A link should explain a relationship. If an essay about memory leads to one about knowledge, the reader needs some reason to expect that the second will help with a question raised by the first. Sharing a broad subject is seldom enough.

Hypertext made it possible to organise a text through connections as well as a single sequence. Ted Nelson’s 1965 proposal described a structure intended to accommodate material that changes and has complex relationships.[^1] The problem remains familiar to anyone who writes notes. An idea may belong in several discussions, and its usefulness may change as those discussions develop.

Niklas Luhmann’s account of his slip boxes describes fixed addresses and connections between notes as part of a working system.[^2] A note could be encountered through relationships beyond its original place. His account describes his method; it is not experimental proof that making similar boxes will produce similar work. The practical point is that a connection can give an existing note another use.

Engelbart’s framework likewise considered tools together with methods, language, and training.[^3] A system for linked writing needs a practice of deciding what the links mean. The software can store them and draw the graph. It cannot establish their intellectual value merely by making them visible.

There are several possible reasons to connect two essays. One may explain an assumption used by the other. It may present an objection, develop a consequence, or examine a related problem from another direction. These reasons differ enough that the writer should know which one applies.

In this series, the main essay introduces the shared question: how changes in our tools affect the definitions we use. The individual essays then examine memory, knowledge, understanding, thinking, mind, essays, books, and authorship. Returning to the main essay helps place each discussion. A few direct connections between the branches are useful as well, such as the relationship between a book’s form and the authorship of its words.

Connecting every branch to nearly every other would make those decisions hard to see. The graph might look busy while saying little about where to begin. A smaller set of deliberate connections allows the leading essay and the important relationships to become visible.

Reading needs its own arrangement. DeStefano and LeFevre’s review describes how navigation through hypertext can place demands on attention and working memory.[^4] Each opportunity to leave a page creates a decision. A reader following an argument should be able to continue without first interpreting an entire network.

This is why a suggested next essay can be useful even when it is not a direct graph connection. It offers a route through the cluster. The graph shows relationships a reader may want to explore; the route gives someone who is beginning a manageable way forward. These two arrangements can serve the same collection without having identical links.

A visible map may help with learning, but its appearance is not a sufficient test. Nesbit and Adesope’s meta-analysis examined varied educational uses of concept and knowledge maps and found benefits that depended on activities and comparison conditions.[^5] In a different study, Karpicke and Blunt found that retrieval practice outperformed the particular concept-mapping activity used in their experiment.[^6] What a learner does matters alongside the form of the representation.

Kirsh’s account of external representations helps explain part of a graph’s practical value: relationships can become available for inspection and manipulation.[^7] A sparse area might reveal an undeveloped question. A prominent node might identify a useful starting point. These observations still require interpretation. Node size and visual proximity cannot stand as independent evidence that an argument is strong.

The collection should also remain readable as individual essays. A piece that sends the reader away whenever a difficulty arises has not done enough of its own work. It needs to establish its claims and explain why its connections matter. Another essay can then extend that work rather than complete an argument that was abandoned halfway through.

As further clusters are published, a route can eventually lead into a new subject through its leading essay. That connection should be earned by the question at the end of the earlier route. A growing archive will need more of these decisions, not simply more links.