Since its conception, hypertext has attracted the attention of educato
rs and psychologists alike. Although a great deal is known about learn
ing from text, little is understood about the process of learning from
hypertext or what benefit it offers over traditional text. This study
is an attempt to (a) determine whether theories of learning from text
may be extended to learning from hypertext, and (b) learn more about
the general effectiveness of hypertext on learning. Seventy-two colleg
e undergraduates participated in a study of hypertext-based learning.
Each participant was assigned to work with one of three hypertext syst
ems. All systems contained the same documents. Two of these contained
the same electronic links (pathways) between documents and the third s
ystem condition served as a control. It was presented as a digitized b
ook (linear text) rather than as a linked system. Participants' naviga
tion behavior was logged electronically as they worked. Posttests incl
uded an essay, a series of short-answer questions, and a concept mappi
ng task. Analyses revealed that learning from hypertext bears many sim
ilarities to learning from text, as the predictions made by Kintsch's
(1988) construction integration model were borne out. System structure
systematically altered what was learned from hypertext, just as chara
cteristics of text alter text-based learning. System structure was als
o relevant to the way in which learners approached the material, as na
vigation behavior was affected. The less structured system seems to ha
ve promoted more active processing and a deeper level of learning. In
addition, hypertext was revealed to have only limited educational bene
fit for users in this study. Although no benefit of either hypertext s
ystem was observed over the linear system on the essay or short-answer
questions, it was revealed that the presence of system links affects
internal representations: Participants who were exposed to the hyperte
xt systems produced concept maps that largely reflected their system l
inks. Results are examined with reference to cognitive theory and the
implication for system design is discussed.