Information Technology Reference
In-Depth Information
employed the cognitive skills that supported asking deep questions in the transfer
task, and they also learned more.
An unexpected finding was that following acquisition those who overheard deep
questions also outperformed those in the monolog condition on free-recall ques-
tions. The difference was only marginally significant, but the effect size (Cohen's
d
)
was 0.44. This suggested that overhearing deep questions embedded in course con-
tent may lead vicarious learners to not only generate more deep questions and learn
more in a transfer task but they might also learn more of the content presented
directly in the context of those deep questions. Thus, an issue that remained to
be determined was: do students not only learn to ask deep questions (Graesser &
Person, 1994), but also comprehend and learn more course content when they over-
hear that content presented in the context of discourse containing deep questions
than they learn when they overhear the exact same content presented in question-free
monolog?
Driscoll et al. (2003) explored whether, given more precise measures, the trend
in favor of the dialog condition obtained by Craig et al. (2000) following acquisition
would prove more robust. Because the between-subject variability was extreme in
the Craig et al. study, an attempt was made to increase precision by making dis-
course type (dialog vs. monolog) a within-subject variable in the Driscoll et al.
(2003) experiments. In Exp. 1 and the one that followed, two computer-controlled
male-animated agents, a virtual tutor and virtual tutee, engaged in both dialog and
monolog discourse. Each vicarious learner overheard four computer literacy topics
discussed in dialog format and four in monolog format. As in Craig et al. (2000)
the virtual tutee asked one broad question at the outset of each of the four topics
presented as monolog, but in the dialog condition he asked a deep question before
each content sentence presented on each topic. Results revealed that learning was
significantly greater on topics overheard in dialog format than in monolog format.
As part of a larger study, Driscoll et al. (2003) designed Exp. 2 to explore several
features of dialog that might have been responsible for the effects that were obtained
in Exp. 1. The study included three dialog conditions with discourse type (dialog
vs. monolog) as a within-subject manipulation in each. The monolog condition was
identical in each dialog condition. Four topics were presented in dialog format and
four in monolog.
In Exp. 2, the virtual tutee's contributions in one condition were deep ques-
tions (e.g., “How does the CPU use RAM when you run applications?”), and in
a second condition they were transformed into shallow (one-word answers) ques-
tions (e.g., “Does the CPU use RAM when you run an application?”). In the
third condition, they were converted into simple assertions (e.g., “The CPU uses
RAM when you run an application.”) spoken by the virtual tutee. It was sug-
gested that if the deep questions used by Craig et al. (2000) and by Driscoll
et al. in Exp. 1 functioned as advance organizers (Lorch & Lorch, 1995), then
shallow questions and assertions should produce the same increments in learning
as deep questions when the dialog conditions were contrasted with the monolog
condition. If questions
per se
were critical, then those in the deep and shallow
question conditions should both outperform those in the assertion condition. After
