Conference PaperPDF Available

Comparing link marker visualization techniques - Changes in reading behavior

Authors:

Abstract and Figures

Links are one of the most important means for navigation in the World Wide Web. However, the visualization of and the interaction with Web links have been scarcely explored, although Links have severe implications on the appearance and usability of Web pages and the World Wide Web as such.This paper presents two studies giving first insights of the effects of link visualization techniques on reading habits and performance. The first user study compares different highlighting techniques for link markers and evaluates their effect on reading performance and user acceptance. The second study examines links-on-demand, links that appear when pressing a dedicated key, and discusses their possible effects on reading and browsing habits.The findings of the conducted studies imply that the standard appearance of link markers has seriously underestimated effects on the usability of Web pages. They can significantly reduce the readability of the text, and alternatives should be carefully considered for the design of future Web browsers.
Content may be subject to copyright.
Comparing Link Marker Visualization Techniques –
Changes in Reading Behavior
Hartmut Obendorf
ASI, Department of Informatics
University of Hamburg, Germany
obendorf@informatik.uni-hamburg.de
Harald Weinreich
VSIS, Department of Informatics
University of Hamburg, Germany
weinreich@informatik.uni-hamburg.de
ABSTRACT
Links are one of the most important means for navigation in the
World Wide Web. However, the visualization of and the interac-
tion with Web links have been scarcely explored, although Links
have severe implications on the appearance and usability of Web
pages and the World Wide Web as such.
This paper presents two studies giving first insights of the effects
of link visualization techniques on reading habits and perfor-
mance. The first user study compares different highlighting
techniques for link markers and evaluates their effect on reading
performance and user acceptance. The second study examines
links-on-demand, links that appear when pressing a dedicated key,
and discusses their possible effects on reading and browsing
habits.
The findings of the conducted studies imply that the standard
appearance of link markers has seriously underestimated effects
on the usability of Web pages. They can significantly reduce the
readability of the text, and alternatives should be carefully con-
sidered for the design of future Web browsers.
Categories and Subject Descriptors
H.5.4 [Information Interfaces and Presentation (e.g. HCI)]:
Hypertext/Hypermedia – navigation, user issues
General Terms
Design, Experimentation.
Keywords
Hypermedia, Input Interaction Technologies, Link Marking
Techniques, Links-On-Demand, Usability Survey
1. A HYPERLINKED WEB
The World Wide Web is today’s largest and most important
online information infrastructure. Its interface and structure is
determined by the concept of the hyperlink, a directed relationship
between two objects. Web hyperlinks usually connect a small
phrase, a graphic, or even only a single word with another
document in the Web address space.
These links form the primary navigational means of the Web.
Studies have shown that following links is the most frequent
action when using a Web browser: Catledge and Pitkow reported
52% of user actions clicks on links [10], Tauscher and Greenberg
reported a value of 42% [32].
Links were initially intended to establish semantic relationships
between related chunks of information. Though the derivation is
far from straight, Web hyperlinks are usually seen as originating
in Vannevar Bush’s historical article “As We May Think” from
1945 [9] and in the conception of nonlinear texts proposed by Ted
Nelson in 1974 [23]. According to their visions, links allow the
associative transition from one piece of information to another,
shaping “paths of thought”.
In the Web, however, links are not exclusively used to express
semantic associations but also to convey structure. The distinction
between associative and structural links can be made both tech-
nically and semantically: Associative links connect words or
phrases embedded in longer passages of text with other chunks of
information that relate to the meaning of the phrases in the
starting context. This user interface concept – clickable words
embedded in continuous text – can be traced back to Shneider-
man’s embedded menus [16]. Structural links, on the other hand,
are usually not embedded in paragraphs but in exposed locations
and are used to express and navigate logical structures [1].
Hereby, they usually form patterns, like hierarchies or sequences
or lead to landmark pages like the homepage of a site, a search
page or an index page, routing the users to other pages [27].
Our observation of current link usage on the Web suggests that
many more links are rather of structural than of associative char-
acter: most Web pages include navigation areas, often explicitly
located in a “navigation bar”. E-commerce sites use links mainly
to structure groups of articles or initiate actions, e.g. displaying of
product descriptions. However, when searching hundred of sites
for appropriate documents for our evaluations, we could find only
very few pages with extended text passages and a substantial
number of embedded associative links. This observation
correlates with the research results of Miles-Board, Carr and Hall,
who programmatically analyzed over 770,000 randomly selected
Web pages for highly linked content passages. Only 576 pages
(less than 1‰) were found that matched their requirements for
Web pages with substantial associatively linked continuous text
[20]. The current Web can be characterized as a sparse hypertext
[25, p. 114].
The rarity of links indicates that highly linked text may be judged
unfavorable by authors and cause readability problems. The basic
concept of the associative hyperlink might thus become less
usable through an inadequate visualization of link markers. At the
same time, enhancements to the Web that extend the simple links
seem to have only slow success; as we argue in the next section,
restrictions of the existing link visualization standard might hinder
this development.
2. NEW DEMANDS FOR THE WWW
The simplicity of the concepts of the Web is probably one of the
factors that helped it to expand and succeed so quickly. However,
Copyright is held by the author/owner(s).
WWW 2003, May 20-24, 2003, Budapest, Hungary.
ACM 1-58113-680-3/03/0005.
the limited linking means of the Web – links have to be embed-
ded, uni-directional and are usually un-typed – have repeatedly
been criticized [7; 8, pp. 39]. In contrast, rich hypertext systems
offer sophisticated support for structuring, editing, annotation and
navigation. All approaches that try to integrate such extended
functionality into the Web have to employ “workarounds” to
handle the weaknesses of the simple concepts of the Web.
In the last years, the XML language family, a series of new Web
standards, has been introduced to challenge these weaknesses and
change the face of the Web from the inside, thus creating new
demands for the user interface of web browsers. One of these
standards is XML Linking, a W3C recommendation that specifies
new hyperlinking facilities for XML documents.
The linking potential of XML linking is based on two key stan-
dards, which are necessary to create and describe links and link
anchors: An XLink consists of an arbitrary number of resources
and arcs. A resource is any addressable unit of information or
service, while arcs create directed relations between two resources
each [11]. Resource Anchors can be defined using XPointer,
which allows addressing different kinds of spans in XML docu-
ments [13]. These spans can vary from points to complex regions
and can even be distributed over the document, e.g. an XPointer
can be used to address a specific string in all citations of an XML
file. Thus, XML Linking allows the separation of structure and
contents by storing link information in link bases, servers or data-
bases dedicated for the storage of links. It becomes possible for
anyone to add links to the read-only material of the Web. Two
possible applications of these features are link bases for specific
topics or user groups, which can help to improve the navigation
on the Web, and personal or workgroup annotation systems. The
density of links could be significantly increased when external
storage of links becomes possible.
Though annotating printed articles and books is an essential prac-
tice for most readers, decent annotation support is still one
missing key feature for Web browsers [26]. As current annotation
implementations that make use of XPointer [34, 39] insert
additional markers that compete for the reader’s attention, both
discernability of links and readability of linked text become
increasingly important.
The XLink standard, allowing for both external link storage and
annotation of documents by readers, can lead to a much richer
interlinking between documents, and thus to a higher link density
within documents. It also significantly increases the chances that
the start of one link marker overlaps with the end of another link
marker, creating overlapping link markers, since different link
authors may choose the same words as part of their link anchors.
Furthermore, XPointer makes it more feasible to link large chunks
of information, such as paragraphs or tables, to other resources,
and the potential to refer to many resources in one XLink allows
to define links with multiple destinations.
Another advantage of XML Linking is its extended typing means
for hyperlinks. Link types describe the relationship between the
source and the destination of a link, often derived from semantic
categories like “explanation” or “example” [33]. They were intro-
duced to help users navigate in hypertext by giving them a better
idea of the link targets. Streitz et al. list semantic link information
as their “first principle of useful hypermedia system design” [31].
XLink defines both machine-readable and human-readable type
information that can be specified for the link as a whole, each
endpoint of a link and for every arc. Obviously, typed links are
only helpful if the user can distinguish the different types. How-
ever, most presentation and behavioral aspects of XLinks have
been deliberately excluded from the model, and so the realization
of a user interface for these concepts does not exist.
In fact, the question of how to visualize link type information in
the Web is not new. Although HTML links don’t appear to be
typed, even early standards, such as HTML 2.0 defined additional
anchor attributes to express link relationships more precisely [4]:
Web authors can set a link title and use the two attributes rel and
rev to set a forward and backward relationship type. HTML 4.0
identifies several of these types as useful like “contents”,
“subsection”, or “alternate" [28]. However, current browsers only
support the title attribute and show it in a little popup; no concept
for the visualization of the other two attributes has been realized
1
.
Looking closer, Web links also have several implicit types: links
can be local or lead to an external site, they can control the con-
tents of another frame or window, and they can use different pro-
tocols like news, ftp, or e-mail. However, this type of information
is hardly visible to the user, a flaw that can cause usability
problems [38]. Already the low discriminability of local and
external links was found to be problematic as users often do not
expect a link to lead to another site [24, pp. 45]. Besides, the two
different apparent link types realized in current browsers – purple
links for recently visited pages and blue for unvisited – already
help the user to navigate and avoid visiting the same page again
and again [24]. Although this shows that the distinction of
different link types is also important for the Web, different link
markers types for different link semantics have barely been
considered or evaluated.
For all these new concepts and use contexts the current underlined
links offer only poor support: Hypertext with a high density of
links becomes poorly readable, and overlapping links are hardly
realizable, as the beginning and end of link markers cannot be
visualized properly. Moreover, there is hardly any support for
typed links: except for blue and purple links, no standard exists
that allows the discrimination of different link types. The need to
develop and evaluate new concepts becomes even more urgent
with the arrival of new techniques and higher usability demands;
the time for a change in link visualization might have come. Still,
there has been little discussion on how the standard user interface
of Web links could be altered in modern Web browsers to enable
extended hypertext features.
3. AN UNDERESTIMATED PROBLEM:
THE VISUALIZATION OF LINK
MARKERS
Current link appearance is not only unsuitable for future require-
ments, it already causes several problems today. Historically, the
origin of the blue underlined links lies in Tim Berners-Lee’s
WWW browser prototype [5] and Marc Andreessen’s Mosaic [2].
The reasons for this choice of link marker appearance were of a
technical nature: it was simple to implement, and at that time most
computers had only either a 16-color or black-and-white display.
Blue was the darkest of the available colors, the closest to black
text; for monochrome displays, the text was underlined [22].
Although later versions also allowed for different link marker
1
On the contrary, they are usually used for technical concepts,
e.g. to include style sheets or to “prefetch” objects as realized in
the Mozilla browser from version 1.2 onwards (see:
http://www.mozilla.org/projects/netlib/)
appearances (boxed, double underline, thin underline, no
underline) from which any user could pick a favorite one, the
standard appearance was adopted by later browsers like Netscape
and Internet Explorer. Even today, the blue underlined links
prevail, and principal Web design guidelines recommend not
changing the appearance of links [17, 24, 30] for the sake of
consistency. They confine the length of link markers instead, e.g.
Nielsen recommends that link markers should be 3 to 5 words
long [24]. This, however, limits the expressiveness of links.
Nevertheless, more and more Web designers style their links, as
the standard blue underlined links exceedingly stand out in the
text. Furthermore, underlining is known to reduce the readability
of text significantly, as it changes the word shape and interferes
with descenders, letters that drop below the line like p, q and j
[18]. It is a well-established custom in Typography to avoid any
underlining and use the more subtle italic and bold typefaces to
highlight text instead [14, pp.35]. The point made here is,
however, still valid, as the standard appearance of links is
unchanged and relatively few authors or readers will take the
necessary steps to change the standard underlined links into some
more readable alternative.
For marking Links in graphics, the standard employed by Web
browsers is even worse: Linked images are framed with a blue
line. This method hardly harmonizes with any page design and so
most Web authors hide the border. Instead graphics have to be
(re-)designed to appear “clickable”. The conflict between the
illustrative function of an image and its interactive functionality
as link also poses problems for image maps. If, for example, a
map or chart shall include hyperlinked areas, they have to be
specifically designed to confer the presence of a link. A common
link standard for text and graphics that could emphasize active
regions without the need for specific image alterations would
decrease the unnecessary design effort and allow more flexibility
for graphical design. It would also help implement links in
interactive audiovisual presentations, as described by the W3C
standard SMIL
2
[29].
Figure 1: Link markers of Intermedia
Former hypertext systems used techniques that avoided a change
in typographic attributes such as underlining, weight or color to
highlight links: One option for marking links are graphic symbols,
in the Eighties, IRIS' Intermedia marked hyperlinks with little
arrow icons located between lines of text and above graphics,
showing the start of the link span but not its endpoint [40].
However, for Web pages this is hardly a feasible standard, as this
method occupies extra screen space and inserting additional ele-
ments can clutter the page layout.
2
The World Wide Web Consortium’s specification of SMIL, the
“Synchronized Multimedia Integration Language” can be found
at http://www.w3.org/TR/smil20/
A more promising method in the context of the Web is the change
of the background color as implemented by Hyper-G's browser
Harmony [3]. This has the advantage that the typeface and style
of the text can be chosen freely, no extra space is needed, and
overlapping links become possible (See fig. 2).
Looking closely, link markers are always a design compromise:
on the one hand the readability of the text ought to be kept as high
as possible; on the other hand the link anchors have to be
distinguishable from other text, as users have to interact with
links. For highly linked hypertext any way to visualize links can
be problematic: “After all, when everything is highlighted, then
nothing is really highlighted anyway.” [25, p. 114]. Not showing
the links could be an alternative. This, however, can also lead to
new problems as with the Symbolics Document Examiner [35]:
link markers were hidden so well that they were only highlighted
when the mouse passed over them. This forced a “hunt and peck”
search for active regions.
An alternative are links-on-demand, a technique that was intro-
duced in Storyspace [12]. This system drew boxes around link
anchors when the reader pressed particular keys, making links
evident on request and thus keeping the text pristine the rest of the
time. In fact, this was also the consensus solution after the Hyper-
text '87 demo sessions, when renowned hypertext designers could
first compare all existing systems side by side [5]. However, when
using links-on-demand, the interface designer must be aware of a
potential disadvantage: since links are not always visible, possibly
distracting mode switches have to be applied.
Figure 2: An overlapping link in Harmony
We listed diverse reasons why an alternative for underlined links
is needed, either to replace it as a standard, or to allow the cus-
tomization for different use scenarios. To be able to select the
appropriate visualization for links, we must find out more about
their intended use. In [37] it was argued that translucent overlays
seem to be appropriate from a conceptual point of view: For text,
they look like a change of background color, they can be applied
to any document without changing the layout and allow over-
lapping links (Fig. 3). Furthermore, overlays do not interfere with
text like underlines, and they can also highlight active regions in
graphics and image maps without the need for image redesign.
Lastly, they can support different link bases or link types by
different colors, and they can emphasize links to different degrees
by variations in the intensity of the overlay.
In the first study of this paper, we empirically compared overlay
link markers with the standard blue underlines; in a second study,
we tried to cast some light on the effects of showing links only on
demand and investigate the change in reading habits this might
induce.
4. STUDY 1: EFFECTS OF LINK
VISUALIZATION ON READING
PERFORMANCE
Our initial research indicated that small variances in the
appearance of link markers can theoretically have important
effects on the readability of the text. To estimate the extent of
these effects, we designed an experiment that could measure the
recognition performance of read phrases, as a closely connected
measure for the reading performance: The participants were asked
to both quickly and thoroughly read text on a Web page within a
limited time and thereafter answer questions relating to it. The
nature of this task was determined both by the necessity to put the
participants under strain, so that they would produce measurable
errors, and our intention of recreating a behavior common in Web
usage: When searching for information, users often only have
little time and read pages only superficially; the first, possibly
hastily executed click will make the current page disappear. We
assumed that the better the participants could perform the reading
task, the better they would recognize the words they had read. To
compare performance for both marked links and remaining text,
50% of the questions related to linked phrases and 50% to
unlinked phrases.
The participants were presented a series of short texts with 90-220
words (mean 155). Each text had 4 to 10 link phrases of 1 to 8
words (most had 3 words) length, so 7 to 11% of the content was
marked as a link. The texts were taken from one of Germany’s
most frequented news tickers and stripped of all navigation areas,
graphics and commercial advertisements (see Fig. 3). We
carefully selected a number of news from 1998 through 2000 that
the participants would neither remember nor identify as old news.
The text difficulty should be consistent and the content of
approximately the same interest to our participant group. Where
appropriate, the text was shortened or links were added to match
our consistency criteria and to achieve a similar link density. We
thus tried to minimize the disturbing factors caused by artificiality
and inconsistency of the Web pages as described by [15].
Different visualization techniques for link markers might change
the reception of the text in several aspects: First, link markers
might draw attention to the phrases highlighted by links, thus
making them stand out and improve the recognition performance
for questions regarding these phrases. At the same time, this
would decrease the overall reading performance and thus word
recognition of the unmarked text. Second, the highlighting could
also make the linked text less readable and thus decrease the
performance for the recognition of linked phrases, especially for
underlined text, as discussed before.
Figure 3: One example text showing overlay link markers
(left) and ordinary links (right). Text source: www.heise.de
Two different link visualization techniques were tested against the
control condition, text without links (PLAIN). The first
visualization (UNDERLINED) used standard blue, underlined
links as employed by all mainstream Web browsers. The second
technique used a light translucent blue overlay (Fig. 3) high-
lighting the words of the link anchor (OVERLAY).
We predicted the different visualizations to effect measurable dif-
ferences in the overall recognition performance: For the PLAIN
condition, the participants should perform best, as no distraction
from the text was present. Both the OVERLAY and UNDER-
LINED conditions should perform worse, possibly with a slight
edge for the OVERLAY condition, as its markers might be less
distracting than the underlined text and have less negative impact
on the readability of marked text. This difference should be most
clearly visible when questions relating unlinked text were con-
sidered; for linked phrases, the OVERLAY and UNDERLINED
conditions should yield a similar or better result as the attention of
the participant was attracted by the phrases marked as link.
4.1 Experiment Setup
12 participants took part in the test, all being unpaid volunteers
and students or staff of the informatics department. The mean age
was 25.3 years, two were female. All participants had extensive
internet experience and use the Web several times a week. Since
we expected the effects of different visualization conditions to be
quite large, we chose to test a comparatively small group in this
pilot study.
To reduce the influence that different degrees of interest in the
test items would have, we selected a very homogeneous user
group. The target group consisted of regular and experienced
internet users, as we wanted to assess the willingness of these
users to adopt changes in the Web interface. Also, familiarity with
the Web browser used in the study and the fact that we invited
only native speakers of German to participate reduced potential
problems not caused by our test tasks.
Figure 4: The evaluation tool for the first experiment
To minimize variations in the experiment environment, we
designed an evaluation tool that both let the participants use the
Web browser in a normal fashion and presented them with the
questions while allowing the input of answers. The Scone
framework [36] was chosen as basis for this tool. Scone supports
the development and evaluation of Web enhancements; here it
was also used to render the different link visualizations. The test
documents were filtered by Scone’s proxy component WBI [19],
which added the appropriate style sheets to every page. Scone
also provided an easily accessible mechanism to control the
browser and to record all navigational actions of the users.
The PC used in the experiment had two 17 inch color monitors
with a screen resolution of 1024 x 768. The evaluation software
was displayed on the left screen (Fig. 4) while the right screen
showed a full screen view of Microsoft
©
Internet Explorer 6 in
standard appearance. The participants were not allowed to change
this setup.
4.2 Experiment Procedure
The test started with a short introduction of the test setup, inclu-
ding the use of the two screens, and the test procedure, which was
in consequence moderated by the evaluation software: The
participants were presented with some training tasks and there-
after with three sets of five tasks. The sequence of visualization
conditions (PLAIN, UNDERLINED and OVERLAY) was altered
between participants, all permutations being tested equally often,
while the order of the test documents remained unvaried. A
change of link marker visualization was announced by a short
message. For each task one of the prepared Web pages was shown
for 35 seconds on the right screen; then the screen was auto-
matically blanked and four questions appeared on the left screen.
Each question had to be answered by choosing a short phrase via
multiple-choice: only one of three offered phrases was included in
the text read before. In addition, the participants had to rate the
certainty they felt when answering each question on a four-point
scale (“very certain”, “quite certain”, “uncertain”, and “guess”).
The answers to two of the questions were contained in phrases
emphasized by links markers (“marked”), the two others could be
found in unmarked text. The experiment thus had a 3x2 design
with three different visualization conditions and two types of
questions.
4.3 Results
In total, the 12 participants read 180 Web pages and answered 720
questions, 240 for each condition. Every single participant read 15
Web pages and answered 60 questions, 20 for each of the three
conditions. The participants could thus score between 0 and 20
correct answers for each condition. The mean number of correctly
answered questions was 14.1 out of 20 (71%) with a standard
deviation of σ=2.4. To reduce the noise introduced by guessing
(for every question, one of three possible answers had to be
selected), the participants’ rating of their certitude was used to
select valid answers: only answers of which the participants were
“very certain” or “quite certain” were finally evaluated as correct.
With these restrictions, the mean number of correctly answered
questions decreased to 10.6 of 20 (53.2%) with σ=3.1.
4.4 Task Performance
The recognition performance showed differences for the different
visualization conditions (Fig. 5). As predicted, PLAIN text
yielded the best results with a mean score of 11.58 out of 20
(57.9%) with σ=2.78. This performance was nearly matched in
the OVERLAY condition with a mean score of 11.25 out of 20
(56.3%) with σ=3.33. In the UNDERLINED condition, the parti-
cipants answered only 9.08 out of 20 answers correctly (45.4%)
with σ=2.87.
A univariate analysis of variance was employed to probe for sig-
nificant differences in the results. The dependant variable was the
number of correct answers, the independent variables were (a) the
visualization, and (b) the type of the task (text questions or link
questions). Both variables induced a significant difference in the
correctness of answers: (a) the visualization p=0.045, F
2,22
=3.256;
(b) task type: p=0.002, F
2,36
=10.460, marked phrases showed a
much better recognition performance. There was no measurable
interaction between the variables (p=0.46, F
2,5
=0.785).
0
2
4
6
8
10
12
14
PLAIN UNDERLINED OVERLAY
Figure 5: Mean number of correct answers
As Levene’s test hinted that the error variance was equal across
groups (p=0.633), we first employed a dependant t-test. The re-
sults hinted at a significant difference between the PLAIN and
UNDERLINED condition (p=0.022, T=2.167 dF=22). This was
confirmed by the conservative Scheffé test (p=0.08) on the assu-
med 10%-niveau. There was no statistical difference between the
PLAIN and OVERLAY conditions. The difference between the
UNDERLINED and OVERLAY conditions was significant only
in the dependant t-test (p=0.046, T=-1.75, dF=22), the Scheffé
test showed no statistical difference (p=0.134).
4.5 Subjective Results
The quantitative data illustrated differences in the recognition
performance of the participants in the experiment. To get a
broader view of what effect the change in link visualization might
have, we collected qualitative data in a semi-structured interview
after the test. Before the interview, the subjects were told that the
test conductors were not the developers of the evaluated tech-
niques, to avoid well-meaning ratings.
PLAIN UNDERLINED OVERLAY
very good (1)
good (2)
OK (3)
bad (4)
very bad (5)
Figure 6: Readability of the three text conditions
First the participants were asked how readable they rated the text
under each visualization condition. On a five point Likert scale
from 1 (very good readability) to 5 (very low readability) the
PLAIN condition was rated best (mean 1.1,
σ
=0.378). Both the
UNDERLINED (mean 2.6,
σ
=0.535) and the OVERLAY (2.8,
σ
=1.165) condition were rated less readable. Noteworthy is the
high divergence of answers for the OVERLAY condition: the
participants gave every answer from “very good” to “very bad”
readability (Fig 6).
Nearly all participants judged text without markers better to read
than text with link markers, except for one who preferred text
with overlay markers as “links highlight important facts”; three
participants rated all conditions equally readable.
Next, the participants were asked which of the two link marking
techniques emphasizes link phrases most. Here only two of the
twelve participants chose underlined links, but six participants
(50%) voted for the overlays. Several of these participants noted
that they found them more conspicuous, as they were “uncom-
mon”. Four participants could not rate one method over the other,
as they saw advantages and disadvantages in both approaches.
Finally the participants were asked how they liked overlay links
compared to underlined links and what characteristics of overlays
they saw as advantages or as weaknesses. The judgment for the
overlays was very heterogeneous. While almost half of the parti-
cipants were pleased with the new overlay method, nearly as
many participants disliked them and preferred the underlined links
(see Table 1).
Table 1: User judgment of overlays
How do you like the overlays compared to the underlined links?
much better better same worse much worse
1 4 3 2 2
The advantage most frequently mentioned for overlays was that
they were “less disturbing”, having “less contrast to the regular
text” and therefore preferable for reading text. Two participants
found the link presentation “more pleasant”. Furthermore, it was
mentioned that underlining could again be used to emphasize text.
Finally, the similarity to using highlighters on paper was per-
ceived as positive.
In contrast to the opinion above, two participants disliked the
overlays for emphasizing links too much and being even more
distracting than underlined links. Also criticized was the un-
familiar character of the overlays: one participant disliked their
"block-like" appearance, another noted that the overlays might be
problematic with colorful pages and a third person found the
alteration of the background of the text disrupting.
4.6 Discussion
As expected, this study shows that underlining words in a text
affects the recognition performance negatively. The underlined
links did reduce the recognition performance for our experienced
readers. This could be due to a combination of two effects: (1) the
readability is reduced, (2) underlines distract attention from the
unmarked text – especially as our participants were experienced,
this could be conditioned behavior. As the underlined links did
not yield a better performance when the participants had to
recognize linked phrases, any positive effect of the added
attention was lost, possibly due to the decreased readability of the
underlined link markers.
The new overlay technique worked very well, it outperformed
underlined links both overall, and in particular for link phrases.
There was no measurable difference in performance between the
overlays and plain text. This is remarkable, since plain text was
almost unanimously rated best to read by the participants. The
assumed performance gain in the OVERLAY condition for linked
text did not become significant, although the mean correctness
rates could suggest an effect here. This calls for further investiga-
tion with a larger number of participants than in this first study. A
possible interpretation for better performance of overlay markers
would be that the overlays directed attention to the marked text,
without having a strong negative effect on its readability.
The observed effect of questions regarding marked vs. unmarked
text in all visualization conditions can partly be explained, as link
markers were often names of companies or people, which the
participants remembered well. A further study could systema-
tically vary the linked phrases.
As this is the first study of its kind, careful interpretation is
needed. The findings are very much what we expected, and the
explanation seems simple, but further studies with a larger
number of participants will have to reinforce and differentiate the
findings presented here. The observed effects will partly depend
on the intensity and color of the overlay – we chose a rather
strong blue for the overlays to give them a distinctiveness com-
parable to underlined links and reduce the potential effects of
different colors. Thus, we may have strengthened the emphasizing
effect more than necessary. On the other hand, being able to
adjust the strength of the overlays is an important advantage com-
pared to underlining: The emphasizing effect can both be tailored
to the user’s preferences and his tasks, as well as be fine-tuned to
match the design and the layout of a Web page.
Our results indicate that the readability of text with underlined
links in Web browsers is impaired. Although our Web pages
could be considered sparsely linked, the effects were measurable
and supported by the subjective results: all participants preferred
text without link markers for reading.
The design implications for the Web are clear: underlined links
should not be used when readability of text is the main concern.
Overlays could be a superior alternative for link markers. Simple
colored text, of which an increasing use can be observed in the
World Wide Web, could be a viable alternative. It was not
included as a visualization condition, as we expected only small
differences to plain black text, that would not have become
significant with our small sample of participants. Also, overlay
markers have the advantage that they work for both text and
graphics.
A degraded readability of hyperlinked text and the strong
emphasizing effects of underlined links could add to the reasons
for people scanning – instead of reading –Web pages. In a study
conducted by Morkes and Nielsen, 79% of the participants always
scanned Web pages first [21]. This is consistent with our results:
ten out of twelve participants described their reading habits on the
Web as scanning, glancing over the text and looking primarily at
headlines and links. Eight participants preferred a printout for
reading.
This fuels doubt if underlined links are an appropriate solution for
a digital paper world. In the long run, their weaknesses might
have added to the reasons for the scarcity of associative links in
the Web. Although there are other reasons for authors not to use
associative links (e.g. they are often more difficult to create and
maintenance is costly), if underlined links hurt reading
performance, they are used sparsely. And, if there are but a few
links, there is less need to change the visualization.
But as the effects of link marker visualization on reading will be
even stronger when extended linking mechanisms are employed
and rich linking becomes more common, a concept is needed to
counter the negative effects of hyperlinks and preserve the
readability of Web pages. In the second study we thus
investigated, whether showing links only on demand would have
effects on Web readers.
5. STUDY 2: IMPLICATIONS OF LINKS-
ON-DEMAND
The first study presents evidence that link marker visualization
does have a significant effect on how people read Web pages:
readability is decreased when underlined links are added to a Web
page. One idea how to reduce the distraction link markers cause
when reading a text – which was in fact already agreed upon as an
optimal solution by hypertext experts – are links-on-demand [6].
This describes a technology where users have to depress a button
to make the links visible. We wanted to evaluate whether hiding
the links would affect the way people interact with hypertext (i.e.
if they would stop scanning over the text), and how this change in
behavior would be received. Thus, we did not only collect data as
part of a formalized task, but also used an evaluation of
qualitative interviews and observation notes.
The study compared the performance impact of links visible only
on-demand, i.e., when a user pressed a key
3
(ON-DEMAND),
with the typical always visible underlined links (ALWAYS). We
designed a set of experimental tasks where the participants were
presented a short text with 80-239 words (mean 158.8, σ=41). In
this text, from 3 to 11 links (mean 5.9, σ=1.96) pointed to other
WWW resources, the links ranged from 1 to 6 words in length
(again, most links were 3 words long). The mean ratio of words
per link was 27, thus 5 to 10% of the text was marked as a link.
The texts were taken from the same news ticker and prepared as
described in study one. The hardware and software setup was
similar, too.
5.1 Experiment Procedure
16 participants took part in the experiment, all unpaid volunteers,
being either student or staff of the informatics department. The
age of the 16 participants ranged from 21 to 36, with an average
of 26.1, three were female. Again, the participants were all native
speakers of German with extensive Web experience.
Figure 7: The evaluator tool for the second study
The participants were instructed to find the answer to a given
question as fast and as accurately as possible. A button labeled
“start” had to be pressed in the test tool to commence a task. Con-
sequently the Web page with the experimental text appeared in
3
In this study the control key of the PC keyboard was used, as it
does not have effects on clicks on links.
the browser on the right screen. The answer to the given question
could be found in two ways: Either it was already contained in the
text given (TEXT task), or a single link had to be followed and
the answer could be found on the next page (LINK task). We
created a TEXT and a LINK task for every page. The task types
and the link rendering conditions were controlled for all
participants, while the order of the test pages remained
unchanged. When the participants felt that they had found the
answer, they had to press a button labeled “stop” and enter the
answer in a text field (Fig. 7).
Performance was measured by correctness of the answer, time
needed to complete the task, and selected links. For the TEXT
tasks, the time between pressing the start button and the stop
button was rated. For the LINK task, time was measured until the
first link had been selected. A TEXT task was answered correctly
when the users entered a correct answer and did not select a link.
LINK tasks were solved correctly, when the user initially selected
the right link. In both conditions, thus, time was gauged only until
the relevant information had been found and disturbing factors
caused by uncontrolled target pages were minimized. Because the
method of measurement differs, however, measurements cannot
be compared across conditions.
We expected a change of reading behavior in the ON-DEMAND
condition: as links are initially invisible, we assumed the par-
ticipants would read or scan the text completely before selecting a
link. Accordingly, for TEXT tasks we predicted a lower number
of erroneously selected links and a faster time to find the answer.
However, for the LINK questions we predicted a slower response
time, as participants would not be tempted to scan the links first.
If the participants failed to find the correct link and instead tried
to answer the question without following a link, the number of
errors would increase.
5.2 Results
The design of the experiment depended on the subjects’ per-
formance in locating information asked in a question. 16 par-
ticipants had to answer 12 questions each. The mean number of
correctly solved tasks was 10.5 of 12 (88%) with a standard
deviation of σ=1.46. In total, the subjects read 192 documents, 96
of both conditions. These two conditions consisted of two task
types each: 48 LINK tasks and 48 TEXT tasks.
5.3 Task Performance
As expected, the participants did make more mistakes when
finding desired information in the TEXT tasks while the links
were ALWAYS visible (13 errors, 27,1%) compared to the ON-
DEMAND condition (6 errors, 12,5%). A
χ
2
test yielded a 1-sided
significance of p=0.0365 (
χ
2
=3.215, df=1). However, the par-
ticipants made significantly more mistakes when answering a
LINK question in the ON-DEMAND condition: 5 of the 48 tasks
(10.4%) were solved incorrectly compared to no errors in the
ALWAYS condition (Fig. 8). As the
χ
2
test could not be used
here due to the lack of errors in the ALWAYS condition, Fisher’s
Exact Test was used; the result had a significance level of
p=0.028.
Percentage of Incorrectly Solved Tasks
27,1%
12,5%
0,0%
10,4%
0%
5%
10%
15%
20%
25%
30%
TEXT Task LINK Task
A
LWAYS ON DEMAND
A
LWAYS ON DEMAND
Figure 8: Comparison of error rates
Comparing the task execution times, the ALWAYS condition
showed excellent results: TEXT tasks were answered in similar
speed in both conditions: the condition ALWAYS had a mean
task completion time of 20.5s (σ=13.4), whereas in the ON-
DEMAND condition, users needed 21.2s (σ=13.4). LINK tasks
were answered in 18.1s (σ=11.7) in the ALWAYS condition,
compared to 23.8s (σ=15.1) for ON-DEMAND links (Fig. 9). A
dependant t-test indicated this difference to be significant
(p=0.044, t-value= -2.014). The high standard deviation in task
completion time was highly affected by interindividual differ-
ences in the participants’ reading habits.
Taks Completion Mean Time
20,5 s
21,3 s
18,1 s
23,8 s
s
5 s
10 s
15 s
20 s
25 s
30 s
35 s
A
LWAYS ON DEMAND
A
LWAYS ON DEMAND
TEXT Task LINK Task
Figure 9: Comparison of task completion times
5.4 Subjective Results
Again, the acceptance of this new technique varied extremely
(Table 2). Four participants rated links-on-demand “much better”,
five “better” than underlined blue links. However, five
participants liked the new technique less than ordinary links and
one judged it to be “much worse”. One was uncertain.
Table 2: User judgment for links-on-demand
How do you like links-on-demand
compared to always visible links?
much better better same worse much worse
4 5 1 5 1
The participants were also asked for their opinion on the advan-
tages and drawbacks of the on-demand technique. 10 of the 16
participants appreciated text without links as more readable and
therefore perceived the links-on-demand as less distracting and
interfering. Several explanations were given for this positive
judgment: Permanently visible links “press” the reader to scan the
text and to click on the links without reading the text. The
unlinked text appears “steadier” and “less distracted” and it is
therefore more motivating to read. Furthermore, hypertext with
links on demand looks more like ordinary text. Finally, three par-
ticipants stated they thought this technique would improve the
design of most Web pages a lot, as the links often look “hideous”.
On the other hand five participants criticized that the mode
switches with the control key were disturbing, as it created an
extra effort to press the key and was an unfamiliar extra inter-
action with the browser. Two participants even found links-on-
demand more distracting, as they felt they had to press the key
almost continuously to be able to solve the tasks.
Finally, six users stated that the on-demand links were not
suitable for the experimental tasks, as they asked for specific
information in the text where the highlighted links were often
useful (for every LINK task); in reality, links would be usually of
less help.
5.5 Discussion
Links-on-demand seem to be a large step away from the hypertext
interaction paradigm of the Web. While some users readily em-
braced this new interaction technique, others showed rejection.
Though two users tried to preserve their everyday practice by
depressing the key continuously, our quantitative data suggests
that an overall change in reading habits was induced:
As we predicted, the ON-DEMAND condition produced less
errors for TEXT tasks; in the ALWAYS condition 10 out of 13
errors were caused by quickly selecting a link before reading the
whole page.
In contrast, when the participants had to follow links to find the
answer, they were more reluctant to do so: 4 out of 5 errors pro-
duced for LINK tasks were produced this way. Instead of scan-
ning the text for links and immediately following a link even if it
seems only remotely connected to the desired information, the
participants often read through the Web page before bringing up
the links. This is consistent with the longer time needed for LINK
tasks in the ON-DEMAND condition, the extra time needed can
also be partly explained by the extra effort of depressing the con-
trol key.
Again, we are faced with a trade-off here: while always-visible
links allow for fast interaction with links, they inhibit reading the
text on the Web page. If a text is to be read completely, it should
not contain always-visible links. Most German online-newspapers
have apparently realized this and now give a table of links at the
end of the text instead of using inline links. Thus, links are used
like footnotes in ordinary text. It seems to be worthwhile to
consider the use of links-on-demand in such cases.
6. CONCLUSION
This paper presented two studies that investigate the effects of
different link visualization techniques. We believe that further
research is needed in this area, as the perception of and interaction
with Web pages is influenced significantly by the look of the
links. To our knowledge, no previous studies exist that try to
investigate the changes in user behavior induced by link appear-
ance. On the contrary, the choice of current link visualization is
quite simplistic and seems almost arbitrary. But, as our studies
show, deliberate design is vital, since small changes in appearance
can cause both measurable and subjectively noticeable differences
in usability.
The first study showed that alternatives to blue, underlined links
exist and compare well: Underlined links seem to substantially
decrease the reading performance on Web pages and may add to
the reasons why users don’t like to read on the Web. Translucent
overlays did not only avoid the disadvantages of underlined links
in our study, they also offer many advantages such as being more
flexible, applicable to graphics and ready for future requirements
like those that will be introduced by the W3C’s XML linking
standard and other rich hypermedia concepts.
Both literature [21] and our participants state that reading on the
Web can often be described as “scanning” for links. We believed
that this behavior might be connected to the fact that links are
very apparent in current Web pages and excessively attract the
attention of the users. To try an alternative, we investigated the
effects of showing links only on-demand. The second study
proved that the interaction of users can change fundamentally if
links are not always displayed: both the measured results and our
observations indicated that users did mostly not scan the texts for
links but started reading when links were initially hidden. The de-
monstrated trade-off will have to be considered when Web pages
are designed to be either quickly scannable or easily readable.
The availability of better link visualization concepts, the change
in reading behavior and the unexpected readiness most partici-
pants displayed in trying out the new interaction approaches show
that the existing blue-underlined link standard does not suffice.
The effected distraction might be part of the reason for the bad
readability of Web pages; it might even cause problems for the
user, as underlined hyperlinks have less than optimal readability,
and navigational targets might not be immediately discernable.
Moreover, as tasks are diverse and user needs change for different
tasks, a single standard for link markers can not meet the
requirements of all users. For some, consistency will rule out all
other considerations; for others, an adaptive link visualization
strategy might be preferable. A wide variety of tasks could also be
supported by the use of overlays that provide for link markers in
both text and graphics and can be adjusted in intensity to fit the
task’s needs. Link style-sheets could be introduced to style the
appearance of certain types of links individually for both users
and authors of links.
ACKNOWLEDGEMENTS
We would like to thank Horst Oberquelle and Winfried Lamers-
dorf for their generous support, and our colleagues at Hamburg
and elsewhere for their ideas and criticism. Also, we would like to
thank all participants for their patience and their comments.
REFERENCES
[1] Allan, J.: Automatic Hypertext Link Typing. Proc. of the 7th
ACM Conference on Hypertext, 1996.
[2] Andreessen, M.: NCSA Mosaic Technical Summary. NCSA,
University of Illinois, 1993.
[3] Andrews, K.: Browsing, Building, and Beholding Cyber-
space, New Approaches to the Navigation, Construction, and
Visualisation of Hypermedia on the Internet, Ph.D. Thesis,
Graz University of Technology, 1996. ftp://ftp.iicm.tu-
graz.ac.at/pub/keith/phd/
[4]
Berners-Lee, T. and Connolly, D. (Eds.): HTML 2.0 Specifi-
cation: The Anchor-Element. W3C, 1995.
http://www.w3.org/MarkUp/html-spec/html-spec_5.html
[5] Berners-Lee, T.: The WorldWideWeb browser, 1990.
http://www.w3.org/People/Berners-
Lee/WorldWideWeb.html
[6] Bernstein, M.: HypertextNow: Showing Links, Eastgate Sys-
tems, 1996. http://www.eastgate.com/HypertextNow/
archives/ShowingLinks.html
[7]
Bieber, M., Vitali, F., Ashman, H., Balasubramaniam, V. and
Oinas-Kukkonen, H.: Fourth Generation Hypermedia: Some
Missing Links for the World Wide Web. Int. J. of Human-
Computer Studies, Vol. 47 (1), pp. 31-65, 1997.
[8]
Bouvin, N.O.: Augmenting the Web through Open
Hypermedia. Ph.D. Thesis, Department of Computer
Science, Aarhus University, Denmark, 2000.
[9] Bush, V.: As We May Think. The Atlantic Monthly. July
1945. Reprinted in Interactions, Vol. III.2, 1996.
[10] Catledge, L. D. and Pitkow, J. E.: Characterizing Browsing
Strategies in the World-Wide-Web. Computer Networks and
ISDN Systems 27(6), pp. 1065-1073, 1995.
[11] DeRose, D., Maler, E. and Orchard, D. (eds.): XML Linking
Language (XLink), W3C Recommendation, June 2001.
http://www.w3.org/TR/xlink
[12]
Eastgate Systems: Storyspace hypertext writing environment
for Macintosh computers, 1991.
[13]
Grosso, P., Maler, E., Marsh, J. and Walsh, N. (eds.): XPoin-
ter Framework, W3C Working Draft, July 2002,
http://www.w3.org/TR/xptr-framework/
[14]
Gulbins, J., Kaufmann, C.: Mut zur Typographie, Springer,
Berlin, 1993.
[15]
Kaasten, S., Greenberg, S., Edwards, C.: How People Re-
cognize Previously seen Web Pages from Titles, URLs and
Thumbnails. Report 2001-692-15; Dept. of Computer
Science, Univ. of Calgary, Alberta, Canada, 2001.
[16] Koved, L. and Shneiderman, B.: Embedded menus: selecting
items in context. Communications of the ACM, 29, pp. 312-
318, 1986.
[17] Lynch, P.J. and Horton, S.: Web Style Guide: Basic Design
Principles for Creating Web Sites, Yale Univ. Print, 1999.
[18] Lythgoe, J.N.: The Ecology of Vision, Oxford University
Press, Oxford, 1979.
[19] Maglio, P., Barrett, R.: Intermediaries Personalize
Information Streams. Communications of the ACM, Vol.
43(8), pp. 96-101, August 2000.
[20]
Miles-Board, T., Carr, L., Hall, W.: Looking for Linking:
Associative Links on the Web. In: Proc. of Hypertext’02,
ACM Press, pp. 76-77, 2002.
[21]
Morkes, J., Nielsen, J.: Concise, SCANNABLE, and
Objective: How to Write for the Web, 29
th
Sept. 2000.
http://www.useit.com/papers/webwriting/writing.html
[22] NCSA: Mosaic(tm) for MS Windows User Guide, National
Center for Supercomputing Applications, University of
Illinois, 1996. http://www.ncsa.uiuc.edu/
/SDG/Software/mosaic-w/releaseinfo/2.1/index.html
[23] Nelson, T. H.: Computer Lib/Dream Machines, Mindful
Press, 1974.
[24] Nielsen, J.: Designing Web Usability: The Practice of
Simplicity. New Riders, Indianapolis, 2000.
[25] Nielsen, J.: Hypertext and Hypermedia, Academic Press,
1993.
[26] O'Hara, K., and Sellen, A.: A comparison of reading paper
and on-line documents. Proc. of the ACM SIGCHI Conf. on
Human Factors in Computing Systems, 1997.
[27] Pirolli, P., Pitkow, J. et al.: Silk from a Sow's Ear: Extracting
Usable Structures from the Web. Proc. of ACM SIGCHI
Conf. on Human Factors in Computing Systems, Vancouver,
Canada, 1996.
[28]
Raggett, D., Le Hors, A. and Jacobs, I. (Eds.): HTML 4.0
specification. W3C, 1998.
[29]
Rutledge, L., Schmitz, P.: Improving Media Fragment
Integration in Emerging Web Formats. In: Proc. of
Multimedia Modeling 2001 (MMM01), Amsterdam, NL, pp.
147-166, 2001.
[30] Spool, J., Scanlon, T., Schroeder, W., Snyder, C., DeAngelo,
T.: Web Site Usability: A Designer’s Guide. Morgan
Kaufmann Publishers, 1999.
[31] Streitz, N., Haake, J., Hannemann, J., Lemke, A., Schuler,
W., Schütt, H., Thüring, M.: SEPIA: A Cooperative Hyper-
media Authoring Environment, Proc. Of the 4th ACM
Hypertext Conference Proc. (ECHT'92), Milan, Italy, pp. 11-
22, 1992.
[32]
Tauscher, L. and Greenberg, S.: How People Revisit Web
Pages: Empirical Findings and Implications for the Design of
History Systems. In Int. J. of Human-Computer Studies
47(1), pp. 97-137, 1997.
[33] Trigg, R.: A Network-Based Approach to Text Handling for
the Online Scientific Community, Ph.D. Thesis, Dept. of
Computer Science, Univ. of Maryland, 1983.
[34] Vatton, I., Kahan, J., Carcone, L. and Quint, V.: Amaya: The
W3C Editor/Browser, 1996-2003.
http://www.w3.org/Amaya/
[35]
Walker, J. H.: Document Examiner: Delivery Interfaces for
Hypertext Documents, Proc. of ACM Hypertext '87 Confer-
ence, Chapel Hill, USA, pp. 307-323, 1987.
[36]
Weinreich, H., Buchmann, V. and Lamersdorf, W.: Scone:
Ein Framework zur evaluativen Realisierung von
Erweiterungen des Webs. In Proc. of KiVS, Springer, Berlin
Heidelberg, 2003.
[37] Weinreich, H., Obendorf, H. and Lamersdorf, W.: The look
of the link - Concepts for the User Interface of Extended
Hyperlinks. In Proc. of the ACM Hypertext 2001
Conference, Aarhus, Denmark, pp. 19-28, 2001.
[38] Weinreich, H., Lamersdorf, W.: Concepts for Improved
Visualization of Web Link Attributes. Computer Networks,
33, pp. 403-416, 2000.
[39]
Wilson, M., Daniels, D. and Philips, J.: Annozilla – Annotea
on Mozilla, 2003. http://annozilla.mozdev.org/
[40]
Yankelovich, N., Haan, J. B., Meyrowitz, N. K. and Drucker,
S. M.: Intermedia: The Concept and the Construction of a
Seamless Information Environment, IEEE Computer, Vol. 21
(1), pp. 81-96, 1988.
... Os links na Web potencializam o rápido deslocamento do interagente através de um hipertexto. São considerados a principal forma de navegação na Web (Obendorf e Weinreich, 2003) e a essência da tecnologia hipertextual (Landow, 1997). Logo, poderia parecer desnecessária uma maior discussão sobre o funcionamento desse elemento tão básico. ...
... O referido potencial do link em estabelecer relações semânticas, conforme lembram Obendorf e Weinreich (2003), tem raízes no histórico artigo "As we may think" 20 de Vannevar Bush (1945). Nesse trabalho, que apresentava sua proposta teórica do sistema Memex, a argumentação partia da crítica de que a tradicional organização hierárquica de informações não reflete o processo mental, que trabalha, segundo aponta, por associações. ...
... 41 Co-links e a construção coletiva de links multidirecionais Conforme a crítica encaminhada até este momento, a funcionalidade dos links textuais na Web permanece muito próxima da formulação original de Berners-Lee. Não se pode deixar de apontar que, conforme lembram Obendorf e Weinreich (2003), existem dois tipos básicos de links: os associativos e os estruturais. Os primeiros aparecem dentro dos parágrafos do hipertexto como "clickable words embedded in continuous text" (Obendorf e Weinreich, 2003, p.1), 47 conectando aquele texto a outras informações, dentro da mesma página ou presentes em outros sites. ...
... Teqniques" [34] og "The Look of the Link" [35] vaert essensiell for avhandlingen. ...
... Dette viser seg å minske lesbarheten betraktelig. I artikkelen "Comparing Link marker visualization techniques" [34], viser det seg at den blå understrekte lenken forstyrrer så mye i teksten at den fordobler feilraten ved tekstbasert oppgaveløsning, i forhold til tekst uten lenker. Annen utforming av lenkene viser å gi et bedre resultat enn de blå understrekte lenkene [34]. ...
... I artikkelen "Comparing Link marker visualization techniques" [34], viser det seg at den blå understrekte lenken forstyrrer så mye i teksten at den fordobler feilraten ved tekstbasert oppgaveløsning, i forhold til tekst uten lenker. Annen utforming av lenkene viser å gi et bedre resultat enn de blå understrekte lenkene [34]. Lenker i dokumenter reduserer lesbarheten til dokumentene betraktelig, design av lenker er derfor meget kritisk når lenker skal integreres i slike system. ...
... We have implemented the system called WPM (Web Page Marker), which provides marking and anchoring functions on ordinary web browsers. WPM is designed to realize a effective system, which is one of Web browsing support systems [9]. WPM users can mark words and phrases on web pages by using their browsers. ...
Article
Full-text available
We propose an online method WFE for editing any Web page by realizing a proxy agent. The main strength of WFE is that the method can provide users with a way to edit any web page regardless of its original author. When a user access a web page via the proxy agent, the proxy agent augments the web page with client-side scripting via JavaScript that provides the user with the editing functions. By utilizing the XHTML DOM, the script included by the proxy agent allows the user to simply select any portion of text on the page and then directly edit the underlying HTML even if the selected text appears in many different places in the document. WFE identifies the specific instance of the selected text to be found. We also describe the system called WPM to show how effectively the method WFE can be used. WPM provides marking and anchoring functions on ordinary web browsers. The users can mark words and phrases on web pages by using their browsers without any extra plug-ins like similar systems. The result shows that the method WFE is effective in developing Web applications such as the system WPM.
... We will come back to the influence of the breadth effect in the theoretical section on information foraging. Obendorf & Weinreich (2003) conducted two studies on the visual aspects of associative links -links that appear in the running text. Their study emerged from the observation that associative links are relatively rare on the Web. ...
... Os portais, repositórios e mecanismos de busca detêm uma grande quantidade de links, com duas funcionalidades principais: expressam associações semânticas ou de estrutura. (Obendof e Weinreich, 2003) Há mais links estruturais que associativos na web, em geral ligados a barras de navegação. No âmbito dos periódicos, a presença de citações de documentos disponíveis on-­‐line, passíveis de serem acessados, é a principal vantagem do artigo eletrônico em relação ao impresso. ...
Article
Full-text available
The present article discusses the electronic scientific journals from the perspective of the new strategies of visibility of science in the web. The analysis is part of the research “The communicative elements of the scientific journals and the relationship with the printed and on-line environments”, which develops a pilot study at the Federal University of Rio Grande do Sul (UFRGS). It argues about visibility, through links, metadata, portals and open files, as the vital condition to the quality and credibility of scientific publications and to the research work. It points out impasses and gaps which are present in the co-living and transition of printed and electronic means.
... Os portais, repositórios e mecanismos de busca detêm uma grande quantidade de links, com duas funcionalidades principais: expressam associações semânticas ou de estrutura. (Obendorf & Weinreich, 2003) Há mais links estruturais que associativos na web, em geral ligados a barras de navegação. No âmbito dos periódicos, a presença de citações de documentos disponíveis on-line, passíveis de serem acessados, é a principal vantagem do artigo eletrônico em relação ao impresso. ...
Article
Full-text available
The contemporary editorial production reveals the tensions between the traditional print world, with its physical limits of information, and the on-line space as a continuous place for information. This article argues about science visibility on the web, addressing topics such as identification, description and exchange of electronic resources in relation to design and to the quality and credibility of scientific publications. The unique characteristics of the editorial process as well as public-readers of these publications are taken into particular account.
Article
Full-text available
Discussions on democratic online participation, cooperation in digital communities and network openness have highlighted the communication processes through services such as chat rooms, forums and weblogs. This article, however, aims to focus on the politics of hypertext. But, instead of solely focusing on the quantity of links offered for a multidirectional reading, this research is dedicated to the possibility of participatory creation of links. Until now, associative links on the Web were still unidirectional vectors (with rare exceptions), and link creation remained normally on the hands of who had access to the server, to the code and knew HTML. Facing these limitations, Alex Primo conceived and Ricardo Araújo programmed the technology of co- links (http://www.co-links.org), aiming to surpass that gap. This new service, which will be here introduced, opens Web pages to cooperative creation of multidirectional links. Co-links technology may then open hypertext to the participatory construction of a social memory on the Web, registering the mental associations made by the participants between that text and others read before. It is also suggested that this system could be of particular interest to educational projects: aiding cooperative processes, promoting group activity and collective writing. In order to test that hypothesis, this paper analyses the use of co-links technology by two study groups, located in two different cities in Brazil. These groups of students using discussion list, blog, a Wiki system and co-links technology, produced a cooperative article on cyberculture and later included links and co-links to the document. This paper presents the results of the investigation that followed the whole process of cooperative construction of knowledge. The debates, the openness of the system to collective writing of the hypertext, the potential and difficulties of participatory online communication are then discussed.
Article
Full-text available
Studies of how users read on the Web found that they do not actually read: instead, they scan the text. A study of five different writing styles found that a sample Web site scored 58% higher in measured usability when it was written concisely, 47% higher when the text was scannable, and 27% higher when it was written in an objective style instead of the promotional style used in the control condition and many current Web pages. Combining these three changes into a single site that was concise, scannable, and objective at the same time resulted in 124% higher measured usability.
Article
This paper discusses methods to generate and display automatically additional hyperlink information to the users of the World Wide Web. Current Web browsers make it hard to predict what will happen if a link is followed: users get different information than they expect, a new window may be opened, a download starts, or the destination object is just not available. Instead of giving an appropriate notification in advance, users have to follow a link, check whether the document contains the expected information, get back, try another link etc. However, usually it is possible to obtain additional hyperlink information from several sources like link anchor tags, the user's history and Web servers. Furthermore, with little enhancements, Web servers may include even more additional information to the hyperlinks in Web documents. These can be displayed before users select a link to improve navigation and reduce the cognitive overhead. In this paper several types of Web hyperlink information are listed, potential methods to present these facts are compared, the prototype implementation of the proposed concept — called by us HyperScout — is presented, and further developments are discussed.