01 · The Question
What if the numbers and narratives tell different stories?
You expected the two components of a mixed-methods study to reinforce each other. Instead, the survey suggests one conclusion while the interviews point somewhere else.
Perhaps most respondents report high satisfaction, yet interview participants describe persistent frustration. A quantitative analysis finds no meaningful group difference, while qualitative accounts suggest that the groups experience the phenomenon very differently. Or a measured improvement appears after an intervention even though participants say little has changed in practice.
The instinct may be to decide which method is right. That is usually too quick. Disagreement can arise from methodological problems, but it can also reveal that the methods captured different constructs, people, contexts, time points, or dimensions of a phenomenon. The discrepancy itself therefore deserves analysis.
03 · What You Need to Know
Discordance can be analytically valuable
Mixed methods does not require convergence
One persistent misconception is that quantitative and qualitative findings are supposed to agree and that disagreement therefore demonstrates methodological failure.
Fetters, Curry, and Creswell describe the fit of mixed-methods integration in terms that include confirmation, expansion, and discordance. Discordance occurs when quantitative and qualitative findings are inconsistent, incongruous, conflicting, or contradictory. They explicitly discuss investigating potential sources of bias, assumptions, procedures, theory, and constructs when discordance appears.
This means that contradiction is one possible outcome of examining how one method relates to the other, not evidence that integration has necessarily failed.
First establish that the findings genuinely disagree
Before explaining a contradiction, verify that there actually is one.
Two findings may appear inconsistent because they address different questions. A survey might measure satisfaction while interviews explore frustration with a particular feature. Participants can be generally satisfied and still describe specific frustrations. There is no logical contradiction.
Likewise, a quantitative analysis might estimate an average effect while qualitative interviews reveal substantial variation among individuals. An average improvement and negative experiences among some participants can coexist.
True discordance
The findings support interpretations that cannot easily be reconciled when addressing the same relevant construct, population, context, and period.
Apparent discordance
The findings differ because the methods answer different questions, operate at different levels, or capture different aspects of the phenomenon.
Check whether the methods measured the same construct
Operationalization is a common source of apparent disagreement.
Suppose a questionnaire asks participants to rate “confidence using AI” on a five-point scale. Interviews ask participants how they verify AI-generated information, protect student data, and decide when AI use is pedagogically appropriate.
High questionnaire confidence and substantial interview uncertainty may appear contradictory. Yet the survey may capture general self-confidence while the interviews probe specific competencies. The instruments may not be measuring precisely the same thing.
Rather than asking which finding is correct, ask what each method actually operationalized.
Check who contributed each form of evidence
Sampling differences can generate discordance even when both analyses are competently conducted.
If survey respondents represent a broad population but interviewees are purposively selected because they experienced unusual difficulties, negative qualitative accounts should not be expected to reproduce the average survey pattern. Conversely, if volunteers for interviews are unusually enthusiastic, their accounts may present a more favorable picture.
Before interpreting disagreement, therefore, revisit whether the quantitative and qualitative samples were appropriately connected.
Check when the data were collected
Timing matters, particularly when the phenomenon changes.
Quantitative data collected immediately after an intervention may show high satisfaction. Interviews six months later may reveal that participants struggled to sustain the intervention. Both findings can be accurate for their respective periods.
Sequential mixed-methods designs may deliberately collect data at different stages. Researchers should distinguish temporal change from methodological contradiction.
Check the quality of each component
Discordance can also expose methodological weaknesses.
A poorly validated questionnaire may fail to capture what interview participants describe. A qualitative sample may omit cases central to the quantitative pattern. Statistical estimates may be unstable. Interview questions may be leading. Coding decisions may suppress disconfirming evidence.
When one component has serious methodological limitations, those limitations become one candidate explanation for the discrepancy. But this requires actual appraisal rather than an automatic preference for quantitative or qualitative evidence.
The appropriate question is how much the weaker component should affect the overall mixed-methods inference.
Check the level of analysis
Some apparent contradictions arise because quantitative and qualitative findings operate at different levels.
Imagine that schools with greater technology investment show higher average student engagement. Interviews reveal that some students within high-investment schools feel less engaged because particular technologies create additional cognitive or accessibility barriers.
School-level association and individual-level experience are not mutually exclusive. Treating them as direct contradictions risks an ecological or cross-level interpretive error.
Look for heterogeneity hidden by averages
Quantitative summaries can conceal subgroups and distributions. Qualitative inquiry may make that heterogeneity visible.
An intervention might produce a positive average effect while interviews reveal that it works well for some participants and poorly for others. The qualitative evidence may therefore challenge the assumption that the average effect describes a typical experience.
This is not necessarily evidence against the quantitative result. It may instead indicate that the average needs qualification.
Revisit analysis when the discrepancy is consequential
If disagreement concerns a central conclusion, additional analysis may be warranted. Fetters, Curry, and Creswell discuss several responses to discordance, including examining potential bias and methodological assumptions, reanalyzing existing data, collecting additional data, seeking theoretical explanations, and reconsidering construct validity.
The appropriate response depends on the study. Researchers might stratify quantitative analyses by a subgroup identified qualitatively, return to transcripts looking for contrasting cases, compare matched participants across datasets, or examine whether an instrument adequately represents the construct emerging from interviews.
Not every discrepancy can be resolved, and methodological archaeology can become creative if pursued long enough. Reanalysis should therefore be theoretically and methodologically justified rather than an exercise in searching until agreement appears.
Joint displays can make disagreement easier to inspect
Bringing corresponding findings into a joint display can reveal exactly where convergence and divergence occur. Researchers can align a quantitative result with relevant qualitative findings and then record the resulting integrated interpretation.
This can prevent vague statements such as “the findings were mixed.” It forces greater specificity: Which findings disagree? In what respect? Is the difference about direction, magnitude, meaning, subgroup variation, or explanation?
Such comparison is part of genuine integration rather than simply reporting two separate analyses.
Sometimes the correct conclusion is that the discrepancy remains unresolved
Researchers are not required to manufacture harmony.
If credible quantitative and qualitative evidence remain genuinely inconsistent after plausible explanations have been examined, the integrated conclusion should say so. The disagreement may indicate uncertainty requiring further research.
An unresolved contradiction is often more scientifically useful than an artificial synthesis that hides inconvenient evidence.
Watch Out
Do not resolve discordance by automatically privileging whichever component produces the more convenient conclusion. Methodological quality, relevance, sampling, measurement, timing, and inferential scope should determine how much weight each finding deserves.
07 · A Quick Checklist
Work through disagreement rather than around it
When quantitative and qualitative findings disagree, check:
Do the findings genuinely contradict one another, or do they address different questions or dimensions?
Were the same constructs operationalized in sufficiently comparable ways?
Did the components study comparable participants, cases, settings, or populations?
Were the data collected at comparable time points?
Could differences in methodological quality plausibly account for the discrepancy?
Could subgroup variation or different levels of analysis make both findings simultaneously defensible?
Have researchers re-examined relevant data, assumptions, or constructs when the disagreement is consequential?
Does the final interpretation preserve unresolved discordance rather than forcing agreement?
