03 · What You Need to Know
A Research Gap Is a Limitation in Knowledge, Not Merely an Empty Search Result
Begin with what the literature already establishes
You cannot identify the boundary of knowledge without first identifying the knowledge on one side of that boundary.
Suppose evidence consistently supports a short-term intervention effect among university students. The unanswered question may not be whether the intervention “works.” It might instead concern durability, transfer, differential effects across populations, implementation at scale, harms, or the mechanism producing the observed benefit.
Good gap identification therefore follows synthesis rather than preceding it. You first establish the strongest defensible conclusion, then ask which consequential questions remain outside it.
Known
A conclusion the current body of evidence can support with an appropriate degree of confidence.
Unknown
A relevant proposition the available evidence cannot yet determine with sufficient confidence.
Do not equate “few studies” with “important gap”
A topic may have only two studies because the question is narrow, trivial, obsolete, or already answerable through adjacent evidence. Conversely, a topic may contain hundreds of publications while leaving an important question unresolved because the studies repeatedly use the same limited design or outcome.
The number of publications therefore does not determine the importance of a gap.
Ask instead what decision, theory, explanation, or substantive understanding remains constrained by the missing knowledge.
One type of gap is simply missing evidence
Sometimes an important question genuinely has not been investigated adequately. Perhaps no studies examine long-term outcomes, a consequential population has been neglected, or harms are rarely measured.
Even here, describe the gap precisely. “There is little research on rural students” is less informative than “Evidence about the intervention's effect on retention comes almost entirely from urban universities, leaving its effectiveness in geographically isolated institutions uncertain.”
Another gap is evidence that exists but is too imprecise
Research may address the correct question yet remain unable to answer it because estimates are too uncertain.
Imagine several small studies whose confidence intervals remain compatible with substantial benefit, little effect, and meaningful harm. The gap is not the total absence of studies. It is insufficient precision to distinguish among consequential possibilities.
Cochrane treats imprecision as one of the domains affecting certainty and emphasizes that wide confidence intervals can leave important effects unresolved.
Inconsistent findings create a different kind of unanswered question
If credible studies produce substantially different results, the gap may concern the conditions responsible for that variation.
Perhaps an intervention works in some contexts but not others. Perhaps outcome measurement changes the apparent effect. Perhaps implementation intensity matters. If the evidence cannot yet determine which explanation is correct, the unresolved question concerns effect modification or context rather than simply whether the intervention “works.”
This is more useful than concluding that “results are mixed and more research is needed.”
Indirect evidence leaves questions about applicability
A literature may provide extensive evidence while still failing to answer the question for the population, context, intervention, or outcome you care about.
For example, dozens of studies may investigate an educational technology among university students in high-income countries. That does not automatically answer how the same approach performs in resource-constrained secondary schools.
Cochrane's discussion of indirectness emphasizes the need to consider how well the evidence matches the population, intervention, comparator, and outcomes to which conclusions are being applied.
Methodological repetition can preserve a gap despite abundant research
Imagine twenty cross-sectional studies consistently reporting an association between academic confidence and use of a learning platform. A twenty-first cross-sectional survey may add another estimate without resolving whether platform use changes confidence or confident students simply use the platform more.
The literature is not sparse. The inferential gap persists because the designs cannot adequately distinguish competing explanations.
This is why “more studies” is often an inadequate recommendation. The useful question is what different evidence is needed.
Watch Out
Do not recommend another study merely because an uncertainty exists. Ask whether the proposed design would actually reduce that uncertainty rather than reproduce the limitation that created it.
Measurement can create unanswered questions
A construct may have been studied extensively but operationalized so inconsistently that findings cannot be compared meaningfully. Alternatively, researchers may repeatedly use convenient proxy outcomes while rarely measuring the outcome that matters most.
Suppose studies of an educational intervention consistently measure satisfaction but rarely assess learning. The literature may establish that students like the intervention while leaving its educational effectiveness largely unanswered.
The gap is therefore not “few studies of the intervention.” It is inadequate evidence about a particular outcome.
Short follow-up leaves durability unanswered
An intervention may produce an immediate improvement while its long-term effect remains unknown. This is especially important when the practical or theoretical claim concerns sustained change.
A literature dominated by post-intervention measurements cannot establish persistence simply by accumulating more short-term studies.
State the temporal gap directly: what duration has been studied, and what duration remains unsupported?
Missing comparisons can leave decisions unresolved
Research may establish that an intervention performs better than no intervention while providing little evidence about whether it is better than an existing alternative.
If practitioners must choose between two active approaches, evidence against a no-treatment control may not answer their actual decision problem.
The unanswered question therefore concerns comparative effectiveness rather than absolute effectiveness.
Mechanisms can remain uncertain after effects are established
Evidence that an intervention changes an outcome does not necessarily establish why it works.
Researchers may propose several mechanisms, each compatible with the observed effect. Unless the studies measure or manipulate those mechanisms appropriately, the causal pathway remains unresolved.
Mechanism questions can matter because they affect theory, adaptation, implementation, and predictions about where an intervention should work.
Harms and unintended consequences are often separate gaps
A literature focused on benefits may contain inadequate evidence about adverse or unintended outcomes. Absence of reported harm is not necessarily evidence that harm does not occur, particularly when studies did not measure it systematically.
Cochrane emphasizes consideration of important outcomes, including adverse outcomes, when interpreting bodies of evidence.
Missing evidence can also result from selective availability
An apparent gap may arise because relevant studies or outcomes have not been published or are difficult to locate. Publication bias and selective outcome reporting can therefore affect not only conclusions about what is known but also perceptions of what has been studied.
Before declaring that “no research exists,” ensure that the search process is adequate for the scope of the claim.
A gap should imply a researchable uncertainty
A useful unanswered question usually contains enough specificity to guide future investigation.
Compare:
“More research on online learning is needed.”
with:
“Although short-term achievement effects have been studied extensively, evidence is insufficient to determine whether the observed gains persist beyond the course in which the intervention is used.”
The second statement identifies the current knowledge, the uncertainty, and the missing temporal evidence.
Not every uncertainty deserves another study
Research resources are finite. Some uncertainties have little theoretical, practical, or policy importance. Others may be impossible or unethical to resolve directly.
Cochrane's guidance on implications for research links future research needs to the specific reasons certainty is limited. For example, risk-of-bias concerns may imply the need for better-designed studies, while indirectness or imprecision point toward different research needs.
The implication is important: future research should target the uncertainty that matters rather than merely increase the publication count.
| Why the question remains unanswered |
What the gap actually is |
What might reduce the uncertainty |
| Too little evidence |
Insufficient information about a relevant question |
Additional appropriately designed studies |
| Imprecise estimates |
Important alternatives cannot be distinguished |
Larger or more informative studies, where appropriate |
| Inconsistent credible findings |
Conditions producing different results remain unclear |
Studies designed to test plausible moderators or contextual explanations |
| Indirect population or context |
Applicability to the target setting remains uncertain |
Direct evidence in the relevant population or setting |
| Repeated weak design |
The literature cannot support the required inference |
A design capable of addressing the unresolved inference |
| Inadequate outcome measurement |
The outcome that matters has not been measured adequately |
Studies using valid and decision-relevant outcomes |
| Short follow-up |
Durability remains unknown |
Longer-term follow-up |
| Unresolved mechanism |
Why the observed relationship occurs remains uncertain |
Studies capable of distinguishing competing mechanisms |
04 · A Practical Example
Turning “More Research Is Needed” Into Specific Unanswered Questions
Hypothetical Example
Research on AI-generated feedback for student writing
Imagine a hypothetical literature containing fifteen studies. Most report that students appreciate rapid AI-generated feedback. Several short experiments report modest improvements in immediate writing scores. Nearly all studies involve university students, follow participants for one semester or less, and evaluate writing produced while students continue to have access to AI assistance.
The generic gap statement
“More research is needed on AI-generated feedback.”
The statement does not identify what the fifteen studies have already answered or what remains unresolved.
Map the boundary of knowledge
What appears reasonably supported? Students commonly report valuing rapid feedback, and some hypothetical experimental evidence suggests short-term improvements in assessed writing performance.
What remains uncertain about durability? The studies provide little evidence about whether improvements persist after AI assistance is removed.
What remains uncertain about transfer? It is unclear whether students become better independent writers or simply produce better work while supported by the system.
What remains uncertain about generalizability? Evidence outside university populations is sparse.
What evidence would address these questions? Longer-term designs assessing independent writing after withdrawal of AI assistance, together with research in currently underrepresented educational contexts, would target the identified uncertainties more directly.
The resulting research agenda follows from the synthesis rather than from a generic desire for more publications.