism remains uncertain.

Similarly, establishing whether a phenomenon occurs, how frequently it occurs, or whether its prevalence changes can be an important research contribution.

Researchers should not expand a focused study into a broad explanatory investigation unless the additional evidence is necessary for the intended purpose.

When an explanatory claim is not required, collecting extensive mechanism or contextual evidence may increase complexity without materially improving the answer.

From Establishing an Intervention's Effect to Explaining How It Worked

Suppose the randomized comparison produces an estimated improvement of six points on a 100-point writing assessment, with appropriate statistical analysis supporting a positive average effect of assignment.

The researcher also observes that intervention participants made more substantive revisions and frequently described using feedback to reconsider their arguments.

These findings would be consistent with the proposed explanation. However, they would not automatically establish that substantive revision mediated the intervention's causal effect.

Students' revision behavior may be influenced by other factors, and the additional explanatory evidence may not identify every relevant causal pathway.

The study could therefore support two conclusions of different strengths: a causal estimate of the intervention's effect, subject to the trial's validity, and evidence consistent with a particular mechanism that remains open to further testing.

This distinction preserves the contribution of both components without requiring the explanatory evidence to establish more than it can support.

Common Mistakes When Moving From Whether to Why

Decide Whether Your Research Needs an Outcome, a Causal Effect, or an Explanation

Before selecting your design, write the central claim you want to make at the end of the study.

If the claim is that an outcome occurred, determine how to observe and measure it accurately. If the claim is that an intervention caused the outcome, determine how to construct a credible causal comparison.

If the claim explains why the outcome occurred, identify which explanatory propositions must be investigated and what observations could distinguish them from alternatives.

One useful test is to ask whether the planned evidence would allow you to distinguish your preferred explanation from another plausible explanation.

If several explanations remain equally consistent with the evidence, your study may still offer valuable findings. However, the explanation should be presented as provisional rather than established.

Conversely, if the intended decision requires only a reliable estimate of whether an intervention made a difference, a focused causal evaluation may be more defensible than an ambitious but underdeveloped explanatory component.

The essential task is to work backward from the claim you intend to make and determine what the research design must be capable of establishing.

Before Choosing Your Evidence, Check What the Question Actually Asks

Frequently Asked Questions About Why and Whether Research Questions

Knowing That Something Happened Is Not the Same as Knowing Why

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.