Hypertext per se does not favor coherence. In fact, it accommodates incoherence very well. That is one of the reasons why it is good for note-taking where you often have no idea what it is that you are going to argue for or what claims will actually be supported by the notes being collected. It is very much like using index cards that way.
On the other hand, hypertext per se does not hinder or prevent coherence either. You can order and re-order the information according to various criteria. And, in a program like ConnectedText you can use categories or tags just as much as you can use outlines. This is again very much like using index cards which can be sorted into different heaps, which can then be organized into a (more) coherent structure.
The beauty of hypertextual note-taking is that it actually does not favor any particular coherent structure and facilitates re-ordering or re-thinking the notes in accordance with different criteria or goals. This is one of its special strengths. It is not unrelated to its ability to tolerate incoherence.
Put differently, hypertext does not per se differentiate between "signal" and "noise." What is "signal" in one context, may just be "noise" in another (and vice versa). That lack of cognitive overhead is (still) one of the reasons why I find applications like Zettelkasten less useful as a desktop wiki like ConnectedText. The main way of connecting information in the Zettelkasten program is keywords which necessitate a decision about what is important and what is unimportant when you enter the information. In order to automatically connect the notes belonging together thematically (thematisch zusammenhängende Zettel) you need to decide right away which words in the notes are significant and which ones are not, thus superimposing a kind of system on what is supposed to be—or perhaps better: is best treated—as raw information.
1. The Website and the Program are written in German. I would go amiss, if I did not point out that the program has been improved a great deal since I last wrote about it. It now also allows for manual linking of different entries, for instance. Furthermore, it now is available for Windows, Unix and the Mac.
2. I am aware that this distinction between the raw and the processed (or data and theory) is far from mutually exclusive in this context. Indeed, it is just because I am aware that our preconceptions have a tendency to influence what we consider as data that I am favoring hypertext over keywords. It is not that it eliminates bias completely, but it does not favor it either.