What Is Peer Review and How It Works

By Kevin Fernando · Sep 8, 2026 · 10 min read

Peer review is a quality-control process in which independent experts in a field evaluate a researcher's methods and findings before publication. It is used by academic publishers to decide whether work should be accepted, rejected, or accepted with revisions. Peer review catches some errors and validates findings, but it is not infallible, it fails to prevent invalid research or fraud, and reviewer disagreement is common.

Peer review is a quality-control process in which independent experts in a field evaluate a researcher's methods and findings before publication. It is used by academic publishers to decide whether work should be accepted, rejected, or accepted with revisions. Peer review catches some errors and validates findings, but it is not infallible, it fails to prevent invalid research or fraud, and reviewer disagreement is common.

When you encounter a peer-reviewed study in your research, you are reading work that has passed through a gatekeeping process. But what actually happened to it? Who evaluated it, and what did they check?

Peer review is not a single credential that means one fixed thing. It is a mechanism, a set of steps, roles, and incentives, that varies in practice across fields, journals, and time. Understanding how it works, and where it systematically fails, lets you read the credential correctly. A peer-reviewed paper has been scrutinized by experts in the field. That is genuine. But it has not been proven correct, and the reviewers may have disagreed sharply about whether to publish it.

What peer review is

Peer review is the process of having a researcher's methods and findings reviewed, usually anonymously, by experts in the same field. Academic publishers use peer review to decide whether work should be accepted, rejected, or accepted with revisions before it enters official publication in journals, monographs, or conference proceedings.

The core function is quality control. Before a paper becomes part of the scientific record, independent specialists examine the work to catch errors in methodology, logic, or interpretation. The reviewers are not the authors' colleagues or competitors they chose; they are selected by an editor specifically to provide expert scrutiny from outside the research group.

Why peer review exists

Scientific publishing would be chaos without some filter. Any researcher could publish any claim. Peer review solves this by requiring that work pass expert evaluation before it becomes part of the official record.

The system protects against obvious errors, a statistical mistake, a flawed experimental design, a misinterpreted dataset. It also protects against fraud: reviewers who know the field can spot claims that contradict established findings or rest on implausible methods. And it creates accountability: an author knows their work will be read carefully by specialists, which creates incentive to do the work correctly.

Peer review emerged as a formal process in 1665, when Henry Oldenburg, founding editor of Philosophical Transactions of the Royal Society, began sending manuscripts to experts for evaluation before publication. It became standard practice because it worked better than the alternative, no evaluation at all.

How the peer review process works

A manuscript moves through several gates. At each one, it faces a decision: move forward, revise and resubmit, or reject. Understanding the sequence helps you see where a paper's credential actually comes from.

Submission and initial screening

When a researcher submits a manuscript to a journal, the editor performs an initial screen. This is not full peer review; it is a desk decision. The editor reads the paper and decides whether it is worth sending to external reviewers.

Editorial screening typically takes a few days for desk rejection or one to two weeks for journals that screen more carefully. The editor is looking for basic fit: Does this paper match the journal's scope? Is it written clearly enough to review? Does it report novel findings, or is it redundant with published work? Does it have obvious fatal flaws?

Many papers are rejected at this stage without ever reaching a reviewer. These desk rejections are not peer review. They are editorial judgment. A paper rejected at desk has not been vetted by field experts; it has been filtered by an editor's assessment of whether it is worth their reviewers' time.

Reviewer assignment and evaluation

If a paper passes the desk screen, the editor invites independent reviewers to evaluate it. At least two reviewers are solicited; many journals request two to three.

Good reviewer selection prioritizes specialists who have recent, relevant publications connected to the manuscript's core claims. Reviewers should be able to comment on methods and statistics, not only the topic. They should represent different institutions and geographies to reduce bias.

Conflicts of interest to avoid include recent co-authors or collaborators, current or recent supervisors or trainees, same-department close colleagues, and anyone with a clear personal or financial stake in the outcome. The goal is to find experts who can assess the work fairly, without incentive to reject it or approve it for the wrong reasons.

Inviting reviewers commonly takes one to three weeks, longer in narrow subfields and around conference deadlines or grant cycles. Once a reviewer accepts, journals typically ask for completed reviews within two to four weeks.

Each reviewer reads the paper and writes a report. They assess whether the methods are sound, the data support the conclusions, the work is novel, and the writing is clear. They may recommend acceptance, rejection, or revision. Critically, reviewers often disagree. For rejected manuscripts, discordance was high at 85.8%.

Decision and revision

After reviews arrive, the editor makes a decision. The editor reads both reviews and decides whether to accept, reject, or ask for revision.

If the reviews are split, one recommends acceptance, one rejection, the editor must choose. The editor may request additional reviewers in cases of controversy or strong disagreement regarding the merits of the work, or when reviewers from several fields are needed.

Many papers are rejected outright. Others are sent back to the authors with reviewer comments and a request to revise and resubmit. The authors then address the reviewers' concerns, revise the manuscript, and submit again. This cycle can repeat multiple times.

The timeline for the full process ranges from about two weeks for a fast desk rejection or accelerated-track journal to well over a year for slow fields or manuscripts undergoing several revision rounds. In fields like physics, a year is common. In fields with slower publication cycles, two years is not unusual.

Once the paper is accepted, it is published. At that point, it carries the credential of peer review: it has been read and approved by independent experts. But approval by reviewers does not mean the paper is correct. It means the reviewers believed the work was sound enough to publish. Across review rounds, discordance declined markedly among published manuscripts to 29.3% in the final round, meaning even at acceptance, roughly one in four published papers still had at least one reviewer who disagreed. Overall alignment between the majority recommendation in the final round and the editorial outcome was approximately 88%, and 6.3% of published manuscripts had a majority recommendation of rejection in the final round.

Types of peer review

Peer review comes in several models. Which one a journal uses affects how much bias can enter the process.

Single-anonymized peer review is the long-standing default across most STEM and biomedical publishing. In this model, reviewers are told the identity of a manuscript's authors, but authors are not told the identity of the reviewers. The reviewer knows who wrote the paper; the author does not know who reviewed it. This asymmetry can introduce bias: a reviewer might be harsher on a competitor's work or more lenient on a friend's.

Of 1,054 MEDLINE-indexed medical journals evaluated, 766 employed single-blinded review and 288 were double-blinded.

Double-anonymized peer review, also called dual-anonymous review, removes this asymmetry. Neither party can identify the other during review: the reviewer does not know who the author is, and the author does not know who the reviewer is. This requires that the manuscript file sent to reviewers has been stripped of author names, affiliations, and other directly identifying material, and that the editorial system withholds reviewer identity from the author throughout. Double-anonymized review reduces the risk that reviewer identity or author identity will bias the evaluation, though it is harder to implement and can slow the process.

Open peer review is an umbrella term that refers to various forms of openness in peer review, often referring to the publication of review reports. One approach is to publish review reports only if the outcome of peer review is positive and the scientific work is considered suitable for publication. Another approach is to always publish review reports, also if the outcome of peer review is negative. Open peer review may also refer to the publication of the identities of reviewers, but this form of openness is more controversial. Publishing reviewer names can discourage harsh criticism if reviewers fear retaliation.

Real limitations of peer review

Peer review is not infallible. It fails in specific, predictable ways. Knowing these limitations helps you read the credential correctly.

Bias and gatekeeping

Reviewer identity and incentives distort outcomes. In single-anonymized review, a reviewer knows who wrote the paper and may be biased for or against the author. Reviewers can suppress novel findings that contradict their own work, or favor findings that support their preferred theories.

This is not always intentional. Reviewers are experts in their field, which means they have developed intuitions about what is true. A finding that contradicts those intuitions may feel wrong, even if the evidence is sound. A reviewer's job is to catch errors, but if the reviewer is convinced the conclusion is wrong, they may scrutinize the methods more harshly than they would for a conclusion they believed.

Gatekeeping also works through the choice of who gets to review. If an editor consistently invites reviewers from a narrow school of thought, papers from other schools will face harsher scrutiny. Peer review can entrench consensus and slow the acceptance of paradigm-shifting work.

The quasicrystal case illustrates this. Dan Shechtman's initial paper on quasicrystals was rejected by Journal of Applied Physics, with the reviewer's assessment being conclusion-based rejection rather than methodology-based. Shechtman was asked to leave his research group at the National Bureau of Standards after reporting his findings to colleagues. Linus Pauling, a two-time Nobel laureate, publicly called Shechtman's findings nonsense, stating there is no such thing as quasicrystals, only quasi-scientists. Shechtman received the Nobel Prize in Chemistry 29 years after his initial discovery, in 2011. A study reviewed instances in which 19 future Nobel Laureates encountered resistance from the scientific community towards their discoveries, and 24 future Nobel Laureates encountered resistance from journal editors or referees to manuscripts dealing with discoveries that later earned them the Nobel Prize.

Speed versus rigor tradeoffs

Publication pressure affects review quality. Reviewers are typically unpaid volunteers. They review papers in addition to their own research and teaching. If a reviewer is busy, they may skim the paper rather than read it carefully. If they are skeptical, they may ask for more experiments than necessary. If they are under time pressure, they may rush their assessment.

Journals face pressure to publish quickly. Researchers want their work published fast. Reviewers want to finish their reviews quickly. These pressures can lead to superficial review. A reviewer might approve a paper without catching an error, or reject a paper without fully understanding it.

The peer review timeline reflects this tradeoff. A two-week desk rejection is fast because the editor made a snap judgment. A one-year review process is slow because the paper went through multiple rounds of revision, or because reviewers took months to respond. Neither guarantees quality. Fast review can miss errors. Slow review can be thorough or can simply reflect bureaucratic delay.

What peer review actually signals

Peer review is a filter, not a proof. It signals that independent experts read the work and believed it was sound enough to publish. It does not signal that the work is correct, that it will hold up over time, or that it is the most important work in the field.

Peer review catches some errors. Analysis of retraction notices for 423 articles revealed that the most common causes of error-related retraction are laboratory errors, analytical errors, and irreproducible results. The most common laboratory errors leading to retraction are contamination and problems relating to molecular biology procedures such as sequencing and cloning. Retractions due to contamination were more common in the past, whereas analytical errors are now increasing in frequency. These are errors that peer review should catch but sometimes does not.

Peer review also fails to prevent fraud. In December 2025, PLoS ONE retracted 10 papers in Obstetrics and Gynecology with the motivation of concerns about compromised peer review. Reviewers can be fooled by fabricated data or can be complicit in the fraud.

Criticisms of peer review include failure to prevent publication of invalid research or scientific fraud, low correlation with future impact, slowness, and lack of accountability. A paper can be peer-reviewed and still be wrong. It can be peer-reviewed and still be ignored by the field. It can be peer-reviewed and still take years to publish.

What peer review does signal is that the work passed a threshold of expert scrutiny. The reviewers believed the methods were sound, the data were presented fairly, and the conclusions followed from the evidence. That is meaningful. It is not nothing. But it is not a guarantee of truth. It is a credential that says: this work was read by people who know the field, and they did not find fatal flaws. Use that information correctly.

Do this automatically

A journal managing hundreds of submissions annually must allocate weeks of expert time per manuscript across multiple independent reviewers, multiplied by the frequency of resubmissions and revisions, creating substantial labor costs that scale with publication volume. Software systems can automate portions of this workflow, from initial manuscript screening through reviewer assignment and tracking, to reduce the manual overhead while maintaining the human expertise that peer review requires. Academia research runs this kind of analysis end to end and cites every figure back to its original source, across as many subjects as you need, with Mimir attaching a citation to every figure. It will not write your paper for you. It gathers and cites the evidence; the argument stays yours, and deliberately so. Every plan includes it, including the free one.

Start your first report and run it on something you already know well.