...all those sardines...

If you ever saw “Noises Off” you will remember the hilarity generated when the plate of sardines was not on the table on the stage as required in the third act…

When I am writing long-ish stories I am always getting lost about the key Annie mentioned to Fred who put it in the cookie jar - did Fred tell Jill before she needed to get into the cottage on the weekend…?

Now I am writing a textbook based on papers I have written over the years - is this the first time I have mentioned Voronoi diagrams, or defined them, or have I repeated myself three times already? How can I make sure that the end result is a linear process (a partial ordering) where I start with the things that don’t need to be described exhaustively before I start?

In other words: what is the Zen for tracking items (or even references) in Scrivener so I can pick out the order I need? All suggestions very welcome before I get too deep into my book!

So long, and thanks for all the fish… Chris

I saw this the other day and have been turning over ideas in my head, but I haven’t had a chance to really go at it and write the full-blown response I want to. (Which in all likelihood is for the best, since I would probably just go on and on for pages.) So this may be a bit lame, but I wanted to give some form of reply. Perhaps it will spark conversation.

I don’t know how much prowling around the forums you’ve already done, but here are a few posts that deal in some form or another with tracking information through stories—mostly these are fiction-related examples, but I think the methods apply to non-fiction just as well.

AndreasE’s post on “How to visualize a subject distributed” talks about using keywords to tag documents and then doing a “reveal in binder” to get an overview of their placement in the project. This was for an earlier version of Scrivener; with 2.0’s enhancements, you could change this up (colored keyword chips are visible in the corkboard now, for instance, and you have arbitrary as well as dynamic Collections at your disposal), but the idea remains a good one.

Someone posted a wish for multiple synopses per document as a way of tracking different “focal points” through a story, and while this isn’t a feature Scrivener offers, people replied with different suggestions for achieving the goal. More on using keywords and the new inspector comments.

Finally, a third idea is one I detailed in another thread where the OP wanted to collate all the references on a given topic, with links back to the original source. The way I do this is to just make new linked notes off the relevant sentence or paragraph and file them topically. It’s easy then to view them together in a Scrivenings session to see how an arc progresses (and how often I repeat the same point); I can make additional notes there, and it’s all automatically back-linked to the original scene. I use it for various fiction elements, but I think the same idea could work for tracking your technical sardines, too.

So those are a few ideas to get you started!


(PS–Bag! Bag bag!..No bag!)

Dear MM,

Thank you so much for replying! I am working my way through your examples.

As a computer scientist, after reading your message I went back and tried to think in first principles. This is what I came up with:

Story Structures

Previous examples:

In the ‘Noises Off’ example I need to track from the initial introduction of the sardines, to where they got moved, and to why they weren’t put on the table in Act Three. This is a linear sequence: on paper I would just read it through and underline when they swam into sight, and check that all the events (actions) followed each other.

In the ‘key to the cottage’ example there is a linear sequence of what happened to the key, as above. However, tracking who knows about the key is not linear: half a dozen people could have heard what was said.

In the ‘textbook’ example, with multiple input sources from various papers, I could search for the term ‘Voronoi’ on the combined document, but in addition I would need to know if this was a definition/description - or just a citation or use of the term while describing something else that needed this background information.

Theoretical Implementation:

To use this approach we need to be able to break the papers (or versions) up into identifiable nodes with titles that can be matched for duplicates. Then some can be discarded. This is obviously manual, and author-specific, drudgery. Each node must have an ID as well as a title/keyword set.

We must have a link collection {node ID, ancestor ID} that means the node is dependent on the ancestor in order to make sense. (There may not be any need for any description here, but it doesn’t hurt.) Any links back to nodes that get discarded must be corrected to show the replacement link. We can make things look cleaner by having a ‘start’ node and an ‘end’ node, connected to the initial and final usages of an idea respectively: these form the start and end of the book. In most stories the activity is what is described in the text – so that will be the node. Links are just for sequence. (For research the text/node is the ‘proof’ or reference – a kind of activity. Links refer to premises.)

This is really a CPA – Critical Path Analysis – chart. Traditionally you make a table of tasks and dependencies: for us node descriptions and dependency links, where each node may have many dependency links. The result is a network diagram (using Activity On Node structure): please see the diagram at

britannica.com/EBchecked/top … od-problem

(The diagram really makes things much clearer!)


To return to Scrivener: we would like each of our texts/nodes to be an index card on the corkboard. We can then pile up duplicates (where we have said the same thing several times and will have to combine/reject later). We then want to draw arrows between piles based on the known dependencies in the original papers, or else from undefined keywords in the texts themselves: each of these must in the end have an arrow from an ‘explanation.’ On a real corkboard I would use thumbtacks and string.

Given this the rest is easy: I have multiple paths from start to finish, but I am writing a linear book, not hypertext, so I must cut some threads – or even better, I keep all the threads but pull all the nodes out into a single linear sequence, based on how I think the story will flow best. The important thing is that every node has ancestors.

Of course, this approach could also be used for designing the story/book in the first place.

So how do I do this – or a reasonable simulation – in Scrivener?

Gratefully, Chris Gold