Research questions · Lesson 3 of 3

Constructs, variables and operational definitions

Name what you will record, and you have turned an argument about ideas into something checkable.

8 min read · Free — no account needed to read

A construct is an idea you cannot observe directly: wellbeing, trust, engagement, productivity, addiction, quality. Everybody agrees these exist and nobody can point at one.

A variable is something that varies and that you can record: hours slept, messages sent, revenue, a score on a questionnaire.

An operational definition is the bridge: the specific, stated way you will measure a construct in this piece of work. "Engagement" becomes "number of sessions per week lasting over two minutes". The definition is not the construct and does not have to be perfect — it has to be stated, so that a reader knows exactly what your finding is about.

Almost every dispute about research findings is a dispute about operational definitions. Two studies of "productivity" reach opposite conclusions because one measured lines of code and the other measured features delivered. Neither is lying, and the disagreement is not really about productivity at all.

This is also where reading gets sharp. When a paper claims something about a construct, the useful question is: what did they actually record? Often the gap between the recorded variable and the claimed construct is the whole story.

See it happen

A study reports that "a four-day week increased employee wellbeing".

The read that arrives on its own

Shorter week, happier staff — that is intuitive and the study confirms it.

The same situation, with this idea

You look for the operational definition. Wellbeing was measured as the average of three self-report items collected in the final week of the trial. That is a legitimate measure and it is also a narrow one: it captures how people felt near the end of a pilot they knew was being evaluated. The finding is real and it is about that measure, not about wellbeing in general.

After this, you can

read any claim about a construct by asking what was actually recorded.

The takeaway

Name what you will record, and you have turned an argument about ideas into something checkable.

Easy to take the wrong way

That a narrow operational definition invalidates a study. It bounds it. The error is only made when the conclusion is stated in terms of the construct as though the bound did not exist.

Try this

Take the central construct in your own question and write three different ways it could be operationally defined. Note how the finding would differ under each.