Fisher M, Friedman S B, Strauss B
Department of Pediatrics, North Shore University Hospital, Cornell University Medical College, Manhasset, NY 11030.
JAMA. 1994 Jul 13;272(2):143-6.
To study whether reviewers aware of author identity are biased in favor of authors with more previous publications.
Randomized controlled trial.
Editorial office of the Journal of Developmental and Behavioral Pediatrics.
Two "blinded" and two "nonblinded" reviewers assigned to 57 consecutive manuscripts submitted between September 1991 and March 1992.
Spearman rank correlation coefficients were used to compare the sum of rating scores of 1 to 5 (1, accept; 5, reject) given by the two blinded reviewers, the two nonblinded reviewers, and the editors to the number of articles published previously by the first and senior authors (as determined from requested curricula vitae). Blinded reviewers were sent a questionnaire asking whether they could determine the identity of the authors, how they knew, and whether they thought binding changed the quality or difficulty of their review.
The Wilcoxon Sign Rank Test disclosed no differences between blinded and nonblinded scores. The number of previous articles by the senior author was significantly correlated (P < .01) with blinded scores (r = -.45) and editors' decisions (r = -.45), but not with nonblinded scores; the number of articles by the first author was correlated (P < .05) with editors' decisions (r = -.35) but not with blinded or nonblinded scores. Fifty (46%) of 108 blinded reviewers correctly guessed the identity of the authors, mostly from self-references and knowledge of the work; 86% believed blinding did not change the quality of their review, and 73% believed it did not change the difficulty of performing a review.
Blinded reviewers and editors in this study, but not nonblinded reviewers, gave better scores to authors with more previous articles. These results suggest that blinded reviewers may provide more unbiased reviews and that nonblinded reviewers may be affected by various types of bias.
研究知晓作者身份的审稿人是否会偏向于支持发表过更多文章的作者。
随机对照试验。
《发育与行为儿科学杂志》编辑部。
两名“盲法”审稿人和两名“非盲法”审稿人，负责评审1991年9月至1992年3月期间连续提交的57篇稿件。
采用斯皮尔曼等级相关系数，比较两名盲法审稿人、两名非盲法审稿人和编辑给出的1至5分（1分表示接受；5分表示拒绝）评分总和与第一作者和资深作者之前发表文章的数量（根据所要求的简历确定）。向盲法审稿人发放问卷，询问他们是否能确定作者身份、如何知晓以及他们认为盲法是否改变了评审的质量或难度。
威尔科克森符号秩检验显示，盲法和非盲法评分之间没有差异。资深作者之前发表文章的数量与盲法评分（r = -0.45）和编辑的决定（r = -0.45）显著相关（P < 0.01），但与非盲法评分无关；第一作者发表文章的数量与编辑的决定（r = -0.35）相关（P < 0.05），但与盲法或非盲法评分无关。108名盲法审稿人中，有50名（46%）正确猜出了作者身份，大多是通过自我引用和对作品的了解；86%的人认为盲法没有改变评审质量，73%的人认为没有改变评审难度。
本研究中的盲法审稿人和编辑，而非非盲法审稿人，对之前发表文章较多的作者给出了更高的分数。这些结果表明，盲法审稿人可能提供更公正的评审，而非盲法审稿人可能受到各种偏见的影响。
