How should one cite previous work of his own? If one paper relies on previous work that author already did I suppose that paper should be cited. Given that one paper is a continuation of the one before should the papers that are cited there as previous work be cited as-well?
If I take your question literally, then this would be my answer:
I citation serves two purposes:
- You avoid that you take credit for work by others, and thus avoid plagiarism. This point also applies to your own work published somewhere else.
- It can act like an "external appendix" (e.g. the interested reader can read the proof in...).
So, if you make a statement in your current article that relies on findings in your previous articles, then you should cite yourself. If your previous articles contain information you would otherwise have put in an (web)appendix, then you can cite yourself. Otherwise, you should not cite yourself.
However, I suspect you want to know how far back you should go when writing the section "previous research". Here the answer is it depends on what is considered normal in your (sub-)discipline. At the very least I would look at that section from the perspective of a reader: Can they understand how you want to place your article within the research that has been done in this area? Discussing the creation of the computer is typically not necessary to achieve that goal...
A paper should be largely self-contained in its citations. Cite everything that is directly needed to understand the context of the current paper. This will typically mean that there are many citations shared between papers, but that is OK.
Let me illustrate by means of three examples. Let's say that you have already written paper A, which presents a method for making widgets. A year later, you write paper B, building on the work.
Case 1: Paper B is about a better method for making widgets: In this case, you need to give the context for addressing the widget-making problem. You should not assume that the reader has read paper A, so the citations of paper B should likely be nearly identical, with the addition of paper A. The only missing will be citations that were supporting the method in paper A but that do not apply to paper B (e.g., an analytical technique used for one but not the other).
Case 2: Paper B is about applications of widgets: In this case, you need to keep the citations supporting why anybody should care about widgets, but can drop all of the citations on alternate ways of making widgets. It is enough to say, "we make the widgets using the method in [cite]", if the paper isn't about making them.
Case 3: Paper B is about using the same method to make gizmos instead of widgets: In this case, you can drop all of the citations motivating care about widgets, since this is about gizmos instead. You will still need citations to compare this method to alternate methods of making gizmos (which might be the same citations as for widgets, depending).
Thus, you may see that the number of shared citations depends on the type of relationship between the papers, because ultimately it boils down to making sure the reader doesn't have to look up another paper in order to find the critical pieces of related work for the one that they are currently reading.