Evidence Gap Detection: A Practical Guide from Draft Audit to Search

You have a draft, a deadline, and a nagging worry that the gap you claim is not really a gap. Evidence gap detection is the work of checking that worry against your own claims, your search record, and the primary sources you cite. This guide connects six moves into one loop: naming the gap type, auditing claims, repairing searches, tracing primary sources, building an evidence matrix, and deciding when to stop. Each section ends with a concrete next action so you can enter at whatever stage your draft is in.

A gap statement is a claim about what prior work does not yet establish. Before you rewrite it, classify it. The classification tells you what evidence would count for or against it.

NIH guidance on rigor and reproducibility treats careful assessment of prior knowledge, strengths, weaknesses, and knowledge gaps as a core part of research practice. That framing is useful here: the gap is not a rhetorical flourish at the end of your introduction. It is a claim you should be able to defend with the same care as a methods paragraph.

Next action: write your current gap statement in one sentence and label it with one row of the table. If you cannot choose a row, the statement is probably too broad.

Once the gap is labeled, list the sentences in your draft that depend on it. These usually include the gap statement itself, the sentence that says prior work has not examined something, and the sentence that says your study is therefore needed.

For each sentence, ask three questions:

1. What is the scope? Does the claim cover all prior work, or only the studies you actually read?
2. What is the source? Is it a primary study, a review, or your own inference from absence?
3. What would falsify it? Name one kind of paper that, if it existed, would weaken the claim.

That third question turns the audit into a search task. If your gap claim is about a population, the falsifying paper is a study in that population. If it is about a method, the falsifying paper is a study using the method you say is missing.

Next action: mark each gap-dependent sentence as supported, needs search, or rewrite. Do not leave any sentence unmarked.

A gap claim is only as good as the search behind it. Cochrane guidance on searching recommends planning searches from eligibility criteria, using multiple sources, combining free-text terms with subject headings, documenting the process, deduplicating records, screening in stages, and rerunning searches before publication. You can apply the same discipline at student scale.

Write your population, exposure or intervention, outcome, and study design as separate blocks. Then build terms for each block. PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, and search details that reveal how terms were translated and combined. Use the search details view to see whether your terms were mapped the way you intended.

Google Scholar documents searching by author and title, date sorting and limits, Cited by, Related articles, citation export, and ways to locate alternative versions of a work. PubMed offers Similar Articles and citation lookup. These are different routes to the same target: the primary report you may have missed.

Record the date, the source, the exact query, filters applied, and the number of records screened. Cochrane guidance treats documentation as part of the search process, and your log is also what lets you rerun the search later without guessing.

If you need a scholarly search interface to run these blocks, WisPaper search is one place to start; for a public entry point to finding academic papers, see WisPaper Scholar Search.

Next action: rebuild your main query as blocks, run it in at least two sources, and paste the exact strings into your log.

Reviews and secondary summaries are efficient, but they compress. A review sentence like "evidence is limited" may hide a handful of small studies, a single underpowered trial, or a body of work in a different population. Tracing means following the citation chain until you reach the report that generated the data.

Use three moves:

• Backward tracing: open the reference list of the review or the paper you are citing and locate the specific study behind the sentence.
• Forward tracing: use Cited by in Google Scholar to find later work that tested, extended, or contradicted the original.
• Version tracing: use Google Scholar's alternative-version links to check whether you are reading a preprint, a conference version, or the final article.

Cochrane guidance also notes the need to link multiple reports of one study. The same trial may appear as a protocol, a conference abstract, and a full report. If your matrix counts them as three studies, your gap claim may overstate the evidence base.

Next action: pick the three sentences in your draft that most strongly assert what prior work shows, and trace each to a primary report.

An evidence matrix is a table with one row per primary study and columns chosen to test your gap claim. Keep it small enough to finish.

Fill the matrix from primary reports, not from review sentences. When a cell is empty, that emptiness is a finding: it may support your gap, or it may mean you have not searched that cell yet. NIH guidance connects transparent methods, validation, and reproducibility to rigorous practice, and the matrix is where those qualities become visible in your literature review.

If you are organizing the papers you have already traced, My Library lets you save or upload papers and organize them in one place. If you want to ask questions based on papers you have uploaded or added from search results, Library QA supports that step.

Next action: build the matrix with five to ten rows before you write another word of the gap paragraph.

Searches can expand indefinitely. Stopping rules make the end of the search a decision rather than a mood.

Cochrane guidance recommends rerunning searches before publication. At student scale, that means one final rerun close to submission, with the date recorded. If the rerun changes nothing, your stopping rule held. If it changes something, you have time to adjust the claim rather than the conclusion.

Next action: write your stopping rules down before the next search session, and date the session.

Readers arrive at different stages. Use this table to pick the first move.

For longer research tasks that involve literature review or inspiration, Scholar Agent is one area of WisPaper built for that kind of work. If you need a workflow that takes a research topic, background, references, and experimental constraints to generate research ideas, Idea Discovery supports that step. If you are setting up a project area, Research Projects is where you can create a new project.

Next action: find your row, do only the first move in the "Enter here" column, and stop for today.

Suppose a student has three days left, a rubric that rewards a "clearly justified research gap," and access to only some journals. The draft says: "Few studies have examined how part-time students use feedback in online courses." The student cannot tell whether this is true or just unsearched.

Day one: diagnose and audit. The statement is a population gap. The audit marks it needs search because the student has read only two reviews. The student writes the falsifying paper: a study of part-time students using feedback in online courses.

Day one: repair the search. The student builds blocks for population (part-time, online), construct (feedback use), and design. They run the blocks in two sources, log the queries, and use search details to check term mapping. They find one small study in a related setting and one protocol without results.

Day two: trace and matrix. The student traces the review sentence to its primary report and discovers the review counted a conference abstract and a full report of the same study as two. The matrix now has one row for that study, not two. The related-setting study fills the translation-gap column but not the population column.

Day two: stopping rules. The student sets saturation and deadline-fit rules. A final rerun is scheduled for the morning of submission.

Day three: rewrite and decide. The gap statement becomes narrower: the student claims a translation gap, not an absolute population gap, and cites the related-setting study as the reason the question remains open in the target setting. The rubric item is addressed with a traceable claim.

Submission decision: submit after the final rerun, provided the rerun does not surface a study in the exact population. If it does, revise the gap sentence to a translation or mechanism gap before submitting. That is a decision rule, not a hope.

• The gap statement names one gap type.
• Every gap-dependent sentence is marked supported, needs search, or rewritten.
• The search log records sources, queries, filters, and dates.
• Key review sentences are traced to primary reports.
• The matrix has one row per primary study, with multiple reports linked.
• Stopping rules are written and the final rerun is dated.
• The submission decision is stated in advance.

If you are checking whether your BibTeX entries point to real records, TrueCite screens BibTeX against academic databases to flag hallucinated references. If you need a general entry point to WisPaper as an AI research assistant for finding, screening, organizing, and working with academic papers, start at the WisPaper homepage.