Mixed Methods Research

What makes a study “mixed methods” research

Mixed methods research is a methodology in which a researcher deliberately collects and analyzes both quantitative and qualitative data within a single study and integrates the two at one or more defined points in the design — not a study that merely happens to contain a survey and some interviews. Johnson, Onwuegbuzie, and Turner’s widely cited definitional synthesis describes it as “a systematic integration of quantitative and qualitative methods in a single study for purposes of obtaining a fuller picture and deeper understanding of a phenomenon,” combining elements of both approaches “for the broad purposes of breadth and depth of understanding and corroboration.”

Three things distinguish genuine mixed methods research from a study that simply uses two methods:

• Both a quantitative and a qualitative strand — not two qualitative methods (e.g., interviews plus document analysis) or two quantitative methods (e.g., a survey plus secondary dataset analysis), which the methods literature classes as multi-method rather than mixed methods designs.
• A deliberate point of integration — the two strands are connected through merging the data, building one strand’s instrument or sample from the other’s results, or embedding one within the other, rather than collected and reported independently.
• A stated rationale for combining methods — typically triangulation (cross-validating a finding two ways), complementarity (each method covering what the other cannot), or development (one strand’s results shaping the other’s design) — rather than “we had time to do both.”

For how mixed methods studies differ from purely qualitative or purely quantitative designs on data type, sample logic, and rigor criteria, see Qualitative Research vs. Quantitative Research: Key Differences and the Qualitative Research dictionary term — this page assumes that distinction and focuses on how the two are combined.

The three core mixed methods designs

Creswell and Plano Clark’s framework, the field’s most widely adopted typology, identifies three core designs distinguished by the timing of data collection and where integration occurs. The shorthand QUAN and qual (or QUAL and quan) indicates relative priority: capitals mark the dominant or primary strand.

• Convergent design — quantitative and qualitative data are collected during roughly the same phase, analyzed separately using each strand’s own conventions, and then merged during interpretation to compare, contrast, or combine what each strand found. Used to cross-validate a finding or produce a more complete picture than either strand alone.
• Explanatory sequential design (QUAN → qual) — a quantitative phase runs first; its results (often an unexpected finding, an outlier subgroup, or a result that needs context) determine the sample and questions for a qualitative phase that follows, to help explain the quantitative results.
• Exploratory sequential design (QUAL → quan) — a qualitative phase runs first, typically because no adequate instrument, variable list, or theory yet exists for the population or phenomenon; its findings are then used to build or adapt a quantitative instrument, which is tested and generalized in a second phase.

Creswell and Plano Clark also describe more advanced variants built on these three — an embedded design (one strand nested inside a larger design built around the other, such as qualitative interviews embedded in a trial), a transformative design (an equity- or advocacy-oriented framework wrapped around any of the core designs), and a multiphase design (multiple sequential mixed methods phases feeding a larger program of research) — but the three core designs above are the starting point for identifying which applies to a given study.

Why researchers use mixed methods

The rationale is almost always some combination of the following, and a well-designed study states which one applies rather than leaving the reader to infer it:

• Triangulation — using quantitative and qualitative data as independent checks on the same question, so agreement across methods strengthens confidence in a finding and disagreement flags something worth investigating further.
• Complementary strengths — quantitative methods typically offer breadth: a large sample, statistical generalizability, and the ability to test a defined hypothesis. Qualitative methods typically offer depth: contextualized, meaning-focused understanding of how or why a pattern exists. Combining them answers questions neither alone can — for example, quantifying how common an outcome is while also explaining the mechanism behind it.
• Development and instrument-building — using one strand’s results to construct or refine the other strand’s data-collection tools, most commonly the exploratory sequential pattern above (qualitative findings shape a survey instrument later tested at scale).

Two worked examples

Example 1: A convergent design evaluating a research-training program

An evaluator studying a research-training program administers a validated skills survey to all participants (the quantitative strand) and, in the same period, conducts semi-structured interviews with a subset of participants about their experience (the qualitative strand). Each strand is analyzed on its own terms — the survey with descriptive statistics and pre/post comparisons, the interviews with inductive thematic coding — and only at the interpretation stage are the two sets of results merged into a joint display comparing what the numbers showed against what participants described. That merging step is what makes this mixed methods rather than two independent studies that happen to share a topic.

Example 2: An explanatory sequential design following up a survey anomaly

A survey of grant applicants finds that one demographic subgroup reports significantly lower satisfaction with a funder’s review process, but the survey data cannot say why. The researcher then conducts a qualitative phase — focus groups or interviews specifically with that subgroup — designed around the quantitative result, to explain the mechanism behind it. The qualitative sample, questions, and analysis are all built directly from the first phase’s findings, which is the defining feature of an explanatory sequential design.

Counter-example: two methods without integration is not mixed methods

A study collects a large quantitative dataset to answer one research question and, separately, conducts qualitative interviews to answer a different, unrelated research question, then reports each in its own section of the paper with no point where the two are compared, merged, or used to inform one another. Even though both a quantitative and a qualitative strand are present, this is not mixed methods research as the field defines it — it is two studies bound together in one manuscript. The methodological literature on this gap is substantial: Bryman’s widely cited analysis found that researchers frequently write up each strand separately and never actually achieve integration in the final account, and identified concrete barriers — different intended audiences, one method being the team’s clear preference, disciplinary and epistemological divides, and journals that favor one method over the other. This is why methods guidance (see the next section) treats integration, not the mere presence of two data types, as the defining test.

Why integration is the hard part

Declaring a study “mixed methods” is easy; actually integrating the two strands is the part the methodological literature keeps finding weak in practice. Fetters, Curry, and Creswell’s widely used framework for achieving integration describes three levels at which it can happen — at the design level (how the study is structured, per the three core designs above), at the methods level (how data collection and analysis procedures connect, such as building a survey from interview findings), and at the interpretation and reporting level (through a joint display, a narrative weaving of both strands’ results, or explicit data transformation, e.g., converting qualitative themes into counts). A study that only manages the first level — correctly labeling itself convergent or sequential — without doing real work at the methods or reporting level has not actually integrated its findings, whatever the design diagram says.

Reporting guidance built specifically to close this gap, such as the Good Reporting of a Mixed Methods Study (GRAMMS) criteria, asks authors to state explicitly what the mixed methods design was, how each component was conducted, and — the criterion most often missing in practice — how the components were actually integrated and what the integration itself contributed to the study’s conclusions. For a research office reviewing a mixed methods protocol, grant application, or manuscript, that last question is the practical test: if the qualitative and quantitative results could be lifted out into two unconnected papers with nothing lost, the study has not achieved genuine integration, regardless of how it is labeled.

Related CASRAI terms

• Qualitative Research — the operational definition of the qualitative strand mixed methods designs draw on.
• Qualitative Research vs. Quantitative Research: Key Differences — the underlying data, sample-logic, and rigor distinctions this page builds on.

References

• Creswell, J. W., & Plano Clark, V. L. Designing and Conducting Mixed Methods Research (3rd ed., 2018), SAGE.
• Johnson, R. B., Onwuegbuzie, A. J., & Turner, L. A. “Toward a Definition of Mixed Methods Research.” Journal of Mixed Methods Research, 1(2), 2007, 112–133.
• Fetters, M. D., Curry, L. A., & Creswell, J. W. “Achieving Integration in Mixed Methods Designs—Principles and Practices.” Health Services Research, 48(6pt2), 2013, 2134–2156.
• O’Cathain, A., Murphy, E., & Nicholl, J. “The Quality of Mixed Methods Studies in Health Services Research.” Journal of Health Services Research & Policy, 13(2), 2008, 92–98 — source of the GRAMMS reporting criteria.
• Bryman, A. “Barriers to Integrating Quantitative and Qualitative Research.” Journal of Mixed Methods Research, 1(1), 2007, 8–22.