What makes a good research question?

When researchers set out to research a topic, before design, analysis, or even hypothesizing, they need to hone in on one or more research questions they want to answer. A lot goes into what makes a research question *good* as opposed to just something you are curious about. I think that a good research question is informed, concise, measurable, specific, and important/relevant to answer. This blog will cover how you can consider these aspects, as well as the aspects of ethics and novelty, when you design research questions.

INFORMED

When you set out to answer a question, you want to understand what people already know about the area you are interested in. Prior findings may influence what you want to ask, and may inform how you think about the topic. It will also tell you if people have investigated your question before so you can contextualize your work, distinguish your question in some way, or choose a more novel question.

CONCISE

A good research question is also simple and focuses in on one question. In many types of common analysis (ANOVA, linear regression), you cannot predict multiple dependent variables (DVs AKA outcomes AKA the thing you are hoping to influence). So your research question should focus on influencing something that is reflected by one variable or one aggregate score derived from multiple variables. If you want to know multiple things, you can ask multiple questions and run different analyses to answer them. A concise research question would be something like “Do people prefer Coke to Pepsi?”

MEASURABLE

The only way you can answer a question is if you can measure it. So it’s important to consider whether you can reasonably answer the question you are asking. If you want to know who people voted for at the individual level without using self-report, there is no way to get that information. You want to ensure that whatever you are asking lends itself to actually being measured. Remember that many things are measurable in a self-report format (AKA asking people), but this data is less trustworthy than observable data (e.g., state records on whether people voted or not) and comes with design requirements (i.e., you have to be able to communicate with your sample directly). In our Coke vs Pepsi question, ways to measure this are clear: blind taste tests, surveys, experiments/RCTs that involve a purchase, etc.

SPECIFIC

This one goes hand in hand with conciseness. The research question should be specific enough that it lends itself to forming a hypothesis (i.e., your educated guess about what the outcome will be). There are only 3 hypotheses available for the Coke vs Pepsi question. They are either people prefer Coke [Pepsi] to Pepsi [Coke] or there is no preference. If the question is “Which soda is everyone’s favorite?” you now have way more possible outcomes available to you, which has implications for your study design. There are at least 50 types of soda at my grocery store alone!

IMPORTANT/RELEVANT

You CAN use the scientific method to answer a lot of different questions, but most researchers want to focus on the most important things to know. Those are the things that advance our knowledge in a way that feels like a significant contribution. “Do dogs prefer Coke to Pepsi” is a potential research question, and it meets most of our criteria: it’s concise, measurable, specific. But who cares? Dogs don’t drink soda and no one would find that information useful (save maybe a small group of irresponsible dog owners). If we were Coca-Cola or PepsiCo, a more important/relevant research question for us is, again, “Do people prefer Coke to Pepsi?” That would give either company actionable and important information they could use to improve or market their product.

OTHER CONSIDERATIONS

These aren’t the only things to consider when you formulate a research question. Though this isn’t as clearly connected to question quality, I would consider ethical research guidelines before moving forward with any research question. I strive to only design studies that would meet the criteria for exempt IRB review, which means only legal adult participants who are fully able to provide consent (18+ in the USA), participation carries no more risk than one would encounter in daily life, there is no chance of lasting negative impact, participants cannot be individually identified in the final dataset, and there is no active deception of participants. The level of risk participants face should be worth the potential outcome. If we are creating drugs to cure cancer, the level of risk tolerance we have for testing is going to be higher than if we want to do some market research about the products people like.

You can also consider the novelty of the question. This is a big deal in academia and for notoriety. But replication is also crucial to identifying reliable interventions. If you do ask a question that someone has already asked before, it’s just important to contextualize your results so that people understand how your investigation fits in with the general knowledge about the topic.

Put these principles to work and see if they help you level up your research questions!