Dividing research findings into the smallest publishable units might increase an investigator's total number of publications, but works against the interests of science. Minimally, this is an inefficient use of scarce resources, including space in journals and the time of authors, editors, and reviewers. Furthermore, fragmentation of one study into many small publications can give the false impression that a line of research has been extensively pursued. Even so, a scientist's reputation is likely to be more respected if based on a few widely respected studies rather than on many small, fragmentary reports. No simple formula can determine the point at which a body of data warrants publication, but scientific literature benefits from the publication of manuscripts that represent new and substantial findings.

Publication of data in more than one location gives the findings more visibility, but it may also mislead readers into believing that the publications represent distinct data sets. In the case of clinical findings, this could contribute to a false impression that some greater number of patients have been studied. In the case of basic research, readers could be misdirected into thinking that the study had been successfully replicated. Any data set, either in whole or in part, should not be published twice without making explicitly clear which of the data have been published previously and where and when the work was published.

Some exceptions to the injunction against redundant publication may be justified. It's acceptable to publish a paper that was published previously only as an abstract, although authors should disclose the prior publication and avoid the suggestion that the two represent distinct results. In some circumstances, the case can be made that two completely different audiences can be reached only by separate publications: for example, when a work warrants publication in two languages. Even in such cases, however, the editor and the publisher for both papers must approve the arrangement. In all of these cases, it is essential that the later publication make clear reference to the earlier work. Without such reference, the second publication constitutes a falsification of the research record.

For the submission of papers, most journals require that the work not be submitted simultaneously elsewhere for consideration. Submission of a paper is tantamount to provisionally giving the selected journal copyright to the work, and it initiates considerable expense of time and effort in reviewing the manuscript. Only when an article has been rejected by or withdrawn from consideration in one journal may it be submitted elsewhere.

Authors take responsibility for both the ideas and words in a publication. For this reason, co-opting an entire manuscript is a clear example of research misconduct. While taking credit for someone else's research findings is clearly a greater wrong than copying their methods section, both are examples of plagiarism-- taking personal credit for someone else's words or ideas. To use the words of another author, either state where the original words can be found or reproduce the original text with clear and well-cited attribution to the original author. Even with proper citation, repeating the words of other authors is constrained by the fair use provisions of copyright law.

A difficult case of plagiarism is when it occurs between colleagues. When text has been co-authored, the question of ownership may be a matter of dispute. In some research groups, jointly written text is assumed to be shared property available for use by any of the original authors. Opinions about the extent of collective ownership of jointly written materials varies and it may be impractical to determine when plagiarism has occurred (see the Office of Research Integrity, 1994). It should never be assumed that it is acceptable to take credit for words written by someone else, and this issue should be openly addressed among collaborators.

Self-plagiarism refers to authors who reuse their own previously disseminated content and pass it off as a ”new” product without letting the reader know that this material has appeared previously. Self-plagiarism is often described in the context of several distinct practices in which some or all elements of a previous publication (e.g., text, data, and images) are reused in a new publication with ambiguous acknowledgement or no acknowledgement at all as to their prior dissemination. Self-plagiarism misleads the reader about the novelty of the material.

The references cited in a research publication form the path that connects new work with the work on which it is built. Just as a thorough reference search is the foundation for responsible research, appropriate reference citation is the foundation for responsible reporting. Because future readers will rely on the references listed, an author has responsibilities to assure the accuracy of each citation so that readers can locate the referenced work, to include adequate references that document the origins of ideas, to verify that referenced works are consistent with the ideas and information credited to them, and to cite original sources whenever possible.

Statistical methods are often used to describe data samples, to summarize results and relationships, and to test hypotheses. Usually, readers will not have access to the raw data and, therefore, will base any conclusions on the outcome of the statistical analyses. Because the assumptions and meaning of statistical tests vary widely, simply reporting a final P value (the probability of a false positive) tells readers very little unless they also know the methods of data collection and analysis. For these reasons, it is essential that authors not only design and analyze experiments appropriately, but also that they clearly and openly describe their methods (American Statistical Association, 1999).

The decision to submit a correction or retraction is not an easy choice to make, and all authors of the paper should be informed. Despite any perceived risks, there are also advantages to an appropriate correction or retraction. Admitting error is typically perceived as a sign of integrity and concern for the highest standards in work entered into the published record. Conversely, failure to admit to an error can be devastating if the problem with the manuscript is first discovered by others. The integrity of the scientific literature is best served by rapid correction of misleading or mistaken information.

If errors are discovered after a manuscript has been published, then authors have several options, depending on the significance of the errors: (1) If minor errors are found to have been included in a manuscript, then a letter (singular erratum or plural errata) describing the errors should be submitted to the journal that published the article. (2) If unintentional errors are great enough to undermine part of a report, then the authors should submit a letter to the journal explaining the errors as a correction to the publication. (3) If unintentional errors are of such a magnitude as to invalidate or seriously undermine the entire report or if misconduct affecting the work on the part of one or more authors is found to have occurred, then the authors should retract the paper by writing to the editor of the publication.