Scientific Research: Frequently Asked Questions

Science generates more questions than answers — that's the point. This page addresses the practical, procedural, and conceptual questions that come up most often when people engage with scientific research, whether they're reading a journal article, designing a study, applying for funding, or simply trying to figure out whether a headline is worth believing. The scope runs from individual researchers and students to institutions, policymakers, and curious non-specialists.

What are the most common misconceptions?

The biggest one: that a single study proves anything. It doesn't. Individual studies produce evidence that either supports or fails to support a hypothesis — confirmation comes from replication and the accumulation of findings across independent research groups. The replication crisis, documented extensively by projects like the Open Science Collaboration's 2015 effort in Science (which found that fewer than 40 out of 100 psychology studies replicated at the original effect size), made this painfully clear.

A close second: that peer review guarantees correctness. Peer review filters for methodological plausibility and logical coherence, not truth. Reviewers don't re-run the experiment. They read, critique, and recommend — a valuable but imperfect process.

Third: that correlation implies causation, a concept so repeated it borders on cliché, yet still behind a remarkable share of misread health headlines. Observational data can generate hypotheses; it rarely settles them.

Where can authoritative references be found?

Primary literature lives in peer-reviewed journals, accessible through databases like PubMed (biomedical), Web of Science, and Scopus. For federally funded research, PubMed Central provides free full-text access — a requirement for most NIH-funded work under the 2023 NIH Public Access Policy update.

Preprints — studies posted before peer review — appear on servers like arXiv, bioRxiv, and SSRN. They're useful for tracking cutting-edge work but carry explicit caveats about unreviewed status. More on preprints and open access research explains the distinctions.

For policy-relevant science, the National Academies of Sciences, Engineering, and Medicine (nationalacademies.org) produces consensus reports that synthesize evidence across disciplines — a different product than a primary study, and often more useful for decision-making.

How do requirements vary by jurisdiction or context?

Significantly. Research involving human participants in the United States falls under the Common Rule (45 CFR Part 46), administered through Institutional Review Boards. Clinical trials have an additional regulatory layer through the FDA under 21 CFR Parts 50 and 56. Institutional review boards operate at the institutional level but must satisfy federal minimums.

Animal research is regulated under the Animal Welfare Act (administered by USDA) and, for federally funded work, the Public Health Service Policy on Humane Care. Roughly 50 species are covered by AWA protections, though mice and rats used in research were explicitly excluded until regulatory pressure began shifting that landscape. The animal research regulations page covers this in detail.

Outside the US, frameworks like the EU Clinical Trials Regulation (EU 536/2014) and GDPR impose different — sometimes stricter — requirements on data collection and participant consent.

What triggers a formal review or action?

Three main triggers activate formal oversight or misconduct review: an allegation from a third party (a colleague, institution, or journal editor), anomalies detected during audit or data review, and self-disclosure by the researcher or institution.

The Office of Research Integrity (ORI) at HHS oversees misconduct in federally funded biomedical and behavioral research. ORI defines research misconduct as fabrication, falsification, or plagiarism — the FFP standard — and distinguishes it from honest error or difference of scientific opinion. An inquiry typically precedes a full investigation, and institutions are required to report findings to ORI within specific timeframes. Research misconduct and fraud outlines how that process unfolds.

How do qualified professionals approach this?

Experienced researchers treat research design and methodology as the non-negotiable foundation. Design decisions made before data collection — sample size calculations, control conditions, pre-registration of hypotheses — determine whether results are interpretable at all.

Pre-registration, now standard practice on platforms like OSF Registries, involves logging the study's hypotheses and analysis plan before data collection begins. This separates confirmatory from exploratory analysis and reduces the risk of p-hacking. Journals including PLOS ONE and the British Medical Journal have adopted registered reports as a publication format specifically to address this.

Statistical analysis in research is treated not as an afterthought but as an integral part of design — because the analysis plan shapes what can legitimately be claimed from the data.

What should someone know before engaging?

Funding shapes research. Conflict of interest in research is a real structural issue — not a conspiracy theory, but a documented phenomenon with measurable effects. A 2017 meta-analysis in PLOS ONE found industry-funded nutrition studies were 5 times more likely to reach conclusions favorable to the sponsor than independently funded research.

Understanding who funded a study, and whether conflicts were disclosed, is basic research literacy. Journals require conflict-of-interest statements; reading them is worth the 30 seconds.

The National Science Authority's home resource provides orientation across the broader landscape of how research functions as a system, not just a set of individual studies.

What does this actually cover?

Scientific research spans basic science (curiosity-driven inquiry without immediate application), applied science (problem-directed investigation), and translational research (moving findings from lab to practice). Types of scientific research maps these categories with more granularity.

Within any category, research can be quantitative or qualitative — a distinction that matters enormously for methodology but is sometimes misread as a hierarchy. Qualitative methods aren't soft science; they're appropriate for different questions. Ethnographic fieldwork answers questions that a randomized controlled trial structurally cannot.

What are the most common issues encountered?

Four recurring problems appear across research contexts:

1. Underpowered studies — sample sizes too small to detect real effects reliably, producing false negatives or unstable effect estimates. Power calculations using tools like G*Power are standard practice for a reason.
2. Publication bias — the tendency for positive results to be published and null results to go unreported, distorting the literature. Systematic reviews and meta-analyses attempt to correct for this through funnel plot analysis and other methods.
3. Data management failures — lost, misformatted, or undocumented data that makes results irreproducible. NIH's 2023 Data Management and Sharing Policy now requires formal research data management plans for most funded grants.
4. Scope creep in interpretation — conclusions that drift well beyond what the data support, often in press releases more than in the papers themselves. A study conducted in 40 college-age participants in one US city does not establish universal human behavior.

These aren't exotic problems. They're the routine friction of real research, and recognizing them is the first step toward reading the scientific literature with clear eyes.