According to Merriam-Webster’s online dictionary, “to plagiarize” is defined as “to use the words or ideas of another person as if they were your own words or ideas.” A key point of plagiarism is the phrase “of another person,” which means that the concept of “self-plagiarism” doesn’t fit into this definition.
So let me give you my definition of self-plagiarism: reusing a lot of words from your already-published paper. That is, for me, self-plagiarism really isn’t plagiarism but repeating already-used strings of words. It is okay to have the same meaning and intent, but when most of a paragraph is identical to your paper published a year ago, then I will ask you to rewrite it.
AGU uses a service called iThenticate to generate “similarity reports,” a document that highlights strings of characters that match those from already-published papers. They maintain a huge database across many publishers; it is essentially impossible to get away with verbatim copying from another paper, yours or someone else’s. Before contacting potential reviewers, the JGR-Space Physics editors check the similar report of the submitted manuscript and make a judgment call on whether to reject it without review. I cannot give you an exact percentage overlap that warrants rejection because it depends on the concentration of the overlap (identical phrases sprinkled throughout the text is fine, but two consecutive full paragraphs is not) and the location of the overlap (in the methodology might not warrant rejection, but significant overlap in the discussion section would). So, we do this on a case-by-case basis. Most manuscripts have overlap numbers of 5-10%, with perhaps at most 2-3% overlap with any one paper. A value as small as 5% could require immediate revision while 35% overlap might be just fine.
The iThenticate company has a guide to self-plagiarism available on the web:
In short, please don’t take the easy way out and simply cut-and-paste text from an existing paper. Even for methodology sections that contain the same essence as several prior papers, write it fresh with every paper you submit for publication.