What Exposure, Intervention, or Phenomenon Was Actually Examined?

What did the researchers actually expose, deliver, observe, or explore?

A paper may claim to study “social media,” “physical activity,” “artificial intelligence,” “stress,” “online learning,” or a named treatment. Those labels identify broad concepts. They do not necessarily tell you what participants experienced or what researchers measured.

The actual study might define social-media exposure as self-reported minutes per day, physical activity through accelerometer counts, AI use as access to one particular chatbot, or an educational intervention as six instructor-led sessions delivered over eight weeks. In qualitative research, the phenomenon may be defined through the particular experience, setting, participants, and questions explored. To understand what the study was actually trying to answer, you need to move from the conceptual label to what was empirically examined.

Identify the operational version of the thing being studied

Move from the conceptual label to the empirical reality

A construct and its operationalization are not the same thing

Many research concepts cannot be observed directly in their full theoretical form. Researchers therefore operationalize them through measurements, classifications, interventions, tasks, observations, or questions.

Consider “academic engagement.” One study might operationalize it as attendance, another as a questionnaire score, another as learning-management-system activity, and another as classroom behavior coded by observers. All may plausibly study aspects of engagement, but they have not measured exactly the same thing.

When appraising a paper, keep both levels visible. The theoretical construct helps you understand the scientific question. The operationalization tells you what evidence the study actually generated.

For an exposure, ask exactly what counted as exposure

An exposure can be environmental, behavioral, social, occupational, clinical, biological, educational, or something else entirely. The label alone is rarely sufficient.

Suppose a study examines “screen time.” Was exposure measured as total daily screen time, recreational screen time, smartphone use, social-media use, nighttime use, or time on a particular platform? Was it self-reported, passively recorded by a device, extracted from administrative logs, or assigned using categories?

Then examine how exposure levels were represented. Researchers might use a continuous measure, categories such as low and high exposure, cumulative exposure, ever-versus-never status, or a threshold derived from some external criterion.

Those choices affect the question being answered.

Ask when and over what period the exposure was measured

Exposure has a temporal dimension. “Exercise” during the previous seven days is different from habitual physical activity over several years. Current occupational exposure is different from cumulative lifetime exposure.

Look for the exposure window, frequency of measurement, changes over time, and whether the measurement occurred before, during, or after the relevant outcome process.

This becomes particularly important when interpreting temporal or causal claims. A measurement collected once may be an imperfect representation of a long-term or changing exposure.

Measurement quality affects what the exposure really represents

Two studies may claim to examine the same exposure while measuring it quite differently.

A self-report questionnaire can be affected by recall, interpretation, and reporting behavior. Device-based measurements introduce their own decisions about algorithms, wear time, thresholds, missing observations, and data processing. Administrative records may capture only activities represented in the underlying system.

STROBE therefore asks authors to provide sources of data and details of assessment methods for variables of interest and, when relevant, describe comparability of assessment methods across groups.

For interventions, the intervention name is only the beginning

Knowing that participants received “cognitive behavioral therapy,” “simulation-based training,” “blended learning,” or a “digital health intervention” may still leave most of the intervention unspecified.

The TIDieR reporting framework was developed because intervention descriptions are often insufficient for replication. It asks researchers to describe why the intervention is expected to work, its materials and procedures, who provides it, how and where it is delivered, when and how much is delivered, tailoring, modifications, and intervention fidelity.

Distinguish the intervention assigned from the intervention received

In a randomized trial, assignment defines the intended contrast, but participants may not receive the intervention exactly as planned. Some miss sessions, discontinue treatment, receive additional care, cross over between groups, or receive different doses.

These facts do not automatically change the primary randomized comparison, which may appropriately estimate an effect of assignment according to the trial's estimand and analysis strategy. They do matter when you are trying to understand what participants actually experienced and when interpreting questions about adherence, implementation, per-protocol effects, or mechanisms.

Intervention fidelity can change what was actually tested

An intervention protocol may specify ten sessions, trained facilitators, individualized feedback, and weekly practice. If most participants received four sessions and feedback was inconsistently delivered, the implemented intervention differs from the idealized protocol.

TIDieR specifically includes both planned and actual fidelity because replication and interpretation require knowing not only what was intended but, where relevant, how closely delivery matched that intention.

This distinction is particularly useful for complex educational, behavioral, organizational, and public-health interventions, where implementation may vary substantially across sites or providers.

For qualitative research, identify the phenomenon rather than inventing variables

Not every study is organized around exposures and interventions. Qualitative research may investigate experiences, perceptions, practices, decision processes, interactions, meanings, or social phenomena.

Suppose researchers interview first-generation university students about their experiences seeking academic support during their first semester. The phenomenon is not simply “academic support.” It is the experience and process of seeking that support within a particular population and context.

Do not force such research into an exposure-outcome template. Instead ask what experience, process, event, practice, or meaning the researchers sought to understand, how they bounded it, and whose perspectives supplied the evidence.

Different groups may experience different versions of the exposure or intervention

Multisite and pragmatic studies deserve particular attention. The intervention called by one name may be implemented somewhat differently across schools, hospitals, communities, or providers. Complex programmes can also be locally adapted.

Such variation is not necessarily a methodological failure. Some interventions are deliberately flexible. But you need to know whether variation was planned, measured, and relevant to interpretation. Reporting frameworks such as TIDieR explicitly request information about tailoring, modifications, and fidelity for this reason.

The comparator helps define what the intervention means

An intervention effect is usually relative to something. “Treatment A works” is incomplete unless you know whether A was compared with placebo, no treatment, usual care, an active alternative, or another dose of A.

Even when your immediate task is identifying the intervention, inspect what the actual comparison was. An intervention cannot be interpreted independently of the condition against which its effect was estimated.

Do not confuse the exposure with the outcome

The distinction can become surprisingly slippery in complex models. Researchers may examine whether stress predicts sleep, whether sleep predicts stress, or whether both participate in a reciprocal process. The same variable can function differently across analyses.

Read the actual model or analytical description. Then identify what outcome was actually measured for the particular question you are evaluating.

Turning “AI-assisted learning” into the intervention that was actually tested

Common mistakes when identifying what was examined

Describe what was examined before interpreting what it means

When reading a study, try replacing the broad conceptual label with a literal description of what researchers observed, assigned, delivered, or explored. If your description becomes much narrower, that is often exactly the point.

Once you can describe the empirical exposure, intervention, or phenomenon precisely, compare it with the broader language used in the conclusions. The key question becomes whether the authors are interpreting the thing they actually studied or quietly generalizing from one operationalization to a much larger concept.

Before interpreting the findings, identify exactly what was examined

Questions about identifying what a study actually examined

The concept in the title is not necessarily the thing that was studied

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.