Background: Peer review is assumed to improve the quality of research reports as tools for scientific communication, yet strong evidence that this outcome is obtained consistently has been elusive. Failure to distinguish between aspects of discipline-specific content and aspects of the writing or use of language may account for some deficiencies in current peer review processes.
Discussion: The process and outcomes of peer review may be analyzed along two dimensions: 1) identifying scientific or technical content that is useful to other researchers (i.e., its "screening" function), and 2) improving research articles as tools for communication (i.e., its "improving" function). However, editors and reviewers do not always distinguish clearly between content criteria and writing criteria. When peer reviewers confuse content and writing, their feedback can be misunderstood by authors, who may modify texts in ways that do not make the readers' job easier. When researchers in peer review confuse the two dimensions, this can lead to content validity problems that foil attempts to define informative variables and outcome measures, and thus prevent clear trends from emerging. Research on writing, revising and editing suggests some reasons why peer review is not always as effective as it might be in improving what is written.
Summary: Peer review could be improved if stakeholders were more aware of variations in gatekeepers' (reviewers' and editors') ability to provide feedback about the content or the writing. Gatekeepers, academic literacy researchers, and wordface professionals (author's editors, medical writers and translators) could work together to discover the types of feedback authors find most useful. I offer suggestions to help editologists design better studies of peer review which could make the process an even stronger tool for manuscript improvement than it is now.