When Should You Start a New Note?

System

When Should You Start a New Note?

The Zettelkasten note-taking method has made book writing and writing scientific papers easy for hundreds of years already.

Read the full story here

Garwyx

Luhmann made use of child and parent notes, i.e. note sequences. This is different from links, which are more loose. How do you use the parent-child concept in your notes?

Sascha

Why should sequences be tighter than links in a digital Zettelkasten?

Garwyx

Because that creates a story and indicates direction. Certain notes follow others, not the other way around, whereas links create nesting rather than hierarchy. So if a note is a contradiction or addition to another, it is a child note. But when it loosely relates to the subject it is a link.

Why do you not distinguish between links and parent/child relations like Luhmann did?

Some of my notes have a lot of links, but only a few of those are tight. It does make it more difficult to get started cf this post. Maybe I should just include fewer links?

ctietze

I wrote about the Folgezettel-as-a-story thing here: https://zettelkasten.de/posts/kinds-of-ties/
Sascha picked up the topic of Folgezettel before that in this post: https://zettelkasten.de/posts/luhmann-folgezettel-truth/

My point of linking to both posts in context is that you can implement stories in different ways. Following links around can create a story just as well. Folgezettel are not a principle (e.g. making stories) but just an implementation or realization of the principle.

In fact, if you say links cannot do this, you'd deny the existence of hypertext fiction altogether. FWIW, there'd be quite a few people who don't agree with you about that

Steph

@Garwyx I dunno, I think a link is a link, and child notes are links like any other. The only difference for me is that if it is a true child note then the link comes at the end of the first zettel, like leading me on through a continued topic, and if it is a looser link then the link may only be found in the 'linkee' zettel (i.e. the 'linked-to' zettel), rather than pointing the way onwards from the first one. Some links will be bi-directional and some one-way.

I would also advise only to make a link when you have described it a bit in words (e.g. 'reminds me of xxxx', 'this implies xxxx is xxxx', 'see xxxx for a similar xxxxx') cos otherwise when you see it again you might not remember what caused you to link them in the first place.

Steph

@Garwyx actually now I am confused cos we have essentially said the exact same thing but drawn different conclusions from it lol.

Perhaps what I do differently from you is that you say 'So if a note is a contradiction or addition to another, it is a child note'. What I do instead is to go back and make a new or different structure note, for the original topic which is disagreed about, with e.g. an explanation saying e.g.

'xxx is a this that and the other. Theories about the origin or implications of xxxx include:
1. [[ link to first one, theory yyyyyy]]
2. [[link to second one, theory zzzzzzz]].
etc

In the first link, I would introduce it as 'Theory yyyy is one explanation of xxxxxxx'.
Then your full zettel of yyyyyyy.

Then at the end: 'see also theory zzzzzzzz'

So, yy and zz both have the same status. They are both children of the structure note.

Garwyx

@Steph that is exactly what I mean and do. The continued topic is at the top of the note, directly under the title, and other links are in a "# Links" section below the note text.

@ctietze and @Sascha I would like to see/know how you implement links in your notes. I have read many of your blog posts but this still eludes me. In some of my cases they can clearly be footnotes because they link to a topic/word in that note. In other cases they are related but end up in the "# Links" section, which can become crowded.

For instance, I just created the note "A random-effects model is a statistical model in which the model parameters are random variables". The next note I created is "Random effects influence only the variance of y". To me, this can be a child note because it adds to the first note. The first note also connects with the note "Linear models consist of one response variable and one or more explanatory variables". This connection is, to me, much looser.

Btw, thanks @ctietze for reminding me about the discussion concerning the folgezettel. It had slipped by me. Based on that first post, do I understand correctly that you do not necessarily create actual backlinks in notes? Indeed the search function will show the first category anyway when you search for the latter.

Steph

@Garwyx I see what you mean now, and I think part of the confusion is that you and I both originally took the maxim of 'single proposition per zettel' too seriously. The way it looks to me, is that you can fit more info on a single zettel. For example you need a zettel called 'random effects model' where you have both of those statements about random effects model, plus loads more about how random effects are those effects which have variance within the category you are measuring, or which varies on the individual level, or whatever it means (argh mixed effects models are horrible!). Plus an example, diagram, graph, etc. It could be your Random Effects structure note

Garwyx

@Steph I indeed make structure notes (I call them overview notes) that aggregate this knowledge. I really like the idea of core notes containing only one idea. That helps me a lot (see this topic for details).

So should I be less strict about these core notes following each other up and backlinking, and rather just add them to the structure note? The structure note about mixed-effects models can be part of a structure note about types of statistical models, but should the structure notes about all these different models then link to eachother? They are already connected via the structure note (same for the core notes of the random-effects model).

This is starting to make more sense .

Steph

yeah yeah I mean the different stats models can link to each other, but they don't have to, as they are all contained on the next structure note up. But you could make links between them to really clarify the key differences between them.

So there can be many different levels of structure notes, and indeed, with something as dense and solid as statistics, maybe the more structure notes the better.

I didn't put my stats stuff in zettels yet, cos I did multi level models last year before I started zettling, but I imagine it could go like:

Level 1 Stats models

Level 2 Mixed effects models

(which consist of:
-fixed effects [[link to level 3 structure note on all the things you want to store about fixed effects]]

• random effects (I guess this is what we started off talking about) [[link to level 3 structure note]]

Level 3 structure note for e.g. random effects:

• things that are peculiar to random effects, what it means for intercept and slope, e.g. how in a variable with a random effect, each variable has a different value.

But now a problem: where to put intercept and slope? You could start new zettels

• intercept [[link to short new zettel of what an intercept is]
• slope [[new zettel all about slopes]]

But then you get the problem that by themselves the concepts are very simple. They are like atoms of concept but what you really want to get at is how the types, quantities and behaviours of slopes and intercepts characterise the different models. So maybe you need a new structure level called 'Slopes' where you have links called
-behavior of slope in random effects - which is then roughly the same text as you just wrote in your random effect zettel so you might end up not with separate cards for slope and intercept - but it shouldn't matter either way.

How much text on a zettel:
So far I've been aiming for roughly about half as much as there is on this post of mine. So, definitely more than a single sentence or proposition, but aiming to make it as 'atomistic' or single-concept as possible, with rather more structure than less.

But obviously, you'll find what works for you, good luck! x

Sascha

@Garwyx said:
Because that creates a story and indicates direction. Certain notes follow others, not the other way around, whereas links create nesting rather than hierarchy. So if a note is a contradiction or addition to another, it is a child note. But when it loosely relates to the subject it is a link.

Now you are just stipulating. Or in more philosophical terms: You are using prescriptive definitions.

My response is quite simple at this point: At no point, you are justifying that links don't create stories or directions. You are just saying that it is like that, which is not enough.

That in fact, leaves your reasoning open to simple counter examples like Christian's:

In fact, if you say links cannot do this, you'd deny the existence of hypertext fiction altogether.

Why do you not distinguish between links and parent/child relations like Luhmann did?

I don't because I did it and experience desolved this distinction to nothing.

Some of my notes have a lot of links, but only a few of those are tight. It does make it more difficult to get started cf this post. Maybe I should just include fewer links?

How do you define tightness?

Garwyx

@Steph thanks, that helps a lot!

@Sascha said:
Now you are just stipulating. Or in more philosophical terms: You are using prescriptive definitions.
My response is quite simple at this point: At no point, you are justifying that links don't create stories or directions. You are just saying that it is like that, which is not enough.

You're right, and I agree that links can create stories. They don't have to be parent/child relations. I was thinking about how outlines are created, where hierarchy is of utmost importance. While I was writing this, I wanted to argue subsequent paragraphs may be loosely related, but realised that they would probably be related to each other in a parent note (i.e. a structure note).

@Sascha said:
How do you define tightness?

The ones that I consider children and parents are tight. In my example, the definition of random effects follows the definition of a random-effects model (child, i.e. tight link). The random-effects model is a kind of statistical model, so that is probably its parent (i.e. tight link). It is related to the fixed-effects model, so I could include a link to it, but that I would consider a "loose" link. This link might be obsolete because they have the same parent, but I can imagine links to other notes to be relavant, too.

Maybe some of my notes contain so many links because they are not atomicised. I tried to explain very broad concepts, which link to a lot of notes, whereas these notes should actually be structure notes, where the links are the essential part.

Will

@ctietze thanks for the reminder to "Search the archive for similar content". Sadly, I don't always do this. Without doing this I can create duplicate or orphaned notes and I'm not growing my personal knowledge through making connections. I can get too excited about the new that I forget that making connections is the real goal.

How can I develop a habit of looking for similar content when adding a note? Ideas? Prompts? Could the software help? Sort of like how a web browser can be set to show related Evernote notes when conducting a search.

I've from time to time sat down not to add new notes but just to massage the whole Zettelkasten. Search random terms. Create links, Serendipitously find connections. For this I created a pair of keyboard macro. When I do this it feels productive but is it? I'm afraid I'm fooling myself.

ctietze

@Will Yeah, I know, I forget about integration of totally random tips I capture on the web as well.

I was working on a script last year to find orphaned notes. It never worked reliably and I didn't fix it, but will revisit this when The Archive supports user scripting: the idea, then, is to enable you to run this maintenance script from the app regularly to generate prompts for e.g. a weekly review (drawing from Getting Things Done here).

A right away helpful tip: don't use ⌘N to create new notes. If you can, search for a place that would want to link to the note you're about to create, then use the Wiki-link method of creating a link to a missing note first, click on it (so it becomes the search term), then create the new note from there (by hitting enter). This way, things are connected from the start. Maybe not as good as they could be if you looked for more places, but 1 is way better than 0

@Steph's example is interesting. You create notes about methods and extract small concepts/technical terms that are repeated a lot ("intercept" & "slope"), and collect all methods in an overview. Then you notice that this hierarchy doesn't provide a comparative overview. I imagine I want to turn the hierarchy I've created through links that way by 90 degrees, to start a different, orthogonal perspective -- and @Steph is spot on, you can just create a new overview that compared methods by their slope/intercept score, "How to pick a method by slope rating" or so, and link to the term up front, then to each method as you shortly discuss it. I like these moments in processing a lot!

Will

@ctietze said

A right away helpful tip: don't use ⌘N to create new notes. If you can search for a place that would want to link to the note you're about to create, then use the Wiki-link method of creating a link to a missing note first, click on it (so it becomes the search term), then create the new note from there (by hitting enter). This way, things are connected from the start. Maybe not as good as they could be if you looking for more places, but 1 is way better than 0

Yes, this would be a start. Your right, this is better than ZERO connections.

I'm a bit slow and dense here. I've heard you talk about this "Wiki-link method" but for the life of me, I can't grasp it. How does creating a link [[201910190447 BLACK STAR]] become a search term? It does seem that highlighting a term or phrase and pressing a hotkey to surround the term in a link and the starting a note would be cool. But how would that then search for other instances of the same term or phrase sans the time stamp? Ideas for a KM macro are swirling around this noggin but may not be any better than this "Wiki-link method" if I could intuit it.

ctietze

I was starting a detailed reply, but figured that a video full of typos would work better:

1. Type [[ + ⌘U to get a new ID + an optional title
2. Click on the resulting link (that goes nowhere)
3. With focus still in the Omnibar, hit Enter to create a note that exactly matches the link anchor text

Will

Thanks for taking the time to put this complete step by step instructional video together. It helped me see how important developing muscle memory by repeatedly doing the same task over and over is in developing expertise.

Connects with the article I just finished reading and will process into Zettelkasten this morning. In it, Oakley and Djikic argue that there are a few explicit ways to use writing as a tool for developing expertise in a field. They focus on the field of writing fiction but what they point to can be applied to other fields. Take away. Key is putting in the time and a focus on expertise development.

@doi:10.1037/1089-2680.12.1.9
Keith Oatley and Maja Djikic (2008): Writing as Thinking, Review of General Psychology 1, 2008, Vol. 12, pg. 9-27 - @Oakey:2008a
PDF

Steph

@ctietze Thankyou, very clear. One thing I can't catch, though, is at the very start. When you create a new note, how did you give it a title so quickly? I mean a title as the top line of the zettel itself, with a '#'. Is that a shortcut (or are you just typing faster than the speed of light?!)

I might not end up using this myself, as so far I like to give the note just a UID to begin with, then go back and rename with the shortest and clearest name that best represents the note, after I have written the note. But I'd still like to add all your tricks to my arsenal, and give them a fair trial

ctietze

@Steph The title I just pasted; I prepared most of the content in the video as a reply here first, but deleted all the text in favor of a 3-point list + video

@Will The Oatley/Djikic paper has a couple of interesting pictures, like "[T]he pen is a machine to think with" (p9), and the list of signal words and concepts to hunt for in texts: schema, model, allegory, analogy, hypothesis, metaphor, representation, simile, theory. (ibid)

Sascha

@Sascha said:
How do you define tightness?

The ones that I consider children and parents are tight. In my example, the definition of random effects follows the definition of a random-effects model (child, i.e. tight link). The random-effects model is a kind of statistical model, so that is probably its parent (i.e. tight link). It is related to the fixed-effects model, so I could include a link to it, but that I would consider a "loose" link. This link might be obsolete because they have the same parent, but I can imagine links to other notes to be relavant, too.

Maybe some of my notes contain so many links because they are not atomicised. I tried to explain very broad concepts, which link to a lot of notes, whereas these notes should actually be structure notes, where the links are the essential part.

So, there is no need for the parent/child-concept. There are tight and not so tight links. Thightness is not something that comes up in the syntax of the method but in the meaning of the surounding material.

@Garwyx wrote:
@ctietze and @Sascha I would like to see/know how you implement links in your notes.

There are three types of (internal links):

1. Linking from a structure note to another note. See the article on structure notes for more information.
2. As a non-explicit link. I write the text and if a statement is expanded via another Zettel I link to him.
3. As an explicit link. Then I write explicitely why the link is there. (e.g. "A similar concept is XY. Both share Z attributes. See: [[ID]]