Browsing Experiments
In [DeVocht-94] a number of browsing
experiments are described, combining different hyperdocument
structures,
different navigation aids
and different browsing strategies.
The purpose of these experiments was to find out which structures,
navigation aids and browsing strategies are best.
Although the range of possible combinations was large, the experiments
lead to some clear conclusions, useful for both authors, readers
and hypertext system builders:
-
The structures that work best for all combinations of navigation aids and
browsing strategies, for long and short browsing sessions, are those that
have about as many cross-reference links as they have structural
(hierarchical) links.
-
The highlighting links
facility works best for improving the browsing efficiency in all cases.
-
The depth-first navigation strategy works best in all cases except
for short browsing sessions in documents with very few cross-reference links.
Furthermore, this optimum is stable: strategies that deviate a bit from
depth-first, like the depth-first strategies 1, 2 and 3, also perform well.
The fish-eye view
facility was not considered in the experiments because
the experiments were aimed at discovering the structure of distributed
hyperdocuments, of which the system cannot show a fish-eye view because
the structure is not known by a single site.
The bread-crumbs
facility was also not considered in the experiments because
the experiments assume that a reader always recognizes a node that was
visited before, without needing bread crumbs as a reminder.