Explanatory Research: Check Evidence and Rival Explanations

Explanatory research asks why or how a pattern occurs. It looks for reasons and processes behind what you observe. That aim can guide a study, but it does not make a plausible story a proven cause.

Start by putting each proposed reason beside the evidence for it, a rival account and the check still missing. The fictional workshop example below shows how this changes a research question.

Atlas can help compare supplied studies and hypotheses; you inspect the sources and decide what the design can support.

What explanatory research tries to explain

A count tells you how many people came. An explanatory question asks why so few came. Was it the time, a missed notice, a barrier to entry, or something else?

The JIBC methods chapter explains how this purpose differs from describing a pattern or finding out what might matter.

An account of why something happens should show a link you can check. For example, the workshop time may clash with work and stop people who want to come.

You need to check both parts: whether staff want to come, and whether the clash stops them. The label “timing” alone does not show how that would work.

The Open University's teaching guide treats explanatory work as connecting ideas to understand why things happen. Write those connections plainly so a reader can see where the evidence is strong and where you are making an inference.

Separate three research purposes

One project can serve all three purposes, but each asks a different question. Exploratory work asks what might matter when you know little about the problem. Descriptive work records what happens and to whom.

Explanatory work asks why or how it happens. Scribbr's guide uses why and how questions to frame that last purpose.

For the workshop, early interviews might reveal barriers you had missed. Session counts might show when few people came. To explain that pattern, you would check whether a barrier fits the facts and what else could fit them. A list of barriers gives you a place to start.

“Explanatory sequential” has a more specific meaning in mixed methods research. Quantitative results come first, followed by qualitative work to help explain them. The NIH-commissioned mixed methods guide describes this sequence. It is one design option, not a synonym for every explanatory study.

Make the proposed explanation checkable

Define the target

Define who, where, when and what you will study before choosing a method. The broad question “Why is attendance low?” leaves those choices open.

You can narrow it to ask how work hours affect whether invited staff come to weekday workshops this term. This names the group, the period and the link you want to check.

State what you expect

Set out what you would expect to find if your proposed reason is sound. In this fictional exercise, hypothesis H1 says work-hour clashes may stop staff who want to come.

You would look for staff accounts of those clashes and records that show the times overlap. Then check whether the clash affected their choice to come.

Keep a rival account

Keep a rival account beside H1. H2 says some invited staff may never see the notice, so they could miss the session even when its time suits them.

A conceptual framework can set out the links you propose. Keep clear which links are part of your model and which you have checked in the sources.

Define what would weaken each account as well. If people knew about a session, had time available and still chose not to attend, timing alone would not explain their choice. Turn the needed checks into research objectives before asking whether interviews, records or comparisons can answer them.

A worked explanation comparison

Read the evidence map

This packet is fictional, with invented documents and locators for the teaching example. Record R has session totals, while interview I contains accounts from invited staff who came.

Notice N describes how staff were invited. Record C covers a later session, giving you a chance to check what else changed.

Each row keeps the observed item separate from the account it might support and the work still needed.

Table 1: The fictional table maps a source item to a proposed account, an inference limit and a next check. It does not rank or confirm the causes of low attendance.

Revise the claim

The last row changes what you can claim. A quick summary might say the new time caused more people to come. But the packet changed three things at once: time, topic and reminder.

The new topic or the reminder could explain the larger count. Keep those rivals in the note as you review H1.

Version 2 of the note might read: “The account at I, turn 6, suggests a time clash can be a barrier. The later session does not show the effect of time alone because topic and reminder also changed. We still need to check whether staff saw the notice.” This keeps the useful clue and what you still need to learn.

Choose evidence that separates the accounts

Choose the next step by asking what would help you tell H1 and H2 apart. More session totals could show the pattern more clearly, but not whether staff saw the notice.

You might learn about both issues from people who missed the workshop. Plan how to reach them, seek consent and protect their data before doing so.

Leave room for other reasons in your interview questions. Ask what happened from the time the person saw the notice to their choice about coming. First check that they saw it; a time clash has a different meaning if they never knew the session existed. Ask about interest too, since you cannot assume everyone wanted the workshop.

If you compare counts, define what differs between the groups and how you measure who comes. Day and evening sessions leave several reasons open when topic, staff and reminders also differ.

You may need a trial or a planned study of existing data to ask about a cause. Igelström's methods glossary explains how the groups you compare and the assumptions you justify affect that claim.

Qualitative work can explore how and why things happened in context, without measuring an effect across a whole population. If you need to test a sequence within one case, process tracing offers a focused route. Choose it when you need to check steps within a case, with each step tied to a source.

Check what a causal claim requires

A causal claim asks what would change if you did something different. For the workshop, you might ask what would happen if you changed the time, with the other conditions defined. Seeing time and attendance vary together leaves a shared cause open. Igelström and colleagues show this other path in their methods diagram.

The excerpt below shows C affecting both A and Y. This shared cause can shape the condition you study and the outcome, giving you another path to check.

Igelström and colleagues (2022), Figure 1, journal page 962, show confounding as a common cause. The row is cropped under CC BY 4.0, with text and diagram unchanged. The workshop materials above are separate fictional examples.

For the workshop, workload might affect both the time chosen and whether staff come. That would make it a possible confounding variable in a study of timing. The diagram shows a type of link to look for. You would need evidence from this setting to say workload had that role.

You can study causes through observed data with a sound design and assumptions you can justify. Adding controls at random does not secure that claim. Explain which groups you compare, how people enter them and how you measure the outcome. The original glossary introduces these requirements; you still need to address bias in your own study.

Compare supplied explanations in Atlas

Use Atlas to build the comparison note from sources you are allowed to upload. Add background studies, draft hypotheses and relevant method guidance through Add a source and Upload files.

Wait for processing readiness and check that the needed pages are readable. Keep names and other sensitive details out when permission or handling rules require it.

1. Mention the selected sources with @. Ask for a table separating each proposed reason, supporting passage, rival account and missing check. Restrict the answer to that source set.
2. Open every numbered citation and read the surrounding passage. Check who was studied, what was measured and whether the source reports a result or proposes an explanation. Precise source positions depend on the information available.
3. Correct the table. A paper about another setting may support the idea behind H1 without showing that H1 explains your workshop. Mark that difference and remove causal wording the design does not support.
4. Create a note through New and Note. Save the checked comparison, hypothesis version, source locators and next checks; confirm Saved after editing.

For the fictional packet, explicitly ask whether C, page 2, separates timing from topic and reminder. If the answer treats the larger count as proof, reject that inference and keep the changed conditions in the note. Atlas supports the source comparison; you decide what further study is needed.

Report the conclusion at its strength

Match your wording to what you have checked. In a planning note you might write, “Timing is a proposed barrier.” If you have checked the interviews, you can say which staff described a time clash. To say changing the time caused more people to come, you need a design that supports that effect claim.

Keep the question, proposed reason, source evidence, rivals and missing checks together. A later reader can then review the account when new evidence arrives. The QuestionPro planning guide links questions with methods and the report. Keep your final claim within what your study can support.