When an author submits a manuscript to be published in a subscription journal, unless expressly stated otherwise, the publisher expects the manuscript to be original work that will make a new contribution to the field of study. Many publishers, as part of the publishing agreement, ask authors to confirm that the work is original, has not been previously published, and that the author is the original copyright owner and is therefore able to transfer copyright to the publisher. See the sample publisher agreements from Taylor & Francis and Elsevier for examples.
If an author signs such an agreement to have work published in a subscription journal, but the new work contains previously-published work by the same author that isn't cited, the publisher may consider that the author has self-plagiarized and has therefore violated the publishing agreement. There are also copyright implications; if the author no longer owns the copyright to the earlier work, the author can't reuse the work without obtaining copyright permission from the publisher.
Here's what the American Psychological Association (2010) recommends to authors regarding self-plagiarism:
Just as researchers do not present the work of others as their own (plagiarism), they do not present their own previously published work as new scholarship (self-plagiarism). There are, however, limited circumstances (e.g., describing the details of an instrument or an analytic approach) under which authors may wish to duplicate without attribution (citation) their previously used words, feeling that extensive self-referencing is undesirable or awkward. When the duplicated works are limited in scope, this approach is permissible. When duplication of one’s own words is more extensive, citation of the duplicated words should be the norm. What constitutes the maximum acceptable length of duplicated material is difficult to define but must conform to legal notions of fair use. The general view is that the core of the new document must constitute an original contribution to knowledge, and only the amount of previously published material necessary to understand that contribution should be included, primarily in the discussion of theory and methodology. When feasible, all of the author’s own words that are cited should be located in a single paragraph or a few paragraphs, with a citation at the end of each. Opening such paragraphs with a phrase like “as I have previously discussed” will also alert readers to the status of the upcoming material. (p. 16)
American Psychological Association. (2010). Publication manual of the American Psychological Association (6th ed.). Washington, DC: Author.