[1]Bolter [7] used the term `episode' the way the term `lexia' [19] is now conventionally used, but his use of the term episode did not catch on; apologies if this new use of the term episode causes confusion. It was tempting to borrow Douglas's [10] term `strand' -- but this seems to prejudge the issue of whether the episode is linear and to be more narrowly suited to the specifics of the node-link model.

[2]This author knows of no literary hypertexts explicitly based on sets as a substructuring method. The complete absence of set-based literary hypertext is both striking and hard to explain (though set-based substructuring is not usually present as an off-the-shelf abstraction in commercially available hypertext software.)

[3]Trellis ([34], [11]) provides a formal basis for dealing with such questions. For a Trellis hypertext one may describe an acteme as any form of hypertext activity which causes the Petri net to fire. If within-component scroll-bars are devices maintained entirely by a client which does not fire the net when they are operated, they would not be considered actemes.

[4]Figure 1 as drawn implies that the lexia is "atomic" with respect to episodes -- i.e. a lexia is either entirely in or entirely out of an episode. Of course an episode may include only part of a lexia; there is no guarantee the reader will read the whole thing.

[5]Zellweger [37] discusses implementation of a similar concept, though her paths are constructed by the author rather than the reader.

[6]At the Spatial Metaphors Workshop at ECHT'94, Mark Bernstein raised the question of how the user of a hypertext might be able to estimate the cost of following a link. Should actemes be coded so that the reader can estimate the cost of activating them?

[7]In [10] Douglas refers to such maps as "cognitive maps". To call the "actual" map of the node-link structure a "cognitive" map is a serious confusion. The map may be structural more than it is cognitive. One might in some cases call a reader's map in the context of an overt gathering interface cognitive; whether the writer's structure map is cognitive or not depends on the circumstances.

[8]This section is heavily indebted to Douglas's paper. While it should not be taken as simply a restatement of her work, most of the ideas in this section were the direct result of reading her very stimulating discussion.