"Classical peer review" has been subject to intense criticism for slowing down the publication process, bias against specific categories of paper and author, unreliability, inability to detect errors and fraud, unethical practices, and the lack of recognition for unpaid reviewers. This paper surveys innovative forms of peer review that attempt to address these issues. Based on an initial literature review, we construct a sample of 82 channels of scientific communication covering all forms of review identified by the survey, and analyze the review mechanisms used by each channel. We identify two major trends: the rapidly expanding role of preprint servers (e.g., ArXiv) that dispense with traditional peer review altogether, and the growth of "non-selective review," focusing on papers' scientific quality rather than their perceived importance and novelty. Other potentially important developments include forms of "open review," which remove reviewer anonymity, and interactive review, as well as new mechanisms for post-publication review and out-of-channel reader commentary, especially critical commentary targeting high profile papers. One of the strongest findings of the survey is the persistence of major differences between the peer review processes used by different disciplines. None of these differences is likely to disappear in the foreseeable future. The most likely scenario for the coming years is thus continued diversification, in which different review mechanisms serve different author, reader, and publisher needs. Relatively little is known about the impact of these innovations on the problems they address. These are important questions for future quantitative research.
Keywords: anonymity; impact metrics; interactive review; non-selective review; open access; open peer review; peer review; preprint servers.