Evidence-Based Writing: Sourcing and Citation Practices That Hold Up to Scrutiny

The Federal Trade Commission's standard for advertising substantiation requires that objective claims about a product or service be supported by evidence at the time the claim is made — not retrieved afterward if challenged, but held in advance of publication. The standard applies to efficacy claims, comparative claims, and statistical claims, and the level of evidence required scales with the specificity and importance of the claim. A claim that a supplement reduces cardiovascular risk requires randomized controlled trial evidence, not a secondary source citing an industry report. The FTC does not grade on a curve for documents that look scientific but are not rigorously sourced.

This standard is more demanding than what most professional document writers practice, and less demanding than what peer review in academic research requires. Between those two poles sits most professional writing: grant applications, policy briefs, white papers, technical reports, regulatory submissions, and research proposals that make claims about the world and cite evidence for those claims, but operate under standards that are neither as strict as peer review nor as lenient as unreviewed publication. Understanding where your document sits on this spectrum — and what evidence quality standard applies to the claims you are making — is the prerequisite for writing that holds up to professional scrutiny.

How Citation Practices Differ by Document Type

Academic peer-reviewed articles, grant proposals, and professional reports make claims and cite evidence, but they operate under different citation standards that reflect different audiences, different review processes, and different consequences for inadequately supported claims.

In academic peer-reviewed articles, the citation standard for empirical claims is primary source citation — the original study that produced the finding, not a review article or textbook that summarizes it. A methods paper that cites a review article for a specific statistical claim rather than the primary methodological paper will receive a reviewer comment asking for the primary source. This is not pedantry; it is quality control against a specific failure mode: secondary sources summarize and sometimes misstate or oversimplify primary findings, and a chain of secondary citations can propagate errors through a literature without anyone returning to verify the original claim.

In grant applications, citation practice is disciplinarily constrained. NIH applications in biomedical fields are expected to cite primary sources for all empirical claims, with review articles acceptable for broad background statements. NSF applications in mathematics and physical sciences may cite fewer sources at higher levels of generality, reflecting the citation culture of those fields. Foundation grant applications often have more flexible citation standards because program officers are not conducting systematic review of the bibliography, but unsupported claims in a grant narrative that are challenged during site visits or progress report review create credibility problems that affect future funding.

In professional reports — policy briefs, white papers, regulatory submissions — citation standards are set by the intended audience and the likely review context. A policy brief intended to influence regulatory decision-making will be scrutinized by staff who know the relevant literature; unsupported claims or miscited statistics will be identified and will undermine the document's credibility. A white paper intended for a general business audience may face less rigorous citation review, but if it makes regulatory compliance claims or statistical claims that become the basis for business decisions, the citation standard should be higher than the audience explicitly requires.

How Expert Reviewers Assess Evidence Quality

Expert reviewers assessing a document's evidence base evaluate along four dimensions: primary vs. secondary sourcing; recency; sample size and study design; and contextual appropriateness. Understanding each dimension clarifies what "adequately cited" means to reviewers in different contexts.

Primary vs. Secondary Sources

The primary vs. secondary distinction is the most fundamental citation quality dimension. A primary source is the original report of a finding — the paper, dataset, or document in which the finding was first presented. A secondary source is any document that reports, summarizes, or interprets a finding from a primary source. Textbooks, review articles, Wikipedia, and news coverage of research are secondary sources. The underlying studies they describe are primary sources.

Primary sources are preferred in high-scrutiny documents for a simple reason: the secondary source's description of the primary finding may be incomplete, simplified, or incorrect, and the reviewer who checks the citation will find this. Secondary sources are appropriate for background statements, consensus positions, and general domain framing, where the specific details of the primary studies are not being relied upon. They are not appropriate for specific empirical claims — statistics, effect sizes, prevalence figures — where the accuracy of the specific number matters.

Recency

Evidence recency requirements vary by document type and by the nature of the claim. Clinical guidelines and regulatory submissions typically require that evidence be from studies completed within a specified time window — often five to ten years — because the field may have changed and older studies may reflect superseded methods or populations. Grant applications that rely primarily on studies from more than ten years ago will receive reviewer comments asking whether the field has moved on. Policy documents that cite social or economic statistics without noting that the data are from a survey conducted fifteen years ago are citing potentially outdated information that reviewers will flag.

The recency requirement is less strict for methodological claims (that a specific statistical method has a specific property) and for foundational conceptual claims (that a theory has a specific implication) than it is for empirical claims about the current state of a phenomenon. The test is whether the passage of time since the cited study is likely to have changed the finding — if the phenomenon being described could plausibly have changed, the citation should be recent enough to reflect the current state.

Sample Size and Study Design

Expert reviewers reading a document that cites empirical evidence evaluate the study design and sample size of the cited studies, not just the finding. A document that cites a single-site study of 40 participants as evidence for a general population claim will draw a reviewer comment about generalizability. A document that cites a cross-sectional study as evidence for a causal claim — that X causes Y — will draw a reviewer comment about causal inference. A document that cites an industry-funded study as the primary evidence for a claim about an industry product will draw a comment about conflict of interest.

Awareness of these design-level concerns is what distinguishes sophisticated evidence use from naive citation. The experienced writer selects studies that are methodologically appropriate for the claim being made and acknowledges when the available evidence has design limitations that affect how confidently the claim can be stated. A claim presented as established when it rests on a single small study is overconfident relative to the evidence; a claim presented as uncertain when it is supported by multiple large trials is underconfident in a way that understates the strength of the evidence. Both mismatches are citation quality failures.

Why Selective Citation Is Spotted Immediately

Selective citation — choosing only the studies that support your position while ignoring studies that complicate or contradict it — is one of the most common and most detectable citation quality failures in professional documents. Expert reviewers who know a literature well will notice when a document cites only a subset of relevant studies, particularly when the omitted studies are prominent in the field. The detection is reliable because reviewers often have the papers you chose not to cite in their own bibliographies.

The consequences of selective citation differ by document type. In academic peer review, selective citation generates reviewer comments that cite the omitted literature and ask why it was not engaged with — a request that requires the author to either address the conflicting evidence or explain why it is not relevant. In regulatory submissions, selective citation can generate formal challenges if reviewers identify that studies unfavorable to the submitted claim were available but not cited. In policy documents, selective citation that is identified by readers from opposing positions becomes the primary rebuttal: "The report ignored X studies showing the opposite" is a more effective rebuttal than "We disagree with the interpretation of the cited evidence."

The professional standard for evidence in documents making specific recommendations is not to cite only supporting evidence but to engage with the full relevant evidence base and explain how the recommendation follows from the totality of that evidence, including the conflicting parts. This standard is more demanding than the instinct to support a position with citations, but it is what produces documents that hold up to expert review rather than ones that are undermined by it.

How to Source Statistical Claims in Non-Academic Contexts

Statistical claims — prevalence figures, cost estimates, effect size claims, trend data — are the most frequently miscited category in professional documents outside academia. The most common errors are: citing a secondary source for a statistic that comes from a primary dataset; citing a statistic from a report that drew on data that has since been updated; citing a statistic without noting the date of the data it is based on; and citing a statistic from a context that differs from the context in which it is being applied.

Government datasets — the Current Population Survey, the National Health Interview Survey, the Bureau of Labor Statistics' employment data, the Census Bureau's American Community Survey — are the appropriate primary sources for most population-level statistics about the United States. Citing a news article that references a government dataset rather than the dataset itself introduces an intermediary that may have described the statistic imprecisely. The primary dataset is always the better citation when it is accessible.

Industry reports and foundation-commissioned surveys are secondary sources for statistical claims unless the report describes original data collection — in which case it should be cited with the caveat that the data source is the commissioning organization, which may have interests in the findings. This disclosure does not necessarily disqualify the citation, but it is what professional review bodies expect to see in documents that take their evidentiary responsibilities seriously.

The FTC Standard vs. Academic Peer Review Standards

Understanding the difference between the FTC's substantiation standard and academic peer review standards is relevant for researchers whose work crosses into commercial or regulatory contexts — a category that includes clinical researchers consulting for industry, scientists contributing to regulatory submissions, and academics writing for policy audiences.

The FTC's competent and reliable scientific evidence standard requires that claims be based on tests, analyses, research, studies, or other evidence that have been conducted and evaluated in an objective manner by qualified individuals and that are generally accepted in the relevant scientific community. This is a contextual standard, not a procedural one: it does not require peer-reviewed publication, but it requires that the evidence methodology be scientifically sound and that the claim reflect the weight of that evidence rather than cherry-picked supportive findings.

Academic peer review's standard for individual papers is narrower: it asks whether the specific claims of the paper are supported by the specific evidence presented in that paper, conducted according to sound methodology, interpreted with appropriate epistemic humility. It does not require that the paper address all related evidence; peer review addresses one paper's claims against one paper's evidence. The FTC standard, by contrast, requires that a claim be consistent with the totality of relevant evidence — a more demanding requirement that is closer to the systematic review standard than the individual paper standard.

Handling Conflicting Evidence in Documents That Make Recommendations

Documents that make recommendations — policy briefs, clinical practice guidelines, regulatory submissions, research proposals advocating for a specific approach — must engage with conflicting evidence in ways that support the recommendation without misrepresenting the evidence base. This is a genuine writing challenge, because the instinct to present a recommendation confidently pulls against the obligation to acknowledge evidence that complicates the confidence.

The professional approach is to address conflicting evidence explicitly, explain why it does not undermine the recommendation, and hedge the recommendation appropriately to reflect the actual state of the evidence. "Based on the available evidence, which includes three large trials supporting X and one smaller trial finding no effect, we recommend X for Y population with the caveat that the evidence in Z subpopulation is less definitive" is a recommendation that is both confident and honest about its evidence base. It will hold up to scrutiny because it demonstrates that the recommending authors have engaged with the full evidence rather than selected only the supportive parts.

Documents that ignore conflicting evidence produce recommendations that look stronger on the page than they are, and that are undermined when expert readers identify the omitted evidence. The perceived strength of the recommendation decreases when the omission is identified, and the credibility of the authors decreases with it — in ways that affect their future recommendations' reception. The short-term gain of a more confident-looking document is purchased at the cost of long-term credibility that compounds across a career.