Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/221116354

Contours of Constructive Hypertexts.

Conference Paper · November 1992


DOI: 10.1145/168466.168517 · Source: DBLP

CITATIONS READS

48 129

3 authors, including:

Michael Joyce David B. Levine


Vassar College Saint Bonaventure University
42 PUBLICATIONS   742 CITATIONS    31 PUBLICATIONS   339 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by David B. Levine on 04 August 2015.

The user has requested enhancement of the downloaded file.


MILANO, NOVEMBER 30- DECEMBER 4, 1992 161

Contours of Constructive EIypertexts

MarkBernstein
Eastgate Systems, Inc.
134 Main Street, Watertown MA 02172 USA

Michael Joyce
Jackson Community College
Vassar College

David Levine
Gettysburg College

Abstract 1 Introduction

Discussion of hypertext rhetoric and criticism has How can we deseribe our experience of large and
foeused on small, freed hypertext that am complex hypertexts-hypertexts intended to
typically used by casual readers for comparatively inform and influence a dedicated, thoughtful
short periods. Here we explo~ complex, reader?
constructive hypertext, intended to inform and to
Discussion of hypertext design, hypertext rhetoric,
infIuence dedicated and thoughtful readers.
and indeed of the merits and faults of hypertext
Recurrence and multivalence, both deplored in
itself, has often fbcused narrowly upon the role of
small hypertext, prove from study of mom
isolated links. Calls for “structured” hypertexk
complex texts to be very valuable. While static,
while emphasizing the complexity of large
graph-theoretic measures facilitate understanding
hypertext, have limited their attention to simple,
of Ioeal hypertext structure, the structures of
recursively defined aggregates. IrI contrast, we
meaning or contours we observe in current
argue that complex hyptxtexts are not simply
hypertext fiction and scholarship do not appear to
expanded versions of small hypertext and propose
reside in static structures, but rather in the complex
some initial steps toward a voeabukwy for better
and dynamic perceptions of the engagedreader.
understanding large hypertext.
Understanding of small-wale hypertextual
The yarns of seamen have a dwect simplicity, incomplete, is
phenomen% while still
the whole meaning of which lies within the
comparatively well advanced. Considerable effort
shell of a cructid nut. But Murlow was not has been devoted to promulgating rules for
typicai (tf his propensity to spin yarns be
effectively placing and explaining hypertext links
excepted), and to him the meaning of an (e.g., [5,25,35]). Indeed, discussion of the merits
episode was not inside like a kernel but and faults of hypertext has, since Conklin’s review
outside, enveloping the tale . . .. [10], centered on the alleged tension between the
Joseph Conrad, Heart of Darkness expressiveness of individual links and the
additional complexity or “cognitive overhead”
which the presence of links imposes on the reader
[8,9,16]. Prescriptive, formal models have been
proposed to discipline link creation and enfome
Permission to copy without fee all or part of this ma- consistency [11,23], taxonomies of tinks have been
terial is sranted movided that conies are not made or
distributed for d;ect commercial ~dvantage, the ACM
proposed [15], and clinical studies (e.g. [20,47,481)
copyright notice and the title of the publication and have sought to compare the effwiency of link
its date appear, and notice is given that copying is by traversal to conventional access strategies. Belief
permission of the Association for Computing Machin- networks [22], retrieval models [12], guides [37],
ery, To copy otherwise, or to republish, requirea a fee
and/or specific permission.
and apprentices [3] have been developed to help
@1992 ACM @89791-547-X/92/0011/0161/ $1.50 readerschoose among links.
162 ACM ECHT CONFERENCE

While these studim have yielded a wealth of 2 Learning From Hypertext Fiction
valuable material, they share (perhaps of necessity)
a simple but unrealistic view of readers and Hypertext rhetoric should be grounded in critical
reading. Empirical studies have emphasized examination of actual hypertext. In the absenceof
retrieval speed and efficiency because retrieval is a hypertext literature, early writers had no choice
easy to measure.Retrieval of facts, however, is but but to extrapolate from their expectations or their
one of myriad ways we use written information, own initial experiments. A purely theoretic
indeed, compared to the influence that an impurtant ~~mh is hazardous, for hypertext are not
textbook or favorite poem may exert on our electronic simulations of printed Iwoks:
character, or an influential research study or To require hypertext to function like a
theoretical analysis upon OK professional lives, book is a bit like expecting a jediner to
retrieval seems neither a very common nor a very behave iike a locomotive: yes it’s very
important activity. The intrinsic complexity of real fasl, but the blasted thing won’t stay on
texts has been obscurd too, by the assumption lhe rails. [42]
that texts have a single purpose or intention. Few
texts are one-dimensional, and truly single-purpose Technological changes may have unforeseen
texts are almost always ephernmd (business cards, impacc Bolter has shown that an innovation in
advextisements, audit records) or archival (laws, binding technology, the codex, transformed the
voter lists). Meaningful documents, even on paper, scale and conception of literary work [7], while
are multivalent and polyvocal [4]. Landow notes that a technical change in page
layout (the introduction of spacesbetween words in
In contemplating the atomic act of link traversaL” manuscripts) changed reading from a vocal and
we may lose sight of larger structures. The public performance to a silen~ personal activity
literature is replete with calls for “structured [36].
hypertext” [16] and taxonomies of hypertext
structure [49], yet the proposed structures are either It is now (barely) possible to ground a discussion
very small or very simple. The rhetoric of of hypertext rhetoric in actual hypertexts-texts
hypertext rhetoric, too, has been confined to small which have been deliberately crafted by writers
scale, Landow, for example, addresses points of who devoted substantial efforts and resources to
departure and arrival [35], while Evenson and the task, and whose reputation rests in some
Rheinfrank propose a visual language for measure with the quality of the resulL The body of
typographical representation of isolated links [211. hypertext literature remains sparse,and some titles
are so specialized in topic (e.g. military equipment
Many of the most exciting developments in maintenance) or intended audience (e.g. preliterate
hypertext systems have applied knowledge children) that it seems dangerous to generalize
representation to create local hypertext structures from them. Many of today’s best-known
and behaviors, but these, too, involve hypertext, monxwer, are works of fiction. Some
comparatively small structures. SEPIA [55], techniques of fiction m specialized, yet we believe
MacWeb [46] and Aquanet [40] let authors express that insights from fiction can often be applied to
relationships among typed nodes and links, technical and scholarly writing.
providing a rich vocabulary of local structure.
Explicit representation promotes local coherence The chasm between fiction and technicrd
and consistency, permitting authors to express communication is less formidable than it
systematic relationships (e.g. between artist and sometimes appears. Effeetive language,
artifact) in a manner which can be grasped and informative organization, dramatic integrity and
anticipated by the reader, searched by a qwxy engagement challenge all writem, regardless of
system, or used as a basis for computation. In some genre or medium. This approach is not particularly
sense these local structures function as rich novel, as Laurel [37] has applied lessons from
metanodes which are themselves linked across a classical drama to user interface design and Peter
network of metanodes, Thus even these subtle and B@gh Anderson [6] uses semiotic concepts from
flexible systems express essentially local relations cinema to understand hypertextual transitions.
among clusters of entities. Earlier, Harnmet and Hemingway adopted the
discourse of journalism to the service of fiction.
Where the novelist seeks an extensible,
MILANO, NOVEMBER 30- DECEMBER 4, 1992 163

constructive hypertext, a “form for what does not 4 Multivalence is not a vice
yet exist” [28], the engineer cites a need for
“information [which] constantly evolves as people Repetition also opens the possibility of multivalent
contribute parts to a conceptually interconnected passages,episodes whose meaning changes when
web.” [38] revisited because the reader has gained new
understanding or insight. Formal, prescriptive
approaches to hypertext have usually regarded
3 Recurrence is not a defect multivalence as undesirable, preferring to assign an
explicitly-~presented role (or role set) to each
As hypertext readers tmvel through a document, writing space (cf. [1123]). In large and complex
they may return to places they have previously hypertext, however, multivalent writing is neither
visited. The impact of unexpected repetition can be undesirable nor, indee& avoidable.
startling; Bernstein [2] introduced the use of
breadcrumbs specitlcally to alert readers to the Consider, for example, the role of Emily Runbird’s
possibility of revisiting a recently-visited place. death in Stuart Moukhrop’s novel of the Gulf War,
Rouet [51] assumesthat revisiting a hypertext node Victory Garden [41]. Emily is a graduate student
is evidence of disorientation. who, by virtue of her ROTC scholarship, finds
herself stationed in Saudi Arabia. In some readings
This perception is fundamentally wrong. In small of Victory Garden, Emily is killed by a SCUD
hypertext, to be sure, recurrence is boring. missile. The meaning of many passagesis changed
Recurrence is unhelpful, too, in hypertextual by Emily’s death; passagesthat had one meaning
encyclopedias and reference manuals which are while Emily is believed to be safely ensconced in
meant to be consulted in brief episodes. In complex the desert mailroom mean something else entirely
hypertext, on the other hand, recurrence is not a after we are led to suspect that Emily has been
defect repetition provides a powerful structural killed. Her sister, Veronica, is brooding and
force, a motif which helps readers synthesize the irritable. Does her unease mirror her professor’s
experience of the reading. Rhythms of recurrence irritation with patriarchal politics, or does it arise
announce patterns of meaning. from her intuitive consciousness that her sister’s
Malloy’s its name was Penelope is a radically life has just ended in flames? Elsewhere, another
“chunky” hypertext a densenetwork of very short, student, Jude, seducesEmily’s lover while wearing
almost stylized texts representing fragmentary a blonde wig. Is Jude a tramp, callously seducing
memories of an artist’s girlhood, professional life, Emily’s lover while Emily is at war? Or is her
and love affairs [39]. Repetitions help modulate the impersonation a reincarnation, a purifying ritual, a
story, superimposing an irregular secondary gesture that life will go on? Meaning emerges not
rhythm of repetition on the uniformity of from a single sequence or interpretation but from
successive,brief passages.Moreover, the passages the coalescenceof multiple readings [17].
themselves contain internal quotations and Hypertext, after all, was originally motivated by
allusiony a passage which the reader has not seen the realization that a single passage may play
before may recall an earlier node, and may itself be several different roles in discourse. Discussion of a
called to mind when the earlier node is revisited. single astronomical object, say M31, may appear in
Similarly, Joyce’s Afternoon [29] exploits textual an astronomy text as an example of a gaseous
cycles to represent (and, simultaneously, to deny nebula, as an object of interest to
the possibility of) closure [181. archaeoastronomy, as the home of a pulsar, or as a
These are not mere artistic effects. Examine, say, a pretty object to look for on a warm summer
university physics text, and you will find evening. Indeed, technical texts deliberately reuse
conservation of momentum and energy expressed examples not becauseof a shortage of material, but
over and over again, in different contexts and in order to establish ~lationships within a
guises, to ensure that students truly understand discipline. The “same chemical reaction may be
Recurrence communicates the existence of cited as an elementary example, may appear
structure, the linkage among ideas which weaves elsewhere as a useful synthetic tool, and be cited in
disparate facts and arguments into a single coherent yet another context as evidence for a mechanistic
work. theory. If a text is large enough and important
164 ACM ECHT CONFERENCE

enough to change the reader, multivalence explicit paths in the hypertext, while others may
expresses that change, and recurrence makes that appear to us only when we think of them, or when
change apparent. a colleague points out an overlooked structure of
meaning. Texts like dictionaries and integrated
Recurrent structures can also be seen as an author’s
circuit catalogs, are essentially flat they may be
instantiation of the ephemeral memories of creating
large and informative, but they convey little sense
a text which itself anticipates the reader’s
of structure. A few texts-business cards and
instantiation of similar (though not identical)
military orders-offer only a single structure and a
structures. This “constantly replaced present tense”
single contour. Most texts, however, m richer, and
of hypertext contours [31] calls to mind Eco’s
present varied contours as we read and reread
intention of the text:
them.
A text is a device conceived in order to
A contour is “a contingent order, a linearity
produce its model reader .. . The
perceived in tension against the possibility of other
empirical reader is only an actor who
lines” [43], While paths [571 may tlilow (and
makes conjectures about the kind of
illuminate) contours, contours are not paths, A path
modei rea&r postulated by the text.
is, in essence,a traceable record of a trajectory. If
Since the intention of the text (intentio
that trajectory is meaningful, we might describe the
operis) is basically to produce a model
path as following a contouq on the other hand, we
reader able to make conjectures about it,
may easily imagine a trajectory which is
the initiative of Ihe model rea&r
convenient or functional but not intrinsically
consists in figuring out a model author
meaningful.
that is not the empirical one and that, in
the end, coincides with the intention of Depth describes the presence of multiple contours
the text. Thus . .. the text is an object that within a text. When a ~ader fiit glances at a
the interpretation builds up in the course hypertext, of course, she sees a set of independent
of the circular eflort of validating itself nodes comected by links, just as when we look at
on the bash of what it makes up as its an abstract painting we initially perceive daubs of
result. [19] color placed on canvas. The perception of structure
arises as we rationalize our initial impmssion. Just
As hypertext dissoIves the distinction between
as adjacent blue patches coalesce into an object, or
readers and writers, each mider forms her own
adjacent figures coalesce into a group, interlined
working model of the intention of the text,
and interrelated nodes coalesce into a contour [1].
appropriating the role of Eco’s model author.
A hypertext will appear to have depth-to offer
many contours-when it can be seen as the
5 Depth and Contour projection of a recurrent structure that is
structurally simpler than a non-recurrent one.
As the text discloses its shape and intention to the
Rather than viewing recurrences as anomalous
reader, the reader creates hypothetical predictions
phenomena, the reader will try to cmt them into
of its shape and possible continuations. These
consistent, large-scale patterns.
expectations change as the reader examines the
work more completely, much as our interpretation A simplified “model railroad” representation of
of a painting changes while we look at it. The eye hypertext contours can begin to illustrate the depth
perceives suggestions of three-dimensional of recurnmt structures. The mder of the hypertext
structure in a two dimensional painting---even a shown in Figure 1, ffom her current perspective (M
non-representational painting. We borrow the in Figure 1), perceives &me distinct contours-the
terms depth and contour from the painter’s contour she has chosen, and two contours she has
vocabulary to describe a similar sense of form (for the present) foregone. One contour lies before
which the mxier gains as she reads. her the other two contours were once readily
accessible, but can no longer be reached by
By conrour, we mean a perceived pattern of
following links. A choice has been made, and the
meaning within the hypertext an observed
reader’s attention is not limited to a specific sector
stnictum that a reader can describe and appropriate
for her own use Some contours may coincide with
MILANO, NOVEMBER 30- DECEMBER 4, 1992 165

of the text. IrI this way, navigational choices lead rejoin, while other links, that once seemed to be
readersto model the text’s intentional structures. mere annotations, may gain significance in
retrospect. The rhythm of these shifts between
recurrent structures takes the form of what Joyce

+5 cl--n has called replacement within contours [31] and

❑-
r E cl---n
Jay Bolter describes as the principle of interactive
oscillation [44]. “Even in ostensibly clear and
instrumental text”, Moukhrop argues, “the
language and hence the hypertextual situation, may
shift into oscillation.”
cl---a-- Cl-WI

Figure 1. The reader at Hperceives three contours.

In actual hypertext systems, of course, the reader


could still t%d ways ‘to reach the sections she has
l=L-
I

WI
1
I
C1--n

bypassed. She might be able to retrace her steps


and follow a different trajectory, or to jump (as if
by magic) to a node whose name she knows.
Though choice is not irrevocable, it influences
perception. Patterns of links suggest patterns of
intention. Figare 3. An added Zink knats contours tagether.

Now let our reader (or author) add links to the In an abstract painting or polyphonic music,
hypertext. We imagine the perspective of a reader complex, textured shaWs seemto oscillate between
who has followed the same trajectory as before, foreground and background [1]. Similarly,
arriving at the sameplace in the hypertext. intersecting contours may give rise to oscillation
and replacement in our mentat model of the
recurring structures of our hypertext. These notions
of shifts, oscillation, and replacement call to mind
Deleuze and Guattari’s notion of “transcoding or
transduction [as] the manner in which one milieu
serves as the basis for another . . . is established
atop another milieu, dissipates in it or is constituted
by it.”[14] The rhythmic recurrence of hypertext
contours enables the reader to form her own
working model, one which in its shifts
Figure 2. An added link provides a shortcut. communicates the intention of the text.
In Figure 2, the reader will still perceive several
distinct contours, for her decisions have made 6 Contingency and Contour
some parts of the texts easily accessible while
others can be attained only by extraordinary means. The hypertext reader not only chooses the order of
In Figure 3, on the other hand, the added link what she reads, but by her choices chooses what it
threads the contours together into a knot. The shall be [17]. Patterns of allusion and reeumence,
reader perceives that she has not been forced (at familiarity and novelty shape themselves into a
this point) to choose among contours, and what transient sense of coherence; for a moment, we
formerly were distinct sections or contours have may look up from the screen and say, “Yes! Now I
coalescedinto a single aggregate cluster. see where this hypertext is going.” What we
During the reading of a complex hypertex~ discover is as much where we are going as it is
contours may emerge and recombine as we what paths the author has laid out.
encounter new links and connections. Contours that ‘Ile text rewritten by the act of choice becomes
once’ seemed to pose irrevocable choices may constructive hypertext [28]—hypertext that is open
166 ACM ECHT CONFERENCE

to change and addition and revision, hypertext in a choice of only two or three links that lead to
which the authorship of the reader is made explicit. dramatically contrasting contours.
Readers constantly make inscriptions in the text—
We can define a crude but computationally-
not ordy by adding bookmarks and margin notes,
tractable measure of the texture T(n) of node n
but also by choosing one link and bypassing
from the following observations:
another, or by remembering an encounter with a
familiar passage. ● Each node we can reach from n contributes to the
texture of n.
A constructive hypertext, like a collaborative
● Nodes that are close to n contribute more texture
design [38], is a foundation for what does not yet
than nodes that can only be reached from n by
exist. Collaborative design and argumentation
following many links.
systems provide a framework that reifies the
transient structures of design and dispute If we define d(n,v) to be the minimum distance
[11 ,53,55]. These collaborative structures are between nodes n and v in the underlying hypertext
continually being replaced as readersperceive (and digraph, a possible definition of T(n) might be:
create) new connections and disjunctions.
Coextensiviry [31] refers to this kind of T(n)=~ 2-d(n’v)
replacement of one writing by another. We vtin
recognize coextensivity in the degree of oscillation
Nielsen [48, p. 169] alludes to a similar measure.
and replacement among recurring structures of a
Some simple examples of fiwniliar substructures
hypertext. Constructive hypertext environments
may clarify texture:
often make these virtual contours explicit through
virtual structures [27] like the Storyspace Path ● The texture of each place in a long, linear
Builder [54] or the implicit links of Perseus [45]. sequenceapproaches 1=(1/2 + 1/4+ 1/8....)
● If A is a footnote connected to B, T(A)= 1/2+
7 Tension and texture T(B)/’2

● T%e texture of the root of a complete binary tree


Linear reading is essentially passiwz the author has is the tree’sdepth.
established an argument which the reader follows.
As we read a hypertext, however, we are constantly The densest places in a hypertext are those where
summoned to choose one path or another, to pursue many contours intersect. For example, consider
one goal and avoid a different goal. On occasion, Bolter’s Writing Space [’7], a monograph on “the
we may know exactly what we want to find, and computer, hypertext, and the history of writing”
bypass the hypertext entirely for a simple query containing of 451 writing spaces and 888 links.
mechanism; more frequently, we do not know The most textured places in the document are
exactly what we seek, and our goak shift as we section headings that describe key topics and
learn. supply a hierarchical navigationat framework:

Not all places in a hypertext afford equal degrees 15.4 Writing Spaces
of choice. Some places are fraught with tension, 14.51 III. Mind
offering many choices which wiU shape our entire 13.23 II. Conceptual
reading; others are flat and simple, offering few or 13,09 I. Visual Space
unimportant choices. Texture measures the 12.99 VIII. Interactive Fiction
capacity for meaning-shaping choice among
elements of a hypertext-the ability of a reader to The least textured places are notes that illustrate or
transform the text and structure. Where contours comment upon specific points in the texfi
intersect and interact, texture is dense, whe~ a
0.75 electronic writing turns
single contour prevails, texture is smooth. The
0.75 unbounded
texture of a writing space is not merely a matter of
1.00 Iamsorrytohearthat
counting linky a writing space with thirty
footnotes, each leading to a bibliographic It can be interesting to keep one eye on the “texture
reference, is flatter than a writing space that offers gauge” while reading a hypertext. Texture climbs
as we approach a major navigational crossroads,
MILANO, NOVEMBER 30- DECEMBER 4, 1992 167

and declines as we pursue more specialized is set too high, we find a only single large
material. User interfaces can give textural cueq for component in each hypertext, accompanied by
example, we may raise the pitch of a sound that many singleton components and very small
signals the turning of a page as the texture grows clusters. This result was not expected, especially in
more dense. Because highly textured regions are Woe, since Woe’s “Mandala” node with 38
clearly critical to the coherence and outgoing links is so central to navigation through
coextensivenessof a hypertext, a software agent or the text’s 63 nodes. Decreasing the texture
a link apprentice (see [3]) might profitably be threshold sufficiently does eventually break up the
programmed to recognize and represent local single strongly-comected component, but also
structures which occur at highly textured spaces. dissects the work into many isolated nodes rather
than a few meaningfid chunks. Thus, if we delete
the ten nodes with texture >5.0, we find one large
8 Static Structure and Dynamic cluster (39 nodes) and ten small clusters . If we
Perception
delete the 24 nodes with texture, M.0, the large
cluster has disappeared, but the 39 remaining nodes
The search for better methods of identifying and
are broken up into 23 components, all small. While
visualizing hypertext structures forms a central
readers describing these hypertext consistently
thread in hypertext research. While the
perceive them in lerms of aggregates like scenes
computational intractability of graph layout
and episodes, these aggregates & not easily
algorithms continues to mandate ~eliance on hand-
emerge from the graph structure.
crafted diagrams [2], interest in layout heuristics
[33,50], interactive manual layout [30, 46], and Nor do these aggregates emerge from simple
visualization [34] has not abated. Other models of reader behavior. In an effort to formalize
investigators have sought to simplify the the notion of contour, we sought to model the
visualization problem by automatically discovering reader’s trajectory through the hypertext as a
reduced or clustered graphs adapted to individual Markov process. Starting at distinct points in the
reader’s interests [13, 22. 24]. Since the placement hypertext, and assuming that readem were equally
of individual links and the organization of small likely to follow any link (and that they completely
neighborhoods can be powerfully expressive, it ignored secondary navigation), we expected that
seems only natural to seek a hypertext’s meaning different probability distributions would emerge,
through study of its underlying graph structure. reflecting differences in contour.

Following Botafogo and Shneiderman [7a], we IIwilr’mmww


sought to identify clusters of meaning in hypertext
by identifying strongly connected components,
sections of the hypertext so interconnected that any
node in the section can be reached from every other
node. We first observe that secondary navigational
modes (bookmarks, history menus, navigational
undo, bmadcrumbs, search, and link-creation by
readers) transform almost any hypertext into a
single strongly connected componenc we are thus Figure 4. Transcripts of two sets of random walks in@ernaan, starting
forced to ignore them, even though Walker [56] from two different nades. The nades of the hypertext are laid out frwm
lcfi to right, and t)w probabiliq disoibwion of each successive
has shown that th~e are common and important generation is shawnfiom back to front. Random walk$ begin at a single
activities even for readers of unalterable technical node (rear of graph) but rapidly spread thraugh ths hypertert until most
nodes share a sin”larly small probability (front of gmph).
documents.
Botafogo and Shneiderman found that they could Much to our surprise, however, the observed
obtain meaningful component clusters in Hypertext distributions become very broad within ten or
Hands-On! [52] by ignoring highly textured nodes. twenty steps. Even in the large, episodic Victoty
Applying the same approach to more densely Garden (993 nodes, 2804 links) we could not find
interconnected hypertext fictions, such as kzme structure in the 15th-generation Markov
Pass [26] and Woe [32], we find it very difficult to distributions; random walks on the hypertext graph
obtain meaningful clusters. If the texture threshold starting horn a bunker in Sandi Arabia and from a
168 ACM ECHT CONFERENCE

campus nightclub each yielded similar probability [6] P. B@ghAndersen, “Towards and aesthetics of
profiles in which hunhds of nodes achieved hype-~xt systems: a semiotic approach,
essentially equal likelihood. Similar mults were Hypertext: Concepts, systetm and
obtained from Afternoon; sku$ng from two very applications, A. Rizk et al., eds., Cambridge
different places---places that readers usually University Press, Cambridge. 1990. pp. 223-
consider completely different episodes-we found 237.
that the random walks spread rapidly and
uniformly through the text. This contrasts starkly [7] Jay David Bolter, Wriiing Space: The
with our own reading experiences, where we often Computer, Hypertext, and the History of
find ourselves striving, to extricate ourselves from Writing, Lawrence Erlbaum and Associates,
an episode. Hillsdale. 1991.

The distributions are not, of course, identical, and [7a] Rodrigo A. Botafogo and Ben Shneiderman,
our failure to identify meaningful patterns in the “Identifying Aggregates in Hypertext
data does not demonstrate that such patterns do not Structures”, Hypertext 91, ACM, Baltimore,
exist. Still, random link choice is a very poor 1991. pp. 63-74
model for hypertext navigation. The experience of
reading these hypertext cannot be separated from [81 peter J. Brow, “Do we need maps to navigate
round hypertext documents?” Electronic
the construction of meaning and perception of
Ptiblishin&- Origination, Dissemination and
emergent contour.
Design 1 (1988) 45-53.

[9] 1Davida Chamey, “Comprehending Non-linear


Acknowledgments Texti The Role of Discourse Cues and
The authors are grateful to Nancy Kaplan, Martha Reading Strategies”, Hypertext ’87
Petry, Stuart Moulthrop, Eric Cohen, and Carolyn Proceedings, ACM, Baltimore, 1987. pp. 109-
Guyer, and to four anonymous referees, for helpful 120
comments on early versions of this paper.
[10] Jeff Conklin, “A Survey of Hypertext”, lEEE
Computer 20(1987) pp. 17 ff.
References
[111 Jeff Cd&I and M. L. Begeman, “gIBIS: A
references to hypertext are preceded by a bullet ~) Hypertext Tool for Exploratory Policy
Discussion”, ACM Trtins. Office Information
[1] Rudolf Amheim, Art and Visual Perception Systems6 (1988) 303-331.
(1974) Berkeley and Los Angeles.
[12] W. Bruce Croft and Howard Turtle, “A
[2] Mark Bernstein, “The Bookmark and the Retrieval Model Incorporating Hypertext
Compass: Orientation Tools for Hypertext Links”, Hyperwxt ’89, ACM, Baltimore, 1989.
Users”, SIGOIS Bulletin 9 (1988) pp. 3445 pp. 213-224.

[31 Mark Bernstein, “An apprentice that discovers [13] Donald B. Crouch, Carolyn J. Crouch, and
hypertext links”, Hypertext: Concepts, Glenn Andreas, “The Use of Cluster
systems and applications, A. Rizk et al., eds., Hierarchies in Hypertext Information
Cambridge University Press, Cambridge. Retrieval”, Hypertext ’89, ACM, BaMmore,
1990. pp. 212-223. 1989. pp.225-238.

[4] Mark Bernstein, J. David Bolter, Michael [14] Gilles Deleuze and Felix Guattari, “1837: Of
Joyce, and Elli Mylonas, “Architectures for the Refrain”, A Thousand Plateaus.Capitalism
Volatile HypertextS”, Hyperlext 91, ACM, and Schizophrenia. Trans. B. Massumi.
Baltimore, 1991. University of Minnesota Press, 1987.

[5] Mark Bernstein, “Deeply Intertwingled [1s1 Steven J. DeRose, “Expanding the Notion of
Hypertext”, Technical Communication (1991) Links”, Hypertexi ’89, ACM, Baltimore,
41-7. 1989. pp. 249-258.
MILANO, NOVEMBER 30- DECEMBER 4, 1992 169

[16] Laura De Young, “Linking Considered [27] Halasz, F.; “Refieetions on NoteCards Seven
Harmful”, Hypertext: Coticepts, systemsand Issues for the Next Generation of Hypermedia
applications, A. Rizk et al., eds., Cambridge Systems’”, Hypertext ’87, ACM, Baltimore,
University Press. Cambridge. 1990. pp. 238- 1987. pp. 345-66.
249.
[28] Michael Joyce, “SiretI Shapes Exploratory
[17] Jane Yellowlees Douglas, “Understanding the and Constructive Hypertext” Academic
Act of Reading: the WOE Beginner-s Guide Computing, November, 1988: pp. 11 ff.
to Dissection”, Writing On The Edge 2 (2)
1991. pp. 112-125. ●[29] Michael Joyce, Afternoon, a story, hypertext
for Macintosh computers, Eastgate Systems,
[18] Jane Yellowlees Douglas, Print Pathways and Cambridge MA, 1990.
Interactive Labyrinths: How Hypertext
Narratives Aflect the Act of Reading, doctoral [30] Michael Joyce, “Storyspace as a Hypertext
dissertation, New York University, 1992. System for Writers and Readers of Varying
Ability”, Hypertext 91, ACM, Baltimore,
[19] Umberto Eco. “Overinterpreting texts”. 1991. pp. 381-388.
Interpretation and overinterpretation, Stefan
Collini., cd., Cambridge University Press, [31] Michael Joyce, “A feel for prose Interstitial
Cambridge. 1992. pp. 45-66. links and the contours of hypertext”. Wriling
On The Edge 2 (2) 1992.
[20] Dennis E. Egan, Michael E. Lesk et al.,
“Hypertext for the Electronic Library? CORE ● [32] Michael Joyce, “WOE”, hypertext for
Sample Results”, Hypertext 91, ACM, Macintosh computers, Writing On The Edge
Baltimore, 1991. pp. 291-298. 2{2) 1992,

[21] Shelley Evenson and John Rheinfrank, [33] Marc Kaltenbach, Franqois Robillard and
“Tow~ds a Design Language for CIaude Frasson, “Screen Management in
Representing Hypermedia Cues”, Ilypertext Hypertext Systems with Rubber Sheet
89, ACM, Baltimore. 1989. pp. 83-92. Layouts”, Hypertext 91, ACM, Baltimore,
1991. pp. 91-106.
[22] Mark E. Frisse and Steve B. Cousins,
“Information Retrieval ti-om Hypertext [341 Patrick Lai and Udi Manber. “Flying Through
Update on the Dynamic Medical Handbook Hypertext”, Hypertext 91 Proceedings, ACM,
Project”, Hypertext 89, ACM, Baltimore, Baltimore, 1991. pp. 123-132.
1989. pp. 199-212.
[35] George P. Landow, “Relationally Encoded
[23] Fqmca Garzotto, Palo Paolini and Daniel Links and the Rhetoric of Hypertext”.
Schwabe, “HDM-a Model for the Design of Hypertext ’87, ACM, Baltimore, 1987. pp.
Hypertext Applications”, Hypertext 91, ACM, 331-44.
Baltimore, 1991. pp. 313-328. [36] George P. Landow, Hypertext: The
[24] Peter A. Gloor, “Cybermap Yet Another Way Convergen~’e of Contemporary Critical
of Navigating in Hypertext”, Hypertext ’91, Theory and Technology, Johns Hopkins Press,
ACM, Baltimo~, 1991. pp. 107-123. Baltimore, 1992.

[25] Robert @ushko.; “Design issues for multi- [37] Brenda Laurel, Computers As Theatre,
doeument hypertext”. Hypertext ’89, ACM. Addison-Wesley Publishing Company, New
Baltimore, 1989. pp. 51-60. York, 1991; Brenda Laurel, Tim Oren, and
Abbe Don, “Issues in Multimedia Interface
●[26] Carolyn Guyer and Martha Petry, “Izme Design: Media Integration and Interface
Pass”, hypertest for Macintosh computers, Agents”, C’HZ90, ACM, Baltimore, 1990. p.
Writing On The Edge 2(2), 133
170 ACM ECHT CONFERENCE

[38] Kathryn C. Malcolm, Steven E. Poltrock, and [49] H, Van Dyke Parunak, “Hypermedia
Douglas Schuler “Industrial Strength Topologies and User Navigation”, Hypertext
Hypermedia: Requirements for a Large ’87, ACM, Baltimore, 1987. pp. 43-50.
Engineering Enterprise”, Hypertext 91, ACM,
[50] Xavier Pintado and Dennis Tsichritzis,
Baltimore, 1991.
“Satellitw Hypermedia Navigation by
● [39] Judy Malloy, its name was Penelope, Affinity”, Hypertext: Concepts, systems and
hypertext for Macintosh and IBM-compatible applications, A. Rizk et al., eds., Cambridge
computers, Eastgate Systems, Inc, 1992. Univemity Press, Cambridge. 1990. pp. 274-
287.
[40] Cathy C. Marhsall, Frank G, Halasz, Russell
A. Rogers and William C. Janssen, Jr., [51] Jean-Frangois Rouet, “Interactive Text
“AquaneE A Hypertext Tool to Hold Your Processing by Inexperience (Hyper-)
Knowledge In Place”, Hypertext 91, ACM, Readers”, Hypertext: Concepts, systems and
Baltimore, 1991. pp. 261-276. applications, A. Rizk et al., eds., Cambridge
University Press, Cambridge. 1990. pp. 250-
-[41] Stuart Moulthrop, Victory Garden, hypertext 260.
for Macintosh computers, Eastgate Systems,
Watertown MA, 1991. . ●[52] Ben Shneiderman and G. Kearsley, Hypertext
Hands On?, hypertext for IBM-compatible
[42] Stuart Moulthrop, “Beyond the Electronic computers, Addison Wesley, Reading” MA,
Book A Critique of Hypertext Rhetoric”, 1989.
Hypertext 91, ACM, Baltimore, 1991. pp.
291-296. [53] Wolfgang Sctmler and John B. Smith,
“Authors Argumentation Assistark A
[43] Stuart Moulthmp, “Toward a Rhetoric of Hypertext-Based Authoring Tool for
Informating Texts”, ECHT 92 Proceedings Argumentative Texts”, Hypertext: Concepts,
(elsewhere in this volume) $ystemsand applications, A. Rizk et al., eds.,
Cambridge University Press, Cambridge.
[44] Stuart Moulthrop,” Notes toward another
1990. pp. 137-151.
statement on hypertext rhetoric 2-15-92”. Talk
presented to the Conference on College [54] Storyspace, hy@ext system for Macintosh
Composition and Communication, Cincinnati, computem, I%stgate Systems, Inc., Watertown
1992. MA, 1991.
[45] Elli Mylonas and Sebastian Heath, “Hypertext [55] Norbert A. Streiz, L Hanneman and Malcolm
from the Data Point of View: Paths and Links TMtring, “From Ideas and Arguments to
in the PerseusProject”, Hypertexm: Concepts, Hyperdocmnents: Traveling through Activity
systems und applications, A. Rizk et al., eds,, Spaces”, Hypertext 89, ACM, Baltimore,
Cambridge University Press, Cambridge. 1989. pp. 343-364.
1990. pp. 324-336.
[56] Janet H. Walker, Emilie Young, and Suzanne
[46] Jocelyne Nanard and Marc I%rE@ “Using Marines, “A Case Study of Using a Manual
Structured Types to Incorpmte Knowledge in Online”, Machine-Mediated Learning 3
Hypertext”, Hypertext 91, ACM, Bahirnore, (1989) pp. 227-41.
1991. pp. 329-344.
[57] Polle T. Zellweger, “Scripted Documentx A
[47] Jakob Nielsen, “The Matters That Really Hypermedia Path Mechanism”, Hypertext 89,
Matter for Hypertext Usability”, Hypertext 89, ACM, Baltimore, 1989 pp. 1-14.
ACM, Baltimore, 1989. pp. 239-248.

[48] Jakob Nielsen, Hypertext and Hypermedia,


Academic Press, 1990. pp.143-162.

View publication stats

You might also like