Quantitative Research Question Examples by Study Design

A quantitative research question has to do more than name a topic — its wording has to match the study design that will actually answer it. A question phrased as a causal claim (“does X increase Y?”) cannot honestly be answered by a design that never manipulates X, and a question phrased as a simple description (“what proportion of X?”) wastes the power of a design built to test a relationship. This guide shows what a well-formed quantitative research question looks like at each of the four major design families — descriptive, correlational, quasi-experimental, and experimental (RCT) — using illustrative example questions, not case studies drawn from any real, specific published study.

For the process of narrowing a broad topic into a question and evaluating it against the FINER criteria, see CASRAI’s guide How to Write a Research Question: From Broad Topic to FINER Criteria. This page assumes that step is done and focuses specifically on how the same underlying topic gets rewritten differently depending on which quantitative design will answer it.

Why study design determines a question’s wording

Every quantitative research question implicitly commits to a design the moment it names a relationship between variables. Four elements recur across all four design families, but which ones are present — and how they’re worded — differs by design:

• Population — who or what is being studied (see Defining the Population).
• Variable(s) — what is being measured, and whether the design treats a variable as independent (manipulated or grouping) or dependent (outcome) — see Independent vs. Dependent Variable.
• Relationship type — whether the question asks about a single characteristic, an association, a group difference, or a causal effect.
• Operational definition — how an abstract concept (e.g., “engagement,” “adherence,” “burnout”) is measured concretely enough to answer the question — see Operational Definition.

The table below shows how the same broad topic — student use of a mobile study app — gets rewritten as four structurally different, equally legitimate quantitative research questions depending on the design chosen to answer it.

Descriptive research question examples

A descriptive design measures and reports a variable, or the distribution of a variable, in a defined population — without testing a relationship between two variables or comparing groups. Descriptive quantitative questions typically use “what,” “how many,” “how often,” or “what proportion” framing, and never imply cause, effect, or association.

• What percentage of clinical trial protocols registered with a national registry report their primary outcome results within 12 months of trial completion?
• What is the average number of co-authors per publication in a given discipline over a five-year period?
• How frequently do early-career researchers report using preprint servers before formal journal submission?

Common pitfall: adding comparison or causal language (“compared with,” “leads to,” “predicts”) to a descriptive question overstates what a purely descriptive design can support — that wording belongs to one of the three design families below.

Correlational research question examples

A correlational design measures two or more variables as they naturally occur, without manipulating anything, and tests whether they are statistically associated. Correlational questions use “association,” “relationship,” “correlated with,” or “predict” language — but not causal verbs like “cause,” “increase,” or “improve,” because a correlational design alone cannot establish causation, even when a strong association is found.

• Is there a relationship between the number of data-sharing statements a journal requires and the average citation count of articles it publishes?
• Does self-reported research-integrity training hours correlate with faculty confidence in identifying research misconduct?
• Is grant application word count associated with reviewer-assigned score in a given funding program?

Common pitfall: reporting a correlational finding using causal language in the write-up (e.g., “training improves confidence”) when the question and design only support association — this is a frequent, avoidable mismatch between what a study asks and what it can honestly conclude.

Quasi-experimental research question examples

A quasi-experimental design compares groups or conditions — like an experiment — but without random assignment to those groups; group membership is determined by a pre-existing or naturally occurring factor (e.g., which institution a participant belongs to, which policy was already in place, or a self-selected choice). This design supports stronger causal inference than a correlational study but weaker inference than a true experiment, because of possible confounding from whatever determined group membership. Questions typically use comparative language (“higher/lower than,” “differ between”) applied to intact or pre-existing groups.

• Do institutions that adopted a structured data management plan (DMP) requirement before grant approval show different data-deposit rates than institutions without that requirement?
• Following the introduction of a mandatory CRediT contributorship statement at a set of journals, did the proportion of authorship disputes reported to editors change compared with the two years prior?
• Do research groups that use electronic lab notebooks show different rates of data-availability compliance than groups using paper notebooks?

Common pitfall: describing a quasi-experimental finding as if random assignment ruled out confounding — a quasi-experimental question and its eventual write-up should name the comparison groups as pre-existing, and the discussion section should address plausible alternative explanations that randomization would otherwise have controlled for.

Experimental / RCT research question examples

A true experimental design — most rigorously, a randomized controlled trial (RCT) — randomly assigns participants to conditions and manipulates the independent variable directly, which is what allows a causal claim. In clinical and health research, this design family is where the Randomized Controlled Trial (RCT) definition and the PICOT structure (Population, Intervention, Comparison, Outcome, Time) from CASRAI’s FINER/PICOT guide apply most directly. Questions use explicit causal or effect language (“does X cause/increase/reduce Y compared with Z”) because random assignment is what earns that language.

• Among early-career researchers randomly assigned to a structured mentorship program versus standard onboarding, does the mentorship program increase the rate of first-author publication within two years?
• In a randomized trial comparing a structured data management plan (DMP) template against an unstructured template, does the structured template reduce time-to-completion for a compliant DMP?
• Among patients randomly assigned to a reminder-based medication adherence intervention versus usual care, does the intervention increase adherence rate at 90 days compared with usual care?

Common pitfall: labeling a study “experimental” or “RCT” in the question when assignment to groups was not actually random — this is the same category error as the quasi-experimental pitfall above, just in the other direction, and it’s one of the more common design-labeling errors reviewers flag.

From vague to precise: worked refinements by study design

The fastest way to see how design shapes wording is to watch the same vague starting point get refined four different ways. Each pair below starts from a genuinely underspecified question a researcher might jot down early in planning, then shows the refined version for a specific design, plus exactly what the refinement changed and why the design requires it. As with every example on this page, these are illustrative composites for demonstrating structure, not questions drawn from a real published study.

Descriptive: refining a vague prevalence question

Vague: How common is research-data sharing?

Refined: What proportion of articles published in a discipline’s top-20 journals over a two-year period include a data-availability statement linking to a deposited dataset?

What changed and why: The vague version names a topic, not a measurable variable — “how common” has no bounded population, no time frame, and no operational definition of what counts as “sharing.” The refined version names the population (articles in a defined journal set, over a defined period), operationally defines the outcome (a data-availability statement linking to an actual deposit, not just a promise to share on request), and asks a single-variable proportion question with no comparison or causal claim — exactly what a descriptive design can support and nothing more.

Descriptive: refining a vague “how much” question

Vague: How much do researchers rely on AI writing tools?

Refined: What percentage of corresponding authors submitting to a set of journals in a given year disclose the use of a generative AI tool in the manuscript-preparation process?

What changed and why: “Rely on” is unmeasurable as written — it implies a judgment about degree of dependence that no single data point captures. The refined version substitutes a concrete, countable event (a disclosure statement being present) for an ambiguous behavior, and bounds the population and time frame so the study is actually feasible to run.

Correlational: refining a vague relationship question

Vague: Does mentorship help early-career researchers?

Refined: Among postdoctoral researchers who have not participated in a structured mentorship program, is self-reported mentorship access associated with first-author publication count in the first two years of appointment?

What changed and why: “Help” is a causal verb, but the vague version describes no manipulation and no comparison — nothing about how the question would actually be answered supports causal language. The refined version keeps the same underlying interest but switches to association language (“is associated with”), measures both variables as they naturally occur rather than assigning anyone to a mentorship condition, and specifies the population and outcome window precisely enough to be operationalized.

Correlational: refining a vague quality question

Vague: Is peer review quality related to reviewer experience?

Refined: Among reviewers for a given journal, is the number of prior reviews completed correlated with the average length, in words, of the substantive comments in a review report?

What changed and why: “Quality” is an abstract concept with no stated operational definition, and “related to” is fine correlational language but was attached to an unmeasurable outcome. The refined version keeps the association framing but swaps in an operationally defined proxy (comment length) for the abstract concept, and specifies both variables as things that can actually be counted from existing records.

Quasi-experimental: refining a vague policy-effect question

Vague: Did requiring ORCID iDs at submission change author behavior?

Refined: Following a journal’s introduction of a mandatory ORCID iD requirement at submission, did the proportion of submitting authors with a linked, verified ORCID record differ between the two years before the requirement and the two years after?

What changed and why: “Change author behavior” names no specific outcome and no comparison structure. The refined version names one specific, measurable outcome (proportion with a linked, verified record), defines the comparison as pre-existing time periods rather than randomly assigned groups (this is what makes it quasi-experimental, not experimental), and uses “differ between,” not a causal verb — because a naturally occurring before/after comparison cannot rule out other things that changed over the same period the way random assignment would.

Quasi-experimental: refining a vague comparison question

Vague: Are open-access articles cited more?

Refined: Do articles published open access in a given journal show a different mean citation count at 24 months post-publication than articles published closed access in the same journal and year, controlling for article type?

What changed and why: The vague version implies a causal comparison (“cited more”) without specifying that publication route was not randomly assigned — authors and editors choose or default into open or closed access, so any observed difference is confounded with whatever drove that choice. The refined version uses neutral comparison language (“different mean citation count”), fixes the outcome window, and names a control variable, all of which flag this as a quasi-experimental comparison between pre-existing groups rather than a randomized test of open access’s effect.

Experimental/RCT: refining a vague intervention question

Vague: Does training reduce research misconduct?

Refined: Among graduate students randomly assigned to a structured research-integrity training module versus a standard institutional orientation, does the structured module reduce the rate of self-reported questionable research practices at a six-month follow-up compared with standard orientation?

What changed and why: The vague version names a plausible causal claim but nothing about how it would be tested. The refined version specifies random assignment to two named conditions (what actually licenses the causal verb “reduce”), operationally defines the outcome (self-reported questionable research practices, measured at a specific follow-up point rather than left open-ended), and names the comparison condition explicitly rather than leaving “training” undefined against no stated alternative.

Experimental/RCT: refining a vague tool-effect question

Vague: Do reminder emails improve survey response rates?

Refined: Among invited survey participants randomly assigned to receive a reminder email at day seven versus no reminder email, is the final response rate at day 21 higher in the reminder-email condition than in the no-reminder condition?

What changed and why: “Improve” already sounds causal in the vague version, but nothing specifies assignment or a defined comparison, so the causal language is unearned. The refined version adds random assignment to two explicit conditions, a fixed intervention timing (day seven) and outcome measurement point (day 21), and states the comparison group plainly — the same underlying idea, now structured so the causal verb is actually supportable by the design. For research questions in clinical and health contexts specifically, CASRAI’s PICO framework guide covers the parallel structured-question format (Population, Intervention, Comparison, Outcome) used to build a question like this one from the ground up, including how PICOT adds a time element for exactly this kind of follow-up-window specification.

A quick self-check before finalizing a quantitative research question

1. Does the verb in the question (“differ,” “associate,” “predict,” “increase,” “cause”) actually match what the design can support? A design without random assignment cannot honestly support a causal verb.
2. Is every variable operationally defined clearly enough that another researcher could measure it the same way? See Operational Definition.
3. Is the population named specifically enough to bound the study? See Defining the Population.
4. For a comparison or group-difference question, is it clear whether group membership was randomly assigned (experimental) or pre-existing (quasi-experimental)? Naming this correctly in the question itself avoids a mismatch that shows up later at the analysis or write-up stage — see Experimental Design.
5. Does the planned analysis match the question? A descriptive question calls for descriptive statistics; a correlational, quasi-experimental, or experimental question calls for inferential statistics that test a relationship or group difference.

Frequently asked questions

What’s the difference between a correlational and a quasi-experimental research question?

A correlational question measures two variables as they naturally occur and asks whether they’re associated, with no groups being compared. A quasi-experimental question compares two or more pre-existing groups (not randomly assigned) on an outcome. Both fall short of a true experiment because neither involves random assignment, but a quasi-experimental design is structured around a specific comparison in a way a purely correlational design is not.

Can a quantitative research question use causal language without random assignment?

Generally, no — causal verbs (“causes,” “increases,” “reduces,” “improves”) should be reserved for designs with random assignment to conditions, because only randomization credibly rules out that some other, unmeasured factor produced the observed difference. Correlational and quasi-experimental questions should use association or comparison language (“is associated with,” “differs from”) instead.

How specific should the population be in a quantitative research question?

Specific enough that another researcher could replicate the sampling frame from the question alone — a vague population (“adults,” “researchers”) is one of the most common reasons a quantitative research question fails the Feasible and Relevant parts of the FINER criteria. See CASRAI’s Defining the Population and the FINER discussion in How to Write a Research Question.

Do these examples come from real published studies?

No. Every example question on this page is an illustrative composite, written to demonstrate correct structure for a design type. None is drawn from, or attributed to, a specific real study, institution, or researcher.

Related CASRAI resources

This page covers quantitative designs specifically. For worked examples across qualitative, mixed-methods and evidence-synthesis questions as well, see the broader research question examples bank.

• How to Write a Research Question: From Broad Topic to FINER Criteria — the process for narrowing a topic and applying FINER/PICOT, upstream of this page.
• Experimental Design
• Independent vs. Dependent Variable
• Operational Definition
• Defining the Population
• Descriptive Statistics vs. Inferential Statistics
• Randomized Controlled Trial (RCT)
• Null Hypothesis vs. Alternative Hypothesis
• Scholarly Writing cluster overview