title_test.gif
Patterns of Hypertext  10/29/04 - Article by Mark Bernstein

Patterns of Hypertext is a clear, concise summary of structures found in hypertexts. The stated purpose of the article is to provide terms for patterns currently found in order to enable and foster discussion of structure in hypertexts. I read it for exactly that purpose, and it did well enough that I've taken notes to come back to later in other projects. Patterns seems to be a touchstone for much later work on narrative, pedagogy, and design in the field.

Structural terminology is important to hypertext

One of the major ways that hypertext differs from [print,linear] text is that structures are more critical to the experience of the work. In a story, you don't necessarily need to know where you are because there is always a line stretching out ahead of you. There are structures underlying the story which are important, sure-- time rarely flows only forward, different plot threads come and go, and characters all have their own ways of thinking about the events of the story.

But in hypertext:

  • the reader is forced to consider (or even speculate or second-guess) structure in order to read through the text
  • in light of the (potentially arbitrary) linear order the reader gives to the lexia, secondary structures --the ways the reader pieces together the work conceptually-- become more significant

Non-hypertextual non-fiction is more like hypertext in that the author generally needs to place a more explicit structure over the work for lack of the easy line of the narrative. We can look at whether a history textbook moves through historical time periods linearly for each culture it examines or surveys all cultures in each time period before moving on, in a sort of geographic spiral.

But we have few terms for discussing those structures even in outside of hypertext, and the need to articulate them is more critical for discussion of hypertext.

Patterns

Cycle

Reader returns to a previously-visited lexia and eventually emerges along a new path.

Joyce's Cycle

Named for Michael Joyce, in this a reader returns to a previously visited portion of the hypertext and reads along a previously visited set of lexia before emerging onto a new path.

Douglas' Cycle

Named for J Yellowlees Douglas, an unbroken cycle (with no emergence) signals the end of a section or an exhaustion of the hypertext.

Web Ring

A Web ring is a grand cycle, connected by topic or more generally by shared readership. The experience of a web ring might be like the experience of one of the other cycles, but is distinguished by the fact that it is a grand structure, one which by necessity incorporates other structures and which may not include revisitation in an average reading.

Contour

A contour is formed where cycles meet, and allows access between cycles. I'm not sure how this is different from a set of connected cycles.

Counterpoint

Two voices alternate. This often communicates structure clearly and is good for interleaving themes or for theme and response.

MirrorWorld

A parallel or intertextual structure that is used specifically to create a different voice or contrasting perspective.

Tangle

A tangle is a structure where the reader has a multitude of links and insufficient information to choose between them. A tangle might be closed within a loop or might branch out into other structures-- the common point is that the reader has insufficient information to choose between the paths.

Sieve

A branching structure which sorts readers out into other structures by a sequence of choices. The choices may be informed (Table of Contents) or uninformed (in which case it looks more like a tangle)

Montage

Several lexia presented together create a montage.

Neighborhood

A Neighborhood establishes an association among [lexia] through proximity, shared ornament, or common navigational landmarks. The common features show that the lexia are "close" in some intentional way.

Episode

In the description of the Neighborhood structure, Bernstein describes "Rosenberg's episodes", which are like neighborhoods but with regard to the reader's perception rather than to meaning in the hypertext.

Split/Join

A split/join knits two or more sequences together. Those individual sequences may be composed of other structures; the point of the split/join is that a connection is made (implying, perhaps, a neighborhood).

Rashomon pattern

A split/join embedded in a cycle creates what Bernstein calls a Rashomon pattern, where the recurrence of the cycle is conducted over different threads (often creating a counterpoint either implicitly or explicitly)

Overview/Tour

A split/join where one side is more detailed than the other (but are rhetorically similar) frequently constitutes an overview or tour of other structures.

Moulthrop's Move

A split where the text "responds ironically" to the reader's apparent expressed interest (indicated by link choice).

Missing Link

An absent (or broken?) link where one is expected can be a meaningful structure when, like an ellipsis, allusion, or iteration, it implies a connection.

Navigational Feint

The offer of a navigational opportunity that cannot or is not followed immediately. It can establish a pattern for later in the reading or provide information about the structure or scope of the text

We can't visualize these yet

Bernstein notes that we don't yet have tools that help us visualize all of these adequately. Analytical tools that help us discover structures are good at examining trees, but often miss cycles. Node-link views (like Tinderbox map view) get individual cycles but don't handle contours well. Hierarchies keep views clean but hide structures that cross hierarchical levels. And massive structures (mirrorworld), negative structures (feint, missing link), and serendipitous structures (montage) are currently hard to display at all.

I think that some of what he's talking about here is what I'd like to see in my vaporware htext engine-- something that is so flexible and facilitates revision so much that seeing the structures doesn't require dragging entire areas out of others, and rearranging them-- the engine does it for you.

I wonder whether Ben Fry's tools might not help with the structures that we currently find hard to analyze.