Information Technology Reference
In-Depth Information
animation showing how the steps were implemented. In Exp. 1 each step in the
animation was described by a male with a standard unaccented American voice or
it was spoken by a similar voice with a Russian accent. Participants presented with
the unaccented narrative showed significantly better comprehension and learning
than did those presented with the accented voice through performance on a task
that tapped deep understanding. Those hearing the unaccented voice also rated the
speaker more positively than those presented the accented voice. In a second study,
Mayer et al. (2003, Exp. 2) reported that those listening to the unaccented voice
also outperformed a machine-synthesized voice on ratings of learning difficulty and
voice quality.
Atkinson, Mayer, and Merrill (2005) extended work on the role of voice qual-
ity during comprehension in two experiments involving 40-min instruction periods
(as opposed to 2 min in Mayer et al., 2003) on four worked-out examples of
math problems. Narratives were spoken by animated agents with English-accented
female voices, either a human voice or a machine-synthesized voice. The nar-
rations were presented in a multimedia display that involved verbal descriptions
of each problem's solution steps along with visual displays of those steps. In
Exp. 1, college students in laboratory settings showed significantly better com-
prehension through performance on versions of the four practice problems, near
transfer problems, far transfer problems, and in ratings of the on-screen agent.
Experiment 2 was conducted in a high-school computer lab with entire classes of
college preparatory students present. Results completely replicated the laboratory
findings obtained in Exp. 1. The authors concluded that their findings were con-
sistent with social agency theory, which posits that appropriate social cues support
comprehension in multimedia presentations by encouraging listeners to engage in
human-computer interactions in ways similar to human-human interactions (Beck,
McKeown, Sandora, Kucan, & Worthy, 1996; Grice, 1975; Reeves & Nass, 1996).
It seems reasonable to conclude that several readily implemented variables of the
input stream improve comprehension of course content. First, they provide multiple
perspectives of the conceptual content. It appears that, when multiple perspectives
are included, whether the content is presented in dialog or monolog format is no
consequence (Fox Tree & Mayer, 2008). Second, the McKendree et al. (2001)
research indicates that introducing new content with questions promotes compre-
hension when contrasted with using simple declaratives (see below, Craig, Sullins,
Witherspoon, & Gholson, 2006; Gholson & Craig, 2006). The McKendree et al.
(2001) data suggest that overhearing content introduced by questions may be at
least as effective as participating in dialog. Third, the coherence of written text plays
an important role in readers' comprehension. So, making ideas in the text explicit
by avoiding potentially ambiguous pronouns, linking concepts together, along with
tying paragraphs to each other, and the topic headings with macro-propositions
improves comprehension of written text. It seems reasonable to assume that these
features would play similar roles when text is spoken. Finally, a personalized voice
style produces better comprehension than a formal style when presenting course
content, and voice quality plays a role in that an unaccented human voice produces
better comprehension than either accented or synthesized voices.
