Bibliography Notes

by Alex Nelson, 25 September 2026

The “toy model” of the Zettelkasten work flow looks something like:

  1. Read article (or book) and take “bibliography notes”
  2. Convert “bibliography notes” into “permanent notes”
  3. Insert “permanent notes” into Zettelkasten

Although this sounds simple, there are some stumbling points with even this simplified process…namely:

How to make a bibliography note

tl;dr How you should make bibliography notes depends on what you are trying to accomplish with your Zettelkasten.

Usually, people abbreviate discussion of “How to construct bibliography notes” to something like, “When reading a text, write notes down as a bibliography note.” The problem with giving any guidance more specific than this is that it depends on (a) what text you are reading, and (b) why you are taking notes.

Nevertheless, people usually adopt one of the following strategies:

Strategy 1. When reading a text, it is natural to write down one’s own observations about the text. A common strategy is to summarize sections or arguments in the text. In other words, one writes down notes to remind one’s self about what the text says.

Strategy 2. When reading a text, what is it not saying? What is excluded when the text makes an assertion? The text selects what is included and thereby implicitly what is not. We are observing how the text makes observations. Making terse notes about these observations constitutes the bibliography notes, which are “grist for the mill” (i.e., used as raw material when writing permanent notes, which expand on these observations).

Strategy 3. I have a particular end-goal in mind (e.g., I am writing a book about proof assistants), and the text provides a unique perspective on some aspect of my end-goal. The notes I take are observations about this. (In reality, this is just strategy 2 repackaged in a particular situation.)

…

So which strategy is “correct”? Well, it depends on how you intend to use the bibliography notes.

There are some texts which I will not have access to again (because they are rare artifacts, or just difficult to obtain articles from out-of-print journals). In this situation, I tend to follow the first strategy: write down a reformulation of (or even a verbatim transcript of passages from) the text.

When I am trying to generate something insightful (as opposed to preserving existing knowledge), I tend to follow something resembling the second strategy.

What do I do?

I realized that I tend to take rather comprehensive notes because I tend to try to form a “rational reconstruction” (in Carnap’s sense of explication) of a given text.

When reading, e.g., Kant’s Critique of Pure Reason, my notes tend to resemble a formalization of the text in first-order logic (or possibly some fragment of higher-order logic). I tend to form a dictionary translating Kant’s epistemology into modern cognitive science terminology.

When reading Mathematics, there are two situations I find myself in:

  1. I am learning new Mathematics. In this situation, I write a lot of scratch work, try decomposing proofs and reconstructing them. The definitions are re-written to fit a standard template. These are necessary for learning Mathematics, but inappropriate for “bibliography notes”
  2. I am reading research Mathematics. In this case, I am trying to extract exercises from the text, I am making note of cited material I should lookup, summarizing what is actually new from what has already been done in other papers, thinking of examples and non-examples, etc.

Reading theoretical Physics is very similar, I tend to focus on extracting exercises and seeing if this is a topic which can be an entry in my encyclopedia.

With Computer Science, Keshav’s paper discusses how to to read a research paper. Keshav advocates reading research papers in three passes:

  1. The first pass skims through the paper. Read the introduction, skim the section titles, skim the Mathematics (if any), then read the conclusion. At this point, you should know what research question the paper addresses, what its contributions are, what background material (previously published papers in the literature) does it build on, and—perhaps most important—whether it is worth spending additional time reading (i.e., commit ourselves to the second pass).
  2. The second pass determines the purpose of the paper, taking about an hour. Importantly, during this step, write down a list of cited papers which you have not read (and why the paper [you are reading] cites them) for future reading. Read through this paper a bit more carefully. You should be able to give an “elevator pitch” of the paper (what’s the problem, what’s been done before, and what’s new in the paper). At the end of the second pass, you now need to decide to (a) set the paper aside and not continue with it or the cited material, (b) return to the paper after becoming familiar with background material, or (c) continue on to the third pass.
  3. The third pass attempts to “virtually re-implement the paper”. If you’ve noticed a theme here, you’ll realize this is precisely a “rational reconstruction” of the paper. What are the implicit assumptions made? Are all the hypotheses necessary? Can the key insights or innovations be applied to other fields or problems? Are there any limitations of the results which the paper does not discuss? Are there any avenues for extension?

At the end of the first pass, I have a one or two sentence summary of the paper. At the end of the second pass, I have some outline of the “argument” or contribution of the paper. At the end of the third pass, I have a more complete rational reconstruction of the paper.

You’ll notice that the upshot of this three-pass reading system is to form a rational reconstruction of the paper.

How did Luhmann use it?

After all, Luhmann points out that (Social Systems (1995) Princeton University Press, p.140; italics are Luhmann’s, bold is mine):

If one begins with the concept of meaning, it is clear from the start that communication is always a selective occurrence. Meaning allows no other choice than to choose. Communication grasps something out of the actual referential horizon that it itself constitutes and leaves other things aside. Communication is the processing of selection. Of course, it does not select in the same way in which one grabs one thing or another off the rack. That would bring us back to the substance theory and the transmission metaphor. The selection that is actualized in communication constitutes its own horizon; communication constitutes what it chooses, by virtue of that choice, as a selection, namely, as information. What is uttered is not only selected, but also already a selection—that is why it is uttered. Therefore communication must be viewed not as a two-part, but as a three-part selection process. It is not just a matter of sending and receiving with selective attention on both sides; instead, the selectivity of the information is itself an aspect of the communication process, because selective attention is actualized only in reference to the very selectivity of information. Selectivity as such attracts further communication: it recruits communications that direct themselves to aspects that selectivity has excluded.

What Luhmann means when he writes, “Communication grasps something out of the actual referential horizon that it itself constitutes and leaves other things aside,” is that communication is a distinction: something is being said, and something is left un-said.

This is important because Luhmann’s “bibliography notes” appears to be Luhmann’s “second-order observations” about what is being left “un-said”, and how it informs his Zettelkasten.

See also: