Ratings in journal peer review can be affected by sources of bias. The bias variable investigated here was the information on whether authors had suggested a possible reviewer for their manuscript, and whether the editor had taken up that suggestion or had chosen a reviewer that had not been suggested by the authors. Studies have shown that author-suggested reviewers rate manuscripts more favorably than editor-suggested reviewers do.
Reviewers' ratings on three evaluation criteria and the reviewers' final publication recommendations were available for 552 manuscripts (in total 1145 reviews) that were submitted to Atmospheric Chemistry and Physics, an interactive open access journal using public peer review (authors' and reviewers' comments are publicly exchanged). Public peer review is supposed to bring a new openness to the reviewing process that will enhance its objectivity. In the statistical analysis the quality of a manuscript was controlled for to prevent favorable reviewers' ratings from being attributable to quality instead of to the bias variable.
Our results agree with those from other studies that editor-suggested reviewers rated manuscripts between 30% and 42% less favorably than author-suggested reviewers. Against this backdrop journal editors should consider either doing without the use of author-suggested reviewers or, if they are used, bringing in more than one editor-suggested reviewer for the review process (so that the review by author-suggested reviewers can be put in perspective).
Citation: Bornmann L, Daniel H-D (2010) Do Author-Suggested Reviewers Rate Submissions More Favorably than Editor-Suggested Reviewers? A Study on Atmospheric Chemistry and Physics. PLoS ONE 5(10): e13345. https://doi.org/10.1371/journal.pone.0013345
Editor: Pedro Antonio Valdes-Sosa, Cuban Neuroscience Center, Cuba
Received: July 15, 2010; Accepted: September 16, 2010; Published: October 14, 2010
Copyright: © 2010 Bornmann, Daniel. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The study was funded by the Max Planck Society. However, the funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
In the research on journal peer review, there are said to be biases, if – independently of the quality of submitted manuscripts – attributes of the reviewers (such as the nomination of a reviewer by the author or the editor) are correlated statistically with the reviewers' ratings . Arkes  defines bias “as any systematic effect on ratings unrelated to the true quality of the object being rated. Thus, bias consists of effects that reduce the validity of ratings through contamination, but not random error” (p. 378). According to Jayasinghe  “a random error is an ‘unexplained’ error whereas systematic bias such as leniency/harshness of reviewers … can be explained or statistically controlled” (p. 35).
Reviewers for a manuscript can be selected by editors (1) on the basis of their personal knowledge and familiarity from past experience, (2) from a database of previous reviewers cross-referenced by name and specialty, (3) from references listed in the manuscript, and (4) based on suggestions made by the authors of the manuscript . For Tonks , an assistant editor at the British Medical Journal (BMJ), the selection of author-suggested reviewers (Ra) “could improve the quality of peer review in two important ways. Firstly, authors are often better placed than editors to know whom to approach for a considered, balanced, and credible opinion in their field of research. The best reviewers are not those with the most experience or eminence and may be unknown to anyone outside the subject. This is a particular problem for editors of general journals, who review manuscripts from a wide range of disciplines. Secondly, nominated reviewers will enrich the BMJs database, keeping us in touch with young active researchers and giving us a broader population of reviewers.”
According to the “Ethical Guidelines for Publication in Journals and Reviews” of the European Association for Chemical and Molecular Sciences , editors have the responsibility “to consider the use of an author's suggested reviewers for his/her submitted manuscript, but to ensure that the suggestions do not lead to a positive bias.” Ra may be biased in favor of the authors . The danger with Ra is that “they can be the authors' best friends”  (p. 15). It is feared that through the use of Ra in addition to editor-suggested reviewers (Re) (meaning reviewers selected by the editor not on the basis of a suggestion by the author), the one (Ra) rates a manuscript systematically more leniently than the other (Re). (We assume this leniency effect, although an Re is not necessarily unknown to the authors.)
A number of studies of different journals showed that this fear is justified. A study by Schroter, Tite, Hutchings, and Black  on the peer review process at 10 biomedical journals found that Ra “tended to make more favorable recommendations for publication” (p. 314) than Re . Similar findings were reported by Scharschmidt, Deamicis, Bacchetti, and Held  for the Journal of Clinical Investigation, Earnshaw and Farndon  for the British Journal of Surgery, Goldsmith, Blalock, Bobkova, and Hall  for the Journal of Investigative Dermatology, Wager, Parkin, and Tamber  for medical journals in the BMC (BioMed Central) series, Rivara, Cummings, Ringold, Bergman, Joffe, and Christakis  for a pediatric journal, and Bornmann and Daniel  for Angewandte Chemie International Edition (AC-IE). In addition, Jayasinghe, Marsh and Bond  found similar results in the area of grant peer review.
In this study we aim to test whether there is a potential source of bias in the manuscript reviewing in public peer review at an interactive open access journal, Atmospheric Chemistry and Physics (ACP), through the use of Ra and Re. Using modern information technology, in particular the Internet, the ACP and other interactive open access journals have now become established in science that work with a “new” system of public peer review , . Compared to the traditional system, the new system of peer review in an electronic environment is seen to have the following advantages, among others: (1) submitted manuscripts are immediately published as “discussion papers” on the journal's website, (2) reviewers' comments on the quality of the content of the manuscript and authors' replies to the reviewers' critical comments are publicly exchanged, and (3) reviewers' arguments are publicly heard, and, if comments are openly signed, reviewers can also claim authorship for their contributions .
Even if all studies so far have found that Ra rate manuscripts systematically more favorably than Re, it would be expected that public peer review at ACP does not show this effect. (With the exception of Wager, Parkin, and Tamber , the aforementioned studies conducted up to now examined traditional peer review.) Public peer review is supposed to bring a new openness to the reviewing process that will enhance its objectivity . Publishing reviews is supposed to lead to reviewers using argumentation and judging solely on the basis of scientific criteria, so that the reviewer's ratings will not be influenced by potential sources of bias. We investigated the extent to which this expectation can be confirmed, taking the example of ACP.
Manuscript review at ACP
ACP was launched in September 2001. It is produced and published by the European Geosciences Union (EGU) (http://www.egu.eu) and Copernicus Publications (http://publications.copernicus.org/). ACP is freely accessible via the Internet (www.atmos-chem-phys.org). It has the second highest annual Journal Impact Factor (JIF) (provided by Thomson Reuters, Philadelphia, PA, USA) in the category “Meteorology & Atmospheric Sciences” (at 4.881 in the 2009 Journal Citation Reports, Science Edition). ACP has a two-stage publication process ,  that is described on the ACP website as follows: In the first stage, manuscripts that pass a rapid pre-screening process (access review) are immediately published as “discussion papers” on the journal's website (by doing this, they are published in Atmospheric Chemistry and Physics Discussions, ACPD). These discussion papers are then made available for “interactive public discussion,” during which the comments of reviewers (usually, reviewers that already conducted the access review), additional comments by other interested members of the scientific community, and the authors' replies are published alongside the discussion paper. The reviewers can be Ra or Re.
During the discussion phase, the designated reviewers are asked to answer to the following questions according to the ACP's principal evaluation criteria (see http://www.atmospheric-chemistry-and-physics.net/review/ms_evaluation_criteria.html, from which the following information is taken): (1) scientific significance (“Does the manuscript represent a substantial contribution to scientific progress within the scope of ACP (substantial new concepts, ideas, methods, or data?”), (2) scientific quality (“Are the scientific approach and applied methods valid? Are the results discussed in an appropriate and balanced way (consideration of related work, including appropriate references)?”), and (3) presentation quality (“Are the scientific results and conclusions presented in a clear, concise, and well-structured way (number and quality of figures/tables, appropriate use of English language)?”). The response categories for the three questions are: (1) excellent, (2) good, (3) fair, and (4) poor. In addition to the principal evaluation criteria, the reviewers are asked to give a final publication recommendation: “Do you recommend acceptance of the manuscript?” Here, the response categories are: (1) yes, without alterations, (2) yes, after minor alterations, (3) yes, after major alterations, and (4) no. Besides giving the formal ratings to the four questions, the reviewers also have the opportunity to write a commentary.
The ratings are submitted in parallel to the commentaries, but they are not open, because they are meant to support the editorial decision rather than the scientific discussion. This policy was introduced in 2001. According to the experiences and the philosophy of ACP's chief-executive editor Ulrich Pöschl, prescribed publication of formal ratings is likely to do more harm than good (e.g., initiation/escalation of unnecessary controversies). Most other journals pursuing public peer review do not prescribe publication of formal ratings either, and some of them explicitly instruct reviewers not to include formal ratings in their public comments (see, e.g., http://adv-model-earth-syst.org/index.php/JAMES/about/faq). At ACP, the editors leave it up to the reviewers if they want to include ratings in their public comments, and sometimes they do (∼30%). With increasing acceptance and spread of public review it may become beneficial and appropriate to prescribe publication of formal ratings. For now, however, the ACP editors prefer a mix of open commentaries and non-public ratings for the discussion phase.
After the end of the discussion phase every author has the opportunity to submit a revised manuscript taking into account the reviewers' comments and the comments of interested members of the scientific community. Based on the revised manuscript and in view of the access peer review and interactive public discussion, the editor accepts or rejects the revised manuscript for publication in ACP. For this publication decision, further external reviewers may be asked to review the revision, if needed. In general, an editor accepts a manuscript for publication in ACP, if – similar to the “clear-cut” rule of the journal AC-IE  – all reviewers rate the manuscript favorably (see here http://www.atmospheric-chemistry-and-physics.net/review/ms_evaluation_criteria.html).
Database for the present study
For the investigation of peer review at ACP we had data for 1111 manuscripts that went through the complete ACP selection process in the years 2001 to 2006 , , . Of the 1111 manuscripts, 1032 (93%) manuscripts were published as discussion papers; 79 (7%) were rejected during access review for publication as discussion papers. Reviewers' ratings on the evaluation criteria and reviewers' final publication recommendations, made during the discussion phase of the reviewing process, were available for 552 (55%) of the 1008 manuscripts. This reduction in number is due to the fact that the ratings have been stored electronically by the publisher only since 2004. Of the 552 manuscripts, 16% (n = 87) have one review, 64% (n = 356) have two, 17% (n = 92) have three, 3% (n = 15) have four, and 2 manuscripts have five independent reviews. Of the total 1145 reviews, 304 (27%) were by Ra and 841 (73%) by Re.
Of the 1111 manuscripts submitted between 2001 and 2006, 958 (86%) were published in ACPD and ACP, 74 (7%) were published in ACPD but not in ACP (here, the editor rejected the revised manuscript), and 79 (7%) were published neither in ACPD nor in ACP (these manuscripts were rejected during the access review). The search for the fate of the manuscripts that were not published in ACP (n = 153) revealed that 38 (25%) were published as contributions in other journals. No publication information was found for 115 (75%) manuscripts, whereby 70 of the 115 manuscripts (61%) were published in ACPD. The 38 manuscripts that were published as contributions in other journals were published in 25 different journals within a time period of five years (that is, between 2005 and 2009). Six manuscripts were published in the Journal of Geophysical Research; three manuscripts were published in Geophysical Research Letters. The other 23 journals published one or two of these manuscripts each .
Normally, when examining the association of a bias variable and reviewers' ratings it is impossible to establish unambiguously whether a particular group of manuscripts receives more favorable reviewers' ratings due to this variable, or if the more favorable ratings are simply a consequence of the manuscripts' scientific quality . For this reason, the statistical analysis should control for the scientific quality of a manuscript . Smart and Waldfogel  call this approach “a clean test for the existence of discrimination“ (p. 5), which in this study was realized through different statistical methods in two independent analysis steps.
To test whether Ra rate more leniently than Re, we used what is called a within-manuscript analysis as a first step. This analysis approach was proposed by Jayasinghe, Marsh, and Bond  for grant peer review research. They analyzed reviewers' gender as a potential source of bias in the Australian Research Council (Canberra) peer review and conducted “a within-proposal analysis based on those proposals with at least one male external reviewer and at least one female external reviewer” (p. 353). Some years later Wager, Parkin, and Tamber  investigated in the area of journal peer review “pairs of reviews from 100 consecutive submissions to medical journals in the BMC series (with one author-nominated and one editor-chosen reviewer and a final decision).”
At ACP between 2004 and 2006 135 of a total of 552 manuscripts (25%) were reviewed by a pair of Ra and Re. Differences in the ratings by the two reviewers of these manuscripts (related paired samples of Ra and Re) were investigated using the marginal homogeneity test , which generalizes the McNemar test from binary response to multinomial response. The method developed in the present release of StatXact  applies to ordered response. As the ACP data for the marginal homogeneity test are sparse, exact p-values were calculated.
As in the within-manuscript analysis only 135 of the 552 manuscripts could be included, an ordinal regression model (ORM) was computed as a second step to analyze ratings of Ra and Re. Using ORM, the association between several independent variables (here: suggestion of a reviewer and citations as an indicator for scientific quality) and an ordinal-scaled dependent variable (here: the reviewers' ratings) can be determined: “As with the binary regression model, the ORM is nonlinear, and the magnitude of the change in the outcome probability for a given change in one of the independent variables depends on the levels of all the independent variables”  (p. 183). For the analysis, the ACP data is a dataset where the assumption of independence between individual ratings of the reviewers may not hold, as the reviews are nested within manuscripts. In order to take the dependencies between individual ratings into account in the estimation of the ORMs, we used the “cluster” option in Stata . Specifying this option leads to robust standard errors in the sense that the estimates provide correct standard errors in the presence of the effects of clustered data . “The performance of the cluster-robust estimator is good with 50 or more clusters, or fewer if the clusters are large and balanced”  (p. 514). In this study we have 552 unbalanced clusters (manuscripts with one to five reviewers).
By fitting an ordinary ORM with robust standard errors for clustered data instead of fitting a variance components model (a multilevel model for ordinal responses), we were treating the within-cluster dependence as a “nuisance” and not as a phenomenon that we were interested in . A Wald test by Brant  was performed to test the parallel regression assumption for each independent variable considered in the ORM . As the test provides evidence that the assumption was violated for the variable “number of citations for a manuscript,” the variable was entered into the regression analysis as a log-transformed variable.
Out of a lack of other operationalizable indicators, it is common in research evaluation to use citation counts as an indicator for scientific quality. According to van Raan  citations provide “a good to even very good quantitative impression of at least one important aspect of quality, namely international impact” (p. 404). According to Lindsey  citations are “our most reliable convenient measure of quality in science – a measure that will continue to be widely used” (p. 201). In the present study we retrieved citation counts for manuscripts accepted by ACP or rejected and published elsewhere for a fixed time window of three years after the publication year. “Fixed citation windows are a standard method in bibliometric analysis, in order to give equal time spans for citation to articles published in different years, or at different times in the same year”  (p. 243). The citation analyses for the present study were conducted based on Chemical Abstracts (CA) (Chemical Abstracts Services, Columbus, Ohio, USA). CA is a comprehensive database of publicly disclosed research in chemistry and related sciences (see http://www.cas.org/).
As the citation counts were captured ex post – that is, after the editors' publication decisions (at ACP or another journal) – they are included in the regression models only as control variables. This means that in the analysis the interest was not the correlation between citation counts and reviewers' ratings but instead the correlation between the bias variable and ratings, when manuscript impact is statistically controlled. In statistical bias analysis this procedure is called the control variable approach .
Table 1 shows the minimum, maximum, mean, standard deviation, and median of the ratings by Ra und Re on the scientific significance, scientific quality, and presentation quality of a manuscript and the final publication recommendation. Whereas the arithmetic average ratings by Re are more negative on all evaluation criteria and for the final publication recommendation than the ratings by Ra, the median ratings of the two groups do not differ on either evaluation criteria or final publication recommendation. The median ratings for the two reviewers groups are always 2. The results shown in Table 1 are not really meaningful, as they do not refer to differences between Ra and Re on one and the same manuscript.
Table 2 presents the results of the within-manuscript analysis. For each evaluation criterion and for the final publication recommendation the table shows the difference between the ratings of reviewers for those manuscripts (n = 135) that were each reviewed by an Ra and an Re. The table shows the number of those manuscripts (row percents) for which the ratings by Ra and Re did not differ (column: “no difference”), the rating by Ra was more positive than the rating by Re (column: “Ra is more positive than Re”), and the rating by Re was more positive than the rating by Ra (column: “Re is more positive than Ra”). As the distribution of the percentage values for all evaluation criteria and for the final publication recommendation show, there are clearly more manuscripts rated more favorably by Ra than by Re than there are manuscripts rated more favorably by Re than by Ra. For instance, 22% of the final publication recommendations made by Ra are more positive than those made by Re. There are more positive recommendations by Re than by Ra for only 11% of the manuscripts (there is no difference between the recommendations by the two reviewer groups for 67% of the manuscripts). Hence, overall for this group of manuscripts Ra rated more favorably than Re more frequently than vice versa. Using the marginal homogeneity test, we examined whether the ratings by Ra and Re also differed statistically significantly. As the results of the test in Table 2 show, the difference is statistically significant only for the final publication recommendation. The differences between the ratings on the evaluation criteria are non-significant.
The differing results of the marginal homogeneity test could indicate that with the same ratings on all evaluation criteria, Ra tend to make a more positive final publication recommendation than Re. To test this hypothesis, in a further analysis we selected those manuscripts among the 135 manuscripts reviewed by both Ra and Re that were rated the same on all evaluation criteria by both reviewers. This was the case for 18% of the manuscripts (n = 24). Table 3 shows the reviewers' ratings on the evaluation criteria and their final publication recommendations for the 24 manuscripts. Whereas the final publication recommendations by both reviewers were the same for 21 manuscripts, for 3 manuscripts the final publication recommendations by Ra were more favorable than the recommendations by Re. No manuscript received a more favorable final publication recommendation by Re than by Ra.
In closing, we tested differences between the ratings by Ra and Re using ORMs. An ORM was computed for each evaluation criterion and the final publication recommendation. Table 4 presents a description of the dependent and independent variables that were included in the total of four ORMs. The independent variables are “Author-suggested reviewer” (Ra or Re) and the log-transformed citation counts. Table 5 shows the results of the ORMs. For all ORMs the variable “Author-suggested reviewer” has a statistically significant effect in the expected direction: If the review is by Ra, the ratings on all criteria as well as the final publication recommendation are statistically significantly more favorable than the ratings, if the review is by Re – independently of the quality of the reviewed manuscript (measured ex-post using citation counts). To be able to assess the size of the effect of the variable “Author-suggested reviewer” on the ratings, after the ORMs we computed percent changes in expected ratings for a unit increase (from rating by Re to rating by Ra) . As the results in Table 5 show, in reviews by Re ratings can be expected that are between 30% and 42% less favorable than the ratings by Ra.
Compared to most of the studies on potential sources of bias in the manuscript reviewing process published up to now, the present study used an optimized strategy with two independent analysis steps. In both steps there was a control for the scientific impact of the research reported in a manuscript in order to be able to determine – independently of their quality – whether manuscripts that were reviewed by Ra are reviewed more favorably than manuscripts that were reviewed by Re. The results of this study are therefore more solid than the results of most of the studies published up to now that did not control for the scientific impact of manuscripts in the evaluation.
In a first step of analysis, we used a within-manuscript approach. Even though this analysis revealed a statistically significant difference between the reviews by Ra and Re only with regard to the final publication recommendation (and not for the evaluation criteria), there is a tendency in the dataset towards more manuscripts that Ra rate more favorably than Re than the opposite case. In addition, with the same ratings on the evaluation criteria, Ra tends towards a more positive than a more negative final publication recommendation than Re. In a second step of analysis, an ORM was computed. This analysis showed that both for the evaluation criteria and the final publication recommendations, more positive ratings can be expected by Ra than by Re. All in all, the results for the journal ACP agree with the results of other studies (see the introduction section) and indicate that the bias variable “Author-suggested reviewer” has an effect on the reviewing process.
However, even though the results of the study indicate that there are differences between the ratings by Ra and Re, the results should be seen as only an indication of a potential source of bias in the ACP peer review process and not as proof of favoritism of certain manuscripts by Ra. Strictly speaking, solid findings on the existence of biases in peer review processes can be produced only by experimental studies in which the research objects (such as manuscripts) are randomly assigned to a treatment and control group (such as Ra and Re) . As a study of that kind would influence the review process, there is a risk of infringing the rules of good scientific practice, as pointed out by critical commentaries on the study published by Peters and Ceci  (see Behavioral and Brain Sciences, 1982, pp. 196–246, and Behavioral and Brain Sciences, 1985, pp. 743–747). In that study manuscripts with fictitious author names and institutional affiliations were submitted to journals for publication.
Regardless of what the results of experimental studies of that kind would be, we can probably assume that there can be no peer review system without the influence of potential sources of bias. Scientists, too, are only human: “Philosophers and sociologists agree that the notion of a truly objective disinterested ‘seeker after truth’ is incompatible with the realities of social existence. We all have personal interests and institutional values that we are bound to promote in our scientific work … It will surely defend objectivity as an ideal, impossible to realize completely in practice but always to be respected and desired”  (p. 754). To obtain an indication of the systematic influence of sources of bias in a peer review process, in research evaluation it is proposed that the process of peer reviewing should be studied continuously and that any evidence of bias in the process should be brought to the attention of the editor for correction and modification of the process , . Hojat, Gonnella, and Caelleigh  demanded “that the journal editors conduct periodic internal and external evaluations of their journals' peer review process and outcomes” (p. 75) to assure the integrity of the process. In the most comprehensive review of research on biases in peer review, Godlee and Dickersin  also concluded that “journals should continue to take steps to minimize the scope for unacceptable biases, and researchers should continue to look for them” (p. 112).
If indications of the effect of sources of bias are found in a peer review process, Thorngate, Dawes, and Foddy  recommend the following measures “to fix the problem … One possible solution is to replace biased judges with neutral ones. Another is to train and to motivate offending judges to mend their judgmental ways. A third is to add more judges in hopes that their biases will counterbalance each other and produce a neutral group consensus. Each is worthy of brief consideration” (p. 55). This study showed, in agreement with all other studies, for the bias variable investigated that independently of the quality of a manuscript, better ratings can be expected from Ra than from Re. Many journals use precautions to avoid biased review from Ra, e.g., by stipulating that reviewers do not work in the same institution, have never published with them, etc. If reviewers have a disqualifying conflict they should excuse themselves or not be used. However, personal relationships are harder to quantify than financial links so they are often overlooked. Journal editors should therefore consider, if Ra are used, bringing in more than one Re for the review process so that the review by Ra can be put in perspective.
We would like to thank Dr. Hanna Joos (at the Institute for Atmospheric and Climate Science of ETH Zurich, Switzerland) and Dr. Hanna Herich (at EMPA, a research institution within the ETH Domain) for the investigation of the manuscripts rejected by Atmospheric Chemistry and Physics and published elsewhere. We thank Dr. Ulrich Pöschl, Chief Executive Editor of Atmospheric Chemistry and Physics, the Editorial Board of Atmospheric Chemistry and Physics, and Copernicus Publications (Göttingen, Germany) for permission to conduct the evaluation of the selection process of the journal, and thank the members of Copernicus Systems + Technology (Berlin, Germany) for their generous technical support during the carrying out of the study. We also thank Dr. Werner Marx and Dr. Hermann Schier of the Central Information Service for the institutes of the Chemical Physical Technical (CPT) Section of the Max Planck Society (located at the Max Planck Institute for Solid State Research in Stuttgart, Germany) for conducting the citation search for citations of the accepted and rejected (but published elsewhere) manuscripts in the literature database Chemical Abstracts. The authors wish to express their gratitude to Liz Wager for her helpful comments.
Conceived and designed the experiments: LB. Performed the experiments: LB. Analyzed the data: LB. Contributed reagents/materials/analysis tools: LB. Wrote the paper: LB HDD.
- 1. Weller AC (2002) Editorial peer review: its strengths and weaknesses. Medford, NJ, USA: Information Today, Inc.
- 2. Arkes HR (2003) The nonuse of psychological research at two federal agencies. Psychological Science 14: 1–6.
- 3. Jayasinghe UW (2003) Peer review in the assessment and funding of research by the Australian Research Council. Greater Western Sydney, Australia: University of Western Sydney.
- 4. Lee K, Boyd E, Bero L (2004) A look inside the black box: a description of the editorial process at three leading biomedical journals; Ottawa, Canada
- 5. Tonks A (1995) Reviewers chosen by authors. British Medical Journal 311: 210.
- 6. European Association for Chemical and Molecular Sciences (2006) Ethical guidelines for publication in journals and reviews. Brussels, Belgium: European Association for Chemical and Molecular Sciences (EuCheMS).
- 7. Perlman D, Dean E (1987) The wisdom of Salomon: avoiding bias in the publication review process. In: Jackson DN, Rushton J, editors. Scientific excellence Origins and assessment. London, UK: Sage. pp. 204–221.
- 8. Anon (2007) Gatekeepers of science. Interview with Peter Stern, editor at Science magazine. BIF Futura 22: 14–18.
- 9. Schroter S, Tite L, Hutchings A, Black N (2006) Differences in review quality and recommendations for publication between peer reviewers suggested by authors or by editors. JAMA 295: 314–317.
- 10. Grimm D (2005) Suggesting or excluding reviewers can help get your paper published. Science 309: 1974.
- 11. Scharschmidt BF, Deamicis A, Bacchetti P, Held MJ (1994) Chance, concurrence, and clustering - analysis of reviewers recommendations on 1,000 submissions to the Journal of Clinical Investigation. Journal of Clinical Investigation 93: 1877–1880.
- 12. Earnshaw JJ, Farndon JR (2000) A comparison of reports from referees chosen by authors or journal editors in the peer review process. Annals of the Royal College of Surgeons of England 82: 133–135.
- 13. Goldsmith LA, Blalock E, Bobkova H, Hall RP (2005) Effect of authors' suggestions concerning reviewers on manuscript acceptance. In: Rennie D, Godlee F, Flanagin A, Smith J, editors. Chicago, IL, USA: 5th International Congress on Peer Review and Biomedical Publication.
- 14. Wager E, Parkin E, Tamber P (2006) Are reviewers suggested by authors as good as those chosen by editors? Results of a rater-blinded, retrospective study. BMC Medicine 4: 13.
- 15. Rivara FP, Cummings P, Ringold S, Bergman AB, Joffe A, et al. (2007) A comparison of reviewers selected by editors and reviewers suggested by authors. Journal of Pediatrics 151:
- 16. Bornmann L, Daniel H-D (2009) Reviewer and editor biases in journal peer review: an investigation of manuscript refereeing at Angewandte Chemie International Edition. Research Evaluation 18: 262–272.
- 17. Jayasinghe UW, Marsh HW, Bond N (2003) A multilevel cross-classified modelling approach to peer review of grant proposals: the effects of assessor and researcher attributes on assessor ratings. Journal of the Royal Statistical Society Series a-Statistics in Society 166: 279–300.
- 18. Bailey CW (2005) Open Access Bibliography. Washington, DC, USA: Association of Research Libraries.
- 19. Pöschl U (2010) Interactive open access publishing and peer review: the effectiveness and perspectives of transparency and self-regulation in scientific communication and evaluation. Liber Quarterly 19: 293–314.
- 20. Koop T, Pöschl U (2006) Systems: an open, two-stage peer-review journal. The editors of Atmospheric Chemistry and Physics explain their journal's approach. Retrieved 26 June 2006, from http://www.nature.com/nature/peerreview/debate/nature04988.html.
- 21. Bingham CM, Higgins G, Coleman R, Van Der Weyden MB (1998) The Medical Journal of Australia Internet peer-review study. Lancet 352: 441–445.
- 22. Pöschl U (2004) Interactive journal concept for improved scientific publishing and quality assurance. Learned Publishing 17: 105–113.
- 23. Bornmann L, Daniel H-D (2009) The luck of the referee draw: the effect of exchanging reviews. Learned Publishing 22: 117–125.
- 24. Bornmann L, Daniel H-D (2010) Reliability of reviewers' ratings at an interactive open access journal using public peer review: a case study on Atmospheric Chemistry and Physics. Learned Publishing 23: 124–131.
- 25. Bornmann L, Marx W, Schier H, Thor A, Daniel H-D (2010) From black box to white box at open access journals: predictive validity of manuscript reviewing and editorial decisions at Atmospheric Chemistry and Physics. Research Evaluation 19: 81–156.
- 26. Bornmann L, Neuhaus C, Daniel H-DThe effect of a two-stage publication process on the Journal Impact Factor: a case study on the interactive open access journal Atmospheric Chemistry and Physics. Scientometrics.
- 27. Budden AE, Aarssen L, Koricheva J, Leimu R, Lortie CJ, et al. (2008) Response to Whittaker: challenges in testing for gender bias. TRENDS in Ecology & Evolution 23: 480–481.
- 28. Laband DN, Piette MJ (1994) Favoritism versus search for good papers: empirical evidence regarding the behavior of journal editors. Journal of Political Economy 102: 194–203.
- 29. Smart S, Waldfogel J (1996) A citation-based test for discrimination at economics and finance journals. Cambridge, MA, USA: National Bureau of Economic Research. NBER working Paper, No 5460.
- 30. Jayasinghe UW, Marsh HW, Bond N (2001) Peer review in the funding of research in higher education: the Australian experience. Educational Evaluation and Policy Analysis 23: 343–346.
- 31. Agresti A (2002) Categorical data analysis. Hoboken, NJ, USA: John Wiley & Sons, Inc.
- 32. Cytel Software Corporation (2010) StatXact: version 9. Cambridge, MA, USA: Cytel Software Corporation.
- 33. Long JS, Freese J (2006) Regression models for categorical dependent variables using Stata. College Station, TX, USA: Stata Press, Stata Corporation.
- 34. StataCorp (2009) Stata statistical software: release 11. College Station, TX, USA: Stata Corporation.
- 35. Nichols A (2007) Causal inference with observational data. Stata Journal 7: 507–541.
- 36. Rabe-Hesketh S, Skrondal A (2008) Multilevel and longitudinal modeling using Stata. College Station, TX, USA: Stata Press.
- 37. Brant R (1990) Assessing proportionality in the proportional odds model for ordinal logistic regression. Biometrics 46: 1171–1178.
- 38. Long JS (1997) Regression models for categorical and limited dependent variables. Thousand Oaks, California, USA: Sage.
- 39. van Raan AFJ (1996) Advanced bibliometric methods as quantitative core of peer review based evaluation and foresight exercises. Scientometrics 36: 397–420.
- 40. Lindsey D (1989) Using citation counts as a measure of quality in science. Measuring what's measurable rather than what's valid. Scientometrics 15: 189–203.
- 41. Craig ID, Plume AM, McVeigh ME, Pringle J, Amin M (2007) Do open access articles have greater citation impact? A critical review of the literature. Journal of Informetrics 1: 239–248.
- 42. Cole S, Fiorentine R (1991) Discrimination against women in science: the confusion of outcome with process. In: Zuckerman H, Cole JR, Bruer JT, editors. The outer circle Women in the scientific community. London, UK: W W Norton & Company. pp. 205–226.
- 43. Shadish WR, Cook TD, Campbell DT (2002) Experimental and quasi-experimental designs for generalized causal inference. Boston, MA, USA: Houghton Mifflin Company.
- 44. Peters DP, Ceci SJ (1982) Peer-review practices of psychological journals - the fate of accepted, published articles, submitted again. Behavioral and Brain Sciences 5: 187–195.
- 45. Ziman J (1996) Is science losing its objectivity? Nature 382: 751–754.
- 46. Geisler E (2001) The mires of research evaluation. The Scientist 15: 39.
- 47. Bornmann L, Mutz R, Daniel H-D (2008) How to detect indications of potential sources of bias in peer review: a generalized latent variable modeling approach exemplified by a gender study. Journal of Informetrics 2: 280–287.
- 48. Hojat M, Gonnella JS, Caelleigh AS (2003) Impartial judgment by the “gatekeepers” of science: fallibility and accountability in the peer review process. Advances in Health Sciences Education 8: 75–96.
- 49. Godlee F, Dickersin K (2003) Bias, subjectivity, chance, and conflict of interest. In: Godlee F, Jefferson J, editors. Peer review in health sciences. London, UK: BMJ Publishing Group. pp. 91–117. 2nd ed.
- 50. Thorngate W, Dawes RM, Foddy M (2009) Judging merit. New York, NY, USA: Psychology Press.