Skip to content

Research Design Taxonomy — Quantitative/Qualitative and Descriptive/Correlational/Experimental

Theory · All theories
Evidence · 8 claims (3 for, 4 mixed, 1 unmarked) · 8 studies (3 associational, 2 review, 2 theoretical, 1 quant-synthesis), q1–q4 · 1 of 8 report an effect size · 8 claims rest on one study

Description

Quantitative research typically starts with a focused research question or hypothesis, collects a small amount of data from each of a large number of individuals, describes the resulting data with statistical techniques, and draws general conclusions about a population. It is comparatively weak at generating novel, interesting research questions or providing rich description of behavior in particular situations. Qualitative research does the reverse: it can generate new questions and hypotheses, collects large amounts of relatively unfiltered data from a relatively small number of individuals, describes that data with non-statistical techniques, and can produce rich, detailed description of behavior in the real-world contexts where it occurs — what qualitative researchers call the "lived experience" of participants. Because the two approaches have complementary strengths, some researchers combine them into mixed-methods research, either by using qualitative work to generate hypotheses that quantitative work then tests, or through triangulation — using both approaches simultaneously on the same question and comparing results. Convergent results reinforce and enrich each other; divergent results raise a genuinely useful further question: why do they diverge, and how can that be reconciled?

A separate dimension classifies research by what kind of conclusion the design can support, regardless of whether the underlying data is quantitative or qualitative. Descriptive research measures and reports a variable without testing any relationship between variables — it can answer interesting and important questions, but not questions about relationships. Correlational research goes a step further and formally tests whether a relationship exists between two or more variables, expressed as a correlation coefficient (r) from −1 to +1: the closer to 1 in either direction, the stronger and more predictable the relationship. Correlations have genuine predictive value — a university admissions committee, for instance, can use the correlation between current students' standardized test scores and their college GPA to help predict applicants' likely success. But correlation is limited in a specific way: establishing that a relationship exists tells us little about cause and effect. A correlation between two variables can arise because one variable is a confounding variable driving both — the classic textbook case is ice cream sales correlating with crime rates, where temperature drives both rather than either causing the other — or because people accept a causal story (do healthier people eat cereal, or does eating cereal make you healthier?) without checking for a confound.

People also perceive relationships between variables that don't actually exist at all — illusory correlations. The persistent belief that a full moon affects human behavior is a textbook case: a meta-analysis of nearly 40 studies found no such relationship (Rotton & Kelly, 1985), yet the belief persists because people notice and remember odd behavior during a full moon while failing to notice that odd behavior occurs at a constant rate throughout the lunar cycle — the same confirmation-bias mechanism (see Evidence-Based Teaching and Scientific Reasoning) that makes personal inquiry generally unreliable. Illusory correlations are not merely a curiosity: research suggests they are involved in forming the prejudicial attitudes toward particular groups that can escalate into discriminatory behavior (Fiedler, 2004).

Experimental research goes further still, randomly assigning participants to conditions so that hypothesis testing can support inferences about causal relationships — see Designing a Valid Experiment for the mechanics. Descriptive, correlational, and experimental research form a genuine hierarchy of what a design can conclude, not a hierarchy of rigor or value: a descriptive study of what a phenomenon even looks like is often the necessary first step before a correlational study is worth running, which in turn can motivate a specific causal hypothesis worth testing experimentally.

Implications

Context

Requirements

  • Matching the research question actually being asked to a design that can support the kind of conclusion it needs — a descriptive or correlational question does not need (and cannot be answered by mis-reading) an experimental design, and vice versa

Constraints

  • A strong correlation, however large r is, never by itself establishes which variable (if either) causes the other, or whether a third variable causes both
  • Readers as well as researchers are vulnerable to illusory correlations — noticing confirming instances and failing to notice the (much larger) set of disconfirming ones

Target Learners

  • Anyone reading, citing, or designing educational research — including this wiki's own claim pages, whose evidence-quality (q) ratings track exactly this design hierarchy (see CLAUDE.md's evidence-quality tiers)

Target Learning Objectives

  • Reading a correlational finding (in a study, a news article, or a claim page) without silently upgrading it to a causal one

Claims

The evidence by question

  • Can a correlational or other non-randomised result be read as evidence that a design choice causes the outcome? The RCT is considered the gold standard of research evidence because randomization ensures only the intervention could cause treatment-control outcome differences [+M]
  • Does a study's design (between- or within-subject, laboratory or field, more or less rigorous) change the size of the effect it reports? Variation in meta-analytic effect sizes stems from synthesists' methodological choices (inclusion criteria, data extraction, formulas, outlier handling) and from study quality [+W]

Examples

Key Sources

  • Rotton, J., & Kelly, I. W. (1985). Much ado about the full moon: A meta-analysis of lunar-lunacy research. Psychological Bulletin, 97(2), 286-306. doi:10.1037/0033-2909.97.2.286
  • Fiedler, K. (2004). Illusory correlation. In R. F. Pohl (Ed.), Cognitive illusions: A handbook on fallacies and biases in thinking, judgement and memory (pp. 97-114). Psychology Press.
  • Arduini-Van Hoose, N. (2020). Quantitative and qualitative approaches to research; Descriptive research; Correlational research. In Educational psychology. Retrieved from https://edpsych.pressbooks.sunycreate.cloud. CC BY-NC-SA 4.0.
Page metadata
Field Value
type theory
title Research Design Taxonomy — Quantitative/Qualitative and Descriptive/Correlational/Experimental
description Two overlapping ways of classifying a research study — quantitative vs. qualitative (or mixed) by data type, and descriptive vs. correlational vs. experimental by what conclusion the design can actually support — determine what a study can and can't tell you, independent of how well it was executed.
status review
canonical True
generated by claude/unspecified on 2026-08-29
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.