The Science: Frequently Asked Questions

Science questions rarely arrive in a vacuum. They tend to show up mid-conversation, mid-crisis, or in the small hours when something you read doesn't quite add up. This page addresses the practical questions that surface most often — about how scientific inquiry works, how findings get classified and reviewed, where reliable information lives, and what separates a well-grounded claim from a well-dressed guess. The answers here draw on established scientific practice, named institutional sources, and the kind of methodological transparency that distinguishes good science from everything else.

What are the most common issues encountered?

The gap between what research shows and what people understand about it is, frankly, enormous — and not because the public is incurious. A 2022 analysis by the Pew Research Center found that 57% of U.S. adults say science has had a mostly positive effect on society, yet confidence in scientists as a group has declined from a 2020 high. That tension — trusting the enterprise while doubting the practitioners — creates fertile ground for confusion.

The most common issues break into three categories. First, misreading statistical significance: a p-value below 0.05 does not mean a finding is important, only that it is unlikely to be random noise. Second, conflating correlation with causation — a pattern that appears in data does not establish mechanism. Third, failing to distinguish between a single study and a body of evidence. A single paper is a data point; the literature is the argument.

How does classification work in practice?

Scientific claims are not all created equal, and the frameworks for ranking evidence quality exist precisely because that matters. The most widely used structure is the hierarchy of evidence, where systematic reviews and meta-analyses of randomized controlled trials (RCTs) sit at the top, followed by individual RCTs, cohort studies, case-control studies, and expert opinion at the base.

The National Institutes of Health (NIH) and the Cochrane Collaboration both use variants of this hierarchy in assessing clinical and public health evidence. In environmental and physical sciences, classification often shifts toward reproducibility and predictive validity — how well a model's outputs match observed data across independent datasets.

The practical upshot: a finding classified as "preliminary" or "exploratory" in a press release is doing different epistemic work than a finding described as "replicated across 12 independent cohorts." Those phrases are not interchangeable.

What is typically involved in the process?

Scientific investigation follows a recognizable structure, even when the subject matter varies wildly. A numbered breakdown clarifies the stages:

1. Observation and question formation — identifying a phenomenon that requires explanation
2. Literature review — establishing what is already known, to avoid reinventing established findings
3. Hypothesis development — generating a testable, falsifiable prediction
4. Study design — choosing methods appropriate to the question (experimental, observational, computational)
5. Data collection and analysis — executing the protocol and applying statistical or qualitative analysis
6. Peer review — submitting findings to independent expert scrutiny before publication
7. Publication and replication — entering the scientific record and inviting independent verification

Peer review, step six, is where much public confusion concentrates. A preprint — posted to repositories like arXiv or bioRxiv before peer review — has not yet cleared that threshold. Preprints are useful for rapid dissemination, but they carry a different epistemic status than a published, reviewed article.

What are the most common misconceptions?

The most durable misconception is that science produces certainty. It does not. Science produces the best available explanation given current evidence, with explicit acknowledgment of uncertainty ranges. The phrase "the science is settled" is more accurately rendered as "the evidence converges strongly enough that continued debate is no longer productive at the margins."

A second persistent misconception: that scientific consensus equals unanimous agreement. Consensus means the preponderance of qualified expert judgment, not 100% agreement. The National Academies of Sciences, Engineering, and Medicine publishes consensus reports that explicitly document dissenting views alongside majority conclusions.

A third: that funding source automatically invalidates findings. Funding source is a factor in assessing potential bias, not a binary disqualifier. Methodology transparency, data availability, and independent replication matter more. For a deeper treatment of these issues, The Science: Myths and Misconceptions addresses the full landscape.

Where can authoritative references be found?

Primary literature lives in peer-reviewed journals indexed by PubMed (biomedical), Web of Science, and Scopus. The NIH National Library of Medicine provides open-access pathways through PubMed Central. For physical and earth sciences, NASA's Technical Reports Server and NOAA's data portals provide both raw data and reviewed findings.

Institutional repositories and government science agencies — NSF, USGS, EPA, CDC — publish technical reports that undergo internal review before release, offering a middle tier between preprint and peer-reviewed journal. The The Science: Trusted Resources page maintains a curated index of these sources organized by domain.

How do requirements vary by jurisdiction or context?

Science itself does not change by geography, but the regulatory and institutional requirements attached to scientific practice do. In the United States, human subjects research must comply with the Common Rule (45 CFR §46), overseen by Institutional Review Boards (IRBs). The Food and Drug Administration applies a separate evidentiary standard — substantial evidence from adequate and well-controlled investigations — for drug approval under 21 U.S.C. §355.

Environmental science standards vary by agency mandate: the EPA's quality assurance regulations under 40 CFR Part 30 govern federally funded research differently than internal agency science. In the European Union, the General Data Protection Regulation (GDPR) imposes additional constraints on research involving personal data, creating a distinct compliance environment even for identical scientific questions.

What triggers a formal review or action?

Formal scientific review is triggered through distinct channels depending on context. In research integrity, a formal investigation typically begins when an allegation of fabrication, falsification, or plagiarism is submitted to an institution's research integrity officer — a process governed by federal policy at agencies including the Office of Research Integrity (HHS). Retractions follow when journals determine published findings are unreliable; the Retraction Watch database tracks over 45,000 retracted papers as of its most recent count.

In regulatory science, formal action is triggered when an agency determines that evidence meets a statutory threshold — for example, when EPA determines a substance presents an unreasonable risk under the Toxic Substances Control Act. In clinical settings, the FDA may issue a safety communication or require label changes when post-market surveillance data reaches a defined signal threshold.

How do qualified professionals approach this?

Working scientists operate with methodological habits that rarely make it into popular coverage. A qualified researcher distinguishes between exploratory analysis — where hypotheses emerge from data — and confirmatory analysis — where a pre-specified hypothesis is tested. Conflating the two inflates false-positive rates, a dynamic documented extensively in the replication crisis literature, including the 2015 Open Science Collaboration study in Science that successfully replicated only 36 of 100 published psychology findings.

Professionals also maintain sharp awareness of effect size, not just statistical significance. A statistically significant finding with a Cohen's d of 0.1 may have no practical relevance. They pre-register study designs on platforms like ClinicalTrials.gov or the Open Science Framework to reduce post-hoc hypothesis adjustment. And they read the methods section — the part most readers skip — because that is where the actual claims live.

For a grounded starting point on these practices, the home page provides orientation across the full scope of scientific domains covered in this reference network.