What Is Peer Review? Process, Types & Decisions什么是同行评审？Peer Review流程、类型、审稿意见、投稿决策与应对方法完整指南

What Is Peer Review?什么是同行评审？

The peer review meaning changes slightly with context. Journals review manuscripts, funding agencies review grant proposals, conferences review submissions, and institutions review appointments or promotions. This guide focuses on scholarly journal manuscripts. In that setting, the process normally happens after submission and before publication, although preprint commentary, registered reports and post-publication review create important variations.

Peer review 的具体含义会随场景略有变化：期刊评审稿件，资助机构评审项目申请，会议评审投稿，机构也会评议聘任或晋升。本指南聚焦学术期刊稿件。通常评审发生在投稿之后、出版之前，但预印本评论、Registered Reports 与发表后评审带来了重要变体。

Peer review is not the same as editing, proofreading or plagiarism screening. A technical check may confirm that files are complete; copyediting improves language and house style; similarity software flags text for human interpretation. Peer reviewers primarily assess scholarly content. Exact responsibilities vary by journal, article type and discipline, so the journal's current policy and reviewer instructions remain controlling.

同行评审不等于编辑加工、校对或查重。技术检查确认文件是否齐全；文字编辑改善语言与格式；相似度软件标记文本供人工判断；同行审稿人主要评估学术内容。具体职责随期刊、文章类型和学科而变，因此应以期刊当前政策与审稿指南为准。

Why Peer Review Matters—and What It Cannot Guarantee同行评审为何重要，以及它不能保证什么

Editors cannot personally master every theory, technique, dataset and clinical or technical context represented in submissions. Reviewers add focused expertise. They can test whether the question matters to the journal's audience, methods fit the claim, analyses are appropriate, results support the conclusions, relevant literature is represented, ethical and reporting requirements are addressed, and limitations are stated honestly. Authors also receive a structured outside reading before the work becomes part of the scholarly record.

编辑不可能亲自掌握投稿涉及的每种理论、技术、数据集及临床或工程语境。审稿人补充专门知识，判断研究问题是否对期刊读者重要、方法是否支撑主张、分析是否合适、结果能否支持结论、相关文献是否充分、伦理与报告要求是否满足、局限是否诚实说明。作者也能在研究进入学术记录前获得结构化的外部阅读。

Peer review cannot independently verify every datum, reproduce an experiment, detect all image manipulation, eliminate conflicts, guarantee novelty, or predict later influence. It also cannot tell an author which journal is best merely because a journal says it is peer reviewed. Assess scope, audience, policy transparency, editorial independence, indexing, cost and access alongside the review model. A Journal Impact Factor likewise cannot reveal review quality or fairness.

同行评审无法逐项独立核验所有数据、重复实验、发现全部图像操纵、消除利益冲突、保证新颖性或预测未来影响。期刊声称采用同行评审，也不能单独证明它就是最佳投稿目标。应同时评估范围、受众、政策透明度、编辑独立性、收录、费用与开放路径。期刊影响因子同样不能反映评审质量或公平性。

Authors, Editors, and Reviewers: Who Does What?作者、编辑与审稿人分别做什么？

A reliable mental model separates authority from advice. Authors are responsible for the work and their response. Editors control the editorial process and final decision. Reviewers provide confidential or public expert assessments under the journal's model. Editorial-office staff may run completeness, identity and policy checks but do not automatically make the scientific judgment.

可靠的理解方式是区分“决定权”与“建议”。作者对研究及回复负责；编辑管理流程并作出最终决定；审稿人按照期刊模式提供保密或公开的专家评价；编辑部工作人员可能执行完整性、身份与政策检查，但不一定负责科学判断。

Reviewer disagreement is normal because experts may emphasize different assumptions or risks. The editor should read the reasoning, seek another opinion when necessary, and decide in the journal's context. Authors should address every transmitted comment, but they may disagree respectfully when a request is infeasible, ethically inappropriate, outside scope, or unsupported. The response should explain the reasoning and offer evidence or an alternative.

专家强调的假设或风险不同，因此意见不一致很常见。编辑应阅读论证，必要时征求额外意见，并在期刊语境中决定。作者应回应所有转达意见，但当要求不可行、不合伦理、超出范围或缺乏依据时，可以礼貌不同意；回复中应说明理由并提供证据或替代方案。

The Peer Review Process Step by Step同行评审流程：逐步说明

1. Submission and administrative checks投稿与行政检查The journal receives the manuscript and checks files, declarations, authorship details, format and basic policy requirements. A returned submission may need correction rather than having been rejected.期刊接收稿件并检查文件、声明、作者信息、格式与基本政策。退回投稿可能只是要求纠正，并不等于拒稿。
2. Editorial or desk review编辑初审An editor assesses scope, article type, relevance, apparent rigor, ethics and priority. Manuscripts can be desk rejected without external reports when they do not fit or are not ready for review.编辑判断范围、文章类型、相关性、表面严谨性、伦理与优先级。范围不符或尚不具备外审条件的稿件可能直接初审拒稿。
3. Reviewer identification and invitation识别并邀请审稿人The editor looks for appropriate subject and method expertise, checks conflicts and availability, and may invite several people before enough reviewers agree.编辑寻找合适的学科与方法专家，检查利益冲突和时间，并可能邀请多人后才获得足够接受者。
4. Independent evaluation独立评价Reviewers read the manuscript, supporting information and journal criteria, then prepare comments for authors and sometimes confidential comments for the editor.审稿人阅读稿件、补充材料与期刊标准，撰写给作者的意见，有时另向编辑提交保密意见。
5. Editorial synthesis and decision编辑综合与决定The editor weighs the reports, their reasoning and their relevance. The decision may be reject, revise, accept, transfer or another journal-specific label.编辑权衡报告、论证与相关性，作出拒稿、修改、录用、转投或其他期刊特定决定。
6. Revision and point-by-point response返修与逐点回复Authors revise the manuscript, identify changes and answer every comment. Unmade changes require a clear explanation, not silence.作者修改稿件、标明变更并逐项回应。未采纳的建议必须清楚解释，不能忽略。
7. Re-review or editor check复审或编辑检查Major revisions often return to reviewers; minor changes may be assessed by the editor. New concerns can arise, and more than one revision round is possible.重大修改通常返回审稿人，小修可能由编辑检查。复审可能提出新问题，也可能经历多轮修改。
8. Final decision and production最终决定与出版制作Acceptance ends editorial evaluation but not author responsibility. Copyediting, proofing, data or license checks and later corrections remain separate stages.录用结束编辑评价，但不结束作者责任；文字编辑、校样、数据或许可检查及后续更正仍是独立阶段。

Timing is variable. Reviewer availability, specialized methods, journal workload, revision complexity and additional rounds all matter. Treat a journal's published timeline as dated evidence, not a guarantee for one manuscript. If a stated period has passed, the corresponding author can send a brief, courteous status request through the journal's official channel.

评审时长差异很大，受审稿人可用性、专业方法、期刊工作量、返修复杂度与额外轮次影响。期刊公布周期是有日期的证据，不是对单篇稿件的保证。若已超过所述时间，通讯作者可通过官方渠道简短、礼貌地询问状态。

Types of Peer Review同行评审有哪些类型？

Names are not perfectly standardized. “Blind” is increasingly replaced by “anonymous,” and “open peer review” can refer to several different practices. Read what the journal actually conceals, discloses and publishes rather than inferring the workflow from one label.

名称并非完全标准化。“blind”正逐渐被“anonymous”替代，而“开放同行评审”可能指多种实践。不要只凭一个标签推断流程，应阅读期刊具体隐藏、披露和公开哪些信息。

Collaborative review, portable review, review cascades and community review add further variants. No model automatically eliminates bias or ensures better comments. For authors, the actionable questions are: who can see identities, who conducts the assessment, whether reports or responses become public, whether supplementary files are reviewed, how conflicts are handled, and who makes the decision.

协作评审、可移植评审、级联评审与社区评审还带来更多变体。没有一种模式能自动消除偏差或保证更好意见。作者真正要问的是：谁能看到身份、谁执行评审、报告或回复是否公开、补充材料是否评审、如何处理利益冲突、最终由谁决定。

What Do Peer Reviewers Evaluate?同行审稿人通常评估什么？

The reviewer should use the journal's criteria and stay within their expertise. A methods specialist may evaluate an analysis deeply but have less authority on a clinical interpretation; a subject expert may understand importance but request statistical input. Good reports distinguish essential issues that affect validity from optional suggestions that merely reflect personal preference.

审稿人应使用期刊标准并限定在自身专业范围内。方法专家可深入评价分析，却未必适合判断临床解释；主题专家理解重要性，却可能需要统计意见。高质量报告会区分影响有效性的必要问题与仅代表个人偏好的可选建议。

• Question and contribution: Is the problem clear, relevant and appropriately situated in prior work?问题与贡献：问题是否清晰、相关，并恰当地置于既有研究中？
• Design and methods: Do sampling, controls, materials, measures and analyses support the stated inference?设计与方法：抽样、对照、材料、测量与分析能否支撑所述推断？
• Results and evidence: Are results reported completely, uncertainty represented and discrepancies addressed?结果与证据：结果是否完整报告、不确定性是否表达、差异是否处理？
• Interpretation: Are conclusions proportional to the evidence, without causal or general claims beyond the design?解释：结论是否与证据相称，是否避免超出设计的因果或泛化主张？
• Ethics and transparency: Are approvals, consent, registration, conflicts, funding, data and code statements addressed as required?伦理与透明：是否按要求处理审批、同意、注册、利益冲突、资助、数据与代码声明？
• Reporting and reproducibility: Could a knowledgeable reader understand what was done, evaluate it and, where appropriate, reproduce or replicate it?报告与可重复性：专业读者能否理解、评价，并在适当时复现或重复研究？
• Journal fit and presentation: Does the manuscript serve the journal's audience, article type and communication standards?期刊适配与表达：稿件是否符合期刊受众、文章类型与传播标准？

A useful report often begins with a short neutral summary, identifies major issues in priority order, lists minor corrections separately, and supports claims with reasoning. Reviewers should not demand citations primarily to inflate their own metrics, rewrite a manuscript into their preferred study, make personal comments, or recommend experiments unrelated to the central conclusions. For a worked example, see the peer review example, which applies these principles to a sample report.

实用报告通常先给出简短中性摘要，按优先级列出主要问题，再单独列小修，并为判断提供理由。审稿人不应主要为提升自身指标而要求引用，不应把稿件改造成自己偏好的研究，也不应进行人身评论或要求与核心结论无关的实验。完整示例请参阅同行评审示例，其中把这些原则应用到一份示范报告中。

What Peer Review Decisions Mean同行评审后的决定分别意味着什么？

Decision labels vary by journal. Read the editor's letter, not just the dashboard status. Reviewers may recommend outcomes, but the editor integrates the reports, policy and journal priorities. A positive reviewer recommendation does not bind the editor; a critical report does not automatically produce rejection.

决定标签因期刊而异，应阅读编辑决定信，而不是只看系统状态。审稿人可以建议结果，但编辑综合报告、政策与期刊重点。积极建议不约束编辑，严厉报告也不自动等于拒稿。

An appeal is not a second revision round. Use the journal's appeal route only when you can identify a material factual, procedural or ethical error, and support the request concisely. Disagreement with expert judgment by itself may not meet the journal's threshold. Never submit the same manuscript to another journal while it remains under consideration unless both policies explicitly permit that arrangement.

申诉不是第二轮返修。只有能够指出重大事实、程序或伦理错误时，才按期刊渠道提出简洁、有证据的请求。仅仅不同意专家判断未必达到申诉门槛。稿件仍在审时，不得同时投往其他期刊，除非双方政策明确允许。

How to Respond to Reviewer Comments如何回应同行评审意见

1. Read the editor's priorities first先读编辑重点The decision letter can resolve conflicting comments and identify mandatory changes. Separate editor instructions from reviewer suggestions.决定信可解决冲突意见并指出强制修改，应区分编辑要求与审稿建议。
2. Build a response matrix建立回复矩阵Copy every numbered comment without altering its meaning. Assign an owner, planned action, evidence needed, manuscript location and completion state.完整复制每条编号意见，分配负责人、计划行动、所需证据、稿件位置与完成状态。
3. Prioritize validity before style先处理有效性，再处理表达Resolve design, analysis, ethics and conclusion issues before polishing wording. Revisit connected sections when one change affects several claims.先解决设计、分析、伦理与结论问题，再润色语言；一项修改影响多个主张时，应检查所有相关章节。
4. Answer point by point逐点回复State what changed, why, and where. Quote only enough manuscript text to make verification easy, and use page, line, table or figure references.说明修改内容、原因与位置，只引用足够核验的稿件文本，并使用页码、行号、表格或图号。
5. Disagree with evidence用证据表达不同意Acknowledge the concern, explain the constraint or interpretation, cite supporting evidence, and revise the manuscript to prevent the same misunderstanding where possible.承认关注点，解释限制或理解，提供支持证据，并尽可能修改正文以防止同样误解。
6. Run a consistency check执行一致性检查Confirm the response, clean manuscript, tracked version, figures, supplements, abstract and conclusions all reflect the same revision.确认回复信、清稿、修订稿、图表、补充材料、摘要与结论反映同一修改。

Avoid emotional language, promises you cannot support, and silent deletions. If reviewers conflict, describe the conflict and follow the editor's direction rather than trying to satisfy incompatible requests invisibly. If a requested experiment is impossible, explain why and narrow the claim, add a limitation, provide an alternative analysis, or ask the editor whether the proposed response is acceptable.

避免情绪化语言、无法支持的承诺与无说明删除。审稿意见冲突时，应说明冲突并遵循编辑指导，不要暗中试图满足不兼容要求。若要求的实验无法完成，应解释原因，并缩小主张、增加局限、提供替代分析，或询问编辑拟议方案是否可接受。

The rebuttal letter guide provides a complete response structure and reusable language. The broader principle is that a response document proves how the manuscript changed; it should not become a performance of automatic agreement.

审稿回复信指南提供完整的回复结构与可复用表达。更重要的原则是：回复文件应说明稿件如何修改，而不应成为无条件同意审稿意见的表态。

How to Evaluate a Journal's Peer Review Policy如何核验期刊的同行评审政策

DOAJ's transparency principles define peer review as obtaining advice from reviewers or experts in the manuscript's subject area and expect journals to state whether content is reviewed, who conducts it, the model used, identity masking, publication of reports, decision responsibility and exceptions. That is a useful verification checklist even for journals outside DOAJ.

DOAJ 透明原则把同行评审定义为从稿件主题领域的审稿人或专家处获得意见，并期望期刊说明内容是否评审、由谁评审、采用何种模式、身份如何隐藏、报告是否公开、谁负责决定及存在哪些例外。即使期刊不在 DOAJ，该清单也适合用于核验。

• Can you find a current, specific peer-review policy on the official journal or publisher domain?能否在期刊或出版社官方域名找到当前、具体的同行评审政策？
• Does it name the review model, decision maker, reviewer role and article-type exceptions?是否说明评审模式、决定者、审稿人角色与文章类型例外？
• Are conflicts, confidentiality, complaints, appeals, corrections and misconduct handled transparently?是否透明处理利益冲突、保密、投诉、申诉、更正与不端行为？
• If timelines are advertised, are they defined and supported rather than guaranteed?若宣传审稿周期，是否给出定义与依据，而非保证？
• Are fees disclosed independently of acceptance, with no unexplained payment destination?费用是否独立于录用透明披露，且没有无法解释的收款方？
• Do recent articles show submission, revision and acceptance dates or published review records where the model expects them?近期文章是否按该模式展示投稿、返修、录用日期或公开评审记录？

A rapid promise does not prove fraud, and a long process does not prove quality. However, guaranteed acceptance, missing editorial responsibility, copied policies, hidden fees, unverifiable board members, unsolicited pressure, false indexing claims or payment to unrelated accounts warrant investigation. Verify journal identity, ISSN and database claims separately. Our Journal Citation Reports guide explains why database metrics and editorial practice are distinct evidence layers.

快速承诺不必然证明欺诈，漫长流程也不证明质量。但保证录用、编辑责任缺失、复制政策、隐藏费用、编委无法核验、强迫式邀稿、虚假收录或向无关账户付款，都值得调查。期刊身份、ISSN 与数据库主张应分开核验。我们的Journal Citation Reports 指南说明数据库指标与编辑实践是不同证据层。

Peer Review Ethics, Confidentiality, Conflicts, and AI同行评审伦理：保密、利益冲突与 AI 使用

COPE, ICMJE and research-integrity guidance converge on core duties: accept only when qualified and available; declare relationships that could bias judgment; keep the manuscript confidential; do not use unpublished ideas for personal advantage; provide constructive, supported comments; respond on time; and alert the editor to suspected ethical concerns through the journal rather than investigating authors independently.

COPE、ICMJE 与科研诚信指南对核心职责具有共识：只有具备能力和时间时才接受邀请；披露可能造成偏差的关系；对稿件保密；不得利用未发表想法谋取个人利益；提供建设性、有依据的意见；按时完成；发现伦理问题时通过期刊提醒编辑，而非自行调查作者。

Confidentiality also applies to delegation. A reviewer should not give the manuscript to a student or colleague, ask another person to draft the review, or contact authors directly unless the editor permits the arrangement. When co-reviewing is authorized, contributors should receive appropriate recognition and understand the same confidentiality obligations.

保密义务同样适用于转交。未经编辑允许，审稿人不应把稿件交给学生或同事，不应请他人代写报告，也不应直接联系作者。若获准共同审稿，参与者应获得适当认可，并承担同等保密义务。

AI output cannot replace the reviewer's independent scientific judgment. It can be wrong, incomplete or biased, and responsibility remains human. Authors should likewise avoid uploading reviewer reports or confidential editorial correspondence without authorization. For InfiniSynapse or any other workflow tool, provide only information you are entitled to process and understand the service's privacy conditions before transmission.

AI 输出不能替代审稿人的独立科学判断，它可能错误、不完整或有偏差，责任仍属于人。作者同样不应未经授权上传审稿报告或保密编辑通信。无论使用 InfiniSynapse 还是其他工具，都应只处理自己有权处理的信息，并在传输前理解服务隐私条件。

Limitations, Bias, and Signs of a Weak Review Process同行评审的局限、偏差与薄弱流程信号

Peer review depends on scarce expert time and human judgment. Suitable reviewers can be difficult to recruit; reports can conflict; fashionable topics, institutions, language, geography, gender, seniority or theoretical commitments can influence evaluation; and anonymity may be incomplete. Double-anonymous review can reduce some identity cues but cannot remove clues from methods, citations, datasets or preprints. Open models increase transparency but can create reluctance where power differences are large.

同行评审依赖稀缺专家时间与人工判断。合适审稿人可能难以邀请；报告可能冲突；热门主题、机构、语言、地域、性别、资历或理论立场可能影响评价；匿名也可能不完整。双匿名可减少部分身份线索，却无法消除方法、引用、数据集或预印本暴露；开放模式增加透明度，但在权力差异较大时也可能让参与者顾虑。

Weak process signals include a report unrelated to the manuscript, unsupported insults, excessive self-citation demands, requests that fundamentally change the research question without justification, identical generic reports, reviewer contact outside the system, leaked confidential details, undisclosed conflicts, impossible deadlines, or an editor who forwards harmful content without mediation. One poor comment does not prove the entire journal is illegitimate, but it merits a documented, calm communication to the editor.

薄弱流程信号包括：报告与稿件无关、无依据侮辱、过量自引要求、无理由改变研究问题、完全相同的通用报告、绕过系统联系、泄露保密细节、未披露冲突、不可能时限，或编辑未经处理直接转发伤害性内容。单条差意见不能证明整个期刊不合法，但值得用记录清晰、冷静地告知编辑。

Authors can reduce avoidable friction by selecting a journal that publishes comparable work, following reporting guidelines, making methods and limitations explicit, checking statistics and references, providing required ethics and data statements, and submitting clean files. These actions improve reviewability, not acceptance probability. A rigorous manuscript can still be rejected for fit, priority or reviewer concerns, and a positive review does not make later correction unnecessary.

作者可通过选择发表相似研究的期刊、遵守报告规范、明确方法与局限、检查统计和参考文献、提供所需伦理与数据声明、提交整洁文件，减少可避免摩擦。这些行动提升可评审性，并不保证录用。严谨稿件仍可能因适配、优先级或审稿关注被拒，积极评审也不能免除后续更正。

Peer Review FAQ同行评审常见问题