Joint Displays for Mixed-Methods Integration

When a manuscript reports a quantitative result in one paragraph and a qualitative theme in another, with no sentence connecting the two, reviewers cannot tell whether the study actually integrated its data or just collected two kinds of data side by side. Bryman’s 2007 survey of published mixed methods research found exactly this pattern was the norm, not the exception: researchers frequently report each strand separately and never demonstrate integration in the final account. The GRAMMS reporting criteria (O’Cathain, Murphy & Nicholl, 2008) responded to that gap by requiring authors to state explicitly how quantitative and qualitative components were integrated — not just that both were used. A joint display is the concrete artifact that answers that requirement on the page.

What a joint display actually is

A joint display is a table (occasionally a figure) that presents quantitative and qualitative findings from the same study side by side around a shared organizing structure — usually a set of research questions, outcome domains, or cases — with one additional column stating the meta-inference: what the researcher concluded by combining the two strands that neither strand supported on its own.

Fetters, Curry & Creswell’s widely cited integration framework (“Achieving Integration in Mixed Methods Designs — Principles and Practices,” Health Services Research 48(6pt2), 2013, pp. 2134-2156) describes integration as happening at three possible levels: design, methods, and interpretation/reporting. A joint display is the standard interpretation/reporting-level integration technique in that framework — it is where integration becomes visible to a reader, regardless of which design (convergent, explanatory sequential, exploratory sequential) produced the data.

The meta-inference column is what separates a joint display from a plain comparison table. A row that simply repeats “quant said X, qual said Y” with no synthesis is not yet a joint display in the sense the methodology literature means — it is a side-by-side listing. The meta-inference is the sentence that says what X and Y mean together.

Why it matters for getting published

Reviewers trained in mixed methods (and journals with GRAMMS-influenced author guidelines) increasingly look for a joint display, or an equivalent explicit integration statement, as evidence the “mixed methods” label in the title is substantive rather than descriptive. A manuscript that reports a quantitative results section and a qualitative findings section with no integration step invites exactly the critique Bryman documented: two studies bound together, not one integrated one. A well-constructed joint display is the fastest way to pre-empt that critique, because it puts the integration work in front of the reviewer rather than asking them to infer it from separated prose.

The recognized joint display types

Guetterman, Fetters & Creswell’s applied review (“Integrating Quantitative and Qualitative Results in Health Science Mixed Methods Research Through Joint Displays,” Annals of Family Medicine 13(6), 2015, pp. 554-561) catalogued six recurring joint display types across published health-science mixed methods studies, noting that many published displays combine more than one type. Four of the six generalize well beyond health-science research and cover most manuscripts:

Side-by-side joint display. Quantitative results and qualitative themes run in adjacent columns, typically organized by research question or outcome domain. The simplest structure and the most common starting point, especially for a convergent design where both strands were collected in parallel.
Statistics-by-themes joint display. Qualitative themes form the rows; quantitative values (means, proportions, or categorical summary scores) sit in the columns next to each theme. Useful when the qualitative coding structure is the more natural organizing frame — for example, when themes emerged first and the quantitative data were collected to test their prevalence.
Comparative matrix joint display with a meta-inference column. The format this guide walks through below: an explicit column, separate from both the quantitative and qualitative columns, stating whether the two strands converge, diverge, or leave a domain where one strand is silent — followed by the meta-inference itself. This is the format most directly aligned with GRAMMS’s integration-reporting requirement, because the convergence/divergence judgment is made visible rather than implied.
Cross-case comparison joint display. Rows are individual participants or cases rather than themes or domains; columns show each case’s quantitative score alongside a qualitative summary or illustrative quote. Common in embedded and multiple-case mixed methods designs, where the unit of integration is the case, not the sample as a whole.

The remaining two types in Guetterman et al.’s typology — mapping qualitatively derived dimensions onto quantitative instrument items during instrument development, and layering a theoretical or conceptual lens across the display — are more specialized and worth knowing exist, but apply to a narrower set of manuscripts than the four above.

Building a joint display, step by step

Choose the organizing rows before touching either dataset. Rows are usually the study’s research questions, its outcome domains, or (for cross-case displays) its individual cases — never an arbitrary list assembled after the fact. Picking the rows first keeps the table from silently steering toward whichever strand happens to have tidier data.
Fill the quantitative column with final, analyzed results — a mean difference, an effect size, a proportion — not raw data. State the statistic and, where relevant, its uncertainty (a p-value, a confidence interval, or an effect size), not just a directional summary.
Fill the qualitative column with condensed themes, generally one or two sentences per row plus a short illustrative quotation if the write-up allows it. This is the point where interview or focus-group coding work (see coding qualitative interview data) gets distilled down to what actually belongs in a summary table, not reproduced in full.
Add an explicit convergence/divergence/silence judgment per row. Convergent rows are where both strands point the same direction; divergent rows are where they genuinely disagree or qualify each other; silent rows are where one strand — usually the quantitative instrument — never asked about something the other strand surfaced. All three are legitimate, reportable outcomes.
Write one meta-inference sentence per row, stating what combining the two columns lets you conclude that neither column supports alone. A meta-inference that only restates the quantitative result is a sign the qualitative column wasn’t actually used in drawing it.
Resist the pull toward a tidy table. If two-thirds of a study’s real findings converge and one-third diverge or go unmeasured by one strand, the table should show that mix. Editing wording until every row reads as agreement misrepresents the study more than an uneven table does — and is the exact failure GRAMMS and Bryman’s critique were written to catch.

Worked example: a convergent-design evaluation (illustrative composite)

Illustrative composite — not a real study. The table below uses a simulated dataset built specifically to demonstrate joint display construction: the quantitative values come from a seeded, reproducible random simulation (mulberry32 PRNG, fixed seed, Box-Muller normal draws), not from a real evaluation or any named institution. The qualitative themes and quotations are illustrative composites in the same sense as CASRAI’s institutional case studies — synthesized to be plausible, not attributed to any real participant, program, or organization. Treat the numbers as an accurate demonstration of how the statistics in a real joint display are computed and reported, not as evidence about any actual training program.

Scenario: a research-data-management (RDM) training evaluation comparing a structured hands-on workshop (intervention, n=32) against existing self-directed institutional documentation alone (comparison, n=32), using a five-point self-efficacy scale (“I feel confident writing a data management plan”) measured before and after, plus semi-structured interviews with a subset of each group. This is a convergent design: both strands were collected over the same period and integrated at analysis.

Quantitative summary (change score = post − pre, 1–5 Likert): intervention group mean change +0.91 (SD 0.89, pre-mean 2.69 → post-mean 3.59); comparison group mean change +0.19 (SD 0.82, pre-mean 2.56 → post-mean 2.75). Welch’s t = 3.35, df = 61.6, two-tailed p ≈ 0.001; Cohen’s d (between-group change score) = 0.84 — a statistically significant, large between-group effect.

Domain	Quantitative result	Qualitative finding	Convergence	Meta-inference
Change in self-efficacy score	Intervention +0.91 (SD 0.89) vs. comparison +0.19 (SD 0.82); Welch t=3.35, df=61.6, p≈0.001, d=0.84	Intervention participants attributed the change to practicing on their own dataset, not to the content itself	Convergent	The hands-on, applied format — not exposure to RDM content alone — appears to drive the measured gain; content delivered without structured practice produced a real but much smaller effect.
Confidence to act independently	Post-workshop scores clustered at 4/5	Several participants scoring 4/5 said they would still want a data librarian to check their DMP before submitting it	Divergent (partial)	A higher self-reported efficacy score did not translate into full self-reported independence — the scale measured comfort with the ideas, not readiness to act without support.
Time as a barrier	Not measured — the pre/post scale did not ask about time or workload	Time away from lab work was the most frequently mentioned obstacle to applying what was learned	Silent (quantitative)	The instrument measured confidence, not the practical conditions that convert confidence into a completed plan — a gap visible only because interviews were included.
Effect of documentation alone	Comparison group change +0.19 (SD 0.82) — small but non-zero	Comparison group participants described consulting the institutional DMP template on their own, without asking anyone for help	Convergent (small effect)	Documentation without facilitation produces a measurable but modest effect, consistent with the comparison group’s smaller quantitative gain — the two strands agree on both the direction and the size of that difference.

Three of the four rows converge; one diverges and one shows the quantitative strand silent on something the qualitative strand raised unprompted. That mix — not uniform agreement — is what a genuine integration exercise usually produces, and Fetters, Curry & Creswell’s own framework treats divergent and silent rows as substantive findings, not table-construction errors to be smoothed away.

Common mistakes that undercut a joint display

No meta-inference column at all. A table with only a quantitative and a qualitative column, however well organized, has not integrated anything — it has arranged two separate results next to each other. This is the single most common way a “mixed methods” manuscript fails GRAMMS’s integration requirement despite technically including a table.
Forcing convergence. Rewording a qualitative theme until it appears to agree with a quantitative result, when the honest reading is more mixed, misrepresents the underlying data and is the exact pattern Bryman’s 2007 review flagged as undermining integration claims.
A meta-inference that only restates the quantitative column. If the meta-inference sentence would be identical with the qualitative column deleted, the qualitative data did not actually inform the conclusion in that row.
Including every measure collected instead of the rows that matter. A joint display with fifteen rows because every survey item and every code got its own line is unreadable and buries the two or three integration points a reviewer actually needs to see.

Where it goes in the manuscript

Most published joint displays sit in the Results section, immediately after the quantitative and qualitative findings have each been reported in enough detail that the table’s shorthand makes sense; some journals and some authors place it in the Discussion instead, using it explicitly as the interpretive synthesis step. Either placement is defensible — what matters, per the GRAMMS criteria, is that the integration step exists and is visible as a distinct part of the manuscript rather than left for the reader to construct.

Frequently asked questions

Do I need a joint display for every mixed methods manuscript?

Not by any formal mandate, but its absence is one of the first things a reviewer familiar with GRAMMS or the broader mixed methods integration literature will notice. For a convergent or explanatory/exploratory sequential design with genuinely separate quantitative and qualitative findings to reconcile, a joint display (or a clearly equivalent narrative integration section) is close to the default expectation in 2026.

How many rows should a joint display have?

Enough to cover the study’s actual research questions or outcome domains, not every individual measure or code. Most published examples run four to eight rows; a table with fifteen or more rows usually means the organizing structure needs revisiting rather than that the study had unusually rich data.

What is the difference between a joint display and a triangulation protocol table?

A triangulation table typically documents which data sources or methods addressed each question, as a design-and-methods-level record. A joint display goes a step further: it presents the actual results side by side and adds the meta-inference — the interpretation-level integration Fetters, Curry & Creswell describe as a distinct level from triangulation’s methods-level integration.

Can a joint display be a figure instead of a table?

Yes. Visual joint displays (for example, a path diagram annotated with illustrative quotations, or a timeline for a sequential design showing where qualitative findings shaped the next quantitative phase) are recognized variants in the methodology literature, though the table format remains the most common and the easiest for reviewers to scan quickly.

What software do researchers use to build one?

Most joint displays are still assembled manually in a word processor or spreadsheet after the quantitative analysis and the qualitative coding are both complete — the display is a synthesis step, not an automated output. Qualitative coding itself is commonly done first in dedicated software; see NVivo or the comparison of NVivo, ATLAS.ti, and MAXQDA for the coding step that feeds the qualitative column.

Joint Displays: Presenting Mixed-Methods Integration

Ask about Joint Displays: Presenting Mixed-Methods Integration