Perceptual Coherence : Hearing and Seeing

238 Perceptual Coherence
quite different from that for linear motion. If the target letter moved upward
and the nontarget letters moved downward, it was easy to detect the target.
Watson and Humphreys concluded that segmenting the array on the basis of
rotation is very difficult, in contrast to segmenting the array on the basis of
linear motion. Previous research of Julesz and Hesse (1970) supports these
results. Julesz and Hesse constructed an array (roughly 80 × 60) of 4,800
“needles” that rotated around their center points. If the target and nontarget
needles rotated at the same speed in different directions, observers could
not perceive the target region. However, if the needles rotated at different
speeds, regardless of direction, the array segmented into different regions.
The authors suggested an explanation based on the organization of local
elements into one or more surfaces. If elements with different motions
form separate surfaces, then detection is made easier because the observer
can attend to one surface. If the local motions do not create different surfaces,
the observer must scan every element, so that detection becomes
much harder. From this perspective, they hypothesized that the rotation of
individual letters (or any small shape) does not give rise to the perception
of a surface, and that accounts for the poor performance.
I think what connects the research of S.-H. Lee and Blake (1999) to that
of Watson and Humphreys (1999) and Julesz and Hesse (1970) is that rotation
direction does not lead to segregation. In Lee and Blake, segmentation
was due to synchrony; in Watson and Humpheys, segmentation was due to
different kinds of movement; in Julesz and Hesse, segmentation was due
to rotation speed. If differences in rotational motion do not give rise to a set
of surfaces such that each surface corresponds to one motion (a failure of
the Gestalt principle of common fate), it demonstrates that visual attention
is not simply afloat in space but is necessarily attached to continuous
surfaces. This is the same argument made by Wandell (1995) in chapter 1
that we perceive motion with respect to dense surface representations.
Auditory Second-Order Patterns
Huddleston and DeYoe (2003) have developed an analogy to second-order
vision patterns by equating movement through pitch with movement
through space. Suppose we start with a set of 10 to 12 tones that range from
300 to 10000 Hz. If we present each tone separately, one after the other,
then we would have the equivalent of a first-order visual stimulus, and we
would expect that it would be trivial for listeners to determine if the tones
were increasing or decreasing in pitch. Now suppose we conceptually start
at Time 0 presenting all of the tones at once, but some of the tones are on
and some are off. As illustrated in figure 5.20, tones 1, 2, 3, 5, 8, and 10 are
on and 4, 6, 7, and 9 are off. Now at each successive time point, we switch
the next highest tone on-to-off or off-to-on. As shown in figure 5.20, at