Sound Source Segregation 137frequencies, and each tone is pulsed on and off such that
when one tone is on, the other is off. Two different percepts
may occur in this situation. In one case, the perception is of
one sound source that is alternating in pitch. In the other case,
the perception is of two sound sources, each associated with
the individual frequencies and with each containing a pulsing
sound. In the latter case, it as if there were two sound sources,
each producing a pulsing tone occurring at the same time, like
two streams running side by side (stream segregation). By
determining the stimulus conditions that yield stream segre-
gation, investigators have attempted to study those stimulus
conditions that promote sound source segregation. In addition
to an appropriate frequency separation between the two
sounds, differences in stimulus complexity, interaural differ-
ences (i.e., spatial differences), temporal amplitude modula-
tion differences, and level differences may promote stream
segregation. The temporal structure of the stimulus context
plays a crucial role in stream segregation. In general, spectral
differences promote stream segregation more than other
stimulus attributes do.
In another methodology involving temporal sequences of
sounds, an auditory pattern of tonal sounds is generated as a
model of complex sounds, such as speech (Watson, 1976). In
many conditions, a pattern of 10 tones presented in succes-
sion, each with a different frequency, is used as a tonal pat-
tern. The frequency range over which the tones vary, the
duration of each tone, and the overall duration of the 10-tone
pattern are often similar to that occurring for many speech
sounds, like words (see the Fowler chapter of this volume).
Listeners are asked to discriminate a change in the frequency
of 1 of the 10 tones in the pattern. In many conditions, the
patterns change from trial to trial in a random manner. In this
case, frequency discrimination of one tone in a pattern of
changing tones is very poor, especially for tones at the begin-
ning and at the end of the 10-tone pattern. However, as the
random variation in the patterns is reduced, frequency dis-
crimination improves, and the differences in discrimination
as function of the temporal order of the tones are also re-
duced. When the same 10-tone pattern is presented on each
trial (i.e., there is no randomization of the pattern frequen-
cies) and only one tone is subjected to a frequency change,
frequency discrimination thresholds for any one tone in the
10-tone pattern is nearly equal to that achieved when that
tone is presented in isolation. These 10-tone pattern experi-
ments show that the uncertainty about the stimulus context
can have a large effect on performance in identifying
complex sounds.
Information masking is used to describe the decrease in
performance attributable to the stimulus context rather than
to the actual values of the stimulus parameters. Thus, the
changes in performance due to certain versus uncertain con-
texts in the 10-tone pattern experiments for the same stimulus
values is due to informational masking. Another example of
informational masking involves a tonal signal and a tonal-
complex masker. If the tonal complex is a 100-tone masker
and all 100 tones are mixed together at one time to form the
masker, a certain signal level is required for signal detection
(assume the signal frequency is in the center of the range of
the frequencies used for the tonal-complex masker). If only 1
of the 100 tones in the tonal-complex masker is chosen at
random and presented alone on each trial and masking of
the signal is measured over the random presentation of the
100 tones, then signal threshold may be elevated by 20 or
more decibels relative to the case when all 100 tones were
mixed together at the same time to form the single tonal-
complex masker. The increase in threshold is referred to as
informational masking due to the uncertainty in the masking
stimulus from trial to trial, despite the fact that the frequency
range over which the masker varies is the same in both con-
ditions, and on many trials the signal should be easy to detect
because its frequency would be very different from that of the
masker on that trial (Neff & Green, 1987).Spatial SeparationThe section on sound localization described the ability of lis-
teners to locate a sound source based on the sound that is pro-
duced. When sound sources are located at different locations,
does this spatial separation aid sound source segregation?
Cherry (1953) stated that spatial separation would aid sound
source segregation when he coined the term cocktail party
effect.That is, spatial separation was a way to segregate one
sound from the concoction of other sounds at a noisy cocktail
party. Spatially separating sound sources does aid in the
identification of the individual sound sources, especially
when there are more than two sound sources (Yost, Dye, &
Sheft, 1996).
The masked threshold for detecting a signal presented
with one set of interaural differences can vary greatly as a
function of the interaural differences of the masker. If the
signal and masker are presented with a different set of inter-
aural differences, then signal threshold is lower than in con-
ditions in which the signal and masker are presented with the
same interaural differences. The decibel difference in masked
threshold between a condition in which the signal and masker
have different interaural differences compared to that in
which they have the same interaural differences is the mask-
ing-level difference, MLD (Green & Yost, 1975; Yost & Dye,
1991). For instance, if the masker (M) and the signal (S) each
have no interaural differences (subscript 0), the condition is