The examination of a single scientific manuscript seldom alerts scientists, reviewers, editors, and scientific administrators to the fabrication and falsification of data and information. This review shows that most documented cases of scientific fraud involve falsification (altering truthful information) and fabrication (inventing information where none previously existed). Plagiarism is much less frequent. The review of published accounts also shows that the publication of scientific papers containing recognizable fraudulent material is very low, probably less than 0.02% and extremely difficult to detect. Because most reported cases of fraud have involved research done at prestigious organizations with distinguished co-authors, and that is published in journals with exacting review processes, it becomes evident that some unscrupulous scientists are adept at fabricating and falsifying data. However, "significant" scientific fraud is detected when scientists repeatedly report results that cannot be independently verified, when colleagues report suspicious behavior, or scientific audits are performed. This review documents and compares many of the better-known cases of scientific fraud. Fraudulent behavior has served as the impetus for the scientific community to develop publication procedures and guidelines that help to guard against not only fraudulent behavior but also against other types of unethical or undesirable behaviors. A companion paper reviews the non-fraudulent issues associated with scientific publication.