What duplicate publication is
Duplicate or redundant publication means publishing the same work, or substantially the same work, more than once without disclosure and without a clear reason. It distorts the literature: the same finding appears to have been established twice, and meta-analyses may count it twice.
The concealment is what makes it a breach. Legitimate secondary publication does exist — a translation for a different language readership, for example — but it is agreed with both editors in advance and carries a clear statement of the original.
Simultaneous submission belongs in the same family. Sending one manuscript to two journals at once wastes reviewer effort and risks two journals publishing the same paper. Essentially every reputable journal prohibits it outright.
The breach is publishing the same work twice without disclosure. Disclosed secondary publication, agreed with both editors, is a normal and accepted thing.
Self-plagiarism and text reuse
Reusing your own sentences is a narrower issue. In the methods section it is often unavoidable and widely tolerated — there are only so many ways to describe a standard protocol, and rewording it for the sake of novelty can make it less precise.
Reusing your own introduction or discussion wholesale is different. That text represents argument rather than procedure, and lifting it verbatim into a new paper implies fresh thinking where there is none.
Similarity software cannot make this distinction. A high similarity score against your own earlier papers is not automatically a problem, and a report needs reading rather than obeying — which is exactly the judgement a raw percentage cannot supply.
Salami slicing
Salami slicing means dividing a single coherent study into the smallest publishable units to inflate output. It is judged by whether the pieces answer genuinely separate questions or whether the split is artificial.
A large study producing several papers is entirely normal where each addresses a distinct question with its own analysis. The same study split so that a single result is spread across three papers, each incomplete, is not.
The test editors apply is whether a reader would be misled by seeing one piece without the others. If the answer is yes, the pieces belong together. If you are unsure, describe the split to the editor at submission — asking is free, and being asked afterwards is not.