Information Technology Reference
In-Depth Information
According to this model, a sensory source is reli-
able if the distribution of inferences based on that
source has a relatively small variance; otherwise
the source is regarded as unreliable. More-reliable
sources are assigned a larger weight in a linear-
cue-combination rule, and less reliable sources
are assigned a smaller weight. (Battaglia et al.,
2003, p. 1391)
having cognitive, attentional, or other origins. This
is especially interesting as there was no degrada-
tion in the quality of the visual percept offered,
which otherwise inevitably provokes the human
perceptual system to rely on other modalities.
To sum up, the combined results of these experi-
ments suggest that there is no clear, generalized
bias of humans toward any of the available mo-
dalities in terms of dominance. Apparently, there
is no such thing as a general dominance of visual
percepts over other stimuli. Instead, whenever such
a bias toward any of the available modalities exists,
this seems to be highly dependent on the context.
Whereas Battaglia et al. (2003) tested subjects for
contradicting localization cues and were presented
with a bias toward the visual percept, Shams et al.
(2000) tested subjects for temporal variations of
cues and were presented with a bias toward the
auditory percept. This actually indicates that the
human perceptual system tends to prefer those
senses (give a higher weight to those percepts)
that promise a higher degree of reliability or
resolution for the presented perceptual problem:
Whereas the horizontal resolution of the human
auditory system is roughly 2 to 3 degrees for
sinusoidal signals coming from a forward direc-
tion (Zwicker & Fastl, 1999), the resolution of the
visual system is at least 100 times as high, about
1 min. of arc (Howard, 1982). On the other hand,
the time resolution of the auditory system allows
to resolve the temporal structure of sounds as close
as 2ms (Zwicker & Fastl, 1999), whereas the hu-
man visual system can be tricked into believing
in a continuously moving object when presented
with only 24 sampled pictures of the continuous
movement per second.
Looking at it this way, visual capture is just a
special case of the MLE model: the highly reliable
percept (the visual cue) is assigned a weight of
one, whereas the less reliable percept (the auditory
cue) is assigned a weight of zero.
Battaglia et al. (2003) describe an experiment
designed to answer the question whether human
observers localize events presented simultane-
ously in the auditory and visual domain in a way
that is best predicted by the visual capture model
or by the MLE model. Their report suggests that
both models are partially correct and that a hy-
brid model may provide the best account of their
subjects' performances. As greater amounts of
noise were added to the visual signal, subjects
used more and more information perceived via the
auditory channel, as suggested by the MLE model.
Yet most notably, according to their analysis, test
subjects seemed to be biased towards using visual
information to a greater extent than originally
predicted by the MLE model. This means that
the model used in the experiments committed a
systematic error by constantly underestimating
the test subjects' use of visual information (thus
overestimating the use of auditory information).
Shams, Kamitani, and Shimojo (2000, 2002)
describe experiments in which visual illusion was
induced by sound, resulting in the auditory cue
outweighing the visual cue. They presented test
subjects with flashes of light and beeps of sound:
whenever a single flash of light was accompanied
by multiple auditory beeps, the single flash was
perceived as multiple flashes. They conclude that
this alteration of the visual percept is caused by
cross-modal perceptual interactions, rather than
AUDItOrY AND AUDIO-
VIsUAL PLAUsIbILItY
In classic room acoustic simulation, the time
necessary to render the room audible (in other
words, to perform the room acoustic simulation
Search WWH ::
Custom Search