03 · What You Need to Know
A Conceptual Framework Is a Starting Proposition, Not a Contract With the Data
Frameworks Can Contain Expectations That Turn Out Not to Hold
A conceptual framework may be carefully grounded in theory and previous empirical research while still containing relationships that are uncertain.
This is not a defect. Research often examines propositions precisely because their applicability, strength, mechanisms, or boundary conditions remain unresolved.
If previous evidence suggested that institutional support should be associated with faculty adoption of generative AI, representing that relationship may have been entirely defensible before data collection. Finding little evidence for it in your sample does not retroactively make the original rationale irrational.
It does mean that the new evidence needs to be taken seriously.
“Not Supported” Does Not Necessarily Mean “Disproved”
Interpret the result at the level your study permits.
A nonsignificant statistical result does not necessarily demonstrate that a relationship is exactly zero. Estimates have uncertainty, and the study may have limited statistical precision, imperfect measurement, restricted variation, or other methodological constraints.
Likewise, failing to observe an expected process in qualitative data does not automatically establish that the process never occurs. Sampling, setting, data-generation methods, and analytical choices affect what the study can reveal.
Not supported in this study
The evidence generated by the present design did not provide the expected support for the proposed relationship or explanation.
Demonstrated to be false generally
A much stronger conclusion requiring evidence capable of ruling out the proposition across the conditions to which the claim applies.
Preserving this distinction prevents both defensive interpretation and overreaction.
First Check the Evidence Before Rebuilding the Framework
An unexpected finding should prompt scrutiny, but not a fishing expedition for a more convenient answer.
Check whether the constructs were measured appropriately, whether the data meet relevant quality standards, whether coding and analytical procedures were implemented correctly, whether important assumptions were violated, and whether the sample or setting differs meaningfully from those in previous studies.
If a coding error or measurement problem explains the discrepancy, the issue may lie in the evidence rather than the conceptual proposition.
But methodological checking should not become repeated analysis until the expected result appears. The purpose is to assess the credibility of the evidence, not negotiate with it.
Unexpected Findings May Reveal Boundary Conditions
A relationship can be plausible in some contexts and weak in others.
Suppose previous studies suggest that perceived usefulness predicts technology adoption. Your study finds that usefulness matters little among instructors who lack institutional permission or reliable access to the technology.
The original relationship may not need complete rejection. The evidence may suggest a boundary condition: perceived usefulness becomes consequential only when particular enabling conditions exist.
Context can therefore explain why a relationship fails to generalize uniformly. This is one reason contextual factors may deserve conceptual attention rather than being treated merely as background description.
The Missing Relationship May Point to an Omitted Concept
Sometimes the original framework is incomplete rather than fundamentally mistaken.
Imagine that AI self-efficacy was expected to predict responsible AI use, but interviews reveal that students with high confidence use AI in very different ways depending on their understanding of academic integrity expectations.
The findings may suggest that an omitted concept, such as integrity beliefs or policy understanding, helps explain why the original relationship was inconsistent.
This can motivate a revised framework.
The important qualification is timing. A concept discovered after examining the findings should be identified as an evidence-informed revision, not quietly inserted into the original framework as though it had always been there.
A Relationship May Need to Be Removed, Reversed, or Reinterpreted
Different findings imply different kinds of conceptual revision.
| Finding |
Possible Framework Implication |
Caution |
| Expected relationship receives little support |
Qualify or possibly remove the relationship |
Consider uncertainty and methodological limitations before concluding that no relationship exists. |
| Relationship appears only under certain conditions |
Add or clarify a boundary condition or moderator |
Distinguish prespecified analysis from post hoc interpretation. |
| Unexpected concept appears important |
Consider extending the framework |
Do not pretend the concept was part of the original model. |
| Direction differs from expectation |
Reconsider directional assumptions |
Ensure the research design can actually establish temporal or causal direction. |
| Qualitative evidence reveals a different process |
Reorganize the conceptual relationships |
Preserve the distinction between the initial framework and the framework generated or refined through analysis. |
Do Not Redraw the Original Framework and Erase the History
Suppose the original model contained:
A → B → C
Your findings instead suggest:
A → C
with little evidence that B performs the expected intermediary role.
It can be useful to present a revised framework reflecting the findings. What becomes problematic is presenting the revised version as though it were the framework that guided the study from the beginning.
The difference between an a priori framework and a framework revised after seeing the evidence matters for interpretation.
Watch Out
Do not retrospectively redesign the framework until it perfectly mirrors the results and then present that model as the study's original expectation. Doing so obscures what was hypothesized beforehand and what was learned from the evidence.
Null and Unexpected Findings Can Be Theoretically Informative
A study does not become valuable only when the arrows survive.
An unsupported relationship may challenge assumptions about generalizability. It may reveal that a construct operates differently in a new population, that a mechanism depends on context, or that previous evidence was less robust than assumed.
Unexpected findings can also expose conceptual ambiguity. Perhaps two constructs thought to be distinct overlap substantially, or perhaps the assumed direction between them was never theoretically secure.
The research contribution may therefore lie partly in explaining why the original framework needs qualification.
Do Not Build an Elaborate Explanation for Every Null Result
There is an opposite danger: treating every unexpected result as evidence of a fascinating new mechanism.
Sometimes the data are simply inconclusive.
If an estimate is imprecise, the sample is small, measurement reliability is limited, or qualitative evidence is sparse, the responsible conclusion may be that the study cannot determine whether the proposed relationship holds.
A restrained interpretation is often more credible than inventing an elaborate post hoc theory to explain noise.
Exploratory Follow-Up Is Useful When It Is Labeled Honestly
Unexpected results often generate new questions. Additional analyses may help investigate plausible explanations.
That is legitimate exploratory work.
The key is to distinguish analyses or interpretations motivated by the observed data from those specified before examining the findings. Transparency allows readers to judge the evidentiary status of each conclusion.
In quantitative research, practices such as preregistration can help preserve the distinction between confirmatory and exploratory analysis. In qualitative and iterative traditions, the relationship between emerging evidence and evolving conceptualization may be expected to be more recursive, but the researcher should still document how interpretations developed.
Framework Revision Should Follow the Methodological Logic of the Study
Not every methodology treats conceptual frameworks as fixed at the same stage.
A confirmatory study may specify a framework before data collection and evaluate whether evidence supports its propositions. An exploratory qualitative study may begin with sensitizing concepts and refine them substantially during analysis. Design-based or longitudinal research may intentionally revise conceptual understanding across cycles.
Accordingly, whether and when a conceptual framework can change during the research depends partly on what role the framework performs in the methodology.
Keep the Original Question in View
When results challenge a framework, researchers can become preoccupied with rescuing or replacing the model.
Return to the research question.
The purpose of the study was not to protect the diagram. The framework was a tool for investigating the phenomenon. If the evidence suggests the phenomenon behaves differently from the initial conceptualization, that difference may be exactly what the research needed to reveal.
